Fields Medalist says ChatGPT 5.5 Pro delivered "PhD-level" math research in under two hours with zero human help

Timothy Gowers says ChatGPT 5.5 Pro produced original, doctoral-level math proofs in under two hours.

The model first found a quadratic bound within minutes, then strengthened a generalized result from exponential to polynomial bounds.

Gowers and another researcher checked the proofs and said the main idea was original and mathematically significant.

If confirmed, the result suggests advanced AI may make genuine contributions to frontier mathematics.