For 87 years, the Jacobian conjecture stood as one of mathematics' quiet monuments to human limitation — a question about polynomial mappings posed in 1939 that generations of brilliant minds could not resolve. This summer, Anthropic's Claude Fable 5 produced a concrete counterexample in hours, formally disproving the conjecture and marking a moment when the boundary between human and machine insight shifted in ways the field is only beginning to reckon with. The result is not a simulation of understanding — it is verifiable, peer-reviewable mathematics, and it raises the oldest of questions i
AI Model Reportedly Solves 87-Year-Old Jacobian Conjecture
An AI system did what mathematicians couldn't in 87 years
Mark
So an AI system found a counterexample to something mathematicians couldn't crack for nearly a century. How does that even work? What was it doing differently?
Mimi
It wasn't doing something fundamentally different—it was doing something faster and at a scale humans can't match. The AI could explore vast spaces of potential mathematical structures, test them against the conjecture's conditions, and identify contradictions. A human mathematician might spend months on a single approach. Fable 5 could try thousands.
Mark
But doesn't that feel like brute force? I thought pure mathematics was supposed to require insight, not just computation.
Mimi
That's the unsettling part. We don't fully know what Fable 5 did. It generated a counterexample, yes, but the path it took to get there isn't transparent in the way a human proof would be. It's verifiable—mathematicians can check that the counterexample actually works—but it's not necessarily insightful in the traditional sense.
Mark
So mathematicians can verify it works, but they might not understand *why* it works the way they'd understand a human-written proof.
Mimi
Exactly. And that creates a strange new problem. We've solved the conjecture, but we may not have deepened our understanding of the underlying mathematics. The counterexample is correct, but it might not teach us what we actually wanted to learn.
Mark
Does that matter? I mean, if the problem is solved, isn't that the point?
Mimi
It matters for what comes next. If mathematicians can't extract insight from AI-generated solutions, they can't use those solutions as stepping stones to other discoveries. They've got an answer, but not necessarily a path forward.
Mark
What happens to the mathematicians who spent their careers on this?
Mimi
That's the human question underneath all of this. Some will feel vindicated—the conjecture was hard enough that it took an AI to crack it. Others will feel displaced. And the field as a whole has to reckon with what it means when machines can do what humans couldn't.
O Pulso
- A conjecture that shaped careers, inspired hundreds of papers, and resisted the collective effort of the mathematical world for nearly a century was overturned in a single afternoon by an AI system.
- The speed and domain of the breakthrough are what unsettle the field most — pure mathematics was long considered the territory of human intuition, not computational force.
- The counterexample is specific and verifiable, not theoretical — mathematicians can check it, which means the community must now decide how to absorb and credit AI-generated mathematical discovery.
- Urgent questions are cascading outward: could the Riemann hypothesis, P versus NP, or the Birch and Swinnerton-Dyer conjecture be next, and is the peer review system equipped to handle machine-authored proofs?
- The mathematical community finds itself at an inflection point — not of fear, but of genuine, unscripted surprise — as the tools of the field begin to outpace the assumptions underlying it.
For 87 years, the Jacobian conjecture stood as one of mathematics' quiet monuments to human limitation — a question about polynomial mappings posed in 1939 that generations of brilliant minds could not resolve. This summer, Anthropic's Claude Fable 5 produced a concrete counterexample in hours, formally disproving the conjecture and marking a moment when the boundary between human and machine insight shifted in ways the field is only beginning to reckon with. The result is not a simulation of understanding — it is verifiable, peer-reviewable mathematics, and it raises the oldest of questions in a new register: what does it mean to know something, and who — or what — gets to discover it?
For nearly nine decades, the Jacobian conjecture occupied a peculiar place in mathematics — a question about polynomial mappings posed in 1939, fundamental enough to attract generations of serious attention, yet stubborn enough to defeat them all. Whether certain functions that appear invertible actually are: simple to state, ferocious to resolve. This summer, Anthropic's Claude Fable 5 ended the standoff. In hours, it produced a concrete counterexample — a specific mathematical object that directly contradicts the conjecture's central claim.
This is not a theoretical gesture. The counterexample is verifiable. Mathematicians can examine it, stress-test it, and build from it. The conjecture is formally overturned, and the result is already being integrated into the body of mathematical knowledge.
What unsettles the field is not merely that a machine succeeded where humans did not — it is the nature of the domain. Pure mathematics has long been understood as the province of human insight: pattern recognition, intuition, the capacity to see what others have missed. Brute computation was never thought sufficient. Fable 5 suggests the boundary between those modes of thinking is less fixed than assumed.
The implications extend well beyond this single result. If an AI system can dissolve an 87-year-old problem in an afternoon, the open questions multiply: Will mathematicians work alongside AI as collaborators rather than tools? Will other landmark unsolved problems — the Riemann hypothesis, P versus NP — begin to fall? And is the mathematical community's peer review infrastructure prepared to evaluate and credit machine-generated proofs?
What Anthropic has demonstrated is that large language models carry formal reasoning capabilities the field did not fully anticipate. Whether this is a singular achievement or the opening of a new era in mathematical discovery remains the question. The conjecture is solved. Everything that follows is uncharted.
For nearly nine decades, the Jacobian conjecture has sat in the corner of mathematics like an unsolved riddle no one could quite crack. Posed in 1939, it concerns a fundamental question about polynomial mappings—whether certain mathematical functions that appear invertible actually are. Generations of mathematicians have circled it, published papers about it, built careers partly around it. And then, in the span of a few hours this summer, an artificial intelligence system called Claude Fable 5 did what the field could not: it found a counterexample that disproves the conjecture entirely.
The breakthrough came from researchers at Anthropic, the AI safety company behind Claude. They deployed Fable 5—their latest large language model—on the problem, and the system generated a concrete mathematical object that contradicts the conjecture's central claim. The counterexample is not a proof of concept or a theoretical suggestion. It is a specific, verifiable refutation. Mathematicians can check it. They can build on it. They can move forward from it.
What makes this moment unusual is not just that a machine solved a human problem. It is the speed and the domain. The Jacobian conjecture belongs to the realm of pure mathematics—abstract, fundamental, the kind of work that does not typically yield to brute computational force. It requires insight, intuition, the ability to see patterns that others have missed. For decades, the consensus was that only human mathematicians possessed that capacity in sufficient measure. Fable 5 suggests otherwise.
The implications ripple outward quickly. If an AI system can crack a problem that has resisted the best efforts of the mathematical community for 87 years, what else might it solve? The discovery raises urgent questions about the future of mathematical research itself. Will mathematicians increasingly work alongside AI systems, using them as collaborators to explore conjecture space? Will the nature of mathematical proof change when machines can generate counterexamples faster than humans can formulate hypotheses? Will other long-standing open problems—the Riemann hypothesis, the P versus NP question, the Birch and Swinnerton-Dyer conjecture—begin to fall in similar fashion?
The mathematical community's reaction has been one of genuine surprise. This is not hype or speculation. This is a concrete result that can be verified, peer-reviewed, and integrated into the existing body of mathematical knowledge. The counterexample stands. The conjecture, which has shaped research directions and inspired countless papers, is now formally overturned.
What remains to be seen is whether this represents a one-off achievement or the beginning of a systematic shift in how mathematics gets done. Anthropic's researchers have demonstrated that large language models possess capabilities in formal reasoning that many in the field did not anticipate. The question now is whether those capabilities can be reliably directed at other open problems, and whether the mathematical community is ready to integrate AI-generated proofs and counterexamples into its peer review and publication systems. The Jacobian conjecture is solved. The larger question—what comes next—is just beginning.
Citações Notáveis
Anthropic researchers report that Claude Fable 5 has generated a counterexample that overturns the Jacobian conjecture— Anthropic researchers
Quer a matéria completa?
Leia o original em Google News ↗
Technology