The OpenAI Podcast: How AI Cracked an 80-Year Math Mystery

A reasoning model just did what seemed impossible: it helped crack an 80-year-old math problem. In this condensed 6-minute summary of The OpenAI Podcast, Andrew Mayne talks with OpenAI researchers Alexander Wei, Hongxun Wu, and Lijie Chen about the breakthrough, the role of test-time compute, and why more thinking can mean more correctness. They discuss the model’s success on math, the International Math Olympiad, the Erdős unit distance conjecture, and what it means when AI becomes a true research partner in artificial intelligence and machine learning. You’ll also hear how the team tested the proof, why the result shocked mathematicians, and what this could mean for education, lifelong learning, P versus NP, and future discoveries in science and philosophy. Listen now to get the key ideas in minutes.