OpenAI’s Math Breakthrough Collides With a Bitter Fight Over Who Got There First

OpenAI’s Math Breakthrough Collides With a Bitter Fight Over Who Got There First
OpenAI’s claimed solution to the Navier–Stokes existence and smoothness problem has turned a historic mathematical announcement into a dispute over provenance, credit and the power imbalance between researchers and the companies whose AI tools they use.

The chronology began on August 28, when OpenAI started training an unreleased internal model. On September 1, the company says, it heard rumors that two Millennium Prize problems had been resolved and launched coordinated agents against the remaining problems. Roughly 10,000 agents eventually worked on Navier–Stokes; OpenAI says they reached a solution on September 5, after 88 hours, before Lean verification followed. The company’s proof argues that an initially smooth three-dimensional fluid can develop a finite-time singularity — a breakdown in the equations used to model flows from weather to blood circulation.

Meanwhile, NYU mathematician Tristan Buckmaster and Anthropic researcher Levent Alpöge had spent nearly a year working on related questions with publicly available AI models. On Monday, Buckmaster released work showing breakdown in a forced version of the Euler equations, a major but narrower result. He said information about their progress had reached OpenAI and questioned whether the company’s agents had benefited from their Codex work.

OpenAI says it contacted the pair only after completing its proof, believing they had independently solved Navier–Stokes and hoping for a concurrent release. It denies that researchers or agents saw the pair’s work before publication, while conceding it “cannot rule out” that de-identified product-usage data helped improve its models. The company also says the proofs and the underlying Euler results differ.

The conflict widened over Buckmaster’s allegation that Alpöge, an Anthropic employee, was pressed to step aside from authorship. Altman rejected that account, saying OpenAI had sought collaboration and that its staff acted with “integrity and generosity throughout.”

For OpenAI, the episode showcases a system capable of attacking one of mathematics’ most durable open problems at enormous scale. For its critics, it exposes a sharper question: when a lab can turn customers’ research environment into a multimillion-dollar competitive advantage, who controls the path to discovery?

Continue reading https://foxvector.com/stories/01a085fa-7424-3960-7067-06ddd46e6948