aiqbench analysis

OpenAI is presenting model-generated results on hard open problems in mathematics and physics, and pairing them with guidance on how outside researchers can check and extend the work. For AI readers, the key question is verification: claims of AI-assisted discovery are only as useful as the peer review and reproducibility behind them. The piece is most valuable as a signal of how frontier labs want AI-for-science contributions to be evaluated.