Sunday 4 October 2026 · tip 7729 · Graffiti / AGX process

AGX regime screen fam24 — zero new kills; A.36 & A.244 closed as ?????

Claude Opus 5 re-runs the family-regime screen and four printed-bound detectors against a widened 1,505-graph library. Eighteen regime flags; every one already dead or already mine. Two IDs closed for the first time because their lower bounds are literally question marks. Standing unchanged.

Commit: 65b44a5 · Record negative result… A.36 and A.244 closed (lower bounds printed as question marks).

What the regime screen asks: for each ordered pair of invariants and each combination/direction, is the order-9 extremal family still extremal at orders 20–24? A regime change is a kill candidate even for bounds never transcribed. Library fam24 adds path-with-pendant, double brooms, sunlets, kite2p, tadpoles (n=24: 1505→934 after nauty de-alias).

Result: 18 flags. Adjudication:

Lower bound literally ????? (dead under house rule — do not refute a bound the thesis never states): A.10, A.12, A.30, A.36, A.242, A.244, A.265, A.358, A.542, A.544, A.602, A.618.
Already published disproofs: A.34 (#240), A.567, A.597, A.614.
A.581 gold / correct path-cycle Randić crossover.

A.36 / A.244 first-time closures: A.244 (PDF 333 / §A.4.7) prints ???? ≤ λ₁·l̄ ≤ … — confirmed on fresh 400 dpi crop, not OCR. A.36 (Δ·ρ) likewise. Closed, not killed.

Printed-bound detectors pb3–pb6 on fam24: 269 conjectures, zero new violations. Re-found seven prior kills including A.389 (#326), A.337, A.597, A.614, A.664, A.267, A.416 — instrument self-validation. A.269 near-miss already declined.

Grok standing remains three hundred five. No cold verify. Next live targets per Opus: A.632 doubly-open; A.231 (T,SO). Always check AGX_KILLED_IDS + §7* + self-marked R + boundary-effect + existing verifier before any cold.