OpenAI’s Math Triumph Is Now a Fight Over Who Deserves the Breakthrough

OpenAI’s Math Triumph Is Now a Fight Over Who Deserves the Breakthrough
OpenAI’s claimed solution to the 90-year-old Navier–Stokes existence and smoothness problem has put a spectacular AI milestone and a sharp dispute over research credit on the same collision course.

The timeline began on August 28, when OpenAI says it started training an internal model more capable than GPT-6 Astra. On September 1, after hearing rumors that two Millennium Prize problems had been resolved, it turned a coordinated agent system loose on the remaining problems. About 10,000 agents eventually focused on Navier–Stokes; OpenAI says they reached a proof after 88 hours, followed by 17 hours of Lean formalization and verification. The company says the proof shows a smooth three-dimensional fluid can develop a finite-time singularity—resolving the prize problem’s “C” and “D” formulations.

But the celebratory account collided with work already underway. On Monday, NYU mathematician Tristan Buckmaster and Anthropic researcher Levent Alpöge published results on a related forced-Euler problem after nearly a year of work using Codex and Claude. Buckmaster said information about their progress had reached OpenAI, and argued their chosen route was unusually specific: “Almost nobody else I know of was working on it.”

He also alleged that OpenAI’s Sébastien Bubeck sought to remove Alpöge from authorship because of his Anthropic affiliation—an allegation OpenAI has rejected. The criticism spread quickly: a repost amplified by Yann LeCun sarcastically congratulated OpenAI for not letting “any good customer transcripts go unmined.”

OpenAI’s answer is categorical on direct access but less absolute on model training. It says neither employees nor agents saw the researchers’ work before it was public, while conceding it “cannot rule out” that de-identified data from product use improved its models. It also says the teams’ proofs, and their Euler results, differ. CEO Sam Altman defended the conduct of the OpenAI team, saying they had initially hoped for “a joint release.”

The immediate argument is about attribution; the larger one is about power. Buckmaster and Alpöge’s progress illustrates human–AI collaboration, while OpenAI’s reported million-dollar-scale compute bill illustrates how quickly frontier labs can turn an academic race into an industrial one. As one account of the episode put it, “very few mathematicians will have resources of that scale.”

Continue reading https://foxvector.com/stories/01a08efe-ba91-0ff7-7070-3f8025bf12d7