GPT-5.6 Pro solves math problems that stumped a Fields Medalist

Fields Medal winner Timothy Gowers said GPT 5.6 Pro solved two math problems he had spent significant time on, each on its first attempt. The result follows OpenAI's earlier claimed refutation of the Unit Distance Conjecture, part of a broader wave of AI-assisted advances in mathematics. Gowers warned of a "possible destruction of mathematical culture" if mathematicians stop building the expertise needed to understand such AI-derived results.

Why it matters: A Fields Medalist calling a model's output a first-try solution to a problem he couldn't crack is a stronger signal than typical benchmark claims, since it comes from someone qualified to independently verify the result rather than just trust it. It also surfaces a real unresolved tension: if AI increasingly does the proving, the human expertise needed to check and build on that work could erode even as output accelerates.