On August 28, OpenAI launched training for a cutting-edge internal model that solved more than 100 world-class mathematical problems within just 24 days—including the notoriously difficult Navier-Stokes Millennium Problem, which it cracked in a mere 88 hours. The research team also made the relevant papers and verification codes publicly available. This breakthrough sent ripples of astonishment through the mathematical community, leading 27 Fields Medalists to issue a joint open letter criticizing AI companies for using mathematical problems as mere benchmarking tools. They argued that this utilitarian, "black-box" approach strays from the true spirit of mathematical inquiry.

In response, OpenAI established an independent Mathematical Advisory Panel composed of nine of the world’s top mathematicians, including Fields Medalists and leading academic figures. These panel members serve on a pro bono basis, retain the right to publicly critique the project, and are tasked with evaluating the significance of research outcomes, advising on publication decisions, and upholding rigorous academic standards. However, they have no authority to influence or direct the progress of AI research itself.

At present, AI’s ability to solve problems far outpaces human capacity for review and evaluation. This mismatch is compelling the mathematical community to rethink and restructure (refactor) its incentive systems and recruitment norms. Mathematicians are increasingly shifting from being problem solvers to becoming arbiters of problem value. Looking ahead, AI’s influence on academia is poised to expand into disciplines such as physics, chemistry, and biology.