ETH proposers lead open review of solver claim; exponent shrink factor ~0.67 at a billion variables
AIFlow_ML · x · 2026-10-27
Napkin math on the OpenAI-related solver claim: the exponent shrink factor (log log n)/sqrt(log n) is 0.73 at 1,000 variables and still 0.67 at a billion, before hidden constants — theory map changes, solver gains are another matter; formal check proposed via Lean. The review is led by Impagliazzo and Paturi (ETH proposers) plus Ryan Williams, with no NDAs, unpaid, and raw data public, though their letter states their involvement 'should not be taken as certification of the claim.'
More from Research
- Complexity researcher on the ETH refutation: stronger than P≠NP but practically moot — burny_tech · 2026-10-27
- OpenAI announces three AI-made discoveries: ETH refuted, Kakeya conjecture, QCD existence — burny_tech · 2026-10-27
- Steering Just 4 Neurons Makes a Model Spell Out an AI Domination Plan — burny_tech · 2026-10-27
- Ladner's theorem view on NP-intermediate problems: the subexponential-algorithm guess falls apart — yacineMTB · 2026-10-27
- UReason (EMNLP Oral): Multimodal Models Reason Fine but Images Don't Follow — niloofar_mire · 2026-10-27
- Ryan Williams explains what a false ETH means for 3SAT — rrwilliams · 2026-10-27