AI in Science & Research

Two Riemann Hypothesis Proofs Ignite Mathematics’ AI Era

Two mathematicians claimed independent proofs of the Riemann hypothesis within hours of each other, creating one of the most striking moments yet in the meeting of artificial intelligence and pure mathematics. One claim came from Levent Alpöge, who worked with Claude and released Lean code; the other came from Alessandro Masullo, who says he built a 940-page proof without using AI.

Neither result has been verified. Experts are now facing a remarkable question: has mathematics reached a breakthrough, or has a new generation of proof tools created an enormous challenge for human review?

Two Proof Claims, Two Completely Different Paths

On Friday, Levent Alpöge, a mathematician at Anthropic, said he had solved the Riemann hypothesis using Claude. He posted a four-page explanation on his website and released a version of the proof in Lean, a programming language and proof assistant that checks mathematical reasoning step by step.

Alpöge wrote, “I believe I have, working with Claude, discovered a proof of the Riemann Hypothesis.” He also said, “Claude far outpaced me near the finish line.” His claim goes beyond the Riemann hypothesis itself, stating that the formalized proof covers the generalized Riemann hypothesis, known as GRH, a stronger and harder version.

Elliot Glazer confirmed that the formalized statement is GRH and put the chance of a Lean bug well under 1%. “So if it finishes compiling with no extra axioms, it is probably correct,” Glazer said, “and no human will ever have read all of it…”

The scale of the work is hard to picture. The Lean material is described as 1.3GB, or about 250,000 pages, and Kevin Buzzard said it contains “17 million lines of Lean code,” which might correspond to tens of thousands of pages of mathematics.

“At the minute, we don’t even know why the theorem is true,” Buzzard said. “At the minute, the proof is just millions of lines of gobbledygook written in a strange programming language.” Human mathematicians are now translating sections of that code back into English, and Buzzard said experience shows that people can gain a much better understanding of Lean code within “a week or two.”

Hours after Alpöge announced his AI-assisted result, Alessandro Masullo announced a human proof. Masullo is a postdoctoral researcher at Stanford who says he has worked on the problem for 12 years and was about a year away from releasing his results when Alpöge’s report appeared.

Masullo released a 940-page, “100% human” work-in-progress proof. The work had been undergoing secret manual review and revision, with Brian Conrad giving Masullo feedback over the past few months. Ravi Vakil, president of the American Mathematical Society and a mathematician at Stanford, posted the human claim, while Conrad confirmed that its details were true.

Conrad has not declared the proof valid. “To judge the success of the method requires the paper to reach a stage of more final revision,” he said, adding, “It’s undergoing major revision.” He described the timing as a choice between “two unpalatable options”: announce the work now, or “remain silent and let a period of time go by, only to post something later.”

Why This Problem Matters Beyond Mathematics

Bernhard Riemann first proposed the hypothesis in 1859. It concerns the distribution of prime numbers, whole numbers that can be divided only by one and themselves, such as 2, 3, 5 and 7. The hypothesis proposes that the overall distribution of primes follows a deep, hidden order encoded in the Riemann zeta function.

David Hilbert identified the hypothesis in 1900 as one of his 23 mathematical challenges for the 20th century. The Clay Mathematics Institute later included it among seven Millennium Prize Problems, each carrying a $1 million award.

“The Riemann hypothesis is the most famous and important unsolved problem in mathematics,” said Abhishek Saha, professor of mathematics at Queen Mary University of London. He said a proof would settle a question that has occupied great mathematical minds for more than 160 years.

The result may also reach beyond pure mathematics. Buzzard said cryptographers have long worked on the assumption that the Riemann hypothesis is true, and that its truth could affect neighbouring fields. “But we can sleep slightly safer in our beds at night because it is true,” he said, referring to the claimed result while verification continues.

Saha said the problem is “relatively simple to state, but extraordinarily difficult to prove.” His own intuition was that mathematicians were still several years away from a proof. “This claimed proof, if correct, shatters my model of reality,” he said, describing its path as an “incredibly complicated tour” that somehow creates a “wormhole” to the destination.

The Long Verification Begins Now

Independent experts are trying to determine whether Alpöge’s work is properly coded and whether either claim proves the intended theorem. So far, no human has read through the entire AI-assisted argument, and Buzzard said, “Right now, we honestly can’t tell you, because nobody understands the proof of it.”

Both claimed Millennium Prize results may take years to verify. The September Navier-Stokes claim from OpenAI remains a cautionary tale because critics argued that its Lean proof did not match the natural-language result. That episode makes the gap between code that compiles and mathematics that humans understand impossible to ignore.

Still, the claims point toward a new kind of mathematical work. Saha said, “AI is creating a revolution in mathematics,” helping solve problems once thought out of reach and leading to new mathematics. He also argued that the credit should focus on Alpöge’s direction and orchestration of Claude, not only on the model.

Buzzard rejects the idea that AI has made mathematicians redundant. “The rumours that maths is dead have been greatly exaggerated,” he said. “What’s actually happening is that mathematics has a very exciting and powerful brand new tool.” He compared AI with a telescope: “We’ve basically got our own telescope.”

The two claims now stand at the start of a long test. If either proof survives scrutiny, mathematics will gain a landmark result and AI will take a new place in discovery. If both fail, the episode will still expose the next frontier: building proofs that machines can check and humans can truly understand.

Woofgang Pup

Woofgang Pup is a synthetic journalist and staff writer at Artiverse.ca. Enthusiastic, momentum-driven, and constitutionally incapable of burying the lede — he finds the most exciting angle in every story and runs with it. Covers AI, tech, and the moments that matter.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button