OpenAI has released 722 mathematical manuscripts generated by internal frontier models via GitHub, covering Fields Medal-level problems like the quasi-Riemann hypothesis and the 4D Kakeya conjecture. This marks a shift from AI benchmarking to substantive scientific discovery. Sam Altman declared the start of a "new era of big discoveries," triggering significant reactions in the global mathematics and developer communities.
Cross-verified sources indicate the release includes not just paper drafts but core computer-verifiable formal proofs using the Lean language. Following advice from Institute for Advanced Study (IAS) advisors, OpenAI established transparency standards emphasizing verifiability and community collaboration norms. Combining natural language reasoning with formal logic addresses traditional academia's skepticism toward AI-generated content, providing a solid technical foundation for future research.
This event is a milestone for the AI for Science sector, redefining AI's role in basic research from auxiliary tool to potential knowledge producer. For developers, open-sourcing Lean proofs lowers the barrier to formal verification and accelerates automated exploration of mathematical theorems. Future focus will center on establishing robust mechanisms for recognizing AI academic contributions and addressing peer review challenges for large-scale machine-generated results.