OpenAI Dropped 400 Math Results and Mathematicians Are Reeling
OpenAI just dropped nearly 400 AI-generated mathematical results across more than 700 manuscripts, covering everything from combinatorics to mathematical physics. Mathematicians are calling it staggering, overwhelming, and pure insanity. The sheer volume has left researchers struggling to digest the material, with some describing entire research programs wiped out overnight. The release raises urgent questions about verification, attribution, and what it means for the future of mathematical discovery.
The Scale of the Drop Is Unlike Anything Before
OpenAI released nearly 400 results spread across over 700 manuscripts, spanning combinatorics, geometry, number theory, theoretical computer science, algebra, topology, probability, statistical mechanics, and mathematical physics. The collection is so vast that OpenAI published guidance on navigating the GitHub repository. Mathematicians told The Verge that simply working through the roughly 40-page table of contents and abstracts took them the better part of an hour. Álvaro Lozano-Robledo, a professor at the University of Connecticut, said just going over the entire list of abstracts is overwhelming. The volume alone makes preliminary assessment difficult, and researchers agreed that fully understanding what was released could take years.
Verification Gaps and the 'Slopocalypse' Fear
Only about 42 percent of the manuscripts have been formalized in Lean, a proof assistant that allows computational verification. OpenAI acknowledged the results are at different stages of verification and said it will update the repository with more formalizations. Kevin Buzzard, a mathematics professor at Imperial College London, said he identified numerous theorems in algebraic number theory but only around six immediately stood out, and few appeared formally verified. He said he either has to read possibly-not-correct slop, wait for others to do the same, or wait for formalization. Researchers had feared a slopocalypse of low-quality AI-generated papers, and while early impressions were better than expected, many papers remain difficult to follow and attribution concerns persist.
Some Results Are Genuinely Impressive, But Disruption Is Real
Despite the chaos, many researchers said the work appeared to be of very high caliber. Stanford mathematician Jared Duker Lichtman said there were tens of results he would put in the category of potentially Fields Medal-worthy, including progress toward the Riemann hypothesis, a special case of the Hodge conjecture, and a solution to the four-dimensional Kakeya conjecture. Scott Armstrong noted many are well-known problems that people have tried for decades. But the release also wiped out research programs. Colva Roney-Dougal said friends and colleagues had grant proposals wiped out. Tristan Buckmaster said he heard of three people whose entire research programs were obliterated. The disruption was concentrated in probability, combinatorics, theoretical computer science, and group theory.
Key Takeaways
- OpenAI released nearly 400 AI-generated math results across 700+ manuscripts, overwhelming researchers.
- Only about 42 percent of manuscripts have been formally verified in Lean, raising quality concerns.
- Some results are considered Fields Medal-worthy, including progress on the Riemann hypothesis and Kakeya conjecture.
- Mathematicians report entire research programs wiped out, with particular disruption in probability, combinatorics, and group theory.
- Understanding the full scope of the release could take years, and researchers fear OpenAI will move on before then.
Keep Reading
