Claude, the artificial intelligence, has shattered expectations in a field previously reserved for the brightest humans. In just 11 days, it accomplished what is estimated to be ten years of human labor. This feat illustrates how AI is revolutionizing mathematics.
Claude formalizes Fermat’s Last Theorem: a staggering acceleration
On September 4, 2026, Anthropic announced a spectacular advancement. Their AI, Claude, formatted the proof of Fermat’s Last Theorem in Lean, a proof language that can be verified by computer. This task was previously estimated to take ten years of hard human work.
Rather than reinventing the mathematical wheel, Claude did not prove Fermat anew. Instead, it translated Andrew Wiles’ proof from the 1990s, which is highly complex, into rigorous machine code. The result: 13 million lines of Lean code generated by the AI, using nearly 29,500 intermediate theorems validated automatically.
Why would this formalization have taken humans ten years?
At first glance, copying 129 pages seems doable in a few months. But formulating the proof entirely for software is another world. Human mathematicians intentionally leave some steps implicit, considered obvious among peers.
Verification software does not allow for shortcuts! Each definition, property, and logical connection must be explicitly detailed in rigorous syntax. This meticulous work can extend over a decade.
This is precisely the challenge Claude has met, achieving in 11 days what was estimated to take so long. It’s not just fast; it’s an enormous shift in scale.
The implications for research and digital technology
This formalization has surpassed the simple mathematical feat. It paves the way for a revolution in scientific verification. Imagine an assistant capable of analyzing proofs, detecting subtle errors, and validating complex results in record time.
Researchers could finally have a tool to transform their ideas into easily verifiable formal proofs. AI Claude thus becomes a bridge between the creative intuition of mathematicians and the relentless rigor of machines.
What does this change concretely?
Mathematics will no longer be just a matter of equations on paper. The capability to formalize this type of work shifts the discipline toward a model where the computer is an indispensable partner. It facilitates the exploration of areas that have previously been too complex to be fully verified.
The role of the mathematician remains central: they invent, understand, propose. But the digital verifier now allows for a much faster closure of proofs. This promises to significantly reduce the risks of human error.
For more details on this advancement, many dedicated articles can be explored, such as those on the Editions Fractales website or Le Monde Sciences.
What are the key figures of this feat?
- 11 days: duration of the formalization by Claude, a real rocket!
- ≈ 10 years: estimated time for an equivalent human effort, a chasm.
- 13 million: lines of Lean code generated for this formalization.
- 29,500: intermediate theorems used for the complete proof.
The impact on digital sovereignty and open source
In a world where technical transparency is key, an AI like Claude also raises fascinating ethical and strategic questions. Its ability to produce open, verifiable, and auditable code is a colossal asset for security and digital trust.
The open formalization aligns with the philosophy of free software: shared knowledge, accessible to all, and not dependent on a black box. It perfectly embodies the vision of French digital sovereignty.
What is the Lean language used by Claude?
Lean is a programming language specialized in formalizing mathematical proofs in a rigorous format that computers can verify step by step.
Why is the formalization of Fermat’s theorem so complex?
Because each logical and mathematical detail must be explicit so that the machine can verify the proof, which is not the case in a classic demonstration written by humans.
Has Claude discovered any new results?
No, the AI translated and formalized the already known proof of Andrew Wiles, but in an entirely automated and computer-verifiable manner without manual errors.
Does this advancement mean the end of human work in mathematics?
Not at all. The AI accelerates verification, but the creativity and intuition of researchers remain irreplaceable and even amplified by these tools.
How can this formalization impact scientific research?
It allows for rapid validation of complex demonstrations, avoiding errors and fraud, and fosters trust in science thanks to automatic rigor.
Source: www.webmanagercenter.com