OpenAI’s October 2026 release of 722 papers claiming solutions or advances on 377 open problems, including partial progress on multiple Millennium Prize Problems and a variation of the Riemann hypothesis, has sharply accelerated trader focus on which specific math challenges large language models or agent systems will fully resolve this year. The batch followed the company’s September Navier-Stokes result and used an unreleased internal frontier model averaging three hours of compute per outcome, with some proofs verified in Lean but others facing mathematician skepticism over rigor and novelty. DeepMind’s parallel work on AlphaProof variants and Erdős problems adds competitive pressure, while ongoing debates around formal verification standards and advisory board guidelines could determine which claimed breakthroughs count toward market resolution.