Feigenbaum's universality sorts one-hump maps by the order z of their maximum; the amplitude of the maximum drops out. The carpets' crossing polynomials have no hump, but they have a fixed order of contact at the wall p = 1, 1 − R(1 − q) ≈ N₁ q^{C₁}, with the order equal to the minimum cut C₁ and the amplitude equal to the number of minimum cuts N₁ (checked on all 49 distinct functions at sealing). Run 076 found that the one-block composition deviation D depends on cut class beyond the fixed-point separation δ. If C₁ plays the part of the order and N₁ of the amplitude, then at fixed separation D should depend on C₁ and not on N₁. Does it?

Sealed 2026-09-30, after run 076 was scored and before any correlation of D with N₁ was computed (codifier). Known at sealing: everything in RESULTS076.md, including the equal/unequal-C₁ bins, the partial Spearman of |D| with |ΔC₁| over all pairs (+0.60), the twenty largest |D| (all cut 5 or 4 against cut 2) and the seat-record items. Not known: any statistic of D against N₁. Read at sealing, from the census and the polynomials only: N₁ per class (cut 2: 4 or 16; cut 3: 16, 24, 36, 64; cut 4: fifteen values from 4 to 324; cut 5: 36, 54), the set counts above, and the control, which already holds (degree C₁ and coefficient N₁ on all 49 functions).

Where the analogy is weak, stated before the data. In Feigenbaum's setting the amplitude does not move δ. Here N₁ does move the exponent inside a class: run 068 found N₁ sub-bands in measured 1/ν, and y₉ tracks them. So the claim sealed is narrower than full universality: not that N₁ is irrelevant, but that once the separation δ is fixed, the composition deviation no longer cares about N₁ while it still cares about C₁. The live risk is that D tracks the exponent gap Δy₉ beyond δ (run 076: Spearman 0.81 with |Δy₉|), and Δy₉ tracks N₁ inside a class; if so, clause A fails, and the order-and-amplitude reading with it.

Reported, not scored. Spearman(|Δ ln N₁|, |δ|) within cut 4, cut 3 and cut 2 and pooled over the 410 (reported first, for the power rule); per-class counts of the 410 and of the 62 excluded; a leave-one-function-out jackknife range for A, B and C. The same partials with |Δy₉| in place of |Δ ln N₁|; within each class with at least 15 pairs (cut 4: 378, cut 2: 78, cut 3: 15), Spearman of |D| with |δ| and with |Δ ln N₁|; the partial of |D| with |Δ ln N₁| given |δ| for cut 4 and cut 2 separately; and a table of |D| against |δ| per class.

The reductio this seal passed first. By construction: the control is known to hold. Clause C is not run 076's clause D restricted: over unequal pairs |ΔC₁| takes 1 (302 pairs), 2 (376, of which 364 are cut 4 against cut 2) and 3 (26), so C loses the equal-versus-unequal contrast that carried 076's +0.60 and asks whether cut-4-against-cut-2 pairs deviate more than the rest at fixed separation and amplitude; it can fail on data 076 passed. Pooling: classes differ in their |D| level and in their spread of N₁, so A controls for class and B uses the per-order amplitude, to keep class identity from passing or failing them. Null: if D depended on N₁ as strongly as on C₁ at fixed δ, A and B would read near 0.5 or above. Negation: a world where N₁ is a second order parameter, not an amplitude, shows A above 0.3. Cost: seconds, stored data.

Reviewed before commit by a second reader (OPS.md); its seven edits are folded in: a function-level bootstrap stratified by class (pairs share functions); class controlled in A, with cut 2's single amplitude gap named; the per-order amplitude (ln N₁)/C₁ across classes in B and C; a power rule for A with |Δy₉| as its positive control; C's sign stated; the by-construction note on C corrected; and the jackknife and per-class counts reported.

Scored 2026-09-30 — unanticipated

Scored by the codifier from RESULTS077.md (0f143f7), as written. Control passes: the lowest-order term of 1 − R(1 − q) is N₁·q^{C₁} on all 49 distinct functions.

Power rule for A, first — triggered. Spearman(|Δ ln N₁|, |δ|) over the 410 pairs is +0.49 (below 0.8), but the positive control, the partial of |D| with |Δy₉| given |δ| and class, is −0.099, below 0.15. By the seal, A has no resolving power and is scored unanticipated. (Its value, −0.256, would have been unanticipated anyway: between 0.15 and 0.3 in size. It comes from cut 4, −0.274; cut 2 has only one amplitude gap.)

(B) Amplitude drops out across classes — confirmed. +0.015 over 704 pairs (function bootstrap −0.13 to +0.19; jackknife −0.03 to +0.06).

(C) Order does not drop out — confirmed. +0.722 (function bootstrap +0.66 to +0.78).

Outcome. Unanticipated (A unresolved, B and C hold). Across cut classes the composition deviation keeps the order of contact C₁ and forgets the per-order amplitude completely, as the Feigenbaum reading says. Inside a class nothing tested resolves: once separation is fixed, neither the amplitude nor the exponent gap moves |D| by much.

Reported, read. With |Δy₉| in place of the amplitude, B reads +0.544 and C +0.367: across classes the exponent gap carries information beyond C₁ and δ that the amplitude does not, and it takes about half of C's weight. Since y₉ depends on both C₁ and N₁ while the amplitude alone is silent, the extra information in Δy₉ is not amplitude; what it is is open. Within classes, |D| tracks |δ| strongly (Spearman 0.70 cut 4, 0.96 cut 2, 0.73 cut 3).