Skip to content
Stage · Developing3 outlets · 3 articlesTech

Source reported

OpenAI says its model refuted ETH, a 1999 hypothesis behind hundreds of papers

OpenAI says its model refuted ETH, a 1999 hypothesis behind hundreds of papers — Newsmesis editorial illustration
— Newsmesis illustration

What happened

OpenAI said on Oct. 26 that an internal model found a disproof of the Exponential Time Hypothesis (ETH), an assumption about how hard Boolean satisfiability problems are that hundreds of computer science papers build on. The company says the model produced the result on Oct. 16 and that it was machine-checked in the Lean proof language the next day, and it has posted the write-up and proof on GitHub. The scientists who proposed ETH, plus MIT's Ryan Williams, are now organizing an independent examination and say their role is not a certification. Per TechCrunch, Anthropic published its own refutation of ETH hours after a company mathematician suggested OpenAI's solutions drew on earlier work.

What we know

TechCrunchtechcrunch.comEngadgetengadget.comThe Next Webthenextweb.com
  • OpenAI published a manuscript and Lean proof claiming ETH is false, per The Next Web and Engadget. ETH has been the basis of hundreds of papers in fields including databases, computational biology and geometry, per Engadget.
  • The examination group is led by Russell Impagliazzo and Ramamohan Paturi of UC San Diego, who first proposed ETH, and Ryan Williams of MIT, per The Next Web and TechCrunch. Their open letter says: "Our involvement should not be taken as certification of the claim."
  • Williams said the group signed no non-disclosure agreements and will not be paid. OpenAI agreed to release raw data and fund a conference for the research community, per The Next Web.
  • Williams told TechCrunch that parts of OpenAI's proof are routine but two or three are "alien-looking". He also said he does not think the algorithm is the best one possible.
  • Per TechCrunch, Anthropic released its own refutation of ETH after Anthropic mathematician Levent Alpöge suggested OpenAI's solutions were based on pre-existing work.
  • Williams told TechCrunch that OpenAI science lead Dan Roberts said the company had checked and was reasonably confident the result implies no cryptographic breaks.

What we don't know yet

  • Whether the refutation survives independent expert scrutiny. The examination group has only just begun and has not published a verdict.
  • Whether Alpöge's suggestion that the OpenAI proofs build on pre-existing work holds. TechCrunch reports it as a suggestion, and no source here documents the specific prior work.
  • Whether the result implies anything for P vs NP or for cryptographic key sizes. The cryptography check comes from OpenAI via Williams and has not been independently confirmed.
  • How credit should be assigned when AI finds results first. The three computer scientists raise this in their letter and offer no answer.

Why it matters

Researchers whose papers assume ETH may have to revisit conclusions about which problems are hard to solve quickly. If the refutation holds, it also tests how scientific fields vet and credit machine-generated proofs, with a funded conference and expert groups as OpenAI's first attempt at a process.

Claims

Related coverage

Hundreds of papers assume ETH. OpenAI says its model disproved it; checks have just begun

Tech ·

Four AI giants pledge one opt-out signal. Only Anthropic honors it today

Tech ·

Anthropic's Claude Duet agent opens to Pro-tier users later; OpenAI's Dots starts at $100 a month

Tech ·

Why you can trust this story

70%

Source map · 3 outlets / 3 articles

How this credibility score is calculated

  • Source reliability
  • Corroboration
  • Primary source
  • Claims checked against the source
  • No contradiction
  • Settled
  • Claim attribution
  • AI disclosure

Checked claims were verified against the cited source text. Weights are fixed and explainable.

Human accountability

  • 3 independent origins / 3 sources
  • Drafted by an automated system; a human editor approves every story before it is published.

Verification ledger · 7

  • Claim: OpenAI claims its model produced a refutation of the Exponential Time Hypothesis.
    Status: UnconfirmedVerification trailUnverifiedChecked by verification engine

    No automated check against the source text is available.

  • Claim: OpenAI says its model found the ETH disproof on Oct. 16 and the proof was checked in Lean the next day.
    Status: UnconfirmedVerification trailUnverifiedChecked by verification engine

    No automated check against the source text is available.

  • Claim: OpenAI published a manuscript on the ETH result on GitHub.
    Status: ConfirmedVerification trailVerifiedChecked by verification engine

    At least one checkable detail was matched against the full text of the cited source.

  • Claim: Impagliazzo, Paturi and Williams are organizing an independent examination of the claim and say their involvement is not certification.
    Status: ConfirmedVerification trailVerifiedChecked by verification engine

    At least one checkable detail was matched against the full text of the cited source.

  • Claim: OpenAI agreed to release raw data on the ETH result and fund a conference, per Williams and OpenAI.
    Status: ConfirmedVerification trailVerifiedChecked by verification engine

    At least one checkable detail was matched against the full text of the cited source.

  • Claim: Anthropic released its own refutation of ETH after OpenAI's announcement, per TechCrunch.
    Status: ConfirmedVerification trailVerifiedChecked by verification engine

    At least one checkable detail was matched against the full text of the cited source.

  • Claim: OpenAI told Williams it was reasonably confident the ETH result implies no cryptographic breaks.
    Status: UnconfirmedVerification trailUnverifiedChecked by verification engine

    No automated check against the source text is available.

Update log · 0

    AI accelerates. High-risk topics wait for a human editor.