On July 20, 2026, researcher Levent Alpoge from Anthropic's lab announced a sensational result: the Claude Fable 5 language model found a counterexample to the Jacobian conjecture — one of the key unsolved problems in algebraic geometry, formulated by German mathematician Otto-Heinrich Keller in 1939.

The essence of the conjecture is simple yet profound: if a polynomial function of several variables has an invertible derivative (Jacobian non-zero) at every point, then such a function must be globally invertible. In other words, local invertibility guarantees global invertibility. The AI constructed a formula that passes the local test but maps three different sets of numbers to the same result — thereby disproving the conjecture. Alpoge provided links to Wolfram Alpha for independent verification of the calculations.

Context and Historical Significance

Over 87 years, the Jacobian conjecture has become a veritable "trap" for mathematicians. As mathematician Qiaochu Yuan noted, this is the most famous open problem ever solved by a large language model. History records numerous erroneous proofs: Beniamino Segre published three incorrect versions, Claude Chevalley accepted a false proof as true, and Igor Shafarevich used the conjecture as a proven theorem. Mathematician T. T. Moh in 2008 analyzed one such attempt in detail.

Mathematician Jared Duker Lichtman called the result "outstanding," recalling that a special case of the conjecture was the topic of Yitang Zhang's dissertation — the future author of a breakthrough in prime number theory.

AI vs. Open Problems: A New Era

The counterexample to the Jacobian conjecture is not the first time AI has ventured into pure mathematics. In October 2025, OpenAI Vice President Kevin Weil claimed that GPT-5 had solved ten Erdős problems, but the claim was not confirmed — the model had found already known solutions. In January 2026, GPT-5.2 Pro, using the Harmonic automatic prover, solved Erdős problem #397, and in May, an OpenAI model disproved the 80-year-old Erdős unit distance conjecture — the first case of such a level, confirmed by nine external mathematicians.

The key difference with Claude Fable 5's result is verifiability. Unlike AI company claims based on benchmarks, a counterexample is a formula that any mathematician can plug in and recalculate. Alpoge, who earned his PhD at Princeton under Fields Medalist Manjul Bhargava, provided full transparency.

My analysis: This case marks a transition for AI from solving "low-hanging fruit" (as Terence Tao put it) to attacking fundamental, decades-old unsolved problems. The main point here is not just the disproof of a conjecture, but a demonstration that modern LLMs are capable not only of generating text but also of constructing nontrivial mathematical structures with verifiable results. This changes the rules of the game for all of academic mathematics.