OpenAI's matrix multiplication ω≤9/4 proof extended to all fields, verified in Lean with GPT-6 assistance
thomasahle · x · 2026-10-07
A new open-source effort extends OpenAI's Lean formalization of the fast matrix multiplication exponent bound ω≤9/4 from the complex numbers to every field. Found and formalized by consumer-grade GPT-6 Astra and GPT-6.1 Sol under Sela Navot's direction, the proof passed Lean verification and is released under Apache-2.0.
- Why it's notable: the new FMM paper takes a strikingly different approach from prior work—instead of powering the CW tensor and trimming it, it defines a potential function on tensors and proves the result by contradiction via the simple convolution tensor.
- Implication: the generalization strengthens the result's universality, and the author speculates n^{9/4} may actually be the right exponent for matrix multiplication.
Related event: OpenAI's Fast Matrix Multiplication Proof Generalized to Arbitrary Fields(2 posts)→
More from Research
- Live Lean 4 benchmark tracks Riemann Hypothesis proof compression across models and agents — flngr · 2026-11-22
- Kalshi traders now put 54% odds on P vs NP being resolved before 2030 — ricklamers · 2026-11-22
- OpenAI proof dump shows first superlinear circuit lower bound, but progress stays slow — thomasahle · 2026-11-22
- One photo, 8 VLAs, only 3 correct: ERQA embodied-reasoning benchmark questioned — giffmana · 2026-11-22
- Fleuret: treat AI as oracles, not explainers — human-comprehensible math will be negligible — francoisfleuret · 2026-11-22
- Vovk paper shows probabilistically nonrandom objects have vanishing universal measure — _onionesque · 2026-11-22