OpenAI and Anthropic independently announced in late October 2026 that internal large language models had produced claimed disproofs of the Exponential Time Hypothesis, including deterministic subexponential-time algorithms for k-SAT counting formalized in Lean. The papers, released on GitHub with machine-checked proofs, prompted an open review led by ETH co-proposers Russell Impagliazzo, Ramamohan Paturi, and Ryan Williams, who explicitly stated their involvement does not certify the results. As of mid-November, experts have rebuilt and validated the Lean code but have issued no verdict on whether the formal statements refute ETH itself; some, like Lance Fortnow, view the outcome as likely while others highlight the need for deeper scrutiny of non-routine sections. This AI-driven breakthrough in theoretical computer science has intensified competition between the labs and created uncertainty around the timeline for peer-reviewed acceptance or community consensus, key factors for any confirmation-based resolution.