OpenAI announced that its AI system has refuted a 30-year-old assumption in computer science, as one of three new discoveries across mathematics, computer science and physics.

What it means

OpenAI has announced that its AI system produced a disproof of the Exponential Time Hypothesis (ETH), a 30-year-old conjecture in computational complexity that states certain classes of problems require exponential time to solve. This is one of three new discoveries the company claims its AI made across mathematics, computer science and physics, with the others involving a novel approach in commutative algebra and a question in quantum gravity. The AI generated the disproof in natural language and then formalized it in the Lean interactive theorem prover, and OpenAI says it shared the result with experts for verification.

Accuracy & sourcesGrounded

The claim rests solely on OpenAI's own announcement, which has not yet been independently peer-reviewed or replicated. The reporting (e.g., New Scientist, Washington Times) is reputable and accurately conveys the company's statement, but the underlying claim—that the Exponential Time Hypothesis is indeed refuted—remains a contested and unverified assertion. An external group is reportedly organizing an open examination of the result, so the academic community has not yet confirmed it.

How new