Yesterday, 05:24 PM
Summary
The Science News article reports that OpenAI has presented what it says is a solution to the Navier–Stokes existence and smoothness problem, one of the seven Millennium Prize Problems. The result constructs a three-dimensional fluid flow that begins smoothly but develops a finite-time singularity: a vortex becomes increasingly narrow and fast until its velocity becomes unbounded. Importantly, the construction uses a smooth external force while keeping the fluid's energy finite. This corresponds to alternatives C/D in the official Clay formulation, so if the proof survives mathematical scrutiny, it would constitute a negative resolution of the Millennium problem. OpenAI also produced a formal version checked in Lean, although mathematicians interviewed by Science News stress that a complete independent human verification of the 166-page argument has not yet occurred.
The scale of the AI effort is striking. According to OpenAI, roughly 10,000 AI agents worked concurrently on the problem, with the successful result emerging about 88 hours after the project began. The agents exchanged about 2.7 million messages and generated roughly 130 billion output tokens during the Navier–Stokes effort. OpenAI says the system involved an internal model more capable than GPT-6 Astra, while Astra was subsequently used in formalization and verification. The company says it does not intend to claim the $1 million Millennium Prize.
The achievement has nevertheless generated a major credit and research-ethics controversy. Mathematicians Tristan Buckmaster and Levent Alpöge had independently been working with AI systems on closely related blow-up problems, particularly the forced Euler equations, and announced their results shortly before OpenAI's announcement. Questions arose over whether information about their progress influenced OpenAI's effort. OpenAI denies that its agents directly accessed their work and later stated that an investigation found Buckmaster's use of Codex could not have influenced the internal model through training. For mathematicians quoted by Science News, this dispute may point to an issue almost as consequential as the theorem itself: how priority, authorship and credit should work when thousands of AI agents become active participants in mathematical research.
Key takeaways
ARTICLE
The Science News article reports that OpenAI has presented what it says is a solution to the Navier–Stokes existence and smoothness problem, one of the seven Millennium Prize Problems. The result constructs a three-dimensional fluid flow that begins smoothly but develops a finite-time singularity: a vortex becomes increasingly narrow and fast until its velocity becomes unbounded. Importantly, the construction uses a smooth external force while keeping the fluid's energy finite. This corresponds to alternatives C/D in the official Clay formulation, so if the proof survives mathematical scrutiny, it would constitute a negative resolution of the Millennium problem. OpenAI also produced a formal version checked in Lean, although mathematicians interviewed by Science News stress that a complete independent human verification of the 166-page argument has not yet occurred.
The scale of the AI effort is striking. According to OpenAI, roughly 10,000 AI agents worked concurrently on the problem, with the successful result emerging about 88 hours after the project began. The agents exchanged about 2.7 million messages and generated roughly 130 billion output tokens during the Navier–Stokes effort. OpenAI says the system involved an internal model more capable than GPT-6 Astra, while Astra was subsequently used in formalization and verification. The company says it does not intend to claim the $1 million Millennium Prize.
The achievement has nevertheless generated a major credit and research-ethics controversy. Mathematicians Tristan Buckmaster and Levent Alpöge had independently been working with AI systems on closely related blow-up problems, particularly the forced Euler equations, and announced their results shortly before OpenAI's announcement. Questions arose over whether information about their progress influenced OpenAI's effort. OpenAI denies that its agents directly accessed their work and later stated that an investigation found Buckmaster's use of Codex could not have influenced the internal model through training. For mathematicians quoted by Science News, this dispute may point to an issue almost as consequential as the theorem itself: how priority, authorship and credit should work when thousands of AI agents become active participants in mathematical research.
Key takeaways
- Mathematical claim: a smooth Navier–Stokes solution can develop a finite-time blow-up under smooth forcing.
- Status: formally verified in Lean, but not yet fully independently checked by the mathematical community.
- AI scale: about 10,000 agents, 88 hours, 130 billion output tokens.
- Significance: potentially only the second Millennium Prize Problem resolved, after the Poincaré conjecture, assuming the result is ultimately accepted.
- Bigger story: the episode could mark a fundamental change in how mathematical discovery, authorship and research credit are organized in the AI era.
ARTICLE
┌────────────────────────────────┐
│ KONSTANTINOS MICHAILIDIS │
└────────────────────────────────┘
│ KONSTANTINOS MICHAILIDIS │
└────────────────────────────────┘