OpenAI posts 722 math papers, incl. Quasi-Riemann Hypothesis

The release is framed as a broad set of mathematical results from an internal frontier model, not a product launch. OpenAI says it consulted the Institute for Advanced Study's independent Advisory Group on Mathematics and AI on how to release them — a step that points to the awkwardness of publishing findings from a model that stays private.

The scale is unusual: 722 manuscripts grouped into 372 families, drawn from an evaluation of about 4,000 problems. Among the highlighted items are a result for integer multiplication faster than n log n and a uniqueness result for the elastic inverse problem that the paper says had been open in 3D since 1994. Commenters also point to partial progress on Riemann, Hodge and BSD.

What the outcome hinges on is scrutiny. Will Depue expects some results should not survive it, and an analysis arguing that roughly 20% are disproofs or counterexamples cuts against the idea that these are mostly brute-force wins. Levent Alpöge's praise came alongside his own note of reported scooping and conflict-of-interest problems involving other labs' users — a sign the social machinery around AI math is straining as much as the mathematics.