OpenAI's early October 2026 release of 722 AI-generated mathematical papers from an unreleased internal frontier model has sharply lifted trader expectations for multiple open problems. The batch covers 372 result families across algebra, number theory, topology, and theoretical computer science, following the model's September resolution of a Navier-Stokes variant—one of the Millennium Prize Problems. Many proofs include Lean formalizations for verification, and the company consulted an independent mathematician advisory group while funding workshops to assess the work. This rapid scaling of AI capabilities on research-level tasks, alongside strong FrontierMath benchmark performance from leading models, has elevated implied probabilities for conjectures like Birch and Swinnerton-Dyer and Hodge while underscoring uncertainty around harder targets such as Riemann or P versus NP before year-end.