OpenAI and Anthropic both announced on October 26 that internal large language models independently produced subexponential-time algorithms for 3-SAT and #3-SAT on October 16, formalized the next day in the Lean proof assistant and posted publicly on GitHub. These claims directly refute the Exponential Time Hypothesis, a foundational assumption in complexity theory that underpins hundreds of conditional hardness results. A review group led by original ETH proposers Russell Impagliazzo and Ramamohan Paturi plus Ryan Williams is organizing public examination, though the experts explicitly noted their involvement does not certify the results. No independent peer review or external reproduction has occurred yet. Traders are watching the review timeline and any planned conference for resolution signals.