OpenAI Claims AI Model Disproves Exponential Time Hypothesis, Resolves Kakeya Conjecture, and Proves Quantum Field Existence

Key points

• OpenAI claims its AI model disproved the Exponential Time Hypothesis (ETH) on October 16, 2026, verified in Lean on October 17.
• The ETH refutation provides a deterministic algorithm for counting satisfying assignments of fixed-width CNF formulas in subexponential time.
• OpenAI also proved the local smoothing conjecture in all dimensions, resolving the Kakeya and Stein restriction conjectures.
• A rigorous proof for the existence of certain four-dimensional quantum field theories, including QCD and pure Yang-Mills theory, was produced.
• Independent expert groups led by Impagliazzo, Paturi, Williams (computer science) and Gross (physics) will examine the results.
• OpenAI released ETH and local smoothing proofs on GitHub but withheld physics materials for the expert group to write up.

Background

This announcement follows OpenAI's October 7, 2026, release of over 700 AI-generated mathematical manuscripts, which included claims of solving 372 open problems but also faced retraction and community backlash over verification and transparency. The current release marks a shift in OpenAI's approach, involving external expert 'shepherds' to guide the interpretation and verification of AI-generated results, a change prompted by criticism from an advisory panel of math advisers.

How outlets are covering it

TNW and The Verge emphasize the significance of the three breakthroughs and OpenAI's new collaboration model with external experts. The Verge notes that OpenAI's advisory panel of math advisers recently recommended against designating certain people to receive output first, citing unequal access and biased selection. Gizmodo highlights the skepticism surrounding the ETH claim, noting that 3-SAT is widely believed not to be solvable in polynomial time, and questions the reliability of Lean verification, referencing previous concerns about OpenAI's Navier-Stokes proof. Kingy AI provides a detailed technical breakdown, noting that the Lean formalization for ETH has a weaker quantitative bound than the manuscript, and stresses the need for independent scrutiny. All sources agree on the core claims but differ in their emphasis on the verification process and the potential impact on academic credit and research norms.

Why it matters

If verified, these results would fundamentally alter theoretical computer science, harmonic analysis, and mathematical physics. The refutation of ETH would invalidate hundreds of papers relying on it for hardness proofs. The resolution of the Kakeya conjecture in all dimensions would settle a long-standing problem in geometry and harmonic analysis. The existence proofs for quantum field theories would address parts of the Clay Mathematics Institute's Millennium Prize problems. The announcement also raises critical questions about how AI-generated discoveries should be credited, verified, and integrated into the scientific process, potentially reshaping academic norms and funding structures.

What to watch

The independent expert groups will examine and explain the ETH and local smoothing proofs, with Impagliazzo, Paturi, and Williams stating their involvement is not an endorsement. The physics group, led by David Gross, will write up the quantum field theory results without releasing the AI-produced materials initially. OpenAI has agreed to release all raw data on the ETH result and fund a conference for the research community. The scientific community will need to verify the Lean proofs and assess the broader implications of the results.