Recent OpenAI releases of hundreds of AI-generated mathematical papers, including claimed partial resolutions of the Hodge conjecture for CM abelian varieties and progress on the Birch and Swinnerton-Dyer leading term, have elevated those problems in trader assessments following the September 2026 Navier-Stokes breakthrough. These outputs rely on multi-agent large language model systems that produce Lean-formalized proofs, yet most results await independent verification and fall short of full Clay Institute certification, sustaining uncertainty about timelines. The close contest between "no solution by end of 2027" and the leading conjectures reflects rapid capability gains in mathematical reasoning benchmarks alongside persistent barriers like rigorous peer review and the distinction between partial advances and complete solutions. Key catalysts ahead include further model iterations and any formal acceptances that could shift implied probabilities.