OpenAI posts 722 AI math preprints, pulls three next day

OpenAI uploaded 722 AI-generated mathematics preprints to a GitHub repository on 6 October 2026, claiming progress on 372 open problems across geometry, algebra and computer science. The manuscripts were produced by an unreleased model, and the company pulled three of them a day later after a sign error invalidated the argument in one paper and the constructions in two others that depended on it, Nature reports.

The mathematics community did not take it well. Alvaro Lozano-Robledo, a number theorist at the University of Connecticut, called the volume "staggering" and asked, "Why is OpenAI trying to solve hundreds of math problems to be released all at once?" Ursula Martin of the University of Oxford described the dump as "tossing the community a messy first draft and expecting us to clean it up." Michael Harris, a number theorist at Columbia University, said he would not read any of it: "I can't bring myself to spend my time studying the outcome of this process."

Not every reaction was dismissive. Levent Alpöge, a mathematician at rival lab Anthropic, said whatever one thought of the release, "it's obviously the most significant moment in mathematical history," while also warning of "sad stories" for researchers whose active projects had been pre-empted. Three of the researchers we follow in our Who's Who directory posted the Nature piece.

OpenAI's spokesperson framed the release as a way to "provide tools for mathematicians and scientists" by evaluating models against difficult frontier problems. Three of the 722 preprints are now withdrawn.