OpenAI and Anthropic each released AI-generated claims in late October 2026 that internal large language models had produced subexponential-time algorithms for 3-SAT counting, formalized in Lean and purporting to refute the Exponential Time Hypothesis. Reviewers including ETH originators Impagliazzo and Paturi plus Ryan Williams are examining the work but have issued no certification as of mid-November, with experts noting that Lean verifies syntax while independent checks on whether the statements match the hypothesis remain incomplete. Trader sentiment reflects this verification gap, as confirmation would upend fine-grained complexity assumptions used in cryptography and algorithm lower bounds, while delays or flaws could preserve the status quo. Key near-term catalysts include any public verdicts from the review group or follow-on papers extending the results.