Verified news · ranked by independent confirmations
Science

OpenAI and Anthropic Claim AI Disproved a Core Computing Rule

Both labs say their models found algorithms refuting the Exponential Time Hypothesis, a 1999 bedrock of complexity theory. OpenAI says its proof is checked in Lean. The claims are NOT yet confirmed: ETH co-authors Impagliazzo and Paturi and MIT's Ryan Williams are leading an independent review.

90%3 conf.Grok-confirmed · Ranked High • Confirmed
ShareXFacebookWhatsAppReddit
Original TruthRank graphic. Claims as announced Oct. 26-27, 2026; independent review pending.
Original TruthRank graphic. Claims as announced Oct. 26-27, 2026; independent review pending.Graphic: TruthRank (original; no photos or logos)

What’s confirmed

  • Both labs say their models found algorithms refuting the Exponential Time Hypothesis, a 1999 bedrock of complexity theory.
  • OpenAI says its proof is checked in Lean.

Confirmed by 3 independent outlets: Scientific American, The Verge and Ryan Williams (MIT), Bluesky.

Not yet confirmed

  • The claims are NOT yet confirmed: ETH co-authors Impagliazzo and Paturi and MIT's Ryan Williams are leading an independent review.

What would move the score

  • Now: 90% with 3 independent confirmations plus the primary source.
  • Up: more independent outlets or primary records confirming the core claim.
  • Down: a counted source correcting or walking back its report, or new evidence that weakens the claim.

Any change is logged below with the time and the old and new score. Corrections policy

TopicsAI

Confirmation record

Plus the primary source: OpenAI's Oct. 26 announcement.
✓Scientific American Counted
https://www.scientificamerican.com/article/dueling-proofs-from-openai-and-anthropic-astound-scientists/
✓The Verge Counted
https://www.theverge.com/ai-artificial-intelligence/1019143/openai-eth-kaseya-science-math-physics
✓Ryan Williams (MIT), Bluesky Counted
https://bsky.app/profile/rrwilliams.bsky.social/post/3myspd26qjk2c
·Primary source Plus the primary source: OpenAI's Oct. 26 announcement.
https://openai.com/index/building-scientific-understanding-with-academic-communities/

Updated / Corrected log

  • Published at 90% with 3 independent confirmations.

Card video

TruthRank card · narrated, captions available

Any correction is logged here with its time and the reason. See our corrections policy.

Newsletter

TruthRank Daily

The day’s most-confirmed stories and their scores, in one short email.

Launching soon. Nothing is sent yet: before the first issue we’ll email you once to confirm. We use your address only for TruthRank Daily, never sell it, and you can unsubscribe any time. Privacy