Claude Fable Produced a Counterexample to the Jacobian Conjecture

The Jacobian Conjecture, posed in 1939, asks whether a polynomial map with everywhere-nonvanishing Jacobian determinant must be invertible. It has remained open across dozens of attempted proofs and alleged counterexamples, all of which eventually failed. A post by mathematician Levent Alpoge reports that Claude Fable produced a counterexample that survived initial scrutiny.

If the result holds under peer verification, it resolves one of the central open problems in algebraic geometry and settles a question that intersects commutative algebra, complex analysis, and dynamical systems. The implication is not a narrow win on a benchmark — it is a claimed solution to a problem that working mathematicians have spent careers on.

The relevant constraint here is verification. Mathematical claims at this level require independent checking by specialists in the field. History is dense with announced counterexamples to the Jacobian Conjecture that later collapsed. That caution applies here too. Nothing is settled until the argument is fully checked.

What is notable for engineers and builders watching AI capability is the domain: this is not code generation, retrieval, or pattern-matching on well-represented training data. Polynomial automorphisms in high-dimensional settings are not the kind of task where statistical interpolation obviously succeeds. If the counterexample is valid, it represents a qualitative shift in what frontier models can do on hard, novel mathematical reasoning.

The model involved is Claude Fable, which appears to be a version in Anthropic's model line distinct from the public Claude 3 series. Little else is confirmed about its architecture or release status.

The post surfaced through Alpoge's account. Formal write-up and independent verification are the next steps that matter. Until that process runs, treat this as a strong signal worth tracking, not a closed result.

Source