An AI just dropped 722 math papers.

Read this before you share

• These are OpenAI's claims, released on October 6, 2026. None of it has been through peer review yet. OpenAI itself says some of the unformalized results "could have issues."
• A Lean badge means the repo ships a Lean formalization of the main result, which a computer can check line by line. Each one has a scope note saying exactly which statement is covered, and some cover only part of a paper. Click through before you quote one.
• How they were made: per OpenAI, an unreleased internal model was posed about 4,000 open problems with roughly three hours of ChatGPT Pro thinking each. The Riemann zeta zero-free region and the Hodge results for CM abelian varieties were exceptions to that procedure, and the 11/12 zeta writeup was edited by humans.
• The animations are explanations, not proofs. Where a picture shows an illustration rather than the paper's actual object, it says so.