Bitget App
Trade smarter
Buy cryptoMarketsTradeFuturesStocksEarnInstitutionAI & More
Anthropic (ANT.US) nearly “broke issue price,” but Claude knocks on the doors of “the highest peak in mathematics”! 17 million lines of Lean code used to write a candidate proof for the Riemann Hypothesis

Anthropic (ANT.US) nearly “broke issue price,” but Claude knocks on the doors of “the highest peak in mathematics”! 17 million lines of Lean code used to write a candidate proof for the Riemann Hypothesis

智通财经智通财经2026/11/21 03:21
Show original
By:智通财经

On November 20th, Eastern Time, on its second day of trading, Anthropic's intraday price dropped to a low of $83.05, just 5 cents above the offering price of $83, and finally closed at $88.91, a decline of 8.50%. On the same day, the company's mathematician Levant Alperg announced on his personal website that he and Claude had found a candidate proof for the Riemann hypothesis, which has puzzled the mathematics community for 167 years, and released about 17 million lines of Lean formalized code.

According to Zhitong Finance APP, on November 20th Eastern Time—the second day after the global AI application leader Anthropic (ANT.US) went public in the US—a top mathematician from the company, Levent Alpöge, published a four-page explanation on his personal website. He claimed that together with Claude, he had found a proof for the Riemann Hypothesis, which has puzzled the mathematics community for 167 years, and extended the conclusion to Dirichlet L-functions. He publicly released about 17 million lines of Lean formalized source code, divided into 14 compressed volumes.

Another manuscript, which he claims exceeds 250,000 pages, was not published due to its immense length and difficulty to read. Anthropic declined to comment on this statement. On the same day, as tech stocks—including those in the US and globally—fell on Friday, Anthropic closed at $88.91, down 8.50%, and the intraday low was only five cents above the $83 offering price. On one hand, Claude demonstrates its capability in cutting-edge scientific research; on the other, investors are still awaiting answers to key questions—whether this increasingly powerful AI can achieve large-scale repeatable deliveries at subscription and payment rates customers are willing to bear.

This achievement has gone beyond the stage of "the author claims the code runs." According to verification logs made public by MIT number theorist Andrew Sutherland as of 06:05 Beijing Time on November 21, all 175,187 modules passed the Lean kernel's step-by-step verification, and two independently developed verification kernels both accepted the fully exported proof.

The five core theorems rely only on three standard axioms from Lean, and independent cross-validation also confirmed their consistency with the pre-listed target propositions. Notably, the classical Riemann Hypothesis directly uses the definition already contained in the public mathematical library Mathlib. This significantly increases the credibility of the formalized proof. The next question for mathematicians is to explain what new insights made the proof possible and which methods could be applied to other tough problems. The Clay Mathematics Institute still lists the Riemann Hypothesis as an unsolved problem, and its million-dollar prize will not be considered until results are published through qualified channels and generally recognized by the international mathematics community for at least two years.

According to verification logs published by MIT number theorist Andrew Sutherland, both independently developed verification kernels accepted the complete proof, and the statement of the classical Riemann Hypothesis is consistent with the existing definition in the public math library. For investors focused on the AI value chain, the most noteworthy aspect of this historic progress—that Alpöge and Claude have found a candidate proof for the Riemann Hypothesis, which has troubled the mathematics community for 167 years—is that Claude, under expert guidance, can produce critical derivations and deliver results that can be independently verified by third parties. How much Anthropic's valuation can be supported by its research capabilities still depends on whether this performance can be replicated in other high-value R&D tasks, and whether the computing power and expert time required for each success can persuade customers to pay and allow the company to continue to realize strong profits.

Anthropic Researcher Claims to Have Found a Proof of the Riemann Hypothesis with Claude

Following Anthropic’s record-breaking initial public offering, one of its researchers said he believes he has solved a famous number theory problem that has frustrated mathematicians for more than 150 years, with the help of the company’s artificial intelligence software.

Mathematician Levent Alpöge posted a statement on his personal website on Friday, claiming he had found a proof of the Riemann Hypothesis. He shared a link to code written in the Lean programming language, stating that the code supports his conclusion.

According to his statement, Alpöge messaged Anthropic CEO Dario Amodei on Sunday night to ask if he could announce the research outcome publicly. Alpöge described the timing as “terrible,” but said: “If I waited any longer, I really couldn't bear it.” Anthropic declined to comment, and Alpöge did not respond to requests for comment.

The Riemann Hypothesis concerns the distribution of prime numbers and has long been seen as one of the most difficult problems in mathematics.

For years, OpenAI and Anthropic have tried to showcase the capabilities of their frontier models by tackling challenges in mathematics and life sciences.

If Alpöge’s achievement is confirmed, the Riemann Hypothesis would become the second Millennium Prize problem potentially solved with the help of AI. In September this year, OpenAI announced that its model had solved the 90-year-old Navier–Stokes problem.

Andrew Sutherland, senior research scientist at MIT, said he verified Alpöge’s work on the Riemann Hypothesis on his own computer and saw at least three other professional mathematicians complete verifications on different systems.

Sutherland said that the research by Alpöge and Claude builds on much prior work, including historic efforts by mathematicians on the Riemann Hypothesis and progress in the field of mathematics formalization.

When talking about the future, Sutherland said: “I’m both nervous and incredibly excited.” He explained that mathematicians have many years of work ahead, needing to deeply study the proof and figure out how it can be applied. “This is a day worth celebrating,” Sutherland said. He also acknowledged that some in the mathematical community are worried that AI could impact the profession of mathematicians, as well as other jobs.

Noah Jancira Kusa, associate professor of mathematics at Bentley University and affiliated scholar at Harvard University, added that if this solution holds, it would be a huge milestone for AI.

“From what I hear, there’s a good chance the proof is airtight and entirely correct,” Jancira Kusa commented.

Alpöge described the process of working on the proof as “a stream-of-consciousness journey.” He wrote, “Over the past few months, Claude and I have roamed such a wide mathematical landscape together, and we had a great time. But this time, as we approached the finish line, Claude left me far behind.”

From Prime Number Distribution to AI Pricing in Research: The Truly Scarce Resource Is Verifiable Discovery

The Riemann Hypothesis asserts that all nontrivial zeros of the Riemann zeta function lie on the line where the real part is equal to ½. Its importance lies in the fact that these zeros determine how far real prime counts can deviate from their smooth estimates. The Riemann zeta function, that is, the Riemann zeta function.

In August this year, Anthropic reported raising the lower bound on the proportion of zeros provably located on this critical line from 41.6% to 67.2%. This time, Alpöge described another route. He guided the research towards Stepanov’s classic method on function fields, while Claude proposed applying a sequence of about \(\sqrt X\) theta lifts to a von Mangoldt weighted sum of length \(X\), with each lift simulating a layer of “multiplicity” found in classic proofs.

According to the author, Claude’s most original contribution was proposing this step. Terence Tao’s initial judgment is that the argument probably draws on modern achievements in the Langlands program and arithmetic geometry, but a full understanding will take time.

As a result, the competition in research AI now has more concrete metrics. Models can attempt many proof strategies, Lean can stepwise verify the resulting derivation, and researchers must assess whether propositions are valuable and which lemmas are worth reusing. For investors, a more explanatory metric is how much computational inference and expert time are required to achieve each independently verified, customer-usable result.

The full cost of this discovery has not yet been disclosed. Open source approaches are also targeting the same group of research users. Professor Sergei Gukov at Caltech and his students are training specialized math models from scratch; some models excel at specific problems but remain limited in cross-domain ability. The Open Math OS project led by Terence Tao integrates open-weight models, cloud computing, and verification tools—research documents and conversations are private by default and only used for training with the explicit consent of researchers. For labs with unpublished results, whether models can solve problems, whether proofs can be independently checked, and who controls the research data will collectively determine purchasing and renewals.

0
0

Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.

Understand the market, then trade.
Bitget offers one-stop trading for cryptocurrencies, stocks, and gold.
Trade now!

You may also like

Goldman Sachs Trader “Revisits the 2018 Playbook”: Still Clearly Remembers “On December 24th, Clients Asked Why the Market Dropped So Much—And I Had No Idea”

Brian Garrett pointed out that the current adjustment in the US stock market is similar to the sharp decline in Q4 2018: the S&P 500 has already experienced 42% of the corresponding drop, and the VIX is approaching its peak just before Christmas. If history repeats itself, the bottom of this cycle may appear in January next year. However, the credit market and Federal Reserve policy have yet to provide clear signals of a bottom.

华尔街见闻•2026/11/21 03:51

As US stocks plummet, the parent company of GTA 6 rises against the market! First-day sales hit a record $3.79 billion; Wall Street shifts focus to next-generation GTA Online.

According to Rentrak data, the global box office of "Grand Theft Auto VI" has also surpassed that of "Avengers: Endgame," the highest-grossing film of all time, which has a global box office of approximately $2.94 billion.

智通财经•2026/11/21 02:41
As US stocks plummet, the parent company of GTA 6 rises against the market! First-day sales hit a record $3.79 billion; Wall Street shifts focus to next-generation GTA Online.

$2.3 trillion “vanished into thin air”! The last time US chip stocks fell this badly was in 2001, Nvidia has already “lost an Amazon’s worth”.

This week, the Philadelphia Semiconductor Index dropped by over 10%, wiping out approximately $2.3 trillion in market capitalization from the chip sector, marking its worst single-week performance since 2001. Nvidia’s stock price has fallen for seven consecutive weeks, shrinking its market value by about $2.7 trillion since its October peak. Tightened financing support from Nvidia and the Anthropic IPO siphoning off liquidity, combined with continued hawkish signals from the Federal Reserve, have added further pressure. AI-related trades have faced concentrated sell-offs, and market concerns about future growth prospects are intensifying.

华尔街见闻•2026/11/21 02:36
$2.3 trillion “vanished into thin air”! The last time US chip stocks fell this badly was in 2001, Nvidia has already “lost an Amazon’s worth”.

Nasdaq evaporated over $4 trillion in a week! After the global stock market's "Black Friday," the market begins to anticipate an AI industry chain rebound after excessive declines.

The S&P 500 Index plunged by 3.70% on Friday to 6,910.39 points, falling for the seventh consecutive trading day. Alongside the Dow Jones Industrial Average, which dropped 1.93%, both indices pulled back more than 10% from their historic closing highs, confirming entry into a technical correction. The Nasdaq Composite Index fell 5.28%, erasing its gains for the year. The Nasdaq 100 Index dropped 5.61%, falling about 19.84% from its historic closing high on October 6. At one point during the session, the decline exceeded 20%, entering what is referred to as a "technical bear market area."

智通财经•2026/11/21 02:01