The Exponential Time Hypothesis Is False [pdf]
Points and comments are a snapshot, not live.
OpenAI claims to disprove the Exponential Time Hypothesis, showing 3SAT has subexponential algorithms.
A preprint hosted on OpenAI's GitHub repository claims to prove that the Exponential Time Hypothesis (ETH) is false. ETH states that 3SAT cannot be solved in subexponential time. The paper is listed among numerous other math and CS preprints in the openai/math repository. The proof's runtime appears to be roughly 2^(n/sqrt(log(n))) with additional fudge factors, which is subexponential but not polynomial.
What commenters are saying
Commenters widely agree the result is huge if correct, but many caution it does not prove P=NP or even NP≠EXP. Several corrections emphasize that subexponential 3SAT does not imply NP is subexponential because polynomial reductions can expand inputs enough to restore exponential bounds. Scott Aaronson is cited raising his P=NP probability from 3% to 5%. A minority worry that many papers assuming ETH now have weakened foundations.