OpenAI Claims Navier-Stokes Math Proof as Users Exit ChatGPT

OpenAI announced that an internal AI system has produced a solution to the Navier-Stokes existence and smoothness problem, a $1 million Millennium Prize challenge, while simultaneously facing a massive user exodus following its decision to accept a Pentagon contract [1, 2]. The dual developments highlight a week of extreme volatility for the company, as it balances a major technical milestone against a public revolt over its role in military operations [1, 2].

The math breakthrough

OpenAI’s internal system utilized approximately 10,000 AI agents working over 88 hours to conclude that fluid motion can develop a singularity—a point where speeds grow without limit—in finite time [2]. The process involved the exchange of 2.7 million messages and the generation of roughly 130 billion output tokens specifically for the Navier-Stokes problem [2]. An additional 17 hours of work by a separate model formalized the result using the Lean verification system [2].

Despite the significance of the proof, the Clay Mathematics Institute has not independently verified or accepted the solution, and OpenAI stated it does not intend to claim the $1 million prize [2]. The achievement follows a trajectory of smaller successes, including a disproof related to the Euler equations produced by 100 agents over 50 hours [2].

The user revolt

While the company touted its technical progress, it faced a sharp decline in its consumer base following reports that it accepted a deal with the Department of Defense for unrestricted access to its models [1]. On February 28, 2026, ChatGPT uninstalls in the United States surged 295% above the typical daily rate of 9% [1]. The "QuitGPT" movement claims 1.5 million users have abandoned the platform, with one-star reviews jumping 775% in a single day [1].

The backlash benefited competitor Anthropic, whose Claude app saw downloads jump 37% on Friday and 51% on Saturday, eventually displacing ChatGPT as the number one app on the US App Store [1]. OpenAI CEO Sam Altman later acknowledged the rollout was "sloppy" and moved to amend the contract, adding language to bar the use of its systems for domestic surveillance [1]. However, the narrative remains complicated by reports that Anthropic’s Claude had previously been integrated into the Palantir Maven platform, which was used in missile strikes in Iran [1].

What to watch

• Independent verification: Whether the mathematical community or the Clay Mathematics Institute validates the Navier-Stokes proof produced by the AI agents [2].
• Platform retention: Whether the surge in Claude’s user base persists or if ChatGPT’s user numbers stabilize following the contract modifications [1].
• Government policy: Further adjustments to the terms of service for AI companies working with the Department of Defense as public pressure regarding autonomous weapons continues [1].

The tension between OpenAI’s pursuit of high-level scientific breakthroughs and its integration into military infrastructure has created a new market variable: users are now actively weighing corporate ethics against product performance. Whether this shift in consumer behavior permanently alters the competitive landscape for AI remains the central question for the industry.