OpenAI and Anthropic both say their AI disproved a 27-year-old computer science rule.
The exponential time hypothesis says some logic puzzles have no real shortcut.
Every new character multiplies the guesses. That is exponential time.
Impagliazzo, Paturi and Zane proposed it. Hundreds of papers built on it.
Its model found the disproof on October 16. A machine checked it on October 17.
A Claude model found the idea on October 16 too, with no human help.
The new method only beats brute force past 10 billion variables. Useless in practice.
No peer review. No human has fully checked either proof yet.
The scientists who created the rule are now checking whether it really fell.
One professor says 50 of his papers may be pointless. He calls it cool and terrible.
OpenAI checked: the new method does not break internet security.
What the proofs say, what is verified, and what happens next.
Get the full breakdown, key numbers, and what happens next.
Read on Rive News