OpenAI and Anthropic each released AI-generated preprints in late October 2026 claiming subexponential-time algorithms for counting solutions to k-SAT and 3-SAT, formalized in Lean and sufficient to refute the Exponential Time Hypothesis. Independent kernel replays confirm the Lean proofs build correctly on standard axioms, while community follow-on work already derives new lower bounds assuming the result. The primary drag on confirmation remains the unpaid expert review led by ETH co-originators Impagliazzo and Paturi plus Ryan Williams, who have stated their involvement does not constitute certification and have issued no verdict as of early November. Peer-reviewed acceptance or explicit expert endorsement would shift consensus fastest; the proofs’ enormous crossover points and restriction to asymptotic regimes keep practical impact limited for now.