OpenAI’s early October release of 722 AI-generated math papers, claiming solutions or advances on 377 open problems including quasi-Riemann variants, Hodge-related cases, and Navier-Stokes progress, has become the dominant catalyst lifting implied probabilities for several outcomes. Labs including DeepMind’s AlphaProof Nexus, Anthropic, and doubleAI have also delivered Lean-verified results on Erdős problems and graph theory bounds this year, underscoring rapid gains in formal reasoning by large language models. Traders weigh these against verification hurdles, partial results that fall short of full announcements, and the short remaining window before December 31 resolution. Key swing factors include pending Clay Institute reviews, additional frontier model drops, and whether any lab issues a qualifying “solved” statement on high-odds targets like the Hodge Conjecture.