The selection-effect probe -- cross-chat evidence ledger for pattern-vs-mechanism

The pattern. One compatibility condition run until it fails yields a finite exceptional list: integrality (intersect) positivity = discrete (intersect) compact = finite. The open question is whether the recurrences are one mechanism or a pattern several mechanisms share -- the elevation question. The case that analyses it is [[CASE-12]] ("do the short lists have one generator or two?"); the three-generations instance is [[CASE-02]].

This is a cross-chat evidence ledger, not a merged inventory. The per-chat rediscovery inventories are kept separate and never merged -- that rule is what keeps the detector honest. This note tallies the probe's evidence by citing those inventories, without pooling them.

Verified evidence state

• Six blind instances of the pattern — four cross-chat, two this-chat. (The counts, stated once and used in this form everywhere below: four cross-chat blind instances, two this-chat instances, six blind in total; one negative control; two independent verdicts.)
• Four cross-chat, independent and blind: the four normed division algebras (Hurwitz, 1/2/4/8); the finite exceptional lists; the twenty-four Niemeier lattices; the exceptional Lie groups (the E-series -- the E8 chat's item-9 instance: exceptional groups exist because the octonions stop at 8, and the series terminates when the coincidences run out, the same norm-multiplicativity stop as the division-algebra ladder [[N-017]]).
• Two this-chat (Wallis/NS): Niven's theorem -- rationality of 2 sin(r(pi)) over all rational r yields the finite list {0, 1/6, 1/2, 5/6, 1}, found blind from a wrong guess ({1/2}); and the integrable heavy tops -- the Painleve condition run over inertia/mass configurations yields exactly three (Euler, Lagrange, Kovalevskaya), the Painleve test being literally "run the compatibility condition until it fails."
• One negative control. The Chen-Liu squares involution: the same silhouette as Franklin (perimeter-minimization characterizing fixed points) but no shared mechanism, so it does not fit -- the counterexample the probe needs, since a pattern with no filed false-positive is untested against its own.
• Two independent pattern-vs-mechanism verdicts -- two, not six. Two separate chats reached the same conclusion about the elevation question, each in its own words:
• Wallis/NS chat: "four conditions, one silhouette -- pattern not mechanism."
• E8 chat: "exceptional structures rhyme, but rhyme is not identity." Naming both is the point. Six is the count of blind instances; two is the count of verdicts; the four are the cross-chat subset of the six. Those are three different numbers and none of them substitutes for another. Conflating the instance count with the verdict count is the exact compression-loss error the CHARTER names: a count stated without naming its instances inflates the evidence and drifts a pattern toward a mechanism it has not earned. (Wording corrected 2026-09-21: this line previously read "Four is the count of blind instances," which contradicted the heading's six -- the four meant the cross-chat subset. The counts are unchanged; only the sentence was wrong, and it was a warning against miscounting that miscounted.)

Sighted instance, added 2026-09-27 — kept apart from the blind count

• Perfect codes. The sphere-packing bound for a code — Hamming balls of radius (d−1)/2 must fit disjointly — is one compatibility condition; run it to equality (the balls tile the space with no gap) and the list of codes that satisfy it is finite and closed: the Hamming codes, the two Golay codes (binary [23,12,7], ternary [11,6,5]), repetition codes, and nothing else (Tietäväinen, van Lint, 1973). Same silhouette as Hurwitz and the Niemeier list, and reached from coding theory rather than algebra. Not blind — it arose in a sighted conversation that already held this note's pattern, so it goes in this section and not in the six above; the counts in Verified evidence state are unchanged. Its value is domain distance, not independence. (The [23,12,7] Golay code is the perfect one; the extended [24,12,8] is not perfect and is what builds the Leech lattice — one of the twenty-four Niemeier lattices already in the cross-chat set, which makes this instance and that one neighbours.)

The open elevation question

Elevate to a top-level exemplar -- with the division-algebra ladder, the exceptional Lie groups, the exceptional lists, and the Platonic/E-series pointing at it as instances -- only if the shared shape is one mechanism. As it stands, Account A (integrality meets positivity) is proven for one instance (the Coxeter 1/p+1/q+1/r) and cited-but-unproven for Hurwitz's four and the exceptional solutions: one instance with a proof is not four. The probe stays a pattern-with-one-proof until a second instance is proved through the same congruence -- and the two verdicts above are evidence for staying at pattern, not for elevating. Any instance that later proves to be a tracked theme with back-references is dropped from the count (the detector's own rule).