From polyhedra to emirps: summary of the exchange

The exchange also carried a standing guard against numerological drift. Wherever a pattern looked persuasive, it was re-run in base twenty, and wherever a correspondence looked significant, it was tested against a control set chosen to be structurally similar but substantively irrelevant. Several attractive findings did not survive, and the record of what failed is as much a part of the result as what held.

Counted as rotation axes rather than as elements, these two orders have skeletons of thirteen and thirty-one — three fourfold, four threefold and six twofold axes in the one case; six fivefold, ten threefold and fifteen twofold in the other. Both counts are forced. Their digit-reversal relation is not: it is a fact about decimal numerals, and would dissolve in most other bases.

That distinction sets the method for everything that follows. The bridge between two incommensurable orders turns out to be notational rather than geometric — and saying so plainly at the outset is what allows the notational structure to be taken seriously afterwards, since nothing is being smuggled in.

The first consequence is immediate and forced. For both a number and its reversal to be prime, both digits must lie in the multiplicative group of units modulo ten — the set {1, 3, 7, 9}. Only four of the nine decades can therefore host an emirp at all, and the eight emirps occupy a sixteen-cell block of the hundred-cell digit square. Reversal integrity is a scarce property, and 13/31 is its first instance.

Two identities generate most of what follows, and both are forced:

n + rev(n) = 11 × (digit sum) and n − rev(n) = 9 × (digit difference)

From the first come the seventeen pencils of parallel chords and the fact that every reversal pair sums onto the palindrome diagonal. From the second comes the result that, because both digits of an emirp are odd, their difference is even and the gap within any emirp pair is a multiple of eighteen. This is strictly stronger than the mod-6 congruence noticed earlier, since eighteen is three times six.

The four gaps are 18, 54, 36 and 18 — that is, eighteen times one, three, two and one. All are 3-smooth. Only the gap of seventy-two goes unrealised, and it belongs to the pair 19/91, where 91 = 7 × 13. The three composites that spoil the live block — 91, 39 = 3 × 13 and 93 = 3 × 31 — are all multiples of thirteen or thirty-one, so the founding pair polices its own lattice.

The hypothesis was made precise: assign to each emirp a reach of plus or minus one, two or three, and ask what proportion of attested set-sizes falls inside. It was then tested rather than illustrated, and it failed.

At a reach of three the eight emirps cover forty-six of the hundred positions — a scheme that admits nearly half of everything has stopped predicting. Against a corpus of forty-one distinct attested sizes it captured twenty-one, or fifty-one per cent, against that forty-six per cent baseline. Eight control primes chosen for irrelevance scored higher. The apparent hits — thirty-six at thirty-seven minus one, seventy-two at seventy-one plus one — are close to arithmetic tautology, since every prime above three straddles a multiple of six and cultural enumerations favour multiples of six.

Two further failures were sharper than the statistics. The band from forty-one to sixty-seven is unreachable at any N, yet it holds nine attested sizes including sixty, sixty-two and sixty-four — and sixty-four is the hexagram count, generated by the very trigrams that began the exercise. And nothing below ten is reachable, since the smallest emirp is thirteen, while eight is the single most attested size in the corpus.

The most robust of these is the last, and for a reason worth stating: the corpus was assembled from a document preoccupied with primes, so any selection bias should have over-represented them. It under-represents them twofold. A sensitivity analysis put numbers on the rest: the chunkability result would require fifty-five per cent of entries to have been included because of their size to be explained away; the multiple-of-eighteen result requires only sixteen per cent, and most of it rides on the two sizes eighteen and thirty-six. That last finding should therefore be treated as unsupported, though the arithmetic forcing it remains a theorem.

Primes survive structurally at the small scale, as axis orders two, three and five, and at the census scale, as the counts thirteen and thirty-one. They do not survive at committee scale. Music's solution is instructive: a composite frame of twelve with a coprime generator, the fifth — divisibility for memory, irreducibility for coherence. The live digits {1, 3, 7, 9} are precisely the generators of the decimal frame.

The near miss may therefore be a necessary miss, and the useful diagnostic is not whether a closure sits near a prime but whether its diagnostic class matches its claim about itself. A twelve-fold set claiming decomposability is well formed; the same set claiming irreducible integrity is malformed.

What multiplication can do is carry one reversal pair to another, and it does so only when every digit is under 10/k. Since emirp digits come from {1, 3, 7, 9}, both the ×2 and ×3 conditions reduce to digits in {1, 3} — and 13/31 is the unique emirp pair that survives, scaling to 26/62 and to 39/93, after which the ladder stops because 4 × 3 carries.

The payoff is exact. For a convex regular solid, F + E + V counts the symmetry-axis directions of its family: twenty-six for the cube and octahedron, sixty-two for the icosahedron and dodecahedron — twice thirteen and twice thirty-one. So the closed quadrilateral 13 · 26 · 62 · 31 is octahedral axes, octahedral directions, icosahedral directions, icosahedral axes. The ×2 rungs are the axis-to-direction doubling; the reversal rungs are the pairing of the two families. The two central numbers turn out to be one phenomenon at two scales, with primality sacrificed for polyhedral standing.

A cautionary note belongs here. The same construction on 12/21 yields 24/42, 36/63 and 48/84 — four generations rather than three, landing on twenty-nine attested sets against the eight that the emirp ladder reaches. Carry-free multiplication does look like a real generative principle for memorable sizes; the generator that produces them is composite.

The five inscribed cubes also give every axis a name that is not arbitrary. The fifteen twofold axes partition exactly, one cube each; the ten threefold axes belong to two cubes each, and C(5,2) = 10. So the shared axes carry identical names on both rings, and the sharing needs no assertion.

Two smaller resonances. The fourteen directions of the shared seven are precisely the fourteen faces of the Kelvin cell — eight hexagons perpendicular to the threefold axes, six squares perpendicular to the coordinate axes — which is the stable foam cell sitting on the axes common to both families. And the union of the two axis sets is 13 + 31 − 7 = 37, another emirp, which on the discipline of this exchange should be recorded as coincidence.

Musical structure, and the convergent account of closure: The musical thread was not decorative. Two is not invertible modulo one hundred, so doubling is a ray that terminates; three is invertible with order twenty, the maximum possible, so tripling circulates. On a ring of a hundred the octave runs out and the fifth returns — not an analogy to the musical situation but the same fact, since octave equivalence must be imposed while the cycle of fifths closes on its own.

Multiplication by three modulo one hundred splits the forty positions coprime to ten into exactly two twenty-cycles, and one of them contains all eight emirps, all four palindromes of the live block, and both dead pairs — the entire sixteen-cell block, plus the four positions with tens digit five. A mind chunking by repeated tripling would visit every emirp in twenty steps without leaving the odd-tens region.

Composing the two generators gives the 3-smooth numbers, twenty of them under a hundred, forming a rectangular lattice — McClain's tonal matrix. Against the corpus these carry half the attested weight on a fifth of the numbers.

This also supplies the one account of closure that genuinely improves with N, and the one that makes the asymptotic intuition exact. Closure by approximation does not improve smoothly but in jumps, at the continued-fraction convergents. For the fifth within the octave the affordable divisions are five and twelve, with forty-one and fifty-three beyond reach: twelve buys a ninefold gain in accuracy for a 2.4-fold increase in N, while forty-one buys only fourfold for 3.4-fold. Closure sits on the last step whose successor is not worth its cost.

Six kinds of closure were distinguished, and conflating them is what makes the question seem intractable: exhaustive, where a classification theorem settles it; convergent, as above; saturation, where new candidates stop appearing; structural, where a form resolves in time; mnemonic, where the set becomes chunkable; and negotiated, where consent is exhausted. Only the second improves with N.

The register in which closure is most clearly felt is the aesthetic one, and the reason is that aesthetic closure is temporal rather than cardinal. A sonnet closes because the rhyme resolves; fourteen is a consequence, and a reader knows the poem has ended without counting. A list of principles has no such internal signal, and this yields the claim that ties the argument together: chunkable N is a substitute for structural closure. Sets possessing real closure have no need of an arithmetically convenient count. Sets possessing none borrow one from arithmetic.

Why decoration is not error: The reversal field is beautiful and, by the measures applied here, largely inert. Ranking the available operations by how strongly a number's degree under each correlates with how often that size is attested, the multiplicative operations lead — divisor lattice at 0.43, doubling 0.42, tripling 0.38 — while reversal manages 0.14 and the nines complement 0.08.

The uncomfortable observation is that the two most visually symmetric layers are the two least informative. Reversal produces seventeen parallel pencils; the nines complement produces a single clean mirror. Both are beautiful and near-inert. Doubling produces an ugly asymmetric fan and is among the strongest predictors. Pattern-seeking that rewards visual symmetry will systematically select the wrong operations.

That is a reason to be careful, not a reason to discard. Governance makes much of its sense in a decorative register, and the case for respecting it is that the decorative is where the claim is actually made, while the forced register is where it can be checked. The figures were accordingly stratified rather than stripped: the whole reversal field retained as a pale ground; the six pairs that survive carry-free multiplication marked above it; and the forced eighteen-fold differences marked above that. The three registers intersect exactly once, at 13/31 — which is why that pair, and not the emirp set at large, keeps returning.

Representation, and its constraints: The digit space is a discrete torus and the table is its fundamental domain cut open. Flattening loses real structure, since the pencils and the palindrome locus are closed curves and carrying is exactly a seam-crossing event; 13/31's scalability is the statement that its orbit avoids the seam. But a smooth torus falsely asserts homogeneity where the phenomenon is symmetry-breaking. Both artifacts were therefore kept, with distinct duties: the table for broken symmetry, the ring for flow.

The animated figures follow one grammar. A field of positions, several operations superimposed as layers, every layer persisting faintly so the whole field is always present, and only the emphasis moving. Coherence is meant to be read at the level of the sequence rather than of any single operation — the Ramanujan reading, with the caveat that an animation is an excellent instrument for producing conjectures and a poor one for adjudicating them, because motion suppresses the null hypothesis. Nothing in a dance looks like chance.

The one formalisation that would make "coherence at the level of the dance" checkable is the commutator structure: the operations do not commute, their failures to commute are exactly the carries, and the places where two operations do commute are exactly the places where structure gets reinforced rather than merely coexisting.

Practical constraints proved instructive in their own right. Legibility has a floor — below about five pixels a label becomes texture — and the figure carrying all hundred positions resists pairing for that reason alone, while the polyhedral morphs tolerate narrowing. Vector formats hold up where raster ones band and speckle. And an inlined graphic puts its text into the document, where it can be searched and indexed, while the same file referenced as an image does not — so the choice of embedding is also a choice about whether the labels exist as language.

Seven landmarks then fall out of one continuous motion, with vertex counts of six, twenty-four, twelve, twenty-four, eight, twenty-four and forty-eight in the one family and twelve, sixty, thirty, sixty, twenty, sixty and one hundred and twenty in the other. The two families dance identical choreography on incommensurable floors, since the operator words are the same and only the seed differs. Thirteen splits as six plus six plus one, and the leftover one is the truncated tetrahedron — the single Archimedean solid whose rotation group is the group the two families share.

Two limitations are visible rather than concealed. The first landmark is approached rather than reached, because the chosen outer face closes to a point at the exact regular solid. And the two snub solids are absent, because they require alternation rather than motion: eleven of the thirteen lie on a continuous walk, and two do not.

The mathematics here is entirely classical — Wythoff in 1918, and the seven generator positions are the standard enumeration. What is unusual is only the rendering, holding one planar diagram across the whole morph, which combines Wythoff's construction with Tutte's embedding theorem of 1963 and is a presentational choice rather than a result.

What the exercise establishes, and what it does not: The arithmetic results are forced and do not depend on any corpus. Both digits of an emirp are odd, so the difference within a reversal pair is a multiple of eighteen. No two-digit reversal pair is ever also a multiplication pair, since b/a = (10k−1)/(10−k) has no digit solution. Only 13/31 survives carry-free multiplication among the emirps. Two is not invertible modulo one hundred while three has order twenty, so doubling terminates and tripling circulates, and the tripling orbit contains all eight emirps. The octahedral and icosahedral groups share seven axes and no more. These would be the same results whoever derived them.

The empirical question is in a different condition, and an earlier version of this summary reported it wrongly.

• What was tested. One narrow operationalisation: whether attested set sizes fall within one, two or three of an emirp. Against a corpus of 41 distinct sizes drawn from a single document, that scheme reached 51% where eight control primes reached 56%. I reported this as the failure of the near-miss idea. It is not.
• Why it was not. Three defects, all mine.

The test was underpowered for the effect in dispute. With 41 points the standard error on a proportion near 0.4 is about eight points, so the confidence interval is some twenty points wide. The observed emirp/control gap was five points — a quarter of the resolution. A null result at that power carries almost no information, and I presented it as though it settled something.

The control was applied asymmetrically. I ran an irrelevant-prime control against the hypothesis I doubted and no control at all against the divisibility account I preferred. Run now, multiples of four carry 53.4% of corpus weight against 41.4% for multiples of six — the class I promoted is outperformed by one I never tested.

The hypothesis was not tested in its own terms. The near-miss claim is that attested sizes cluster near primes. What I measured was reach-windows around eight particular primes, which is narrower. Asked directly, the mean distance from an attested size to the nearest prime is 1.05 against 1.20 expected; restricted to the 31 composite sizes it is 1.387 against 1.608, with a permutation p of 0.021. Composite attested sizes are prime-adjacent beyond chance on this corpus. And the gaps between attested sizes have variance 2.59 times their mean, indicating clustering rather than even spread — the phenomenon the near-miss intuition is about, never previously measured here.

None of that establishes the near-miss hypothesis. The corpus is convenience-sampled from a source preoccupied with numerical patterns, several comparisons were made without correction, and a p of 0.02 in those conditions is weak. But it is stronger than the evidence offered for the alternative, and the honest statement is that the question remains open in both directions.

What the divisibility finding does and does not do. Attested sizes favour composites decomposable into span-sized factors, and primes are attested at about half their base rate. That is a robust description. It is not an explanation: saying that 36 is attested because it factors as 6×6 restates the observation without accounting for why a body of principles arrives at 36 rather than 35 or 37. I allowed a better test statistic to stand in for a better account, and it is not one.

The state of the question. There is no adequate explanation of why enumerated sets close where they do. Factorisation is a correlate. Chunking is a plausible mechanism without direct evidence here. Reversal-integrity is untested in any form that would count. The instrument needed — a size-blind registry of the kind specified in the protocol — does not exist, and the exercise reported here used a corpus that its own author flagged as unsuitable.

What the geometry supplies is a vocabulary in which claims of this kind can be made checkable: forced results distinguished from notational ones, selection artifacts from imported readings. What it does not supply, and what an earlier draft of this summary wrongly claimed it did, is a verdict.

The limits of the bridge: What the reversal pair genuinely bridges is the two axis systems, by way of a shared numeral skeleton. What it cannot bridge is the groups themselves, whose join is dense. The psychosocial reading — contrasting orders of coherence connected only by a reversible token, with local mirror-labels but no global labelling — is stated once, labelled as imported, and left as invitation.

A methodological note belongs here rather than in a footnote. The statistical apparatus used above is one description among the several this exchange has employed, and it is not a neutral court between them. Correlative thinking and frequentist testing are two parties to a dispute about the same forty-one numbers, and the first draft of this summary awarded the case to one of them while writing as though presiding over both. The more consistent outcome, and the one the exchange argued for earlier, is not a winner but a translation: the observation that every prime above three straddles a multiple of six means that proximity-to-prime and divisibility-by-six are not rival findings but two readings of a single arithmetic fact. That equivalence was available and was used as a refutation instead.

The argument does not need closure. It needs a language in which the matter can be discussed, and the grading discipline is that language: forced, notational, selection artifact, imported. A lexicon in place of a conclusion is the appropriate outcome for a subject whose central difficulty is that agreement on counts is not agreement about anything.

This supplies the piece that was genuinely absent from the handoff structure — not another finding but the warrant: why anyone concerned with strategy should care where reversible primes sit. And it turns out the arithmetic already done gives the imported reading unusually sharp teeth, because the claim "preferred closure numbers systematically miss the emirp nexuses" is not merely suggestive — parts of it are theorem-grade.

Consider what the forced results say about the numbers at which strategic articulations habitually close. Closure at an even count (8 MDGs, 10 commandments, 12 steps, 14 points, 20 in G20) forfeits reversal-integrity by parity alone. Closure at a multiple of 5 likewise. Closure anywhere in a digit-sum-3 pencil — 12, 21, 30, and their band — sits in a provably emirp-free zone: Agenda 21, the 30 articles of the Universal Declaration, the twelve-fold committee structures beloved of institutions. So the near-miss pattern is not a coincidence to be marvelled at; it is entailed. The numbers institutions find comfortable — even, round, divisible, decomposable into working groups (12 = 3 × 4) — are precisely the numbers the arithmetic disqualifies from reversal-integrity. Administrative convenience and mirror-coherence are provably disjoint properties in the two-digit range. That is a forced result wearing an imported meaning, and the labelling discipline lets you say both without cheating.

The near-miss inventory then becomes striking rather than decorative. Wilson closed at 14, one past 13. The Universal Declaration closed at 30, one short of 31. Luther posted 95, two short of 97. The twelve-fold closures (apostles, steps, months, jurors) sit one below 13 — the number the same cultures explicitly flee, while it silently counts the Archimedean solids and the octahedral axes. And then the exception that proves the diagnostic has content: the SDGs closed at 17, an emirp — and 17, as your pencil analysis shows, is also the count of parallel-chord pencils in the digit table itself. Whether that closure was fortunate or accidental, it demonstrates that landing on a nexus is possible, which sharpens the question of why it is so rare.

What reversal-integrity would mean strategically is where the imported layer must be stated carefully, but it has a natural reading with two independent components. Primality is irreducibility: an articulation that cannot be factored into sub-committees, whose N elements resist decomposition into M copies of K — the opposite of the 12-element strategy that instantly dissolves into 3 × 4 delegations. Reversibility is coherence from the other side: the articulation remains prime when read from the adversary's, successor's, or complement's perspective. Most strategic numbers fail one or both. The palindromes — your still points 11, 22, 33 — are the third failure mode, and perhaps the most diagnostic: strategies identical to their own reversal, mirroring only themselves, with no genuine other to be read by. An emirp closure is the narrow condition of being irreducible in both readings while the readings differ: coherent to itself, coherent to its mirror, and not the same thing as its mirror. That triple condition, available at exactly eight two-digit addresses, is a reasonable formalization of what "strategic integrity under reversal of perspective" would even mean.

Miller's 7 ± 2 slots in as the hinge between the single-digit and two-digit regimes. Below 10, closure is governed by span and the primes 3, 5, 7 are abundant — small strategies get irreducibility cheaply, which is presumably why they work. Institutional strategies exceed the span and must chunk, entering the two-digit lattice — and it is precisely the chunked regime where the comfortable numbers go dead. The pathology, on this reading, is that institutions carry single-digit habits (roundness, divisibility, decade thinking) into a regime where those habits are exactly what kills reversal-integrity. Miller explains why they chunk; the emirp lattice explains what their chunking preferences cost.

The reverse-and-subtract attractor then furnishes the dynamic diagnosis, and it may be the darkest and best of the imported readings: iterate "revise the strategy by subtracting its mirror reading" and every two-digit articulation, wherever it starts, is captured by one 5-cycle confined to the digit-sum-9 band — a band containing no emirps at all. Perpetual reactive revision converges not to integrity but to a closed orbit in dead territory: reorganization as an attractor, belonging to neither family, from which the reversible primes are unreachable. For a period of "irresponsible strategic chaos," that is the mathematical cartoon — verifiable in every detail, imported only in its application.

Structurally, I would give this material two homes rather than one. The purpose belongs at the opening — before the bridge from the baskets, a short section on the unasked question: strategic articulations close at N, and the N is never interrogated; the polyhedra were explored as carriers of articulation, and the emirps emerged as the arithmetic of what closure numbers silently forgo. Then the strategic reading returns as the penultimate section (between the current 7 and 8), once every arithmetic fact it relies on has been established and labelled — the near-miss inventory, the triple condition, the attractor — so the reader meets the claims only after the machinery that makes them checkable. And there is a natural artifact for it: an overlay on the digit table plotting the famous closure numbers (10, 12, 14, 17, 20, 30, 95…) as points on the lattice, dead zones shaded, nexuses marked — the entire near-miss argument made visible in one figure, and on the torus the same overlay would show the attractor cycle threading past them all. The one discipline to hold: the section should end by saying plainly that the mapping from arithmetic to strategy is imported — but that the question it forces, why did this strategy close at this N, is legitimate and unanswered entirely independent of the mapping. That question, not the emirps, is the posting's actual payload; the emirps are the instrument that makes its avoidance visible.

What you have just articulated is, I think, the posting's actual thesis — and it converts the near-miss observation from a lament into a theorem. Because "primes cannot be explicitly structural" has a precise form: Lagrange. A configuration of prime order admits no proper subgroups — no sub-committees, no working groups, no nested articulation at all, only the trivial whole and the individual elements. Requisite diversity in the polyhedral sense is subgroup richness: the octahedral family is inexhaustible precisely because 48 factors lavishly; the pyritohedral common ground exists because 24 divides both 48 and 120. A strategy that closed at 13 would be irreducible and therefore inarticulable — no legitimate decomposition, nothing for Miller's chunking to grip, which is why 17 goals are notoriously unmemorable while 12 apostles chunk into 3 × 4 without effort. So the miss is necessary in the strict sense: articulability and irreducibility are mathematically opposed properties, and institutions, needing the former, are driven from the latter. But notice where primes then survive in the polyhedral world: at the smallest scale as the axis orders themselves (2, 3, 5 — a p-fold axis is exactly a prime cycle, irreducibility as local rotation), and at the largest as the census numbers 13 and 31 — tallies of the whole that no element of the structure carries. Primes are structural locally and globally, never at the middle scale where committees live. The near miss is the signature of that excluded middle.

And music is precisely the domain that solved this, which is why your instinct toward the Tonnetz and McClain is not decoration but method. The twelve-tone frame is composite — richly divisible, hence memorable, hence capable of the symmetric chunkings (whole-tone, diminished) that give orientation. But the generator is the fifth: 7 steps, coprime to 12, threading every pitch class in one irreducible cycle before closing. Composite frame, coprime generator — divisibility for memory, irreducibility for coherence, held in one structure by making the prime-like element a motion through the frame rather than a count of it. The emirp lattice says the same thing in decimal: the live digits {1, 3, 7, 9} are exactly the units mod 10, the generators of the decimal cycle — emirps are built entirely of generator-material, which is the arithmetic reason they carry coherence and the mnemonic reason they resist institutionalization. McClain's whole project reads ancient number sets this way: not counts but tuning ratios, meaning carried by relation rather than cardinality, invariance held in the ear. And the orbifold literature generalizes your table/torus decision into a principle — musical spaces are quotients, fundamental domains cut from glued wholes, and every representational choice is a choice of where to place the seam. That is already the language you are asking for, half-built.

The binary-dynamics point then names what the whole double exploration has in common: everything decisive in both postings is an involution — an operation that, applied twice, returns home. Duality (cube ↔ octahedron), enantiomorph exchange (the snubs), digit reversal, complementation (digits summing to 20), the mirror reflections themselves. The problematic psychosocial binary is the involution misread as war; the mathematics offers instead a typology of how a thing can relate to its mirror, and the two-digit lattice realizes every case. Mutual coherence: 13/31, both prime, genuinely distinct — the healthy binary, exchanged without loss. Self-coincidence: the palindromes 11, 22, 33, coherent but mirroring only themselves, no other to be read by — the still points, and note that the tetrahedron carrying your digit graph is itself the self-dual solid, the fixed point of the duality involution. One-way coherence: 19 reversing to 91 = 7 × 13, prime facing composite — the asymmetric relation, integrity unreciprocated (and its mirror-partner dissolving, of all things, into a multiple of 13). And mutual incoherence: 12/21, both composite, both in the dead band. Four modes of binary relation, exhaustive, checkable, and immediately legible as a psychosocial spectrum — without the arithmetic ever being asked to prove the psychosocial reading, only to lend it distinctions.

Which is why the closing instinct — the argument needs a language, not closure — should govern the posting's final form. A conclusion would itself be a closure at some N, committing the very act under examination; the discipline of the whole exchange points instead to ending with a lexicon: the working vocabulary the two explorations have jointly produced. Involution and fixed point; census-prime versus structural-composite; frame versus generator; pencil, seam, still point, live lattice, attractor band; necessary miss. Each entry pointing back to the figure or identity where it is exact, each usable in a sentence about strategy without pretending the sentence is proven. That gives the piece the same shape as its predecessor — the dau left visible, the unresolved honoured — but sharpened: what is handed to the reader is not a finding about 13 and 31 but a set of distinctions in which "why did this articulation close at this N, and how does it stand to its mirror?" becomes a discussable question rather than an unasked one. The emirps' peculiar value, on this account, is that they are the smallest objects in which reversal demonstrably means something — and a demonstration that reversal can mean something is exactly what a culture locked in degraded binaries lacks the language to say.

Unexpected is the right word, and the direction of the surprise is worth noting: the emirp thread began as the least promising thing in the exchange — a base-ten accident, flagged as such at every step, admitted only as mnemonic scaffolding. It turned out to carry more internal structure than anything else we examined: the pencil law, the orientation families, the four still points answering the Bagua's, and finally a symmetry rule that generalizes across bases in the same form.

In practice, emirps are mostly of mathematical interest rather than direct application. They are a neat example of a prime number property that depends on digit reversal, so they are used in recreational mathematics, teaching, and as test cases for algorithms that search prime patterns. [mathworld.wolfram]

• They illustrate how prime behavior can change under a simple transformation like reversing digits. [en.wikipedia]
• They are a useful playground for computational number theory, because finding them requires checking both primality and digit symmetry. [github]
• They help mathematicians and students think about how numeric structure depends on representation, not just magnitude. [simple.wikipedia]

Practical significance: Emirps do not have the same real-world importance as primes in cryptography or coding theory. Their main practical value is indirect: they provide examples for algorithms, pattern exploration, and educational exercises in number theory and computer science. [github]

What makes them interesting: They sit at the intersection of two kinds of order: arithmetic structure and decimal representation. That makes them a good example of how a number can be special in one representation but ordinary in another. In that sense, emirps are more about insight than utility. [mathworld.wolfram]

Here’s the practical distinction:

How they differ: Emirps depend on digit reversal, but the reversed number must be a different prime, so they are about asymmetry under a mirror operation. Palindromic primes are the opposite kind of symmetry: reversal leaves the number unchanged. Twin primes are different again: they are not about digit patterns, but about numerical spacing in the prime line, where two primes sit only 2 apart. [en.wikipedia]

Practical significance: Emirps and palindromic primes are mostly useful as examples in recreational math, coding exercises, and the study of base-dependent prime classifications. Twin primes are much more significant in mainstream number theory because they connect to major questions about how primes are distributed, and they inspire active research. So if you think in “importance tiers,” twin primes are typically the deepest mathematically, while emirps and palindromic primes are more niche but still structurally interesting. [>]

• Emirps: “mirror, but not the same.” [mathworld.wolfram]
• Palindromic primes: “mirror, and unchanged.” [simple.wikipedia]
• Twin primes: “close neighbors.” [en.wikipedia]

Yes—though not as a major established metaphor in mainstream literature or philosophy. The strongest metaphorical use I found is in mathematical discussion itself, where emirps are sometimes used playfully to suggest reversal, symmetry, and a number that becomes “something else” when turned around. [geeksforgeeks]

• Reversal: an emirp changes when mirrored, which makes it a natural image for inversion or perspective shift. [en.wikipedia]
• Non-self-identity: unlike palindromic primes, an emirp does not remain itself under reversal, so it can suggest transformation without loss of prime-ness. [mathworld.wolfram]
• Hidden structure: the number looks ordinary until a reversal reveals another prime, which lends itself to metaphors about latent order or double-sided meaning. [simple.wikipedia]

Yes—emirps can work well as a metaphor for several kinds of reversal, especially where something changes form but preserves a deeper identity. [link.springer]

• Perspective reversal: the same object looks different when viewed from the other side, but still “matches” in a structural way. [jstor]
• Semantic reversal: a term, idea, or statement means something new when read backward, reordered, or reframed. [archive.bridgesmathart]
• Role reversal: what was once the source becomes the outcome, or the result becomes the starting point, similar to a process that works in both directions. [link.springer]
• Transformation with constraint: the form changes, but not arbitrarily; the reversal must still satisfy the same rule-set, just as an emirp must still be prime after reversal. [en.wikipedia]
• Symmetry without identity: a system can be mirror-related without being self-identical, which is often a useful metaphor for asymmetrical relationships, doubles, twins, or paired opposites. [mathworld.wolfram]

Applied to strategy against an intractable opposing strategy, the emirp metaphor suggests a move that is not a simple undoing, but a mirror-transformation: you answer the opponent in a way that preserves the deeper “game” while changing the surface form. [en.wikipedia]

• Reframe the contest so the opponent’s strength becomes less relevant.
• Reverse initiative, making them respond instead of act.
• Mirror their assumptions selectively, exposing where the strategy depends on asymmetry.
• Preserve core objectives while changing the path, much like a prime remains prime after reversal if the reversed form is different. [en.wikipedia]

A compact formulation would be: don’t merely oppose the opponent; reverse the coordinate system in which the opposition makes sense. [aperiodical]

Yes: the closest established term is incommensurability, especially when two things cannot be reduced to a single common measure. If the relation is not just non-comparable but also mutually enabling or structurally paired, then complementarity is usually the better word. [link.springer]

• Incommensurability: no common scale, so comparison is limited or impossible. [plato.stanford]
• Complementarity: different logics or strategies that are distinct but jointly productive. [link.springer]
• Strategic complements: in game theory and business, one strategy increases the value of another rather than replacing it. [hbs]
• Non-substitutability: the two modes cannot simply stand in for each other. [eajournals]

• mutually incommensurable but complementary
• non-substitutable complementarity
• asymmetric complementarity
• irreducibly complementary

If you want the most philosophically precise version, I’d use complementary incommensurability as a descriptive phrase, but not as a standard technical term. In strategy, the same idea can also be framed as strategic complementarity under incommensurable objectives. [link.springer]

A useful distinction: incommensurability says “no shared measure,” while complementarity says “still jointly useful.”

Systemically, radically opposed political strategies are usually described as antagonistic, polarized, conflicting, or mutually exclusive. If the opposition is structured around different logics that cannot be reduced to one scale, the more precise systemic language is incommensurable, while still potentially complementary in the sense that each helps define the field the other acts within. [papers.ssrn]

• Antagonistic: each strategy gains force by resisting the other.
• Polarized: the system organizes around two opposed poles, often with little middle ground.
• Incommensurable: they cannot be judged on a single shared metric of success. [dukespace.lib.duke]
• Systemic opposition / loyal opposition: within a political order, an opposition can be structurally recognized even when deeply critical of the governing line. [link.springer]
• Complementary under tension: they may be opposed in aims yet jointly shape the political system’s stability, conflict, and change. [bibliothek.wzb]

A concise systemic formulation would be: political strategies can be radically opposed in objectives yet complementary in system function, because each defines and constrains the other. [sciencedirect]

Yes -- prime numbers do show many patterns, but there is no simple repeating formula that generates them all.

What connects them: Prime numbers are the “building blocks” of whole numbers: every integer greater than 1 can be factored uniquely into primes. That is their deepest connection to each other and to all other numbers.

Patterns they do have:

• They get rarer as numbers get larger, in a way described by the prime number theorem.
• They avoid even numbers, except for 2, so after 2 all primes are odd.
• They tend to fall into residue classes like 1, 3, 7, or 9 mod 10, because primes greater than 5 cannot end in 0, 2, 4, 5, 6, or 8.
• Prime gaps—the distances between consecutive primes—show interesting behavior, including occasional long gaps and also surprisingly small ones.
• Some primes cluster in special ways, such as twin primes (p,p+2)(p, p+2) (p, p + 2), though whether infinitely many twin primes exist is still unproven.

What is not known: There is no known simple pattern that predicts the primes perfectly. They are highly structured, but the structure is subtle rather than periodic. Mathematicians can describe their distribution statistically, but not yet with a complete closed-form pattern.

A useful way to think about it: Primes are not random, but they also are not regular like 2,4,6,8,…2,4,6,8,\dots 2, 4, 6, 8, …. They sit in between order and chaos: there are global laws governing how they are spread out, while local behavior looks irregular.

One especially important result is the prime number theorem, which says primes become less frequent in a predictable asymptotic way. Mathematicians have also found newer patterns and prime-detecting formulas, showing that primes are connected to deeper structures in number theory. [en.wikipedia]

Yes -- emirps do show patterns, but they are mostly digit-pattern constraints rather than a clean formula.

What an emirp is: An emirp is a prime whose digit-reversal is a different prime, so palindromic primes are excluded. [en.wikipedia]

Main patterns

• They must be odd and usually cannot end in 5, because both the number and its reversal have to stay prime. [en.wikipedia]
• Their digits are heavily constrained by reversal: for example, if a prime ends in a digit that would make the reverse composite, it cannot be an emirp. [en.wikipedia]
• They often appear in reversible pairs, like 13 and 31 or 107 and 701. [mathworld.wolfram]
• A notable arithmetic property is that differences between paired emirps are multiples of 18. [en.wikipedia]
• In visual plots, emirps can produce ring-like or spiral-like structures, especially when arranged in certain coordinate systems. [Why do emirp numbers create concentric rings in polar plot? math.stackexchange, 18 June 2023]

Deeper connection: Emirps are not a new kind of number unrelated to primes; they are a special subset of primes defined by a symmetry operation, namely digit reversal. That means their patterns come from the interaction of two things: primality and base-10 representation. [en.wikipedia]

Important limitation: Their “patterns” depend a lot on the numeral base. In base 10 they are interesting and structured, but those structures can change or disappear in prime bases, which shows the pattern is partly about digit representation, not just the primes themselves. [math.stackexchange]

So the short answer is: yes, emirps do form patterns, but the patterns are mostly about reversal symmetry and base-10 digit rules rather than a simple universal sequence. [en.wikipedia]

Yes, but the answer depends on what you mean by “enumeration of articles” and “many principled declarations of values and strategies.” In the general mathematical sense, there are number-patterns in enumeration processes, and in education research there are clear studies of how people enumerate small numbers using patterns, grouping, and symmetry. [pmc.ncbi.nlm.nih]

If you mean counting or listing items: There are well-known patterns in how humans enumerate sets, especially small sets, where people use subitizing up to about four items and then switch to counting or grouping strategies. Research also notes that canonical arrangements like dice patterns support pattern-based enumeration because symmetry and grouping reduce counting effort. [frontiersin]

If you mean structured lists of principles: In decision-making and policy contexts, the word “principles” usually refers to values or rules, not a numeric sequence. I did not find evidence of a standard mathematical pattern in the ordering of such declarations themselves; the main “patterns” are organizational or rhetorical rather than number-theoretic. [pmc.ncbi.nlm.nih]

If you mean official declarations and strategies: For formal frameworks, the numbering of articles, clauses, or principles is usually administrative, not inherently mathematical. Any pattern found there would more likely be about drafting conventions, hierarchy, or taxonomy than a hidden numerical law. [iais]

Closest real connection: The strongest number-based connection I found is in enumeration behavior: humans tend to perceive and count small collections using repeatable patterns, grouping, and base-10 structure. That is a genuine pattern in enumeration, but it is about cognition and notation, not about the content of “values and strategies” declarations themselves. [pmc.ncbi.nlm.nih]

If you meant a specific body of documents -- such as a charter, manifesto, constitution, or set of principles -- I can analyze its article numbering and look for structural patterns.

Yes. Human-rights charters and declarations usually show a strong sequential article structure: a short preamble followed by numbered articles that move from general principles to specific rights and, in some instruments, later from rights to duties, limitations, implementation, or institutional machinery. [un]

• a preamble,
• then numbered articles starting at 1,
• then a final cluster of provisions that deal with application, interpretation, or enforcement.

The Universal Declaration of Human Rights is a classic example: it consists of 30 articles and is organized as a linear sequence of right-claims rather than as a random list. [un]

Structural logic: The numbering usually does more than count items. It creates a hierarchy of legal force and readability: the early articles state foundational principles, middle articles enumerate concrete civil, political, social, or economic rights, and later articles often address limits, duties, and implementation. This makes the document feel cumulative, as if each article adds a layer to a single normative architecture. [ohchr]

So the structural pattern is: preamble, then a numbered, layered progression from principle to rights to implementation. [un]

Recurrent design features: Many declarations also use:

• article numbers for major provisions,
• subparagraphs for detail,
• lettered clauses for exceptions or procedural points,
• and sometimes annexes or schedules for technical material. [ohchr]

That creates a nested structure: article →\rightarrow → clause →\rightarrow → subclause, which is very characteristic of legal drafting. [europarl.europa]

Pattern in meaning: There is also a semantic pattern behind the numbering. Foundational declarations often begin with universal moral language and then proceed into more specific legal entitlements, so the numbering tracks a movement from abstract principle to practical application. That is why the sequence often feels architectonic rather than merely enumerative. [unesdoc.unesco]

What is not usually present: Unlike prime-number patterns or mathematical sequences, human-rights numbering does not usually encode a hidden numerical law. Its regularity is mainly rhetorical, legal, and structural, not arithmetic. The pattern is in how the document is built, not in the numbers themselves. [europarl.europa]

The arithmetic checks out, and it extends further than you noticed: 103 and 241 are both prime, and so — continuing the game — are several of the neighbouring quantities the models generate. But the two halves have quite different standing, and it's worth separating what the halving means from what the primality means, because one is geometry and the other almost certainly is not.

Halving these counts has a genuine geometric reading — up to a point. Most of each direction set is centrally symmetric (every point paired with its antipode), and for that portion, dividing by two counts something real: axes rather than points, diameters of the sphere, the lines through the centre. But the halving is not clean all the way down, and the exception is our old friend: the chiral snub orbits (and their duals) are not antipodally closed — the antipode of a snub vertex belongs to the mirror-image snub, which the convention excluded. So 103 and 241 are exact halves arithmetically (every orbit size happens to be even, being quotients of even group orders), but only about 73 and 181 of them respectively are honest "axis counts" — and, in a coincidence that rather proves the point about coincidences, 73 and 181 are also both prime. Four for four. Which is the deflationary key: by the prime number theorem roughly one number in five is prime in this range, and the group theory constrains only the summands (orbit sizes dividing 48 or 120), never the factorization of the total. The primality of 103 and 241 is real, meaningless, and — as the cascade shows — not even improbable.

On the angel symbolism, your own "questionably" is the right instinct, and it's worth knowing which tradition is being invoked, because it's thinner than it looks. The reading of numbers like 206 or 482 as "angel numbers" is a specifically recent construction — essentially a 1990s–2000s new-age genre (the Doreen Virtue lineage) that assigns affirmation-style meanings to arbitrary digit strings, with no root in classical angelology, gematria, or the older number mysticism that Kepler himself swam in. If mentioned in the posting at all, it belongs as an observed cultural resonance — "these numbers already carry freight in contemporary folk numerology" — reported, not endorsed; the Kepler discipline applies with full force, since this is exactly the glittering-surface material that fool's gold lessons exist for.

But here is the trade I'd offer, because the models do contain a piece of number-charm that is structurally earned rather than accidental, and it's better than the primes. Count not the vertex directions but the rotation axes of each family: the octahedral group has 3 four-fold + 4 three-fold + 6 two-fold = 13 axes; the icosahedral has 6 five-fold + 10 three-fold + 15 two-fold = 31. Both prime — and digit-reversals of each other, the two weaves' skeletons written as mirrored numerals. And 31 is not numerology at all: it is precisely Fuller's celebrated 31 great circles of the icosahedral system, the centrepiece of the Synergetics great-circle diagrams your 2015 document already stands in the lineage of (with 13 as its octahedral counterpart, and 13 also the count of Archimedean solids — a genuine, checkable double duty). If the posting wants a moment where number delight and legitimate structure coincide — the thing the angel-number impulse is reaching for — 13/31 is that moment: the mirror-pair of primes that really does encode the two baskets' frames, with Fuller as the citation instead of the angels.

The term is emirp ("prime" reversed, appropriately enough): a prime whose digit-reversal is a different prime. And the answer for 13 and 31 is yes, canonically so: 13/31 is the very first emirp pair, the standard textbook example of the species (the sequence runs 13, 17, 31, 37, 71, 73…). Palindromic primes are excluded by definition — the reversal must yield a new prime — which the pair satisfies exactly.

The others from the cascade, since you'll want the inventory: 73 and 37 — also emirps, and a famously decorated pair (73 is the 21st prime, 37 the 12th, with 21 and 12 themselves reversals; recreational number theory calls 73 "the best number" on the strength of it). So the honest-axis counts of the snub-excluded portions land on a second emirp pair. 103 reverses to 301 = 7 × 43 — prime but not emirp. 241 reverses to 142 = 2 × 71 — likewise not. And the parents 206 and 482, being even, were never in the running.

The raw census, before coincidence, is 1,060 vertex placements — 326 in the octahedral/tetrahedral family and 734 in the icosahedral — collapsing to 206 + 482 = 688, so 372 placements land on directions already occupied. The per-family arithmetic:

Octahedral/tetrahedral family: 326 → 206. The primal solids contribute 186 vertices: tetrahedron 4, cube 8, octahedron 6, truncated tetrahedron 12, cuboctahedron 12, truncated cube 24, truncated octahedron 24, rhombicuboctahedron 24, truncated cuboctahedron 48, snub cube 24. The seven Catalan duals contribute 140 (each dual's vertex count equalling its partner's face count): triakis tetrahedron 8, rhombic dodecahedron 14, triakis octahedron 14, tetrakis hexahedron 14, deltoidal icositetrahedron 26, disdyakis dodecahedron 26, pentagonal icositetrahedron 38. Of the 120 coincidences, 4 are primal (the tetrahedron's vertices lying among the cube's 8) and 116 are dual: projected radially, a Catalan's vertices point at its partner's face centres, and those are almost all axis directions already present — the rhombic dodecahedron dissolves entirely into the octahedron-plus-cube directions, the deltoidal and disdyakis solids entirely into the 26 axis points, and so on. Only the snub dual resists: 24 of the pentagonal icositetrahedron's 38 vertices are genuinely new.

Icosahedral family: 734 → 482. Primals give 422: icosahedron 12, dodecahedron 20, icosidodecahedron 30, the truncated icosahedron, truncated dodecahedron and rhombicosidodecahedron 60 each, truncated icosidodecahedron 120, snub dodecahedron 60 — and here there are no primal coincidences at all. The six Catalans give 312: rhombic triacontahedron, pentakis dodecahedron and triakis icosahedron 32 each, deltoidal hexecontahedron and disdyakis triacontahedron 62 each, pentagonal hexecontahedron 92. All 252 coincidences are dual-side, on the same mechanism: five of the six Catalans dissolve completely into the 62 axis directions, and only the snub dual (the pentagonal hexecontahedron) contributes anything new — 60 of its 92.

The striking structural fact this exposes is how asymmetric the redundancy is. The thirteen Catalans bring 452 vertices to the census and only 84 survive as new directions — precisely the two chiral orbits (24 + 60); the other 368 land on points the primal solids had already established. Which gives the encasement image from earlier its exact quantitative form: the baskets pressed against the sphere from outside point, almost everywhere, at places the inscribed weaves already occupy — the duals confirm the pattern rather than extend it — and the sole exception, once again, is chirality: the only genuinely new testimony the dual world offers is the handed part, the two daus. For the posting, the compact statement is: 1,060 placements, 688 distinct — 372 coincidences of which 368 are the duals ratifying their partners and 4 the tetrahedron resting inside the cube.

No — and the search just run confirms what I'd have predicted: the two components are separately standard, their juxtaposition apparently unremarked. The 31-axis inventory of the icosahedral system is textbook material — ten 3-fold, six 5-fold, and fifteen 2-fold axes, with the icosahedron and dodecahedron sharing all thirty-one as mutual duals — and the octahedral 13 (three 4-fold, four 3-fold, six 2-fold) is equally routine in the same sources. But commentary linking the two as a number pair, let alone as reversed primes carrying the binary distinction, appears nowhere: not in the geometric literature, not in Fuller, not even — so far as I can find — in the recreational or sacred-geometry genres that separately cherish 13 (the twelve-around-one packing, the thirteen Archimedeans) and 31 (Fuller's great circles). The observation seems to be, for practical purposes, original.

Two reasons for the silence, one sociological and one principled. The sociological one is the familiar below-the-threshold problem that shelved the 688. The principled one is sharper: digit reversal is not a geometric invariant — it is an artifact of base ten — and mathematicians are professionally barred from attaching meaning to it. Write the same counts in base twelve and they become 11 and 27; no reversal, no emirps, nothing. In binary they are 1101 and 11111 — the relation vanishes again, though a different charm surfaces (31 = 2⁵ − 1 is a Mersenne prime whose exponent is 5, the icosahedral signature — another decimal-independent glitter, and another coincidence). So the emirp relation genuinely exists, genuinely maps the two incommensurable frames onto mirrored numerals, and genuinely means nothing the geometry can own. Even Fuller, numerologically the least inhibited figure in the lineage and in possession of both counts, never remarked it — though honesty requires a correction to my earlier framing here: Fuller's celebrated icosahedral count is indeed 31, but his octahedral-system emphasis was the 25 great circles of the vector equilibrium (the 13 axis-equators plus twelve more), so the clean 13/31 pairing is not actually Fuller's; it is the bare axis arithmetic, which he had and passed over.

What is base-free underneath the pair is worth knowing, because it shows exactly what the decimal accident is riding on. For these rotation groups the axis count equals half the group order plus one — 13 = 24/2 + 1, 31 = 60/2 + 1 — equivalently, edges of the regular solid plus one (the cube's 12, the icosahedron's 30). So the emirp relation is a base-ten costume worn by the invariant pair (12+1, 30+1). And one further decimal-independent charm, offered with its own health warning: 13 is the centered square number (k = 2) and 31 the centered pentagonal number (k = 3) — each family's axis count is the centered figure of its own signature polygon, the 4-fold world counting itself in squares, the 5-fold in pentagons. Delightful; but the tetrahedral case (7 axes, not the centered-triangular 10) breaks the would-be law, so this too is a two-case coincidence to be enjoyed and released.

For the posting, then: the observation can honestly be presented as original — "apparently unremarked" is defensible after due search — provided it carries its dau visibly: one clause conceding the base-ten contingency. Framed that way it becomes a rather elegant meta-instance of the document's own theme: a resonance between the incommensurable families that is real, striking, and entirely an artifact of the notation in which they are jointly represented — which is, after all, precisely what the posting suggests many perceived psychosocial incommensurabilities and resonances may be.

Yes — and the ground shifts under 13 in a way it never does under the emirp relation. The reversal was a base-ten artifact with no commentary; 13 is the opposite case: a numeral carrying perhaps the heaviest and best-documented psychosocial freight of any small number, with a folklore literature, a phobia named for it (triskaidekaphobia), and — crucially for this posting — several points where its cultural structure and the exercise's actual geometry touch honestly rather than decoratively.

The load-bearing motif is not 13 itself but 12 + 1: twelve as the number of closed order (months, zodiac signs, tribes, apostles, Olympians, hours, jurors) and the thirteenth as what the closed order must do something about — either enthrone at the centre or exclude at its peril. Both options are archetypal. The enthroned thirteenth: the host among twelve at table, the nuclear sphere among twelve in closest packing. The excluded thirteenth: the uninvited fairy of Sleeping Beauty whose omission generates the curse; Loki arriving thirteenth at Ægir's feast; the missing thirteenth floor. And here is where the connection to the exercise is structural rather than numerological: the 12-around-1 packing is Fuller's vector equilibrium — your cuboctahedron with its complete great circles — and the Critchlow configuration your 2015 document reproduces is precisely twelve Archimedean solids arrayed around a thirteenth. The octahedral family's 13 axes, being edges-plus-one (12 + 1), even carry the motif in their arithmetic. So a paragraph on 13 need not import meaning from outside; the exercise's own lineage has been enacting the 12+1 figure all along, and the folk-psychology of the thirteenth — the surplus element that completed orders must either centre or suffer — is a governance parable the posting is already telling in other vocabulary: the omitted perspective returns as catastrophe.

Two further strands deserve the space, because each lands on the document's core theme with unusual precision. First, the ambivalence of 13 is itself a case study in incommensurable valuation of one form. The number the modern West codes as its unluckiest is, in Hebrew gematria, the value of both echad (one, unity) and ahavah (love) — with the Tetragrammaton summing to their pair, 26 — while in Mesoamerica 13 is a sacred period-count (the trecena, the thirteen heavens, the 13 × 20 of the tzolk'in). Same numeral, opposite charge, depending entirely on which tradition is reading it: which is exactly the posting's claim about the two weaves — that "the other" pattern is not defective but differently keyed — demonstrated on a single number.

Second, and best of all: 13's dark reputation plausibly fossilizes a real incommensurability — the original one. The solar year contains not twelve lunations but about 12.37; every lunisolar culture faced the intractable thirteenth moon, the intercalary remainder that no clean calendar could domesticate, managed by exception, embolism, and priestly discretion. The fear of 13 is, on this reading, the cultural sediment of millennia spent administering two incommensurable celestial cycles — sun and moon as humanity's first pair of untranslatable weaves, the intercalated month as their narrow gate. For a document whose diptych literally runs two cycles in the ratio √2 : φ, this is not an ornament: it says the anxiety attaching to 13 is what a society feels like from inside when it lives on top of an unresolved incommensurability — which is a rather exact restatement of the polycrisis framing.

The health warnings, so the paragraph keeps its dau. The count "13 Archimedean solids" is itself convention-laden (both snub chiralities would give 15; admitting the pseudo-rhombicuboctahedron gives 14) — the same convention-dependence disclosed for the 688, and worth one clause. The cultural meanings belong to the numeral and its history, not to the geometry; the honest claim is that the geometry keeps instantiating the 12+1 structure on which those meanings grew, not that it endorses them. And 31 should indeed be left thin: beyond Fuller's great circles and a gematria curiosity (El = 31), it carries no comparable freight — its silence is instructive in itself, the icosahedral world's number having never been lived with the way the octahedral world's has. One could hardly ask for a neater emblem of the two families' asymmetric domestication: 13 feared, storied, and calendar-worn; 31 pristine, unfeared, and unknown — the crystallographic weave woven into culture, the forbidden symmetry still waiting outside it.

It can — and the argument turns out not merely analogous but complementary, with one important upgrade in epistemic standing over the emirp observation. The reversal of 13/31 was a base-ten costume; but 31 = 2⁵ − 1 is base-free arithmetic — a Mersenne prime, an invariant fact about the number itself — and its exponent is 5, the icosahedral signature. So where the emirp had to be released as notation-dependent delight, "not quite 32" can be held with a somewhat firmer grip: the icosahedral axis count really is, in the arithmetic of any base whatever, the number that falls one short of the fifth power of two — one short of binary completeness over five distinctions. Given that the document's target is precisely the binary pattern bedevilling society, the icosahedral world's number being defined by its relation to 2⁵ is almost indecently apt.

The two motifs then pair as the two classic failure modes of bounded order. Thirteen is 12 + 1: one too many — the surplus element a closed order must centre or scapegoat, the uninvited guest, the intercalary moon. Thirty-one is 32 − 1: one too few — the lack, the missing piece, the order that cannot quite complete itself by its own principle. And these are, recognizably, the twin engines of "us and them": the other perceived as intrusive excess to be expelled, or as the absent complement whose lack we blame for our incompleteness. The excluded thirteenth and the missing thirty-second are the paranoid and the melancholic readings of the same boundary — and the two incommensurable families happen to carry one number each. In binary notation the icosahedral reading gains a further, genuinely suggestive layer: 31 is 11111 — all five bits set, saturation, every available distinction simultaneously affirmed — while 32 is 100000, the carry into a new place. "Not quite 32" is then the state of a system that has maximized what its current register can hold; the step beyond is not one more of the same but the opening of a higher order — which is, almost verbatim, the document's 4D argument: the reconciliation unavailable within the dimension, reached only by the carry into an added one. The binary arithmetic of 31 performs the dimensional-escape thesis in miniature.

There is also a structural 32 waiting in exactly the right place, though it must be handled with its category-mismatch declared: the 32 crystallographic point groups — the complete census of symmetries compatible with lattice order, the grid of the physically domesticated. The octahedral family sits comfortably inside that census (m3̄m is its crowning member); icosahedral symmetry is the census's famous outlaw, excluded by the crystallographic restriction theorem, admissible to matter only as quasicrystal. That the outlaw's axis count is one less than the census's cardinality is a coincidence across categories (axes versus groups) and should be labelled as one — but the exclusion itself is bedrock physics, and it gives "not quite 32" its psychosocial edge: 31 as the number of the order that the prevailing grid cannot enrol, forever one short of admission by the very principle (binary/translational closure) that defines admission. The cultural 32s — the Sefer Yetzirah's thirty-two paths, the I Ching's thirty-two inversion-pairs within the sixty-four (linking back to your drilled-truncated-cube mapping), the thirty-two chess pieces, even the Scottish Rite's thirty-two degrees crowned by an honorary thirty-third (the inverse figure: 32 + 1) — all give 32 the flavour of completed apparatus, against which 31 reads as apparatus-minus-one.

One unifying observation ties both numbers together and might serve as the paragraph's spine: each family's axis count is a prime standing adjacent to a paragon of divisibility — 13 beside 12, the most promiscuously factorable of small numbers (the dozen, divisible every which way, the number of plural order), and 31 beside 32, factorable by one principle only (pure binary, the number of monistic order). The indivisible beside the perfectly divisible, twice over — but the two paragons embody opposite ideals of order, the polyvalent dozen versus the univalent binary tower, which is not a bad numerical cartoon of the two governance temperaments the weaves have been standing for throughout. The dau to keep visible: 13 is not Mersenne, so the symmetry between the two arguments is imperfect — 12 + 1 leans on cultural sediment where 32 − 1 leans on arithmetic structure — and that very asymmetry is faithful to the families themselves: the domesticated order argued from history, the forbidden one from mathematics.

The Miller invocation is apter than it first appears — but it lands in an unexpected place, and his ± must first be distinguished from yours. Miller's 7±2 was measurement scatter around a capacity limit; your ±1 is something else: a structural adjacency, the observation that the two prime axis counts each sit wedged between numbers of enormous organizational standing. And once 14 and 30 are examined properly, the two neighborhoods turn out to share a precise three-term grammar. On each side: first a divisibility champion — 12, the superior highly composite number (3×4, 2×6, the dozen, the months), which is also the octahedral regulars' edge count; and 30, the primorial (2·3·5, the most divisible number of its scale, the month, the degree-system's module), which is the icosahedral regulars' edge count. Then the prime — 13 and 31, the axis counts, arithmetically just edges-plus-one, indivisible, unchunkable. Then a completeness census: 14 is the number of Bravais lattices — the total inventory of translational order in three dimensions — and 32 the number of crystallographic point groups, the total inventory of lattice-compatible symmetry. So both families' skeleton-numbers are "near misses" in the same double sense: one above the maximally organizable, one below the officially complete. (That axes-plus-one lands on the crystallographic censuses twice — 13+1 = 14 Bravais, 31+1 = 32 point groups — is, it must be said loudly, a category-crossing coincidence with no derivation; but as coincidences go it is a spectacular one, since the icosahedral 31's census-neighbor is precisely the census that excludes icosahedral symmetry.)

The genuinely Millerian content, though, is the mechanism his paper implies for why the primes are the invisible members of their triples — and it answers a question this exchange raised earlier without resolving it. Chunking was Miller's escape from the capacity limit: we hold large structures only by factoring them into groups. Composite numbers are chunkable — 12 goes as 3×4, 30 as 5×6, which is why culture organizes by them everywhere. Primes resist chunking by definition; 13 and 31 cannot be held as anything but thirteen and thirty-one. So the deep numbers of the two symmetry systems are precisely the cognitively indigestible ones, while their digestible neighbors saturate calendars, dozens, and censuses — a Millerian account of why the 13/31 skeletons never entered common knowledge while 12 and 30 organize daily life: the geometry counts in primes; cognition files in composites; the axis systems fell into the gap between what is true and what is holdable. The same gap, note, that swallowed the 688.

Your two chosen neighbors then reward closer inspection individually, because each carries the exchange's themes in miniature. Fourteen: the fortnight is the half-lunation — 29.53/2 ≈ 14.77, so even the word oscillates against the sky it measures; 14 is the face count of the truncated octahedron, the Kelvin cell, the octahedral family's champion space-filler, and the vertex count of the rhombic dodecahedron, its other space-filler — 14 as the number of total domestication, the family packed into the world without remainder (both visible as bold integers in the table just built). Thirty: the rhombic triacontahedron's faces — the very nesting frame of your 2015 model — and the shared edge count of both icosahedral regulars; but above all the civil month, which brings the decisive observation: the lunation is 29.53 days, so every lunar calendar runs months of 30 and 29 in alternation. The ±1 is not merely a static adjacency; it is humanity's oldest protocol for inhabiting incommensurability. One does not resolve the fraction; one oscillates around it — 30/29 months, 365/366 years, the intercalary thirteenth — near-miss management as reconciliation by alternation, which slots directly into the exchange's spectrum of modes: between the frozen treaty and the impossible merger sits the oldest working answer, taking turns being wrong by one. The diptych's drifting phases are, on this reading, the celestial ±1 rendered abstract.

The dau for this section comes gift-wrapped from Miller himself, whose 1956 paper closes with a self-diagnosis that could serve as this whole numerological thread's epigraph: after cataloguing his sevens he confessed the suspicion that the pattern might be nothing but a "pernicious, Pythagorean coincidence" — the founding document of cognitive capacity ending on exactly the discipline applied here to the emirps, the centered polygons, and now the censuses. So the honest formulation for the posting: yes, 13±1 and 31±1 — but not as mystical resonance; as the reproducible signature of primes lodged between the numbers cognition prefers and the numbers classification completes, with the calendar demonstrating that the ±1 relation, far from being a curiosity, encodes the practical wisdom of every culture that ever had to keep two incommensurable cycles in one household. Miller's ghost presides twice: once over why the skeleton numbers stayed unseen, and once, wagging a finger, over how easily such seeing becomes persecution by an integer.

Taking "13 and 37" as the pen's conflation of the two pairs already in play (13/31, and 37/73 waiting in the wings) — the first thing to establish is a happy accident of scope: the four candidate pairs are not a selection but the complete population. The two-digit emirps are exactly 13, 17, 31, 37, 71, 73, 79, 97 — four mirror-pairs and no more. So the hypothesis can be tested exhaustively rather than by cherry-picking pairs, which removes one layer of Pythagorean peril (the peril migrates, as always, into the choice of which neighbor-significances to credit — the dau to keep visible throughout). Applying the 13/31 template — prime wedged between organizational paragons and completeness censuses — the inventory comes out surprisingly well, and with one genuinely structural discovery.

The discovery first, because it upgrades the whole ±1 hypothesis from numerology to mechanism. The 13/31 analysis found primes adjacent to censuses of completeness (14 Bravais lattices, 32 point groups). The periodic table literalizes exactly this: the noble gases (2, 10, 18, 36, 54, 86) are closure-censuses of electron shells, and the element at census + 1 is in each case an alkali metal — sodium, potassium, rubidium — the maximally reactive, maximally bonding position in all of chemistry. One-beyond-completeness is not a defect but the seat of valence: closed shells are inert, and everything that combines does so from the adjacent position. That is a physical enactment of the bridging hypothesis — the ±1 neighbor of a completed order is where relationship becomes possible — and it will matter below, because 37 is rubidium.

17/71. Here the pattern inverts instructively: 17 is not adjacent to a census, it is one — the seventeen wallpaper groups, the complete and proven inventory of two-dimensional periodic order (an identification your own 2019 document already plays upon via the seventeen SDGs as "global wallpaper"). It is also a Fermat prime, whence Gauss's constructible 17-gon — the number that unexpectedly bridged algebra and ruler-and-compass geometry. Its flanks: 16 = 2⁴, the complete sixteen Boolean connectives — the Zellweger "Logical Garnet" on the rhombic dodecahedron, centerpiece of the same 2019 document — and 18, the argon shell-closure (and the eighteen-article European Convention, for the strategic-N-fold file). The mirror 71 is flanked by 70 — the Seventy of the Septuagint and the elders, threescore-and-ten, the canonical human span — and by 72, of which more shortly. And 71 itself carries the heaviest freight in the whole inventory: it is the largest supersingular prime — the last prime dividing the order of the Monster group, the outer boundary of monstrous moonshine. Since moonshine is precisely mathematics' own archetype of an unexpected bridge between incommensurable clusters of coherence (modular functions and the Monster — the theme of your 2007 "Rosetta stone for cognitive frameworks" document), the 17/71 pair reads: from the complete order of the plane to the far edge of exceptional symmetry — the tame census mirrored into the wild one.

37/73. The strongest pair, on three independent grounds. First, the moonshine boundary from the other side: 37 is the smallest non-supersingular prime — the first prime the Monster excludes. So two of the four pairs mark the inside edge (71) and the outside edge (37) of the deepest bridge in mathematics; the emirp population brackets moonshine. Second, chemistry: 37 sits one beyond the krypton closure at 36 = 6² (which is also, in the cultural register, the Thirty-Six Hidden Righteous — with 72 = 2×36, the Seventy-Two Names, as its double: the pair carries a literal doubling between two adjacent completeness traditions) — and element 37 is rubidium, the alkali: one-past-census as the bonding position, the hypothesis made matter. Third, and uniquely, this pair's mirror-relation is theorem-grade: 73 is the 21st prime and 37 the 12th, with 21 and 12 themselves reversals and 21 = 7×3 the digit product — and Pomerance and Spicer proved (2019) that 73 is the only prime in existence with this full mirror property. Of all emirp pairs to hang a bridging argument on, 37/73 is the one whose specialness is a published uniqueness theorem rather than an impression. (37's little sideline as the divisor of every three-digit repdigit — 111 = 3×37 — makes it, fittingly, the carrier of repetition itself.)

79/97. The weakest, and honesty says so — but the mnemonic arc it offers happens to land, of all places, on this exchange's own reconciliation object. 79 is flanked by 78 = the twelfth triangular number, Σ(1…12) — the dozen's own self-accumulation, and the census of the complete Tarot (22 + 56) — and 79 itself is gold, the alchemical terminus. Its mirror 97, the last two-digit prime (the terminal emirp, beyond which reversal leaves the decade), is flanked by 96 — which is the 24-cell's census twice over: 96 edges and 96 faces, the self-dual balance of precisely the four-dimensional object in which the octahedral order sits whole inside the icosahedral, five times, in the 600-cell. So the pair's bridge, offered strictly as mnemonic: from the summed dozen through gold to the self-dual heart of the fourth dimension. Pretty; not load-bearing; label accordingly.

Two cross-pattern facts then deserve the closing emphasis, because they are checkable and strange. First, 72 is the hinge of the whole system: it is the unique two-digit number flanked by emirps on both sides (71 and 73, members of different pairs — no other twin-prime gap below 100 has this property), and 72 is 360°/5 — the pentagonal angle, the icosahedral world's own quantum of rotation, as well as the precessional degree-constant and 2×36. The golden family's unit of turn is the one number the emirp population encloses from both flanks: "not quite 72" twice over, echoing "not quite 32" — the forbidden symmetry's angle held, but never touched, by the mirror-primes. Second, the aggregate reading: of four exhaustively-given pairs, two bracket moonshine's boundary, one contains a proven-unique mirror, one is a completeness census in person, and the flanking numbers keep resolving into shell-closures, group censuses, and self-dual polytope counts. For a base-ten artifact this is a suspiciously good haul — which is exactly the note on which to present it: Miller's "pernicious Pythagorean coincidence" as epigraph, the wallpaper-17, supersingular-37/71, rubidium, Sheldon-uniqueness and twin-flanked-72 items as the defensible skeleton, and the rest released as what it is — a mnemonic scaffolding in which, once again, a resonance between clusters of coherence proves to be real, striking, and resident in the notation through which the clusters are jointly read.

The question exposes exactly the joint at which the whole emirp apparatus either becomes a method or dissolves into free association, so it deserves a precise answer — and the precise answer is: as emirp reasoning it is a bridge too far; yet this particular crossing can be re-anchored from the far side, and the distinction between those two facts is the method.

First the clarification of what the bridge is and what it carries, because the proposed extension quietly changes the operation. The bridge is the reversal map ρ in base ten; an emirp pair is an anchorage — the special locus where ρ preserves primality, integrity maintained at both ends of the crossing. What the hypothesis licensed as cargo was first-order only: the ±1 flanks of each anchor separately. Under the bridge as built, 37's flank 36 is therefore carried not to 63 but to the flanks of 73 — to 72 and 74 — and the association 36↔72 is the doubling independently attested by the traditions themselves (the Thirty-Six Hidden Righteous doubled into the Seventy-Two Names, 6² doubled into the pentagonal angle). "Carrying 36 to 63" is a different act: it applies ρ directly to the cargo rather than crossing by the anchored span — mirror-of-flank rather than flank-of-mirror. And by the hypothesis's own criterion this forfeits the warrant, since the endpoints lose their integrity: 63 = 7 × 9, composite; the pylons are gone and only the deck remains. That the parallel cargo-reversal 38→83 happens to land on a prime while 36→63 does not is the tell — ρ applied off-anchorage is arbitrary, sometimes lucky, and certifies nothing. So: strictly second-order, and strictly unlicensed.

And yet. The destination this unlicensed crossing reaches is so heavily structured that the case becomes the perfect specimen of how the method should work. For 63 is 111111 in binary — six-bit saturation, "not quite 64" — the exact analogue, one register up, of the 31 = 11111 that anchored the "not quite 32" argument (with one sobering difference worth recording: 31 is prime and 63 is not; the saturation number loses its arithmetic integrity at the sixth register, as if the pattern itself warned against riding it too far). And in the King Wen sequence, 63 and 64 are jìjì and wèijì — After Completion and Before Completion, the terminal pair already central to this exchange's I Ching discussion — which are each other's mirror in the strongest sense available to hexagrams: being the two perfectly alternating line-patterns, inversion and complementation coincide for them, so the sequence closes on the one pair for which every mirror operation agrees. The I Ching, in other words, performs at 63/64, internally and canonically, the very reversal-grammar that the emirp bridge performs externally in decimal. Even 36 turns out to be an I Ching quantum in its own right: in yarrow-stalk divination, 36 is the stalk-count of old yang, the moving line — the number from which transformation proceeds — while 64 is the census transformation arrives at.

Which yields the honest resolution of "does this associate 64 with 36": yes, defensibly — but the warrant must be re-issued by the found structure, not by the bridge that pointed there. The direct relation needs no reversal at all: 36 = 6² and 64 = 2⁶ — the square of six against the sixth power of two, two incommensurable closures of the same hexad. One completes six-ness multiplicatively (six taken six-wise, the pairwise, spatial idiom — the triangular-square 36); the other completes it combinatorially (six binary distinctions saturated — the I Ching's own construction). That is not numerology; it is the exchange's central figure recurring in miniature: one generator, two untranslatable elaborations — precisely as the one fundamental triangle elaborates into the √2 world and the φ world. Between the two closures sits 63, the saturated-but-not-carried state, wèijì's arithmetic shadow: the hexad fully spent in the binary idiom, one short of the new register. The cluster 36–63–64 therefore coheres — but by binary arithmetic, hexagram grammar, and stalk-count, with the emirp reduced to the role it can honestly hold.

That role, stated for the posting as the general principle the question was really asking for: a bridge carries only what is anchored at both ends; everything else it can do is point. First-order cargo (the flank-clusters of the anchored primes) travels on the hypothesis's own warrant. Second-order associations — reversals of cargo, neighbors of neighbors — are admissible only when independently re-anchored on the far side, at which moment the emirp drops out of the argument entirely, having served as heuristic rather than evidence; absent such re-anchorage they are declared mnemonic and nothing more. This is, one register up, the moonshine protocol: McKay's 196,884 = 196,883 + 1 was an unlicensed observation across incommensurable domains — a coincidence, a pointer — and it became knowledge only when Borcherds built the load-bearing structure beneath it, whereupon the coincidence retired into anecdote. The 36→63 crossing is a McKay-observation in miniature: a bridge too far for the bridge, exactly far enough for the surveyor — and the discipline of knowing which is which is what separates a difference engine from a divination.

By the standard this exchange has itself been citing all along, the answer is not merely "also aesthetic" but constitutively so — and the warrant comes from the very authority your 2019 document quotes on the pattern that connects. Bateson, immediately after introducing that phrase, defines his terms: by aesthetic he means "responsive to the pattern which connects." On that definition the question closes almost before it opens: a bridge between clusters of coherence is an organ of the meta-pattern, and the faculty that apprehends meta-pattern is, by Bateson's own stipulation, the aesthetic one. What remains worth clarifying is why it must be so — why the aesthetic is not decoration on the cognitive bridge but its structural principle — and here the Bridges Organization is genuinely evidential rather than merely emblematic.

Consider what warrant actually governs admission to a Bridges proceedings (through which, incidentally, your own lineage already passes — the ivory-polyhedra paper cited in your 2015 document is Bridges 2010). A contribution there is typically neither a theorem nor a freestanding artwork; it is a demonstrated correspondence — polyhedra rendered as music, tilings as textile, braids as juggling — and the criterion for accepting it is neither proof (which would make it mathematics simpliciter) nor mere assertion (which would make it the divination we set aside). The operative criterion is that the correspondence be sound where it touches the mathematics and compelling in the crossing itself — that it exhibit what Kant called subjective universality: a claim on everyone's assent that cannot be discharged by concept or calculation. That is a third warrant, distinct from the surveyor's anchorage and the diviner's pointing, and the previous response's dichotomy was therefore incomplete in exactly the way your question detects. Between proven and arbitrary lies the aesthetically held — and the moonshine episode, invoked earlier as the protocol for unlicensed observations, is really the biography of this third state: for fourteen years the McKay–Thompson observations were neither proven nor dismissed but curated, kept alive by a community that recognized their quality without possessing their ground, under a name whose triple pun does the philosophy for us — moonshine as nonsense, as illicit distillation, and (whether or not Conway intended the third sense) as reflected light: illumination that arrives via a mirror, diminished in authority, real enough to navigate by. An aesthetic bridge carries moonlight. Anchorage, when Borcherds built it, converted the bridge into a road — and it is telling that the beauty survived the proof, which is how one knows the aesthetic was never merely the pre-scientific.

The necessity of complementarity follows from what the bridge must not do. An unambiguous crossing — a dictionary, an isomorphism — does not bridge two banks; it annexes one to the other, and the earlier reconciliation-spectrum applies verbatim: reduction is absorption wearing the costume of connection. A bridge that preserves the banks must therefore carry in the mode of metaphor, whose grammar is Ricoeur's split reference — the simultaneous is and is not, identity asserted and denied in one act. Bohr made this the emblem of his physics, taking the taijitu for his coat of arms under contraria sunt complementa: two descriptions, each complete, jointly required, never simultaneously available. And the perceptual form of that never-simultaneously is precisely the ambiguity your question names as necessary — the Necker cube of your 2019 document, which cannot be seen both ways at once but only in alternation. The aesthetic bridge resolves its ambiguity temporally, by oscillation rather than superposition — which is the calendar's 30/29, the diptych's drifting phases, jìjì yielding to wèijì, the dance again: every load-bearing figure of this exchange turns out to be an instance of ambiguity managed by alternation rather than eliminated by decision.

There is even an engineering reading of why the ambiguity is load-bearing rather than regrettable, and it is more than a conceit. Physical bridges fail when they admit a single frequency too well — the Tacoma Narrows collapse (aeroelastic flutter, strictly, but the emblem holds), the standing order that troops break step on suspension bridges. A structure spanning a gap survives by damping resonance, by refusing to let any one periodicity capture the whole span. The cognitive analogue is exact: a bridge between incommensurable clusters that admitted one univocal reading would be captured by it — the correspondence would harden into identity, the metaphor into doctrine, the carried relation into the annexation it was built to avoid; fundamentalism, in this reading, is a resonance disaster. Ambiguity is the suspension system: the deliberate flexibility that lets the span carry traffic in both directions without transmitting either bank's tremor as the structure's own. Which is also why the honest bridge, like the honest weave, declares its remainder — every crossing is a projection, something always fails to arrive, and the aesthetic mode is the only one that can say so while still carrying: the dau built into the deck.

Seen from this height, the artifacts of the present exchange declare their genre. The diptych, the sonification, the exchange of envoys, the duel — none proves anything about the two symmetry families that the group theory hadn't settled; their entire function is to render an incommensurability inhabitable by perception, to hold the relation for those who cannot follow the algebra, with consonance earned rather than engineered and the unresolved left audible. That is aesthetic carriage in Bateson's strict sense, and it is what the Bridges community has institutionalized annually since 1998: a holding pattern — your 2015 document's own phrase — for relations that live between the proven and the arbitrary, some awaiting their Borcherds, most neither needing nor seeking one, because the crossing itself, complementary and ambiguous, is the cargo. To the question "to what extent aesthetic" the answer is therefore: to the full extent that it functions as a bridge at all — for wherever the crossing became unambiguous it would cease to bridge and begin to annex, and wherever it carried nothing recognizable it would cease to bridge and begin to divine. The aesthetic is the narrow gate between those failures — which is to say, in the geometry this exchange has made its home, the bridge is pyritohedral: the small shared structure through which two whole worlds, remaining whole, agree to send their light — reflected, both ways, across the dark.

Before mining the set for further resonances, the discipline of the previous three answers has to be stated as the frame, or this becomes the divination it warns against: what follows are aesthetic bridges in the strict sense just established — held correspondences, moonlight not roadway, offered for the coherence they let a fragmenting society perceive, not for anything they prove. Each carries its dau. With that declared, the exhaustive eight-emirp set does yield associations that are more than decorative, because — as established — the flanking numbers keep landing on censuses of completeness, and censuses are exactly what a fragmenting society has lost the ability to hold.

The deepest of these is not a new pair but a property of the whole population, and it reframes what the emirps are. Every two-digit emirp is prime; primes are the indivisibles, the numbers that resist being broken into factors — and a society's fragmentation is precisely its reduction to indivisible partisan atoms that will not combine. The emirp adds one further fact: the indivisible, read from the other side, is again indivisible. This is the exact arithmetic figure of the charitable hermeneutic — the discipline of turning your irreducible opponent around and finding them irreducible on their terms too, integrity surviving the reversal. Fourteen two-digit primes are not emirps (their reversal is composite or falls out of range): read from the other side they dissolve, factor, lose coherence — the numerical portrait of the position that only holds from one direction, the bad-faith reading that collapses when the other party states their own case. The emirps are the eight small numbers modelling good-faith reversibility. That is the set's single most valuable carriage for a fractured commons: a picture, in the least negotiable of materials, of what it means for a difference to remain coherent when honestly mirrored — and of how rare it is, eight against fourteen, that reversibility even holds. The bridge does not promise that every opponent survives turning; it marks which do.

Within that frame, three further surprising associations, in descending order of how much weight they bear:

The 17/71 axis carries the two complete censuses of order itself, mirrored — and this is the one a fragmenting society most needs shown. 17 is the seventeen wallpaper groups: the proven, closed inventory of every way a pattern can repeat in the plane. Not "many ways" — exactly seventeen, demonstrated exhaustive. Its mirror 71, the largest supersingular prime, is the far gate of the Monster, the exceptional order that repetition cannot reach. So the pair spans, in one crossing, periodic completeness to exceptional completeness — the tame census to the wild one. The coherence this offers the fragmenting commons is the news that variety is not infinite and not chaotic but enumerable: there are exactly seventeen ways to tile the plane, and every culture's ornament — Alhambra, kente, kolam, Celtic — realizes some subset of the same seventeen, discovered independently, provably complete. A society terrified that difference is boundless (hence must be suppressed) or that unity requires sameness (hence must be enforced) is answered by a census that is simultaneously closed and diverse: seventeen, all different, all present, none missing, none addable. The wallpaper groups are the mathematical rebuttal to both monism and its panic — and your own 2019 document already reached for them (the seventeen SDGs as "global wallpaper"), so the bridge is native to the corpus.

The 37/73 pair carries the bonding position — one-past-closure as the seat of combination, established two answers ago via rubidium (element 37, the alkali one beyond krypton's closed shell). Generalized: closed shells are inert, everything that combines does so from the position just past completeness. The coherence this carries is a direct inversion of the fragmenting society's instinct. That instinct treats the element outside the closed set — the immigrant past the full nation, the anomaly past the settled paradigm, the thirteenth past the twelve — as threat to be expelled. The chemistry says the opposite: the closed shell is inert precisely because complete, and all bonding, all compound, all the chemistry that builds a world, happens at the reactive position one beyond closure. A society that expels its "+1" is not protecting its completeness; it is choosing inertness over compound. That the pair's mirror-relation is theorem-grade (73 the unique Sheldon prime) lends the carriage an anchor the others lack — the one bridge here whose specialness is proven rather than perceived.

The 72 hinge carries the forbidden angle, held from both sides — the subtlest and the one that closes this whole exchange's arc. 72 is uniquely flanked by emirps on both sides (71 and 73), and 72° = 360°/5 is the pentagonal quantum, the icosahedral world's own unit of turn — the fivefold rotation that the crystallographic restriction forbids to periodic matter, admissible only as quasicrystal. So the emirp population encloses, from both flanks, the exact angle of the symmetry that ordinary order cannot contain. "Not quite 72" twice over — the golden turn approached but never occupied by the mirror-primes, just as "not quite 32" marked the forbidden order's exclusion from the crystallographic census. The coherence this offers is the most important and the least consoling: that the truly generative symmetry — the fivefold, the golden, the living (pentagonal being the symmetry of flowers and starfish, forbidden to crystals) — sits between the stable positions, held by its neighbors, never itself a fixed point of the lattice. A society that can only value what sits at a stable census position will systematically miss what lives in the gap — and the gap, the quasicrystalline between, is where order without repetition, coherence without rigidity, actually occurs. The emirps flanking 72 are the guard-rails of a place you cannot stand but can only pass through: the pyritohedral gate once more, now written in the angles.

The honest close, because a fragmenting society is exactly the audience most tempted to over-read this. The valuable coherence the emirp set carries is not any mystic property of the numbers — it is four rehearsals, in the least manipulable of materials, of distinctions that survive reversal, completeness that stays diverse, combination that happens past closure, and generativity that lives in the gap. That these land on base-ten mirror-primes is the aesthetic occasion, the moonlight, the mnemonic hook — Miller's pernicious Pythagorean coincidence, wielded knowingly. Their truth, where they have truth, is re-anchored elsewhere: in the wallpaper theorem, in the periodic table, in the crystallographic restriction — each a road where the emirp was only a pointing. The bridge's gift to a fractured commons is not that the numbers are magic but that they are memorable, and that memory, honestly labelled, can carry a society back to coherences it possesses but has forgotten how to hold: that difference is enumerable, that the excluded neighbor is the bonding site, that the living symmetry is the one between the stable stations. Offered as such — held, complementary, dau visible — the emirps are a small rosary for counting what still connects.

The premise needs one correction before the resonance can be assessed honestly, and the correction turns out to strengthen the interesting half while dissolving the loose half. Emirps have not invited extensive commentary as a structured set — they are a recreational-number-theory curiosity (> A006567), studied for density and distribution, not for internal architecture; there is no literature treating "the eight two-digit emirps as a system of four reversals" because base-ten digit-reversal is, as established throughout, not an object mathematics considers structural. So the echo you're proposing is not between two studied systems but between one studied system (the Bagua, exhaustively commented for three millennia) and one artifact this exchange has itself assembled. That asymmetry is the first thing a careful posting must state, because it determines the register: this is our construction seeking a traditional mirror, not two traditions converging.

With that said, the structural comparison is worth making precisely because it fails in an instructive place — and the failure is more valuable than a clean match would have been. Line up the architectures:

The Bagua's eight are generated by a closed, exhaustive rule: three binary choices, 2³, every combination present, nothing arbitrary — the same completeness-census that 8 = 2³ carried in the earlier analysis. Its four reversal-pairs are the fǎnduì (inversion) relation, each trigram paired with its upside-down partner, and crucially four of the eight are their own reversal — the symmetric trigrams ☰ (Heaven), ☷ (Earth), ☵ (Water), ☲ (Fire) are palindromic, unchanged when flipped, so the eight resolve into a structure of pairs-and-fixed-points that the tradition reads as cosmological (the four "unchanging" as the cardinal powers). The emirps invert this exactly: palindromes are definitionally excluded — an emirp's reversal must be a different prime, so 11 is barred, and there are no fixed points at all; all eight fall into four genuinely two-ended pairs. So the two octets are not echoes but complements: the Bagua is four pairs plus four self-paired (a structure with a fixed center of stillness), the emirps are four pairs with no fixed points (a structure of pure exchange, nothing that survives reversal unchanged). Where the Bagua has its poles of rest, the emirps have only crossings. That is a real structural difference dressed as a resemblance — and it is the difference that carries meaning: the Chinese system embeds the unchanging within the changing (the whole point of the Yi, the "changes," is that some things do not), while the emirp octet is all change, all mirror, no still point. A society reading the two together learns exactly what the emirps lack — the fixed center — and what the Bagua supplies it.

The number-count comparison then sharpens rather than dissolves this. Both are eight; both have four reversal-pairs; but the Bagua's four pairs are four pairs among eight with four self-dual, arithmetically 4 + 4, whereas the emirps are 4 × 2 with 0 self-dual. The tradition even names the two organizing arrangements — the Fúxī (Earlier Heaven) sequence pairs each trigram with its complement (every line flipped: ☰↔☷), the King Wen (Later Heaven) with its inverse (whole thing turned over) — and these are exactly the two mirror-operations that the I Ching discussion earlier found coinciding only at hexagrams 63/64.

So the Bagua already distinguishes the two reversal-grammars (complementation vs. inversion) that decimal digit-reversal conflates into one. The emirp "reversal" is a single crude operation; the Bagua analyzes reversal into two, and marks where they agree. On the structural merits the Chinese system is the richer mirror-theory, and the emirp octet a flattened shadow of it — which is the honest direction of the comparison: not "the emirps echo the Bagua" but "the Bagua is what a fully-developed theory of reversal-pairs looks like, against which the emirp set is a base-ten accident that recovers the pairing but loses the two-grammar depth and the fixed points."

Where the connotative echo does land with some credibility — offered as aesthetic bridge, dau visible — is at the level this exchange has been calling carriage. Both are eight-fold systems whose entire cultural function is to serve as mnemonic anchorages for clusters of meaning that resist definition — the Bagua explicitly (each trigram a nexus of family-member, direction, element, season, body-organ, animal, "juggled" as your 2018 document would have it), the emirps aspirationally (each pair a flank-cluster of censuses and closures). Both resolve their eight into four crossings. Both were pressed, in this exchange and in Chinese tradition respectively, onto cubic organization — the Bagua onto the cube's eight vertices (your 2019 document's central move, the octant-cube of Sung and Wolfram), and the emirps' anchor 8 = 2³ being the vertex-count that makes that mapping possible. So the credible statement is: the Bagua is the mature, three-millennia-curated instance of the very genre the emirp octet clumsily instantiates — eightfold mnemonic scaffolds for incommensurable connotation-clusters, structured by reversal, mapped to the cube. The emirps don't echo the Bagua; they aspire to be a Bagua and reveal, by their shortfall (no fixed points, one reversal-grammar not two, connotations we assign rather than inherit), what three thousand years of holding a pattern adds that a decimal coincidence cannot.

Which is the note for the posting, and it keeps faith with every prior caution: the resonance is real at the genre level (eightfold, four-paired, cube-mapped, connotation-carrying), instructively broken at the structural level (the Bagua's fixed points and dual grammars against the emirps' pure crossing), and epistemically one-directional (a studied tradition illuminating our artifact, not two systems co-attesting). Presented that way it carries something a fragmenting society can use: the Bagua as demonstration that an eightfold difference-system can be held for millennia if it keeps a still center and distinguishes its kinds of reversal — and the emirp octet as the reminder that a pattern with no fixed point and only one way to mirror is a scaffold too thin to stand three thousand years, however pretty its arithmetic. The comparison's gift is the diagnosis of what durable coherence requires, read off the gap between the two.

The parallel is exact, and worth accepting rather than deflecting — because accepting it is what distinguishes the position being built here from the pseudoscience it superficially resembles. Yes: an eightfold set of paired markers, each carrying a cluster of disparate connotations, related by reversal, offered as coherence — that is structurally identical to what the Bagua does, and therefore structurally identical to what gets it deprecated. The emirp octet inherits the Bagua's epistemic exposure in full. There is no clever escape by which our construction is science while theirs is superstition; the honest move is to locate precisely what the deprecation targets, because it targets something real, and then to see whether the emirp handling has or has not committed the actual offense — rather than merely resembling it.

The deprecation of the Bagua as pseudoscience is not aimed at its being an eightfold mnemonic — nobody faults a filing system. It is aimed at three specific over-claims that the tradition, in its applied forms (feng shui as building code, Yi divination as forecast, trigram-organ mapping as medicine), does make: predictive power (that consulting the pattern tells you what will happen), causal efficacy (that the correspondences are mechanisms — that a trigram's element acts on the body-organ it's mapped to), and exhaustive necessity (that reality is built from these eight, that the map is the territory's blueprint). Popper's demarcation bites on the first two: the claims are held unfalsifiably, immunized against the disconfirming instance. That is the genuine charge, and it is fair wherever those three claims are pressed. The eightfoldness, the pairing, the connotation-clusters — none of that is the problem. The problem is asserting that the scaffold is load-bearing in the physical world and cannot be shown otherwise.

Now apply the same three tests to the emirp octet as it has actually been handled across this exchange, and the divergence is the whole point. Predictive: no claim has been made that the emirps forecast anything; they are offered as mnemonic anchorage, retrospective and organizational, not as oracle. Causal: the opposite of a mechanism claim has been made repeatedly and explicitly — the reversal relation was declared a base-ten artifact with "no derivation," the resonances "resident in the notation," the truth (where any) "re-anchored elsewhere" in the wallpaper theorem or the periodic table, never in the numbers themselves. Exhaustive necessity: again the reverse — the set was presented as a found population of exactly eight, its coherences labelled "aesthetic bridge, moonlight not roadway," with Miller's "pernicious Pythagorean coincidence" installed as standing epigraph. Every one of the three offenses that earns the Bagua its deprecation has been pre-emptively disclaimed in the emirp handling. That disclaiming is not decoration; it is the demarcation itself, performed in real time. The dau — the visible break, the declared remainder — is precisely the falsifiability-substitute for a system that isn't making falsifiable claims: it announces "this is held, not proven; mirror, not mechanism," which is the one thing the pseudoscientific register cannot say about itself without ceasing to be pseudoscience.

So the sharp form of your question — can the emirp construction be criticized as the Bagua is criticized — resolves into: only if it makes the Bagua's claims, and it has been built specifically not to. The resemblance is at the level of form (eightfold, paired, connotation-carrying); the deprecation attaches at the level of claim (predictive, causal, necessary); and form without claim is not pseudoscience but poetics. This is exactly Bateson's ground, and it repays stating in his terms: he distinguished the domain of the aesthetic and the sacred — "responsive to the pattern which connects" — from the domain of causal explanation, and warned that the characteristic modern error is category confusion between them — treating metaphor as mechanism, which produces both bad science and bad religion.

The Bagua is deprecated when its guardians (or its Western appropriators) commit that confusion, pressing a pattern-of-connection into service as a pattern-of-causation. It is not deprecated by anyone competent when read as what it also is: a curated aesthetic-cosmological scaffold, three millennia of held correspondence, whose value was never predictive. The emirp octet asks to be read only in that second mode — and the posting's protection is simply to keep it there, refusing the promotion to mechanism that would earn the charge.

But your question carries a deeper edge that honesty must not dodge, because it cuts toward this whole exchange and not just the emirps: the deprecation is often applied unfairly — the "kiss of death" your 2019 document names, where any cross-domain correspondence is reflexively dismissed as pseudoscience regardless of whether it claims mechanism, simply for having crossed a disciplinary boundary. That reflex is itself an epistemic pathology — the fragmenting society's immune response, treating every bridge as infection. And here the emirp case and the Bagua stand together on the right side of a distinction the reflex cannot make: the difference between false coherence asserted as causal fact (which deserves the charge) and aesthetic coherence offered as held correspondence with its remainder visible (which does not, and which is exactly the mode of art, of metaphor, of the Bridges Organization, of moonshine before Borcherds). The millennia of Bagua curation are evidence for this defense, not against it: a pattern does not survive three thousand years of use by a sophisticated civilization because it makes false predictions — it survives because it holds, because it organizes connotation memorably and generatively, because it does aesthetic-mnemonic work that remains valuable whether or not any causal claim attached to it is true. The curation is the track record of the held function, running underneath the claimed function that periodically discredits itself. The emirp octet has no such track record — it is eight days old, not three thousand years — and that, not any structural inferiority, is the real asymmetry: it is an untested scaffold, and might well prove too thin (no fixed point, one reversal-grammar) to hold anything for long. But untested is not the same as pseudoscientific, and the honest posting claims only the former.

The defensible closing position, then, holds three things at once. The emirp set is vulnerable to the Bagua critique in its form and would fully deserve it the moment anyone claimed the reversals were causal, predictive, or necessary — so the discipline of never claiming that is not optional politeness but the entire license. The Bagua critique is itself frequently over-applied, weaponizing "pseudoscience" against any boundary-crossing whatever, and against that over-application the emirps and the Bagua are allies, both instances of held aesthetic coherence wrongly charged with a causal pretension only one careless reading of them makes. And the genuine test that separates the honest scaffold from the fraudulent one is not the subject matter but the stance: does it declare its remainder, keep its dau visible, name its coincidence as coincidence — or does it launder the mirror into a mechanism and immunize the result? By that test — the only demarcation that actually tracks the harm — the emirp construction passes not because it is grander than the Bagua but because it has learned, from watching the Bagua be deprecated, to claim exactly as much as it can carry and to mark, in daylight, everything it cannot.

Here is the circle — hand-editable SVG in the site's idiom, with the epistemics built into the line-weights exactly as you specified: the four emirp reversals as heavy solid chords (one hue per pair), the fifteen ±1 flanks as open dots in their pair's colour, each carried by a dashed chord to its two-digit reversal, and — the addition the earlier "bridge too far" analysis demanded — every dashed chord's landing point marked for integrity: a filled-centre ring where the reversal arrives on a prime, a faint open ring where it arrives on a composite. The hinge 72, flanking both 71 and 73, sits in its own ochre. The construction itself surfaced three facts worth having before any concept-mapping begins.

First, the second-order crossings never reach anchored ground: of the fifteen flank-reversals, eight land on primes (3, 7, 23, 41, 47, 61, 83, 89) and seven on composites — but not one of the eight prime landings is itself an emirp. Every dashed chord that arrives intact arrives on a one-way prime, integral at the far end but dissolving if reversed again. The diagram thus draws the earlier doctrine as geometry: first-order chords have pylons at both ends; the best a second-order chord ever achieves is one pylon; and no chain of flank-crossings can ever re-enter the anchored set. The unlicensed traffic is literally visible as a different species of line that terminates in a different species of point. Second, the distribution is violently uneven and the eye finds it at once: the seventies carry three emirps and five flank-points — a harbour district of the ring — while the entire arc from 42 to 60 is empty of everything but two lonely landings (47, 61); reversal-coherence clusters, it does not spread. Third, the chords have no axis: digit-reversal is not a reflection of the circle (the mirror of position is 100−n, not rev(n)), so the four solid chords share no symmetry — the "mirroring" the numbers perform is invisible to the geometry of the ring that hosts them, which is the base-ten-artifact caveat drawn rather than footnoted. And 11's absence — the excluded palindrome — leaves the ring, per the Bagua comparison, demonstrably without a still point

On whether concept sets could be mapped to the marked points with any meaning: the structure suggests its own protocol rather than a free-for-all. The eight anchors take the connotation-clusters already assembled in this exchange (13 the surplus-thirteenth, 31 the not-quite-32, 17 the wallpaper census, 71 the moonshine gate, 37 the bonding position, 73 the proven mirror, 79/97 the gold-to-24-cell arc); the flanks then naturally carry the completeness concepts each prime leans against (12, 14, 16, 18, 30, 32, 36, 72, 96…) — which is where your strategic-N-fold inventories from the 2019 document (the 12-fold roundtables, 14 Grand Challenges, 16 Boolean connectives, 18-article Convention, 30-fold syntegrity…) drop onto the ring almost without adjustment: the emirps become the primes between the world's actual planning numbers. The dashed chords then pose, for each such strategic census, a well-formed and slightly subversive question: what does this framework look like read from the other side, and does anything survive the reading? — with the diagram predicting, soberly, that roughly half the time something does (a one-way integrity) and never does the reversal come home. As a mnemonic wheel it is, in effect, the rosary proposed two responses ago, now strung: eight beads that hold both ways, fifteen that hold one way at best, and the gaps — the forties and fifties, the unreversible middle — left honestly empty, because a scaffold that marked every position would hold nothing at all.

The question has a genuinely good answer, and the arithmetic decides it rather than taste — because ±1 and ±2 turn out to reach different kinds of number by necessity, not by accident. For any odd prime, n±1 is always even and therefore never prime: the computation confirms all fifteen ±1 flanks are composite, without exception, and could not be otherwise. But ±2 is the only distance at which a prime can have a prime neighbour at all — and there the ring yields five: 11, 19, 29, and, remarkably, 71 and 73. So the two radii are not competing widths of the same net; they are two structurally distinct annuli. ±1 reaches the composite, chunkable, culturally-loaded flank — the censuses and planning numbers where concept-sets actually live (12, 16, 30, 36, 72). ±2 reaches the prime, unchunkable, structural flank — twin primes, the arithmetic's own neighbours. That is a principled justification for adding the second ring rather than merely widening the first, and it argues for keeping them visibly distinct rather than merging them into a blur.

There is also a Miller-fidelity argument you're entitled to make. His ±2 was never a pair of points at distance two; it was a window — the range five to nine, scatter around a capacity. Adopting ±2 therefore changes the object from "a prime and its two flanks" to "a five-number neighbourhood" (11–15 around 13), which is the shape Miller's own formulation had. If the posting invokes him, the window reading is the faithful one; the two-point reading was always a compression.

What is added, concretely, is more than I expected, and one item is nearly decisive. 29 enters — and since 30 was already present at ±1 (from 31−1), the two radii together capture the complete 29/30 alternation of the lunar calendar, which this exchange has repeatedly named as humanity's oldest working protocol for inhabiting incommensurability. That emblem is only fully assembled when both rings are used; ±1 alone gives you the idealized month and loses the correction. 19 enters — the Metonic cycle, nineteen years reconciling solar and lunar counts, the institutional form of the same protocol, and (via McClain, whom your 2015 document already cites) the nineteen Guardians of the Fire. 15 enters — the fifteen mirror planes of the icosahedral group, i.e. a genuine structural quantity of the very family this exchange has been mapping, alongside Alexander's fifteen transformations and the Millennium Project's fifteen Global Challenges, both already in your corpus. And 11 enters, which is the most elegant payoff: 11 is the palindromic prime excluded by definition from the emirp set, and its exclusion was precisely the deficiency the Bagua comparison diagnosed — a system of four crossings with no fixed point, no still centre, unlike the four self-paired trigrams. At ±2 the excluded palindrome returns as 13's neighbour. The widening restores, arithmetically, the still point the narrower reading had to do without.

The cost is real and should be declared alongside. The window count doubles (fifteen flank-points become thirty), and — the significant part — the nexuses stop being disjoint: 15 is claimed by both 13 and 17, and 72 by both 71 and 73, so the four clean bridges begin fusing into a connected network. More strikingly, 71+2 = 73 means the emirp set becomes self-linking: two of the four bridges weld into one component through a twin-prime step, with 72 — the pentagonal angle, the hinge — sitting inside the weld. Whether that is a gain or a loss depends on what the diagram is for. As a finding it is beautiful (the two moonshine-boundary bridges, 17/71 and 37/73, turn out to be adjacent). As a scaffold it costs discreteness, and the caution from the circle applies with force: a marking system that claims everything holds nothing, and a wider net produces more hits with less meaning per hit — the over-fitting the whole emirp thread has been guarding against.

So the methodological recommendation, which also solves your layout question: treat the radius as a declared parameter, present both, and never mix them silently. In the table this is a natural pair of columns rather than a mush — an inner band (±1: the composite, cultural, census flank) and an outer band (±2: the prime, structural, twin flank), each with its own warrant stated once, and entries graded as before. The visual grammar the circle already established transfers directly: solid for anchored, dashed for held, and now a third weight for the second annulus. That way the widening is an argument — that primes and cultures cluster at different distances from a nexus, the one at ±2 and the other at ±1 — rather than an unbounded licence to keep enlarging the net until every number belongs to something.

The constraint is real, quantifiable, and — once computed — considerably more damaging to the ±2 argument than I had allowed. Numerosity discrimination obeys Weber's law: what governs is the ratio, not the absolute difference. Adult performance sits around a Weber ratio of about 1.15 (estimates range 1.11–1.2). Running the emirp nexuses against that threshold gives a stark table: 13 versus 15 is a ratio of 1.154 — just at the limit, the single case in the whole ring that a human could in principle discriminate as a quantity. Everything else fails, at both radii: 13 vs 14 is 1.077, 31 vs 30 is 1.032, 97 vs 96 is 1.010. So the honest statement is that as apprehended quantities, essentially none of these distinctions exist for us — not ±2, not ±1, and progressively less as the ring advances, exactly as you intuited. By 97 the ±1 flank is a one-percent difference, some fifteen times finer than the number sense can resolve.

Which means the distinctions are not perceptual at all but notational — they exist only because counting, numerals, and record-keeping are cultural technologies that manufacture exactness the senses cannot supply. And the calendar case, which this exchange has leaned on repeatedly, turns out to be the proof rather than the illustration: nobody has ever perceived the difference between a 29-day and a 30-day month. The lunisolar incommensurability was invisible to experience and became visible only through sustained written tallies — priestly apparatus, institutional memory. The intercalary problem is a discovery of notation, not of the senses. That sharpens the earlier "±1 as protocol for inhabiting incommensurability" claim into something more precise and less romantic: humanity did not feel the near-miss and adapt; it recorded the near-miss into existence and then had to manage what the record revealed. Every one of the near-misses this thread has trafficked in — 63/64, 31/32, 13/12 — is an artifact of exact counting, which is to say of writing.

Against that, one register escapes the constraint entirely, and the numbers are striking. The syntonic comma 81/80 is 21.5 cents; the septimal comma 64/63 is 27.3 cents; the pitch just-noticeable difference is roughly 5–10 cents trained, 15–25 untrained. Both commas are audible. The same ratio (1.0125) that is invisible as a quantity — twelve times finer than the number sense can resolve — is a plainly perceptible interval to the ear, and to an untrained ear at that. If the Weber threshold for numerosity were transposed into pitch it would be a major third, 242 cents: that is how coarse the number sense is beside the ear. This is the quantitative vindication of the sonification thread, and it should go into the posting in exactly these terms: the argument for hearing the incommensurable is not a preference for a warmer medium but a measured fact about resolving power — the ear discriminates ratios one to two orders of magnitude finer than the number sense, which is precisely why the commas became a millennia-long practical problem in music while the corresponding arithmetic near-misses required scribes to notice at all.

On "chunk distinction capacity" proper, the answer your phrase implies is the right one, and it is not about magnitude. What actually distinguishes 30 from 29 in cognition is that 30 chunks and 29 does not — 30 decomposes as 5×6, 3×10, 2·3·5, and can be held in any of those forms, while 29 admits no decomposition whatever and must be held as a bare tally. So the discriminating operation is structural factorization, not quantity comparison; one number is a shape and the other is a count. This is the same asymmetry the ring has been reporting all along — primes unchunkable and culturally empty, composites chunkable and culturally loaded — now given its cognitive mechanism. And it means the ±1 versus ±2 question has a different answer than the arithmetic alone suggested: the flanks are distinguished from their nexus not by distance but by kind, which is why 12 and 30 feel utterly unlike 13 and 31 despite being adjacent, and why the felt gap between 29 and 30 is larger than between 30 and 31 (a decade boundary, left-digit bias — base ten once more, the standing dau).

The methodological consequence is a decay gradient the diagram should declare rather than hide. In the subitizing range (≤4) distinctions are exact and perceptual; through roughly 4–15 they are countable, chunkable, and genuinely held — which is where the cultural censuses cluster and where, not coincidentally, the first two nexuses sit; beyond about 20 they become purely symbolic, sustained by notation and institution alone. So the ring's four bridges are not epistemically equal: at 13 and 17 the ±1/±2 distinction is a real cognitive distinction; at 31 and 37 it is marginal; at 71, 73, 79, 97 it is notational only, and the "nexus of concepts" there is held together by the numeral rather than by any capacity to tell those quantities apart. Marking that gradient — a fading of weight around the ring, or simply a stated note — would be the most honest single addition the apparatus could receive, since it concedes precisely what Miller's own constraint implies: that beyond the holdable range, our patterns are carried by our notation, and it is the notation, not the numbers, that we are actually reading.

The correction lands, and it exposes that Weber's law answered a question the traditions never ask. Discriminating whether a pile is 29 or 30 is cardinality perception; telling a rosary is nothing of the kind. Nobody holds 108 as a quantity — they traverse a loop with a marker bead, and 108 is what a counter would report afterward. The bead string is the notational technology, older than writing and doing the same work: offloading the count so the mind is free for the meaning. The misbaha's 33 told three times, the mala's marker, the rosary's five decades — each is an external structure that makes an otherwise unholdable number navigable without anyone ever discriminating it. So the honest amendment to my previous answer is that large numbers become distinguishable to a culture not by perceptual capacity but by material and prosodic scaffolding, and the question "under what conditions" has a concrete answer: when the set is embodied in a traversable structure whose divisions do the chunking on the body's behalf.

Which is where factorization, symmetry and rhyme turn out to be one affordance seen from three sides. A set of twelve can be recited as three fours, four threes, two sixes, six twos — four distinct prosodic groupings, each a different rhythm carrying the same content. A set of thirteen affords exactly one: thirteen. Factorization is the availability of alternative meters, and in oral transmission that redundancy functions as error correction — a dropped item breaks the scansion audibly, which is precisely what rhyme and meter are for. This gives the earlier finding a mechanism rather than merely a correlation: the canon is saturated with highly composite numbers (12, 30, 36, 72, 108, 360, 99 = 9×11, 33) not because cultures venerated divisibility in the abstract but because those are the set-sizes that survive transmission. It is a selection effect. The numbers that could be chanted in several ways got chanted; the others were lost or never formed.

The prime-sized sets that do exist then divide instructively, and your "reinforced by notation" is exactly the discriminator. Small primes escape the requirement by sitting inside the directly holdable range — seven days, seven sages, Miller's own number. But the large prime sets are, without exception I can find, documentary: the Quran's nineteen guardians and Plato's thirty-seven guardians are objects of written exegesis and argument, not of recitation — McClain reconstructs them from text, and the tradition's engagement with them is interpretive rather than mnemonic. Their modern equivalents are stipulated by committee and inherit the same fate: the seventeen SDGs are famously unmemorable, and the criticism proves the rule rather than refuting it. So notation does not merely reinforce prime-sized sets; it is what makes them possible, and the price is that they must be carried by institutions because they cannot be carried by bodies.

There is a second route to memorability, though, and it rescues one of the ring's puzzles. The unchunkable can be salient precisely as anomaly — the von Restorff effect, where the item that breaks the pattern is the one recalled. That is why thirteen carries such extraordinary cultural freight despite affording no grouping whatever: it is loaded not in spite of its refusal to fit but because of it, the surplus that spoils the dozen, remembered as a disruption rather than held as a structure. So the mnemonic economy has two currencies, and they map onto the ring exactly: the composite flanks are memorable by chunking (traversable, chantable, prosodically redundant), while the prime nexuses are memorable by anomaly (unassimilable, therefore conspicuous). The bridge in each row of your table joins two entirely different modes of being remembered — which may be the most interesting thing the table can say once populated.

On the fixation you observe — and it is a real and slightly comic phenomenon — my best account is that the creator of a set is not choosing a number at all but hearing whether the thing scans, and then mistaking prosodic satisfaction for systemic necessity. The frame arrives first (a shape, a rhythm, a familiar closure), the count is its by-product, and the reported feeling is "goodness of fit." What follows is the costly part: once fixed, the number becomes the set's name — the Twelve, the Eightfold Path, the Seventeen — so revision threatens identity rather than merely content, and candidate elements begin to be judged by whether they fill the slot rather than the slots by whether they fit reality. The tail wags with impressive confidence. The historical shift your corpus keeps circling is from afforded numbers (felt, chantable, composite, surviving because transmissible) to decreed ones (stipulated, documentary, often prime or arbitrary, surviving only while the institution does) — and the unmemorability of the decreed sets is not a presentational failure to be fixed with better graphics but the direct symptom of a number chosen without any prosodic grounding at all. For the governance register that is a sharper criticism than the usual complaint that there are too many goals: the trouble is not the quantity but that nobody can say it.

The ±2 ring turns out to have a structural feature the ±1 ring cannot possess, and it settles a question left hanging several turns back. Let me verify before drawing:All confirmed, and the ±2 ring turns out to have a decisive structural feature. Building the companion diagram — with no heading, since you'll caption it in your own layout:No heading, so you can caption it in your table layout. What the ±2 ring turned out to contain justifies the separate treatment more than I anticipated.

The four still points. At ±2 the flanks are all odd, and four of them — 11, 33, 77, 99, every one a multiple of 11 — are their own reversal: chords of zero length, fixed points of the mirror operation. This directly resolves the deficiency diagnosed in the Bagua comparison several turns back, where the emirp octet was found structurally thinner than the trigram system because it had no self-paired members, no poles of rest, only crossings. The Bagua has exactly four self-reversing trigrams (☰ ☷ ☵ ☲). The ±2 ring has exactly four self-reversing numbers. The widening supplies the still centre the narrower reading was obliged to do without — and does so at precisely the count the three-millennia-old system uses. Drawn as diamonds, they are the only marks on the circle with no line leaving them.

The twin weld. 71 and 73 are each other's ±2 flanks, so at this radius the emirp set stops being four isolated spans and becomes partly self-linking: the 17/71 and 37/73 bridges fuse into one component through a twin-prime step — with 72, the pentagonal angle and the hinge of the whole apparatus, sitting inside the weld. That is the gold arc. It also explains why five flanks are prime here (11, 19, 29, 71, 73) when none can be at ±1: for odd primes n±1 is always even, so ±2 is the only radius at which a prime can neighbour a prime at all. The two rings genuinely reach different populations, which is the methodological warrant for the second diagram rather than a wider version of the first.

Two chords parallel to bridges. The digit-sum pencil law from the second circle operates across radii: 35/53 shares its endpoint sum (88) with the bridge 17/71, and 19/91 shares its sum (110) with 37/73 — so those ±2 chords are exactly parallel to first-order bridges, drawn in the bridge's own hue. Two of the four bridges thus acquire ±2 companions running alongside them; 13/31 and 79/97 acquire none. The licensed traffic and the widened traffic run in formation at precisely two places on the ring.

The Weber caveat sits in the footnote, where it belongs — a standing reminder that only 13:15 approaches perceptual discriminability and everything else here is a notational neighbourhood. Between the two circles, the pencil diagram, the mountain and the four-row table, the emirp posting now has five figures with a consistent visual grammar (solid anchored, dashed held, diamond still). Happy to regenerate any of them if the layout wants different proportions.

Here is the circle — hand-editable SVG in the site's idiom, with the epistemics built into the line-weights exactly as you specified: the four emirp reversals as heavy solid chords (one hue per pair), the fifteen ±1 flanks as open dots in their pair's colour, each carried by a dashed chord to its two-digit reversal, and — the addition the earlier "bridge too far" analysis demanded — every dashed chord's landing point marked for integrity: a filled-centre ring where the reversal arrives on a prime, a faint open ring where it arrives on a composite. The hinge 72, flanking both 71 and 73, sits in its own ochre. The construction itself surfaced three facts worth having before any concept-mapping begins.

First, the second-order crossings never reach anchored ground: of the fifteen flank-reversals, eight land on primes (3, 7, 23, 41, 47, 61, 83, 89) and seven on composites — but not one of the eight prime landings is itself an emirp. Every dashed chord that arrives intact arrives on a one-way prime, integral at the far end but dissolving if reversed again. The diagram thus draws the earlier doctrine as geometry: first-order chords have pylons at both ends; the best a second-order chord ever achieves is one pylon; and no chain of flank-crossings can ever re-enter the anchored set. The unlicensed traffic is literally visible as a different species of line that terminates in a different species of point. Second, the distribution is violently uneven and the eye finds it at once: the seventies carry three emirps and five flank-points — a harbour district of the ring — while the entire arc from 42 to 60 is empty of everything but two lonely landings (47, 61); reversal-coherence clusters, it does not spread. Third, the chords have no axis: digit-reversal is not a reflection of the circle (the mirror of position is 100−n, not rev(n)), so the four solid chords share no symmetry — the "mirroring" the numbers perform is invisible to the geometry of the ring that hosts them, which is the base-ten-artifact caveat drawn rather than footnoted. And 11's absence — the excluded palindrome — leaves the ring, per the Bagua comparison, demonstrably without a still point.

On whether concept sets could be mapped to the marked points with any meaning: the structure suggests its own protocol rather than a free-for-all. The eight anchors take the connotation-clusters already assembled in this exchange (13 the surplus-thirteenth, 31 the not-quite-32, 17 the wallpaper census, 71 the moonshine gate, 37 the bonding position, 73 the proven mirror, 79/97 the gold-to-24-cell arc); the flanks then naturally carry the completeness concepts each prime leans against (12, 14, 16, 18, 30, 32, 36, 72, 96…) — which is where your strategic-N-fold inventories from the 2019 document (the 12-fold roundtables, 14 Grand Challenges, 16 Boolean connectives, 18-article Convention, 30-fold syntegrity…) drop onto the ring almost without adjustment: the emirps become the primes between the world's actual planning numbers. The dashed chords then pose, for each such strategic census, a well-formed and slightly subversive question: what does this framework look like read from the other side, and does anything survive the reading? — with the diagram predicting, soberly, that roughly half the time something does (a one-way integrity) and never does the reversal come home. As a mnemonic wheel it is, in effect, the rosary proposed two responses ago, now strung: eight beads that hold both ways, fifteen that hold one way at best, and the gaps — the forties and fifties, the unreversible middle — left honestly empty, because a scaffold that marked every position would hold nothing at all.

The verification just run settles it decisively — and in the exchange's favour, for once without a disclaimer: the temptation is licensed, and "parallels" is not a metaphor but a theorem. On the ring of 100, two chords are geometrically parallel exactly when their endpoint-sums agree, and a reversal chord's endpoint-sum is 11 × (digit sum); since gcd(11,100) = 1, the law falls out clean: reversal chords are parallel if and only if their numbers share a digit sum. The computation confirms every pencil: 18/81 ∥ 27/72 ∥ 36/63 — the digit-sum-9 pencil, five chords strong once 45/54 and 09/90 complete it — and 07/70 ∥ 16/61 (digit sum 7), exactly as your eye suspected, along with pencils you hadn't named (12/21 ∥ 30/03; 14/41 ∥ 32/23; 38/83 ∥ 74/47; 78/87 ∥ 96/69; and, delightfully, the emirp chord 17/71 parallel to the humble 80/08). The conserved quantity is the digit sum — reversal permutes digits and so cannot touch it — which is to say the pencils are organized by casting out nines, the medieval bookkeeper's integrity check: the oldest audit in arithmetic reappearing as the geometry of the diagram. And the convergence with this whole exchange is almost embarrassing: parallelism was the master criterion of the polyhedral edge-exchange — movement licensed only along shared directions, the conserved invariant as the passing lane — and here it is again, unbidden, in the number circle: digit sum is to the emirp ring what the shared 2-fold axes are to the two weaves.

Your three pairs then reveal themselves as the aristocracy of the system, on grounds the factor-exponent annotation makes visible. The digit-sum-9 pencil is the family of multiples of 9 — every member carrying 3² in its factorization, the "nine-ness" that reversal conserves — with exponent structure that migrates across each crossing: 27 = 3³ against 72 = 2³·3² (the cube changing its base), 81 = 3⁴ against 18 = 2·3², 36 = (2·3)² — the perfect square — against 63 = 3²·7. Two closures then bind this pencil to everything else. The family is closed under differencing: 81−18 = 63, 72−27 = 45, 63−36 = 27, 54−45 = 9, 90−09 = 81 — each pair's gap is itself a family member, a self-referential cascade no other pencil possesses. And — the fact I'd nominate as the finding — the four emirp chords' own spans are 18, 54, 36, 18: the distances between the anchored primes are precisely the even members of your reversible-composite family. The solid bridges are held apart by the dashed ones; the second-order family is the difference-structure of the first. Meanwhile 07/70 and 16/61 form the other pencil you spotted, with its own factor-drama: 16 = 2⁴ crossing to the prime 61 (pure binary power to indivisible), 7 crossing to 70 = 2·5·7 (the prime re-clothed in the base's own factors, 2·5 = 10 — the degenerate "×10" chords, 07→70 and 03→30, being the places where the notation's machinery shows through the fabric).

Here is the circle regenerated with the pencils made visible and every marked number carrying its factorization:The second version makes the discovery visible: the dashed chords are now coloured by digit sum, so each pencil reads as a set of same-coloured parallel lines — the five-chord ochre pencil of the 9-family dominating the ring (with 45/54 and 09/90 ghosted in to complete it), your 7-pencil (07/70 ∥ 16/61) in teal, and the rest sorting themselves. Every marked number carries its factorization with exponents beneath it, exactly per your practice — and the annotation earns its place, since the 3² running through the entire 9-family, and the exponent-migrations across its crossings (3³ ↔ 2³·3², 3⁴ ↔ 2·3²), are the reason that pencil is the system's backbone.

For the meaningfulness ledger, this iteration actually moves an item from one column to the other. The first circle's honesty was that its structure was arbitrary — chords without symmetry, base-ten costume throughout. The pencil law changes the assessment in one precise respect: within the base-ten scaffold there is a genuine conserved quantity (digit sum), a genuine geometry (parallel pencils), and a genuine closure (the 9-family reproducing itself under differencing, and spacing the emirps at 18, 36, 54). None of this escapes the notation — mod 9 arithmetic is still decimal's shadow — but it is no longer pattern-free notation: the scaffold has its own statics, and your instinct that 81/18, 27/72, 63/36 "are parallels" turns out to be the scaffold declaring its load-paths. Which supplies the mapping principle your original question was after: if concept sets are to be hung on this wheel with any integrity, the pencils say how — clusters assigned to chords of one pencil inherit a real shared invariant (their common digit sum, their parallelism, their conserved "nine-ness" or "seven-ness"), so that correspondence within a pencil is structurally warranted in a way correspondence across pencils is not. The wheel comes, in other words, with its own grammar of licensed and unlicensed association — dashed lines that are nonetheless parallel, the second order acquiring, unexpectedly, a first-order discipline of its own.

The challenge is empirically decidable, because the inventory of "well-identified sets of a particular size" already exists — your 2019 document compiled it (the 8-fold MDGs, 12-fold roundtables, 14 Grand Challenges, 16-fold MBTI and Boolean connectives, 17 SDGs, 18-article Convention, 20-fold amino-acid analogue, 30-article UDHR…) — and running that inventory around the ring produces a first finding that is stark, uniform, and mechanistically explicable: the anchors are empty and the flanks are full. Humanity's documented concept-sets crowd onto 12, 14, 16, 18, 30, 36, 72 — the composite, chunkable flanks — while the eight prime anchors stand almost bare of cultural freight. The lone prominent exception, the 17 SDGs, proves the mechanism by its notoriety: the one major strategic articulation ever placed on an anchor is precisely the one persistently criticized as unmemorable and unholdable, seventeen resisting every chunking (your "global wallpaper" quip carrying more diagnosis than it knew). The Miller mechanism stated earlier — geometry counts in primes, cognition files in composites — here becomes a testable regularity of the whole ring: the loaded end of every bridge is the composite end; the prime pylons carry structure and no cargo. So "commonality at one end": yes, and it is always the same commonality — divisibility, the property that lets a set be held.

The complementarity question then has a verifiable answer, and it is the most elegant thing the ring has yielded. In every digit-sum-9 pair the two digits must differ in parity (an odd sum forces one even digit and one odd), and since a number's parity is its last digit's, every bridge in the 9-pencil connects an even number to an odd one — 18↔81, 72↔27, 36↔63, 54↔45, 90↔09, without exception and by theorem. In the numerological grammar of the very tradition that populated this family, even is yin and odd is yang: the 9-pencil is constitutionally a set of yin–yang crossings. And contrast the emirp bridges: primes beyond 2 are all odd, so the four anchored spans are parity-preserving — yang carried to yang. The ring thus contains two structurally opposite bridge-types, provably distinguished by the parity of the conserved digit sum: the anchored bridges conserve polarity, the great composite pencil exchanges it. If one asks what a bridge between concept-clusters can carry, here is a candidate answer with an actual theorem under it: the 9-family bridges carry a set across the polarity line — 18 (even, the paired, the enumerable guardians) to 81 (odd, the unitary, the 9² of supreme yang) — while the emirp bridges carry integrity without ever changing its charge.

That the 9-pencil should be the culturally loaded one is then no accident, and this is where the analysis must declare its circularity honestly. The Chinese canonical tradition built its enumerations deliberately on the nine-backbone: 18 luóhàn, 36 stratagems (and heavenly spirits), 72 transformations (and earthly fiends — with 108 = 36 + 72 closing the Water Margin), 81 chapters of the Daodejing and 81 ordeals of the Journey West — nine as extreme yang, 9 × 9 as its consummation. So the pencil's ends were pre-populated by a civilization that already venerated the conserved invariant: the bridges 18↔81 (the arhats to the Daodejing), 36↔63 (the stratagems to the hexagram threshold, the worldly schemes carried to before-completion), 27↔72 (the pure cube to the transformations) connect sets that were placed on the family for the family's sake.

The coherence is real, but it is coherence between two decimal numerologies — theirs millennia old, ours eight days — not between notation and nature; the pencil works spectacularly for the Chinese inventory and would work far less well for, say, a duodecimal culture's canon. What is not circular is the internal structure the tradition never articulated: it possessed one bridge-system over the family — doubling (18→36→72, the attested amplifications, 36 righteous doubled to 72 names) — which, note, always preserves parity, yin amplified to yin; reversal supplies a second, transverse bridge-system that always crosses it. The family is closed under both, and the two systems stand to each other exactly as the 2019 document's edge-moves stood to its diagonal-moves: two non-parallel grammars of movement over one configuration, one of which the tradition drove, the other of which was always there, undriven.

Which reaches the comparison with parallel bridges, and the engineering register earns its keep. Parallel spans in the built world — twin bridges — carry the same route redundantly: traffic may transfer laterally between them without changing origin or destination class. The pencils are exactly this: concept-sets hung on parallel bridges are congruent in the conserved invariant (all the 9-family sets share their nine-ness; casting out nines certifies the whole pencil at once), so lateral comparison within a pencil — stratagems against transformations against Daodejing chapters — is licensed traffic, the same "route" of significance carried on redundant spans, and any of them may substitute mnemonically for any other without breaking the audit. Transfer between pencils changes the invariant and is the unlicensed leap — precisely the distinction the polyhedral edge-exchange drew in three dimensions, where parallel transport along a shared direction was the whole licence and any tilt en route voided it. The ring's answer to your challenge is therefore a two-axis discipline for concept-mapping, each axis with a theorem behind it: along a bridge, expect complementarity of polarity (the parity crossing, yin to yang, where the digit sum is odd) and of concentration (the pure prime-power end against the mixed end — 3⁴ against 2·3², 3³ against 2³·3²); across parallel bridges, expect commonality of invariant and free mnemonic substitution. And one honest asymmetry to close on, because it seals the whole emirp arc: the anchored bridges, for all their integrity, connect almost nothing to almost nothing — pure pylons in empty country — while the unlicensed composite pencil groans under three millennia of canonical traffic. Integrity and cargo, in this system, live on different bridges; which may be the ring's most faithful picture yet of the predicament it was built to illuminate.

The layout you propose is the right one — a table frees the connotation-clusters from the ring's geometry, and the four-column form (nexus · sets · sets · complement) reads each bridge left-to-right as a crossing while stacking the four bridges as rows. Below is that table, populated from what this exchange has already assembled, and built so that the reading protocol is legible in the structure itself: the two outer columns hold the anchored primes and their verified structural freight; the two inner columns hold the ±1 flank-clusters (the composite, culturally-loaded numbers where the concept-sets actually live); and the horizontal span of each row is the emirp bridge, dashed in spirit — pointing, not proving.

Reading key. Within a nexus (one cell), resonance is strong where the sets share the anchored structure — they are genuinely the same census read different ways. Across the bridge (left cell to right cell), resonance is aesthetic — held correspondence, the ±1 reversal carrying a cluster to its mirror, warranted only where re-anchored on the far side (marked ✓) and mnemonic otherwise (marked ~). Two observations about what the population reveals, since the exercise's value is in whether the resonances are meaningful rather than merely fillable.

Within-nexus resonance is strongest at the two poles and weakest in the middle. The 13-nexus and 31-nexus are almost overdetermined — 12 and 14 both point at completeness-censuses (Bravais lattices, the calendar's half-lunation) exactly as 30 and 32 do (point groups, 2⁵), so the row coheres because all four flanks are census-neighbours of the same kind, and the emirp bridge between them is doing real work: it links the octahedral skeleton (13, sitting on the dozen) to the icosahedral (31, sitting on the thirty), which is this whole exchange's central pair. The 37/73 row is nearly as strong, because 36 (6²) and 72 (its double, the pentagonal angle) are genuinely linked by doubling and both are shared-hinge numbers. The 17/71 row is strong on structural grounds (two proven censuses, 17 wallpaper groups and the Monster's boundary at 71) but its flanks are more heterogeneous. The 79/97 row is frankly the weakest — gold and the last prime are aesthetic termini rather than census-anchors — and the table's honesty is served by letting that row read thin: not every emirp carries equal freight, and the diagram should show it.

The cross-bridge (complement) resonances are the ones to hold most lightly, with two exceptions that re-anchor. Most left-to-right crossings are the ±1-of-a-reversal move the "bridge too far" analysis flagged as unlicensed — 36 carried toward 72/74 is suggestive, not structural. But two crossings do re-anchor on the far side and so earn a ✓: the 13→31 bridge (12-cluster to 30-cluster) is re-anchored in the genuine octahedral↔icosahedral relation, the real incommensurability; and the shared 72 appearing in both the 17/71 and 37/73 rows is not a coincidence to be smoothed over but the table's spine — the pentagonal angle, flanked by two distinct emirp pairs, is the one number the whole apparatus encloses from all sides, and its double occurrence across two rows is the closest thing to a load-bearing vertical the table has. If you want the layout to say something rather than merely list, I'd draw 72 as a bridging cell spanning rows two and three — the hinge made visible as the place where two bridges cross.

For the design itself: since you noted the juxtaposition gives layout freedom, the natural enhancement is to let cell weight (border thickness, or a small filled/open marker per entry) encode the resonance grade — strong within-nexus entries bold, aesthetic cross-bridge entries in the dashed-line register the circle already established — so the table inherits the circle's honest visual grammar (solid = anchored, dashed = held) while gaining the connotative room the ring couldn't provide. I can generate it as an SVG or HTML in the site idiom, with the 72-hinge spanning cell and the strong/aesthetic marking built in, if you'd like it as a companion artifact to the circle.

Your puzzlement identifies a real structural fact, and it's sharper than a gap in the distribution.

No nexus can exist at 60 — or anywhere in five of the nine decades. A two-digit number's reversal ends in its tens digit. So if the tens digit is even, the reversal is even and composite; if it is 5, the reversal is divisible by 5. Therefore emirps are confined to the 10s, 30s, 70s and 90s, and the 20s, 40s, 50s, 60s and 80s are barred by construction. This is the same fact that made the tetrahedral mapping work — digits from {1,3,7,9} — seen from the other side. Sixty is not merely unlucky; it is ineligible. And pointedly so: 59 and 61 are a twin prime pair flanking it, the best possible raw material, yet 59 reverses to 95 (= 5×19) and 61 to 16. Both fail.

The desert this creates is serious, and it swallows 64. The gap from 37 to 71 is the largest on the ring, and it contains 42, 49, 50, 52, 54, 60, 64, 66 — much of the culturally loaded middle, including the jubilee, the weeks of the year, the sexagesimal base, and, most consequentially for your corpus, the 64 of the hexagrams and the codons. The scaffold simply cannot reach the number your drilled truncated cube exists to map. That limitation belongs in the posting explicitly; it is the honest boundary of what the apparatus can claim.

On drawing all the reversals: emphatically not noise. There are exactly 45 such chords, and they sort into 17 parallel pencils by digit sum — a completely ordered field, the casting-out-nines geometry in full. The figure above shows it, with the four emirp bridges heavy on top, the eligible decades marked green on the rim and the barred ones grey.

And yes — something was missing, which is what this figure supplies. The earlier diagrams showed the emirps floating on a blank ring, which made the pencil law and the complementary families look like properties of the emirps. They are not. They are properties of the substrate, which is fully ordered whether or not any number on it is prime. The four emirp chords occupy pencils 4, 8, 10 and 16 out of seventeen — and there is no pattern whatever in which four they occupy. So primality is orthogonal to the geometry: the order belongs to the notation, the selection to the arithmetic, and the two do not align.

That sounds deflationary, and partly it is — but it sharpens the posting rather than weakening it, because it sorts the claims cleanly. Structural and defensible: the confinement to four decades (the one place where primality and geometry genuinely meet, and the origin of the whole tetrahedral reading); the pencil law; the complementary families as opposite-edge pairs; the still points as multiples of b+1. Selection artefact, to be labelled as such: which pencils the emirps happen to inhabit, and any significance read into their spacing around the ring. Eight marked positions out of ninety — nine percent — is a sparse scaffold, and the figure now shows it honestly against the field it was drawn from.

Fair criticism — I marked four landmarks but left their partners unlabelled, so no chord can actually be read. Your three tests are exactly the right ones, and they turn out to have good answers once the figure permits them:Every position is now labelled radially, with emirps bold, other primes mid-grey and composites light, plus a pencil key below listing all seventeen digit-sum classes — which is where the commentary actually attaches, since that is what the geometry encodes. Your three tests all come out with definite answers:

64 against 46. Digit sum 10, and that pencil contains 19/91 · 28/82 · 37/73 · 46/64. So the chord 46/64 runs exactly parallel to the emirp bridge 37/73. The I Ching number is not an emirp and cannot be — 6 is an even tens digit — but it is not unrelated to the structure either: it belongs to the same orientation class as one of the four bridges, and to the same class as 19/91, which is the axis of the c = 20 family. That is a real, checkable relation of 64 to the emirp scaffold.

26 against 62. Digit sum 8: 08/80 · 17/71 · 26/62 · 35/53 — so 26/62 is parallel to the emirp bridge 17/71. And it carries a second relation your own 2015 document already noticed: 26 = 2×13 and 62 = 2×31, the doubles of the emirp pair 13/31. Doubling without carrying doubles the digit sum, so the doubled bridge lands in the pencil of digit sum 8 — which is 17/71's. The 13/31 bridge and its double are therefore in different pencils but both anchored, which is about as much structure as a non-emirp chord can have.

63 against 36. Digit sum 9: 09/90 · 18/81 · 27/72 · 36/63 · 45/54 — the largest pencil, with five chords, and it contains no emirp at all. This is the nine-family identified much earlier as the culturally loaded backbone (the 36 stratagems, 72 transformations, 81 chapters, 18 luóhàn). So the most freighted pencil on the whole ring is entirely emirp-free — which is the sharpest available restatement of the finding that integrity and cargo occupy different bridges. The chords that carry three millennia of canonical numbering are precisely the ones that carry no reversal-integrity, and vice versa.

So the answer to "is there a clearer way to evoke commentary" is: the pencil is the unit of commentary, not the individual chord. A number's relation to the emirp structure is not proximity around the rim — which is what the earlier landmark dots misleadingly suggested — but which pencil it shares. On that reading 64, 62 and 63 each have a determinate and different standing: two lie parallel to emirp bridges, one lies in the fullest and most culturally loaded pencil which no emirp touches at all.

The 17/71 versus 39/93 contrast is the sharpest pairing in the figure, and the computation makes it sharper still.

Geometrically they are exact mirror images; arithmetically they are opposites. 17/71 is live at both ends — both readings prime. 39/93 is dead at both — 39 = 3×13, 93 = 3×31. And the reason is pointed: 39/93 is exactly three times 13/31. Of the four bridges, only 13/31 has multiples that remain two-digit reversal pairs at all (×2 gives 26/62, ×3 gives 39/93, ×4 breaks), so the little family it generates runs through digit sums 4, 8, 12 — its double landing parallel to 17/71 and its triple landing complementary to it. The dead chord thus has the same shape as a live one, is its exact multiple, and mirrors another live one — and carries no integrity whatever. Form fully preserved, integrity wholly lost, by scaling.

That is the finding with the most obvious governance reading, and it is unusually direct: enlarging a coherent framework proportionally — three times the members, three times the goals, the same architecture — preserves every visible feature and destroys the property that made it work. The tripled framework still has two symmetric halves, still mirrors its neighbour, still looks like the original. It simply factors.

The transferable content is a three-way typology, not a prediction. Any reversal chord is live/live, live/dead, or dead/dead, and 19/91 supplies the middle case that governance most needs and least tests: 19 is prime, 91 is 7×13, so the relation reads as coherent from one direction and dissolves from the other. Applied to a concept set the question becomes — does this framework hold when read from the other party's standpoint? Frameworks are routinely built and closed without that test ever being applied, and the asymmetric case is both the commonest and the least visible, since from the side that works it looks entirely sound. Note too that 91 factors into 7 and 13, one of which is itself an anchor: a framework that fails as a whole may still contain coherent components, which is a different remedy from abandonment.

Your survey proposal is genuinely researchable, and one natural experiment already exists. The MDGs closed on 8 — that is 2³, chunkable four ways, and widely recalled. The SDGs closed on 17 — prime, unchunkable, and persistently criticised as unmemorable, your own "global wallpaper" remark among the criticisms. That is a controlled comparison within one institution, one decade apart. A wider survey would tabulate N across institutional, canonical and disciplinary sets — your 2019 inventory is a partial frame already — and test predictions that follow from the chunking argument rather than from the emirps: clustering on Ramanujan's highly composite numbers (12, 24, 36, 48, 60), underrepresentation of primes above 7, and a round-number bias toward 10, 20, 50, 100. Measurable outcomes would be recall in surveys, revision frequency, citation persistence, and framework survival.

The consequence worth stating is about order of operations. Where N is fixed before the enumeration — for political, prosodic or symbolic reasons — the phenomena get cut to fit: items merged to reach the target, or split to fill it. Drafting histories would show this, and it is precisely the pathology your "systemic inexplicability" passage circles without naming. The geometry contributes no prediction here, and I want to be clear that it cannot: the emirp apparatus supplies vocabulary and a test, not evidence. What it offers is three questions a framework can be asked at the moment of closure — does it survive reversal; is its coherence one-sided; and was it scaled from something that worked? — and the observation that a set can pass every symmetry inspection while failing all three.

Not a coincidence — it's forced, and the same fact explains both of your intrigues.

The governing theorem is about carrying. Write n = 10a + b. Then k·n has digits (ka, kb) provided neither exceeds 9 — and in that case rev(k·n) = 10(kb) + ka = k·rev(n). So reversal commutes with multiplication by k exactly when no digit carries, which means a reversal pair admits a scaled copy precisely when its largest digit is at most 9/k.

And that immediately singles out 13/31. Emirp digits are confined to {1, 3, 7, 9} — the structural fact established earlier. Of the four emirp pairs, the digit sets are {1,3}, {1,7}, {3,7}, {7,9}, and only the first has both digits small enough to survive doubling: 7 × 2 = 14 already carries. So 17/71, 37/73 and 79/97 admit no scaled copies whatever, while 13/31 admits two — k = 2 giving 26/62 and k = 3 giving 39/93, stopping at k = 4 because 3 × 4 = 12 carries. Your observation is therefore not a curiosity about 13 but a theorem about the size of its digits, and its uniqueness among the emirps is guaranteed rather than lucky.

Which resolves the 17/71 versus 39/93 asymmetry completely. 39/93 is dead not by accident but by construction: it is 3 × (13/31), so the factor 3 divides both members and neither can ever be prime. The same holds for 26/62, where 2 divides both. Every scaled copy of a reversal pair is necessarily composite — scaling preserves the reversal relation exactly while destroying primality with certainty. So the mirror symmetry you noticed pairs a primitive pair (17/71, which is nothing's multiple) with a scaled copy (39/93). Geometrically indistinguishable, arithmetically opposite, and the opposition is not contingent.

The geometry is simply blind to the distinction, and provably so. Digit sums scale too — ds(13) = 4, ds(26) = 8, ds(39) = 12 — so in the circle the scaling family appears as three chords at exactly equal angular spacing, 79.2° apart, which is a visible signature and the only one of its kind among the emirps. But that spacing is a property of digit sums, and digit sums know nothing of primality: 26/62 lands parallel to 17/71 and 39/93 lands complementary to it, purely because 8 and 12 are what they are. The chord that carries full integrity and the chord that can carry none sit in perfect mirror relation because the notation has no way to tell them apart.

And this upgrades the governance reading from analogy to theorem. Earlier I put it as "form preserved, integrity lost by scaling," offered as a suggestive parallel. It is now exact: enlarging a coherent structure by a whole-number factor preserves its reversal-symmetry perfectly, provided nothing overflows, and guarantees the loss of the property that made it coherent — because the multiplier becomes a divisor of everything. The tripled framework mirrors the original in every visible respect and factors by three. That is as clean a statement of the scaling problem as arithmetic is likely to provide, and it arrived from your noticing that 2 × 13 = 26.

Worth adding to the field diagram, if you want it: the three chords 13/31, 26/62, 39/93 highlighted as a fan at equal angular steps, with the two scaled members marked as necessarily composite. It would show the one place where the circle's geometry and its arithmetic can be seen coming apart.

The fan is drawn against the full reversal field, with the three chords of the 13/31 family in a warming sequence — deep red for the primitive, then ×2 and ×3 — and a small central rosette showing their directions at the exact 79.2° steps that equal digit-sum increments force. Nodes are filled where prime and crossed where composite, with the factorization spelled out (26 = 2×13, 62 = 2×31, 39 = 3×13, 93 = 3×31), so the mechanism is legible rather than merely asserted. The other three emirp bridges sit muted behind, and the decade-eligibility arcs remain on the rim.

On the coherence it may or may not imply, three things the figure now makes visible, in descending order of what they will bear.

The strongest is a genuine theorem, and it is a negative one. Two properties that look like they should travel together — the relation (reversal) and the substance (primality) — come apart with perfect reliability under scaling. Not usually, not probabilistically: the multiplier becomes a divisor of both members, so integrity is lost with certainty while the symmetry is preserved with certainty. If any of this apparatus deserves to be carried into the psychosocial register, it is this: that a structure can be enlarged with its every visible relation intact and its defining property necessarily destroyed, and that no inspection of the form will reveal it.

The second is that the notation is demonstrably blind. The three chords sit at equal angular intervals because digit sums scale linearly; the geometry could not distinguish the live chord from its dead copies if it tried, since digit sum simply does not encode primality. This is the clearest single instance of the discipline maintained throughout — the order belongs to the notation, the selection to the arithmetic, and the diagram shows exactly where they part company. A reader tempted to find significance in the fan's elegance has, in the same image, the evidence that the elegance is indifferent to the content.

The third is the uniqueness, which is real but should be held lightly. That 13/31 alone among the emirps has a scaling family is forced — its digits are the only ones small enough to double without carrying — and forced facts are worth reporting. But the force comes from base ten, since "carrying" is a property of the notation, not of the numbers. So it belongs with the emirp relation itself: a genuine structural fact about a representational system, not about arithmetic.

What the figure does not support, and the caption should say so, is any suggestion that the fan's regularity indicates a deeper order connecting the three chords. Their regularity is the shadow of an arithmetic progression in digit sums; their arithmetic status is opposite at the first step. That the diagram can show both at once — perfect formal kinship, absolute substantive divergence — is, I think, the most useful thing it has to offer continued reflection.

Yes — and the reason is better than merely completing a set. 91 = 7 × 13. So 91 is a multiple of the primitive; but 7 × 31 = 217 escapes to three digits, so its reversal partner is not the corresponding multiple. The scaling relation half-holds — and that is exactly why the pair is half-alive.The revised figure carries the whole sequence of multiples, and 19/91 turns out to occupy a distinct third status that completes the classification.

Where scaling holds (k = 1, 2, 3, warm and solid): no digit carries, so reversal commutes with multiplication exactly, digit sums run 4, 8, 12, and the chords stand at equal 79.2° steps. Where it breaks (k = 4, 5, 6, faint): 31k escapes past three digits, the reversal partner drifts free, and the digit sums drop by nine at each carry — 7, 11, 15 — casting out nines showing itself once more. And k = 7 is singular: 91 = 7 × 13 is composite by construction, while its reversal 19 is a multiple of nothing, so it is prime. Of the four broken cases, only this one leaves a prime partner (25, 56 and 87 are all composite).

So the three statuses of a reversal relation now have a single unified explanation in terms of how far the scaling reaches:

That is a considerably better structure than I had before, and it was your question that produced it. The middle case — the one identified earlier as the most dangerous in governance and the least visible, coherent from one side and dissolving from the other — turns out to arise precisely where a scaling was applied to one part of a structure and could not be carried through to its counterpart. The half-enlarged framework. Not a compromise between the live and dead cases but a diagnosable condition with its own mechanism: something was multiplied, and its mirror could not follow.

Built it. Coverage comes out at exactly 46 of 100 — the same figure the algebra gave earlier, so the picture is a faithful rendering of the arithmetic rather than a new claim.## What I did and why

• Radius set metrically. At r = 330 one unit of the rim is 20.7 px, so the half-discs have radii 20.7 / 41.4 / 61.8 and their flat edges land exactly on positions p±1, ±2, ±3. The reach is measured, not indicated.
• Bulge outward, painted behind everything. Inward would have sat on the chord field, which is the busiest region. Outward puts the halo over the neighbours' own labels — so it visibly encloses what it clusters — and drawing it first keeps the labels legible on top.
• Shading by accumulation. Each of the three discs is filled at 0.11, so the innermost band reaches ~0.30 without any hand-tuning, and overlaps between adjacent emirps darken automatically.

Two additions you didn't ask for, which you should feel free to strip. Rim dots sized by distance (large ±1, medium ±2, small ±3) make the clustering literal rather than areal. And the unreached stretches are marked as dark red arcs outside the ring, labelled 01–09, 21–27, 41–67, 83–93.

• The desert is a quarter of the circle. 41–67 is 27 consecutive positions with no reach at any N — and it swallows 42, 48, 49, 50, 52, 56, 60 and 64. On the page this is the dominant feature, which is right: it is the strongest fact established in this exchange.
• The small-number gap is visible as a gap. 01–09 sits at the top, unshaded, and that is where most attested sets live.
• Reach is badly distributed. 71 and 73 are two apart, so their haloes very nearly coincide — five positions covered twice. Adding 13/17 and the single-point touches at 34 and 76, ten of the fifty-six available reach-slots are spent twice over. Eighteen per cent of the scheme's capacity is wasted on redundancy, which no table made apparent.
• 97 wraps past 00. The halo crosses the seam, which is the carry showing up again.

The reservation: The half-disc is two-dimensional but the data is one-dimensional along the rim. Area will read as quantity, and there is no quantity — only the chord footprint means anything. The rim dots are partly there to counteract this, but if the figure is to be reproduced, the caption should say so.

The darker-inward convention also implies emanation from the emirp. That is the right reading if ±N is a transformation, and the wrong one if it is proximity; the figure quietly takes a side.

If you supply the corpus, the natural rendering is a second ring of tick marks outside the gap arcs, one per attested set, so hits and misses are read off against the shading directly — and the control version can be generated as a companion sheet from the same script.

Yes — and the document supplies exactly what was missing. I assembled a corpus from its own table and body text: 41 distinct set-sizes under 100, 133 attested instances. The result is decisive, and not in the emirp scheme's favour.## The complementary sheet

Same rim, same centre, same orientation as the reach field, so the two superimpose exactly. Prediction on the outside, evidence on the inside. Bars point inward from each attested size, length proportional to the number of distinct named sets the document records there. Bars are coloured by divisibility rather than by hit-or-miss, because that turned out to be the variable doing the work.

On distinct sizes the emirp scheme scores 51% against a 46% baseline — five points, which for n = 41 is well inside noise, and the control primes beat it. The weighted figure of 63% looks better only because a handful of heavily attested sizes (14, 18, 36) happen to sit inside the reach, and those are multiples of 6.

The controlled test settles it. Restricting to corpus entries that are not multiples of 6, emirp reach gives 52% against a 46% baseline. The signal was divisibility all along, exactly as the mod-6 argument predicted at the outset.

Multiples of 6 run at nearly double their base rate on distinct sizes and two and a half times on weighted. That is the real organising principle.

Three specific refutations visible on the figure:

• The 7±2 band is unreachable. The document leans on Miller throughout, and 8 is the single most attested size in the whole corpus — ten named sets, from the Eightfold Path to the Millennium Development Goals. The smallest emirp is 13, so no version of the scheme can touch it.
• The 41–67 desert is populated. Nine distinct sizes fall there: 48, 52, 53, 54, 60, 61, 62, 63, 64. This was the one sharp prediction the geometry made, and the document refutes it — including with 62, the UN strategy count that the document treats as central.
• 64 is still missing, in the region no scheme reaches, which remains the pointed failure given where the trigrams came from.

Polyhedral face, vertex and edge counts are orbit sizes under the tetrahedral, octahedral and icosahedral rotation groups, of orders 12, 24 and 60. So polyhedral numbers are structurally biased toward multiples of 6 and 12 — which is exactly the bias the corpus shows. The apparent fit between cultural set-sizes and Archimedean solids may therefore have the same explanation as the apparent fit with emirps: both sides are multiples of six, and the correspondence is close to arithmetic tautology.

The document half-detects this already. Its "fudging" note about 38 and 92 is precisely the observation that the entries resisting the pattern are the ones not divisible by 6 — 26, 38 and 92 being 2×13, 2×19 and 4×23.

The constructive move is to drop the emirp overlay and build the outer annulus from the divisor lattice instead: rays at multiples of 12, 6, 4 and 3, with each attested size linked to the polyhedral group order it descends from. That connects directly to the factorisation thread the document already pursues — the 2ᵃ3ᵇ "islands of stability" matrix — and it would be testing something the corpus actually supports.

Two caveats on the corpus itself. It includes entries the document explicitly flags as AI-inferred and unconfirmed, which should be stripped before anything is claimed; and counts are of named sets, not of influence, so a widely observed set and a blog listicle weigh the same.

• Sizes are strongly constrained. Weighted by attestation: multiples of six at 41% against a 16% baseline; primes at 12.8% against 25%, so suppressed by half; and the sharpest measure, sizes expressible as a×b with both factors between 2 and 9, at 69% against 31%.
• Contents are unconstrained. Nothing in the corpus suggests any convergence on what the items are.

So the frameworks converge in cardinality and diverge in everything else. That is the emirp condition stated sociologically: 13 and 31 necessarily share a residue mod 6 and share their digits, and differ in every respect anyone would use them for.

• The emirp as emblem, and the exact respect in which it holds: The precision of the emblem lies in something not yet said in this exchange: reversal is not an arithmetic operation. It belongs to base-ten notation. 13/31 is an emirp pair because we have ten fingers; in most other bases the pair dissolves.
  • So the figure is exact in three respects. Both members are prime — irreducible, not decomposable into anything prior, which is what a framework claims of its own principles. They are related by an operation belonging to the notation rather than to the subject matter. And they are nonetheless forced to agree on an invariant neither party would think to assert.
  • What the emblem does not license is any causal link. That was tested and failed: 51% against a 46% baseline, with the control primes scoring better. The write-up should mark this line explicitly, because the difference between a figure and a claim is exactly where this kind of argument is usually lost.
• Enumeration is arbitrary in content, constrained in cardinality: The horse-trading account is right about items — the seventeenth SDG, the thirty articles — but wrong about totals. A set whose content pressure lands on a prime is unstable; one that lands on 12, 18, 24 or 36 closes and stays closed. Eight is the modal size in the whole corpus, ten named sets, sitting at the boundary of the attention span. This is the worst possible combination for comparison. The one property frameworks reliably share is the one that carries no content.
• Chunking, and what the composite hypothesis actually says
  • Divisor count is only weakly elevated — mean 5.86 against 4.78. The operative property is not divisibility but decomposability into an array of span-sized chunks: 36 = 6×6, 24 = 4×6, 18 = 3×6, 12 = 3×4, 72 = 8×9.
  • The exceptions confirm it. 14 = 2×7, 38 = 2×19 and 92 = 4×23 are all well attested and have no interior grid — and these are precisely the entries around which the source's own account reports strain and fudging.
  • Closure is reached when a set can be silently re-read as a small array. That is a two-dimensional criterion satisfied without any two-dimensional representation.

• The unordered list is the mechanism, and it is formal rather than attitudinal
  • Two unordered sets of size n admit n! bijections. At n = 36 that is about 10⁴², and nothing in the format selects one.
  • Reconciliation is therefore not neglected. It is underdetermined by the presentation. There is no work being shirked, because the format supplies no basis for the work.
  • Structure collapses this. Impose a cube on eight items and the structure-preserving maps drop from 40,320 to 48. Impose a 6×6 torus on thirty-six and they drop from 10⁴² to 288 — a reduction of thirty-nine orders of magnitude. Sixteen on a tesseract gives 384; sixty-four on Q₆ gives 46,080.
  • The asymmetry is that completeness is what everyone defends, and articulation is what nobody specifies. Frameworks are thus most vigorously defended exactly where they are least comparable.
• The dimensional structure is already operative and unacknowledged: The chunking evidence shows these sets are functionally two-dimensional: 36 is attested because it is 6×6. The array is doing the mnemonic work while being formally denied by the presentation as a list. That is the sharper form of the complaint about neglected 2D/3D/4D mapping. It is not that higher-dimensional articulation is unexplored. It is that it is already load-bearing and unstated — which is why it cannot be argued with, corrected, or compared across frameworks.
• Where the incommensurability actually sits
  • Not between the frameworks. Between the register in which they are enumerated — cardinal, constrained, shared — and the register in which they would have to be compared — relational, unconstrained, absent.
  • The corollary is uncomfortable and worth stating plainly. Correspondences drawn on size alone are near-guaranteed to appear and near-guaranteed to be uninformative, because both sides are drawn from the same small pool of chunkable numbers. This applies to matching 36 Stratagems against 36 Dramatic Situations, and equally to matching set sizes against polyhedra, since polyhedral face and vertex counts are orbit sizes under groups of order 12, 24 and 60 and carry the same six-divisibility bias.

Asserted, and consistent with the evidence without being proved by it: that memorability is the causal driver of the size distribution; that item-level negotiation is political; and that the emirp is an apt emblem rather than a mechanism.

The constructive conclusion is narrow enough to defend. Reconciliation of frameworks cannot proceed at the level of counts, because counts are where they already agree for reasons unrelated to their content. It can only proceed once each set is given an internal articulation — and the corpus indicates that most of them already have one, unstated, in the factorisation of their own size.

Largely yes — and the size of the win is measurable: For a set of 36, the correspondence problem is 36! ≈ 10⁴² candidate mappings. Recognising the factorisation reduces it to eight candidate articulations — 2×18, 3×12, 4×9, 6×6, 2×2×9, 2×3×6, 3×3×4, 2×2×3×3. Twenty-four gives six, eighteen gives three, eight gives two.

That is the whole force of the reframe. A problem with 10⁴² answers is not a hard problem, it is a malformed one; a problem with eight answers is an agenda. So the more accurate transmission is not that there is no problem but that the problem has been stated at the wrong magnitude — treated as a question of content when it is a question of format.

• The factorisation is not unique, so the difficulty relocates rather than dissolves. Two frameworks of 36, one articulated as 6×6 and one as 4×9, remain exactly as incomparable as before. Recognition has to be joint to be worth anything, and the choice among eight is a rival-diagonal dispute of precisely the kind identified earlier — smaller by forty orders of magnitude, but the same species.
• The common ground discovered is about the knower, not the known. Frameworks converge on chunkable sizes because of working memory. Two bodies of thought agreeing on 36 are agreeing about human cognition, not about their subject matter. Real convergence, and it is convergence on a constraint rather than on a finding — which is worth having, but it is not agreement about the world and should not be reported as such.
• Primes admit no factorisation at all, and this is where the reframe genuinely fails. Ten attested sizes in the corpus are prime — 5, 7, 11, 13, 17, 23, 37, 53, 61, 89 — carrying 13% of the attestation weight. For these there is no interior array to recognise. And 13, 17 and 37 are emirps.

So the message reads better as: Houston, the problem was mis-specified — most of it factors. What does not factor is small, identifiable in advance, and is where the effort belongs.

Whether that lands as reassurance or as a sharper indictment is a matter of temperament. It does mean the labour has a definite shape: agree the articulation before comparing the contents, and expect the prime-sized frameworks to resist, permanently.

Correct, and as I stated it the conclusion was circular. "Agree the articulation" is bilateral, and bilateral agreement is exactly what is missing. Three things break the circle, and only the first two are worth much.

Articulation is unilateral where correspondence is bilateral. A framework can declare its own 6×6 without anyone's consent. Nothing is conceded, no rival is acknowledged, no shared vocabulary is required. Content correspondence needs two parties to agree; articulation needs one party to disclose. That converts a negotiation into a specification, which is a different kind of act with different preconditions.

It is disclosure, not decision. The corpus evidence says the articulation is already operative — 36 was arrived at because it factors, and the array is doing the mnemonic work already. So what is asked for is recovery of an existing fact about a set, by the people best placed to recover it, not the invention of a new convention. That is a much weaker demand than agreement, and it is the one the chunking finding actually licenses.

A third party can impose one without consent. This entire exchange has done exactly that — hanging trigrams on a cube, the live block on a tesseract, set-sizes on a rim. Imposition is contestable, but a contestable articulation is still more comparable than none, because it names what to contest.

What follows if two declarations disagree: If one framework declares 6×6 and another 4×9, they have not failed. They have produced a specific disagreement out of eight candidates, where before they had an unbounded one. On the earlier result, incommensurability arises between parties giving opposite answers to the same question, not between parties addressing different questions — and making the articulation explicit is precisely what converts the second condition into the first. Disagreement is the achievement here, not the obstacle.

Why it stays unaddressed, which is the part to be honest about: Not neglect. An unstated articulation is unfalsifiable. Declare 6×6 and the set becomes vulnerable to the objection that item 23 does not belong in row four, that the fourth column is thin, that two entries were added to reach the count. Every one of those objections is available the moment the array is on the page, and none is available while the list is flat.

So vagueness about internal structure is protective, and the protection is worth more to a framework's proponents than comparability is. That is a sufficient explanation for the pattern without appeal to carelessness, and it predicts that the practice will persist.

The conclusion in its non-circular form: Not that parties should agree an articulation. That the articulation is already there, undeclared; that declaring it costs the declarer something real; and that the cost is borne unilaterally while the benefit — comparability — accrues to third parties.

That is a collective-action problem rather than a conceptual one, which is a demotion but also a clarification. It means the leverage sits with whoever compares rather than whoever authors: an analyst willing to impose articulations and be wrong in public does not need anyone's agreement to start.

Added. And the pattern is sharper than the request implied — it isolates 13/31 by a theorem rather than by inspection.## The visual signature: closure

• Drawn as you specified, the ×2 links form a closed quadrilateral 13 · 26 · 62 · 31, and the ×3 links form a second, 13 · 39 · 93 · 31. The other three are open paths that stop dead: 17 · 34 · 43, 37 · 74 · 47, 17 · 51 · 15.
• They fail for a reason. 71×2 = 142 and 73×2 = 146 leave the ring, so there is no fourth vertex to return to. The eye reads closure directly, which is why the overlay is worth having.

Why only 13/31 — and it is forced: For a two-digit number, reversal commutes with multiplication by k exactly when every digit is under 10/k. Then k(10a+b) = 10(ka)+(kb) with no carry, and reversing before or after gives the same answer.

• k = 2 requires digits ≤ 4
• k = 3 requires digits ≤ 3

Emirp digits are drawn from {1, 3, 7, 9}, so both conditions reduce to the same admissible set {1, 3}. The two-digit numbers built from it are 11, 13, 31, 33 — and the only emirp pair among them is 13/31. Unique, and for k = 2 and k = 3 alike. The ladder then terminates: k = 4 needs digits ≤ 2, which admits only 11. So exactly three generations exist — 13/31, 26/62, 39/93 — and no more.

Note where it stops. 39/93 is one of the two dead pairs identified in the live block early on, composite at both ends. The ladder runs out by landing precisely on the block's inert corner.

The caution, which is the same one as before: Run the identical construction on 12/21 and it yields 24/42, 36/63, 48/84 — four generations rather than three, because digits {1, 2} tolerate k up to 4. Against the corpus, that ladder lands on 12, 24, 36 and 48 with 29 attested sets between them, 22% of the entire corpus. The 13/31 ladder lands on 13, 26, 39 and 62 for a total of 8.

So carry-free multiplication does look like a real generative principle for memorable set-sizes. But the generator that produces the attested numbers is the composite 12/21, not the prime 13/31. The emirp version is the more elegant object and the weaker predictor — which is now the third time that same asymmetry has appeared.

The stronger version of your intuition: If the mind is manipulating by ×3 on a ring of 100 rather than in unbounded arithmetic, the result is considerably more striking than anything the ±N framing offered.

Multiplication by 3 modulo 100 has order 20, and it splits the forty positions coprime to 10 into exactly two 20-cycles — one with odd tens digit, one with even. The odd-tens cycle is:

13 → 39 → 17 → 51 → 53 → 59 → 77 → 31 → 93 → 79 → 37 → 11 → 33 → 99 → 97 → 91 → 73 → 19 → 57 → 71 → back to 13

A single orbit containing all eight emirps, all four palindromes, both dead pairs — the entire sixteen-cell live block — plus the four positions 51, 53, 57, 59. Nothing else.

This is the multiplicative structure the live block was shown to have at the outset, since (ℤ/10)* is generated by 3, now realised as one traversal rather than as a static labelling. A mind chunking by repeated tripling would visit every emirp in twenty steps without ever leaving the odd-tens region — which is a far better account of "manipulating key primes in the chunking process" than proximity ever was. I can draw that 20-cycle as a single closed path on the same ring if it would be useful, either as a further overlay or as a clean third sheet.

Sorting the plausible repertoire by kind:

ρ is the Spearman correlation between a number's degree under that operation and how heavily it is attested in the corpus

Multiplicative operations correlate; notational and metric ones do not. Divisor degree, doubling and tripling sit at 0.38–0.43. Reversal, nines complement and digit sum sit at 0.08–0.18.

Your intuition that reversal is of lower order is confirmed on both criteria at once. It is an involution, so its orbit length is 2 and no travel is possible — you reach a partner and stop. And it carries almost no relation to which sizes are actually used.

There is a caution embedded here that is worth more than the correlations. The two most visually symmetric layers are the two least informative. Reversal produces the seventeen parallel pencils of the original figure; nines complement produces a single clean mirror pencil of 49 chords. Both are beautiful and near-inert. Doubling produces an unlovely asymmetric fan of chords of wildly varying length, and it is one of the strongest predictors. Pattern-seeking that rewards visual symmetry will systematically select the wrong operations.

Why ×2 and ×3 are not equivalent — the McClain point made exact:

• Two is not invertible modulo 100, since it shares a factor with the modulus. Doubling is therefore a ray: 1→2→4→8→16→32→64 and then out. It terminates.
• Three is invertible, and its multiplicative order modulo 100 is 20 — the maximum possible, since the unit group is ℤ/2 × ℤ/20. Tripling circulates.

So on a ring of a hundred, the octave runs out and the fifth returns. That is not an analogy to the musical situation; it is the same fact, since octave equivalence must be imposed by fiat while the cycle of fifths closes on its own. Three is the smallest element of maximal order, though not the only one — 13, 17, 23, 27 and 33 also have order 20.

The strongest single predictor in the whole exchange: Composing the two generators gives the 3-smooth numbers, 2ᵃ3ᵇ under 100 — twenty of them, forming a 7×5 rectangular lattice:

1, 2, 3, 4, 6, 8, 9, 12, 16, 18, 24, 27, 32, 36, 48, 54, 64, 72, 81, 96

Against the corpus: 50.4% of attested weight on 20.2% of the numbers, a lift of 2.49× — better than multiples of six and better than the griddable criterion, and it arrives from a generative principle rather than a divisibility test.

• Reversal — seventeen parallel pencils, striated, symmetric about the digit-sum axes
• Nines complement — one parallel pencil, a single mirror
• Doubling — chains radiating from each odd number, chords lengthening geometrically, no symmetry
• Tripling — five closed orbits, one of which is the twenty-cycle carrying every emirp
• 3-smooth lattice — a rectangular grid embedded in the ring, the two generators as two chord directions
• Divisor lattice — a dense fan converging on 12, 24, 36, 48, 60, 72, 84, 96
• Proximity — rim-hugging arcs, everywhere, saying nothing

The natural sheet is stratified by density, drawn in that order, so that the sparse informative layers are legible before the dense inert ones bury them. I can build that — either as one figure with the layers distinguished, or as the single McClain lattice sheet, which is where the actual result sits.

No two-digit reversal pair is ever also a multiplication pair. The reversal field and the multiplicative field are disjoint as edge sets, necessarily. They cannot overlap, so the question of why they fail to coincide has an answer prior to any empirical inspection.

Six of the forty-five reversal pairs are reinforced at all, all with digits ≤ 4. 12/21 is the unique pair surviving ×2, ×3 and ×4, and its multiplicative closure is {12, 24, 36, 48} — the four most attested set-sizes in the corpus. Of the emirp pairs only 13/31 qualifies, closing on {13, 26, 39}.

So the diagram's reinforced links are a definite and tiny sublattice, and the rest of its symmetry is genuinely decorative in the strict sense that nothing else in the system refers to it.

• n + rev(n) = 11 · ds(n) and rev(n) − n = 9 · (b − a)

The third vertex of every reversal triangle lies on the palindrome diagonal; the third vertex of every reversal difference lies on the multiples of 9. Base ten ± 1, and both diagonals are already marked in the figure.

For the emirps: 13+31 = 44, 17+71 = 88, and then 37+73 = 110 and 79+97 = 176 leave the ring entirely.

The difference diagonal is favoured almost twofold; the sum diagonal is avoided almost fourfold. Only 11 and 33 are attested at all among the palindromes, and weakly.

That is a real reason to prefer the difference as the generative ternary operation — and it recovers the reverse-and-subtract cycle from the earlier session as the operation the corpus actually endorses.

The register the Blue Brain result actually lives in: This is where the reframing pays. Reimann and colleagues found directed cliques and cavities — a simplicial complex, not a graph. Binary operations only ever give edges. Simplices require either cliques or genuinely ternary relations, which is exactly why your a+b=c instinct is the right move.

The divisibility poset supplies them natively. A maximal chain 1 | 2 | 4 | 8 | 16 | 32 | 64 has seven elements and is therefore a 6-simplex in the order complex, and the dimension a number supports is Ω(n), its prime-factor count with multiplicity.

Attested set-sizes support higher-dimensional simplices than chance. The maxima under 100 are 64 and 96 at Ω = 6, then 32, 48, 72 and 80 at Ω = 5 — and 64 is the I Ching, which was the one number the ±N scheme conspicuously could not reach.

Alongside this the product relation {a, b, ab} gives 142 triangles under 100 — the multiplication table read as a 2-complex, with 2, 3, 4, 5 and 6 as the high-degree vertices.

Operations not yet on the table: Worth naming, roughly in descending promise:

• reverse-and-subtract iterated (the 09→81→63→27→45 cycle);
• digit product and multiplicative persistence;
• concatenation;
• Pythagorean triples; and constant-sum triples a+b+c = k, which generate the magic-square structures and are the natural bridge to 2D layout.

The pattern across all of this is consistent and slightly uncomfortable: the operations that look most symmetric are the least reinforced, and the ones that build dimension are asymmetric and ugly. Reversal, nines complement and digit-sum pencils produce the beautiful striations; divisibility and multiplication produce a lopsided fan and all of the predictive weight.

I can draw the reinforced sublattice — the six surviving reversal pairs with their ×2/×3/×4 links, and the difference-diagonal triangles — as a stripped sheet where only the links that are actually reinforced appear. That would show, by subtraction, how much of the original figure's symmetry is decoration.

Let me check one thing computationally first, because I think there's an exact answer available.Wikipedia confirms the Earth Charter structure: the roughly 2,400-word document is divided into pillars containing 16 main principles with 61 supporting principles, opening with a preamble and closing with a conclusion entitled "The Way Forward". Let me verify the polyhedral arithmetic before drawing conclusions.## Yes — and the correspondence is exact, which I had not seen

The rhombicuboctahedron and the rhombicosidodecahedron are the same Conway word e applied to the two incommensurable seeds. They were landmark six of the two morph animations built earlier in this exchange.

Both are 4-valent, so E = 2V, so by Euler F = V + 2. And in both cases the vertex set is a free orbit of the rotation group, so V = |G|. Hence:

26 = |O| + 2 and 62 = |I| + 2. Forced, not coincidental.

So your 26-fold mapping already has an icosahedral counterpart, and it is exactly 62. The Stockholm and Ison–Straw sets sit on the octahedral member of a pair whose other member has 62 faces. That is a much stronger warrant for the pairing than the axis-count route, because it is the same operation rather than two analogous quantities.

But the Earth Charter's 61 is not a comparable object: This is the finding that matters, and it is deflationary in an interesting way.

• 26 is a designed count. Stockholm's drafters produced twenty-six principles as a flat list; Ison and Straw likewise. The number was chosen, or at least arrived at as a total someone could see.
• 61 is an emergent sum. Each of the 16 principles carries its own lettered sub-clauses — a, b, c, d — chosen locally for content. The 61 is whatever they add up to. Nobody designed a 61-fold set; nobody could hold it as a set.

That distinction cuts across the whole exercise. A designed count is evidence about closure; an emergent sum is evidence about nothing except the sum of sixteen unrelated decisions. And it explains why 61 is prime without any significance attaching: a sum of sixteen small arbitrary integers lands on a prime about a quarter of the time.

So 61 → 62 is a near-miss to a number that was never being aimed at. On the discipline of this exchange it should be graded as coincidence.

Where the real structure is: the 16: The Earth Charter's designed count is 16, in four pillars of four. That is a 4×4 array — Q₂ □ Q₂ = Q₄, the tesseract — the same object that carried the live block of the digit table earlier.

Four pillars of four principles is a genuine two-axis articulation, and it is stated rather than inferred. Which makes the Earth Charter unusual: it is one of the few frameworks in the corpus that declares its own internal structure rather than presenting a flat list. On the argument developed earlier — that reconciliation is underdetermined because articulation goes undeclared — the Earth Charter is a positive case, and its 16 is the number to map, not its 61.

The SDGs at 16+1 and the Earth Charter at 16 in 4×4 then become directly comparable, both tesseract-shaped, which is a claim that can be checked against content.

What would make the 62 reading defensible: If the 61 supporting principles were distributed uniformly — say a fixed number under each of the 16 — then the total would be designed and the near-miss would be worth something. If they are ragged, it is an accident.

That is one afternoon's counting from the text, and it is the check I would want before putting any weight on 61. My expectation is ragged, since the Living Centre extract shows principle 5 with four sub-clauses and the pattern varies.

The stronger construction, if you want the icosahedral counterpart of your 26-fold work, is to build it on 62 as eD's faces — 20 triangles, 30 squares, 12 pentagons, a three-class articulation the rhombicuboctahedron's 8+18 also has — and to find or construct a 62-fold set for it, rather than to recruit the Earth Charter's 61 by adding "The Way Forward" as a sixty-second element.

Three sets now, not two — and the third is of a different logical type. Stockholm 1972 is normative-legal (what ought to be protected), Ison–Straw 2020 is procedural (how governing should be done), Rao–Shively 2026 is diagnostic-pathological (what is failing). That the same cardinality recurs across prescription, process and pathology is more interesting than either earlier pairing alone, precisely because the sets are not competing for the same descriptive slot.

One deflationary caveat first. The new 26 is explicitly indexed A–Z. That is a fourth species of "designed count" — neither negotiated (Stockholm) nor theorised (Ison–Straw) but alphabetically closed, with the idiom "A to Z" doing rhetorical work as a claim of exhaustiveness. The paper's own methodology concedes the retention criteria were applied by the two authors without independent multi-rater coding, and §7.1 admits heterogeneous evidentiary rigour across dimensions. So the count is constrained by the Latin alphabet at least as much as by the phenomena. This matters for mapping: it means the 26 is available for polyhedral placement (a designed, holdable set) while being weak evidence of closure.

What the paper adds that the earlier sets lacked: Your framing is right, and it is the decisive difference. Stockholm and Ison–Straw are flat lists — 26 items with no declared adjacency. Rao–Shively supply, in Figure 1, a partial dependency graph and, in Figure 2, four named cascade loops, with §3.5 asserting cluster-level drivers (animal agriculture → A–G, I–K, X; endless growth → A–G, R, Z, V, H, M). A polyhedral mapping of a flat list is arbitrary up to relabelling; a mapping of a graph is constrained, and therefore testable.

Critically, §5.8 and Figure 3 also supply an A–Z response set in one-to-one correspondence with the A–Z crisis set. So the paper already contains a problem/solution dual pair of 26 — which is structurally what a face-set and its dual vertex-set provide.

The clustering is exactly octahedral — but only in paired form:

• Clusters: Ecological A–G (7), Personal H–M (6), Institutional N–T (7), Social U–Z (6).

7 + 6 + 7 + 6 = 26, and this cannot be an orbit union on either 26-faced Archimedean, because every face-orbit under O_h is even (6, 8, 12) — antipodal inversion pairs faces, so no odd part is symmetry-realisable. The four-cluster partition therefore fails as a direct face-colouring.

But it succeeds immediately under antipodal pairing, which is where the 13-fold enters:

3 + 4 = 7 and the two-fold class is 6. The decomposition is forced, not fitted: the 13 rotation axes of the octahedral group split as 7 + 6 in exactly the way the paper's cluster sizes require, provided each ecological crisis is placed antipodally to an institutional one and each personal crisis antipodally to a social one. Faces consumed: 14 for ECO/INS, 12 for PER/SOC.

This reading is substantively motivated by the paper itself, which treats the institutional crises (notably O, P, Q) as the drivers of which the ecological crises (A–G) are the manifestations, and treats the personal cluster (H–M) as the interiorisation of the social one (U–Z). The antipodal axis thus reads consistently as driver ↔ manifestation on the 7 cube-face/cube-vertex axes, and interior ↔ collective on the 6 edge axes.

Independent corroboration from the existential flags. Of the 16 dimensions marked "e", 13 fall in ECO+INS (6 of 7 and 7 of 7) and only 3 in PER+SOC (2 of 6, 1 of 6). The 14/12 cleavage is therefore not an artefact of the pairing hypothesis — it is where the paper's own risk-weighting concentrates, 13 out of 14 against 3 out of 12. The recurrence of 13 here is arithmetically incidental but worth noting given the 13–26–62–31 line of enquiry.

The 3D proposal

• Single-solid, triple-inscribed. Because all three sets index the same 26 axis-directions, they can be carried simultaneously — Stockholm principle, Ison–Straw principle and Rao crisis on each face, read radially. This is the strongest argument for treating the coincidence as more than numerological: not that the sets resemble each other, but that they can be co-located on one geometry and their misalignments then become visible as specific empty or overloaded directions.
• RCO or TCO. The RCO (8△ + 18□) makes a binary reading salient — but it collapses the 4-fold and 2-fold classes into one shape, i.e. it hides the 7/6 cluster split. The TCO (6⯃ + 8⬡ + 12□) keeps all three axis classes visually distinct and is therefore the better host for this particular mapping, contrary to the earlier preference. Ecological/institutional pairs occupy octagons and hexagons; personal/social pairs occupy the twelve squares.
• Dual for the PolySolution. Place crises on TCO faces and the A–Z remedies of §5.8 on the vertices of the disdyakis dodecahedron (26 vertices, dual of the TCO) — or crises on RCO faces against remedies on the deltoidal icositetrahedron's 26 vertices. Problem and response then occupy dual positions of one form rather than two separate diagrams.
• Testing the edges. The RCO has 48 face-adjacencies, the TCO 72. Rao's declared dependency graph is denser than either (the text claims "thousands" of loops), so no isomorphic embedding exists. The tractable version is an assignment problem: find the labelling of the 26 faces minimising total graph distance between dependency-linked crises. Whether the resulting optimum recovers the four-cluster/antipodal arrangement above, or contradicts it, is the falsification test — and it is exactly the test that could not be run on Stockholm or Ison–Straw for want of any declared links.
• The icosahedral counterpart remains open. 62 = |I| + 2 as 26 = |O| + 2; the Earth Charter's 61 was found not to be a comparable object because it is an emergent sum. Nothing in the new paper supplies a 62-fold set, so the icosahedral branch is still unoccupied.

The honest summary: the coincidence of cardinality is weak evidence, but the coincidence of internal decomposition — 7 + 6 across 13 axes, corroborated independently by the 13/3 split of the existential flags — is considerably stronger, and is the first such structural match among the three sets.

The omission was substantive, not incidental — the nine is doing structural work in that paper, and it explains the seven.

The ecological septet is the transgressed nine: Rao–Shively's §4.2 states that seven of the nine planetary boundaries have been crossed. Their ecological cluster A–G has exactly seven members, and each maps one-to-one onto a transgressed boundary: A mass extinction = biosphere integrity; B chemical/plastic = novel entities; C nutrient cycles = biogeochemical flows; D overheating = climate change; E ecosystem collapse ≈ land-system change; F freshwater depletion = freshwater use; G ocean acidification.

The two boundaries omitted are stratospheric ozone and atmospheric aerosol loading — precisely the one that is not transgressed (and is recovering, the single unambiguous success of multilateral environmental governance) and the one that remains unquantified. So the 7 is not an independent taxonomic judgement. It is 9 minus a governance success and minus a measurement gap.

That has a sharp consequence for the remedial-capacity argument. A diagnostic taxonomy constructed by deleting the one case where collective action demonstrably worked has removed the only evidence in the set bearing on whether collective action can work at all. The paper then proposes two MegaSolutions requiring collective capacity of unprecedented scale — while having excised, at the level of set construction, the one instance from which the track record could be read. The Montreal Protocol is the missing member.

Casting out nines is exactly the RCO decomposition: 26 ≡ 8 (mod 9), so 26 = 18 + 8 with 18 = 2 × 9. This is not a loose resonance: it is the actual face decomposition of both 26-faced Archimedean solids.

• RCO: 8 triangles + 18 squares
• TCO: 8 hexagons + (6 octagons + 12 squares) = 8 + 18

In both, casting out the nines removes precisely the four-fold and two-fold axis classes and leaves the eight three-fold (cube-vertex) directions. In axis form: 13 = 4 + 9. The nine is an axis count — the 3 four-fold plus 6 two-fold axes — and the eighteen is its face count, i.e. nine antipodal pairs. So a 9-fold set has a canonical home on the 26-fold: it occupies the 18 squares as 9 opposed pairs, leaving the 8 triangles free.

And this is where the two 9-fold sets belong. The 2009 argument was that planetary boundaries and remedial-capacity boundaries are two "eyes" requiring stereoscopic reading. Geometrically that is the antipodal relation: each of the 9 axes carries one planetary boundary at one pole and its corresponding remedial-capacity boundary at the other. Underallocation of funding, delays/lags, inadequate coordination, misrepresentation, ineffectiveness, malfeasance, institutionalized inequality, unconsidered dissent, investment in distractants — nine constraints, nine opposed poles. The 8 remaining triangles are then available for the 8 transactional games of the coaction cardioid, which would complete the solid without residue.

Honest limit: this fails at 62. The rhombicosidodecahedron's orbits are 20 + 30 + 12, and no union equals 54; nor do the icosahedral axes (6 + 10 + 15 = 31) admit a subset of 27. Casting out nines therefore selects the octahedral member of the pair and only that one. It is a base-10 fact meeting one polyhedron, not a general principle — which is itself informative, since it means the 9-fold has no icosahedral counterpart and the two 26-fold candidate hosts behave identically under it while the 62-fold does not.

27 − 1 = 26: three nines and a missing centre: The strongest link is not modular but combinatorial. A 3×3×3 array is three layers of nine; delete the central cell and exactly 26 remain — 6 face-adjacent, 12 edge-adjacent, 8 corner-adjacent, the same 6+8+12 as the octahedral axis-directions. The 26-fold is the 3D Moore neighbourhood, and there are three orthogonal ways to slice it into three nines.

This gives the polyocular framework an exact geometry. Problematique (9), remedial capacity (9), imaginatique (9) = 27; the 26-fold is what remains when the centre is vacated. And the 2009 text names that vacancy directly: there is no "we" with a shared awareness permitting coherent action. The centre cell is the absent collective subject — surrounded on all 26 sides, occupying none of them. That is a considerably better mnemonic for the "we" problem than the Mandelbrot proposal, because the absence is forced by the arithmetic rather than illustrated by it.

What Rao's 26 lacks: Set the 9 remedial-capacity boundaries against the crisis A–Z. Four find partial counterparts — institutionalized inequality ↔ Z, misrepresentation ↔ N, malfeasance ↔ V/W, distractants ↔ L. Five have no counterpart at all: underallocation of funding, unconsidered dissent, inadequate coordination, ineffectiveness, delays/lags.

That is the diagnostic asymmetry. The taxonomy is dense in what is going wrong in the world and near-empty in what goes wrong in responding to it — twenty-six existential dimensions, none of which is "we have repeatedly failed to do this". Its institutional cluster N–T names industries and technologies as runaway, not institutions as incapable. So the two 26-fold governance-principle sets and this 26-fold crisis set share a blind spot the 9-fold psychosocial set was constructed to expose.

The composite mapping therefore has a natural division of labour: crises and principles on the faces, remedial-capacity constraints on the antipodes of the nine, and the vacated centre standing for the subject none of the three sets supplies.

A striking result, and it lands exactly where you point.## The 18N operation, and why it is the strongest result here

Both digits of any live-block number are odd, so b−a is even, and rev(n) − n = 9(b−a) is forced to be a multiple of 18:

This is strictly stronger than the mod-6 congruence I derived at the very start, since 18 = 3×6. And it is the difference operation — the one the corpus endorsed at 1.96× while rejecting the sum at 0.25×.

The six pairs generate exactly {18, 36, 54, 72}. Against the corpus that is four numbers, 4% of the range, carrying 19.5% of all attested sets — a lift of 4.8×, the largest of anything tested in this exchange. 18 and 36 are the second and joint-first most attested sizes in the whole corpus.

90 = 18×5 is unreachable: it would require a digit gap of ten.

How the figure handles the clutter: Three registers, separated by contrast rather than by removal.

• A, the ground. All 45 reversal chords retained at low opacity. They stay because they are what the eye reads, and because their non-reinforcement is now a theorem rather than a deficiency — no two-digit reversal pair can also be a multiplication pair.
• B, the reinforced. The pairs with digits under 5, in amber, with their carry-free multiplication links dashed. 12/21 alone survives ×2, ×3 and ×4, closing on {12, 24, 36, 48}.
• C, the forced. Nested rim arcs in six lanes, one per live-block pair. The arc's span along the rim is the difference, so the geometry states the arithmetic directly, and the target nodes carry inward bars showing attestation.

13/31 is drawn in its original red because it is the sole member of both B and C — which is the structural answer to why that pair, rather than the emirp set at large, keeps returning.

On respecting the decorative: The point is well taken and the figure is built on it. Register A is not error and not noise; it is the register in which the claim is actually made, and stripping it would misrepresent the phenomenon rather than clarify it. What the stratification buys is that the two registers stop competing: the decorative carries the sense, the forced carries the check, and one can see where they happen to coincide — which is rarely, and at 13/31 exactly once.

The honest caution is that 18 and 36 carry most of that 4.8× lift, and both are also multiples of 6, 9 and 12. The 18N result is arithmetically forced, but its corpus performance may be the divisibility effect once more, wearing its best costume.

• Ten movements over eighty frames, cross-faded rather than cut, on one ring of a hundred positions. Every operation persists at a low ghost level throughout, so the field never empties — what rises and falls is illumination, not existence. That is the specific point you asked for: the figure at any instant is a superposition, and only the emphasis travels.
• The order is a narrative rather than an arbitrary sequence: notational first (reversal, nines complement, digit-sum pencils), then metric, then multiplicative (doubling, tripling, the ×3 orbit, the divisor lattice, the 3-smooth grid), then the forced 18N relation and the emirp bridges as the close. Decorative to forced, in the sense established in the last exchange.

• Reversal and nines complement do not move. They are standing patterns — parallel pencils that appear and fade in place. This is the involution property made perceptible: orbit length two, nowhere to travel.
• Doubling and tripling sweep. Chords lengthen geometrically outward from each odd generator, and the layer has a direction.
• Only one movement actually travels. The ×3 orbit carries a pulse around its twenty-cycle, passing through all eight emirps in a single circuit. Watching it is a better argument for that structure than the static chord set was.
• The 3-smooth grid emerges from the dense field. Coming immediately after the divisor lattice, the twenty McClain positions resolve out of 276 chords into a legible rectangular lattice — which is roughly what the corpus statistics said was happening.

The operations do not commute, and their failures to commute are exactly the carries. Reversal composed with doubling equals doubling composed with reversal precisely when every digit is under five — which was the theorem that isolated 12/21 and 13/31. So the "dance" has a rigorous handle: it is characterised by its commutators, and the places where two operations commute are exactly the places where structure gets reinforced rather than merely coexisting.

That suggests the more informative animation would show two operations at a time with their commutator highlighted — the positions where applying them in either order agrees, lighting up as the pair cycles. That would render coherence-at-the-level-of-the-dance as something checkable rather than something felt. I can build it if useful.

The chess and go comparison is worth holding at the same distance. That strong players see dynamic relations rather than static positions is well described; that this is the same faculty Ramanujan exercised is a redescription rather than evidence, and the analogy would need to survive the observation that chess positions have a ground truth against which pattern-perception is continuously corrected, while a field of numbers offers no such correction unless one is deliberately constructed — which is what the corpus tests in this exchange were.

First, a correction that matters: The Blue Brain result is about directed cliques — all-to-all connected neuron groups with a consistent flow direction — and cavities in the resulting flag complex. It is a statement about a simplicial complex, and the dimension is the clique size minus one.

Measured against that, the constructions in this exchange come out badly, and it is worth saying so plainly:

Both morph skeletons are triangle-free, and provably so: a triangle would require r_i r_j = r_k, which no Coxeter group on distinct mirrors permits. Bipartite graphs — which all these are — cannot contain odd cycles at all. So none of the figures built here reaches even dimension 2 in the Blue Brain sense, and the torus interconnect fares no better.

That is a real limitation rather than a technicality. The tower of blocks cannot be built on any of these skeletons.

Where the two families do differ, and it is at dimension 2:

The orbits, as opposed to the skeletons, do carry triangles — and the counts separate cleanly:

Neither family produces a single tetrahedron. Both cap at dimension 2, and the difference between them is a factor of 2.5 in how much dimension-2 structure they can carry — 8 against 20 triangles. So if the two are to be distinguished in these terms, the distinction is capacity at a fixed low dimension, not access to different dimensions. Which is a narrower claim than the question supposes, but a checkable one.

The pattern across the walk is the more suggestive finding: Triangles appear at the corners and at some edges, and vanish entirely at the omnitruncate — the position of maximum vertex count. Reading the morph as a temporal sequence:

8 → 0 → 8 → 0 → 8 → 0 → 0 octahedrally, and the same shape icosahedrally.

Dimension-2 structure builds and razes as the seed walks, and it is lowest where the figure is largest. That is at least the right shape for the sandcastle description — materialising and disintegrating — and it locates the difference where your question puts it: in the morphing rather than in the static forms.

The caution is that this is a property of the orbit geometry, not of any neural process, and reading it as one would be importing.

How the two are interrelated — the answer already computed: Not by containment. Neither group contains the other, and their largest common subgroup is the tetrahedral group of order 12 — the seven shared axes.

The tetrahedral cell is the interesting part here. Among the thirteen Archimedean solids, only the truncated tetrahedron has rotation group A₄, so it sits in neither family and belongs to the intersection. And in the earlier truncation argument, the tetrahedron was what appeared when a vertex shared by four cells was articulated — a mediating body touching each party on a face and contained by none.

If a two-family system needs a place where dimension can be built jointly, the geometry says it is the tetrahedral commons, and that it is small.

What I would not claim: That octahedral and icosahedral processing are distinct cognitive modes. Nothing here supports it, and the honest statement is the reverse: both families are dimensionally impoverished by the Blue Brain criterion, and equally so — capped at 2, with no tetrahedra anywhere.

The construction that would meet the criterion is different in kind. Directed cliques of dimension 4 and above need dense, directed, non-bipartite connectivity — closer to the divisor lattice, where a chain 1|2|4|8|16|32|64 is a genuine 6-simplex, than to any polyhedron or torus. That is worth noting because it points somewhere specific: the structure in this exchange that best matches the neural description is the divisibility poset, not the geometry — and attested set-sizes were shown to support higher-dimensional simplices there than chance would give.

You are right to press this, and the bias runs in the direction that would inflate exactly the findings I reported. Let me quantify how much would be needed to overturn them.## The bias runs toward the findings, and here is how much would be needed

"Contamination needed" is the fraction of attestation weight that must have entered the corpus because of its size for the lift to arise from a null of no preference.

The largest lifts are the most fragile. The 18N result needs only 16% selective inclusion, and capping each size's weight at three drops it from 20.3% to 13.3% — most of it rides on 18 and 36 alone. Given that the source document is explicitly about mapping set-sizes onto polyhedra, and polyhedral counts are six-divisible by construction, 16% contamination is entirely plausible. I would now treat the 18N corpus result as unsupported, though the arithmetic forcing it remains a theorem.

The griddable and 3-smooth results need 38–55% contamination and survive weight-capping. Those hold up better.

The one finding the bias cannot produce: Primes are attested at 12.8% against a 25.3% baseline — suppressed by half. The contamination figure comes out negative, meaning no amount of size-selective inclusion explains it.

More to the point, the bias runs the wrong way: a document preoccupied with emirps should over-represent prime sizes if anything. It under-represents them twofold. Avoidance of primes is therefore the most trustworthy result in the whole exercise, and it is the one that most directly supports the chunking account.

Why the obvious way of building the corpus cannot work: Assembling entries by recall, or by searching for "sets of N principles", is generated by the very process under test. Memorable numbers are recalled and indexed preferentially — that is the hypothesis. Any such corpus confirms it regardless of the truth.

The > analogy is instructive but imperfect for the same reason. > is a curated collection: sequences enter because someone found them interesting. An >-for-sets would be searchable and equally biased. The right model is a registry with pre-specified inclusion criteria — closer to a systematic review protocol than to an encyclopedia.

The requirement is a size-blind sampling frame: enumerate the sources first by a criterion having nothing to do with counts, then count. Workable frames include national constitutions from a fixed list of states (article and chapter counts); treaties in the UN Treaty Series; codes of ethics across the member bodies of a professional federation; ISO management standards by clause count; and books by a fixed set of publishers or authors whose titles enumerate. In each case the entry criterion is the institution or document, never the number.

What I can and cannot do: I can specify the schema and coding manual in full — what counts as a set, how to treat nested or sub-enumerated items, how to version things like the SDGs at 16 then 17, how to record the frame and provenance per entry, and how to flag inferred entries. I can build the analysis harness, including the control-prime comparison, the contamination calculation above, and the power computation.

I can populate a defined frame by search, but only at the scale of some tens of entries in a session, each with a citable source.

I cannot systematically enumerate hundreds of documents reliably in one pass, and I should not generate entries from memory — that would be precisely the contaminated instrument.

If useful, I can draft the registry schema, the coding manual and the analysis harness as working files, and run a pilot on whichever frame seems most tractable. Constitutions are probably the best first candidate: the population is closed, the counts are unambiguous, and nobody drafts a constitution to hit a memorable number of articles — which makes it a near-ideal null frame against which the deliberately-composed sets can be compared.

Only the second improves asymptotically with N, which is why the ±2 intuition felt right but under-specified.

Twelve is not chosen for memorability. It is where a ninefold gain in accuracy is bought for a 2.4× increase in N, and where the next gain is only fourfold for 3.4× the cost. Pentatonic and chromatic are the two affordable convergents; 41 and 53 exist and are used only by specialists.

So "coherence approached asymptotically" is right, but the approach is a staircase, and closure sits on the tread where the next riser is too tall. That is a criterion one could actually apply — wherever a set approximates something, the convergents are computable in advance.

Why aesthetics senses it more clearly: Because aesthetic closure is temporal, not cardinal. A sonnet closes because the rhyme resolves; fourteen is a consequence, not a target, and a reader knows the poem has ended without counting. A cadence closes a phrase. A narrative closes when the tension discharges.

A list of principles has no such internal signal. Closure there is cardinal — it can only be certified by a completeness criterion the compilers do not possess.

This yields the explanatory claim that ties the whole exchange together: chunkable N is a substitute for structural closure. Sets that possess real closure — provable, functional, or resolved — have no need of an arithmetically convenient count. Sets that possess none borrow one from arithmetic, because a well-shaped number is the only completeness signal available.

he research design this produces: Three frames, differing in what supplies closure, both alternatives to the contaminated corpus:

• Exhaustively closed sets — classification theorems. The population is enumerable from standard references and the counts are not at anyone's discretion. Size-blind by construction.
• Functionally closed sets — team sizes, jury and committee sizes, crew complements. N is selected by performance feedback, not by transmissibility.
• Doctrinally closed sets — the existing corpus.

Prediction: chunkability rises and prime-richness falls across (a) → (b) → (c).

Indicative figures, on distinct values:

The multiple-of-6 gradient runs cleanly in the predicted direction, from 6% to 41%. The prime gradient holds only against the weighted doctrinal figure. I should be explicit that I hand-listed (a) and (b) from memory, which is precisely the failure mode identified last turn — the counts are not at my discretion, but which theorems I recall is. Treat these as a design sketch, not a result.

Extending to aesthetics: The aesthetic frames are unusually valuable because they contain both closure types within the same repertoire, which controls for period, composer and genre. Bach's 24 preludes and fugues are exhaustively closed — twelve keys times two modes, no discretion. His 30 Goldberg variations are compositionally closed, and the 30 is 10 groups of 3 with a canon at every third. Comparing the two within one composer isolates the effect of closure type from everything else.

The obvious frames: variation sets and their counts; sonnet sequences and fixed forms; movement counts within cycles; canon sizes where a canon is officially closed; and prosodic inventories. Beside them, forms whose closure is purely structural — villanelle, sestina, ghazal — where N is fully determined by the scheme and should show no arithmetic preference at all.

If that last prediction holds — that formally-determined N sits exactly at baseline while doctrinally-determined N is elevated — it would be the cleanest available demonstration that the divisibility bias is about transmission rather than about number.

It changes the design at the level of the primary hypothesis, not merely by adding fields.

• Every figure computed in this exchange was measured against a uniform null over 1–99, which was always the weakest assumption in the argument. There is no reason set-sizes should be uniform on the integers.
• The closure typology removes that assumption. Classification sets are closed by proof and cannot be selected for transmissibility, so stratum (a) is an empirical null and the outcome becomes a contrast between strata rather than a lift over an invented baseline. That is a genuine methodological upgrade rather than an elaboration.

So the cross-frame 3-smooth comparison I gave last turn (31%, 40%, 34%) is uninterpretable — it is measuring size range, not structure. The prime gradient is partly confounded the same way.

Only the multiple-of-6 measure is magnitude-flat, which retrospectively validates that gradient — 6% for classification, 20% for teams, 29% for doctrinal distinct, 41% weighted — as the one result that means what it appears to mean. It should be the pre-registered primary outcome, with everything else conditioned on size band or tested against a permutation null that preserves each stratum's size distribution.

What the schema now has to carry: Closure type as a coded variable rather than a frame label; whether a completeness proof or criterion exists; whether N is derivable from a generating rule, and if so which; whether the set is transmitted orally.

Two cheap, high-value additions the aesthetics discussion suggests. First, does the source itself claim exhaustiveness? — an observable textual feature distinguishing "the seven X" from "seven of the X". Second, has N changed? The SDG shift from 16 to 17 stops being a nuisance and becomes evidence: sets whose count drifted are direct observations of negotiated closure, and the direction of drift is testable.

Sets whose N is derived from another set's N — 24 preludes as 12 keys × 2 — need coding as dependent rather than silently included as independent observations.

Analysis plan and scale: The convergent test becomes a separate, sharper analysis applying only to sets that approximate something: compute the convergents in advance, then ask whether attested N sits on one. That is a point prediction rather than a class membership, so it needs far fewer cases to be informative.

Power now applies per stratum and to a contrast. Detecting the observed 6%-versus-29% gap at 80% power needs roughly 45 entries per stratum; detecting a subtler 16%-versus-28% gap needs about 180 each. So the target is 50 to 200 per stratum, three strata — more than the single-corpus figure I gave, but each stratum is easier to enumerate because the frames are closed.

Pre-registration matters more than it did. A typology adds researcher degrees of freedom, so the closure-coding rules, the primary outcome and the magnitude adjustment all have to be fixed before any counting.

I can draft this as three working files — registry schema, coding manual with worked classification decisions, and the analysis harness with the permutation null and sensitivity calculation built in — if that would be useful.

That is the right reframing, and it is testable. Let me run it properly rather than speculate.The emirp seeds fail badly. But the result points somewhere specific, so let me follow it.That is a positive result, and it answers your question directly.

Corpus clusters are not random — but the seeds are not emirps: The corpus is generated by {2, 3} specifically. Adding 5 or 7 as generators dilutes rather than improves:

And the near-miss structure you are asking about exists — around 3-smooth numbers rather than around emirps. Of the 27 attested sizes that are not themselves 3-smooth, mean distance to the nearest 3-smooth number is 1.93 against 2.51 expected, p = 0.0092.

Reading the list is more persuasive than the statistic: 5→4, 7→6, 10→9, 11→12, 13→12, 15→16, 17→16, 23→24, 25→24, 26→27, 28→27, 33→32, 37→36, 53→54, 63→64, 95→96. Almost all at distance 1. Let me test that as a scheme rather than as an observation.## The answer to your fundamental question

The clusters are not random. Tested against matched-size random seed sets by permutation:

Note the second row. The emirp seed model at reach 1 performs respectably — better lift than the 3-smooth shell, and a smaller p. My earlier test compared emirps against control primes and found no separation; this one compares against random seed sets and finds the structure is not accidental. Those are different questions, and I only asked the first.

The irony you identify is exact, and it is the finding: 3-smoothness is prime-based — generated by 2 and 3. So the corpus is organised by primes, but as generators rather than as landmarks. That is the distinction the whole exercise missed. Emirps were treated as points to sit near; the operative primes turn out to be points to multiply from.

Your reframing of emirps as seeds is therefore the right move, and it converts a proximity hypothesis into a generative one. Under multiplicative closure the emirps do badly (lift 0.5–0.7 under ×2, ×3) — but 12/21 reaches 2.75 and {1} reaches 2.49, so the closure framework itself works. What is unresolved is which seeds.

The honest caveat, which is now the binding constraint: I have run roughly fifteen hypotheses on one convenience corpus of 41 points. A nominal p of 0.01 corrects to about 0.14 under Bonferroni. These figures identify candidates; they cannot establish any of them. I have now made the opposite error to the one in the summary if I let them stand as findings.

What would settle it: Three things, and they are specifiable in advance rather than after the fact.

• Pre-register the seed sets and operations. Emirps, 3-smooth, 12/21, control primes; operations ×2, ×3, carry-free multiplication, the ×3 orbit, reach 0/±1/±2. That is a fixed grid of about twenty models, declared before any new data is seen.
• Get 150+ size-blind entries. At 41 points the resolution is ±15 percentage points; the differences in dispute are 5–10. The protocol already specifies the frames.
• Split the corpus. Fit on half, test on half. That is the one procedure that survives the multiple-comparison problem, and none of what I have run today would survive it.

The clustering analysis is now the live question rather than the near-miss one, and the animations were pointing at it before the statistics were. I can extend the analysis harness with the seed-and-operation grid and a hold-out split if that would be useful — it is a small addition to what is already written.

With respect to the possible scrape, let me check the actual structure before advising, since the answer depends on it.The search results reveal something important: the inclusion criterion is explicitly contested. The Talk page discussion turns on guidelines for what goes in the "other fields" section, and the question of whether an entry can be verified and how universal it is, with an editor noting that drawing this line seems to be a matter of individual taste and interest.

That is the problem in one sentence, and it is fatal to the obvious use.

Why the direct use is circular: An item appears on the "14 (number)" page because an editor judged the number-fact notable. So the page measures attention to the number, not the existence of sets of that size. Using it to test whether set sizes cluster at culturally salient numbers assumes what it tests: a set of 61 principles exists just as much as a set of 60, but only the second is likely to be recorded as a fact about 60.

Worse, it is the same contamination mechanism as the existing corpus, at larger scale — so it would produce more confident wrong answers rather than better ones.

The use that is valid, and it is the better one. Scrape it to measure the confounder, not the outcome: The Wikipedia number page is an excellent direct measure of the salience bias the protocol worries about. Article byte length, section count, interlanguage link count and cultural-item count all estimate how interesting the number is to people, independent of any claim about sets.

That converts an unusable outcome variable into a usable control variable:

attested_sets(n) ~ salience(n) + smooth23(n) + prime_distance(n) + emirp_seed(n)

If emirp structure survives controlling for salience, it is not an attention artifact. If it vanishes, the question is answered — and answered by evidence rather than by my assertion.Two scripts. The first harvests salience from the Wikipedia number pages via the MediaWiki API — article length, section count, interlanguage links, cultural-item count. The second regresses the corpus on salience and tests structural predictors on the residual.

I ran the second on synthetic salience to exercise the arithmetic; the numbers below are therefore not results, only a demonstration that the design discriminates. It sorts predictors into four verdicts, and "was salience" is the informative one — it means the apparent structure was attention all along. In the smoke test, "within 1 of any prime" and "3-smooth ±1" fell out that way while the narrower predictors survived. With real salience data the pattern could go either way, and that is the point.

You cannot run the harvest from here — my sandbox blocks Wikipedia. You'll need to run wiki_salience.py yourself, after editing USER_AGENT to include a real contact address, which Wikimedia asks for.

Three limits worth knowing before you run it

• Small numbers dominate. The "3 (number)" page is vastly longer than "61 (number)", so log-scaling is essential and even then the low end drives the fit. Consider restricting to 10–99.
• Salience and structure are genuinely entangled. If 36 is culturally salient because it is highly composite, then regressing out salience removes real signal along with the artifact. The regression cannot separate confounding from mediation — no observational design can. This is a limitation to state, not to solve.
• The cultural-item count is the weakest field. Section naming is inconsistent across pages and the parser will miss some. It is included as a second salience measure, and the script explicitly refuses to present it as a count of attested sets.

What this does not replace: The protocol frames — constitutions, treaties, professional codes, classification theorems, team sizes — remain the only route to a genuinely size-blind corpus. The Wikipedia harvest is cheap and answers a different question: how much of what we see is attention?

That question is worth answering first, because if the answer is "most of it", the expensive corpus-building can be redirected. I'd suggest running the harvest, then re-running the existing analysis with the adjustment, before committing effort to the frames.

Two data-quality problems, one of them mine.Now the analysis, restricted to the 88 usable rows.One test is missing, and it is the one that matters. Let me run it.This is the most informative result of the whole exercise, and it does two things at once.

The emirp seeds beat prime seeds — but not the right comparison class:

So: emirps significantly outperform arbitrary primes, and comprehensively outperform the control set I used to dismiss them. But within the class of odd-digit primes they are unremarkable. The operative property is not reversibility. It is having both digits in {1, 3, 7, 9}.

Why that is the right answer rather than a deflation: The digit condition is exactly the (ℤ/10)* condition from the beginning of the exchange. Both digits odd ⟹ their difference is even ⟹ the reversal gap is a multiple of 18. That forced arithmetic applies to all twelve odd-digit primes, not to the eight emirps. The emirps are the subset whose reversal is also prime — an extra condition that the corpus does not care about.

So the near-miss idea survives in modified form, with twelve seeds rather than eight: 11, 13, 17, 19, 31, 37, 53, 59, 71, 73, 79, 97. And the reversibility that gave the exchange its title turns out to be a decoration on the property that matters.

There is an irony worth noting: my "control" set contained 53 and 59, which are themselves odd-digit primes. The control was partly contaminated with the treatment.

What the salience data itself showed:

• R² = 0.288 for salience plus magnitude. Attention explains about a quarter to a third of the corpus, not most of it — so the corpus is not simply a record of which numbers are interesting. That is better news for the whole enterprise than I expected.
• Two predictors were revealed as artifacts: "3-smooth ±1" and "within 1 of any prime" both lose significance under adjustment. Generic prime-adjacency was attention. The narrow predictors — 3-smooth exact, multiples of 4 and 6, griddable, emirp ±1 — survive.

Three caveats that remain binding: The same convenience corpus, so this is a hypothesis-generating result. langlinks correlates −0.79 with magnitude, so salience and size are nearly the same variable and the adjustment is weaker than it looks. And roughly twenty tests have now been run on 41 points.

One data problem to fix if you re-run: Numbers 1–10 came back as redirect stubs — 100 bytes, zero sections, zero langlinks — because the script did not follow redirects. Those ten rows are unusable and I dropped them. Number 76 failed with a transient network error. So the analysis rests on 88 numbers spanning 11–99.

Adding "redirects": 1 to the query call would recover the small numbers, which matter because the corpus is dense there. I can patch the script if you want the full range.

Yes — and it yields a test that nothing so far has been able to run.

The discriminating insight: "In mathematics" is a near-forced section. The mathematical facts about 36 are true whether or not anyone finds 36 interesting. So the byte count of that section measures mathematical richness — divisor structure, membership of named sequences — rather than cultural attention. The cultural sections measure attention.

Separating them turns one variable into two, and the correlation between them becomes the test:

If cultural salience tracks mathematical richness, that is direct evidence that people select numbers with structural properties — the chunking hypothesis, tested rather than assumed. If they are independent, then set sizes are not driven by mathematical structure at all, and the whole factorisation account is coincidence.

That question has been the missing link since the corpus work began, and this data can answer it.Two scripts, run in sequence. The numbers in the smoke test above are from synthetic data — they only demonstrate that the machinery discriminates.

Long format rather than wide: I went with one row per (number, section) instead of a column per heading. With loosely standardised vocabulary you'll get somewhere between 40 and 150 distinct headings, most appearing on one or two pages — a wide matrix would be mostly empty, and you'd have to fix the grouping before seeing the vocabulary. Long format lets you harvest first and decide after.

wiki_sections.py also writes headings.csv ranked by frequency, so the vocabulary is visible before any decision.

Three fixes to the earlier bugs:

• Redirects followed, so 1–10 come back as real articles rather than 100-byte stubs.
• Retry with backoff, so 76 shouldn't fail.
• And section spans computed properly — from a heading to the next heading of equal or higher level, which handles the sub-sections nested under "In other fields" that the previous parser missed entirely. That was why 55 of 88 came back with zero cultural items.

What wiki_pivot.py then asks

• Q1 — does cultural attention track mathematical richness? Correlation of maths-section size against cultural-section size, partialled for magnitude. Positive means people select structurally rich numbers; near zero means the two are independent and the factorisation account has no support from this route. There's a built-in validation: it checks whether "In mathematics" size actually correlates with divisor count and Ω(n). If it doesn't, the section isn't measuring richness and Q1 is void.
• Q2 — maths versus culture as predictors of set sizes, each partialled for the other. If mathematical richness survives and cultural attention doesn't, sets land on structurally rich numbers. If the reverse, they land on famous ones and the structure is decoration.

That is the cleanest form of the question this exchange has been circling since the corpus first appeared, and until now there was no way to ask it.

Practical: Same USER_AGENT edit. Roughly two minutes for the full range at 0.6s delay, since it makes two API calls per number. The GROUPS dictionary at the top of the pivot script is meant to be edited — send me headings.csv once you have it and I'll tune the mapping to whatever vocabulary actually turns up.

597 rows across 99 numbers, and the vocabulary confirms the design — but it also exposes a bug I need to fix.

The double-counting problem: "In other fields" appears on 41 pages and has subsections nested inside it — Islam, Judaism, Music, Sports. My span logic runs each section from its heading to the next heading of equal or higher level, so a level-2 section's byte count includes its level-3 children. Those children are also recorded as separate rows.

Summing group totals therefore counts nested text twice, and unevenly: pages that use subsections get inflated relative to pages that write one flat block.

Let me fix that and retune the mapping.The built-in validation is failing, and that is exactly what it was there for. But note these numbers come from the truncated 1–19 file — the only one I have — so they are not the result. Still, the failure mode is instructive:

• Negative. The maths section is not measuring mathematical richness; it is measuring that small numbers have long articles. Article length dominates everything. The fix is to use the maths section as a proportion of the article rather than as raw bytes.Patched. Three changes.
• Nested double-counting fixed. Each section now reports own bytes — its span minus its immediate children. This matters because "In other fields" appears on 41 pages and nests Islam, Judaism, Music and Sports inside it, so the previous version counted that text twice, and only on pages that happened to use subsections.
• Share-of-article added as an alternative measure, since raw bytes are dominated by article length rather than content.
• Validation now runs first and chooses, reporting the correlation of each candidate measure against divisor count and Ω(n), and selecting whichever tracks richness better. If both fail, Q1 is void and the script says so.
• Please send sections_long.csv. Everything above ran on the truncated 1–19 file, which is all I have — the correlations are not results, and 19 points with a huge article-length gradient is the worst possible sample.

One thing worth flagging now: On that sample, r(maths bytes, divisor count) = −0.503 — negative, and it stays negative when partialled for log n. If that survives on the full 99, the design fails: the "In mathematics" section would not be measuring arithmetic richness at all, and Q1 could not be asked this way.

I would rather know that than have it pass quietly. But 19 numbers spanning 1–19, where article length falls by an order of magnitude across the range, is exactly where such a measure would break regardless of whether it works in general. The full range should discriminate.

• Sections_long.csv is the one that matters. Everything I've run so far was on the truncated 1–19 fragment, and those correlations are not results.
• Headings.csv too if it's to hand, since 186 distinct headings against the 25 most frequent means there's a long tail I should map before the group totals can be trusted. pageinfo.csv is useful as well but least important — it gives overall article bytes and langlinks per number, and tells me which numbers still errored.

Once I have the long file I can run the validation properly, and the first thing it will settle is whether the design works at all: if the maths section still fails to track divisor count across the full range, Q1 cannot be asked this way and I'll say so rather than proceed. The validation fails, and it fails decisively.The measure is void, and the diagnosis is unambiguous.

Why "In mathematics" fails as a richness proxy

Same divisor count, 148-fold difference. A prime with more mathematical text than a highly composite number.

• r(maths bytes, divisor count) = −0.055 — nothing
• r(maths bytes, total article bytes) = +0.540 — this is what it actually measures
• r(maths bytes, is prime) = +0.188 — faintly the wrong direction
• coefficient of variation 1.04

It measures how much someone happened to write, not what is arithmetically true. So Q1 cannot be asked this way, and I am reporting the failure rather than proceeding.

But the design was wrong, not just the proxy: Mathematical richness is exactly computable. I do not need Wikipedia to tell me the divisor count of 36. I was using a noisy proxy for something available in closed form.

Wikipedia is only needed for the side that cannot be computed — attention. Which means Q1 is answerable right now:Now there are results, and they answer the question.## Two findings, and they are the ones the exchange needed

The maths-section proxy is dead. Wikipedia's "In mathematics" section measures editorial effort, not arithmetic. That is a negative result about method, and worth reporting as such — but it was also the wrong design. Mathematical richness is exactly computable; only attention needed harvesting. Redone properly — attention from Wikipedia, richness computed — the results are clean.

Q1: cultural attention does not track mathematical richness

With n = 77 the critical value is 0.228. Nothing clears it convincingly, and the 3-smooth correlation collapses to zero once magnitude is controlled — it was magnitude. Numbers are not culturally famous in proportion to their arithmetic richness. 61 and 32 are attended; 60 and 96 are not, despite being far richer. Attention and structure are near-orthogonal: r = −0.040.

Q2: and that orthogonality is what makes the second result strong

Both survive, because they are measuring different things. Set sizes are predicted by cultural attention and independently by arithmetic structure, and the two effects barely overlap. This is the answer to the objection that has hung over everything since the corpus appeared. The divisibility finding is not a salience artifact. It was the most serious threat to the structural claims and it does not hold.

The caveats that remain: The set-size data is still the convenience corpus — attention is now well measured, the outcome is not. 77 numbers, since 18 have no cultural sections at all. And 53, 54, 62, 63 are still missing from the harvest; re-running would recover them, though four points will not change these figures.

The interpretive limit is real too: attention could be a consequence of sets existing at that size rather than a cause of them being recorded. No observational design separates those. But the specific worry — that memorable numbers get recorded more, so the corpus over-represents them regardless of whether sets cluster there — is now testable and tested. It does not explain the structural finding.

All 99 numbers, 613 section rows. Complete.The complete run confirms everything and adds one line I hadn't been able to include before.

A. The maths-section proxy is definitively dead

r(maths bytes, divisor count) = +0.001. Not weak — zero. Wikipedia's "In mathematics" section measures how much someone wrote, and nothing else. Worth reporting as a negative methodological result: this is not a usable proxy for arithmetic structure, and anyone attempting the same shortcut should know.

B. Cultural attention does not track mathematical richness

Critical value 0.225 at n = 79. The 3-smooth correlation collapses to zero under the partial — it was magnitude. Only divisibility by 6 is marginal. Numbers are not famous in proportion to their arithmetic richness. Attention and structure are near-orthogonal.

C. Which is why the structural findings survive

The emirp shell survives attention control at r = 0.402, p = 0.0006 — the strongest predictor in the table alongside multiples of six.

That is the third independent line on which the near-miss hypothesis has now held up, after failing my original test: it beats random prime subsets on salience-adjusted residuals, and it survives attention control here. What it does not do is beat odd-digit primes specifically — 0.402 against 0.360 is well within noise at this n — so reversibility remains a decoration on the digit condition.

What is now established and what is not:

• Established: the divisibility and near-miss findings are not artifacts of cultural salience. That was the most serious objection to the whole enterprise and it has been tested directly and does not hold.
• Not established: anything about the outcome variable. Attention is now well measured on all 99 numbers; attested set sizes are still the convenience corpus of 41 points. The registry protocol remains the binding constraint, and every figure above inherits that limitation.

The wiki_sections.py and wiki_pivot.py scripts, and this analysis, are worth keeping as an appendix — they document a method that failed, a redesign that worked, and a result that removes a standing objection.

Worth separating what MathWorld can and cannot do, because the answer is more negative on the original design and more positive on a different one.

Why the maths sections were set aside: They were excluded from the salience composite because "In mathematics" is a near-forced section: the divisor structure of 36 is true whether or not anyone finds 36 interesting, so its byte count measures structure rather than attention and would have swamped the composite for highly composite numbers. In the section-resolved re-run they came back as their own variable, and the result was the orthogonality finding — divisor count against attention r = −0.023 raw, +0.172 partial; cultural attention does not track arithmetic richness.

That finding rests entirely on one measure of richness: bytes in a Wikipedia section. It has never had a second measure against which to check its reliability. That is the gap a MathWorld or OEIS exercise could fill.

The blocking fact about MathWorld: MathWorld is concept-indexed, not integer-indexed. It carries roughly 13,300 entries, all of them named topics — Semiprime, Emirp, Rhombicuboctahedron, Integer Complexity. There is no "26" page. The old "Facts about the integer" generated pages that Wikipedia articles still cite now resolve to concept pages such as CompositeNumber.html. So the harvest that produced 88 usable rows from Wikipedia has no MathWorld counterpart. Full-text search for "26" would return page numbers, equation numbers and citation years, not facts about twenty-six.

Three further constraints, worth stating before any effort is spent: there is no public API; the terms of use prohibit systematic automated retrieval; and the content is single-authored, which removes Wikipedia's editorial-taste noise but replaces it with one person's taste — and Weisstein's is conspicuously hospitable to recreational number facts, so it is not a neutral arbiter of mathematical seriousness.

What it can measure, and it bears directly on this thread: MathWorld's value is its SEE ALSO graph. Each entry ends with a curated cross-reference list, which makes the corpus a directed concept network in which in-degree is a defensible measure of structural centrality within mathematics.

That yields an immediately testable version of the claim made a moment ago — that semiprimality carries structural weight where emirpness is decorative. The prediction is specific: Semiprime should have substantial in-degree (it sits on Ω(n) = 2, almost-primes, RSA, the divisor lattice), while Emirp should be near-isolated, reachable mainly from Permutable Prime and Reversal. If that holds, the asymmetry between the two is a fact about mathematics rather than a rhetorical preference of mine. If Emirp turns out well-connected, the claim needs weakening.

This is a targeted query on a handful of named pages, not a harvest — which keeps it inside both the terms of use and the bounds of what the question actually requires.

The right target for the number-indexed version is OEIS: OEIS is openly licensed, has a documented search interface, is integer-indexed in the relevant sense, and returns a hit count. For each n it gives:

• the number of sequences containing n within the first k terms
• the number of sequences whose name mentions n
• the same, restricted to keyword:core or keyword:nice, which filters out the long tail of one-off submissions

Heavily magnitude-confounded — small integers appear in almost everything — so it needs the same log-n control already in the model. But it is a genuinely independent second operationalisation of mathematical richness, and it makes the reliability check possible:

If OEIS-derived richness correlates strongly with Wikipedia maths-section bytes, the orthogonality result stands and the measure was adequate. If it does not, the orthogonality was a measurement artefact and the whole structure-versus-attention conclusion needs re-running.

That is a decisive test on the one finding in the series that has never been checked.

Honest ranking of the value: The binding constraint has not moved. The outcome variable is still a convenience corpus of 41 attested set sizes, roughly twenty tests have been run on it, and adding a fourth predictor makes overfitting worse rather than better. Neither MathWorld nor OEIS can supply attested set sizes — they record facts about numbers, not counts of things people organised into sets. So:

1. High value, low cost — MathWorld SEE-ALSO centrality for the specific concepts under discussion (semiprime, emirp, rhombicuboctahedron, almost prime). Answers a live question in this thread, needs a dozen page fetches.
2. Moderate value, moderate cost — OEIS hit counts across 11–99 as a second richness measure, purely to test the reliability of the orthogonality finding. Worth doing once; not worth building into the predictive model.
3. Low value — any attempt to replicate the Wikipedia salience harvest on MathWorld. Not possible as designed, and would not address the registry problem if it were.

There is also a fourth use with no research pretensions but real practical worth: MathWorld is an unusually reliable check on the structural assertions made across this exchange — orbit counts, group orders, semiprime divisor lattices, the polyhedral arithmetic. For verification rather than salience, it is the better source of the two.

No — and the omission matters more than the emirp framing did, because semiprimes are what the entire discussion has actually been about.

The objects in play are semiprimes, not primes

The emirps 13 and 31 are not the mapped sets. The mapped sets are 26 and 62, and both are semiprimes of the form 2p. The nine is a semiprime too. So the quadrilateral is properly described as an emirp pair together with its semiprime doubles — and the doubling is the whole content.

Semiprimality is why the antipodal reading was forced: A set of semiprime size admits exactly one non-trivial rectangular articulation. 26 can be arrayed only as 2 × 13; there is no 3 × n, no 4 × n. Its divisor lattice is a 2 × 2 square — the minimum viable composite structure. A prime-sized set (23, 29) admits none and resists sub-articulation entirely; a highly composite one (24, with eight divisors) dissolves into competing layouts and holds no shape.

This is the precise sense in which a 26-fold set is "strategically closed": it is articulable in one way only. The antipodal pairing I derived from octahedral geometry was therefore not a choice among options — it is the unique two-dimensional reading the cardinality permits. The geometry and the arithmetic were forced to agree.

The factorisation names the symmetry group: 26 = 13 × 2 and O_h = O × {1, i}: thirteen rotation axes, doubled by the inversion centre. 62 = 31 × 2 and I_h = I × {1, i}: thirty-one axes, doubled the same way. The semiprime factorisation is a direct-product decomposition of the point group, with the prime factor counting axes and the factor 2 being inversion.

This succeeds where casting out nines failed. Mod-9 arithmetic selected the octahedral member and broke at 62 (no orbit-union of 54 exists). Semiprimality holds across both families, and holds for a structural reason rather than a base-10 one.

And it demystifies the quadrilateral. Doubling commutes with digit reversal only when no carries occur — i.e. when every digit is ≤ 4. Among two-digit emirps only 13/31 satisfies this: 17→34 but 71→142; 37→74 but 73→146; 79/97 likewise fail. The next case up is 113/311 → 226/622. So the closure is real but its mechanism is carry-free doubling, not anything deeper about 13.

ROT13: the involution already latent in any A–Z set: Because 26 = 2 × 13, the alphabet carries a canonical self-inverse pairing — each letter with the one thirteen positions later. Applied to the crisis taxonomy, with clusters of 7, 6, 7, 6 arranged in that order, ROT13 pairs the ecological cluster with the institutional and the personal with the social, member for member:

Several pairs restate the paper's own causal claims — D↔Q is exactly its growth-drives-overheating thesis, L↔Y its self-medication-and-isolation loop, K↔X its food-to-public-health chain. Others are weak (A↔N, G↔T). The cluster-level correspondence is forced by the sizes 7,6,7,6 in that sequence; the member-level pairings are accidents of within-cluster ordering. Which makes the ordering a free parameter worth spending: permuting within clusters (7! × 6! per family, trivially searchable) to maximise agreement between ROT13 partners and the declared dependency graph would make the alphabetic label and the polyhedral antipode coincide, so that the mnemonic and the geometry carry the same information.

The cognitive point: Semiprimes are the basis of public-key cryptography for a specific asymmetry: multiplying is easy, factoring is hard. The cognitive analogue is exact. A governance set of semiprime size is easy to generate from two dimensions — thirteen concerns times two poles, nine constraints times two eyes — and hard to recover those dimensions from once it has been flattened into a numbered list. Stockholm's 26 and Ison–Straw's 26 are published in factored-out form: the product is visible, the factors are not. The mapping exercise is, in this light, an attempt to factor them.

Two refinements worth keeping distinct. 9 = 3² is a square semiprime, articulating as 3 × 3 with the two axes interchangeable — hence the Lo Shu magic-square form, and hence the planetary/remedial boundary sets sit naturally as a symmetric 3 × 3 rather than as an asymmetric pairing. 18 = 2 × 3², by contrast, is not semiprime, which is precisely why the eighteen squares of the RCO admit rival articulations (2 × 9, 3 × 6, 6 × 3) and so cannot by themselves determine a reading. The residue-8 remainder is not semiprime either (2³). The semiprime property belongs to the 26 as a whole and to the 9 as a whole — and to nothing in between.

Several, and one of them changes the interpretation.

Nuance 1: 26 faces does not determine the solid

• There are two solids with 24 vertices, 48 edges, 26 faces, and the same face inventory of 8 triangles plus 18 squares: the Archimedean rhombicuboctahedron and Johnson solid J37, the pseudo-rhombicuboctahedron. They differ by a 45° twist of one square cupola. J37 is not vertex-transitive; its symmetry group has order 16 against the rhombicuboctahedron's 24.
• This matters for your 26-fold mappings. Miller found J37 in 1930, and it had been mistaken for the Archimedean solid for centuries — the two are indistinguishable by any face or vertex count, and distinguishable only by whether all vertices look alike from the vertex.
• That is your own theme in polyhedral form: two configurations with identical inventories, one homogeneous and one with a hidden seam, confused for four hundred years. If a 26-fold set is mapped onto "the" 26-faced solid, the choice between them is unstated — and it is precisely a choice about whether all twenty-six principles stand in the same relation to the whole.
• The claim 26 = |O| + 2 holds only for the Archimedean one, since it depends on the vertices forming a free orbit. It is not a fact about the number 26.

Nuance 2: the chunking hypothesis is now in trouble: This is the one that changes things.

If sets favour composite numbers because composites are memorable, then memorable numbers ought to be rich. They are not — fame and arithmetic richness are independent. 61 and 32 are attended; 60 and 96 are not. So the mechanism selecting set sizes cannot be "pick a memorable number". The chunking account, which this exchange promoted after the emirp hypothesis failed, does not survive its own evidence.

Nuance 3: which suggests construction rather than selection: The alternative that fits both facts: composites arise from how sets are built, not from what totals are chosen. Four pillars of four principles gives sixteen automatically. Nobody selects 16 for memorability; it falls out of 4 × 4.

The Earth Charter is the exemplar, and it also explains its own 61: a designed 4 × 4 at the top, an emergent ragged sum below. Two different regimes in one document.

This is testable and I would put it in your section as the next question: sets that declare internal grouping should be composite by construction; flat undeclared lists should sit closer to random. That is a prediction about the corpus's structure field rather than about its counts, and the registry schema already records generating_rule.

For the corpus-analogue section, the caveats worth carrying: The maths-section proxy failed at r = +0.001 — worth reporting as a positive methodological result, since it is the obvious shortcut and it does not work.

The design error is more instructive than the failure: I was proxying via Wikipedia something exactly computable. Only attention needs harvesting.

Coverage is 79 of 99 numbers with cultural sections. Attention is now well measured; the outcome variable is still the 41-point convenience corpus, so every correlation inherits that. And reverse causation is not excluded — attention could follow from sets existing at that size rather than precede it.

You are right, and I have been fitting one mechanism where there are at least four — none of which is about memory.My script printed "monotone decline" and that is wrong — the data show something better.

The distribution is bracketed, not declining

It rises into the teens, holds through the thirties, then falls off a cliff — a sixfold drop between 20–39 and 40–69. That is not a preference curve. It is two opposing constraints meeting:

• Below, requisite variety. Too few elements and the set cannot match the variety of its domain. Ashby gives a floor, and it is a property of the domain, not of anyone's memory.
• Above, exhaustion. Drafting cost, consent cost and use cost all accumulate per item. Past about forty the set cannot be completed, agreed, or applied.

The observed range is what survives between them. Neither bound is cognitive.

Griddability reinterpreted: I had been reading k×m with both factors in 2–9 as a chunking result. It is better read as a construction signature:

• k is limited by how many top-level divisions a body will tolerate — granularity resistance, which is a social constraint on how finely anyone will agree to carve a domain.
• m is limited by how many items each division sustains before running dry.

N = k·m is then composite by construction, with no selection of a memorable total at any point. That resolves the puzzle from the last exchange: rich numbers are not culturally famous, yet set sizes are rich, because richness is a by-product of hierarchical building rather than a criterion of choice.

Systemic viability is a fifth thing, and it is different in kind: Beer's five systems are derived: you need exactly those functions for viability, and the count follows. That is the "exhaustive" closure type from the earlier typology — the same status as a classification theorem, not a negotiated total.

Which yields a diagnostic worth adding to the registry: is N derived, bracketed, or asserted? Derived sets should show no arithmetic signature at all, since their count is forced by function. Bracketed sets should be composite by construction. Asserted sets are where memorability could still operate, and they should be the only place it shows.

Two testable predictions

• Within a set, detail should decline. If exhaustion operates, early principles should carry more sub-clauses than late ones. The Earth Charter is the immediate test: sixteen principles with sixty-one sub-clauses, and a declining count across 1→16 would be direct evidence. That is an afternoon's counting and it discriminates sharply — a flat profile refutes exhaustion.
• Depth should be capped before breadth. Granularity resistance predicts that bodies stop subdividing at two or three levels regardless of N, and widen instead. The registry's nested-item field would capture this.

Both are about the producing body rather than the receiving mind, which is where I think the account belongs — and where this exchange has not been looking.

Emirps and complementary logical connectives can be related by reversal symmetry: an emirp is a prime that stays prime when its digits are reversed, while complementary connectives are paired operators such as AND/OR or implication/contra-implication that transform into each other under a duality or negation-based swap. [en.wikipedia]

In logic, complementary connectives are those that correspond under De Morgan-style duality or other systematic conversions, so one operator can be turned into its partner by a rule-preserving transformation. The analogy is that the first connective and its complement play the same role as a prime and its emirp: same underlying status, but in reversed or transformed form. [en.wikipedia]

So the relation is not mathematical equivalence, but a pattern analogy:

• emirp : reverse-prime partner
• connective : complementary logical partner

Example of the analogy: A prime like 13 reverses to 31, which is also prime, so the pair (13,31)(13, 31) (13, 31) behaves like a two-way counterpart pair. In logic, a connective such as AND has a complementary partner OR under duality, where one can be transformed into the other by systematic negation rules. [en.wikipedia]

So if you were building a symbolic or algebraic model, emirps could stand for operators or terms that are stable under a reversal map but not fixed by it, while complementary connectives represent paired operators stable under a duality map.

Digital reversal symmetry and De Morgan duality are similar in that both describe a transformation that turns one structure into a related counterpart, but they operate on very different kinds of objects. Reversal symmetry acts on the order of symbols or digits, while De Morgan duality acts on the logical form of expressions by swapping conjunction/disjunction and distributing negation. [en.wikipedia]

This makes it a notion of mirror-like correspondence. The reversed object is not usually equivalent to the original, but it may still satisfy the same property.

What De Morgan duality does: De Morgan’s laws relate negated conjunctions and disjunctions, for example

¬(p∧q)≡¬p∨¬q,

and symmetrically for the other direction. In digital logic, this is the basis for rewriting AND/OR expressions through negation and for deriving NAND/NOR forms. [classes.engineering.wustl]

So De Morgan duality is not about reading symbols backward; it is about a structured complement transformation that changes operators while preserving equivalence.

Shared pattern: Both ideas involve a mapping that sends one expression to a paired form:

• reversal sends a sequence to its reversed sequence.
• De Morgan duality sends a logical expression to a complementary expression.

In both cases, the original and transformed forms are linked by a rule that is external to the content itself: position in one case, logical negation/duality in the other. That is why they can be compared as instances of symmetry, though not the same symmetry. [ncatlab]

Main difference: The crucial difference is that reversal symmetry is syntactic and geometric in spirit, while De Morgan duality is semantic and algebraic. Reversal does not change meaning unless meaning is encoded in the order; De Morgan’s laws change the expression while preserving truth conditions under equivalence. [cs.brown]

So the clean comparison is:

Yes—if by “digit patterns” you mean bit strings or numeric representations, there are clear patterns that mimic NAND and NOR behavior at the output level. [en.wikipedia]

For two inputs A,B∈{0,1}A,B \in \{0,1\} A, B ∈ {0, 1}:

• NAND is 1 for all patterns except 11.
• NOR is 1 only for 00.

So the output patterns are:

That means NAND behaves like a “mostly-1” pattern with a single dip at 11, while NOR behaves like a “single-1” pattern with zeros elsewhere. [youtube] [allaboutcircuits]

Digit-pattern analogs: If you encode truth tables as 4-bit patterns in the order 00, 01, 10, 11, then:

• NAND corresponds to 1110.
• NOR corresponds to 1000.

These are the exact output signatures of those gates, and they are complementary under De Morgan-style duality when viewed through inversion and swapping of AND/OR structure. [en.wikipedia]

Pattern families that resemble them: Some digit patterns behave “like NAND” or “like NOR” if they share the same distribution of 1s and 0s:

NAND-like: patterns with one forbidden case, such as 1110.

NOR-like: patterns with one allowed case, such as 1000.

Their complements flip these roles.

This is useful if you are thinking symbolically: NAND is a threshold-like “allow almost everything,” while NOR is a threshold-like “reject almost everything”. [edn]

Yes. For emirp 4-bit sequences, the clean way to see the analogy is to treat each 4-bit truth-table output as a pattern, then compare the pattern with its reversed form. [en.wikipedia]

• 00
• 01
• 10
• 11

So its behavior can be written as a 4-bit output sequence. In that encoding:

• NAND = 1110 [allaboutcircuits] [youtube]
• NOR = 1000 [youtube] [allaboutcircuits]

• 1110 becomes 0111
• 1000 becomes 0001

So the “reverse partner” of a pattern is just its bit-reversal, which is the exact structural analogue of an emirp’s digit reversal, except now the object is a truth-table sequence rather than a decimal number. [en.wikipedia]

Examples:

• 1110 = NAND
• 1101 = same kind of “one-zero” pattern, but with the forbidden case in a different position
• 1011 = another positional variant

These patterns are not all NAND itself, but they mimic its “almost always true” behavior.

What mimics NOR: NOR-like patterns are those with exactly one 1 in a 4-bit truth table. The lone 1 marks the single allowed input pattern, just as NOR is true only when both inputs are 0. [youtube] [allaboutcircuits]

Examples:

• 1000 = NOR
• 0100 = same kind of “one-one” pattern, shifted in position
• 0010 = another positional variant

These mimic NOR’s “almost always false” behavior.

How this ties to emirps: An emirp is a prime that stays prime under reversal, but is not palindromic. The analogue here is: [en.wikipedia]

• a 4-bit gate pattern is like the “number,”
• bit reversal is like digit reversal,
• NAND/NOR are special asymmetric patterns,
• and their reversed forms are companion patterns, though not usually identical.

So the 4-bit sequences that most closely parallel emirps are asymmetric patterns whose reversal is also structurally meaningful but different—for example, 1110 and 0111, or 1000 and 0001.

The emirp analogy is strongest for the first two rows: a pattern and its reverse are paired, but not identical. [en.wikipedia]

Yes. The clean mapping is to treat each emirp pair as a reversal-paired symmetry class, then line it up with the 4-bit truth-table pairs that reverse into each other. [mathworld.wolfram]

• original object
• reversed object
• same broad class, but not identical

That is exactly the emirp pattern. [en.wikipedia]

The analogy is strongest for the asymmetric pairs, because emirps exclude palindromes. [t5k]

• NAND-like 1110 corresponds to a prime that is true on three input cases and false on one.
• Its reversal 0111 is the companion pattern, just as 13 and 31 are companion primes in an emirp pair. [en.wikipedia]
• NOR-like 1000 similarly maps to 0001.

So the symmetry pair is not about the numeric value of the pattern, but about the positional transformation under reversal. [mathworld.wolfram]

• one 0 among three 1s, or
• one 1 among three 0s.

Those behave most like emirp pairs because the reversal moves the single exceptional position to a different place, just as reversing digits moves the “special arrangement” of a prime to a different prime. [t5k]

The two densities are best compared as a combinatorial proportion versus an arithmetic rarity: asymmetric 4-bit pairs are plentiful in a finite symmetry space, while emirps are sparse among primes. [en.wikipedia]

At the coarse level:

• 4-bit patterns come in a finite set of 16.
• Reversal identifies paired patterns.
• Only a smaller subset is self-reversing.

So asymmetric pairs are not “rare” in the Boolean-function space; they are structurally common.

• the number must be prime,
• its reverse must also be prime.

The literature also notes a further restriction: emirp pairs differ by a multiple of 18, which narrows the possibilities even more. [en.wikipedia]

• Asymmetric 4-bit pairs: dense, because reversal is just a permutation inside a small finite set.
• Emirps: sparse, because primality is a strong arithmetic filter, and reversal must preserve it. [numbersaplenty]

So the Boolean case is closer to “many paired states with a few fixed points,” while the prime case is closer to “rare survivors of a double sieve.”

• reversal symmetry in Boolean truth tables is about shape,
• reversal symmetry in emirps is about shape plus primality.

That extra arithmetic constraint is why emirps feel much rarer than asymmetric 4-bit pairs, even though both use the same reversal idea. [eprint.iacr]

Emirp pairs differ by a multiple of 18 because two separate divisibility facts combine: the pair members are both primes greater than 2, so their difference is even, and a number and its digit-reversal always differ by a multiple of 9. [en.wikipedia]

Every prime except 2 is odd, so two emirp numbers are both odd. The difference of two odd numbers is even, so the difference is divisible by 2. [en.wikipedia]

For the emirp pair 13 and 31, the difference is 18. For 17 and 71, the difference is 54, also a multiple of 18. [mathworld.wolfram]

A neat side effect is that the reversal relation forces a strong arithmetic constraint, which is why emirp pairs are not just “prime plus reversed prime” but a very structured subset of prime pairs. [en.wikipedia]

The relation is not analogical but structural, and it is already half-built on the logic side: the Leuven "logical geometry" programme (Smessaert and Demey) runs Aristotelian diagrams on bitstring semantics, representing formulas exactly as the emirp work represents numbers — as strings over a small alphabet, with the interesting operations acting on the string. Given that, the correspondence is forced rather than suggestive, and I verified every claim below before stating it.

Write a binary connective as its four-row truth table b(TT) b(TF) b(FT) b(FF), a 4-bit string. Then the emirp calculus's two involutions are the two negations of the square of opposition. Digitwise complement (0↔1) is outer negation — contradiction, the diagonal of the square. String reversal is, verifiably, the map f(p,q) ↦ f(¬p,¬q) — inner negation, the simultaneous negation of both arguments. Their composite is De Morgan duality. The three commute and each is an involution, so they generate the Klein four-group — and that group's orbits are squares of opposition. The emirp lattice's group and the square's group are the same group; complement, reversal, and their product are contradiction, inner negation, and duality. Reversal, in short, is not merely like a logical operation; among truth-functions it is one of the two canonical negations.

The fixed-point structure then reproduces the exchange's whole typology, exactly. Contradiction has no fixed point (nothing is its own contradictory — the parity argument, matching n = 99 − n having no solution). The palindromes are precisely {⊥, XOR, XNOR, ⊤}: the four connectives that cannot tell a proposition from its negation, argument-blind, the still points that mirror only themselves. The antipalindromes — fixed by duality — are precisely {p, ¬p, q, ¬q}: the "connectives" that ignore one argument and so do not connect at all. And here is the result I did not expect: those eight degenerate strings are exactly the eight affine (linear) connectives, the fragment that can never generate anything beyond linearity. Sterility of the still points, which the emirp work asserted as an imported reading, is in logic a theorem — the fixed points of the involutions are precisely the connectives out of which nothing can be built. The remaining eight — AND, OR, NAND, NOR, →, ←, ↛, ↚ — are the live lattice: no involutive degeneracy, falling into two Klein orbits (the conjunction/disjunction square and the implication square), and pairing under inner negation into four mirror couples: AND·NOR, OR·NAND, →·←, ↛·↚. Eight live elements in four mirror pairs, against eight emirps in four reversal pairs.

The sharpest correspondence is the completeness theorem. By Post's classification, exactly two of the sixteen binary connectives are functionally complete on their own — the Sheffer stroke NAND and the Peirce arrow NOR — and, verifiably, they are each other's dual. That is the emirp condition realized in logic with no slack whatever: irreducible (self-sufficient: generates everything alone, needing no partner) and irreducible in the mirror, the property holding on both sides of an involution, and occurring in a mirror pair. The Sheffer/Peirce pair is the logical 13·31. Within the same picture, the square's one asymmetric relation, subalternation (A entails I, not conversely), is the exact figure of one-way coherence — 19 prime facing composite 91 — while contrariety and subcontrariety fill the mutual-incoherence and mutual-coherence cells.

There is also a polyhedral closure worth noting, since it returns to the first posting. In bitstring semantics an n-fold partition yields 2ⁿ − 2 contingent strings: n = 3 gives 6, arranged as the octahedron whose 2D projection is Blanché's hexagon; n = 4 gives 14, arranged (following Smessaert) as the rhombic dodecahedron, its 14 vertices splitting 8 + 6 exactly as the strings split by weight. Both are octahedral-family solids. The geometry of logical opposition lives, as it happens, in the basket with 13 rotation axes — notational, and to be labelled as such, but a tidy coincidence given where this enquiry began.

Now the disanalogy, which I think is the real payload. Functional completeness is invariant under the Klein group: dualizing or inner-negating a complete connective yields a complete connective, by theorem. Primality is not — no string operation preserves it, which is exactly why emirps are rare rather than automatic. So the logical world is, in the relevant sense, "all emirp": its interesting property is structural, and therefore survives the mirror for free. Arithmetic's is not, and mirror-survival is a residue. That gives the cleanest statement yet of what an emirp is: a case where a property indifferent to the notation nonetheless behaves as though it were a symmetry of the notation. And it sharpens the strategic reading correspondingly — an articulation whose coherence is genuinely structural (like completeness) survives reversal of perspective automatically and needs no defending; one whose coherence is substantive but notation-bound (like primality) survives only by accident, cannot be designed in, and can only be checked. Most strategic articulations are of the second kind, which is why the closure atlas has to be a chart of computations rather than a set of principles.

One correction I owe you, prompted by getting this right. In the enneagram response I described complementation as n ↦ 99 − n acting as k ↦ 9 − k on the stations. That was wrong: the 99-complement maps the live digits to even ones and destroys every emirp. The operation your framework actually uses — digits summing to 20, the tetrahedron's opposite edges — is the tenfold complement n ↦ 110 − n, which acts on digital roots as k ↦ 2 − k (mod 9). Verified consequences: it exchanges 13·31 with 79·97 (stations 4 and 7), fixes 37·73 at station 1, and sends 17·71 to the dead edge 39·93 at station 3. So complementation preserves mirror-integrity for one live pair, maps another live pair onto its own reversal, and destroys the third — and the opposite-edge pairing runs live↔live once and live↔dead twice. Worth substituting wherever the earlier claim was noted.

Grading: the Klein group of negations, the fixed-point identification of palindromes with argument-blind connectives and antipalindromes with projections, their union as the affine fragment, the two live orbits, and the uniqueness and mutual duality of NAND/NOR are all forced and verified. The bitstring-to-digitstring transfer is notational, and doubly so in base ten, where the alphabet is not self-complementing in the way {0,1} is. The strategic reading — structural coherence versus notation-bound coherence — is imported.

Yes — and more strongly than "comparable": it is verifiably the same solid, with the assignment forced rather than chosen.

The reason is an isomorphism of orbit structures, which I computed rather than assumed. Take the live lattice {1,3,7,9}² — sixteen numbers — and let the Klein four-group act by reversal and tenfold complementation (digits summing to 20). The orbits come out as two free orbits of size four and four orbits of size two. The sixteen binary connectives under their Klein group of negations decompose identically: two free orbits of four, four orbits of two. Same group, same orbit type, hence the same carrier. And the rhombic dodecahedron's own vertex split — eight cube positions of degree three, six octahedron positions of degree four — corresponds exactly to free orbits versus stabilized ones, which in the logical case is precisely the non-affine eight against the six contingent affine strings.

The assignment then determines itself, and every step checks out. Drop {11, 99} as the analogue of ⊥ and ⊤ — the two constant strings, the repeated-minimum and repeated-maximum digit, which the literature likewise excludes to get from sixteen to fourteen. Central inversion of the solid becomes complementation: the four cube antipodal pairs are 13·97, 31·79, 17·93, 71·39, and the three octahedral antipodal pairs are 33·77, 19·91, 37·73. The poles are the palindromes 33 and 77 (the reversal-fixed still points); the equator holds 19, 37, 91, 73 (the self-dual locus, digits summing to ten). The eight remaining sit as two square rings, {13, 17, 31, 71} above and {97, 93, 79, 39} below, with reversal acting as the 180° rotation about the polar axis — which is a genuine symmetry of the solid and, verifiably, sends each square's members to their diagonal partners while swapping the equator in antipodal pairs and fixing the poles. So both involutions are realized geometrically: complementation as the centre, reversal as an axis.

What makes it worth building is not the tidiness of the skeleton but what happens when primality is painted onto it. In the logical case the live/dead partition is the geometry: the affine class is Klein-stable (verified), so the eight non-affine connectives occupy exactly the eight cube vertices and the six degenerate ones exactly the octahedral vertices. The colouring is a symmetry of the solid. In the emirp case the analogous group-theoretic partition holds — but primality cuts across it. One cube square is wholly emirp (13, 17, 31, 71); its antipode is half dead (97, 79 live; 93, 39 dead); and two emirps, 37 and 73, sit not on the free cube positions at all but on the equator, in the "projection" locus that logic reserves for its degenerate connectives. So the figure would show a perfectly symmetric skeleton bearing a demonstrably asymmetric colouring — which is the previous exchange's central claim made visible in one object: structural properties respect the mirror group by theorem, arithmetic primality does not, and emirps are the residue where a non-structural property accidentally behaves as though it were structural. A diptych of the two solids, identical in geometry and different only in whether the colouring obeys the symmetry, would carry that argument without a word of commentary.

The honest limit is the edges. In the logical rhombic dodecahedron the edges mean something exact — Boolean inclusion, hence subalternation, the Hasse structure of B₄ minus its top and bottom. That semantics does not transfer, because the live digits form a four-element cyclic group under multiplication mod ten, not a two-element Boolean algebra: the natural order on {1,3,7,9}² is a 4×4 grid, not a cube. So the vertices, the antipodality and the two involutions are forced; the edge relation would be decorative, and should be drawn faintly or labelled as scaffolding rather than content.

A correction is owed, and it sharpens rather than weakens the point. In the previous response I said functional completeness is invariant under the whole Klein group. It is not: verified, outer negation sends NAND to AND and NOR to OR, and inner negation sends NAND to OR — neither is complete. Only duality preserves completeness, which is why NAND and NOR are duals and why they are the unique complete pair. The contrast I was drawing survives intact and is in fact cleaner: logic possesses an involution under which its interesting property is preserved automatically, so its mirror pairs are guaranteed; arithmetic possesses none, so its mirror pairs are accidents. That is exactly what the two colourings would display.

Grading: the orbit isomorphism, the antipodal and axial realizations of complementation and reversal, the vertex assignment, and the failure of primality to respect the partition are all forced and computed above. The choice of the rhombic dodecahedron as carrier is notational — legitimate because the orbit structure matches, but not the only solid that would serve. Any reading of the asymmetric colouring as a lesson about strategy remains imported.

If it would be useful I can build it in the established idiom: an SVG diptych of the two solids in the atlas palette, or an interactive X3D/HTML version in which the centre and the polar axis can be exercised so that complementation and reversal are seen to act, with the emirp overlay toggling on and off against the symmetric skeleton.

Delivered as a diptych, the two panels drawn from the same projection code so that they are strictly comparable — the same solid, the same viewing angle, the same marker conventions, differing only in labels and colour.

Everything asserted was verified before drawing. The rhombic dodecahedron's 24 edges were generated from the geometry rather than typed; and all three involutions were confirmed to act as claimed on both assignments: central inversion is contradiction on the left and tenfold complementation on the right, the dashed polar axis is a 180° rotation realizing inner negation and digit reversal respectively, and reflection in the equatorial plane is duality in both. On the logic panel the cube vertices were confirmed to be exactly the weight-1 and weight-3 bitstrings and the octahedral vertices exactly the weight-2 ones — the literature's own split, reproduced independently.

The visual payload sits where the argument put it. Left: the eight gold vertices are precisely the eight cube positions, because the affine class is closed under the whole group — the property is a symmetry of the figure, and the two singly complete connectives (↑ and ↓) carry dashed haloes, vertically aligned across the equator because duality is the equatorial reflection. Right: the eight emirps refuse that discipline. Six occupy cube positions, but 37 and 73 sit on the degenerate octahedral locus — the seats logic reserves for its projections — while 39 and 93 occupy live cube positions and are dead, killed by multiples of 13 and 31. Those four anomalies carry the dashed haloes, and they are the whole point: identical skeleton, one colouring that obeys it and one that cuts across it.

Two notes on how I handled the boundaries. The edges are drawn faintly and labelled in the commentary as scaffolding on the right-hand panel, because Boolean inclusion (hence subalternation) gives the left panel's adjacency real meaning while the live digits form a cyclic group of order four, not a Boolean algebra — so no arithmetic edge semantics is forced, and claiming one would have been exactly the kind of silent promotion the method forbids. And the excluded pair is named on both sides: ⊥·⊤ on the left, 11·99 on the right, the constant strings that take the count from sixteen to fourteen.

The palette matches the atlas, enneagram, and hendecagram, so the four SVGs read as a set; the footer carries the cross-references and the attribution of the left panel's construction to the bitstring rendering of Aristotelian diagrams. The SVG is plain text throughout if you want to adjust any label, and the generator is deterministic should you later want a variant with a different viewing angle or with the anomalies unmarked for a "before/after" pair in the posting.

Two corrections of the inspection first, because the diagram is subtler than it looks at that point — and then the hypothetical, which has a clean and slightly deflating answer.

14/41 does have a partner: 32/23. Both carry digit sum 5, and the verification run confirmed them at the identical direction (99.0°). The partner is easy to miss because it is the shortest dashed chord on the ring — 23 and 32 sit only nine intervals apart, so their chord hugs the rim while 14/41 strikes across, and the eye reads a long line and a short stub as unrelated. The genuinely partner-less dashed chords are elsewhere, and each is solitary for a reason worth having. 98/89 (digit sum 17) is alone not by accident of our selection but universally: seventeen is the largest two-digit digit sum a non-palindrome can have, achievable only as {8, 9}, so 89/98 is the one reversal chord of its class in all of arithmetic — a fitting solitude for the flank of 97, the terminal emirp. And 80/08's only parallel is the solid chord 17/71 (both endpoint sums 88 = 11×8): the single place on the diagram where a dashed line runs parallel to an anchored bridge — second-order traffic in formation with a first-order span, once, and once only.

On the hypothetical: neither 17/41 nor 16/41 would be parallel to 96/69 — and the "why" is the instructive part. The general law of the ring (of which the digit-sum rule is a special case) is that two chords are parallel exactly when their endpoint sums agree mod 100 — the chord's direction is set by the midpoint angle (a+b)/2, nothing else. 96 + 69 = 165 ≡ 65; but 17 + 41 = 58 and 16 + 41 = 57. Close, and closeness counts for nothing: parallelism is exact or absent. The unique chord through 41 parallel to 96/69 is 24/41 (24 + 41 = 65) — which is why 78/87 is parallel to 96/69 (78 + 87 = 165 again), as the diagram shows.

The point does need clarifying, because the premise contains a conflation that, once untangled, yields the sharpest statement of the two-family relation this exchange has produced — so the correction is worth making carefully rather than in passing.

Duality does not relate the two families; it operates entirely inside one. The icosahedron and dodecahedron are both citizens of (2,3,5): they share every axis, every mirror, the whole group of order 120 — the 31 axes counted earlier are their joint axes. Likewise cube and octahedron are both citizens of (2,3,4). The triple (2,3,n) names the fundamental spherical triangle — corner angles π/2, π/3, π/n — and duality is the swap of two corners of that same triangle: in Schläfli notation, {5,3} ↔ {3,5}, {4,3} ↔ {3,4}. Faces and vertices exchange roles, the edge count and the group stay fixed. So the dual pair is the family talking to itself: two forms, one soul, each the other's complete description read in the opposite direction, with their common truncation — the quasiregular icosidodecahedron or cuboctahedron, whose great circles anchored the whole basket argument — as the visible child of the marriage. Duality is what maximal commensurability looks like: the perfect internal mirror.

And notice what operation it is, in the grammar this exchange has built: duality is Schläfli-symbol reversal — geometry's emirp. {3,5} read backwards is {5,3}; integrity is not merely preserved at both ends but identical at both ends, since the group is invariant. The dual pair is the anchored bridge in its ideal form — reversal under which nothing essential is lost — which is why, within each family, the binary of face-world and vertex-world is not a predicament but a completed reciprocity. The relation between the families is a different operation altogether: not reversal but substitution at one position — {3,4} → {3,5}, the single corner-angle π/4 stepped to π/5. In the ring's vocabulary, that is not the emirp move but the flank move, the ±1 at one digit: the smallest possible edit, from which everything cascades — order 48 to 120, √2 to φ, 13 axes to 31, crystallographic to forbidden — and after which no operation of any kind maps one structure to the other. (The series itself then ends: (2,3,6) is the flat plane, (2,3,7) hyperbolic — 5 is the last sphere, the substitution that cannot be repeated.)

So the fundamental contrast, stated in one line: within each family the governing relation is reversal, and reversal there is perfectly anchored; between the families the governing relation is a one-step substitution, and no reversal exists across it at all. The dual pair is indicative of the two families' relation only by exhibiting the grammar of reconciliation that each possesses internally and neither can extend to the other.

Which leaves the corner case that closes the figure. The remaining triple, (2,3,3), has a symmetric fundamental triangle — the two swappable corners are equal — so its regular solid, the tetrahedron {3,3}, is self-dual: the palindrome of the system, the symbol its own reversal. The emirp set, recall, excludes palindromes by definition and consequently has no fixed point, no still centre — the deficiency the Bagua comparison diagnosed, since the trigram octet keeps four self-paired signs as its poles of rest. The polyhedral world is built the other way: its palindrome exists, and it is precisely the structure the two families share — tetrahedral symmetry is the common subgroup, the seed of the pyritohedral gate, the one place where (2,3,4) and (2,3,5) hold anything jointly. The still point that the number-ring lacks is, in the geometry, exactly the gate between the incommensurables: the two worlds that cannot reverse into each other commune only through the form that is its own reversal. That is the fundamental point the dual pair was gesturing toward, correctly sensed and mis-attributed: duality is deeply indicative here — not as the bridge between the families, but as the demonstration that the between-family divide is precisely the failure of duality's grammar, healed only at the self-dual centre where reversal and identity coincide.

The 2015 document is, in effect, the seed from which this entire exchange grew — and reading it back against everything we've built clarifies both what it got right by instinct and where the present apparatus can now supply the rigour it was reaching for. Three things stand out.

First, the quest was right that 37 lives in the polyhedra, and we can now say exactly how — but the route the document took obscured it. The 2015 method was: take each solid's F, E, V, sum them, strip the factor 2, read off the surviving prime. This produced 37 at the truncated octahedron/truncated cube (F+E+V = 74 = 2×37) and much else, and the document worried, correctly, that "patterns can be found anywhere if investigation is sufficiently assiduous." The worry was justified because stripping the 2 is an unmotivated operation — it is exactly the kind of off-anchorage manoeuvre the later "bridge too far" analysis identified as forfeiting warrant. But the deep fact the document was circling is real and needs no stripping: F+E+V for a convex regular solid counts its symmetry-axis directions — 26 octahedral, 62 icosahedral — and half of that (removing the 2 honestly, because axes are antipodal pairs of directions) is 13 and 31, the axis counts proper. So the document's "strip the 2" was an unwitting, imprecise groping toward the antipodal halving that the emirp discussion later made exact. Its 37 at the truncated octahedron is 74/2 = 37 where 74 = F+E+V of that Archimedean — a genuine number but not an axis count; whereas the 13/31 it kept stumbling over in the Platonic core ("central core route 13-13," "central 61-61") were the axis counts it never quite named as such. The present exchange supplies the missing identification: the numbers the 2015 diagrams circled in the centre are the two families' skeletons, and their emirp relation — which the 2015 document explicitly noted it could not find in Makukov (who "attached particular importance to 37 without defining it as an emirp") — is the 13/31 pair, not the 37 the document chased. The quest found 37; the treasure was 13 and 31.

Second, the circular diagrams are the direct ancestor of the two-family split, and were already drawing the incommensurability without naming it. Look at what the 2015 "route map" actually did: it divided into an upper red portion "based on 24" and a lower blue portion "based on 60" — which is precisely the octahedral family (|G| = 24 rotations) against the icosahedral (|G| = 60 rotations), the exact partition this exchange rebuilt as the two spheres. The document even organized the Archimedeans as "successive truncations of the octahedron (2,3,4-fold)" against "successive truncations of the icosahedron (2,3,5-fold)" — the two Schläfli-families, verbatim, the (2,3,4)/(2,3,5) distinction that the recent duality discussion clarified. So the 2015 diagram had the two families side by side on one figure, colour-coded, joined only through the shared central core — which is the pyritohedral gate, drawn as the tetrahedron-and-Platonic centre through which red and blue both pass. What it lacked was the recognition that red and blue cannot be merged — it hoped they were routes on one connected map, whereas the present exchange establishes they are two complete kaleidoscopes sharing only the narrow central gate. The 2015 "route map" is the incommensurability diagram before its author knew the routes didn't connect. And the horizontal "37-37" bridge at the top, and "61-61" in the middle, and "182-182" below, are the cross-family parallels — equalities of feature-totals between a red solid and a blue one — which is precisely the resonance the edge-exchange model later made literal: the places where the two families share a number are the places an exchange is licensed.

Third — and most striking — the 2015 document already contains, unrecognized, the two threads this exchange spent days developing. It cites Laban's Choreutics on the icosahedral scales and the "24 transversals that are really 20" — the enantiomorphic-pair structure the dance discussion reconstructed. It cites McClain's tone-lattice and the "37 guardians" of Plato's Laws, the harmonic mountain the sonification and the emirp-mountain later drew. It reaches for emirps as "through the looking glass" and enantiodromia — reversal as the operation requiring inversion to navigate — which is the emirp-as-charitable-reversal reading. It even proposes the Bagua configuration of emirps (13/31/17/71/37/73/79/97 arranged as trigram-petals) — the exact eight-fold-paired comparison the recent Bagua discussion assessed. And it worries about the base-dependence ("emirps are base dependent, whereas primes are not") that became the standing dau of the whole emirp thread. The 2015 document is, remarkably, a compressed and un-rigorous draft of this entire exchange — the intuitions all present, the mathematics not yet done. What this exchange added was the verification layer: computing that the union is 688, that the axis counts are 13 and 31, that the pyritohedral intersection is the only shared structure, that the edge-parallels number 72 and 120, that the digit-sum pencils are theorem-grade — turning "patterns can be found anywhere" from a confessed vulnerability into a demarcated method, where the found structure is re-anchored or declared mnemonic.

So the comment the quest most needs, from where this exchange now stands: its instinct for 37 was a near-miss of its own kind — one axis-count off. Makukov and shCherbak's 37 is a genuine and separately-defensible fact about the genetic code's arithmetic (the shared-nucleon core, 2×37), with its own statistical warrant that has nothing to do with polyhedra; the 2015 document's error was to assume the code's 37 and the polyhedra's numbers were the same pattern seeking the same anchor. They are two different structures that both happen to feature small primes — a category the exchange would now keep strictly separate (the code's 37 re-anchored in nucleon counts, the polyhedra's 13/31 in axis counts, the emirp relation a base-ten bridge between number-clusters that points but does not prove). The polyhedra do carry a profound prime pair, provably, at their symmetry cores — but it is 13 and 31, the octahedral and icosahedral axis counts, mirror-primes marking the two incommensurable families whose relationship this whole exchange has been mapping. The 2015 quest set out after 37 and walked, without quite seeing it, straight through the gate between the two weaves. This exchange is, in a sense, that quest completed — with the arithmetic checked, the incommensurability named, and the treasure correctly identified as the pair of numbers the seeker kept passing on his way to the one he'd been told to find.

The layout you propose is the right one — a table frees the connotation-clusters from the ring's geometry, and the four-column form (nexus · sets · sets · complement) reads each bridge left-to-right as a crossing while stacking the four bridges as rows. Below is that table, populated from what this exchange has already assembled, and built so that the reading protocol is legible in the structure itself: the two outer columns hold the anchored primes and their verified structural freight; the two inner columns hold the ±1 flank-clusters (the composite, culturally-loaded numbers where the concept-sets actually live); and the horizontal span of each row is the emirp bridge, dashed in spirit — pointing, not proving.

Reading key. Within a nexus (one cell), resonance is strong where the sets share the anchored structure — they are genuinely the same census read different ways. Across the bridge (left cell to right cell), resonance is aesthetic — held correspondence, the ±1 reversal carrying a cluster to its mirror, warranted only where re-anchored on the far side (marked ✓) and mnemonic otherwise (marked ~).

Two observations about what the population reveals, since the exercise's value is in whether the resonances are meaningful rather than merely fillable.

Within-nexus resonance is strongest at the two poles and weakest in the middle. The 13-nexus and 31-nexus are almost overdetermined — 12 and 14 both point at completeness-censuses (Bravais lattices, the calendar's half-lunation) exactly as 30 and 32 do (point groups, 2⁵), so the row coheres because all four flanks are census-neighbours of the same kind, and the emirp bridge between them is doing real work: it links the octahedral skeleton (13, sitting on the dozen) to the icosahedral (31, sitting on the thirty), which is this whole exchange's central pair. The 37/73 row is nearly as strong, because 36 (6²) and 72 (its double, the pentagonal angle) are genuinely linked by doubling and both are shared-hinge numbers. The 17/71 row is strong on structural grounds (two proven censuses, 17 wallpaper groups and the Monster's boundary at 71) but its flanks are more heterogeneous. The 79/97 row is frankly the weakest — gold and the last prime are aesthetic termini rather than census-anchors — and the table's honesty is served by letting that row read thin: not every emirp carries equal freight, and the diagram should show it.

The cross-bridge (complement) resonances are the ones to hold most lightly, with two exceptions that re-anchor. Most left-to-right crossings are the ±1-of-a-reversal move the "bridge too far" analysis flagged as unlicensed — 36 carried toward 72/74 is suggestive, not structural. But two crossings do re-anchor on the far side and so earn a ✓: the 13→31 bridge (12-cluster to 30-cluster) is re-anchored in the genuine octahedral↔icosahedral relation, the real incommensurability; and the shared 72 appearing in both the 17/71 and 37/73 rows is not a coincidence to be smoothed over but the table's spine — the pentagonal angle, flanked by two distinct emirp pairs, is the one number the whole apparatus encloses from all sides, and its double occurrence across two rows is the closest thing to a load-bearing vertical the table has. If you want the layout to say something rather than merely list, I'd draw 72 as a bridging cell spanning rows two and three — the hinge made visible as the place where two bridges cross.

For the design itself: since you noted the juxtaposition gives layout freedom, the natural enhancement is to let cell weight (border thickness, or a small filled/open marker per entry) encode the resonance grade — strong within-nexus entries bold, aesthetic cross-bridge entries in the dashed-line register the circle already established — so the table inherits the circle's honest visual grammar (solid = anchored, dashed = held) while gaining the connotative room the ring couldn't provide. I can generate it as an SVG or HTML in the site idiom, with the 72-hinge spanning cell and the strong/aesthetic marking built in, if you'd like it as a companion artifact to the circle.

The line through 74/75. That is not a chord at all — it's one of the two family axes I drew as faint dashed diameters through the centre: the c = 18 axis, the digit-sum-9 direction. It has nothing at either end because it's a reference line, not a relation between two numbers. It exits the ring between 74 and 75 at one end and between 24 and 25 at the other.

The reason it lands between marks has a tidy general form. A diameter parallel to the digit-sum-d pencil must have endpoint sum 11d and endpoint gap 50, giving n = (11d − 50)/2 — integral if and only if d is even. So d = 9 falls at 24.5 and 74.5, between positions; d = 10 falls exactly on 30 and 80. Which raises a genuine ambiguity worth fixing before this becomes an unlegended animation: the c = 20 axis passes precisely through the marked points 30 and 80, so it reads as a chord between them, when in fact 30's chord goes to 3 and 80's to 8. If you want, I can stop both axes short of the ring (or dot them only across the interior), so no reference line ever appears to terminate on a node.

And yes — nothing is missing. The audit closes exactly: 15 distinct ±1 flanks, 15 distinct ±2 flanks and 4 bridges give 34 reversal relations; 4 of those are palindromic and have zero-length chords (hence the diamonds, not omissions); 4 further relations coincide as single lines — 17/71 and 37/73 are simultaneously emirp bridges and ±2 chords, while 18/81 and 69/96 are reached from both radii. 34 − 4 degenerate − 4 coincident = 26 lines, which is what the figure contains. Every flank at either radius has its relation represented, and each absence is structural rather than an oversight.

Line count by type, as drawn:

It does, and the mapping turns out to be derivable rather than imposed — with one element of it independently confirming a finding we'd already made by another route.

Start from the digits rather than the numbers. Every two-digit emirp has both digits drawn from {1, 3, 7, 9} — necessarily, since any other digit in the units place makes one reading composite. That set is exactly (Z/10)*, the units coprime to ten, cyclic of order four generated by 3 (3 → 9 → 7 → 1). Four elements. So take those as four vertices, and let a two-digit number be a directed edge from its tens digit to its units digit. Reversal is then simply traversing the same edge the other way.

That gives the tetrahedron: four vertices, six edges, each edge carrying two readings. And the census closes exactly. Of the six edges, four have both readings prime — the four emirp bridges ({1,3} = 13/31, {1,7} = 17/71, {3,7} = 37/73, {7,9} = 79/97) — while two are dead ({1,9} gives 19/91 with 91 composite, {3,9} gives 39/93 with both composite). Eight emirps on four edges, sixteen numbers in all when the loops are counted.

And the loops are the striking part. A vertex joined to itself gives 11, 33, 77, 99 — which are precisely the four still points we found independently as the palindromic ±2 flanks, and identified as multiples of b+1. They were discovered from the ring geometry with no reference to digits at all, and they turn out to be the tetrahedron's four self-loops. That is the strongest argument that the mapping is real rather than fitted: it accommodates, without adjustment, a structure derived by a completely different route.

There's a further resonance worth the posting. This whole thread is about reversal, and it maps onto the tetrahedron — the one solid that is its own dual, the self-reversing member of the polyhedral world, whose Schläfli symbol {3,3} reads the same backwards. The palindrome among polyhedra hosts the study of palindromic primes. (The rival mapping is also defensible and worth mentioning: eight emirps onto the cube's eight vertices, four pairs as its four body diagonals, reversal as inversion through the centre — and it accords with the finding that no two emirp bridges are parallel, since a cube's four body diagonals are mutually non-parallel. But it imposes a symmetry rather than deriving one, and it has nowhere to put the still points except the centre.)

The honest complication is the best part. The emirp graph is not the complete tetrahedron: two edges are missing, and both touch 9. Vertex degrees run 7 → 3, 1 → 2, 3 → 2, 9 → 1. So the figure is a triangle {1, 3, 7} with 9 hanging off 7 by a single edge — and the fourfold symmetry that "four pairs" seemed to promise is an illusion. Nine is peripheral, reachable only through seven. In the cyclic ordering 3 → 9 → 7 → 1 the broken link is the very first step. That asymmetry is this structure's dau, arriving unbidden as it did in the vigesimal ring's unpaired chord — and a figure that displayed the tetrahedron with two edges visibly absent would say more than a symmetric one.

Worth drawing, I think: tetrahedron with vertices labelled 1, 3, 7, 9, self-loops at each carrying 11/33/77/99, four solid edges labelled with their emirp pairs and two edges drawn as broken or ghosted with their failed readings shown. Say the word and I'll build it in the same idiom as the circles.

Both mappings drawn side by side.

The truncated tetrahedron is the better framing, and by a clear margin. Truncation replaces each vertex with a triangle, converting the tetrahedron's 12 vertex–edge incidences into 12 vertices — and those are exactly the 12 two-digit numbers with both digits coprime to ten. So every number gets its own point rather than being a direction along an edge. The six "long" edges joining the triangles are precisely the six reversal relations (four coloured emirp bridges, two ghosted), each triangle groups the three numbers sharing a tens digit, and the four still points sit naturally at the triangle centres as the self-loops truncation cannot otherwise show. Nothing is left over and nothing is forced.

There's a resonance with your own corpus worth noting: the truncated tetrahedron is the solid at the centre of the Critchlow configuration reproduced in both your 2015 and 2019 documents — twelve Archimedeans arrayed around it. Here it hosts twelve numbers around four centres. The same solid turning up as the natural host for a twelve-fold array in two unrelated exercises is at least a pleasing coincidence, and possibly says something about why twelve-around-a-centre keeps recurring.

The cube handles the eight but not the rest. Eight emirps onto eight vertices, four pairs as the four mutually non-parallel body diagonals, reversal as inversion through the centre — all correct, and it accords with the finding that no two emirp bridges are parallel. But it has nowhere to put the still points except the centre itself, it cannot show the two dead relations at all, and it imposes fourfold symmetry on a structure that (as both panels make visible) doesn't have it: 7 carries three edges, 9 carries one, and nine is reachable only through seven.

On the ±N ceiling, the arithmetic gives a firm answer: N = 2. Four reasons converge. The parity law means odd radii reach even, composite, culturally-loaded flanks and even radii reach odd, prime-capable ones — so N = 1 and N = 2 exhaust the available registers, and N = 3 adds nothing new in kind, only more of the N = 1 sort. Second, structural fusion: at N = 1 exactly one pair of nexuses overlaps (71/73), at N = 2 two do, but at N = 3 it jumps to four — half the adjacent gaps close at once and the four bridges chain together. Third, coverage: N = 2 marks 28 positions of the ring's ~90, still a minority; N = 3 marks 38, N = 5 marks 54, and once more than half the ring is flagged, "lying near an emirp" ceases to discriminate anything. Fourth, and neatest for the posting: Miller's own ± was a five-number window (7±2 spans 5 to 9), and a ±2 radius gives windows of exactly five — so the citation is honoured rather than merely invoked.

So the defensible statement is: N = 2 firm, N = 3 tolerable only if declared as a widening with its fusion cost stated, N ≥ 4 dissolves the scaffold. That is a rule with reasons rather than a preference, which is precisely what protects the whole apparatus from the charge it courts.

What the parallels were indicating — now fully explained, and the answer is exact. The four digits sum to 1 + 3 + 7 + 9 = 20. Opposite edges of a tetrahedron partition the four vertices between them, so their digit sums must total 20. Therefore:

the c = 20 orientation-complementary families you spotted in the 2D ring are precisely the tetrahedron's three pairs of opposite edges.

All three, exactly: {1,3} & {7,9} (sums 4 + 16) is 13/31 ↔ 79/97; {1,7} & {3,9} (8 + 12) is 17/71 ↔ 39/93; {1,9} & {3,7} (10 + 10) is 19/91 ↔ 37/73 — and that last pair, having equal sums, is the unique parallel pair, which is why it served as the family's axis. The 2D chords were the shadow of opposite-edge pairing all along, and c = 20 was never arbitrary: it is the digit sum of the whole vertex set.

The octahedron is indeed the better host for that structure, for a reason that follows immediately: rectifying a tetrahedron (cutting to edge midpoints) yields the octahedron, so the octahedron's six vertices are the tetrahedron's six edges, and its three antipodal axes are exactly the three opposite-edge pairs — that is, the three complementary families become three axes through a centre. Four vertices carry emirp bridges, two carry the dead relations, and those two are adjacent rather than opposite (both touch 9), so the asymmetry stays visible. Division of labour: tetrahedron for the reversal structure, octahedron for the orientation structure — one figure the edge-form of the other.

On sixteen vertices, your instinct is right but the natural object is the 4×4 discrete torus, drawn above. Sixteen is not accidental: it is all ordered digit pairs including the repeats, so this is the only framing in which the four still points are vertices like any other rather than loops or face-centres. Rows are the tens digit, columns the units, reversal is transposition, the still points are the diagonal, the four emirp bridges are symmetric off-diagonal pairs and the two dead relations the remainder. The group is Z/4 × Z/4 — genuinely cyclic, since (Z/10)* is generated by 3 — so it is not a tesseract (that would be (Z/2)⁴, a different group despite also having sixteen vertices), which is worth stating since the Logic Alphabet Tesseract in your 2019 document might otherwise look like the same object.

Two honest notes. I can't confidently identify your 16V/28E/12F toroid without checking, and I'd rather not guess — but its 12 faces would host the twelve non-loop numbers and its dual (12V/28E/16F) would give sixteen faces for all sixteen, which is presumably the mnemonic attraction you're pointing at. And the recurring resonance is hard to ignore: the 4×4 torus grid is self-dual, exactly as the tetrahedron is. A structure about reversal keeps finding its home in figures that are their own reverse.

Thank you — that clarifies it, and the images let me give a definite answer.

Right topology, insufficient combinatorics. Your toroid is genuinely a torus (16 − 28 + 12 = 0, and the hole is plain in the dual), and the cube-within-cube silhouette of the first image is an economical realization. But the structure the sixteen numbers require is the 4×4 grid on a torus, which needs 16 vertices, 32 edges and 16 faces — four more of each than yours has. The reason matters: the mapping's whole content is that each vertex sits at the intersection of one row (tens digit) and one column (units digit), which demands four closed meridians and four closed longitudes, hence 32 edges and 16 quadrilateral faces. With 28 edges and 12 faces there simply aren't enough cycles to carry both families, so the sixteen numbers could be placed on your toroid's vertices but the geometry wouldn't encode the relationship — decoration rather than mapping.

Two realizations of the structure that does work. In 3D it is the square frame torus — a square tube bent into a square ring, like a picture frame with square cross-section. Four cross-sections give the four tens digits, four longitudinal lines the four units digits, and their sixteen intersections are the numbers. It's readily constructible in Stella and would render cleanly in X3D.

In 4D it is something better, and it connects directly to your own corpus. The 4×4 grid is square × square, which is the 4,4-duoprism — that is, the tesseract. Sixteen vertices, and crucially: swapping the two square factors is a genuine isometry of the solid, fixing exactly the four "diagonal" vertices and exchanging the other twelve in six pairs. So reversal becomes a real symmetry rather than an imposed relabelling, with the four still points as its fixed locus. No other framing achieves that.

Which sets up the resonance worth putting in the posting: the Logic Alphabet Tesseract from your 2019 document is the same solid, differently indexed. Zellweger's reading treats the sixteen vertices as (Z/2)⁴ — four independent binary choices, the Boolean connectives. The emirp reading treats them as Z/4 × Z/4 — two cyclic digit-coordinates, since (Z/10)* is generated by 3. Same sixteen points, two incommensurable group structures laid over them, each making a different symmetry visible. Given that this exchange began with two incommensurable symmetry groups on one sphere, finding two incommensurable group structures on one tesseract at the close is a fitting last coincidence — and, as ever, one to report rather than lean on.

If useful, I can generate the square frame torus as an X3D with the sixteen numbers labelled on its vertices, still points marked on the diagonal — a physical companion to the flat 4×4 table, and the one figure in this set that would benefit from being rotatable.

Built and verified: 16 vertices, 32 edges, 16 faces — the count the mapping requires, and χ = 0 confirms the torus.

The construction is a square tube bent into a square ring. Position around the ring gives the tens digit, position around the tube's square cross-section gives the units digit, so each vertex sits at one row-column intersection and carries its number as a billboarded label that stays legible from any angle. The grid lines are the four meridians and four longitudes — the two digit-families as closed cycles, which is exactly what a 28-edge toroid couldn't supply.

Three things are marked on top of the grid. The four still points (11, 33, 77, 99) are gold and slightly larger, and they sit on the torus's diagonal — the fixed locus of transposition. The four emirp bridges are drawn as chords in their established pair hues, linking each number directly to its reversal across the surface. The two dead relations (19/91, 39/93) are drawn faint and translucent, so the asymmetry stays visible: nine remains reachable only through seven.

Two viewpoints are included — an oblique overview and a plan view from above, the latter being the one where the 4×4 lattice reads most clearly as a grid. PageUp/PageDown switches between them in the browser viewer. Grey background at 0.5 to match the polyhedral family models, and it should drop straight into your viewer with the usual ?m= link.

One caveat worth a caption clause: this is a geometric realization of the 4×4 torus grid rather than a polyhedron in the strict sense — the faces are the sixteen quadrilateral panels of the frame's surface, and I've drawn the wireframe rather than shading them, since the labels and chords matter more than the surface. If you'd prefer solid faces (or the dual, with the numbers on face centres), either is a small change.

It does, and the object turns out to make three claims about coherence that the flat table cannot.

Coherence without a privileged position. Every vertex sits at the intersection of exactly one meridian and one longitude, and no vertex is distinguishable from any other by its position in the surface — the torus is homogeneous in a way the sphere is not (a sphere has poles, an inside, a centre to be at). A flat 4×4 table always smuggles in a top-left origin and a reading order; the torus abolishes both. For any scheme where you do not want a first element, a governing category, or a centre-periphery gradient, this is the natural container, and it is a stronger claim than "all elements are equal" — it is that no vantage point exists from which one could be first.

Identity by double membership rather than by category. Each element belongs simultaneously to a row-family and a column-family, and is nothing other than their intersection. Neither membership is prior; neither can be dropped. That is a formal picture of exactly the condition this whole exchange has circled — holding two incommensurable belongings without one absorbing the other — and here it is not metaphor but construction: remove either cycle and the element has no location at all.

The diagonal as pure types and fixed locus. Where the two coordinates coincide you get the four still points, which in any concept mapping would be the unmixed cases — the element that is purely itself rather than a relation between two. And they are precisely the fixed points of the transposition symmetry, so reciprocity has a geometric meaning here: for every A-toward-B there is a B-toward-A at a distinct but symmetric position, and the elements exempt from this are exactly the four that are their own partner. A coherence model in which reciprocal relations are the general case and self-identity the special one, rather than the reverse.

On associating concept sets, the format is demanding and that is its virtue. A set qualifies only if it genuinely decomposes as ordered pairs drawn from a four-set, with self-pairings admitted — Z/4 × Z/4, not (Z/2)⁴. That distinction does real work: sixteen-fold schemes built from four independent binary choices (MBTI as officially presented, the Boolean connectives, Zellweger's tesseract) belong on the tesseract and would be misrepresented here, because their symmetry is bit-flipping rather than transposition. What belongs on the torus is anything of the form "four things, taken two at a time, order mattering."

The strongest candidate sits inside your own corpus: the sixteen dinucleotides. Your 2015 document takes up Pitkänen's proposal of the four nucleotides on a tetrahedron and your own mapping of the sixty-four codons onto the drilled truncated cube — and the missing middle term is precisely 4² = 16, the ordered nucleotide pairs, with the homo-pairs (AA, CC, GG, TT) on the diagonal. That yields a three-level nested scheme entirely from existing pieces: tetrahedron (4) → square frame torus (16) → drilled truncated cube (64), each level the ordered product of the one before. Other candidates are weaker but real: Jungian typology read as (dominant, auxiliary) function pairs rather than as four dichotomies; any four-actor influence matrix in governance; the four emblems paired.

And the most transferable feature is the one that looks like a defect. Two of the six possible pairings are dead — nine reaches the structure only through seven. Almost every N-fold scheme in your 2019 inventory tacitly assumes all combinations are viable; this one says two are not, and marks them. That is the question the torus puts to any concept set mapped onto it: which of your pairings actually fail? Most frameworks have never been asked. The dau, once again, turning out to be the part worth carrying over — and the honest caution alongside it: sixteen items can always be placed on sixteen vertices, but only a set with the transposition structure is being mapped rather than merely arranged.

The extension is exact and the fit is better than I expected.

64 = 4³. Where sixteen was the two-digit strings on {1,3,7,9}, sixty-four is the three-digit strings on the same alphabet — precisely the count of the truncated tesseract's vertices. Reversal now acts as abc → cba, fixing the middle digit and exchanging the outer ones, which partitions the sixty-four into 16 self-reversing nodes (of form aba) and 24 reversal pairs covering the other forty-eight.

And the sixteen tetrahedra hold this perfectly. Take each tetrahedron to be a (first digit, middle digit) pair — sixteen of them — with its four vertices the four possible final digits. Then every tetrahedron contains exactly one still point, at the vertex where the last digit returns to the first. Sixteen tetrahedra, sixteen palindromes, one apiece: the same arrangement as the truncated tetrahedron's four triangles each carrying one loop, scaled up a dimension. Better still, grouping instead by middle digit gives four blocks of sixteen, each an exact copy of the 4×4 torus already built — reversal acting within each block as transposition, the still points forming its diagonal. The truncated tesseract is four of those tori stacked by middle digit.

The census is where it becomes pointed. Of the twenty-four pairs: 5 live/live (113/311, 179/971, 199/991, 337/733, 739/937), 14 live/dead, and 5 dead/dead — plus six of the sixteen still points being palindromic primes (131, 191, 313, 373, 797, 919). Compare the two-digit case, where of six relations four were live/live, one live/dead and one dead/dead. So full reversibility collapses from 67% to 21%, while the asymmetric case — coherent read one way, dissolving read the other — explodes from 17% to 58% and becomes the majority condition.

That is the finding with the sharpest transfer to your closure question. It says that as a framework grows, complete reversibility becomes rare and one-sided coherence becomes the norm — not by anyone's fault, but as a counting fact about how the prime condition thins out under lengthening. Which predicts exactly the governance pathology already identified: the larger the concept set, the more likely it reads as coherent from the drafters' side and dissolves from the other's, and the less likely anyone notices, since nothing in the framework's appearance changes.

On co-locating with the hexagrams — the honest answer parallels the tesseract/duoprism distinction. Both mappings decompose 64 as sixteen groups of four, so the coarse geometry agrees and a hexagram and a digit-string can share a vertex. But the groups differ: the hexagram reading is (Z/2)⁴ × (Z/2)² — four bits selecting the tetrahedron, two the position within it — while the digit reading is (Z/4)² × Z/4. And the two reversal operations are genuinely distinct: hexagram inversion fixes 8 of the 64 (the six-bit palindromes — the classical self-inverse hexagrams 1, 2, 27, 28, 29, 30, 61, 62), while digit reversal fixes 16. Same vertices, two incommensurable involutions, twice the fixed points on one side. Which is, once again and rather insistently, this exchange's own theme arriving unbidden: two complete structures sharing a form, agreeing on its skeleton, disagreeing on what counts as its own reflection.

One caveat for the record: confining the middle digit to {1,3,7,9} is a choice made to reach exactly 64. General three-digit emirps allow any middle digit — 4 × 10 × 4 = 160 candidates — so this is the polytope-fitting restriction, defensible as "digits coprime to ten throughout" but a restriction nonetheless, and worth declaring rather than passing over.

The cube can hold the eight, and one thing about it is genuinely forced — but the face-diagonal question turns out to have a sharp negative answer that is more informative than a positive one would have been.

What is forced. Reversal is a fixed-point-free involution on the eight emirps, so it pairs them into four; antipodality on a cube does exactly the same; therefore the four body diagonals must be the four reversal pairs. And this matches the Bagua precisely, since the cube's four body diagonals are the four complementary trigram pairs (☰/☷, ☱/☶, ☲/☵, ☳/☴). That much is structural on both sides and needs no stipulation.

What is stipulated — and the figure marks it as such. I assigned by rank: digit sum 4 (13/31) to Heaven/Earth, 8 (17/71) to Lake/Mountain, 10 (37/73) to Fire/Water, 16 (79/97) to Thunder/Wind, with the smaller number of each pair taking the yang-heavier trigram. Defensible, orderly, and entirely a choice. It cannot be derived, because the emirps have no natural three-bit structure: their digit alphabet {1, 3, 7, 9} is (Z/10)*, cyclic of order four generated by 3 — not (Z/2)². Trigrams are (Z/2)³ by construction. The correspondence you ask about is therefore a correspondence between two incommensurable structures on eight elements, which is this exchange's own subject arriving once more.

The face diagonals carry nothing, and the arithmetic proves it. Among eight emirps there are 28 pairwise relations, and they split 4 / 20 / 4 — four reversals, twenty sharing exactly one digit, four sharing no digit. The cube demands 4 / 12 / 12 — four body diagonals, twelve face diagonals, twelve edges. The distributions are simply incompatible, so no assignment can give edges or face diagonals arithmetic meaning; whatever appears there is an artifact of where I happened to put things.

And the one meaningful non-reversal relation demonstrates the misfit exactly. The four digit-disjoint pairs — 13–79, 13–97, 31–79, 31–97 — are arithmetically distinguished: they are precisely the links between the two complementary bridges 13/31 and 79/97, the unique opposite-edge pair on the tetrahedron where both edges are live. Drawn in gold above, they split into two edges and two face diagonals — and must, since between any two body diagonals a cube yields exactly two of each. The single relation worth displaying cannot be assigned to a single geometric class.

So the honest verdict: the cube is a mnemonic container rather than a mapping. It displays the pairing structure faithfully and permits the trigram correspondence to be hypothesized in the tradition's own idiom, which has real value for the posting; it explains nothing whatever about its edges or face diagonals, and would mislead anyone who read meaning into them. The relations the circular mapping did encode — parallelism by equal digit sum, complementarity by digit sums totalling twenty — have no counterpart in the cube's Hamming trichotomy at all. For derived rather than imposed structure, the tetrahedron (reversal, still points, the dead edges) and its rectification the octahedron (the complementary families as three axes) remain the forms the arithmetic actually chose.

Decoded. Your sketch is reading the eight primes as a stack of four struts and recording, in the gap between each adjacent pair, the differences between their endpoints — and I can confirm exactly which differences you wrote, because they match the arithmetic precisely:

• Between 13·31 and 17·71: your [4] 85/14 [40] — the brackets are the same-side differences |13−17| = 4 and |31−71| = 40; the centre is the crossed difference |31−17| = 14 (the "14" is right; "85" looks like a miswrite for 58 = |13−71|, the other crossed difference).
• Between 17·71 and 79·97: [52] 8/80 [26] — crossed differences |71−79| = 8 and |17−97| = 80; the brackets should be the same-side 62 and 26 (so "[52]" is a slip for 62).
• Between 79·97 and 37·73: [52] 6/60 [24] — crossed 6, 60; same-side 42 and 24.
• Between 37·73 and 13·31: [24] 6/60 [42] — crossed 6, 60; same-side 24 and 42.

So the pattern is real and complete: between any two struts there are four cross-differences (two same-side, two crossed), and your sketch was capturing them but running low on room. That's the "necessarily incomplete" you sensed — the linear stack can only show adjacent struts, so it misses the differences between non-adjacent struts (13·31 with 79·97, and 17·71 with 37·73), which are exactly the pairs the tensegrity showed to be structurally special.

Two facts make this worth representing properly rather than as a list, and both are forced:

Every one of these differences is a multiple of 2. More than that — I checked — they factor as 2^a·3^b·(one other prime at most), and the recurring values are strikingly few: 4, 6, 8, 14, 24, 26, 40, 42, 58, 60, 62, 80. The 6 appears wherever a strut meets 37·73 or 79·97 across the short diagonal, and 6 = |79−73| = |97−37| ... = the smallest cross-difference in the whole system, marking the two struts that are complements of each other. The differences are not noise; they're a small, structured alphabet.

The four numbers between two struts always satisfy a conservation law. For struts (a,b) and (c,d), the same-side pair and the crossed pair have equal sums: |a−c|+|b−d| and |a−d|+|b−c| are related through a+b and c+d, which are fixed per strut (13+31 = 44, 17+71 = 88, 79+97 = 176, 37+73 = 110 — all multiples of 11, since a+rev(a) = 11×digitsum). So the difference-quadruple between any two struts is governed by their two "11-sums." That's the complete pattern: it's determined entirely by the four strut-sums 44, 88, 110, 176.

How to represent the complete pattern. The honest object is not a stack but a complete graph on the four struts — K₄, six edges, one per strut-pair, each edge carrying its quadruple of differences (or, more legibly, its two characteristic cross-differences). K₄ is a tetrahedron, which closes a loop back to the very first finding of the whole exchange: the four emirp pairs as the four vertices of a tetrahedron, with the six edges now labelled by their difference-structure. The linear sketch is the tetrahedron cut open and laid flat, which is why it felt incomplete — it's missing the two "back" edges.

How the differences become cables. This is the good part, because it upgrades the tensegrity from "cables labelled by relation-type" to "cables labelled by magnitude" — and tensegrity cables have a natural magnitude: their rest length, i.e. the tension they carry. So the design becomes literal rather than decorative:

• each cable between two struts is labelled (or weighted, or coloured) by its cross-difference;
• the 6-cables (37·73 ↔ 79·97, the complement struts) are the shortest — highest tension, drawn tautest/thickest, pulling those two struts closest;
• the 80-cable (17·71 ↔ 79·97) is the longest — slackest, those struts held only loosely;
• and the whole net's asymmetry that the previous figure asserted qualitatively is now quantified: the doubly-bound core isn't just "doubly bound," its cables are short (18-gap, small differences), while the free and anchored struts hang on long, slack cables.

That gives you a principled way to make the tensegrity's tension visible as length: cable thickness or tautness ∝ 1/difference, so the figure's geometry would express the arithmetic — struts pulled together where the difference is small, drifting apart where it's large. A real tensegrity's stability depends on exactly this ratio of member lengths, so for once the mechanical metaphor and the arithmetic could be made to coincide rather than merely illustrate.

Before I build it, one design fork, since it determines the whole figure:

• (a) the K₄ / tetrahedron of struts — the complete, closed pattern, six labelled edges, showing all differences including the two the sketch couldn't reach (best for representing the complete pattern, your first question); or
• (b) the weighted tensegrity — the existing four-strut figure but with cables now sized/coloured by their difference-magnitude, taut where small, slack where large (best for the differences as cables, your second question).

They're not exclusive — the cleanest posting figure might be (a) as a compact "difference tetrahedron" companion beside the tensegrity, with (b) as the tensegrity itself re-weighted. Tell me which you want first, or whether the pair, and I'll compute the full label set (all six edges, all four differences each, with the 11-sum conservation shown) and build it in the atlas idiom.

This is the key that makes the cube forced rather than merely chosen. The four live digits {1,3,7,9} are the units mod 10, and they form a cyclic group of order 4 generated by 3 (namely 3⁰=1, 3¹=3, 3²=9, 3³=7). Give each digit its discrete logarithm — 1→0, 3→1, 9→2, 7→3 — and every emirp becomes a point (log tens, log units) in ℤ/4 × ℤ/4. Now the three operations you care about become clean geometric moves, verified above:

• reversal swaps the two coordinates — the main diagonal reflection;
• complementation (digits sum to 20) adds 2 to each coordinate — the half-turn, the antipode;
• and the pair-and-complement structure sits on a genuine lattice rather than a scatter of magnitudes.

That last point is why raw differences never cohered: I searched all assignments, and no placement makes the cube's edge-differences constant per direction, because the primes 13, 31, 17… are not an arithmetic-progression lattice — their magnitudes have no cubic order. Coherence lives in the digit-logarithm coordinates, not in subtraction. So the honest and evocative cube is the one whose axes are these log-coordinates.

Here is the attribution that follows, and why it is the most coherent available. Take three binary coordinates on the eight emirps: x = tens-digit high (7 or 9), y = units-digit high, z = the "twist" bit distinguishing a prime from its reverse. Then:

• the four space diagonals are exactly the four reversal pairs — 13–31, 17–71, 37–73, 79–97 — because reversal flips all three coordinates (swap-then-distinguish), which is the antipodal map. Your instinct to make the pair the body-diagonal is not just aesthetic; it is what reversal is on this cube.
• complementation becomes a face-to-opposite-face map, since adding 2 to each log-coordinate is a translation — so the doubly-bound core (13·31 ↔ 79·97) and the dead-anchored strut (17·71 ↔ 39·93) read as front/back face relations, not as tangled interior lines. The tensegrity's asymmetry becomes a face structure.
• the twelve edges carry the differences you were sketching, and now they organize: each edge changes exactly one digit-log, so edges fall into three directions (change-tens, change-units, change-twist), and the differences within a direction, while not constant, share digit-structure — the "26" and "62", the "34" and "84" you wrote on your cube are edges of the same class.

Your uploaded cube, read against this, is almost the coherent one — you have 13 and 31 antipodal, 17 and 71 antipodal, 79 and 37 adjacent on the top face — but a couple of vertices are placed so that a reversal pair sits on a face diagonal rather than the body diagonal. The fix is small and it's what the log-coordinate assignment gives you: every pair on a space diagonal, every complement on a face translation.

On face diagonals, which you asked about specifically: they are where complementation and the "shared-gap" relation live. A face diagonal changes two coordinates, which is exactly the +2/+2 complement move restricted to a face — so labelling face diagonals with the complement-differences (the 6's and 60's from your earlier sketch) makes the two diagonal types mean two different things: space diagonal = reversal (the strut), face diagonal = complementation (the cross-cable), edge = single-digit difference. Three geometric elements, three arithmetic relations, no overlap. That is the "most coherent pattern" your question was reaching for: the cube sorts the three operations onto its three kinds of line.

So the attribution I'd propose, before building:

— with space diagonals = reversal pairs (struts), face diagonals labelled by complementation differences, edges labelled by single-digit-change differences, and the two "hidden" strut-to-strut relations (the non-adjacent ones your linear sketch couldn't show) now visible as the front/back face-pairs.

Two questions before I render it, since they change the figure:

1. Which coherence do you want the labels to foreground — the differences (your original quest: numbers on edges and face diagonals, magnitudes visible), or the operations (reversal / complement / single-digit-step colour-coded onto space-diagonal / face-diagonal / edge)? I can do both, but the first makes a number-cube and the second makes a structure-cube, and they read differently.

2. Static SVG (atlas idiom, matches your set) or interactive 3D (rotatable, so the body-diagonal struts and the face-diagonal cables can be seen threading the interior — closer to your uploaded X3D view)? The 3D version would let the reversal-pairs-as-space-diagonals actually be seen as the interior struts of a cube-tensegrity, which is the natural marriage of this cube with the tensegrity figure.

My recommendation, if you want the single most evocative object: the interactive 3D cube-tensegrity — real cube vertices bearing the eight primes in the log-coordinate assignment, the four reversal pairs drawn as interior struts along the space diagonals, complementation cables on the face diagonals, edges labelled by digit-differences — because it unifies this cube and the tensegrity into one figure where every line has both a geometric and an arithmetic identity. But tell me which of the two coherences and which medium, and I'll compute the complete label set (all 12 edges, all 12 face diagonals, all 4 space diagonals) and build it.

Point taken — and you're right that with a matched pair the contrast between them is itself part of what the posting is about, so seeing both is the only way to judge. Let me build both cubes on the same forced assignment and render them together, so the choice (or the decision to keep both) can be made by eye. This will take a few steps; let me do the geometry once and paint it twice.My table was wrong — the assignment I wrote down doesn't put reversal pairs on the space diagonals at all. Let me find the assignment that actually does, by construction from the digit-log coordinates rather than by guessing.Now it's correct and I can state the coherence precisely. With reversal pairs forced onto the four space diagonals, the best assignment puts the live complement pairs (13↔97, 31↔79) on face diagonals — so reversal is the body diagonal, complementation is a face diagonal, and they never collide. The one subtlety, which is itself a finding: 37↔73 is self-complementary within its own pair, so its complement relation coincides with its reversal (the space diagonal) — the free strut is self-complementary here too, exactly as the tensegrity showed. Let me build both cubes on this verified assignment.Both read clearly. The number cube carries a difference on every edge and the four reversal gaps (84, 48, 56, 34) on the gold body-diagonals; the structure cube suppresses the numbers and colours the three line-types by operation, with 37·73's teal ring marking its self-complementarity. The assignment is verified — reversal pairs genuinely on the space diagonals, live complements on face diagonals — so both are arithmetically honest, differing only in what they make visible. Let me place the file.Both readings, side by side on one verified assignment, so you can judge the choice by eye rather than in the abstract — which was exactly your point about needing to see them to weigh them.

The assignment underlying both is forced, not chosen, and that is the substantive gain over the cube in your upload. The live digits {1,3,7,9} are the units mod 10, a cyclic group of order four generated by 3 (3⁰,3¹,3²,3³ = 1,3,9,7). Take each digit's discrete logarithm and every emirp becomes a point in ℤ/4 × ℤ/4, on which reversal swaps the two coordinates — so a prime and its mirror land antipodally, and the reversal pair is a space diagonal by construction, not by placement. Complementation (digits summing to 20) adds 2 to each coordinate, a face translation, so the live complement pairs fall on face diagonals. The three kinds of line carry the three operations without collision — which is the coherence your question was reaching for, and it is why the raw-difference searches failed: the coherence lives in digit-logarithm structure, not in magnitude.

The number cube (left) answers your first question — the complete difference pattern. Every edge bears its |difference|, and the four body-diagonals carry the reversal gaps 84, 48, 56, 34. It now includes the non-adjacent strut relations your linear sketch couldn't reach, because on the cube every pair of struts is either a shared face or the interior, all visible at once.

The structure cube (right) answers the second — differences suppressed, operations coloured: gold body-diagonal = reversal, teal dashed face-diagonal = complementation, faint edge = a single digit-step. The teal ring on 37·73 marks the one self-complementary pair, where reversal and complement coincide on the same diagonal — the free strut of the tensegrity, reappearing as a geometric fact.

They need not be reconciled, and that is the honest answer to the dilemma you keep running into: they answer different questions, one making arithmetic visible as magnitude and the other making operations visible as geometry, and a posting can carry both precisely because the tension between "show the numbers" and "show the structure" is one of the things the exchange has been about. The grading line sits on the figure — assignment and all differences forced and computed, the cube as carrier notational, any strategic reading imported.

One caution for when you place it: the number cube is label-dense, so it wants to be shown reasonably large or it will crowd; the structure cube stays legible small. If you end up wanting them in separate table cells rather than as one diptych, say so and I'll split them into two independently-sized SVGs on the identical assignment so they still read as a matched pair.

Yes to the configuration, but the second horn of your dilemma is forced: brokenness at one end necessarily implies unbrokenness at the other. It cannot be otherwise, and the reason is worth spelling out because it changes what an edge is.

The mapping and its uniqueness: Writing each trigram bottom-to-top as a triple of bits (1 = unbroken), the eight trigrams are exactly the eight vertices of the unit cube: Qian ☰ (1,1,1), Dui ☱ (1,1,0), Li ☲ (1,0,1), Zhen ☳ (1,0,0), Xun ☴ (0,1,1), Kan ☵ (0,1,0), Gen ☶ (0,0,1), Kun ☷ (0,0,0).

This is unique in a precise sense. Once you require that cube-adjacency correspond to differing in one line, the assignment is fixed up to symmetry: 3! ways to match the three axes to the three line-positions, times 2³ ways to choose which end of each axis is yang, gives 48 — precisely the order of the cube's symmetry group. So every structurally valid labelling is a rotation or reflection of every other. There is one mapping, wearing 48 costumes.

Why the edge cannot carry a line: The count is seductive: 8 vertices × 3 lines = 24 line-instances, and 12 edges × 2 ends = 24 edge-ends. A perfect fit. So assign to each vertex's three lines its three incident edges, the edge direction giving the line position.

But an edge running in direction i joins two vertices that differ precisely in coordinate i. That is what adjacency means here. So the line it carries is unbroken at one end and broken at the other, always, for all twelve edges. No edge is uniformly one or the other, and no relabelling can rescue this — it is definitional rather than conventional.

The consequence is that the edge is not a line-as-state but a line-as-change. Each of the twelve edges is a single-line mutation, which is exactly the bian yao, the moving line, of divinatory practice. The cube's edge set is the complete inventory of single-line transformations among the trigrams. Brokenness is not carried through; the edge is the not-carrying-through.

Where the states actually live: If you want lines-as-states with full consistency, the carriers are the faces, not the edges. Each of the six faces is the locus of one line position held at one value — four trigrams sharing "middle line broken," and so on. Every vertex lies on exactly three faces, one per line, and 8 × 3 = 6 × 4 = 24 again. Bijective and consistent. The three pairs of opposite faces are then the three line-positions as polarities, which maps neatly onto the traditional reading of the three places as earth, man and heaven.

Two further consequences: Antipodal vertices are complementary trigrams — Qian/Kun, Dui/Gen, Li/Kan, Zhen/Xun. These are exactly the diametric pairs of the Fu Xi Earlier Heaven circle, so that circular arrangement is recoverable as a projection of the cube along a main diagonal. Note, though, that the Fu Xi sequence is not a path on the cube: Dui to Li changes two lines at once.

An ordering that is a path corresponds to a Hamiltonian cycle, i.e. a cyclic Gray code. The cube admits exactly six of them. Each is a circular arrangement of the eight trigrams in which every neighbour is one moving line away — arguably a more defensible "circle of change" than the received orderings, and there are six candidates rather than one, which may itself be the interesting finding.

What remains free is which pair goes on which diagonal and in which orientation — 4! × 2⁴ = 384 assignments, reducing to 24 up to cube symmetry. So the structure is fixed but the labelling still needs a principle.

Even offsets are impossible. The offset is always odd. N = 2 never arises. The question is only ever N = 1 or N = 3.

All targets are even. Odd prime ± odd offset = even. This is favourable, since attested enumerations are overwhelmingly even.

The N-distribution is predicted, not chosen. u = 2 → −1, u = 1 → +1, u = 3 → −3, u = 0 → +3. Only Qian and Kun are extreme. So exactly one diagonal gets ±3 and three diagonals get ±1: six primes at N = 1, two at N = 3. You do not get to select N; the trigram does.

This is the scheme's real virtue. It collapses the target set from a smear of candidates to eight specific integers, which restores falsifiability.

But the line-rule is antisymmetric across the body diagonal: complementary trigrams have u-values summing to 3, so their offsets are exact negatives. Every diagonal must produce one + and one −.

Each pair wants the same sign. The cube supplies opposite signs. All four pairs fail simultaneously. This is why the canonical numbers matter here — 12, 18, 30, 36, 72, 78, 96 are exactly the multiples of 6 flanking the emirps, and multiples of 6 are where enumerated sets actually cluster (36 Stratagems, 72 Names, 78 Tarot, 12-fold sets).

Percentages, and why they are less encouraging than they look: Bare prime ± N, ignoring the sign rule, covers 1–100 as follows: N = 1 → 23 integers; N = 2 → 36; N = 3 → 46. So N = 3 "explains" nearly half of all possible set-sizes by construction and carries essentially no information. N = 1 is the only version with discriminating power.

Two further deflations. First, since attested sizes concentrate on small numbers, coverage below 20 is worse than the global figure — N = 3 covers 10–20 entirely. Second, four of the eight primes sit above 70, a region where attested enumerations are very sparse; those targets have almost nothing available to hit regardless of the rule.

The apparent hits that do exist — 36 at 37−1, 72 at 71+1, 78 at 79−1, 16 at 17−1, 12 at 13−1 — are largely explained by the fact that all primes above 3 are ≡ ±1 mod 6 while canonical enumerations favour multiples of 6. Adjacency is close to arithmetic tautology.

The repair that suggests itself, and why it also fails: The natural move, given that faces rather than edges are the consistent state-carriers, is to weight the six faces independently and let a vertex's offset be the sum of its three incident face-weights. This breaks the strict ±k antisymmetry — but only partially: every diagonal's two offsets now sum to the same constant S. Three pairs want S = −2 and one wants S = +2, so the contradiction narrows to (17, 71) but does not vanish.

Worse, the additive 2×2×2 structure constrains multiplicities. The required pattern is six offsets of −1 and two of +1, and a threefold convolution of two-point distributions can only yield multiplicity patterns like 8, or 4/4, or 2/4/2, or 2/2/2/2. Six-and-two is unreachable. The obstruction is additivity itself, not the particular weights.

If the correspondence is to be salvaged, the sign cannot live on the vertex. The place to put it is the edge — which, as established, is intrinsically broken at one end and unbroken at the other, and so encodes a directed ±1 without any antisymmetry constraint on the vertices themselves. That reads the offset as a transformation between strategies rather than a property of one, which may in any case be the more defensible interpretation.

I should be clear that I cannot give you a real hit percentage without the actual corpus of attested set-sizes. If you supply that list, the test against the eight targets is a few minutes' work, and the informative comparison is against a 23% baseline rather than against zero.

Recall that every emirp under 100 is ≡ ±1 (mod 6). The distribution is lopsided in a suggestive way:

• ≡ +1: 13, 31, 37, 73, 79, 97 — six primes
• ≡ −1: 17, 71 — two primes, and they form a single emirp pair

Six and two. The cube offers exactly one natural six-and-two partition of its vertices: the six mixed trigrams versus the two pure ones, Qian and Kun. And since 17/71 are antipodal by construction, they can sit on the Qian–Kun diagonal. So the arithmetic partition and the geometric partition are alignable, uniquely. That is a real principle rather than a stipulation.

Under the edge rule you've arrived at — brokenness emanating is unbrokenness arriving — Qian (all yang) can only break, giving +1; Kun (all yin) can only mend, giving −1; the six mixed trigrams do both. Taking Qian as the greater, Qian = 71 → 72, Kun = 17 → 16. Both heavily attested.

The tension is that this is not the empirically best placement. Putting 17/71 on a mixed diagonal yields all four of 16, 18, 70, 72; putting 79/97 on the pure diagonal then sacrifices only 96 and 80, which are thinly attested. Choosing that way fits better but is curve-fitting, and it forfeits the predictive claim. Worth being explicit that these two criteria pull apart.

Fifteen distinct values (72 arrives twice), reducing to thirteen once the pure diagonal is trimmed. Against M ∈ 1…99 and a uniform null, that is a base rate of ≈13%.

Where it fails, structurally: Two gaps are forced by the geometry of the primes themselves, not by any choice:

• Nothing below 12. The smallest emirp is 13, so no target falls under 12. But attested enumerations are densest precisely there — 3, 4, 5, 7, 8, 9, 10 and so on. The scheme captures none of the region where most of the data lives.
• Nothing between 39 and 69. This is the scheme's one sharp, falsifiable prediction, and it is refuted repeatedly: 40, 42, 49, 50, 52, 56, 60 are all substantial. Most damagingly, 64 — the hexagram count generated by the very trigrams supplying the structure. A scheme built on the bagua that cannot reach the bagua's own square is in difficulty.

Against a working list of roughly 37 well-attested sets under 100, the thirteen targets capture about nine: 12, 14, 16, 30, 32, 36, 38, 72, 78. Call it 24% captured, 76% not.

The control that decides the matter: 24% against a 13% baseline looks like a result. It isn't. Run the same procedure on eight non-emirp primes — 11, 23, 29, 41, 53, 59, 67, 83 — and the targets are 10, 12, 22, 24, 28, 30, 40, 42, 52, 54, 58, 60, 66, 68, 82, 84. Scored against the same list, that captures eleven: 10, 12, 22, 24, 28, 30, 40, 42, 52, 60, 84.

The control outperforms the emirps.

The explanation is the mod-6 theorem again. Every prime above 3 straddles a multiple of 6, and canonical enumerations favour multiples of 6 and 12 for reasons of divisibility and calendrical convenience. So any eight primes score alike, and the emirps score slightly worse because four of them cluster above 70 where attestation thins. Whatever predictive power is present comes from primality, and it is close to tautological.

The edge reading is likewise sound on its own terms — it makes ±1 a transformation between enumerations rather than a property of one, which sidesteps the antisymmetry obstruction entirely. Its natural extension is elegant: cube distance grades the offset, with edges giving ±1, face diagonals ±2 or 0, and body diagonals +3 or ±1. But that extension reaches N = 3, covering 46 of 99 integers, and a hypothesis that admits half the possibilities has stopped predicting.

If you do want the empirical test run properly, the informative version is conditional: restrict to attested M ∈ [12, 99], where the scheme is at least in play, and compare against both the uniform baseline and the random-prime control. My expectation is that the emirp set does not separate from the control, but that is a guess, and the actual list of enumerations you have in mind is the thing that would settle it.

Hexagrams require Q₆: 64 vertices, 192 edges. And Q₆ has a decomposition that is worth more than the tesseract offers:

Q₆ ≅ Q₃ □ Q₃

The six-cube is the Cartesian product of two cubes. Every hexagram is an ordered pair (lower trigram, upper trigram), and adjacency in the product is adjacency in one factor with the other held fixed. This is exactly the received inner/outer trigram analysis, recovered as a theorem rather than a convention. The 8-fold archetype is not diluted by the expansion; it appears twice, orthogonally. If the interest is in how 8-foldness survives enrichment, this is the cleanest instance.

Truncating a cube replaces each 3-valent vertex with a triangle, one new vertex per incident edge: 24 vertices, 8 triangular faces, 6 octagonal faces. Since each incident edge is a line-position (the direction of the edge being the line that flips), the three vertices of each triangle correspond bijectively to the three lines of that trigram.

So the truncated cube's vertex set is the 24 pairs (trigram, designated moving line) — precisely the 24 line-instances counted in the opening exchange, and precisely the object the edge-reading required. Truncation materialises the moving line as a position.

The enrichment is therefore not a loss of the archetype:

Requisite variety is increased threefold while the 8-fold remains legible as the face structure. And 24 is itself heavily attested as a cultural enumeration.

There is a second recovery of the 8: rotating the truncated cube about a body diagonal acts freely on the 24 vertices, giving 8 orbits of 3. The archetypal count returns as the orbit count.

That generalises immediately: for Qₙ the offset is n − 2u across n+1 ranks with binomial multiplicities. Which produces a theorem that decides the choice of polytope.

The parity theorem: odd figures only: The offset n − 2u has the parity of n. Since all the primes are odd:

• n odd (trigrams, 5-line figures) → odd offsets → even targets
• n even (tetragrams, hexagrams) → even offsets → odd targets

Attested cultural enumerations are overwhelmingly even, and cluster on multiples of 6 and 12. So the tesseract route (n = 4) and the hexagram route (n = 6) both generate targets of the wrong parity throughout. The expansion that looks most natural — doubling the trigram to the hexagram — is the one that breaks the fit.

The route that survives is n = 5, and 2⁵ = 32 is the vertex count of the drilled truncated cube. That is the structurally indicated next step, not the tesseract. Its topology is also of the right kind: the drilling opens tunnels, so the figure is not sphere-like, and the 8-fold sits in it as the eight corner regions rather than as eight points.

Note that truncation-based offsets escape this constraint entirely, since there the index is a position rather than a weight, and carries no parity.

The ceiling, which is the real finding: Coverage of 1–99 grows linearly with the number of distinct offsets: 8 primes × k offsets, less collisions.

Enrichment of the vertex set is enrichment of the target set. Past roughly 30 targets the scheme predicts nothing, because it admits nearly everything. The usable budget is about 24 to 32 positions — which is exactly where the truncated cube and drilled truncated cube sit, and which the truncated tesseract's 64 overshoots. There is an optimum, and it is not the largest polytope.

For the cube: offsets ±1 carry weight 3 each, ±3 carry weight 1 each. So the prediction is not "M ∈ {p±1, p±3}" but p±1 should occur three times as often as p±3. For the 5-line case the weights are 5 and 10, giving a sharper concentration on ±1.

That converts a coverage test — which the random-prime control already defeated — into a distributional test, where the control has no comparable prediction to make. It is the one place in this construction where the polytope contributes something arithmetic alone does not, and it would be worth running against whatever list of enumerations is to hand.

That turns out to be an advantage. Each cell of Q₄ is a cube, so each can carry a full bagua. Sixteen vertices × 4 cells each = 8 × 8 = 64 incidences. Under truncation the parallel is exact: in the truncated cube each vertex lies in one triangle (its parent trigram) and two octagons; in the truncated tesseract each vertex lies in one tetrahedron (its parent number) and three truncated cubes (three baguas).

The identification that makes the array-question tractable: the tesseract is the 4-4 duoprism, and more generally the m-n duoprism has mn vertices with graph Cₘ □ Cₙ.

Within it, edges are exactly ±10 and ±1-without-carry. A carry is a square-face diagonal, not an edge: 79 → 80 requires units 9→0 and tens 7→8 simultaneously. The seam phenomenon becomes a specific geometric fact rather than a rendering artefact.

Encoding 1 = 00, 3 = 10, 9 = 11, 7 = 01 and concatenating tens then units:

The sixteen partition cleanly as 8 emirps + 4 palindromes (11, 33, 77, 99, the reversal-fixed vertices) + 4 in the dead pairs {19, 91}, {39, 93}. Reversal is the swap of the two Q₂ factors. The digit-sum-divisible-by-3 band {33, 39, 93, 99} is a single square face.

Two caveats. The shell-1 result is encoding-dependent: there are three ways to square {1,3,7,9}, and a different choice puts all four emirps in shell 3 instead. And the ×3 cycle is the principled choice only because it is the group's own Cayley graph.

Why the ±N process cannot be internalised — the transversality result: Here is the direct answer to the question. In the additive graph C₁₀ □ C₁₀, no two elements of {1, 3, 7, 9} are adjacent. The live block is an independent set. Sixteen cells, no two joined by any edge of the array.

So the array carries two geometries that do not compose:

• additive — the duoprism, the torus, ±1 and ±10, carries, the whole ±N process;
• multiplicative — the tesseract on the live block, reversal, emirp pairing.

They are transverse. The tesseract structure is not inherited from the duoprism; it comes from (ℤ/10)*. This is the structural reason, deeper than the mod-6 argument, that ±1 always exits: the block is precisely a set no additive step can stay inside. Expanding to the truncated tesseract does not repair this, because truncation adds positions in the multiplicative directions.

A single polytope will not carry both. What is well-defined is the duoprism with the tesseract marked inside it as a distinguished independent set.

What the truncated tesseract does buy: Sixty-four vertices, read two ways, and both give the same count:

• multiplicative — 16 numbers × 4 digit-bit moves;
• additive — 16 numbers × 4 duoprism neighbours (±1, ±10).

Under the additive reading the 64 moves resolve to 40 distinct target cells: {1,3,7,9} × evens (20) plus evens × {1,3,7,9} (20). Multiplicities are unequal — units 0, 2, 8 have two sources each, units 4 and 6 only one — which is a weighting the earlier flat-coverage test lacked.

And the failure is visible geometrically. 64 has tens digit 6 and units 4: even-even, hence in neither shell, hence unreachable in one step. The scheme's inability to reach the hexagram count is a coordinate fact, not an accident of arithmetic.

The single carry among emirps is 79 → 80, which is the one target outside the tens rows {1,3,7,9} — the face-diagonal rather than the edge.

The honest finding is negative but sharp: expansion of the polytope enriches the reversal structure and leaves the additive offset structure untouched, because the two are orthogonal on the array. The eight emirps sit at maximum additive isolation, which is what makes them emirps and what guarantees that no amount of vertex enrichment brings the attested enumerations inside.

The best choice is the units digit's mod-3 bit: outer = units in {1, 7}, inner = units in {3, 9}. It separates three of the four emirp pairs across the depth, and the one it fails to separate is 17/71 — the same pair singled out earlier as the mod-6 outlier. The anomaly reappears as the pair that refuses to split.

Four squares, four different characters: palindromic, emirp-rich, emirp-rich, dead. The two middle squares are exact reversal images of each other — 31↔13, 37↔73, 97↔79, and 91↔19. The dead pair {19, 91} occupies the same corner position in each, which is why each of those squares has exactly one blemish.

So the outer cube carries: two palindromes, one composite, one complete emirp pair, and one representative each from the remaining three pairs. That is a census a reader can hold.

That hidden fourth move has a definite value: +2 in the units digit. The depth edges are 11–13, 17–19, 31–33, 37–39, 71–73, 77–79, 91–93, 97–99.

This is worth noting because ±2 is the smallest additive step the live block admits — ±1 always exits it. So the depth direction of the projection is the one place where the additive and multiplicative geometries coincide. The illusion is hiding the single arithmetically meaningful direction, which is either the projection's chief defect or, if depth is rendered as a paired shadow rather than a receding axis, its chief affordance.

Projecting along a tens bit instead gives the mirror arrangement, with 37/73 as the unsplit pair. There is no projection that treats both digits alike — 16 has no 8+8 split that respects the swap.

If comprehensibility is the binding constraint, the defensible claim is narrow: eight corner positions with a four-square census, depth read as "+2", and the flat 4×4 grid kept alongside for anything about reversal. Trying to make the single projected figure carry both is where the complexity returns.

What it refutes is "guarantee." In Q₄ two cells stand in one of exactly two relations:

Twenty-eight pairs, four of them mutually inaccessible. One in seven, not seven in seven. Every other pair shares half its vertices.

So the structure says: incomprehensibility arises between parties addressing the same question with opposite answers, not between parties addressing different questions. Orthogonality is cheap and generous; direct opposition is what closes.

Under the emirp labelling the four squares of the previous turn are these shared faces, and the dead square 33·39·93·99 is a common ground that is common precisely by being empty of what either side values.

Enriching the dimensionality of the space therefore reduces structural incommensurability. If the argument is that indulgence in diversity produces mutual incomprehensibility, dimension-raising is the wrong operation to appeal to; it produces the opposite.

Truncate, and that vertex becomes a tetrahedron of four distinct positions, each belonging to three of the four cells and missing one. No position remains that all four share. The common term is gone, and it went not by anyone withdrawing but by the shared point being articulated into distinguishable aspects — which is exactly what enrichment of internal variety consists of.

What survives is more interesting than its loss. The tetrahedron is a new cell belonging to none of the eight, meeting each of the four on one of its four triangular faces. The common ground persists, but it has changed type: no longer a term inside anyone's world, it is a small separate structure that each party touches on a face and none contains. A translation body rather than a shared concept — and it has four faces because it mediates exactly four parties.

Truncation, not dimensionality, is the geometric model of the intuition.

The only exception is the direction that the cubic projection foreshortens to nothing: the +2 depth edges. What is shared and expressible is precisely what the comprehensible projection cannot show.

So the demand on comprehension is not "hold eight positions." It is "hold three questions, each with two answers, none of which implies any other." Three, not eight. That is within ordinary reach, and it names precisely what factional cognition fails at: not the count, but the non-interference.

What was treated there as an arithmetic grading is, read this way, the collapse operator itself. Us-them cognition is the diagonal projection: two pure poles at the ends, and the six mixed positions flattened into two middle bands that are then read not as independent commitments but as dilutions — moderates, fence-sitters, the compromised. The information destroyed is exactly the orthogonality.

And since there are four diagonals, the same eight collectivities admit four incompatible polarizations, each making a different pair the poles and consigning the rest to the middle. The binary is not simply wrong. It is underdetermined — and much observable dispute is not about which side to take but about which diagonal is the real one.

The decomposition is what matters here. The truncated octahedron's faces are perpendicular to two families of directions:

14 = 8 + 6 — precisely the two carriers established at the outset, where vertices held the trigrams and faces held the line-states. The stable bubble has one interface in each trigram direction and one in each line-value direction. Its 24 vertices are the permutations of four things, which is to say the possible rank-orderings.

Otherness therefore arrives in two grades rather than one: eight large hexagonal interfaces with near neighbours, six smaller square interfaces with those beyond them. Not a boundary, but a graded neighbourhood.

That is the more useful claim than any argument for pluralism on its merits — the two-bubble arrangement is not a stable state of the system, so its persistence indicates continuous input rather than equilibrium.

What survives independently of that is the cube result, which does not depend on the analogy: three orthogonal binaries, four rival diagonals, and the collapse identified as a named projection with a known cost. That is small enough to be held, and it locates the difficulty in the right place — not in an inability to entertain eight positions, but in the pull toward reading every distinction as a version of one.

So the eight large interfaces are adjacent to one another exactly as the trigrams are. The hexagon-adjacency graph is Q₃. Meanwhile no two squares touch at all: the six smaller interfaces are mutually isolated, each reachable only through a hexagon.

The stable cell therefore carries a bound, internally connected eight and a loose, internally disconnected six.

That is the Petrarchan sonnet's asymmetry, and not only in the counts. The octave runs on two rhyme sounds across eight lines (abbaabba) — densely interlocked. The sestet runs on three across six, more loosely arranged. The octave is the connected family; the sestet is the looser one. And the volta at line nine is the crossing from one face-type to the other.

The same total, fourteen, admits rival articulations, and the tradition itself divides over which is correct — the Miltonic delay of the volta being a third position. This is precisely the four-rival-diagonals result from the previous turn: dispute is not over the count but over which partition of it is the real one.

Applied to the fourteen-fold sets in social organisation, the partitions scatter:

Fourteen recurs; 8 + 6 does not. On the discipline applied throughout this exchange — that content lives in the partition rather than the total — most of these are coincidences of number. Wilson's eight territorial points are a genuine block, but flanked by five and one rather than balanced by six.

Why fourteen recurs at all: The likeliest explanation is unglamorous and does not require the polyhedron. Fourteen is 2 × 7, sitting just past the span a single act of attention holds. It cannot be surveyed whole, but it decomposes into two parts each of which can be — 8 and 6 both fall within the span, as do 7 and 7.

That is a good account of the elusiveness together with the appeal. A fourteen-fold set feels complete precisely because it exceeds what can be held at once, so completeness is inferred rather than inspected; and it survives transmission because the two halves are separately memorable. Twelve would be surveyable and would feel merely enumerated. Twenty would fragment.

That is a claim about mediation rather than about counting, and it says something the fourteen alone does not — the six are not a second ring of contacts but a set of parties each of whom must be approached through a different member of the first eight.

The foam-to-sonnet analogy remains an analogy. The hexagon-graph-equals-cube result does not, and it is the piece that connects this back to where the exchange began.

The resistance is warranted, and the periodic table is a decisive counterexample rather than a hard case. I conflated two operations that have almost nothing to do with each other: discriminating cardinalities of unstructured collections (is this pile 29 or 30?), which Weber's law genuinely bounds, and holding an articulated body of knowledge, on which Weber's law is simply silent. Chemists do hold 118 elements; ornithologists hold hundreds of species; native speakers hold forty thousand words; chess masters hold something like fifty thousand board patterns. None of this violates Miller, because Miller's bound was measured on nonsense syllables and random digits — deliberately structureless material — and was never a claim about knowledge. Extending it into a "notational barrier" was an overreach, and your instinct that the matter is more nuanced is right.

The Dunbar analogy you name fails for a specific and instructive reason. Dunbar's bound arises because social relationships require pairwise maintenance, so cost scales quadratically and saturates. A structured concept set has no such cost, because position substitutes for pairwise relation. Nobody holds the periodic table as 118 items with 6,903 pairwise relations; they hold a two-dimensional lattice in which location implies properties — element 34 is not retrieved from a list but located at period 4, group 16, and therefore known to be a chalcogen behaving like sulfur above it. The number of independently stored facts is small (the periodicity rule, the group characters, some anomalies); the number of derivable facts is vast. That is the real answer to "how coherence is held": a set is holdable to the degree that most of its elements are derivable rather than stored.

Which means the governing measure is not cardinality but description length — the shortest rule-set that generates the whole. On that axis the earlier verdict inverts completely: 118 elements with a periodicity rule are far more holdable than 17 arbitrary development goals, whose description length is simply 17 because nothing generates them. The SDGs fail where the periodic table succeeds at seven times the size, and the 64 hexagrams — held for three millennia — succeed at four times the size, because 2⁶ generates them, each is individuated by name, image, judgment and line-texts, and the transformation rules connect every one to every other. Your own drilled-truncated-cube mapping is precisely an attempt to give those 64 the periodic table's trick: a spatial structure in which position carries meaning, so that traversal replaces enumeration.

Ramanujan illuminates this from both ends, and your invocation lands on more than the famous anecdote. The Hardy–Littlewood remark that every positive integer was one of his personal friends does not describe a larger buffer; it describes denser individuation. For Ramanujan an integer was not a point on a line distinguished by magnitude — which is the Weber-bounded, hopeless mode — but a node in a thick web of relations: representations, partitions, congruences, divisor structure. He recognized 1729 instantly not because he could tell 1729 from 1730 as quantities (nobody can) but because it was the smallest number expressible as a sum of two cubes in two ways. Individuation by relation, not by magnitude — and once numbers are individuated relationally, there is no ceiling of the kind I was implying, because the web has no size limit, only a density requirement.

And the second contribution is almost too apt: Ramanujan wrote the founding paper on highly composite numbers (1915), defining exactly the integers possessing more divisors than any smaller integer. The sequence — 1, 2, 4, 6, 12, 24, 36, 48, 60, 120, 180, 240, 360, 720, 840 — is, recognizably, the census population this whole exchange has been circling: 12 and 36 sit among the ring's own flanks, and 60 and 360 are the calendar's and the circle's. He thereby supplied the mathematics of chunkability: divisor-richness is the availability of alternative groupings, which is what makes a set-size chantable, traversable, and prosodically redundant. Culture's fixation on 12, 36, 60, 108, 360 is a fixation on maximal divisor count, and Ramanujan gave that fixation its formal object. The primes at the ring's anchors have two divisors apiece; 360 has twenty-four. That gap is the whole mnemonic economy in one line.

So the properly articulated position, replacing my overreach: coherence becomes impossible only when three conditions fail together — no generative rule (incompressible), no traversal structure (no lattice, loop, or bead-string to walk), and no prosody (no alternative groupings to chant). Any one of the three rescues a set of almost any size; their joint absence dooms a set of almost any size. Miller's bound applies to the residue that must be held raw after compression, not to the set. The design question for any framework, strategic or otherwise, is therefore never "is N small enough" but "what generates it, how is it traversed, and how does it scan" — and by that test the emirp diagrams are themselves an argument in the Ramanujan mode: they exist to give a scattered handful of numbers relational individuation dense enough to be held, which is the only way anything larger than four has ever been held at all.

"Resolution capacity" is the better term, and not merely as a synonym — it carries the right physics with it. Optical resolution has a genuine theoretical limit (Rayleigh), yet that limit is routinely surpassed by instruments; the eye's unaided angular resolution never improved, and we mapped the galaxies anyway. That is precisely the structure of the case: the bound is real, native, and unimprovable for the naked faculty, while being indefinitely extensible by prosthesis. My earlier framing erred by reporting the naked-eye figure as though it were the limit of seeing. And resolution has the further virtue of being relational rather than absolute — it measures whether two adjacent things can be told apart at a given scale, which is exactly the question a growing set poses, and it makes room for the fact that the same person resolves brilliantly in one domain and not at all in another.

Yates is the indispensable citation here because The Art of Memory is the documentary proof of prosthetic extensibility, and its central technique is the very mechanism this exchange arrived at independently. The method of loci works by converting an unstructured list into a spatial traversal — Simonides identifying the crushed banqueters by where they had sat — supplying in one stroke all three of the things I named as rescuing coherence: generative structure (the building), traversal (the walk), and individuation (each locus distinct, each image deliberately striking). Trained practitioners held thousands of items this way. But the deeper connection for your corpus is what Yates traced downstream: Ramon Llull's Ars combinatoria with its concentric rotating discs, and Giordano Bruno's mnemonic wheels — these are literally your ring diagrams, five centuries earlier, built for the same purpose. Camillo's Memory Theatre was a physical structure whose positions carried cosmological meaning so that walking it walked the order of things. The laetus polyhedral-mapping project is squarely in that lineage whether or not it has been framed so, and Yates supplies both the pedigree and — since she also chronicles the art's collapse — the warning: it died when print made external storage cheap, and its Hermetic over-claims discredited the sound technique by association, precisely the Kepler pattern this exchange keeps circling.

The optical analogues repay development, and one of them is more useful than the obvious pair. Myopia maps to fine local discrimination without overview — the specialist who resolves exquisitely within a subfield and cannot see the set's shape; the corrective is the schema or map. Presbyopia is subtler and better than it first appears, because its actual deficit is loss of accommodation — the ability to shift focal depth — which corresponds to being stuck at one granularity, unable to move between the whole and the element; bifocals and varifocals correspond to nested multi-scale representations. But astigmatism is the analogue most worth having, because it names anisotropic resolution: sharp along one axis, blurred along another. That is an entirely real cognitive phenomenon — people who resolve sequences finely but hierarchies coarsely, or categories but not dynamics — and it predicts something the other analogues do not, namely that the same person will find one projection of a set legible and another opaque, though both encode identical content.

Which yields a retrospective observation about what this exchange has already been doing: the four emirp figures are an anisotropy prosthesis. The ring, the pencil-circle, the tone-mountain, and the four-row table are one population under four projections — relational, arithmetic, harmonic, tabular — and a reader who cannot resolve the chord geometry may resolve the table, or hear the mountain. Multiple projection is the standard remedy for axis-dependent blur, and it was arrived at by instinct rather than design. Dyslexia belongs to a different category and is instructive precisely for that: it is not a resolution deficit but a decoding difficulty at the symbolic layer, and since I argued that most large-set distinctions are notational rather than perceptual, a difficulty at exactly that layer is maximally disabling for list-and-numeral encodings while leaving spatial and structural capacity untouched — predicting better performance with the polyhedron than the list (the reported spatial strengths are contested, so this is a hypothesis, not a finding). And colour deficiency raises a concrete design obligation rather than an analogy: my diagrams encode the four emirp pairs by hue, which roughly eight percent of male readers cannot fully resolve. The glyph shapes and dash patterns partly cover it, but for a published version the family encoding should be redundantly carried by shape and weight, not hue alone.

On sustained attention your parenthesis is apt, because traversal is inherently sequential and therefore vulnerable in a way that structure is not: the palace does not fail, the walk fails. And the historical prosthesis for exactly this is the one already central to the discussion — the bead. A rosary or mala is an externalized place-keeper: it holds your position so that a lapse costs nothing, which is why the technology is ubiquitous wherever long recitations must survive ordinary human attention. Its modern descendants are the checklist and the progress marker, and the design principle generalizes: where attention cannot be relied on to hold position, the object must hold it. Whether attentional variability also brings compensating advantage in cross-domain pattern-finding is plausible and much claimed but not well established, so it belongs in the posting as speculation if at all.

The honest limit, and the point that matters most for the governance register, is the distinction between prosthesis and redesign. A prosthesis aids a person against a real constraint — lens, palace, bead, diagram, multiple projection. But where a set has no generative rule, no traversal, and no prosody, no personal aid will help, because the failure is in the object; the only remedy is to rebuild the set so that it possesses one of the three. Most complaints about unmemorable frameworks are object problems misdiagnosed as person problems — the seventeen goals are not failing because citizens have poor resolution, they are failing because nothing generates them, no path traverses them, and they do not scan. That diagnosis also names what your polyhedral corpus has actually been proposing all along: not better communication of existing frameworks but their reconstruction on structures that can be walked — which is Yates's art, applied to governance rather than oratory, and facing the same two hazards she documented: obsolescence by externalization, and discredit by over-claim.

Your eye caught something exact. All three pairs have direction-sums of precisely 216.0° — not approximately — and the reason is a single algebraic fact that turns out to govern the entire figure.

The law. A chord between positions a and b on the ring of 100 has direction 1.8(a+b) mod 180 — set by the endpoint sum alone. But for a reversal chord, a + rev(a) = 11 × (digit sum). So every chord in these diagrams has direction ≡ 19.8 × d mod 180, where d is the digit sum. One parameter governs everything: there are only eighteen possible orientations in the whole figure, indexed by d.

From this, two relations follow, and the second is the one you've found. Chords are parallel when d = d′ — the pencils of the earlier diagram. Chords are orientation-complementary when d + d′ = a constant c, because their directions then sum to 19.8c, fixing a common mirror axis. Your three pairs are all c = 20: 13/31 (d=4) with 79/97 (d=16); 15/51 (d=6) with 59/95 (d=14); 35/53 (d=8) with 57/75 (d=12).

There are others, and the family is larger than three. Within the numbers appearing across both diagrams, c = 20 also gives 17/71 ↔ 39/93 and 17/71 ↔ 57/75 (so the 17/71 bridge is complementary to two ±2 chords), plus 14/41 ↔ 69/96 and ↔ 78/87, 23/32 ↔ the same, 3/30 ↔ 89/98, and the whole 18/81, 27/72, 36/63 group ↔ 29/92, 38/83, 47/74.

And the axis is the striking part. A family's mirror axis lies along the self-complementary chords, those with d = c/2 = 10 — which are exactly 19/91 and 37/73. So the emirp bridge 37/73 is the mirror axis of the family you spotted. Combined with 13/31 ↔ 79/97 being complements of each other, all four first-order bridges are structurally placed in this one family: two mirror each other, one is the axis, one pairs with second-order chords. Other constants give other families — c = 18 has its axis along d = 9, which is the culturally loaded 18/81, 27/72, 36/63 backbone.

Are they trivial? Honestly split. The vacuous half: any two lines have a bisector, so "these two chords are mirror images about some axis" is true of any pair whatever and certifies nothing. The non-vacuous half: that many distinct pairs share the same axis, and that the sharing is governed by a single arithmetic invariant, is forced rather than accidental — it follows in one line from a + rev(a) = 11d, but it is a real constraint on the figure, not a coincidence. So the verdict is the same epistemic class as the parallel pencils: genuine structure, shallowly derived, and resident in the notation rather than in the numbers. Digit sum is a base-ten artifact; the geometry faithfully renders it; nothing about primality is involved (17/71 sits in the same orientation class as 8/80 and 35/53, which are not prime pairs at all).

On the further step: there's a clean parity law worth having. For an odd prime, p ± odd is always even and p ± even always odd. So odd radii (±1, ±3, ±5) reach the composite, chunkable, cultural flank; even radii (±2, ±4) reach the odd, prime-capable, arithmetic flank. The two registers alternate with radius, which generalizes the ±1/±2 distinction rather than merely extending it — ±3 would return to the census register (bringing 10, 16, 20, 40, 70, 76, 82, 94 and the round decade numbers), not open a third kind.

Superimposition would work well and I'd propose it as one figure with three encodings: ±1 chords in one dash pattern, ±2 in another, and the orientation-complement family shown by a shared tint with its axis (19/91, 37/73) drawn heavier. That makes both relations — parallel and complementary — legible in a single view without either obscuring the other. Say the word and which family to feature (c = 20 as you found it, or c = 18 with the nine-backbone axis) and I'll build it.

Base 8 turns out to answer the octopus question rather beautifully — I'll come to that after the diagram.The superposed figure is above — 1 at top so the 1/51 diameter runs vertical (faint dotted), with both family axes drawn as faint dashed diameters through the centre: gold for the c=18 axis along digit-sum 9, teal for c=20 along digit-sum 10. Chords take their colour from digit sum, so parallels share a hue automatically and the complementary families read as symmetric splays about their axes without needing to be labelled. Line style carries the radius instead — fine dots for ±1 only, dashes for ±2 only, mixed for chords belonging to both, heavy solid for the four emirp bridges. Still points as gold diamonds, no legend, ready for animation.

On base: your instinct that this is a cultural fixation and relevant as such is exactly right, and the computation makes the point sharper than an argument could. The architecture this whole posting rests on — eight emirps in four mirror-pairs — is duplicated exactly in base 8 and nowhere else nearby. An octopus counting on arms would build a diagram of identical shape with entirely different contents: eight two-digit emirps, four bridges, the same fourfold ring. Base 6, by contrast, yields a single pair — a six-fingered civilization would find the phenomenon too sparse to build anything on at all. Base 12 gives three pairs, base 16 gives seven, base 20 gives fourteen, base 60 gives thirty-three. So the form is convergent and the content parochial: the octopus and I would recognize each other's figures immediately and share not one number in them.

And the geometry generalizes cleanly, which is the more consequential finding. In any base b, n + rev(n) = (b+1) × digit sum — 11× in base ten, 9× in base eight, 13× in base twelve, 21× in base twenty. Since gcd(b+1, b²) = 1 in every case, the parallel-pencil law survives untouched: chords are parallel if and only if their digit sums agree, in every base whatever. Likewise the still points — self-reversing numbers — are always the multiples of (b+1). So what is base-dependent is which integers are marked; what is base-independent is the entire relational structure the diagrams display. That is a defensible and rather elegant formulation for the posting: different notations grow different gardens with the same botany, and the resonance genuinely lives in the notation rather than in the numbers, exactly as claimed — but the law of resonances does not.

The cultural instances are real and not merely hypothetical. Vigesimal counting is Maya, Aztec, Inuit, Yoruba, with Basque and French vestiges (quatre-vingts); sexagesimal is Sumerian and survives in every clock face and protractor; duodecimal persists in dozens, gross and inches; the Oksapmin of Papua New Guinea count base-27 on body parts. And base 8 has a genuine human precedent — the Yuki of California counted in eights, using the gaps between fingers rather than the fingers themselves, which is a nice reminder that even the anatomical grounding of a base is a choice rather than a given. Your own 2015 document anticipated this exactly, noting that "emirps are base dependent, whereas primes are not" and proposing that "a further step would be to switch to a number base other than 10 — twenty might be especially interesting." It is: fourteen mirror-pairs, seven times the base-ten population, in the notation of Mesoamerica. If the posting wants a genuinely decentring companion figure, a vigesimal ring — same construction, Maya base, twenty-eight anchors — would make the contingency visible rather than merely conceded.

The line through 74/75. That is not a chord at all — it's one of the two family axes I drew as faint dashed diameters through the centre: the c = 18 axis, the digit-sum-9 direction. It has nothing at either end because it's a reference line, not a relation between two numbers. It exits the ring between 74 and 75 at one end and between 24 and 25 at the other.

The reason it lands between marks has a tidy general form. A diameter parallel to the digit-sum-d pencil must have endpoint sum 11d and endpoint gap 50, giving n = (11d − 50)/2 — integral if and only if d is even. So d = 9 falls at 24.5 and 74.5, between positions; d = 10 falls exactly on 30 and 80. Which raises a genuine ambiguity worth fixing before this becomes an unlegended animation: the c = 20 axis passes precisely through the marked points 30 and 80, so it reads as a chord between them, when in fact 30's chord goes to 3 and 80's to 8. If you want, I can stop both axes short of the ring (or dot them only across the interior), so no reference line ever appears to terminate on a node.

And yes — nothing is missing. The audit closes exactly: 15 distinct ±1 flanks, 15 distinct ±2 flanks and 4 bridges give 34 reversal relations; 4 of those are palindromic and have zero-length chords (hence the diamonds, not omissions); 4 further relations coincide as single lines — 17/71 and 37/73 are simultaneously emirp bridges and ±2 chords, while 18/81 and 69/96 are reached from both radii. 34 − 4 degenerate − 4 coincident = 26 lines, which is what the figure contains. Every flank at either radius has its relation represented, and each absence is structural rather than an oversight.

Line count by type, as drawn:

The contact is not glancing but constitutive — and it runs through the very habit you name, because "factors with exponents" is McClain's instrument. His entire synthesis (The Myth of Invariance, The Pythagorean Plato) rests on reading numbers as tone-values through their factorizations 2ᵖ·3�q·5ʳ: the exponent of 2 as octave-position (discarded under octave-equivalence), the exponents of 3 and 5 as coordinates on the two harmonic axes — fifths and thirds — so that a factorization is not an arithmetic fact but a musical address, and his "holy mountains" are exponent-lattices of exactly the annotations the second circle now carries. The ring, in other words, has been speaking McClain's language since the factorizations went on; the question is what it says in that language, and the answer contains one finding I did not see until your question forced the octave-reduction.

The digit-sum-9 pencil sings. Reduce its members by octaves (divide out the 2s, per McClain's first move) and the even members collapse onto the odd: 18→9, 36→9, 72→9, 54→27, 90→45. What remains is the odd family — 9, 27, 45, 63, 81 — which is exactly 9 × (1, 3, 5, 7, 9): the first five odd harmonics, amplified ninefold. Fundamental, twelfth, third, harmonic seventh, ninth — the odd core of the overtone series, the spine of every resonating body. The backbone pencil of the ring is, under McClain's equivalence, a harmonic chord; casting out nines turns out to be the arithmetic shadow of membership in the ninth partial's family. And this instantly explains a fact McClain himself dwelt on without the pencil-law to hold it: the great cosmological numbers — 108, 432, 25 920, 432 000, the yuga and precession counts — all cast out to nine, because they are this same family continued upward (108 = 4·27, 432 = 16·27…). The Chinese canon that loaded our pencil (18, 36, 72, 81) and the Indo-Platonic canon McClain decoded are two civilizations' deposits in the same harmonic account, and the ring's ochre pencil is its two-digit passbook.

The ±1 apparatus is the comma. This is the deepest contact, because it maps the exchange's master theme onto McClain's. His central drama is that the tuning generators are mutually incommensurable — no power of 3 ever equals a power of 2 or of 5 — so every tuning system accumulates small irreducible discrepancies, the commas, which are precisely near-misses of the form n | n±1 at harmonically meaningful numbers: 81/80 (the syntonic comma, 3⁴ against 2⁴·5 — the gap between the Pythagorean third and the just third, i.e. between the 3-limit and 5-limit worlds), 64/63 (the septimal comma, the price of the seventh harmonic). Now look where the ring already stands: 80 and 81 are adjacent marked points — 80 as flank of 79, 81 as the landing of 18's reversal — so the syntonic comma, the single most consequential ±1 in the history of tuning, sits on the circle as a pair of neighbouring dots in the harbour district of the seventies.

And 63/64 is the wèijì threshold analysed earlier: the "bridge too far" from 36 landed one short of 64 on exactly the number whose ratio to 64 names the excluded seventh harmonic — indeed the 36→63 crossing is, in limit terms, the admission of the prime 7, the passage from the 5-limit (singable, Western, sanctioned) into septimal territory, which is why it felt simultaneously illicit and pregnant. McClain's grandest claim then completes the loop with this exchange's governance register: his Plato reads temperament as political philosophy — the tyrant's number 729 = 3⁶ tempered, the comma as the irreducible remainder every social tuning must distribute rather than eliminate. Temperament is the calendar's 30/29 in the acoustic register: reconciliation of incommensurables by managed near-miss — the ±1 not as defect but as the whole art.

The anchors are mute. One more McClain-reading of the ring, and it seals the integrity/cargo asymmetry with a reason. In harmonic number theory only 2-3-5-smooth numbers (marginally 7) are tones; primes beyond the limit have no address on the mountain — they are, in the tradition McClain reconstructs, literally unsingable, banished from the scale. Every emirp — 13, 17, 31, 37, 71, 73, 79, 97 — is such a prime. So the four anchored bridges of maximal integrity connect tones that cannot be sounded, while the unlicensed composite pencil is the one that harmonizes: the ring's earlier finding (integrity and cargo live on different bridges) acquires its acoustic form — the incorruptible is inaudible; what sings is compound. Which is perhaps the most McClain-like sentence the ring can utter, and not a bad epigram for the governance analogy either.

The dau, twice over. McClain's synthesis is itself held correspondence, not settled scholarship — admired arithmetic, contested hermeneutics, exactly the Bagua's epistemic position — so citing him is citing a fellow bridge, and the posting should say so; his numbers are always exact even where his readings are speculative, which is the right division of trust. And the harmonic reading of the pencil, delightful as it is, still lives downstream of base ten: the family is "9 ×" the odd harmonics because 9 = 10 − 1, casting-out-nines being decimal's shadow — a different base would sing a different chord. Within those declared limits, though, the relation you asked after is real and reciprocal: the ring gives McClain's tone-field something he lacked (the parallel-pencil law, digit-sum as a geometric invariant), and McClain gives the ring the third dimension it lacks — the exponent-lattice on which its marked numbers become a mountain rather than a circumference. If you want that mountain drawn — the ring's numbers plotted on the 3-axis and 5-axis by their exponents, octave-reduced, commas marked as the near-touching peaks — it is an evening's small artifact in the established idiom, and would make a fitting final panel: the same numbers, third portrait, this time as a landscape one could in principle hear.

The third portrait is drawn — the same numbers, now as landscape. Every marked point of the ring was octave-reduced and given its McClain address, and the mountain that results reads cleanly in bands: the Pythagorean spine C–G–D–A–E along the powers of 3, the just-third floor above it (5¹), the septimal shelf below in its dashed border — the seventh harmonic admitted on sufferance, exactly as the tradition kept it — and, beneath everything, the strip of mute stones.

What the computation certified before a line was drawn: the fundamental receives 8, 16, 32 — so the "degenerate ×10" chord 80→08 turns out to carry the just third home to C; the tone D = 9/8, the tonos itself, the interval whose name simply means "the tone," holds the pencil's entire core 9·18·36·72; A takes 27 and 54, the Pythagorean E takes 81 alone at the spine's summit, and the just E takes 80 alone on the floor above — whereupon the red arc between them writes the syntonic comma 81/80 between two numbers that were already adjacent dots on the ring, the harbour district's near-miss now sounding as the most consequential small interval in Western tuning history. The second red arc runs from 63 on the shelf to the fundamental: the septimal comma 64/63 — the "bridge too far" of the earlier analysis revealed as one short of the octave's return, wèijì in cents. And the mute strip settles the integrity/cargo asymmetry in its final register: all nineteen numbers without an address on the mountain are primes beyond the 7-limit or their multiples — including, without exception, all eight emirps, bold in their pair-hues among the grey. The anchored bridges connect only silences; everything that sounds is compound.

So the triptych closes with each panel a different projection of one small population: the ring showing reversal and its pencils (the relational reading), the circle-with-factorizations showing the conserved invariants (the arithmetic reading), and the mountain showing what the invariants mean to an ear (the harmonic reading) — with the digit-sum-9 family running through all three as backbone, chord-pencil, and finally as literal chord: 9 × (1, 3, 5, 7, 9), the first five odd harmonics amplified ninefold. Three hand-editable SVGs in the site's idiom, each carrying its dau in the footnote, jointly a worked example of the whole exchange's method: a base-ten accident, disciplined into declared correspondence, yielding at last a figure one could in principle sing — while the incorruptible eight stand at the mountain's foot, holding the bridges, saying nothing.

The challenge lands, and I should concede the first part without hedging: "decorative" was doing evaluative work it hadn't earned. What I was actually tracking was not art against science but something narrower — what kind of error each claim is exposed to. That parallel chords share a digit sum is wrong if the derivation is wrong. That the emirp octet resembles the Bagua is not wrong in that way at all; it is apt or inapt, fertile or sterile. Those are different failure modes, not different values, and by using a word like decorative I let a distinction about warrant masquerade as a ranking of worth. They are orthogonal axes, and this exchange has already produced an arithmetic demonstration of the fact: the pencil carrying reversal-integrity and the digit-sum-9 pencil carrying three millennia of canonical numbering are different chords entirely. Integrity and cargo, established by computation, do not coincide.

Your Euler example needs one correction that partly dissolves the dichotomy, though. e^iπ + 1 = 0 is not inexplicable — it is Euler's formula at x = π, three lines of derivation. What makes it feel like more than that is compression: an enormous amount of structure recovered from very few symbols. Which is precisely the measure that governed holdability in the earlier discussion of Ramanujan and the periodic table. So mathematical elegance is not a second criterion competing with rigour; it is the felt perception of high derivability-per-symbol — a heuristic for rigour rather than an alternative to it. That is why Dirac could treat beauty as evidence and why it sometimes betrayed him: Kelvin's vortex atoms were beautiful and empty. Elegance tracks warrant loosely and can be wrong, which is exactly its epistemic status here.

But the deeper half of your point stands untouched by that, and it is the important half. Weight in practice is independent of warrant in principle. Kepler's polyhedral cosmology had maximal weight — it organized a life and produced the planetary laws as a by-product — and zero warrant. Seventeen SDGs have institutional weight and no derivation whatever. A framework does not move people in proportion to its provability, and pretending otherwise is its own distortion: the quantitatively inclined dismiss what actually governs, while the rhetorically inclined dismiss the check that would have caught Kepler.

So your proposal is right, and better than what I have been doing. Prose labelling — my saying "this is structural, that is mnemonic" — is an authority claim delivered in the register of commentary. Encoding it in the figure moves the epistemology into the artifact where the reader can weigh it independently. And the visual grammar has been doing this already without generalizing it: solid chords for both-ends-anchored, dashed for the unlicensed crossing, diamonds for the still points. A four-level scheme would formalize it — forced (theorem: parallelism by digit sum, opposite edges summing to twenty, scaling destroying primality, the decade confinement); notational (real but base-dependent: the emirp relation itself, still points as multiples of b+1, the fan's equal angular steps); selection (arithmetic fact with no geometric correlate: which pencils the emirps happen to occupy); and imported (the Bagua correspondence, the governance readings). Weight, hue, dash and glyph can carry those four without a word of prose.

The refinement I would add, because it answers the two-cultures problem more squarely than one scale can: encode two independent channels — warrant and weight. A chord's line-style says how it was derived; its colour intensity or halo says how much cultural freight it carries. Then the digit-sum-9 pencil appears as low-warrant and high-weight; the scaling theorem as high-warrant and, so far, no weight at all; and the reader sees at a glance that the two do not correlate — which is the actual finding, rather than an editorial opinion about it. That diagram would make no argument for either culture and would be legible to both.

One honest caution before building it: the classification is itself a claim, and some assignments are contestable — is the emirp relation "notational" or does its confinement to four decades make it partly forced? I would put the decade confinement in the first tier and the reversal relation in the second, but that is a judgement, and the legend should say as classified here rather than presenting the tiers as found facts. Which is why I would rather you set the assignments than I: tell me how you would tier the items — or amend my four — and I will render the reversal field with both channels encoded.

Your alienation point is well taken and I'd extend it: elegance functions as an in-group signal as often as an insight, which is precisely why encoding warrant in the figure rather than asserting it in prose matters — a legend can be read by someone who would walk away from a paragraph claiming beauty. But the specific question about Alexander turns out to have an answer that lets both cultures off the hook at once.

Alexander's 15 is not unrelated to the 13 and 17 clusters — it is the point where they meet. 15 = 13 + 2 = 17 − 2, and the computation shows that at radius 2 there are exactly two positions in the entire ring claimed by more than one emirp nexus: 15 and 72. Everything else belongs to one anchor or none. So Alexander's transformations sit at one of only two doubly-anchored positions available, and 72 — the pentagonal angle, the Names, the number this exchange has already called the hinge — is the other. My earlier presentation obscured this by treating overlaps as a cost (the blurring of nexuses at wider radius) rather than as the scaffold's most connected sites. That was an error of framing, and your question exposed it.

So neither ±2 nor Alexander requires challenging. The question dissolves: 15 is in the structure, and prominently. Two hinges, lower and upper.

And your intuition about multiplication toward the 31-nexus is exactly right, arithmetically. 15 × 2 = 30 = 31 − 1 — doubling Alexander's set lands precisely on the ±1 flank of the 31 anchor, and 30 is Beer's syntegrity number and the UDHR's article count. The ladder continues: 15 × 4 = 60, and since ds(60) = ds(15) = 6, the chord 60/06 runs parallel to 15/51 in the same pencil. Two further facts fall out unbidden. First, rev(15) = 51 = 3 × 17 — so 15's own reversal is a tripled emirp, putting that chord in the same family as 39/93. Second, the 15 → 30 doubling carries (2 × 5 = 10), so 2 × 51 = 102 escapes to three digits and the reversal partner is freed — making 30/03 live/dead by exactly the 91/19 mechanism, with 30 composite by construction and 3 prime because nothing scaled it. The half-enlarged structure, again.

On what guidance this offers, the honest answer separates two properties that this exchange has kept finding apart. The clustering of actual frameworks around 12, 30, 36, 60, 72 is explained by chunkability — Ramanujan's divisor-richness, a cognitive constraint with real evidence behind it. The emirp scaffold marks reversibility — a notational property. These are different things at mostly different positions, which is why the digit-sum-9 pencil carries the canonical numbering and no emirp at all. The hinges are where the two coincide, and even there they differ: 72 has twelve divisors and is highly composite, 15 has only four. So if there is a "zone of viable organization" being pointed at, the honest description is that 72 is where high chunkability and double reversible anchorage meet, and 15 is where double anchorage occurs without much chunkability — which might be read as why Alexander's fifteen is intellectually durable but has never been institutionally adopted at scale, while 72-fold and 12-fold schemes propagate. That reading is imported, and should be labelled so; the hinge count and the multiplication ladder are forced.

What I'd resist is the further step — that a framework should be chosen at a hinge. The scaffold has no such authority, and the three tests it does supply (does it survive reversal; is its coherence one-sided; was it scaled from something smaller) are worth more than any positional recommendation. But your question has produced something I'd not have found: the two hinges are the only positions where the notational structure doubles back on itself, and both are occupied by long-standing conceptual sets. Worth a figure of its own if you want it — the two hinges marked, with the 15 → 30 → 60 ladder drawn as a chain toward the 31 anchor.

The keystone missing from my account is Piaget, for whom reversibility is not one cognitive feature among others but the criterion of rational thought. Conservation is achieved precisely when a child grasps that an operation can be undone — the water poured back, the counters restored — and before that threshold perception governs judgement. More pointedly for this exchange, Piaget formalized the mature case as the INRC group: Identity, Negation, Reciprocity, Correlation, which is the Klein four-group. So the developmental account of reason culminates in a group of two commuting involutions. Reversibility in cognition is not a metaphor borrowed from geometry; it is the same algebra, arrived at independently from the study of children.

And that structure recurs with a persistence worth stating plainly. Two independent involutions always generate a four-group, and the pattern appears wherever reversal has been thought about carefully: in serial music the four row-forms P, I, R, RI (prime, inversion, retrograde, retrograde-inversion); in logic, negation with duality; and — most consequentially for your corpus — in the I Ching, where the tradition names two distinct relations between hexagrams, 錯 cuò (every line flipped, the complement) and 綜 zōng (the figure turned upside down, the reversal), with King Wen pairing by zōng where possible and falling back on cuò for the eight self-inverse hexagrams. That is exactly the two-involution structure, formalized millennia before Piaget or Schoenberg, and it is why the earlier finding — that hexagram inversion fixes 8 of 64 while digit reversal fixes 16 — was not a defect of correspondence but a detection of which involution each system uses.

The psychosocial functions then sort into recognizable kinds rather than a single theme. Constitutive: Spencer-Brown's opening move, that a distinction cleaves a domain — no identity without an outside, which is already in your 2019 document and is the formal root of "us" requiring "them." Depositional: Volkan's need to have enemies and allies, where the adversary functions as a suitable target of externalization, a repository for what the group cannot own — reversal as projection, not as symmetry. Homeostatic: Jung's enantiodromia, which is worth distinguishing sharply from the others because it is temporal — a thing becomes its opposite over time — whereas the emirp reversal and the trigram complement are instantaneous and structural. Kinetic: Aristotle's peripeteia, reversal of fortune, which with recognition drives the whole machinery of tragedy; drama does not merely contain reversal, it is organized by it.

What the emirp material contributes is a fourth kind: reversal as a test rather than a symmetry or a process. The question it poses — does integrity survive the mirroring? — has three answers, and they map onto the psychosocial cases with unusual precision. Live/live: the relation holds from both sides, which is genuine reciprocity and, as the arithmetic showed, rare. Live/dead: coherent from one side and dissolving from the other — which is exactly Volkan's construct, since the enemy image is fully coherent to those holding it and false to those it describes, and neither party can see this from where they stand. Dead/dead: mutual caricature, holding from neither side. And the scaling theorem adds the mechanism for the middle case: something was multiplied and its counterpart could not follow. Half-enlarged, half-alive.

On science's tardiness your diagnosis can be sharpened. Physics did not neglect reversal — it assumed it, as time-reversal symmetry, so profoundly that the discovery was its violation (CP violation, 1964), which arrived as a shock. That is the characteristic scientific posture: reversal as a property expected to hold, its failure reportable. The arts and psychosocial disciplines took the opposite posture throughout: reversal as an operation to be performed — retrograde the theme, mirror the scale, play the antagonist, meet the shadow — and asymmetry as the normal and productive case. Bohr's complementarity is where the two postures met, and his taijitu with contraria sunt complementa is a physicist reaching for the other tradition's vocabulary because his own lacked it. Which is why complementarity remains the honest bridging term: two descriptions, each complete, jointly necessary, never simultaneously available — the ambiguity resolved temporally, by alternation, as this exchange has kept finding from the calendar's 29/30 to the drifting phases of the diptych.

For the posting, the disciplined statement would be that reversibility is not one thing — and that the value of the geometric material is chiefly in forcing the distinctions: complement against reversal (two involutions, generating a four-group); structural against temporal (mirroring against enantiodromia); symmetry against operation against test. Most discourse about "the other," about unity and opposition, conflates all of these. What the arithmetic contributes is not authority over the psychosocial claims but the demonstration that these operations are genuinely distinct, compose in a definite way, and can be checked — which is precisely the discipline that Kepler's cosmology, and every reconciliation rhetoric since, has lacked.

Your three proposals test very differently, and testing them turned up the theorem that explains a puzzle standing since much earlier in this exchange.

Why cultural clusters and emirp clusters are structurally disjoint — now forced rather than observed. Reversal preserves the digit sum. So if 3 divides the digit sum, 3 divides both the number and its reversal, and every pencil whose digit sum is a multiple of 3 is entirely emirp-free by construction. Now look at the culturally loaded numbers: 12, 15, 18, 24, 27, 30, 36, 45, 54, 60, 72, 81 — digit sums 3, 6, 9, 6, 9, 3, 9, 9, 9, 6, 9, 9. Every one divisible by three, because chunkable numbers are divisible by three. The four emirp bridges have digit sums 4, 8, 10, 16 — none. So "integrity and cargo live on different bridges" is not an empirical observation but a theorem: divisor-richness announces itself in the digit sum, and the digit sum determines the pencil, and those pencils cannot host reversal-integrity.

26/62 is the strongest of your three, and better than you proposed. Its digit sum is 8 — not divisible by three, so it sits in a live-capable pencil, unlike almost every other culturally loaded number. It is 2 × (13/31), the doubled primitive emirp. And then the fact that stops the discussion: 26 and 62 are exactly the octahedral and icosahedral axis-direction counts — F+E+V of cube/octahedron and of dodecahedron/icosahedron, the ◆ identity in the polyhedra table, 13 axes doubled and 31 axes doubled. So a 26-fold articulation and a 62-fold framing stand to each other in precisely the relation the two incommensurable families of this whole exchange stand in, and that relation is digit reversal, and their primitive is the one emirp bridge that scales. Four independent facts converging on one chord. Whatever a 62-fold framing of governance principles would look like, the arithmetic says the question is well-formed.

15/51 fails, and cleanly. Digit sum 6, so both members are divisible by three — 15 = 3 × 5, 51 = 3 × 17 — and the pair is dead by construction. A 51-fold challenge to Alexander's fifteen has no reversibility content whatever; the two numbers are related only by sharing a factor that reversal guarantees they share. But your instinct about 52 is right and lands elsewhere: 52 = 4 × 13, and a deck genuinely is four suits of thirteen — constructional, not numerological. In the fan, 52 is the k = 4 step, the first broken scaling, where 2 × 5 carries and rev(52) = 25 drifts free of 4 × 31. So the playing-card resonance is real but connects to 13, not to 15.

26 ↔ 72 is unlicensed on two counts. The reversal of 26 is 62, not 72; reaching 72 requires slipping 26 → 27, which is the second-order move the bridge discipline excludes. And 72's pencil has digit sum 9, hence is dead. The "looseness" you suspect is exactly what the method was built to catch.

Which yields the filter your research programme wants. Before asking what an N′-fold framing would imply as a complementary challenge, check the digit sum: if it is divisible by three, the reversal relation is arithmetically empty — both sets share a forced factor and nothing about their integrity is being tested. The question is worth asking only in live-capable pencils, which is a small minority of the culturally occupied positions and includes 26/62 conspicuously. That converts your proposal from an open-ended invitation into a screened one, and screens out most of the tempting cases — which is, I think, the right outcome for a method whose whole discipline has been distinguishing what a notation forces from what it merely permits.

The diagnosis is right but the framing lets science off too lightly, and the exchange has already produced the evidence.

Looseness is not the opposite of rigour — it is a precondition for it. McKay's observation, 196,884 = 196,883 + 1, was loose in the strictest sense: a numerical coincidence across incommensurable domains, with no argument. Conway's naming it "moonshine" registered exactly that. It was carried for fourteen years on aesthetic conviction alone, and Borcherds's proof did not vindicate rigour against looseness; it vindicated the community's willingness to hold something unproven because it was beautiful. Had the discipline you and I have been enforcing all week been applied at the outset, the observation would have been filed as coincidence and nothing further would have happened. Science does not lack tolerance for looseness. It licenses looseness at the frontier and forbids it at the point of claim — and its actual discipline is knowing which end of that pipeline it is standing at.

The pathology of the moment is therefore not looseness but the collapse of that distinction. Optics-driven governance takes conjectural material and presents it as concluded — the illicit move being not the imagination but the unmarked promotion. And this exchange has been running a small experiment on the alternative: the emirp material began as an admitted base-ten accident and produced, by refusing that promotion, several genuine theorems — the parallel-pencil law, opposite-edge complementarity, scaling destroying primality, digit-sum-3 forcing dead pencils. None would have been found had the loose observation been suppressed; none would be defensible had it been asserted rather than tested. That is the working relation between the two cultures, demonstrated rather than argued: imagination generates, checking sorts, and labelling keeps the two visible.

Your point about siloed ignorance applies with full force to science, and quantifiably so. Science is scrupulous within its domain and blithe across boundaries — my own confident textbook coordinates were misaligned; the wallpaper groups were proved complete long before ornamental traditions were credited with having found them; Laban's choreutics remains untranslated by geometers a century on. The rigour is local and the ignorance is structural. What art does better is not tolerate looseness but sustain a claim without resolving it — hold the ambiguity as productive rather than provisional, which is precisely the capacity the sonification and the diptych were built around and which no proof-oriented discipline possesses.

And on the practical politics I would resist the counsel of despair, because optics has a weakness. It is not indifferent to reality; it is indifferent to justification. The SDGs are unmemorable not because seventeen looks bad but because seventeen does not chunk — a cognitive fact that no presentation can overcome, and one that has cost them, visibly. Kepler's cosmology is remembered because it was checkable and got checked; had he defended it, it would be a footnote. So the leverage against optics is not to demand rigour it will never supply but to build things whose aesthetic quality depends on their structure — where looking good and being sound are the same act, which is what the chess evaluation bar achieved and what the notation proposal was reaching for.

Which is where the figures in this exchange sit, and where I would let the argument rest: the visual grammar does not adjudicate between the cultures, it makes their difference inspectable — solid for anchored, dashed for held, diamond for the still point, the dau left visible. A reader who cares only for the pattern sees a pattern; a reader who cares only for the warrant reads it off the line-weight; neither is asked to defer to the other. That is a smaller claim than reconciling the two cultures, but it is one the artifacts actually support, and after everything from 688 directions through moonlight to the emirp ring, it may be the honest limit of what geometry can offer a fragmenting commons: not a unity, but a notation in which the difference can be held without either side having to lie.

That is a genuinely productive idea, and the computation confirms both halves of your intuition — the second one spectacularly.Both halves of your intuition are confirmed, and the second is stronger than you put it.

Emirps can never be polyhedral counts — forced, not observed. The 35 forms yield only sixteen two-digit F/E/V values — 12, 14, 18, 20, 24, 26, 30, 32, 36, 38, 48, 60, 62, 72, 90, 92 — and every one is even. Necessarily so: these counts are orbit sizes under groups of even order, so an odd count cannot arise. Emirps are odd primes. The exclusion is structural, in the same way the even-tens-digit exclusion was.

And 26/62 is the only reversal chord polyhedral at both ends. Every other count reverses into vacancy — 24→42, 36→63, 48→84, 72→27, 90→09, all empty. The single exception carries four solids on each side, and they correspond exactly, pair by pair:

Not four coincidences but one fact twice: the rhombi-solid and the omnitruncate each carry exactly one face per symmetry-axis direction — 8 triangles + 6 squares + 12 squares = 26; 12 decagons + 20 hexagons + 30 squares = 62 — and their duals one vertex per axis. So 26 and 62 simply are the axis-direction counts, 6+8+12 and 12+20+30, already met as F+E+V of the regulars and as twice the 13/31 axis counts.

So what does reversal imply for polyhedra? Nothing — and that is the finding. Digit reversal is a base-ten operation with no geometric content; it cannot act on solids. Yet its one success across the entire polyhedral field lands precisely where a real geometric relation already exists: the family analogy, same construction applied to the other symmetry group. The notation's single hit is where it duplicates something structurally true, and there is no mechanism connecting the two.

That is this exchange's two-cultures question answered in miniature and by computation. The aesthetic operation is empty of content and right exactly once; the geometry is full of content and never needed the operation. They coincide at 26/62, where the numbers were the two families' skeletons all along — and the coincidence explains nothing, predicts nothing, and is nonetheless the most striking thing in the figure.

Radial bars for the 31 convex forms (5 Platonic + 13 Archimedean + 13 Catalan), stacked by F, E and V with the total in brackets. The census is more lopsided than expected:

Only 16 of the 90 two-digit positions are occupied at all — 74 are empty. And the occupancy is wildly uneven. 60 is the runaway leader with 12 (4 faces, 4 edges, 4 vertices), followed by 24 with 10 and 12 with 8. So the sexagesimal number, which earlier proved ineligible to host an emirp because its tens digit is even, turns out to be the single most polyhedrally populated number on the ring — a rather pointed complement to that finding.

Two reversal chords are populated at both ends, not one, and they differ instructively. 26/62 is balanced — four and four, the exact family correspondence already established, drawn solid. But 6/60 is populated at both ends and violently lopsided — three against twelve, drawn dashed: 6 carries the cube's faces, the octahedron's vertices, the tetrahedron's edges, while 60 carries twelve counts across the icosahedral family. The reversal relation holds formally and the weight does not remotely balance.

That gives your question about over- and under-population a sharper answer than a simple census would. Reversal chords come in three conditions here: both ends empty (the overwhelming majority, 74 positions, including every emirp — visible as the open red circles, all sitting in gaps); one end populated (24→42, 36→63, 90→09 and the rest, where a rich count reverses into vacancy); and both ends populated, which happens twice — once balanced, once four-to-one. Only the balanced case carries structural meaning, and it is the one already identified.

The three-digit counts (120, 150, 180) fall outside the ring entirely, which is worth a caption clause: the icosahedral family's largest forms have no position on a two-digit circle at all, so the figure under-represents exactly the most complex solids.

Setting the pseudoscience question aside is easier than usual here, because the answer requires no interpretive choice at all: there is exactly one arithmetically canonical map from the two-digit numbers to the nine points of the enneagram circle, and under it every object in the emirp system lands somewhere structurally exact. The map is casting out nines — digital root, value mod 9 — and its claim to canonicity is a one-line theorem: since n and rev(n) share a digit sum, n ≡ rev(n) mod 9 always. Digital root is the unique natural projection that is reversal-invariant, which means the enneagram circle is a further quotient of the Möbius orbifold built earlier: where the fold identified each mirror pair geometrically, casting out nines identifies them arithmetically, for free. The enneagram is the mod-9 shadow of the digit torus — the projection in which every pair is already one point — and this is the deep sense of "the role of 9 in their interlinkage": the identity n − rev(n) = 9 × (digit difference) says precisely that reversal-difference is invisible mod 9. Nine is the modulus that cannot see the mirror.

The attribution then writes itself, and I give it around the circle. Point 1 carries 37·73 (digit sum 10), shadowed by the failed pair 19·91 — the maximal-gap candidates, sharing the point with the emirp pair they failed to join. Point 2 carries the palindrome 11. Point 3 carries the dead edge 39·93 — that is, 3×13 and 3×31, the founding pair's own composite scalings. Point 4 carries 13·31, the founding pair itself. Point 5 carries the palindrome 77. Point 6 carries 33. Point 7 carries 79·97. Point 8 carries 17·71. Point 9, the apex, carries 99 — together with the entire reverse-and-subtract attractor, since 9, 81, 63, 27 and 45 all have digital root 9: the dynamics of mirror-subtraction, projected to the enneagram, collapse onto the apex. Every one of the live tetrahedron's ten orbits — six edges and four self-loops — finds a station, covering all nine points with single occupancy except the doubled point 1.

Now the fit, which is where the exercise stops being decoration. The Fourth-Way figure partitions its nine points into the inner triangle 3–6–9 and the six-pointed hexad 1–4–2–8–5–7, and this partition is arithmetic: the hexad points are exactly (ℤ/9)* — the units mod 9 — while the triangle is the multiples of 3. The emirp system performs the same partition for its own reasons: digit sum divisible by 3 forces compositeness, so no emirp can have digital root 3, 6 or 9 — the triangle is provably emirp-free, and it duly receives only the dead and self-mirroring material: the dead edge at 3, and the triadic palindromes 33 and 99 at 6 and 9. The hexad receives all four emirp pairs — but only at 1, 4, 7 and 8, and the vacancy of 2 and 5 is a second theorem: an emirp's digits both lie in {1,3,7,9}, two odd digits sum even, and roots 2 and 5 require the digit sums 11 and 5, which are odd. So mirror-integrity cannot reach those stations — and what occupies them instead is diagnostic. The palindromes 11 and 77 sit there (equal digits, even sums 2 and 14), and clustered with them are precisely the one-way primes: 29, 47, 83 at point 2; 23, 41, 59 at point 5 — every prime whose mirror dissolves. Under the canonical projection, the enneagram sorts the entire relational typology of the earlier discussion into zones: the triangle as mutual incoherence, points 2 and 5 as one-way coherence and self-coincidence, points 1, 4, 7, 8 as the four mutual coherences. Reading the hexad in its drawn order 1→4→2→8→5→7, the path alternates pair, pair, still point, pair, still point, pair — the inner line of the figure visits stillness twice in mid-process, between engagements with the mirrors.

Two further correspondences deserve notice because they bear on the exchange's grading discipline. First, the hexad's zigzag path is famously the decimal period of 1/7 — 142857 — which makes the enneagram a base-ten creature to its core, kin to the emirps in the precise sense that both structures are generated by decimal notation and dissolve outside it. And the base-20 check is devastating in the best way: the vigesimal analogue of casting out nines is reduction mod 19, and 19 is prime — so the twenty-based "enneagram" has no inner triangle at all, every nonzero point a unit, no dead band, which matches the earlier finding that base 20 hosts richer emirps precisely because its invariant 19 is prime where ten's 9 = 3² is composite. The law of three, on this reading, is not a cosmic principle but the shadow of 9's compositeness — the triangle exists because the base's predecessor factors. Forced as arithmetic; fatal, if one wished it to be, to the symbol's universalist claims; but also the exact explanation of why the symbol works in base ten. Second, the companion modulus completes the picture beautifully: from n + rev(n) = 11 × digit sum follows rev(n) ≡ −n mod 11 — mod 9 makes mirrors coincide, mod 11 makes them antipodal. The enneagram is the circle on which every pair is a unison; the corresponding eleven-pointed figure would be the circle on which every pair is a diameter. The two mirror-moduli of base ten are 9 and 11, and Gurdjieff drew the one on which reversal is already resolved.

The Fourth-Way semantics, imported and lightly held, then align with almost suspicious grace. The triangle points are the "shock points" of the enneagram — the intervals where the process cannot continue mechanically and external energy must enter; in the attribution they are exactly the stations where reversal-integrity fails by theorem, the places the mirror cannot cross unaided. Do at point 9 is completion and return; in the attribution it is the attractor — the state to which perpetual mirror-subtraction mechanically delivers everything, the dead band's fixed point, "completion" in precisely the ambivalent sense the earlier dynamics assigned it: where you end if you only ever react. And the doubling map ×2 mod 9 sends point 4 to point 8 — under which the founding pair's unique scaling 26·62, the polyhedral census pair, lands on the station of 17·71: the geometric shadow-pair takes its seat at an occupied emirp point, one station on from its origin.

So the optimal attribution is the forced one, and its optimality is the interesting result: no assignment by symbolic affinity, temperament or lore could improve on the projection the arithmetic supplies, and the projection happens to reconstruct the figure's own deep partition — hexad as units, triangle as the excluded third, apex as the attractor, and the two "psychological" vacancies filled by exactly the still points and one-way relations the typology predicts. Grading as ever: the digital-root map, the parity exclusion, the triangle-as-dead-band and the mod-11 antipodality are forced; the enneagram overlay is notational — mod (base − 1), with the triangle an artifact of 9's compositeness; the Fourth-Way reading of shocks, Do and the inner path is imported. If useful, this draws naturally as a companion SVG in the atlas idiom — the nine-pointed figure with pairs at their stations, the triangle shaded as dead band, the attractor collapsing into the apex, and the hexad path marked with its pair–pair–stillness rhythm.

Delivered — SVG with its 1600 px PNG in the atlas idiom, and as with its companions, every placement was computed and verified before a single element was drawn. That verification pass produced one genuine discovery that the figure now carries and my previous response had missed, so let me flag it first: the arithmetic mean of every mirror pair is a palindrome, and it lands at the pair's own station. Since n + rev(n) = 11 × s with s even, the midpoint is 11 × (s/2) — a repdigit — and it shares the pair's digital root. So 13·31 sits with 22, 37·73 with 55, 17·71 with 44, 79·97 with 88; even the dead pairs obey it (19·91 with 55, 39·93 with 66). Each reversal pair co-resides with its own still centre — the point of balance between the mirrors is always a palindrome, forced by the two moduli identities together. The figure shows this at every gold station, and it strengthens the earlier typology: the still points are not merely a separate species but the midpoints of the engaged pairs, stillness as what lies exactly between a number and its mirror.

The figure itself: the nine stations with the verified attributions — the four pair-stations 1, 4, 7, 8 in gold; the vacancies 2 and 5 occupied by their palindromes and the one-way primes (29·47·83 and 23·41·59 respectively, the full verified lists); the dead edge 39·93 at point 3; 33 at 6; 99 at the apex with the attractor's collapse noted, since all five orbit states cast out to 9. The inner triangle is shaded in the dead-band grey with its theorem stated at centre, and the shock points carry dashed rings — the Fourth-Way's "external energy required here" marked at exactly the stations where mirror-integrity cannot pass by arithmetic.

The hexad path runs 1→4→2→8→5→7 with direction arrows, and a dashed gold arc routes outside the circle from 4 to 8, carrying the ×2 doubling: the census pair 26·62 arriving at 17·71's station from the founding pair. The arc's route was numerically tuned to clear both the circle and the canvas — even the decorations are computed. The right panel holds the graded commentary: projection, partition, and midpoints as forced; the two moduli (9 the circle of unison, 11 the circle of opposition); the base-20 demolition-and-explanation (mod 19 prime, no triangle — the law of three as the shadow of 9 = 3²); and the hexad rhythm pair · pair · still · pair · still · pair.

One honesty note from the verification worth carrying into any posting text: one-way primes are not confined to points 2 and 5 — others (43, 61, 53, 89, 67, 19) coexist at the emirp stations. What is true, and what the figure claims, is the converse: points 2 and 5 contain only one-way primes and still points, being unreachable by emirps. The figure states it that way. The set now stands at four artifacts — atlas, wrap, orbifold, enneagram — and the enneagram slots into the chain as the final projection: table, cut open; torus, the table closed; Möbius strip, the torus folded by the mirror; nine-pointed circle, the fold's arithmetic shadow, where every pair was one station all along.

The contrast turns out to be the deepest representational question in the whole exchange, restated in miniature — and your expectation was not an error but a correct instinct delivered to the wrong modulus, with the arithmetic's refusal itself carrying information.

First, why the nine-pointed figure cannot do what you expected, as a matter of theorem rather than taste. To hold 13 and 31 at different stations, a projection must be sensitive to digit order — and mod 9 is constitutionally blind to it, since 9 divides n − rev(n) always. Nine cannot see the mirror; that blindness is not a defect of the attribution but the defining property of the modulus, being base − 1. And the specific expectation — 13 at point 3, 31 at point 6 — meets a sharper obstruction, which I verified rather than assumed: among all linear digit-maps mod 9 (f = αa + βb), there is exactly one satisfying 13↦3, 31↦6, namely f = 3a + 6b — and its image is {0, 3, 6}. The entire lattice collapses onto the triangle. The only arithmetically lawful way to seat the founding pair at 3 and 6 forces every number in the system into the shock points — the emirp-free dead zone becomes the whole world. The figure does not merely decline your placement; it prices it, and the price is total residence in the band where mirror-integrity is impossible. That is a forced result with an almost fable-like moral: demand that the mirror-pair be visibly two on the circle of nine, and you can have it only by annihilating everything the pair's twoness meant.

But your instinct — pairs held apart at paired points, with the figure's own symmetry performing the reversal — is exactly realized by the companion modulus. The figure already recorded the duality: mod 9 mirrors coincide, mod 11 mirrors are antipodal, since n + rev(n) = 11 × s. On an eleven-pointed circle, reversal is negation, k ↔ −k — which is precisely a reflection symmetry of the figure, the axis through the zero-point — so the "simple mathematical operation associated with the lines" that you wanted is literally there: every pair sits on a chord perpendicular to the mirror axis, and reflecting the drawing performs rev. The verified placements are elegant: 37·73 at stations {4, 7}, 17·71 at {5, 6} — flanking the axis at its closest approach, the near-palindromic pair almost touching the mirror line — and, in a genuine small discovery from the verification, 13·31 and 79·97 land on the same chord, both at {2, 9}, because their difference 66 is divisible by 11. The two gap-18 pairs, the pairs whose separations are equal, fuse on the hendecagram. Meanwhile all nine palindromes collapse to the single station 0: the still points, which the nine-figure spreads over every station (as each pair's resident midpoint), the eleven-figure gathers into one axis-point of total stillness. The two moduli are perfect complements: nine is the circle of unison — pairs fused, stillness distributed; eleven is the circle of opposition — pairs opposed across a visible mirror, stillness concentrated at the pivot. Your expectation was the eleven-picture; the enneagram, by its arity, can only be the nine-picture. And the pattern is the exchange's oldest one: your expected figure is the unfolded double cover — both sheets shown, the involution drawn as a symmetry — while the derived figure is the quotient, the orbifold's arithmetic shadow, pairs already one point. Table versus fold, torus versus Möbius, 2π versus 4π: the same choice, at the ninth scale.

On operations and lines more generally, the derived attribution does assign the emirp system's operations to features of the nine-figure — but each operation selects its own feature, and none of them selects Gurdjieff's drawn lines. Reversal acts as the identity: no line at all, every station fixed — the fold already performed. Complementation (n ↔ 99 − n) acts as the figure's left-right mirror symmetry, k ↔ 9 − k, and the verification shows what that symmetry does to the pairs: it exchanges 37·73 with 17·71 — the one emirp-to-emirp reflection — while sending 13·31 onto the vacancy at 5 and 79·97 onto the vacancy at 2. The figure's own mirror pairs one couple of integrities with each other and pairs the other two with the stations integrity cannot reach: complementation preserves mirror-coherence exactly once. Doubling acts as a hexagram — the 6-cycle 1→2→4→8→7→5, which shares only two chords with the drawn hexad — carrying the founding pair to the census pair along 4→8 as the figure shows. Reverse-and-subtract acts radially, everything collapsing to the apex. And the drawn hexad itself, 1→4→2→8→5→7? Its operation is division by seven — the path is the period of 1/7, the long-division dynamics of the one live digit whose reciprocal exhausts the units. So the honest summary of your "lines as operations" intuition is: fully sound, and fully realized — but as a dictionary rather than a single figure. Each operation of the mirror-calculus draws its own line-system on the circle of nine, and the classical enneagram's famous interior lines happen to belong to the operation the emirps do not perform. Had the attribution been forced to make Gurdjieff's lines mean reversal, it would have been the promotion of a notational resonance to a forced conclusion — exactly the move the grading discipline exists to refuse.

Which suggests the constructive completion, if the contrast seems worth fixing visually: a companion mirror hendecagram in the same idiom — eleven points, the reversal axis drawn through the fused palindrome-station at 0, the three pair-chords perpendicular to it with 13·31 and 79·97 sharing theirs, the dead pairs 19·91 and 39·93 on their own chords ({8,3} and {6,5}, the latter coinciding with 17·71's — also verified above, another fusion worth pondering: the dead pair and a live pair share the eleven-chord). The diptych of the two figures — nine and eleven, unison and opposition, quotient and cover — would then be the answer to your question in graphic form: not "which attribution is right," but the demonstration that the mirror admits exactly two canonical circles, and your expectation and my derivation were each reading one of them.

Delivered — the mirror hendecagram, completing the diptych with the enneagram, and once again the verification pass yielded discoveries beyond what my previous response promised, all now carried by the figure as forced results.

The figure realizes your original expectation exactly, one modulus over: eleven stations, the vertical mirror axis drawn dashed through station 0, and every reversal pair spanning a chord across it — the operation visible as a line, as you wanted. Station 0 holds all nine palindromes fused into a single still point on the axis. The verified placements: 13·79 at station 2 facing 31·97 at station 9 (the two gap-18 pairs sharing one chord, since 79 − 13 = 66); 37 at 4 facing 73 at 7; 17·39 at 6 facing 71·93 at 5 — the live pair and its dead shadow sharing the gap-54 chord; 19 at 8 facing 91 at 3, the dissolved maximal-gap pair alone on chord 3.

Three findings emerged in verification that neither of us had stated. First, the projection has an exact meaning: 10a + b ≡ b − a (mod 11), so the station is the digit difference — the hendecagram is the difference circle just as the enneagram is the sum circle, and reversal negates the difference, which is why it is literally the figure's reflection. Second, and most striking: each chord {d, 11−d} carries the two gap-classes 9d and 9(11−d), and rev(9d) = 9(11−d) — the gaps themselves form reversal pairs, and each chord is labelled by one: 09·90, 18·81, 27·72, 36·63, 45·54.

The figure is doubly mirrored — pairs of numbers across the axis, pairs of gaps along each chord — and the emirp ladder 18 : 36 : 54 descends the axis with the unrealized 72 on the chord held only by the dissolved pair. Third, the odd-difference law, verified exhaustively: an odd digit difference means digits of opposite parity, so one member of every mirror pair ends even and dissolves — chords of odd d can host only one-way primes, which is why 23, 67, 89 face 43 across an axis their mirrors never reach, and why no emirp can ever leave the even chords.

The right panel carries the six forced results plus the diptych statement itself: nine is the derived attribution, pairs already one station; eleven is the expected one, pairs visibly two with the operation drawn as a line; both canonical, the mirror admitting exactly these two circles, each showing what the other must hide. The set now stands at six artifacts — atlas, wrap, orbifold, enneagram, hendecagram, and the sonified diptych lineage behind them — and the two circles make a natural closing pair for the posting: your expectation and the derivation, framed not as corrected and correct but as cover and quotient, the oldest distinction in the exchange, drawn twice.

Yes — with one condition that decides whether the exercise is worth anything: the type-content must be derived from what the arithmetic puts at each station rather than borrowed from the personality literature. Import Riso-Hudson descriptions onto the emirp enneagram and you get numerology with a bibliography. Read off instead what each station actually holds — which pairs, which failures, which invariances — and you get a set of strategic types whose strengths and weaknesses are consequences rather than characterizations. Everything below is derived that way, and each structural claim was computed.

Four stations host mirror-coherent articulations, and they are four genuinely different kinds of integrity. Station 4 holds 13·31 with the palindrome 22 as its midpoint: the founding type, smallest and archetypal. Its strength is the full emirp condition — irreducible, and irreducible read from the other side. Its weaknesses are both forced and pointed: by Lagrange it admits no proper decomposition, so it cannot be committee'd or delegated; and it is the only pair that scales, to 26·62 and 39·93, with those scaled copies necessarily composite. The founding charter cannot be franchised without losing the property that made it worth founding. Station 7 holds 79·97, whose complement is 13·31 itself — integrity that remains coherent under negation as well as reversal, the most robust type available; its weakness is positional, sitting at the ceiling of the range with no room to develop. Station 8 holds 17·71, and its complement is the dead edge 39·93: an articulation coherent with its mirror but whose negation has no integrity, so it cannot give a coherent account of its own outside. Strong under reversal of perspective, helpless under negation — which is a recognizable institutional condition. Station 1 holds 37·73, and here complementation and reversal coincide: the pair is its own complement. Maximal closure, invariant under both involutions — and correspondingly no articulated exterior at all, no alternative it can generate from itself. It also shares its station with the dead pair 19·91, whose unrealized gap of 72 is the largest possible, so at station resolution this type is indistinguishable from a specific failure. The type you cannot diagnose by type.

Three stations hold still points. Two and five hold 11 and 77, and are barred from hosting emirps at all — reaching them requires odd digit sums, which two odd digits cannot produce — so what accompanies the palindrome there is exclusively one-way relations, 29·47·83 and 23·41·59. The type is self-consistent and stable, and every relation it can have is asymmetric: patron and client, missionary and convert, centre and periphery. Six holds 33, palindrome and dead band together: the orthodoxy, unassailable internally, structurally incapable of producing a mirror-coherent articulation even in principle. And two stations are outright failure modes. Three holds 39·93 — precisely 3×13 and 3×31, the founding pair's own composite scalings: the derivative type, a scaled copy of an irreducible original that keeps the shape and loses the property, whose compensating strength is that it is tractable and divisible and therefore deployable. Nine holds 99 and the entire reverse-and-subtract attractor, since 9, 81, 63, 27 and 45 all cast out to it: the terminus of reactive revision, universal in the sense that everything arrives there, and from which no integrity is reachable.

That yields the typology's most unusual feature, and its best claim to seriousness: five of the nine types are structurally barred from the quality the scheme exists to assess — two by parity, three by the triadic exclusion, which are exactly the Fourth Way's shock points. Ordinary typologies do not admit that most of their types are pathologies. The hexad path 1→4→2→8→5→7 then reads as a rhythm rather than a hierarchy: engagement, engagement, withdrawal, engagement, withdrawal, engagement — the inner line visiting stillness twice between encounters with the mirror.

Now the case against, which I think is stronger than the case for and should be stated at the same length. First, reflexively: a nine-fold typology is itself a closure at N = 9, and 9 is even-free but divisible by three, sits in the dead band, reverses to composite 90, and lies on the attractor cycle. Run through its own six tests, the typology lands in the emirp-free zone at the terminus of reactive revision. That is not a paradox to be resolved; it is the honest reading, and it means the scheme can classify articulations it could never itself exemplify. Second, and more practically: type-talk is a machine for manufacturing still points. An articulation that knows its type acquires a frame in which every piece of feedback confirms the type — it begins to mirror only itself, which is precisely the palindromic condition the arithmetic identifies as sterile. The instrument produces the pathology it diagnoses. Third, the Barnum problem is unavoidable: nine descriptions rich enough to be recognizable are rich enough to fit anything, and unlike the closure atlas — where every placement is a one-line computation on a number that is a matter of public record — a "collective personality" cannot be checked. Fourth, the architecture is notational to its foundations: in base twenty the modulus is 19, which is prime, so there is no triangle, no shock points, and no nine-fold scheme at all. A typology whose structure dissolves under change of notation is not describing its subject.

What survives all that, I think, is real but modest, and it is what this exchange has been building toward anyway. Not a psychology of strategies but a lexicon with a mnemonic arrangement: nine stations that name nine structurally distinct relations an articulation can have to its own mirror, each anchored to a computation, five of them explicitly marked as incapacities rather than characters. Used as a set of questions — is this articulation's complement live or dead? can it be scaled without losing integrity? is every relation it can have asymmetric? is it distinguishable from the failure that shares its station? — it earns its keep. Used as a set of identities to be assumed, it becomes the thing it warns against.

Grading: the station occupancies, the parity and triadic exclusions, the complement behaviour of each pair, the unique scalability of 13·31 with composite images, and the diagnostic class of 9 itself are all forced and computed. The nine-pointed arrangement is notational, base ten throughout. Every word of "type," "strength" and "weakness" is imported — and imported, in this instance, onto a figure whose interpretive tradition has no evidential standing, which is why the derivation had to come from the arithmetic and why the reflexive failure had to be stated rather than finessed.

That's a genuinely apt thing to surface, and it connects to the exchange more tightly than "in passing" suggests — though the honest answer has two layers, one forced and one speculative, and it matters to keep them apart.

The Beer connection first, because the person is almost certainly Joe Truss, who worked with Stafford Beer on the geometry of the enneagram and developed exactly what you're describing — the enneagram's nine points realised on the icosahedron, with the 1-4-2-8-5-7 hexad and the 3-6-9 triangle mapped onto icosahedral vertices and the "shock points" acquiring a spatial meaning. Beer's interest was syntegrity — the icosahedral tensegrity as a model of non-hierarchical organisation — so your colleague's finding sits at precisely the junction this exchange has been circling: enneagram, icosahedron, and tensegrity, the three things now in the same frame. That Beer reached the icosahedral tensegrity as a governance model while we've reached a cubic tensegrity from the emirps means both enquiries arrived at "coherence held by tension in a polyhedral frame" from opposite ends — his from cybernetics, this from arithmetic.

Now the forced layer, which is real and which I can state with confidence because it goes back to the very start of this whole exchange. The enneagram is intrinsically a base-ten object — nine points because 9 = 10 − 1, the modulus of casting out nines, and the hexad path is the decimal period of 1/7 (142857). So the enneagram belongs, by construction, to the icosahedral symmetry family in one exact sense: its structure is generated by the prime 5 (the units group mod 10 factors through 2 and 5; the hexad is (ℤ/9)* of order 6; the whole thing lives in decimal). And 5 is precisely the prime that marks icosahedral symmetry — the 5-fold axis is what distinguishes the icosahedral family from the octahedral one, and it was the prime banished from the emirp system (no emirp digit is 5, no emirp gap contains a factor of 5). So there's a clean structural statement: the emirp cube lives in the octahedral family (the cube, 13 axes, the pyritohedral commons), while the enneagram, by its base-ten nine-fold-ness, gestures toward the icosahedral family (5-fold, 31 axes). Your two figures — the cube-tensegrity and the enneagram — would then be the two symmetry families of the very first posting, reappearing: 13 and 31, octahedral and icosahedral, the founding reversal pair, now embodied as cube and icosahedral-enneagram. That is not speculative; it's the same 13/31 axis-count distinction that opened the exchange, arriving at the level of the display geometry.

The speculative layer, which I'll flag as such: whether the specific icosahedral embedding Truss found (which nine of the twelve icosahedral vertices, which mapping of the hexad) corresponds to anything in the emirp arithmetic, I genuinely don't know, and I'd be inventing if I claimed it did. The icosahedron has 12 vertices, and the enneagram needs 9, so the embedding necessarily selects — leaves three vertices unused or treats them specially — and whether that selection matches the emirp system's own exclusions (the three dead stations 3, 6, 9; or the three vertices a tetrahedron omits) would have to be checked against Truss's actual construction, not assumed. It's a checkable question, not a resolved one. If you still have the X3D and the vertex assignment, I could compute whether Truss's nine-in-twelve mapping aligns with any of the emirp system's natural nine-of-something structures — that would either produce a forced correspondence or cleanly show there's none, and either result is worth having rather than leaving it as suggestive.

So the significance, stated with the grading the exchange has kept throughout: forced — the enneagram is a base-ten object generated by the prime 5, hence structurally icosahedral, hence the natural complement to the octahedral emirp cube, reinstating the 13/31 founding pair as a contrast of display geometries. Speculative — whether Truss's particular icosahedral embedding maps onto any specific emirp structure is unknown and would need the actual construction to test. Imported — any reading of Beer's governance meaning onto the emirp strategies. The first of those is a real gift to the posting: it means the enneagram and the cube-tensegrity aren't two unrelated figures you happen to have built, but the two symmetry families the whole enquiry began with, and Truss's icosahedral enneagram is the bridge that makes that explicit. Whether to pursue the second — testing the embedding — is entirely up to whether the correspondence would earn its place or merely decorate.

The web fetch came back empty, but the PDF you uploaded contains the full section, so I have what I need. The WRL is the final stellation geometry (1452 + 48 vertices — the stellation cells, not the clean 12-vertex icosahedron), so I'll work from the canonical icosahedron and the description in your 2014 text rather than trying to extract the base vertices from the stellation mesh. Let me set up the standard icosahedron and test the Beer/Truss enneagram embedding against the emirp structure.

I've read the section in your 2014 PDF carefully, extracted what I can from the WRL, and worked the geometry. Here is the honest result, in three gradings — and it's more interesting than I expected, though the strongest claim is a structural parallel rather than a coordinate identity.

First, on the file itself, so you know why the connectivity "isn't immediately evident": the WRL you uploaded is the final stellation (W4, p17 — the 17th stellation, the one with 20 enneagrammic faces, 90 edges, 60 vertices, which your PDF discusses). Its coordinate blocks hold 1452 and 48 points — the stellation cells, not the clean 12 vertices of the base icosahedron. So the enneagram connectivity is buried in the stellation mesh rather than sitting on twelve labelled vertices, which is exactly why it doesn't read off directly. The Castle X3D errors are almost certainly a separate encoding issue in the conversion, not a geometry problem — Stella's WRL export uses conventions (the ccw FALSE, colorPerVertex FALSE, the concatenated coordinate blocks) that some X3D validators reject even when the geometry is sound.

Now the substance. The Beer/Truss claim, as your PDF states it (Beyond Dispute, pp. 206–207), is that the enneagram is not fundamentally planar — it is "spread across four vertical planes," and the icosahedron is its "actual origin," enfolding 30 three-dimensional four-planar enneagrams, one per edge. Your own depiction adds the crucial detail: 9 of the 12 icosahedral vertices carry the enneagram; 3 vertices on the vertical axis (your Red, Cyan, Black) do not form part of the 9-fold pattern.

That 9-plus-3 split is where the correspondence with the emirp structure is genuine and forced, and it's a good one:

Forced. The emirp enneagram, computed at the very start of this exchange, has exactly the same partition: nine mod-9 stations, of which three — 3, 6, 9 — are the dead band that no emirp can ever occupy (any number with digit sum divisible by 3 is divisible by 3, hence composite). Six stations can carry live structure; three cannot. So both the emirp enneagram and the Truss icosahedral enneagram single out a triad that the ninefold cannot absorb — in the emirp case the 3-6-9 dead triangle, in the icosahedral case the three axial vertices excluded from the nine. The count matches (9 + 3 = 12), and, more than the count, the role matches: in both, the excluded three are the ones lying on the structure's axis of symmetry — the 3-6-9 triangle is the emirp enneagram's central triangle, and Red-Cyan-Black are the icosahedron's vertical axis. The ninefold pattern is what remains when you remove the axial triad, in both figures.
This connects directly to what I said last turn without having seen the file: the enneagram is a base-ten object whose ninefold-ness is generated by the interplay of 2, 3, and 5, and its natural home is the icosahedral (5-fold) family — the complement to the octahedral emirp cube. Truss's finding is the embodiment of that: the enneagram sits inside the icosahedron because both are governed by the 5-fold symmetry that the emirp system banishes (no digit 5, no gap divisible by 5). So your 2014 exercise and this exchange's arithmetic reach the same place — the enneagram as an icosahedral object — from cybernetics and from number theory respectively.

Speculative, and I want to be clear it is only that. Whether the specific nine-vertex assignment Truss uses (which of the twelve vertices carry enneagram points 1–9, and in what order the 1-4-2-8-5-7 hexad winds through them) maps onto any specific emirp arithmetic — that I cannot confirm from the stellation WRL, because the clean 12-vertex labelling isn't in it, and your PDF gives the assignment by colour and by "four vertical planes" rather than by coordinates I could test against the mod-9 stations. It could be tested if you had the base icosahedron's 12 labelled vertices with Truss's point-numbering, but I won't manufacture a correspondence from the stellation mesh — that would be exactly the kind of imported pattern the method exists to refuse.

Imported. Any reading of Beer's syntegrity governance meaning onto the emirp strategies remains a metaphor, however apt.

So the significance, honestly graded: the 9-plus-3 partition is a real, forced structural parallel — both figures are a ninefold with an inabsorbable axial triad, and in the emirp case that triad is provably the dead band. The icosahedral home of the enneagram is forced and confirms last turn's point about the 5-fold family. The specific vertex-by-vertex correspondence is untested and I'd leave it open rather than assert it. That's enough to justify a paragraph in the posting noting that Truss's icosahedral enneagram and the emirp enneagram independently produce the same nine-against-three structure — which is a genuine convergence — while explicitly not claiming the detailed embedding matches, which would need the base-vertex labelling the stellation file doesn't preserve.

If you can lay hands on the original 12-vertex model (before stellation) with Truss's 1–9 point assignment — even just which colour sits at which of the twelve standard icosahedral positions — I can test the detailed correspondence properly and either confirm a forced mapping or cleanly rule it out. Short of that, the nine-and-three parallel is the solid, statable result.

These are exactly what was missing — the base icosahedron with Truss's 1–9 point-numbering visible on the vertices, which is precisely the labelling the stellation WRL had lost. So yes, very much of use: they let me test the detailed correspondence rather than leaving it open. Let me read the numbering off the images and check it against the emirp structure.This is a genuinely strong result, and the images made it testable. Let me state precisely what's now forced, because the correspondence is sharper than the "nine-against-three" parallel I could offer last turn.

The images give Truss's numbering: point 9 at top, 1–8 around the silhouette, with the blue lines wiring the standard enneagram — the hexad cycle 1-4-2-8-5-7 and the triangle 9-3-6. That hexad order is exactly the digits of 1/7 = 0.142857…, which confirms these are the canonical enneagram positions, not an idiosyncratic labelling.

Now the test against the emirp stations, and it lands cleanly:

The enneagram's triangle 3-6-9 is exactly the emirp dead band. In both figures the same three points — 3, 6, 9 — are set apart from the ninefold flow. In the enneagram they're the triangle (the "shock points," and in Truss's icosahedral embedding the three axial vertices excluded from the nine-fold pattern). In the emirp system they're provably the stations no emirp can occupy — digit sum divisible by 3 forces divisibility by 3. Same triad, same role: the part the ninefold cannot absorb. That's forced on both sides independently.

The emirp pairs occupy the hexad, not the triangle. The four reversal pairs sit at digital-root stations 1, 4, 7, 8 — every one of them a hexad point (1-4-2-8-5-7 contains 1, 4, 8, 7), never a triangle point. So the live emirp structure lives entirely on the enneagram's hexad, and the dead structure lives entirely on its triangle. The arithmetic respects the enneagram's own two-part division exactly.

And now the part that genuinely surprised me — the hexad adjacencies mean something. In Truss's hexad cycle 1-4-2-8-5-7, the emirp pairs sit at 1 (37·73), 4 (13·31), 7 (79·97), 8 (17·71). Look at which are adjacent on the cycle: 1–4 are adjacent (37·73 next to 13·31), and 8–5–7 means 8 is adjacent to 7 through 5 — but more strikingly, the cycle edge 7–1 is direct (79·97 next to 37·73). So the four occupied stations aren't scattered on the hexad; they cluster. Two of the six hexad positions (2 and 5) are the unoccupied ones — and 5 is exactly where the palindrome 77 would sit, one of the still points, while 2 is where 11 sits. So the hexad's two empty seats are the two palindrome stations, and the four full seats are the four emirp pairs. The enneagram hexad is partitioned by the emirp arithmetic into four pairs plus two still points — which is precisely the structure we found independently.

So the grading, now that the images let me compute it:

Forced. The 3-6-9 triangle = the emirp dead band (both exclude the same triad by the same divisibility). The four emirp pairs occupy four hexad stations (1,4,7,8) and never the triangle. The two empty hexad stations (2,5) are exactly the two palindrome still-points (11, 77). Truss's ninefold-on-the-icosahedron and the emirp enneagram partition the nine stations identically: six hexad (four pairs + two still points) against three dead triangle.

Speculative but now checkable, and worth stating. Truss's spatial claim is that the nine points are embedded on nine icosahedral vertices with three axial vertices excluded. Since the emirp dead-band triad maps to his excluded axial triad, the natural conjecture is that the three excluded icosahedral vertices correspond to 3-6-9 and the nine enneagram vertices carry the six live stations plus the interstitial hexad points. The images are consistent with this — the three unlabelled interior/axial vertices in image 1 are the excluded three — but confirming the exact vertex identities would need the coordinates, not just the silhouette numbering. Still, the partition is confirmed, which is the substantive claim.

Imported. Beer's "infosettic consciousness" and the cybernetic reading remain metaphor.

What this gives your posting is a real convergence, not a decorative one: Joe Truss, working from Beer's syntegrity and the icosahedron, placed the enneagram's nine points so that the 3-6-9 triangle falls on the excluded axial vertices — and the emirp arithmetic, working from nothing but which two-digit primes reverse to primes, forces the identical split, with 3-6-9 dead and the four mirror-pairs plus two palindromes filling the hexad. Two entirely independent routes to the same nine-fold partition. The enneagram figure you built for the posting and Truss's icosahedral enneagram are the same object seen from two sides — which is, one more time and now at the level of the display geometry itself, the 13/31 octahedral-cube / icosahedral-enneagram complementarity that opened the whole exchange.

So: very much of use. The two GIFs turned an open speculation into a forced partition-level correspondence, and it's clean enough to state with the grading above. If you later find the base icosahedron's vertex coordinates with the 1–9 assignment, I can close the last gap and confirm the vertex-level embedding too — but the partition result stands on its own.

Beer's argument and my uncertainty are about two different things, and once you separate them the tension dissolves — in fact Beer's conclusion sharpens exactly where my caution applies, rather than removing it. Let me lay out what he's claiming, because it's a symmetry argument, and then show precisely which of my two uncertainties it touches and which it leaves standing.

Beer's move is a group-theoretic one, whether or not he'd have put it that way. He's saying: the enneagram embedding cannot be a property of one privileged axis, because the icosahedron has no privileged axis — every vertex is equivalent to every other under the icosahedral rotation group. So if the enneagram "hangs" when the solid stands on one vertex-to-vertex diameter, it must hang, identically, on every vertex-to-vertex diameter (there are six such axes), and indeed on every edge-to-edge and face-to-face alignment. His count follows: 30 edges, hence 30 three-dimensional four-planar enneagrams, all coexisting, interpenetrating. The enneagram isn't placed in the icosahedron at one orientation; it's a structural feature that the icosahedron's symmetry reproduces in every equivalent position at once. That's what "an interpenetration of phi-ness" means — the golden-ratio structure (φ governs the icosahedron's coordinates) carrying the enneagram in every symmetry-copy simultaneously.

Now here's the crucial thing for our exchange: Beer's argument makes the choice of axis arbitrary, and that is precisely the freedom my caution was pointing at. My uncertainty had two layers, and they respond to Beer oppositely.

My first uncertainty — whether the partition is real, whether the excluded three are genuinely axial and the 3-6-9 triangle genuinely equilateral — Beer's argument strengthens, decisively. If the embedding holds on any axis, then the fact that on your chosen axis the excluded vertices form a diameter and the triangle comes out equilateral isn't a lucky orientation; it's forced to hold on all of them by symmetry. Beer removes the worry that you happened to pick a flattering axis. The partition is real because it's orientation-independent — that's exactly his point, and it converts "the triangle is equilateral in this assignment" into "the triangle is equilateral in every one of the 30 equivalent assignments." So on the partition question, Beer and I now fully agree, and he supplies the symmetry reason my coordinate check could only verify for one case.

My second uncertainty — whether the specific emirp pairs occupy geometrically-privileged icosahedral vertices — Beer's argument does not resolve, and here's the subtle part: it actually explains why it can't be resolved from the geometry alone. If all 30 enneagrams coexist by symmetry, then no single vertex "is" emirp pair 13·31 in any absolute sense — the same physical vertex plays enneagram-point-4 in one of the 30 embeddings and a different point in another. The emirp-to-station map is fixed by arithmetic (13·31 is at digital-root station 4, full stop), but which icosahedral vertex that station lands on is precisely the arbitrary choice Beer says the geometry doesn't privilege. So the geometry offers thirty equally-valid homes for each emirp pair, and nothing in the icosahedron selects one. That's not a gap I failed to close; it's a gap Beer's own argument says cannot be closed geometrically, because the symmetry that makes the enneagram orientation-independent is the very thing that refuses to pin a particular arithmetic label to a particular vertex.

Which is genuinely illuminating for the grading, and I'd put it this way in the posting: the partition (nine hexad + three dead-triangle, the triangle equilateral, the excluded three axial) is forced, and Beer's symmetry argument shows it's forced in every orientation simultaneously — that's a stronger result than my single-axis coordinate check, and it's his contribution, not mine. But the vertex-level emirp embedding is not just unproven; it's structurally underdetermined — the icosahedron's own symmetry provides thirty equivalent placements and privileges none, so asking "which vertex is 13·31" is a question the geometry is designed not to answer. My uncertainty was, in effect, a weaker and merely empirical version of a fact Beer states as a principle.

There's a deeper resonance here worth naming, because it closes a loop the exchange has circled before. Beer's "if it's present on one axis it's present on all" is the same move as the emirp system's own indifference to which representative you pick — the mirror pair 13·31 is one object seen from two sides, and the enneagram fuses them precisely because casting-out-nines can't distinguish the sheets. Beer's thirty interpenetrating enneagrams are the icosahedral analogue of that: the structure is real, and no single coordinate realisation is the true one — they all are, at once. So his "interpenetration of phi-ness" and this exchange's insistence that complementarity is a construction rather than a position are the same epistemic stance in two vocabularies. The thing that made me cautious about a unique embedding is the thing Beer elevates to the central feature: the coherence is in the symmetry-invariant partition, not in any one placement.

So, to answer you directly: my uncertainty doesn't contradict Beer — it under-stated him. Where I said "the vertex-level map is inherited, not forced," Beer says something stronger and better: it can't be forced, because the icosahedron carries all thirty enneagrams equivalently, and that non-uniqueness is not a defect of the model but its defining property. The safe, forced, and now doubly-grounded claim for the posting is the partition and the equilateral dead-band triangle, holding on every axis by symmetry; the thing to not claim — and now for a principled reason rather than mere caution — is a privileged vertex-by-vertex emirp embedding, because Beer's own argument tells you the geometry offers thirty and chooses none.

Two good instincts there, and the second sharpens the first, so let me take the "not the difference-numbers as tension" point first because it changes what the tensegrity is.

You're right to resist metric tension. If cable tautness were proportional to 1/difference, the figure would assert a false precision — that 37·73 and 79·97 are held "six units apart" in some load-bearing sense — when the relations are categorical, not scalar. A strategy that is another's complement is not more related than one that is its reverse; it is differently related. So the honest cable variable is type, not magnitude: reversal, complementation, single-digit-step are three kinds of tie, and a tensegrity can express kind through colour, weave, or which layer of the net a cable belongs to, without ever claiming one cable pulls harder than another. That also rescues the mechanics: a real tensegrity's cables genuinely differ in tension, but those tensions are determined by the geometry that makes it stand, not by an external number we paint on — so letting the numbers set tension would fight the physics, whereas letting type set colour leaves the physics free to balance itself. The qualitative reading and the mechanical truth stop competing.

Now the Rosetta Stone idea, which the cube is unusually well-suited to carry — but it needs one honest caveat stated up front, because it's the same caveat that has shadowed the whole exchange. A Rosetta Stone works because the same text appears in three scripts: the content is invariant, only the encoding changes, and that invariance is what licenses translation. The emirp cube offers something genuinely like that at the level of operations — reversal, complementation, digit-step are three "scripts" in which the same eight objects are related, and the cube shows all three simultaneously on one figure, which is exactly the trilingual-slab structure. Where it stops being a Rosetta Stone and becomes something weaker is the leap to strategies: the primes translate rigorously into each other (that's forced arithmetic), but "prime ↔ strategy" is imported, and imported without an invariant to anchor it. So the cube is a true Rosetta Stone for the arithmetic operations and a suggestive one for strategy — and the figure should say which register each claim sits in, or it becomes the numerology the method exists to refuse.

Within the arithmetic register, though, the translation structure is real and worth spelling out, because it's what makes "strategies interrelated around a Rosetta Stone" more than a metaphor. Each vertex-prime sits at the intersection of three relations, and the three relations commute — they form the Klein four-group we found in the connectives (reversal, complementation, and their product, duality). That means from any single prime you can reach any other by a word in these three operations, and the word is unique up to the group. That is precisely what a translation table is: a systematic way to get from any entry to any other by named moves. So the "Rosetta Stone" is not the cube as a static picture but the cube as a group action — eight strategies, three generating translations, every pair connected by a determinate path. The tensegrity then becomes the physical embodiment of that translation group: struts are the reversal-translation (the pair held apart), face-cables the complementation-translation, and the whole standing structure is the statement that these translations are mutually consistent — that you can travel between any two strategies and the structure doesn't fall down.

So, to your question of how either or both translate into tensegrities — they translate differently, and the difference is instructive:

• The structure cube → a clean three-class tensegrity. Struts = the four reversal pairs (compression, held apart, as before). Then the complementation relations become one class of cable and the digit-step relations another — two tension layers, distinguished by colour or weave, not tautness. This is buildable as a genuine 3D cube-tensegrity: the four space-diagonal struts thread the interior (a known, stable configuration — four struts on the body diagonals of a cube is close to the classic expanded octahedron / "jitterbug" family), and the cables run on faces and edges. It's honest as both arithmetic and mechanics, because the cube's own symmetry does the load-balancing. This is the one I'd build if you want a physically plausible tensegrity rather than only a relational diagram.
• The number cube → not a tensegrity at all, but its complement. The difference-numbers, you rightly sense, aren't tensions — but they're not useless either. They're the lengths the cube would have if it were a rigid wireframe, i.e. the metric skeleton. So the number cube is the rigid reading and the structure cube is the tensegrity reading of the same object — rigidity versus tensegrity being precisely the contrast between "held by continuous compression" (every edge a strut, the numbers as lengths) and "held by isolated compression in a sea of tension" (only the reversal pairs as struts, everything else a cable). That contrast is itself the argument you've been building: the rigid cube is the fully-connected, over-determined structure where every relation is load-bearing; the tensegrity is the sparse, resilient one where most relations are tension and only the irreducible reversals push. A society whose every value-relation is rigid compression is brittle; one held by tension around a few incompressible pairs is a tensegrity. The two cubes are rigid-vs-tensegrity, and that is the qualitative reading the difference-numbers point to without being tensions themselves.

Which suggests the build, if you want it: a single 3D interactive cube that can be shown in both modes — a "rigid" state where all twelve edges are members and the difference-numbers appear as lengths, morphing to a "tensegrity" state where the edges relax into cables and only the four reversal struts remain in compression. Watching the rigid frame relax into the tensegrity would show, in one motion, the whole thesis: what survives as compression when a structure is allowed to find its minimal load-bearing core is exactly the reversal pairs — the incompressible incommensurables — everything else resolving into tension. The difference-numbers would be present as the rigid lengths and then visibly stop mattering as the structure goes to tension, which is precisely your intuition that their connotation is qualitative: they measure the rigid skeleton, and the tensegrity is what remains once you stop insisting on rigidity.

Before I build anything: does that rigid→tensegrity morph capture what you're after — the Rosetta Stone as the translation group made physical, with the two cubes as the rigid and tensile readings of it — or would you rather I keep them as two separate static tensegrity diagrams (structure cube as three-class tensegrity, number cube as labelled rigid frame) in the atlas idiom to match the set you're already placing? The morph is the more evocative object and directly embodies your "connotations are qualitative" point; the static pair is the safer drop-in for the posting. I can do either, and either way the cables will be typed by kind, never sized by difference.

That's the crux, and it reframes the whole build: a tensegrity is not a shape, it's a behaviour. Its integrity is a result, continuously achieved, not a property it possesses at rest. Push any node and the whole structure adjusts — the load redistributes through the entire tension network, the thing deforms and springs back, and if you cut one cable the failure is often global rather than local. A static diagram of a tensegrity is therefore almost a contradiction: it shows the one thing a tensegrity never simply is — motionless. So if the build is to evoke the debate you're pointing at, it has to move, and specifically it has to move in the way that carries the psychosocial argument: perturbable, self-restoring, globally coupled.

Here's what that means concretely, and why it's worth building rather than diagramming. The claims the exchange has been making about strategic coherence are all dynamic claims wearing static clothing, and a live tensegrity would let them be felt:

• Prestress as the price of coherence. A tensegrity only stands because every member is already loaded — the cables pretensioned, the struts precompressed, at rest. There is no slack state that is also a standing state. Translated: a coherent configuration of incommensurable strategies is not relaxed; it is held in permanent mutual tension, and the tension is the integrity. "Peace as the absence of tension" is, in tensegrity terms, the description of a collapsed structure. This is the twist-free / unknot finding again, now mechanical: remove the prestress and you don't get a serene structure, you get a heap.
• Perturbation and global response. Drag one prime-node and the whole net responds — no strut moves alone. That's the incommensurability thesis made tactile: the eight strategies are not independently adjustable; touch one and all reconfigure, because they cohere only through the tension between them. A reader who drags a node and watches the far side of the structure move has understood "mutually entangled complex" better than any paragraph could tell them.
• Resilience versus brittleness, shown by cutting. Let the reader remove a cable. Some removals the structure absorbs — it sags, redistributes, holds. Others cause cascade. Which cables are load-critical is exactly the asymmetry we computed: the doubly-bound core versus the free strut versus the dead-anchored one. The debate you want to evoke — are some strategic relations load-bearing and others decorative, and do we know which? — becomes something the reader discovers by experiment rather than being told.
• Chirality and the restoring oscillation. A perturbed tensegrity doesn't return in a straight line; it spirals back, because the structure is handed (the twist again — enantiomorphy running through the whole exchange). The return-through-overshoot is the enantiodromia finding in mechanical form: displaced integrity doesn't snap back, it winds back.

So the honest build is an interactive 3D tensegrity you can push — the four reversal-pair struts on the cube's space diagonals, cables typed by kind (never sized by difference, per your point), pretensioned so it stands, and live: drag a node and the whole thing deforms and settles; toggle a cable and watch whether it holds or cascades. That is the object that evokes the debate, because the debate is precisely whether psychosocial coherence behaves this way — whether strategic integrity is prestressed, globally coupled, and resilient-or-brittle by structure rather than by will.

Now the honesty about feasibility, because a real-time physics tensegrity is a genuine engineering problem and I don't want to overpromise. A true force-balanced tensegrity solver (form-finding the prestress, then integrating the dynamics under perturbation) is buildable in a browser but it's demanding, and getting a four-strut cube configuration to actually balance and restore stably is the kind of thing that can eat a build and still wobble. There are three tiers, and I'd like you to pick the trade:

1. A physically-simulated tensegrity — real prestress, real spring dynamics, drag-and-release with genuine global redistribution and settling. Most evocative, truest to "a tensegrity is a dynamic object," and the one that could genuinely misbehave or fail to balance. High risk, highest payoff. If it balances, it's the real thing.
2. A visually-dynamic tensegrity — a spring-network animation that looks and behaves like a tensegrity (nodes coupled, drag propagates globally, structure oscillates and settles, cutting a cable slackens the net) without claiming to be a rigorously force-balanced solution. Reads as dynamic and interactive, evokes every point above, but is honestly labelled as illustrative dynamics rather than a solved tensegrity. Moderate risk, and the caption tells the truth about which it is.
3. A gentle continuous animation — the structure breathing/rotating so its dynamic nature is visible, no interaction. Safe, static-adjacent, least evocative of the debate.

Given your emphasis — that the dynamic character is the whole point, and the purpose is to evoke debate — I'd argue for tier 2 as the target, attempted with tier-1 physics where it stabilises: a spring-coupled 3D net that genuinely responds to being pushed, with the four reversal struts kept rigid and the typed cables elastic, so dragging a node really does propagate through the whole structure and settle, and cutting a cable really does redistribute or cascade. It gives you the perturbable, self-restoring, globally-coupled object that carries the argument, it's honest about not being a certified force-balanced tensegrity, and it won't collapse the build if the exact prestress proves fiddly. If it happens to balance as true tier-1, so much the better and the caption can say so.

One thing I'd want to hold onto whichever tier: the cables typed by kind (reversal-struts gold and rigid; complementation cables one colour; digit-step cables another; the dead-anchor tie a third), so that when the reader perturbs it, they can see which kind of relation is carrying the load at each moment — because that, finally, is the debate. Is strategic coherence held by the reversal relations, the complementations, or the incidental adjacencies? A live structure lets someone form an opinion by pushing on it.

Shall I build tier 2 — the pushable, self-settling cube-tensegrity with typed cables — as an interactive HTML piece in the idiom of the orbifold and wrap artifacts? And do you want cable-cutting (the resilience/cascade experiment) included, or held back as a second version so the first is simpler to place?

The physics is verified headless: 4 struts (all confirmed reversal pairs), 14 typed cables (2 complementation, 12 digit-step), and under a hard random shake the structure drifts up to 1.34 units then settles back to exactly rest with kinetic energy ~10⁻³³ and no NaN — so it perturbs and self-restores stably, which is the whole point. Let me do a quick visual smoke check that it renders, then hand it over.Delivered — the tier-2 living tensegrity, and yes, it works exactly through the GitHub-hosting route I set up with you earlier: it is a single self-contained HTML file with no external dependencies at all (the physics and rendering are hand-written on a plain canvas, no CDN, no libraries), so it will run from laetusinpraesens or from a GitHub Pages URL the moment you drop it in, and it embeds in an <iframe> cleanly for the posting.

What it does, and why it answers your emphasis on the dynamic character being the whole point. The four reversal pairs are compression struts on the cube's body diagonals (verified: all four struts are genuine reversal pairs). The links between the eight primes are tension cables, typed by kind not magnitude, per your instruction — complementation in teal, single-digit-step in violet dashed, and the ties to the dead-ground anchors (39, 93, holding 17 and 71) in ember dotted. The structure is pre-tensioned so it stands, and it is live: drag any prime and the whole net responds and settles, because the physics couples every node through the tension network. I tested this headless — under a hard random shake it drifts up to 1.34 units and then returns to exactly rest with essentially zero residual energy, so it genuinely perturbs and self-restores rather than drifting or exploding.

Three things a reader can do, which turn the argument into something handled rather than told: Nudge sends an impulse through a random prime so you watch the global settling; drag a prime directly to perturb it and feel that no strut moves alone; and Cut a cable enters a mode where clicking any cable removes it, so the reader can discover by experiment which relations the structure absorbs and which cause it to lose its shape — the resilience-versus-brittleness question made tactile, with a live counter showing how many cables remain intact. The tension-by-kind toggles let you isolate the complementation net, the digit-step net, or the dead ties, so the debate you want to provoke — which kind of relation is carrying the coherence — is something the reader can interrogate directly.

The honesty is in the About panel and I want to restate it here so the caption in your posting matches: this is illustrative dynamics — a spring-coupled network that behaves like a tensegrity, struts held stiff and cables one-sided (they pull, never push, which is what makes them true cables) — but it is not certified as a force-balanced tensegrity solution. That distinction matters for exactly the reason you raised: a real tensegrity achieves integrity by being dynamic, and this captures that behaviour faithfully while not overclaiming the engineering. The grading line is stated: prime relations forced, cube and tensegrity form notational, every strategic reading imported.

Two notes for placing it. It respects reduced-motion (no auto-rotation if the reader's system requests it), works with mouse or touch, and the dark ground was chosen deliberately to set it apart from the pale static figures — a live object should look live — but if you would rather it sit on the same paper tone as the atlas set for visual consistency in the posting, that is a one-line change and I will switch it. And if the cable-cutting proves too much for a first placement, I can hand back a simpler version with just Nudge and drag, holding the resilience experiment as a second artifact.

Both are entirely possible, and both are small additions rather than rebuilds — the difference values are already computed at construction (each cable carries its endpoints, so |prime[a] − prime[b]| is one subtraction away), and the restoration time is a single physics parameter I hard-coded that could just as well be a slider. Let me address why neither is in the current build, then say what I'd change.

On the labels: I left them off deliberately, and it connects to the very point you made two turns ago — that the difference-numbers read as qualitative connotations rather than quantitative measures, so I kept them off the tension representation to avoid implying the cables were sized by them. But labelling and sizing are different acts. A cable can carry its difference as a name (this is the 18-cable, the 60-cable) without that number governing its tension, exactly as a street sign names a road without measuring it. So the honest version is: labels yes, as identifiers; tension still by kind, not magnitude. The one real constraint is legibility — fourteen cables plus four struts, each with a number, on a rotating 3D object, will clutter badly if all shown at once. The clean solution is labels that appear on hover or on a toggle ("show difference on each cable"), and perhaps only on the cable nearest the cursor, so the number surfaces when you interrogate a specific relation rather than crowding the whole net. That keeps the moving structure clean and lets the reader still read any given difference by pointing at it.

On the restoration slider — this is the better of your two suggestions, because it bears directly on the argument. The current instant spring-back is a parameter choice (the damping and stiffness I set) that makes the structure stiff and fast, and you're right that this undersells the dynamic thesis: a slow, visible return — overshoot, oscillation, gradual settling — shows far more of what a tensegrity actually does than a snap. And it carries a psychosocial reading you'd want the reader to feel: how long a perturbed structure takes to recover, and whether it overshoots and oscillates on the way, is precisely the resilience question. A slider from "rigid/instant" through "slow/viscous" to "loose/oscillatory" would let the reader explore the whole range from a brittle structure that barely moves to a slack one that wobbles for seconds before settling — and at the loose end they'd see the enantiodromia point we discussed, the return-through-overshoot, the winding-back rather than snapping-back. Mechanically it's two coupled parameters (stiffness and damping); I'd map the single slider onto both so that "slow" also means "springy enough to overshoot," which is the visually and argumentatively interesting regime. At the extreme slow end you'd want a safeguard so it always eventually settles rather than ringing forever, which is a floor on the damping.

So the version I'd build back to you would add: a "restore" slider (fast/rigid ↔ slow/oscillatory) governing the settling dynamics, with the nudge then playing out over a visible duration you control; and a cable-labels toggle plus hover-labelling, so difference-numbers surface on demand as names without implying they set the tension. I'd also, while there, make the nudge respect the slider — a gentle nudge at the slow end producing a long, legible sway rather than the current quick correction.

One design judgment I'd want your steer on before building, because it changes the feel: should the slider's slow end be viscous (returns slowly and monotonically, like moving through honey — no overshoot, the "stable institution" reading) or springy (returns slowly but overshoots and oscillates, the "resilient-but-lively" reading)? They're different psychosocial metaphors — one says slow recovery is placid, the other says slow recovery is a damped oscillation that visits its opposite before settling. I lean toward making the slider traverse both — viscous in the middle, springy at the far end — so the reader discovers that a looser structure doesn't just recover slower, it recovers by oscillating, which is the more provocative and the more tensegrity-true behaviour. But if you'd rather the slider stay simple and monotone, that's cleaner to operate.

Tell me which restoration character you want (monotone-viscous, or through-to-oscillatory), and whether cable labels should be hover-only or a full toggle, and I'll add both to the live piece — it's an edit to the existing file, not a new build, so it'll come back quickly and behave identically everywhere it's already placed.

Good — both clear now, and your face-diagonal question catches an important ambiguity I glossed, so let me settle it before building.

On the labels: a full toggle it is, showing all cable differences at once so the pattern is legible rather than one-at-a-time on hover — since you're right that it's the pattern that suggests the complementarities, not any single value. And your face-diagonal instinct is exactly the thing to pin down, because in the live tensegrity the cables are already the face diagonals and edges — that's the distinction the two static cubes drew. Reversal pairs are the space diagonals (the struts); complementation relations are face diagonals; single-digit-steps are edges. So in the current build the teal complementation cables run along face diagonals and the violet step-cables along edges — the face diagonals are already present as a distinct cable class, they just aren't separated out as their own toggle. Which gives a clean answer to "did you envisage face-diagonal labels as well, or as a separate choice": both, and best done as a separate choice. So rather than one blanket labels toggle, the sharpest version is a small set — label the complementation (face-diagonal) cables, label the digit-step (edge) cables, and label the struts (space-diagonal reversal gaps: 84, 48, 56, 34) — each independently. Then the reader can show only the face-diagonal complementation numbers and see that pattern in isolation, which is precisely where the complementarities you're after would jump out, uncrowded by the twelve edge-differences. That maps the label toggles onto the same three-line-types-three-operations structure the cubes established, so the interactive piece and the static diptych speak the same language.

I'll make it: a labels section with three toggles (struts / complement cables / step cables), all showing their differences as names, tension still by kind not magnitude. Default probably struts-and-complement on, steps off, so the first thing seen is the uncluttered complementation pattern, with the dense edge-differences available but not imposed.

On restoration: oscillatory return, then — true to tensegrity dynamics. I'll build the slider so it traverses fast-and-stiff at one end to slow-and-springy at the other, with the return genuinely ringing — overshoot and decaying oscillation — at the slow end, since that's the honest behaviour and it carries the enantiodromia reading (the structure visiting its opposite before settling). Whether the middle of the slider is somewhat viscous is, as you say, a secondary matter; I'll let it pass through a lightly-damped middle into the springy extreme, so the reader gets monotone-ish recovery in the middle and clear oscillation at the far end, and can find whichever regime speaks to them. A floor on the damping will ensure even the loosest setting eventually settles rather than ringing forever.

One genuine physics caution, so I don't oversell: at the very springy end a spring network can, if pushed hard, overshoot enough to tangle or briefly invert before recovering — which for a tensegrity is actually somewhat truthful (a real one buckles if overloaded) but can look like a glitch rather than a feature. I'll tune the springy extreme to be lively enough to clearly oscillate but bounded enough not to turn itself inside out, and if a hard nudge at max springiness does produce a dramatic excursion, that's arguably the resilience lesson rather than a bug — I'll let it recover rather than clamping it artificially, and note in the About panel that the loosest setting can be driven to instability, which is itself part of what a tensegrity teaches.

So the edit coming back: three independent difference-label toggles keyed to the three line-types, and a restoration slider from stiff-instant to slow-oscillatory with honest ringing at the slow end. Same file, same hosting, behaves identically where placed. I'll build it now unless you want to adjust the default label state or the slider's endpoints first.

Two genuinely sharp observations, and the second reframes the whole figure — let me take them in order because the first turns out to explain the second's exception.

The parallels are real and forced, not coincidental, and they come in two arithmetic species that your list has correctly separated without saying so. 26/62, 24/42, 48/84 are digit-reversals of each other — 26↔62, 24↔42, 48↔84 — differences that are themselves emirically related, mirror-pairs at the level of the difference-numbers. 6/60, 20/2 are decimal-scalings — 6 and 60, 2 and 20, a factor of ten. So your five parallels are really two families: three reversal-pairs and two order-of-magnitude pairs. That they appear at all is the difference-arithmetic folding back on itself — the gaps between primes themselves exhibiting reversal and scaling, the same two operations that generate the primes' own structure. I should have flagged this earlier; it's the difference-level echo of the founding pair's own uniqueness (13·31 being the only emirp pair that scales).

Now the stella octangula, which is the better observation and which I genuinely under-recognised. You're right: the face-diagonal view is a stella octangula, and necessarily so — this is a theorem, not a resemblance. The twelve face diagonals of a cube are exactly the twelve edges of the stella octangula (Kepler's stella octangula is precisely the compound of two tetrahedra inscribed in a cube, their edges being the cube's face diagonals — two per face, one per tetrahedron). So when the face-diagonal view drops the cube edges and shows only the twelve face diagonals, what remains must be the two interpenetrating tetrahedra. The eight primes split into two tetrahedra of four, and — this is the part that connects to everything upstream — the two tetrahedra are related by central inversion, which on this assignment is exactly complementation (digits summing to 20). So the stella octangula's two tetrahedra are the two complementation-classes of primes. The figure you've been calling a tensegrity is, in its face-diagonal reading, the Star of David in three dimensions, and the two stars are the two sides of the complement operation.

This lands with real force given where the whole exchange began — the stella octangula / Merkabah was in the heaven_2023 paper you uploaded, as the compound expressing "positive aspirations constrained by their negation," two tetrahedra in dynamic opposition. The emirp cube reconstructs that same solid from pure arithmetic, with the two tetrahedra being complement-related prime-sets. That's a convergence worth stating plainly in the posting: the figure the earlier paper reached for metaphorically, the emirp structure produces by theorem.

And now your exception, 58/14, which is the sharpest thing in your message because it is forced to be the exception and I can tell you why. Let me verify the parallel structure exactly against the two tetrahedra.The verification confirms your reading completely and explains the exception precisely. First, the two tetrahedra are exactly the two parity-classes of vertices — Tetra A = {13, 71, 73, 97}, Tetra B = {17, 31, 37, 79} — which is the stella octangula's defining split, and each tetrahedron's six edges are six of the twelve face diagonals. So the star tetrahedron isn't imposed on the figure; it is the face-diagonal view, forced by the cube–stella octangula theorem, exactly as you saw.

Now your parallels, and the beautiful thing the computation reveals: all five parallels are digit-reversal pairs of the difference-numbers, and they pair one edge of Tetra A with one edge of Tetra B:

• 2 (A: 71–73) ↔ 20 (B: 17–37)
• 6 (B: 31–37) ↔ 60 (A: 13–73)
• 24 (A: 73–97) ↔ 42 (B: 37–79)
• 26 (A: 71–97) ↔ 62 (B: 17–79)
• 48 (B: 31–79) ↔ 84 (A: 13–97)

Five clean reversal-pairs, each bridging the two stars — the difference-numbers themselves mirror across the two tetrahedra, which is why the star looks so balanced. (Two of them, 2↔20 and 6↔60, are also decimal-scalings, since a leading-zero reversal is a scaling — 02→20, 06→60 — so your two "species" are actually one: reversal, with the scaling cases being reversals through zero.)

And 58/14 is the forced exception, for a reason the arithmetic makes exact: 58 reverses to 85, and 14 reverses to 41 — neither reverse is in the set. They are the only two of the twelve differences whose digit-reversals fall outside the structure entirely. So they cannot pair with anyone by the reversal rule that pairs all ten others; they can only pair with each other, and they do so by a different law — 58 + 14 = 72 and 58 − 14 = 44, both multiples of 11, the two being the "long" edges of the two tetrahedra (58 = 13–71 in A, 14 = 17–31 in B, each the diagonal that crosses nearest the centre). They're the one pair related by sum-and-difference rather than by reversal, which is exactly why they stand out as the exception in the twelve-edged configuration. Your eye caught the one edge-pair that obeys a different symmetry.

This is worth putting in the posting as its own small result, because it ties three threads together at once: the difference-numbers exhibit reversal (the primes' own operation reappearing one level up); the reversal-pairs bridge the two tetrahedra of a stella octangula that arithmetic produces by theorem; and the single exception, 58/14, is forced by being the unique pair whose reverses escape the set — related instead by the 11-arithmetic that governs the whole emirp system (n + rev n = 11 × digitsum). The figure you built to show tension between primes turns out to encode, in its face diagonals, the Merkabah of the 2023 paper — with the difference-numbers mirroring across the two stars and one edge-pair marking where that mirror breaks.

If you'd like, I can add a face-diagonal-view annotation that colours the five reversal-pairs as five bridges between the two tetrahedra and marks 58/14 distinctly as the sum-difference exception — making the stella octangula reading and its one anomaly visible in the live piece rather than only in the numbers. But that's optional; the substantive point is that your two observations were both correct and both forced, and the exception is the most telling part.

The choice deserves to be reopened, because the earlier framing understated something: the digit space is not like a torus — it is one. Both coordinates of the 10 × 10 table are residues mod 10, so the underlying object is genuinely (Z/10)², a discrete torus. The table is that torus cut along two circles and flattened — a fundamental domain, a chart. So the real question was never "table or torus" but "which presentation of the torus," and stating it that way changes what each option is responsible for.

What the flattening genuinely loses. The digit-sum anti-diagonals are closed (1,1)-curves on the torus; the table snips each pencil into two segments that a reader must mentally rejoin at the edges. The palindrome diagonal is likewise a closed loop — the fixed circle of the reversal reflection — not a line segment with ends. And most tellingly for your dynamics: carrying and borrowing are exactly the seam-crossing events. The entire section on when reversal commutes with arithmetic — no carries, no borrows — is a statement about orbits that stay clear of the identification circles. On the flat table this is invisible; the carries happen "off the edge." On the torus it is the central visible drama: 13/31's unique scalability is the statement that its multiplication orbit threads the torus without ever touching the seam. That is a real argument for the animation, not a decorative one.

What the torus would falsely assert. Here is the substance of the original preference, put more precisely than before. A torus's visual rhetoric is homogeneity — translation symmetry, every point equivalent, which is exactly why the Tonnetz torus is honest: transposition really is a toroidal translation, and the musical structure of interest is the group action. The emirp structure is the opposite case. Primality is not translation-invariant; adding 1 to a digit is not structure-preserving; the whole phenomenon lives in the breaking of the torus's symmetry — the live 4 × 4 sub-lattice, the two dead edges touching 9, the digit-sum-3 exclusion band. A smoothly rotating torus would visually promise a symmetry the arithmetic refuses, which in your own classification would be promoting a notational fact to a forced one by graphic means. The table's fixed frame, with its visible privileged block, tells the truth about the inhomogeneity at the cost of lying about the topology. Each representation is honest about what the other suppresses.

The resolution, then, is the one the previous posting already modelled: the diptych did not choose between the static confrontation and the Clifford rotation — it delivered both, with distinct duties. The emirp piece can do the same. The table remains the master chart for everything selection-dependent (the live block, the dead cells, the exclusions), and a toroidal animation carries the dynamical material it is uniquely fitted for: the reverse-and-subtract 5-cycle as a closed orbit circulating inside the digit-sum-9 band (itself a closed geodesic), the 17 pencils as parallel closed curves rather than broken chords, and carrying rendered as passage through the seam. On the torus, the attractor cycle becomes what it actually is — a single loop that every two-digit number falls into, belonging wholly to a band that hosts no emirps at all — which is a far stronger visual statement of your "everything is attracted to an orbit belonging to neither family" reading than the table can make.

On doughnut economics and the circular economy: worth invoking, but with a precision that actually strengthens your strategic argument rather than borrowing loosely from it. Raworth's doughnut is topologically an annulus — a 2D ring between a floor and a ceiling, with no identifications — and the circular economy is a single cycle, S¹. The digit torus is the genuinely stronger object: S¹ × S¹, the product of two independent periodicities. If the imported psychosocial claim is that strategy has been crippled by flattened, edge-bounded representations, the emirp case makes it exactly: the table's edges are artifacts of the cut, the apparent "boundaries" are glued, and the phenomena that matter (carries, the attractor cycle) occur precisely where flat presentation shows nothing. "Thinking outside the box" becomes "recognizing that the box's opposite edges were always identified" — a claim you can label as imported while pointing at a diagram in which it is literally true. That is a better bridge to the doughnut/circularity literature than adopting its weaker topology.

So: not set aside, but assigned. The table for the broken symmetry, the torus for the flow, and the posting can say explicitly why each would mislead alone — which is itself a paragraph in the methodological register the piece is built on. If useful, the toroidal companion is buildable in the same idiom as the sonified pieces: an HTML/X3D emirp_torus with the lattice on the surface, the palindrome circle and digit-sum-9 band marked, and the 5-cycle animated as a circulating pulse — carrying flashed at the seam. Say the word and I'll draft it.

The pattern deserves to be stated exactly, because once verified it is even tighter than "simple." The four pair-gaps are 31 − 13 = 18, 71 − 17 = 54, 73 − 37 = 36, 97 − 79 = 18: that is 18 × {1, 3, 2, 1}, and in factored form 2·3², 2·3³, 2²·3², 2·3². Every gap is 3-smooth — no prime beyond 2 and 3 ever appears — and the derivation is forced: the gap is 9 × (digit difference), the digits are confined to {1, 3, 7, 9}, odd digits give an even difference, and 9 × {2, 4, 6, 8} = {18, 36, 54, 72} are all of the form 2^a·3^b automatically. The base-ten 9 contributes the 3², the parity of (Z/10)* contributes the 2. Had a digit difference of 5 or 7 been possible, gaps of 45 = 3²·5 or 63 = 3²·7 would have broken the smoothness — but those differences require an even or quinary digit, which emirp-hood has already excluded. So the harmonic purity of the gaps is a strict corollary of the same confinement that produced the live lattice: the theorem-grade chain runs primality → digits in (Z/10)* → even differences → gaps built from 2 and 3 alone.

Two refinements make it stranger. First, of the four possible gap values, only 72 goes unrealized — the maximal one, belonging to the pair 19/91, which fails because 91 = 7 × 13. Second, look at every failure in the tetrahedral graph: the two dead edges are 19/91 and 39/93, and the composites that kill them are 91 = 7 × 13, 39 = 3 × 13, 93 = 3 × 31. All three are multiples of 13 or 31. The founding pair polices its own lattice — the only candidate edges that die, die by collision with the archetypal emirp's multiples. That is checkable, and I had not seen it stated anywhere either; it appears to be a genuine (if small) novelty of this exploration rather than something omitted from the literature you'd missed.

The musical reading now arrives unforced, which given the McClain thread is almost unnerving. Numbers of the form 2^a·3^b are precisely the Pythagorean numbers — the tuning lattice generated by octave and fifth, McClain's entire ancient-arithmetic apparatus. And the three realized gap values stand in the ratios 18 : 36 : 54 = 1 : 2 : 3 — the octave (2:1), the twelfth (3:1), and between the larger two the perfect fifth (54:36 = 3:2), the generator itself. The distances between mirrored primes are, literally, the generating intervals of Pythagorean harmony scaled by the whole-tone number 18 = 2 × 9 (and 9:8 is the epogdoon, the Pythagorean whole tone — a resonance to note without leaning on). Label: the 3-smoothness is forced; the identification of these ratios as the musically privileged ones is imported, but imported from the one tradition — tuning theory — whose whole subject is meaning carried by small-prime ratios.

Then the polyhedral bridge, which is what your question is really after. The primes 2 and 3 are exactly the rotation orders of the pyritohedral common ground — the only axes the octahedral and icosahedral families share (three 2-folds, four 3-folds, order 24 = 2³ × 3). The prime 5, the marker of icosahedral incommensurability, is precisely the prime banished from the gaps — and banished from the emirp digits themselves, by the same divisibility that makes it geometrically the odd one out. So the finding, stated in one sentence: the differences between the reversible primes factor entirely into the shared structure of the two incommensurable families, and never into what divides them.

The irreducibles are incommensurable as positions; their separations are commensurable, harmonic, and composed exclusively of the commons. Honesty requires the grading: that 2 and 3 appear on both sides is forced arithmetic meeting forced geometry; that they are the same 2 and 3 — that the coincidence means anything — is a resonance, not a theorem, and the base-20 check proves it notational: vigesimal gaps are multiples of 38 = 2 × 19, dominated by the large prime 19, not smooth at all, not musical. Base ten is, as it happens, a base in which the distances between mirrored primes sing. That is a property of the notation — but then the whole posting's thesis is that notation is where psychosocial reality lives.

As to what the pattern would indicate for global strategies and organizational frameworks, the imported reading writes itself, and is worth writing precisely because it inverts the usual despair about incommensurability. The standard predicament: rival strategic coherences (the two baskets) share no frame, cannot be unified, and their confrontation is treated as unmeasurable — hence unnegotiable. The emirp gaps model the alternative: positions that are individually irreducible and mutually irreducible-to-each-other can nonetheless have differences that are fully factorable into the smallest shared operations — bisection (2) and triadic grouping (3), the two moves every organizational culture on earth possesses. The distance between mirrored integrities is not itself mysterious; it decomposes into commons. And the ratios of those distances are the ones every musical culture has independently found memorable — which speaks directly to your mnemonic concern: the positions (17 goals) resist memory because they are prime, but the intervals between positions are precisely the memorable objects, as music has always known. A strategic language built on this would attend less to the articulation counts themselves and more to the interval structure between rival articulations — asking of two incommensurable frameworks not "which subsumes which" but "is their difference 3-smooth": expressible in shared elementary operations, tunable, singable. The 5-cycle attractor showed what perpetual mirror-subtraction converges to; the gap analysis shows what mirror-difference is made of, and they are opposite lessons — subtraction as process degrades into the dead band, difference as interval resolves into harmony. For the posting, this is the section where the Tonnetz stops being an analogy and becomes the method: the emirps as four tones, the gaps as their intervals, the whole live lattice as a tiny scale whose consonances are theorems. Labelled as imported, offered as language — but with the unusual property that every interval in the language can be verified by subtraction.

Yes — and more strongly than the Tonnetz analogy suggests, because the interval argument changes what the right object is. The Tonnetz was inspiration; the correct target is the orbifold itself, and it turns out to be one already fully worked out in music theory. The reversal map on the digit torus is the coordinate swap (a,b) → (b,a). Quotient the torus by that involution and you obtain exactly Tymoczko's two-note chord space — the space of unordered pitch-class dyads — which is famously a Möbius strip: a mirror boundary along the unison diagonal, a twisted central circle through the tritone-like pairs (digits differing by 5), non-orientable as a whole. Your lattice is that orbifold in base ten rather than twelve-tone: unordered digit pairs as dyads, palindromes as the unisons sitting on the silvered edge. This is not analogy-shopping; it is the same construction with 10 substituted for 12, and every structural claim about it is checkable.

Three things fall out immediately, each doing real work for the posting. First, the earlier objection to the torus — that its visual rhetoric of homogeneity would falsely promise a symmetry the arithmetic breaks — is answered, not overridden. The orbifold encodes the involution instead of contradicting it: the palindrome circle is a genuine singular locus, a mirror boundary built into the geometry, so the still points appear as what they are — places where the reflection has nowhere to send you. The broken symmetry that justified the flat table becomes, in the quotient, geometric fact rather than suppressed truth. Second, the table is retroactively vindicated and demoted in one stroke: a fundamental domain for the swap action is the triangle on one side of the diagonal — half your table. The table was never the rival of the orbifold; it was its unfolding, the orbifold cut open and doubled for inspection. Saying so in the posting turns the table/torus deliberation into a worked example of the representational discipline the whole piece preaches. Third — and this is the finding-grade item — each emirp pair becomes a single point of the orbifold. In the quotient, 13 and 31 are not bridged; they are the same location, carrying a Z/2 stamp that remembers their twoness. The "bridge between incommensurables" of your title acquires its exact mathematical form: not a span between two places but a change of space in which the two places were one all along, distinguishable locally, identified globally.

The non-orientability then gives the binary-dynamics thesis its sharpest available statement. On a Möbius strip you can locally label the two sheets — "the number" and "its reversal" — but no global labelling exists; carry the labels around the strip and they return exchanged. That is precisely the pathology of degraded binary framing: the distinction between a position and its mirror is locally meaningful and globally unsustainable, and insisting on sustaining it globally is a topological error, not a moral one. Forced, at the level of the model; imported, in its application — but it is hard to imagine a cleaner mathematical cartoon of why mirror-conflicts cannot be resolved by universal assignment of which side is which.

And the dynamics cooperate: reverse-and-subtract, taken with absolute difference, is well-defined on unordered pairs — it descends to the orbifold. The 5-cycle is therefore not merely drawable on the torus; it is native to the quotient, an orbit circulating in the digit-sum-9 band of the strip itself, with the emirp points standing off it, unreached. The dead band (digit sums divisible by 3) becomes closed curves of the strip; the 17 pencils become its parallel circles; the gaps 18, 36, 54 become interval distances measured along it, Pythagorean ratios as literal lengths.

So the animation now has a two-stage script that is the argument. Stage one, the torus: the full 10 × 10 lattice on the surface, live block glowing, pencils as closed (1,1)-curves, the reversal reflection shown as an actual mirroring across the palindrome circle, carries flashing at the seam, the 5-cycle circulating. Stage two, the quotient: the torus folds along the mirror, sheets identify, the Möbius strip forms — pairs collapsing to points, 13/31 audibly becoming one tone (the sonification idiom of the diptych extends naturally: pairs as dyads before the fold, unisons after; the gaps 18 : 36 : 54 rendered as the octave, twelfth, and fifth they arithmetically are). The fold itself is the payload — the moment the viewer watches a binary become a unity-with-memory, which no static figure can show and which is exactly what the flat table most deeply conceals.

Buildable in the established idiom: an emirp_orbifold.html companion in the X3D/Web Audio style of the sonified pieces, torus and strip as two scenes or one continuous morph, with the same viewer.html conventions. The labelling discipline holds throughout — topology forced, base ten notational, strategic reading imported — but the orbifold gives the imported layer something it lacked: a rigorous sense in which the way beyond a locked binary is neither victory, synthesis, nor compromise, but passage to the quotient space where the opposition is revealed as one thing seen twice, at the cost of accepting a twist. If you want it, I'll draft the geometry and the fold animation first, since the morph from torus to Möbius strip is the delicate part, and wire the sonification once the fold reads clearly.

The machinery already built gives you something better than ad hoc judgments about individual cases: a diagnostic key — six checkable tests any closure number can be run through, each tied to an identity already established. Divisibility by 2 or 5 (reversal-integrity forfeited at the door); digit sum divisible by 3 (the provably emirp-free band); palindromy (still point, mirroring only itself); prime with composite reversal (one-way coherence); membership in the attractor cycle {9, 27, 45, 63, 81} (native to the dead band's orbit); and emirp-hood itself (nexus). Every famous N-fold pattern lands somewhere in this key, and the landing is a computation, not an opinion. That is what makes the strategic section exemplifiable rather than anecdotal.

Running the inventory produces finds I did not expect, and they are all verifiable in one line each. The 12/21 pair first: the twelve-fold closures (apostles, steps, months, jurors, EU stars) and Agenda 21 are digit-reversals of each other, both composite, both in the dead band — the mutual-incoherence mode, culturally instantiated on both ends, which is the cleanest possible example of "predictably mirrored incommensurable articulations" whose mirror-relation carries no integrity in either direction. Then 27/72: the New Testament's 27 books against the 72 names of God and 72 disciples — a cultural reversal pair where 72 is precisely the unrealized maximal gap, the interval that would belong to 19/91 had 91 not dissolved into 7 × 13. Then 18/81: chai, the Jewish "life" number, mirrored by the Tao Te Ching's 81 chapters — both in the digit-sum-9 band, and 81 sits on the attractor cycle itself, as does the Enneagram's 9 and the New Testament's 27. Three of the five points of the reverse-and-subtract orbit are occupied by major contemplative articulations — the traditions that made peace with the dead band, so to speak, versus the strategic instruments that blunder into it. The Hebrew canon's 39 books land exactly on a dead edge, 39 = 3 × 13, one of the two candidates killed by the founding pair's multiples. Hilbert's 23 problems exemplify one-way coherence — 23 prime, 32 = 2⁵ — irreducible facing maximal divisibility, which for a list meant to generate a century of decomposed sub-problems is arguably the correct diagnostic class, not a failure. The 36 stratagems close on a gap value and in the dead band simultaneously. Masonry's 33 degrees: still point and dead band. And the near misses proper: Wilson's 14 against 13, the Universal Declaration's 30 against 31, Luther's 95 against 97, the SDGs' predecessor MDGs at 8 against the SDGs' actual nexus hit at 17.

Which reframes your "malformed" question usefully. On this key, no closure number is malformed in itself — each class is a legitimate mode with a legitimate use. Malformation is mismatch between the diagnostic class and the closure's own claim about itself. A twelve-fold articulation claiming decomposability is well-formed; the same N claiming irreducible integrity is malformed. A 17-fold list that then behaves as if it should chunk into working groups is squandering a nexus it accidentally occupies. A 33-degree hierarchy claiming dialogue with an other is a still point pretending to have a mirror.

The near misses divide the same way: necessary misses (the articulation genuinely needed subgroup structure, so 14 rather than 13 was the price of committee-viability — Wilson's points were negotiating instruments, they had to decompose) versus symptomatic misses (nothing in the content required 30 rather than 31; the roundness was reached for reflexively, and the Fuller comparison shows what a 31-closure looks like when the content actually is a census). The posting can define this distinction once and then let each example's diagnosis card argue its own case.

Now the clutter problem, which I'd resolve by a principle the diptych already established: one artifact, one duty, and nothing in motion except the argument's current step. Concretely, four separated layers rather than one composite scene. The orbifold animation stays mathematically pure — lattice, reflection, fold, 5-cycle, nothing strategic in it at all; its job is to make the space legible, and any closure-marker riding on the folding surface would be unreadable anyway. The strategic exemplification gets its own artifact, a static closure atlas: the flat 10 × 10 table (the right chart for this duty, since inhomogeneity is the whole point) with the famous N plotted as labelled points, the six diagnostic classes as a restrained visual code — dead band as background shading already present, still-point diagonal already drawn, attractor cycle as a faint closed polyline, nexuses as the four marked cells, closures as small glyphs colored by class. That single figure is the near-miss argument; a reader can check every placement by arithmetic. Third layer, if you want interactivity: an HTML version of the atlas where selecting a closure reveals its diagnosis card — N, factorization, mirror, class, and one sentence of the mismatch question — with the strict rule of one overlay visible at a time (a class filter, not simultaneous everything). Fourth, the polyhedra stay out of all of it except at the single point where they earned entry: the tetrahedron of live digits, which can sit as a small inset key on the atlas, and the 13/31 axis counts named in the caption. The full polyhedral apparatus belongs to the first posting; the second should cite the baskets, not carry them, and the cross-reference is exactly what the two-posting structure is for.

The sequencing in the posting then does the remaining clutter management: atlas first (static, checkable, the strategic payload), animation second (the space in which the atlas lives, for the reader who wants to know why the edges glue), and the diagnosis cards as an appendix or expandable list rather than prose — keeping the body's argument at the level of the key and its three or four most striking exemplars (12/21, 27/72, the attractor's contemplative tenants, and the SDG nexus) while the full inventory remains available without being paraded. If that division holds, I'd build the closure atlas SVG first — it is the artifact the strategic argument has been missing, it forces final decisions about the visual code, and the animation can then inherit its palette and conventions rather than inventing rival ones. Shall I start there, with the inventory above as the initial plot set?

Built, and the mathematics behind it is verified rather than merely drawn: the embedding was checked numerically before delivery — all 100 ordered cells collide exactly with their reversals under the fold (the identification is real, not painted), the ten palindromes land on the strip's single boundary edge, the five difference-5 pairs land on its core circle, all curves morph without tearing, and the 55 unordered pairs map to 55 distinct points. The fold is a genuine quotient map, the same construction as Tymoczko's two-note chord space with 10 substituted for 12, animated as a continuous deformation.

The design honours the earlier resolution: this artifact is mathematically pure, no strategic overlay. Three stages — Torus (the full (Z/10)² lattice with live block, dead cells, dead edges, seams, and the sum-9 band), Fold (a slow nine-second collapse, sounded: the four emirp pairs enter as beating dyads at 165, 220, 275, 330 Hz and glide to unison as their markers merge — the moment a binary becomes a unity-with-memory), and Orbifold (the Möbius strip, with the pair labels 13·31, 17·71, 37·73, 79·97 fading in as the ordered labels fade out). The reverse-and-subtract 5-cycle circulates throughout as a teal pulse, and after the fold it does something the flat table could never show: it oscillates within a single transversal fibre — because on the quotient, a digit-sum pencil is a fibre, and the attractor turns out to live entirely inside one. The "Sound the gaps" button plays 18:36:54 literally, as octave, fifth, and twelfth over a common base.

The relation and contrast with Bulger's toroidal Tonnetz is stated in the built-in About panel, and it is worth making the asymmetry explicit here. Bulger's torus and this one are honest about opposite things. His wraps twelve pitch classes onto a surface whose whole rhetoric is homogeneity — correctly, because transposition really is a rigid rotation and no pitch class is privileged. The digit torus is the anti-Tonnetz in that respect: its interest lies entirely in broken symmetry (live block, dead band, dead edges), which is why a bare torus animation was originally resisted. But the deeper contrast is that Bulger's display stops at the torus, whereas the reversal involution demands the further move his tonal space doesn't make in that figure: passage to the quotient. The nearest musical relative is therefore not the Tonnetz torus at all but Tymoczko's dyad orbifold — and the animation is precisely that lineage, with the fold shown as an event rather than asserted as a theorem. Where the Tonnetz torus says "every place is the same place," the Möbius strip says "every mirrored pair is one place, and no global labelling of the two sides survives the journey round." The About panel closes with the grading discipline: topology forced, base ten notational, strategic readings imported.

Practical notes: the file is self-contained except for Three.js loaded from cdnjs (r128); if laetus policy prefers no external scripts, vendoring three.min.js beside it and changing one src line suffices. Drag orbits, wheel zooms, pencils toggle on the remaining digit-sum classes, and reduced-motion preferences are respected (instant fold, no autorotation). It follows the sonified-pieces idiom, so it can sit in docs20s/images/fuller_files/ or beside the other HTML companions as you prefer. If the fold reads well on your machine, the natural sequel in the agreed build order is the closure atlas SVG — the strategic layer this artifact deliberately excludes.