AI Trends

Anthropic Mathematician Says He Cracked the Generalized Riemann Hypothesis With Claude

Cailian PressUpdated Nov 21 03:04

AI summary

Anthropic mathematician Levent Alpöge says he worked with Claude to find a proof of the Generalized Riemann Hypothesis and used Lean to formally encode the mathematical reasoning. Other experts still need to verify whether the encoding and derivation are valid; if confirmed, a large number of mathematical results premised on the conjecture would thereby be established.

Why read this

AI score 72

The material presents a specific case of Claude participating in mathematical research and Lean formalization, while also explaining the key steps for experts to verify the result.

Topics

Tags

Body · AI translation

CLS Financial News Agency

Open App

Breakthrough in a Century-Old Unsolved Problem? Anthropic Mathematician Claims to Have Solved the Riemann Hypothesis

CLS, Zhao Hao; Editor: Lin Qi; November 21, 2026, 02:10

① Anthropic mathematician and employee Levent Alpöge claims that, with the help of a large language model, he has successfully proved mathematics’ most famous problem—the Riemann Hypothesis; ② If valid, the proof would establish numerous mathematical results premised on this important conjecture, with potential implications extending far beyond mathematics.

CLS, November 21 (Editor: Zhao Hao)—Anthropic mathematician and employee Levent Alpöge claims that, with the help of a large language model (LLM), he has successfully proved mathematics’ most famous problem—the Riemann Hypothesis.

“I believe that, working together with Claude, I have found a proof of the Riemann Hypothesis,” Alpöge wrote in a statement published Friday on his personal website.

The Riemann Hypothesis was first proposed in 1859 by mathematician Bernhard Riemann. It holds that although the distribution of prime numbers along the number line may initially appear random, it is actually governed by an intricate pattern of profound mathematical significance.

The Riemann Hypothesis subsequently became a guiding light for the entire field of number theory and mathematics’ most famous unsolved problem. It is also one of the select mathematical problems for which the Clay Mathematics Institute has offered a $1 million prize for a solution.

Progress on the Riemann Hypothesis has long been so rare that most leading mathematicians have devoted little serious time to the problem, because no clear path to proving it has ever been found.

The first sign of a major breakthrough in the problem emerged in October this year, when OpenAI released a large collection of mathematical research results, including one proof that made significant progress toward understanding patterns in the distribution of prime numbers.

Proof Completed with Claude, Mathematical Reasoning Verified Using Lean

Now, this long mathematical campaign finally appears to be approaching its conclusion.

Alpöge published the proof he completed with Claude’s assistance and formally encoded it using the Lean programming language. Lean is a tool capable of verifying the correctness of mathematical arguments step by step.

Although Lean itself may also have issues, other experts should not need long to determine whether the relevant conjecture has been encoded correctly and whether the entire proof holds.

If ultimately confirmed, it would provide a resounding conclusion to a year of rapid advances in artificial intelligence’s mathematical capabilities.

Alpöge said that his proof actually establishes the Generalized Riemann Hypothesis.

The Generalized Riemann Hypothesis is an extraordinarily sweeping extension of Riemann’s original conjecture. It implies that a vast number of as-yet-undiscovered deep structures may underlie the entire mathematical system.

If valid, the result would have an immeasurably transformative impact on mathematics and immediately establish numerous mathematical results premised on this important conjecture.

“If this proof is correct and can be understood by people, it will establish a complete set of correspondences between different branches of mathematics where previously the connections between them were only partial,” mathematician Hector Pasten said.

“It is a bit like the Rosetta Stone of mathematics,” he added. The Rosetta Stone is an important artifact that helped scholars decipher ancient Egyptian hieroglyphs. Pasten used it as a metaphor for the crucial role the proof could play in bridging different fields of mathematics.

Alpöge’s statement carried a note of melancholy. He acknowledged that many mathematicians might feel a sense of loss because of the breakthrough, while also expressing optimism about the future.

“It is natural to feel that this is some kind of ending,” he wrote. “For me, it means losing one of my most important guiding lights.”

“But what I saw while making progress astonished me so deeply that I believe this is only the dawn of an even greater era across all of science.”

It is worth noting that Alpöge had already attracted media attention on several previous occasions.

In July this year, he posted on social media a Claude-generated solution to the famous mathematical problem known as the Jacobian Conjecture, kicking off an unprecedented summer of breakthroughs in mathematics.

In September, he again became a key figure in a controversy involving OpenAI and the Navier-Stokes equations. At the time, New York University mathematician Tristan Buckmaster suggested that OpenAI had stolen research produced by him and Alpöge.

If this proof of the Riemann Hypothesis is ultimately confirmed, its impact will extend far beyond mathematics.

The result will not immediately yield concrete real-world applications, upend the global banking system, or suddenly make the internet run faster.

Over the past year, however, people have watched large language models surpass one benchmark after another in mathematical ability. Now, it is difficult to imagine any achievement more astounding than proving the Riemann Hypothesis.

MIT mathematician Andrew Sutherland once remarked: “If there were an AI capable of proving the Riemann Hypothesis, that is absolutely not an AI I would want to encounter. If it really happened, whether mathematicians could keep their jobs would probably be the least of our worries.”

For Alpöge, however, this is undoubtedly a moment worth celebrating.

He wrote: “I hope this argument will offer us a glimpse of the miraculous possibilities that lie ahead and show us the joy and wonder we are about to discover in the simplest things.”

Global Market Intelligence

Artificial Intelligence

CLS Statement: The content of this article is for reference only and does not constitute investment advice. Investors act on this information at their own risk.

Business Cooperation

Popular Unlocks

[Institutional Dragon-and-Tiger List Analysis] Physical AI + embodied intelligence + AI applications: building a physical AI data foundation around four categories of data while also developing comprehensive solutions for embodied-intelligence training grounds, covering simulation and validation, data collection and annotation, and transfer and delivery to physical systems; this company received net buying

① Physical AI + embodied intelligence + AI applications: building a physical AI data foundation around four categories of data while also developing comprehensive solutions for embodied-intelligence training grounds, covering simulation and validation, data collection and annotation, and transfer and delivery to physical systems; this company received net buying. ② Printed circuit boards + copper-clad laminates: orders for copper-clad laminate products are plentiful, product prices are rising, and profitability has improved significantly; high-performance HDI capacity has reached 240,000 square meters per year, while 200G–400G optical modules are in small-batch mass production; institutions made large net purchases of this company’s shares.

[Intraday Scoop] Demand has improved significantly while supply is recovering slowly; amid synchronized global trends, the industry’s price midpoint is rising, and this company possesses abundant resource reserves

Demand has improved significantly while supply is recovering slowly; amid synchronized global trends, the industry’s price midpoint is rising. This company possesses abundant resource reserves, while another company benefits from an integrated industrial chain.

[Market Trend Research Report · Reporter Dispatch] Synthetic soda ash capacity has begun to clear; industry insiders expect the soda ash sector may enter a short upcycle, with this product category continuing to benefit

Against a backdrop of severe oversupply, high-cost ammonia-soda process producers will face greater operational pressure and may have to reduce operating rates, suspend production for extended periods, or even exit the market. Thanks to low raw-material costs, a simple production process, low energy consumption, and limited environmental-protection investment, this category of products enjoys an overwhelming cost advantage over products made using the ammonia-soda and combined-soda processes.

Related Reading

Down from 4.7% to 3.7%! U.S. consumer inflation expectations cool after the midterm elections

CLS, Shi Zhengcheng, 2026-11-21 00:05

① The University of Michigan’s consumer survey released Friday evening showed that U.S. one-year consumer inflation expectations fell from 4.7% to 3.7% in November, their lowest level since February, while consumer sentiment also improved. ② Shifting expectations after the midterm elections and falling gasoline prices provided relief, but the high cost of living remains consumers’ primary concern.

After the S&P 500 quickly broke below a key moving average, what signs of a bottom are technical traders watching?

CLS, Shi Zhengcheng, 2026-11-20 22:59

① As of Thursday’s close, the S&P 500 had closed below its 200-day moving average for three consecutive sessions, but historically, after similar occurrences, the index was higher 21 trading days later 71% of the time. ② Some technical analysts believe a near-term bottom may be approaching, with potential support levels including 7,160, 7,080, and 7,000 drawing attention.

Anthropic’s record-breaking IPO of approximately $93.5 billion fails to revive enthusiasm for AI stocks

CLS, Li Ying, 2026-11-20 22:51

① Even before the excitement surrounding the record-breaking IPO of approximately $93.5 billion had faded, Anthropic’s shares weakened alongside global technology stocks, at one point falling more than 9% at the open on their second day of trading and nearing the $83 issue price. ② Analysts bluntly said that the AI narrative is “starting to crack.” Investors are locking in profits while becoming increasingly selective, no longer believing that a rising AI tide will lift all boats and imposing ever more stringent demands on companies’ growth and profitability.

Source

Cailian Press