OpenAI has announced three claimed breakthroughs across mathematics, theoretical computer science and physics, while using a different disclosure process for each result. The company says expert “shepherds” are helping decide how the material should be shared after researchers criticized an earlier bulk release of mathematical preprints.
For mathematics, OpenAI released AI-generated output that it says resolves the Kakeya conjecture in all dimensions. A Brown University professor told The Verge the claim looked well supported but had not yet checked its Lean formalization. For computer science, OpenAI says it disproved the exponential time hypothesis and plans to convene outside researchers to explain the result. The physics work concerns quantum chromodynamics and Yang–Mills theory, but the underlying material has not been released.
The topics are consequential, yet announcement and verification are not the same. Experts described the results as potentially major while stressing that complete evaluation would take time even with full information. Withholding the physics evidence and relying on selected groups to interpret other outputs limits independent scrutiny. The immediate news is therefore not settled proof of three breakthroughs, but a set of serious claims entering different and incomplete verification paths.