On July 20, mathematician and Anthropic researcher Levent Alpöge announced a potential counterexample to the Jacobian Conjecture, a foundational problem in algebraic geometry originally proposed in 1939. The discovery was generated by Anthropic’s Claude Fable 5, which autonomously constructed a three-variable polynomial mapping that satisfies the conjecture’s premises but directly contradicts its conclusions. The initiative began when Alpöge’s colleague Akhil submitted the problem to the model, which subsequently executed an extended computational search and returned the candidate mapping. The proposed counterexample features a polynomial transformation from three-dimensional complex space to itself with a Jacobian determinant that remains identically equal to −2. Because −2 is a non-zero constant, the mapping fulfills the conjecture’s core requirement. However, when three distinct input points are processed through the transformation, they converge to a single output coordinate. This non-injective behavior definitively violates the conjecture’s claim that the mapping must possess a polynomial inverse. The mathematical community has moved quickly to validate the finding. Researchers affiliated with Stanford University and contributors on platforms like MathOverflow have launched independent verification efforts using computational algebra systems such as Wolfram Alpha and SymPy. Preliminary results confirm both critical mathematical conditions: the determinant remains constant and non-zero, and the collision of multiple inputs to a single output is mathematically sound. Should the counterexample withstand further scrutiny, it would conclusively disprove the Jacobian Conjecture for three-dimensional and all higher-dimensional cases. Since the conjecture requires universal validity across dimensions, its failure in three dimensions invalidates the entire proposition, leaving the two-dimensional scenario as the sole remaining open question. Historically, the Jacobian Conjecture has resisted resolution for nearly nine decades, frequently attracting unproductive efforts from both academic mathematicians and independent researchers. Fields Medalist Stephen Smale notably ranked it among the most consequential mathematical challenges of the coming century. This breakthrough, however, diverges from traditional proof methodologies by leveraging AI-driven exploration of an uncharted search space. Alpöge, who holds advanced degrees from Harvard and Princeton before transitioning to AI research at Anthropic, emphasized that the model successfully isolated a novel mathematical object rather than verifying human-generated hypotheses. The development marks a significant shift in computational mathematics. Large language models are increasingly moving beyond assisting with formal proof validation to actively navigating abstract mathematical landscapes. By autonomously generating candidate structures that human researchers have not identified, AI systems are demonstrating the capacity to challenge established conjectures and redefine the boundaries of algorithmic discovery in pure mathematics.