No primary paper or preprint was located, so this article does not treat the counterexample or its verification as established.
A one-line counterexample can be checked quickly; what it teaches mathematics is a separate question.
Reports say Harvard mathematician Levent Alpöge presented a concrete counterexample to the Jacobian conjecture with help from Anthropic's Fable 5. If the framing holds, the event is notable not only for AI capability, but for the way it exposes the difference between finding an answer and producing mathematical understanding.
Velocity
The counterexample spread fast and was reportedly checked quickly by many mathematicians.
Verifiability
The claimed map is explicit enough that the algebra is easier to verify than it is to discover.
Implication
The key question is no longer just whether the answer works, but what theory and intuition come out of it.
New Scientist reports that on July 19, Levent Alpöge posted that the Jacobian conjecture is false and gave a 216-character counterexample, while crediting his “close friend fable,” apparently referring to Anthropic's Claude Fable 5. The article quotes mathematician Abhishek Saha calling it perhaps the biggest conjecture so far in which AI has played a significant role, while also stressing that the method behind the result is not yet well explained in public. New Scientist
Fortune pushes the story further by saying the result was verified rapidly and even checked in Lean, while still emphasizing that mathematicians ultimately want more than a correct output: they want a story about why the answer is true or false. That distinction matters because AI can compress search, but explanation remains a different layer of value. Fortune
Mashable adds a useful caution through Andrew Blumberg's comments: producing a compact counterexample is not the same thing as generating the kind of proof or conceptual framework that opens new mathematics. In that reading, the event is impressive but not identical to building a deep theory. Mashable
The mathematics blog sbseminar grounds the discussion in the explicit 3-variable polynomial map at issue, noting that its Jacobian is constant while the map is generically three-to-one rather than bijective. That is what makes the story more than generic AI hype: the claim lands on a specific object that working mathematicians can examine directly. sbseminar
So the most careful reading is two-layered. First, the reported Fable-assisted counterexample is a major AI-math headline because it appears both concise and quickly checkable. Second, the deeper meaning of the result — what it teaches, what structure it reveals, and whether it changes mathematical practice — still depends on human interpretation, explanation, and follow-on work.
Related Issues
Issue 05
Distillation Is Standard Engineering; Unauthorized Extraction Triggers Contract and Sanctions Risk
Why capability gains quickly raise compliance questions.
Issue 06
The Legal Boundary of Knowledge Distillation
Where law may try to catch up with expanding model capabilities.
Feature 01
US-China AI Competition
The larger race in which research capability becomes strategic leverage.