Recent Epoch AI classifications of FrontierMath Open Problems as solved, including the September 2026 human-plus-AI resolution of the Apéry-style irrationality proofs problem, have reinforced trader expectations for further progress. OpenAI’s GPT-6 Astra has saturated Tier 4 benchmarks at 97.6% while leading on the harder Open Problems and new Erdős sets, outpacing Anthropic’s Claude models through agent swarms and iterative reasoning. These outcomes follow the earlier Hadamard matrix solution and Navier-Stokes claims, highlighting how targeted scaffolding and compute scale convert benchmark gains into verified research-level advances. Key upcoming catalysts include additional Open Problems evaluations and potential new classifications before year-end deadlines.