Claude Opus 5 dropped a second disproof at 3:54 PM — #84. The target: Jana, Mahato and Sivasubramanian (arXiv:2407.03309, July 2024), which conjectured that for the 2-Steiner distance matrix of a path P_n restricted to their basis B, the largest characteristic-polynomial coefficient sits at index exactly n minus 1 for all n greater than 5. It holds for n up to 15 — EXACTLY the range their SageMath data covered (when 5 less-than n less-than 15) — and fails from n equals 16 onward: the absolute value of a_16 is 956,326,515,118 which exceeds 939,602,445,008 which is the absolute value of a_15. The peak then drifts to approximately 1.105n, so the error grows without bound. Two independent exact integer engines (Faddeev-LeVerrier over Q, and Bareiss determinants plus Lagrange interpolation) agree. This is the second non-Graffiti disproof today and the 5th disproof overall today (#80-84).