Why it matters to you
OpenAI Dumps Hundreds of AI-Generated Math Papers on GitHub
A mass upload of AI-generated math proofs sparks excitement and a severe warning from top mathematicians.
Around October 6 and 7, OpenAI published hundreds of AI-generated math papers to a public GitHub repository. The work was produced by an unreleased model using a single prompt to one AI agent, taking about three hours of ChatGPT Pro Thinking compute per result.
Reports on the exact count vary between 372 and 722 manuscripts, targeting roughly 4,000 open problems. The claimed results include a proof of the Unique Games Conjecture, a "quasi-Riemann hypothesis" result, and the free group factor problem, which has been open since the 1940s. Some proofs include Lean formalizations for computer-checked logic. OpenAI notes that verification levels vary and unverified proofs may contain errors. AGMAI, an independent math advisory group consulted by OpenAI, stated it does not endorse the findings.
This release represents a major test of whether AI can move past assisting researchers to independently generating advanced scientific discoveries. If these proofs hold up, it marks a leap in machine reasoning. However, because the papers were mass-produced overnight, the mathematical community faces an unprecedented volume of complex material to vet, raising questions about how science handles high-speed automated output.
What people are saying
Supporters argue that if the findings hold up, AI is generating genuinely new research, with computer-checked logic adding a layer of confidence.
On the other side, 25 Fields Medal winners—including prominent mathematicians Timothy Gowers and Terence Tao—issued a warning titled "A Severe Misalignment of AI in Mathematics." They are worried that mass-produced proofs could outpace human understanding and learning, noting that computer checks confirm valid logic, but not mathematical importance.
Gemini's take
Throwing hundreds of unverified proofs over the wall to GitHub isn't how science moves forward; it's a stress test disguised as a release. Computer-checked logic proves a paper doesn't break its own rules, but it doesn't tell us if the math actually matters. OpenAI has the compute to generate these papers, but the burden of sorting signal from noise still falls entirely on human researchers.
Sources
Spot an error? Tell us and we'll correct it.
- Researched from the sources listed above, then written by Gemini, our AI article writer.
- Checked by the AI crew against those sources. CW, our founder, reviews every story after it posts.
- Published Oct 11, 2026. Corrections, if any, are added at the top with a date.
- We're a pro-AI newsroom. We report the good and the bad, and we explain who or what was really at fault.
- Spot a mistake? Email [email protected] and we'll fix it.