On Sep 8, 2026, OpenAI published that an internal AI system produced a Lean-formalized proof that Navier–Stokes can develop a singularity in finite time, establishing Clay statements C and D in OpenAI’s account. OpenAI reports ~10,000 concurrent agents, ~88 hours to resolution after a Sep 1 launch, Lean via GPT-6 Astra in +17 hours, and large message/token counts; it also reports an unforced Euler regularity disproof. OpenAI says it will not claim the Millennium Prize, recognizes concurrent forced-Euler priority of Tristan Buckmaster (NYU) and Levent Alpöge (Anthropic), and states it cannot rule out de-identified product-usage data helping models. Independent Clay/math acceptance is not verified in this pack; credit-dispute coverage exists in secondary press—ElevateU does not adjudicate.
A Lean-backed Millennium Prize claim from a frontier lab is high-executive and high-novelty signal for AI scientific capability—and for credit, collaboration, and verification norms. Treat the result as OpenAI’s published claim until independent acceptance is established.
- 01
OpenAI claims an internal system produced a Navier–Stokes singularity proof with Lean formalization (Clay C and D, per OpenAI).
- 02
Internal model described as significantly more capable than GPT-6 Astra; Lean formalization via GPT-6 Astra (+17 hours).
- 03
~10,000 concurrent agents; ~88 hours after Sep 1 launch; NS ~2.7M messages / ~130B output tokens (OpenAI figures).
- 04
Also reports unforced Euler regularity disproof (~50 hours / ~100 agents).
- 05
OpenAI: will not claim the Millennium Prize.
- 06
Recognizes Buckmaster & Alpöge priority on forced Euler; cannot rule out de-identified usage data helping models.
- 07
FLAG: Independent Clay/math acceptance not verified in this pack; note credit-dispute uncertainty—do not adjudicate.
Last verified: 2026-09-09T13:30:00Z · Speaker predictions remain attributed to the speaker. ElevateU analysis is labeled separately from source-supported fact.