OpenAI’s large mathematical research release awaits scrutiny

OpenAI released 722 manuscripts across 372 result families, with Lean proofs for many but not all; the model identity is undisclosed. Independent expert review and formal verification are still needed to establish the work’s validity and significance. A Google DeepMind talk separately emphasizes mechanically checkable proofs as one foundation for trust in AI-driven mathematics, but offers few technical details.