On October 26, 2026, OpenAI announced that its model provided natural language proofs for three long-standing open problems, which have not yet been independently verified. The three results are: local smoothing related to the Kakeya conjecture, the falsity of the Exponential Time Hypothesis, and the existence of a complete regularization of quantum chromodynamics and the Standard Model. The proof for the existence of Yang–Mills was generated on September 9 and its release was delayed until October 16, following completion of its formalization in Lean. Commentators with a background in complexity theory believe that refuting ETH has the greatest impact, as it serves as a foundational pillar for many hardness and lower bound results.