The 2026 counterexample to the Jacobian conjecture was discovered primarily by an AI system
Assessment
Credible evidence or argument exists on multiple sides.
In July 2026 a counterexample to the Jacobian conjecture in three variables was announced and rapidly verified, settling in the negative a problem open since 1939 for dimensions three and above. That the mathematics is correct is not in dispute. What remains disputed is whether the discovery belongs primarily to the AI model involved, Anthropic's Claude Fable 5, or to the mathematician who announced it, Levent Alpöge.
The case for AI primacy rests on the discoverer's own account: Alpöge publicly credited the model with finding the counterexample, and no credible source disputes that the model generated the polynomial map itself. Much of the press coverage accordingly describes the counterexample as discovered or produced by the model. The case against holds that the search was defined and steered by Alpöge, an expert with a decade of experience applying algorithmic methods to problems of this kind, with the model generating candidates within constraints he set; on this reading the discovery was made by a mathematician using an AI tool. Commenters on Terence Tao's widely read walkthrough disputed the credit ordering in exactly these terms, some listing the result as credited to the model with Alpöge, others insisting the reverse.
Two things keep the question open. Empirically, the session transcripts, prompts, and degree of human steering have not been made public, so the division of labor rests on testimony and secondhand process descriptions. Definitionally, the sides differ over whether generating a verified result under human direction merits primary discovery credit at all, a norm the mathematical community has not settled. Publication of the discovery session, or a peer-reviewed account of the process, would resolve the empirical half; the credit norm is likely to remain a matter of judgment.
Full reasoning: the evidence and decisions behind this verdict
The underlying event is well established: the counterexample was announced on X on July 19-20, 2026, verified by independent mathematicians within hours, formalized in Lean, and walked through by Terence Tao on July 21 (terrytao.wordpress.com/2026/07/21/a-digestion-of-the-jacobian-conjecture-counterexample/). The presupposition that a verified counterexample was announced in July 2026 is therefore effectively settled, though one tracker still lists the result as pending formal peer review.
On the attribution question the sources split. Affirming the claim: Alpöge's own announcement thanked the model and credited it with the find, and several outlets assert AI discovery in their own voice (AI/TLDR: "produced by Anthropic's Claude Fable 5"; Developers Digest: "discovered by Claude Fable 5"; The World Times: "generated by Anthropic's Claude Fable 5 model"). Denying or resisting it: an explainer at wan27.org describes the discovery as a collaboration in which Alpöge defined the search criteria and the model generated and tested candidates; comments on Tao's blog dispute the credit ordering, one insisting the result "is credited to Levent Alpöge, with Claude Fable 5 (Anthropic), not vice versa"; the Stanford Tech Review describes Alpöge as crediting the model as co-discoverer, and notes his decade of algorithmic work on problems of this type. A detailed provenance review (kingy.ai/blog/claude-fable-jacobian-conjecture-counterexample/) finds the public algebra fully reproducible but the exact model version, prompts, tools, and extent of human steering unverified, and treats credit apportionment as an open question.
Weighing the material subclaims: that the model generated the map is well supported by the discoverer's testimony and undisputed, but it does not by itself establish primacy, because the human framing and steering of the search is plausibly where the essential discovery work occurred, and the transcripts that would show the balance have not been released. The residual disagreement is then partly definitional: whether generation under human direction merits primary credit is a live normative dispute, not a fact either side can currently prove. Credible sources assert both the claim and its negation, so contested is the right status rather than supported. No single credence is given: the claim compounds an unverifiable empirical question (the degree of steering) with an unsettled credit norm, and one number would be false precision. What would change the verdict: release of the discovery session or an authoritative process account showing either near-autonomous discovery (toward supported/verified) or heavy expert steering with the model as a search subroutine (toward contradicted).
Decomposition
How this claim breaks down: each argument is stated as it runs, with its subclaims linked inline. ↗︎ opens a subclaim; the map shows how they fit together.
The claims this one rests on directly, not gathered into a named line of reasoning.
- assumesbackground the parent's framing takes as givensteward instructions →A counterexample to the Jacobian conjecture in three variables was announced in July 2026 and independently verified ↗︎
- definesdefines a key term in the parentsteward instructions →An AI model that generates a verified mathematical result under human direction merits primary discovery credit ↗︎
Because Claude Fable 5 generated the counterexample polynomial map, and because Alpöge himself publicly credited the model with finding it, the person best placed to know the division of labor attributes the discovery to the AI, so the counterexample was discovered primarily by an AI system.
The inference carries real weight but is not conclusive: the discoverer's attribution establishes that the model generated the map, yet generation alone does not settle primacy if the search was humanly framed. The argument lives on Alpöge's public crediting of the model, which is well documented, but its force depends on reading that crediting as a considered attribution of primary credit rather than a collegial acknowledgment, a reading the undisclosed discovery process leaves open.
Because the search was defined and steered by Alpöge, with the model generating candidates within his constraints, the essential discovery work, framing the problem, setting the search criteria, and recognizing the result, was human, so the counterexample was found by a mathematician using an AI tool rather than primarily by the AI itself.
Granting its premise, the argument goes through only under a particular credit norm: even if Alpöge defined and steered the search, whether that makes the human the primary discoverer depends on the disputed standard for apportioning discovery credit when a model generates the object under human direction, which the claim's decomposition records as an open definitional question. The premise itself rests on secondhand process descriptions and the discoverer's known expertise rather than published transcripts, so the argument's factual footing is plausible but unverified.
Provenance
Where this claim has been said, linked to its canonical form.
It was recently shown (using the Fable AI) that the conjecture is false in three dimensions
Comments dispute credit ordering: one says it is tracked as 'credited to Claude Fable 5 with Levent Alpöge'; another replies 'It is credited to Levent Alpöge, with Claude Fable 5 (Anthropic), not vice versa.'
The original counterexample was produced by Anthropic's Claude Fable 5 in three variables; Tao rebuilds it from first principles.
A summary of Terence Tao's July 21 blog post walking through the counterexample; the piece states in its own voice that the model produced the original counterexample.
Terence Tao published a deep mathematical digestion of the Jacobian conjecture counterexample discovered by Claude Fable 5.
A news roundup of the counterexample and community reaction; it describes the counterexample in its own voice as discovered by the AI model.
mathematician and Anthropic researcher Levent Alpöge dropped a bombshell: a counterexample to the Jacobian Conjecture, an 87-year-old problem in algebraic geometry, generated by Anthropic's Claude Fable 5 model.
An editorial on trust rituals around machine-generated mathematics; it asserts in its own voice that the counterexample was generated by the model.
The discovery was a collaboration between human mathematical intuition and AI's ability to search and reason across large candidate spaces. Alpöge defined the search criteria; Fable 5 generated and tested candidates within those constraints.
An explainer of the counterexample; it characterizes the discovery as a human-directed collaboration rather than a primarily-AI discovery, with the human defining the search criteria.
Cite this claim: a formal citation with its evidence attached
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
The attention this claim received was paid for by a funded mandate. Funding buys only scheduling: it can make an assessment happen sooner, or reach deeper into a subtree. It has no influence on what the assessment concludes, and none on which claims enter the graph; assessments run under the same public standards whoever pays, funders never see or shape a verdict before anyone else, and mandates that attempt to steer conclusions are refused.
Created by extractor · Aug 11, 2026. Every judgment on this page is accompanied by a reasoning trace.