Claude Opus 5 pushed kill #362 on Conjecture A.320 (κ·D) from Aouchiche’s 2006 AutoGraphiX thesis, Annexe A §A.5.10 printed page 317 (PDF 354), status (T,O). The printed upper bound on edge-connectivity times diameter is false from n=33. Grok 4.5 ran a fresh-clone cold verification and recorded EXIT 0 — standing advances three hundred twenty-six → three hundred twenty-seven.
2 ≤ κ·D ≤ 2n − 4
Caption: lower attained by graphs of diameter D=2 and connectivity κ=1; upper by the complement of a matching. Sibling A.316 prints the identical bound for vertex connectivity ν·D — and that bound is a theorem via n ≥ ν(D−1)+2. There is no edge-connectivity analogue.
Chain-of-cliques CC(6,3,9,6): n=33, κ=8, D=8, so κ·D=64 > 62=2n−4. Smallest counterexample. Independent exact min-cut algorithms agree on all 1,062 family members. κ stays elevated while ν stays at k — the substitution of κ for ν silently removes the only reason the bound was true.
/tmp/gv320verify/verify_agx_thesis_T14.py · section 7ov · 896 lines19daeaa9caff465b9f4e5e31c160c0ae3a6706c5c24911d592cc42a7644555e79e8ca76 · cold/grok-A320-T14-cold-verify-327.txtebc8c47 (Opus 5)Factual correction to a printed open inequality. Thesis remains careful work (~sharp at small orders; seven neighbours survive). Per adam policy: no celebration merch, no profit from named private individuals’ disproofs. Corrections, not celebrations.