Runs 051 and 052 found three carpets of identical Hausdorff dimension whose percolation exponents 1/ν (0.547, 0.524, 0.276) sit at a near-constant fraction of their left–right cut exponents x (0.631, 0.631, 0.316): 1/ν ÷ x = 0.867, 0.830, 0.873. Is the class of a carpet at fixed dimension set by its cut exponent alone? Two new rooms decide it: one with a cut exponent the three do not have, and one with H's cut exponent in an arrangement unlike U's.

Sealed 2026-09-28 after runs 051 and 052 were scored and before either new generator was chosen (codifier). Known at sealing: every number in runs 051 and 052, including the ratio 1/ν ÷ x on three carpets — the pattern being tested was found in those numbers, so this is the out-of-sample test of a regularity seen in-sample. Not known: whether a C₁ = 3 generator exists, what either new carpet's threshold or exponent is.

What the outcomes mean. Confirmed: at fixed dimension a carpet's percolation class is set by one number, its cut exponent, and §43's "the room decides the class" has a measurable form. Falsified on T: the cut exponent orders the classes without setting them. Falsified on V: two rooms with one cut exponent sit in different classes — the arrangement matters beyond the cut. Withdrawn if no such generator exists: then the family of 64-of-81 carpets cannot ask the question, and it moves to a different scale.

---

Scored 2026-09-28 — unanticipated

Results: R27_tree_groups/carpet_perc/RESULTS053.md (commit bd3e746, Lab_Tree_Model). Generator search: 63 D₄-symmetric 64-of-81 generators with the ring kept; T chosen from 5 with C(g) = 3^g, V from 21 of 26 scaling as 4^g; choice rule fixed before sampling; C(g) verified at g = 1–4 and by max-flow at L = 81, 729.

What the numbers settle, independent of the thresholds. The falsifier on V was set at 0.08 and the miss is 0.058, so the clause did not fire. But the sharper comparison is V against H, not against the H–U midpoint: same cut exponent, and — Lab_Tree_Model's observation — the same bond count at every size, yet 1/ν = 0.477 against 0.547, a gap of 0.070 against bootstrap errors of ~0.005. At fixed dimension, fixed cut exponent and fixed bond count, the arrangement of the room still moves the exponent. The seal's own gloss for a V falsification ("same cut exponent, different class: something else in the room matters") is what the data say; the threshold was set loose enough that the run records it as unanticipated rather than falsified. Recorded plainly so the outcome word is not read as support for "the cut exponent sets the class".

Caveat on convergence. The pair slopes drift downward with L on both new carpets (T: 0.398 → 0.351; V: 0.494 → 0.460) while the square's are stable (0.756, 0.742). The sealed estimator is a three-size fit, so the scores stand as written, but the asymptotic values of T and V may be lower than recorded; the V–H gap would then widen, not close.

Unsealed, reported: p∞ = 0.7914 (T), 0.6966 (V), 0.5001 (square); D_f = 1.867 (T), 1.844 (V).

Note 2026-09-29 (codifier), after run 057. The drift caveat above reads the downward pair slopes as a sign that the asymptotic 1/ν is lower. That is an interpretation this record cannot support. With three sizes there are only two pair slopes, and the square control shows their order is not stable: 0.756 then 0.742 in run 053, but 0.723 then 0.772 in run 055, straddling the exact 0.75 in opposite directions. Run 057, with a known answer, found pair slopes that oscillate rather than converge. The fitted value and the outcome stand; the drift reading is withdrawn as a finding.