GPT-5.6 Pro helps crack a mathematical hypothesis unsolved for 30 years

Four researchers have produced a counterexample disproving the HRT hypothesis in harmonic analysis, a problem that had stood unsolved for three decades. ChatGPT running on the GPT-5.6 Pro model helped develop the key configuration and proof strategy, though its reasoning contained gaps that required human correction. The result is one of the first cases where AI has meaningfully contributed to a significant mathematical breakthrough.