An OpenAI model solved a famous math problem that stumped humans for 80 years

OpenAI just announced something wild. One of their internal AI models disproved the Erdős unit distance conjecture, a famous geometry problem that's stumped mathematicians for 80 years.

This isn't just another benchmark score. Tim Gowers, who won the Fields Medal (basically the Nobel Prize of math), called it "a milestone in AI mathematics." That's high praise from someone who's seen plenty of overhyped AI claims.

What makes this different is that the AI worked autonomously. University of Toronto professor Daniel Litt said this is the first AI-generated result he finds "exciting in itself, as opposed to as a leading indicator." Translation: the math itself matters, not just what it signals about AI progress.

OpenAI gave several mathematicians early access to review the proof before going public. That's a smart move, since AI-generated proofs need serious verification before anyone can trust them.

For anyone building with AI, this matters because it shows these models moving beyond pattern matching into genuine reasoning. If AI can crack problems that stumped humans for decades, the ceiling for what's possible in your work just got a lot higher.

The announcement came in mid-May, though OpenAI hasn't shared which specific model produced the result or when it might become publicly available. For now, it's a proof of concept that the frontier of AI capabilities keeps pushing forward.