OpenAI Says AI Solved Decades-Old Navier-Stokes Math Problem

OpenAI says it used thousands of AI agents to solve the Navier-Stokes existence and smoothness problem, a fluid dynamics puzzle that has stood for decades, according to Wired. The claim has set off a debate among mathematicians about whether brute-force computing threatens a field long treated as a creative discipline.

What did OpenAI claim to solve?

The Navier-Stokes equations date to the 19th century and describe how viscous fluids move, Wired reports. Engineers use them to model airflow for aircraft design. Mathematicians, separately, have spent decades trying to prove whether solutions to the equations always exist and remain smooth, a question known as the Navier-Stokes existence and smoothness problem. OpenAI says its agents solved it. Jared Speck, a mathematician at Vanderbilt University, told Wired that mathematicians' interest in the equations was never about engineering. "Mathematicians' main interest in the equations was certainly not engineering," Speck said, pointing instead to the "mathematical richness, the puzzle aspect of it."

Why has math been called an art form?

English mathematician G.H. Hardy made that case in his 1940 essay "A Mathematician's Apology," written during World War II to argue for math pursued for its own sake. "A mathematician, like a painter or poet, is a maker of patterns," Hardy wrote, comparing a painter's shapes and colors and a poet's words to a mathematician's ideas. Hardy held up number theory, the study of integers, as an example of beautiful, useless math, citing Carl Friedrich Gauss. Hardy's framing of math as art has held up, Wired notes, even though his specific example did not: number theory later became central to the encryption protocols that secure email and banking today.

Does the AI proof have practical use?

No, according to Wired. The Navier-Stokes puzzle OpenAI's agents solved will not help engineers build a more aerodynamic airplane wing. Speck compared it to studying the equations of a cylinder without expecting the work to improve cake-baking. The problem's appeal, he said, resembles the draw of Sudoku or chess: no utility beyond being fun and intellectually demanding. "When problems resist solution, they take on a bit of lore," Speck told Wired.

What are mathematicians worried about?

Juspreet Singh Sandhu, a mathematician at Colorado State University, told Wired that math is now facing what musicians and visual artists already confronted with AI tools such as Suno and Tilly Norwood. "The artists and the musicians have already gone through this," Sandhu said. Mathematicians typically build proofs through a deliberate, exploratory process akin to inventing a puzzle, Wired reports. OpenAI's approach relied on scale and speed instead, a shortcut mathematicians say risks undermining human understanding of why a proof works, not just that it works.

What happens next for AI in mathematics?

Wired does not report a next step from OpenAI or a peer-review timeline for the claimed proof. The dispute for now centers on process versus outcome: whether a correct answer produced by thousands of automated agents carries the same value as one built through the kind of patient, pattern-making work Hardy described eight decades ago.

Artiglio is A coming-soon iPhone chief of staff for briefings and drafts. Not on the App Store yet.