The Exponential Time Hypothesis (ETH), a core assumption underpinning hundreds of computer science papers and critical to computational complexity theory, faces a potential refutation by an OpenAI-developed AI model. The claim, announced in October 2026, is undergoing careful scrutiny by prominent researchers directly involved with the original ETH proposal.
- OpenAI AI model claims to disprove the Exponential Time Hypothesis
- Independent review underway by ETH originators and MIT experts
- AI also produced breakthroughs in mathematics and physics, including the local smoothing and Yang-Mills conjectures
What happened
OpenAI announced that an internal AI model has found a claimed refutation of the Exponential Time Hypothesis (ETH), a widely accepted conjecture concerning the difficulty of solving Boolean satisfiability problems. The discovery was reportedly made on October 16, 2026, and was verified with a Lean formal proof system the following day. OpenAI published the AI's write-up and the formal proof on GitHub, making the claim transparent for independent verification.
This announcement coincided with additional AI achievements by OpenAI in mathematics and physics, including proofs related to the local smoothing conjecture, Kakeya conjecture, and a major proof in quantum field theory addressing the existence of pure Yang-Mills theory in four dimensions. These results have drawn extensive interest across computer science, mathematics, and physics communities.
Why it matters
The Exponential Time Hypothesis has been a foundational assumption in theoretical computer science, shaping our understanding of computational hardness and influencing results in graph theory, databases, and biology. If the AI model’s refutation holds true, it would necessitate fundamental revisions in complexity theory and open new directions for algorithms and computational limits.
The involvement of key figures such as Russell Impagliazzo, Ramamohan Paturi, and Ryan Williams—original ETH proposers and prominent complexity theorists—indicates the seriousness of the claim. They have initiated a transparent, independent examination without endorsing the result outright and called for open data and community engagement, illustrating emerging challenges and opportunities around AI's role in mathematical discovery and intellectual credit.
What to watch next
The verification process by independent experts is ongoing and will be critical to confirm the validity of the AI-generated refutation. The outcome may redefine theoretical boundaries and influence computational complexity research globally. Meanwhile, discussions are underway about attribution and recognition, as AI technologies increasingly produce original research outputs.
Further developments are also anticipated from the related AI discoveries in mathematics and physics, where OpenAI has supported workshops and conferences through donations to advance collective understanding. How the academic community incorporates AI-assisted findings, balances credit, and adapts peer review will be important to monitor moving forward.