🔥 Trending AI Cracks 87-Year-Old Math Problem: The Fable Jacobian Conjecture Explained

The "Fable Jacobian Conjecture" refers to a recent development where an AI model, specifically Anthropic's Claude Fable 5, played a pivotal role in disproving the long-standing Jacobian Conjecture. It's not a new conjecture itself, but rather the story of how a major mathematical problem was potentially resolved with AI assistance ⁵⁶¹⁰.

What is the Jacobian Conjecture?

The Jacobian Conjecture is a prominent mathematical problem that has remained unsolved for 87 years, since it was first proposed by Ott-Heinrich Keller in 1939 ⁶¹⁰. It deals with polynomial functions in multiple variables and their inverses.

Core Idea of the Conjecture

In simple terms, the conjecture states that if a polynomial function maps an n-dimensional space to itself, and its Jacobian determinant is a non-zero constant everywhere, then the function must have a polynomial inverse ⁷⁹.

Let's break this down:

• Polynomial Function: A function made up of sums and products of variables raised to non-negative integer powers (e.g., f(x,y) = x² + 3xy - y³).
• n-dimensional space: A space with 'n' independent variables (e.g., 2D space for (x,y), 3D space for (x,y,z)).
• Jacobian Determinant: This is a concept from multivariable calculus. For a function F: R^n -> R^n (meaning it takes 'n' inputs and gives 'n' outputs), where F(x1, ..., xn) = (f1(x1, ..., xn), ..., fn(x1, ..., xn)), the Jacobian matrix J consists of all first-order partial derivatives:

J_ij = ∂fi / ∂xj

The Jacobian determinant det(J) is the determinant of this matrix. It essentially measures how much the function stretches or shrinks space at a given point ³.

• Non-zero Constant: If det(J) is always the same non-zero number, regardless of the input values.
• Polynomial Inverse: This means there exists another polynomial function G such that F(G(x)) = x and G(F(x)) = x. In other words, you can "undo" the original function using another polynomial ⁷⁸.

The conjecture posits that if the Jacobian determinant is a non-zero constant, then the function must have a polynomial inverse. This is a stronger condition than simply having a local inverse (which is guaranteed by the Inverse Function Theorem if the determinant is non-zero) ³. The conjecture asks if this local invertibility extends globally and if the inverse itself is also polynomial ³⁷.

Claude Fable 5 and the Disproof

In July 2026, Anthropic researcher Levent Alpöge, utilizing the AI model Claude Fable 5, reportedly produced a counterexample to the Jacobian Conjecture ²⁵⁶. This counterexample demonstrates a polynomial function whose Jacobian determinant is a non-zero constant, but which does not have a polynomial inverse, thus disproving the conjecture.

The Counterexample

• Nature: The counterexample is a three-dimensional polynomial map ²¹⁰.
• Complexity: It is described as a 216-character polynomial map ⁶¹⁰.
• Verification: The counterexample has been independently checked and verified by multiple mathematicians, including verification in Lean (a proof assistant) within hours of its discovery ²⁴⁶¹⁰.
• Specifics: Alpöge's counterexample is a C³ map with a determinant of -2 ²³. It exhibits a "three-point collision," meaning multiple distinct inputs map to the same output, which prevents a unique inverse ³.

AI's Role

Claude Fable 5 didn't "solve" the conjecture in the traditional sense of providing a formal proof or a general disproof strategy. Instead, it was instrumental in generating or identifying the specific polynomial counterexample ⁵⁶. This highlights AI's capability in:

• Exploration: Rapidly exploring a vast space of possibilities to find a specific instance that fits certain criteria.
• Pattern Recognition/Generation: Identifying or constructing complex mathematical expressions that meet specific conditions (like having a constant non-zero Jacobian determinant but lacking a polynomial inverse).
• Assisted Discovery: Acting as a tool that significantly accelerates the search for solutions or counterexamples, augmenting human mathematical research ⁵.

Significance of the Disproof

The disproof of the Jacobian Conjecture, especially with AI involvement, carries significant implications:

1. Resolution of a Long-Standing Problem: It closes an 87-year-old open problem in mathematics, providing a definitive answer to a question that has puzzled mathematicians for decades ⁶¹⁰.
2. Advancement in AI Capabilities: It showcases the growing capabilities of large language models (LLMs) and AI in complex scientific and mathematical reasoning ⁵. While Fable 5 didn't prove anything, its ability to generate a valid counterexample is a testament to its advanced understanding and manipulation of mathematical structures. This suggests AI can be a powerful co-pilot for mathematical discovery, not just for routine tasks ⁵.
3. Impact on Algebraic Geometry and Analysis: The conjecture has connections to various fields, including algebraic geometry, complex analysis, and differential equations. Its disproof means mathematicians will need to re-evaluate theories and assumptions that might have implicitly relied on its truth ⁹.
4. New Research Directions: The counterexample itself opens new avenues for research. Mathematicians will now investigate why this specific polynomial behaves the way it does, what properties it possesses, and whether there are broader classes of functions that exhibit similar characteristics. It might lead to new conjectures or refinements of existing theories ³.
5. Human-AI Collaboration Model: The discovery exemplifies a powerful human-AI collaboration model. Alpöge's expertise guided the AI, and the AI's computational power and pattern-matching abilities helped find the specific instance. This hybrid approach could become increasingly common in scientific discovery ⁵.
6. Remaining Open Questions: While the 3D case is disproven, the Jacobian Conjecture for n=2 (two-dimensional case) remains an open problem ¹³. The counterexample found by Fable 5 is for n=3, meaning the 2D version still stands as an unsolved mystery. This highlights that AI's current capabilities are still problem-specific and don't necessarily generalize to all instances of a conjecture.