OpenAI posts 722 AI math papers via Advisory Group on Mathematics and Artificial Intelligence

OpenAI's release follows consultation with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study (AGMAI). OpenAI says it drew on AGMAI's advice, including a recommendation on the responsible publication of AI-generated mathematical information. The repository also lays out procedures for revising and citing the papers.

The 722 manuscripts are grouped into 372 result families and sorted by mathematical field. Many include the associated proof materials, and OpenAI has published formalizations of many proofs in Lean, a programming language that lets computers check mathematical proofs. OpenAI says it plans to update the repository as new formalizations become available.

Reaction from outside has been intense but mixed. Quantum information researcher Isaac Kim called the list of problems in his field shocking, saying things are changing so fast he cannot imagine what will happen in a few months, let alone a year. The X account AI Notkilleveryoneism Memes said AI has solved 90 of the 500 most important open problems in mathematics, listing partial progress on high-ranked problems such as the Riemann Hypothesis, the Hodge Conjecture, and the Birch and Swinnerton-Dyer Conjecture. Another X user, imjustnewatai, picked four major claims including the quasi-Riemann hypothesis and the unique games conjecture, but noted that none have been confirmed by outside mathematicians yet.

The stakes now seem to hinge on whether mathematicians can find genuinely new ideas inside these proofs, or whether the results mostly recombine existing techniques. Hundreds of proofs may take months to analyze, and not every manuscript has a Lean formalization, so some problems remain unverified. How the mathematical community judges the novelty and reliability of this batch is likely to shape how AI-generated results are released and reviewed in the future.