OpenAI unleashes hundreds more math results upon a field in shock

OpenAI released 372 mathematical results produced by an internal, unreleased model that resolved or advanced major open problems in mathematics and theoretical computer science. The batch includes a solution to the four-dimensional Kakeya conjecture, improvements to key algorithms, and progress toward the Riemann hypothesis. Many of the proofs have been formalized in the Lean proof assistant, which mechanically verifies logical steps. OpenAI reports that nearly every result emerged from a single prompt to a single AI agent, a stark contrast with its earlier Navier-Stokes solution that was generated by a costly, 10,000-agent computational swarm.

The release triggered intense debate because the model, exact prompts and full compute details remain private. An independent advisory panel recommended public disclosure of models, prompts and compute resources; OpenAI provided only average compute times and limited statistics and says it is working toward a responsible public release. Leading mathematicians express deep skepticism and demand reproducibility before acceptance; OpenAI acknowledges some results required multiple attempts and that many proofs are not yet fully understood even by its own researchers. The episode forces a community choice between rapid, opaque proliferation of AI-generated proofs and slower, transparent validation, while OpenAI argues rapid progress is necessary to demonstrate advancing AI capability.