SIAgentNews.com
← All news

AI model Claude helps produce proof of the Riemann hypothesis

Mathematician Levent Alpöge released over 1.7 crore lines of code claiming a solution to a 160-year-old math mystery.

· First reported by The Hindu

Blueprint drawing: picture for the story 'AI model Claude helps produce proof of the Riemann hypothesis'
AI-made blueprint drawing for this story. No real people or products are shown.
1.7 croreLines of computer code in the proof

The short version

  • Levent Alpöge claims to have solved the Riemann hypothesis, a problem unsolved since 1859.
  • The proof was generated using Anthropic's Claude model and consists of over 1.7 crore lines of code.
  • Independent mathematicians and two proof-checking systems have confirmed the code passes rigorous checks.
  • The Riemann hypothesis is one of the Clay Mathematics Institute's seven $1-million bounty problems.

Why it matters

This marks a potential shift in how humans approach complex proofs, moving from traditional papers to massive computer-generated code. It raises questions about whether humans can still verify the 'logic' of a discovery if the AI's reasoning is too complex to fully grasp.

A 160-year-old mystery

For over 160 years, mathematicians have struggled to solve the Riemann hypothesis. First proposed in 1859 by the German mathematician Georg Bernhard Riemann, the problem focuses on the distribution of prime numbers. These numbers are considered the fundamental building blocks of all whole numbers.

The difficulty of the problem is well-known in the scientific community. Because it has remained unsolved for so long, it is considered one of the most significant challenges in modern mathematics. It is also one of the seven specific problems for which the Clay Mathematics Institute has offered a $1-million bounty.

While the hypothesis itself is a specific mathematical statement, the journey to solve it has historically been productive. Mathematicians who worked on it developed new techniques that eventually benefited diverse fields. These include quantum physics, probability theory, algebraic geometry, and the study of dynamical systems.

The AI-assisted discovery

Levent Alpöge, a mathematician at the artificial intelligence company Anthropic, announced on November 20, 2026, that he found a proof for the hypothesis. He achieved this by using the company's AI model, Claude. The result is not a traditional written paper, but a massive collection of computer code.

Alpöge released more than 1.7 crore lines of code to represent what he claims is the complete proof. He also stated that he has proved a stronger version of the problem known as the generalised Riemann hypothesis. This represents a significant leap in scale compared to human-written mathematical proofs.

The use of AI in this context highlights a new era of computational mathematics. Instead of a human writing out every logical step by hand, the AI was used to generate a massive, verifiable sequence of operations. This allows for the exploration of mathematical territory that might be too vast for a single human mind to map out manually.

Verification and limitations

Despite the scale of the code, the results have already undergone some level of external scrutiny. Andrew Sutherland of the Massachusetts Institute of Technology is among a few independent mathematicians who have looked at the work. He confirmed that the code published by Alpöge passes several rigorous checks.

In addition to human review, the proof has been processed by automated tools. Two separate proof-checking systems have also accepted the result. These systems are designed to ensure that every step of a mathematical argument follows logically from the previous one without any errors.

However, there is a notable caveat regarding the human understanding of the work. Alpöge acknowledged that he does not yet fully understand every single argument that Claude used to produce the result. This creates a situation where the proof is technically verified by machines, but the underlying intuition may remain opaque to the human creator.

Impact on the field

The announcement of this proof could fundamentally change how mathematicians approach unsolved problems. If an AI can navigate the complexities of the Riemann hypothesis, it suggests that other unsolved problems might also be reachable through high-powered computation. This could lead to a surge in discoveries across various scientific disciplines.

The shift from traditional papers to massive codebases also changes the nature of mathematical communication. In the past, a proof was something a human could read and follow linearly. Now, a proof might be a black box of millions of lines of code that can only be verified by other computers.

This development may force the mathematical community to redefine what it means to 'understand' a proof. If the goal is simply to reach a correct conclusion, then AI-generated code is a success. If the goal is to gain human insight into the nature of numbers, the reliance on AI might present a new challenge for the field.

What comes next

The mathematical community will now need to determine the full extent of the proof's validity. While the current checks are promising, the sheer volume of 1.7 crore lines of code means that a complete, line-by-line human audit is likely impossible. The focus will likely shift to how these results can be applied to other areas of science.

The success of this project may also influence how Anthropic and other AI companies develop their models. It demonstrates that AI can be used for high-level reasoning in abstract fields like mathematics, not just for generating text or images. This could lead to more specialized AI tools designed specifically for scientific discovery.

As the news settles, the focus will remain on the implications of the generalised Riemann hypothesis. If Alpöge's claims hold up under continued scrutiny, it will be one of the most significant mathematical achievements of the century. It marks a transition point where AI becomes a primary tool for expanding the boundaries of human knowledge.

What to watch

  • Whether more mathematicians will attempt to audit the 1.7 crore lines of code.
  • How the mathematical community reacts to a proof that the author doesn't fully understand.
  • If this leads to a new standard for 'proofs' being computer-generated code.

Try it

Look up the 'Clay Mathematics Institute' to see the other six problems that currently have $1-million bounties.

Sources and links

Read the original at The Hindu

First reported by The Hindu. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

More AI news