AI-Assisted Navier-Stokes Proof Sparks Controversy as Researcher Claims Two Major Problems Within 24 Hours

A researcher at NYU and a researcher at Anthropic released an AI-assisted proof on Mastodon claiming that the Navier-Stokes equations can exhibit finite-ti

On November 20, 2026, Tristan Buckmaster of New York University and Levent Alpöge, a researcher at Anthropic, released an AI-assisted proof on Mastodon, claiming that the Navier-Stokes equations can undergo finite-time blowup in the absence of external forces.

Factual Reconstruction

According to reports from news.lavx.hu, this proof targets the “artificial forcing” loophole for which OpenAI's September result had been criticized, attempting to demonstrate blowup under natural conditions. Buckmaster and Alpöge did not immediately release a formal certificate and only stated that follow-up would come. That same day, Alpöge also announced a claim regarding the Riemann Hypothesis.

Background shows that OpenAI's earlier proof was criticized for introducing specific external forces, which was said to be inconsistent with real fluid conditions; the new work directly responds to that criticism.

Mechanism Breakdown

Initial information mentions that the paper was completed jointly using Claude and Codex, based on 600,000 lines of Lean formal verification. In this process, Claude took on the role of generating extremely long formal code and checking logical consistency. Compared with OpenAI's approach, the main difference lies in whether dependence on external forces is completely eliminated.

Because the certificate has not yet been made public, the specific execution path and error-handling details of the Lean verification remain unconfirmed.

Industry Impact

This incident pushes the application of AI tools to Millennium Prize Problems toward stricter evaluation standards. With formal verification reaching 600,000 lines, it highlights the practical capability boundary of current models in long-context consistency.

The same researcher leading two claims within 24 hours reflects the accelerating pace of AI-assisted mathematical research and may also intensify disputes over academic priority.

Strategic Judgment

[Analysis, not fact] If the certificate ultimately passes peer review, AI's role in mathematical formalization will shift from an auxiliary tool to a core verification partner; if it does not pass, it may reinforce the industry consensus that “AI-generated content requires secondary human verification.”

[Analysis, not fact] The comparison with OpenAI's approach shows that after eliminating “artificial forcing,” models need to handle more complex boundary conditions without external forces, which places higher demands on the training data and verification processes of future AI mathematical systems.