Anthropic's AI Model Solves the 87-Year-Old Jacobian Conjecture
Key Takeaways
- ▸Anthropic's AI model disproved the Jacobian conjecture by finding a counterexample after 87 years of remaining unsolved
- ▸Part of a rapidly accelerating trend of AI solving longstanding mathematical problems, with multiple breakthroughs since mid-2025
- ▸AI provides correct answers but lacks the explanatory reasoning mathematicians consider essential for true understanding
Summary
Anthropic's language model has successfully disproven the Jacobian conjecture, a mathematical problem unsolved since 1939. The breakthrough was announced by Anthropic employee Levant Alpöge on X, where the post drew over 20 million views. The AI demonstrated a counterexample—a mathematical "map" with a constant Jacobian determinant of −2 everywhere that maps three different starting points to the same output—proving the 87-year-old conjecture false.
This achievement marks another milestone in an accelerating series of AI-driven mathematical breakthroughs since mid-2025, when AI first solved five of six International Mathematical Olympiad problems. Subsequent victories include OpenAI's disproof of an 80-year-old Erdős conjecture in May. The trend has prompted 16 researchers to publish the Leiden Declaration on Artificial Intelligence and Mathematics, calling for guardrails around transparency, attribution, and peer review before AI reshapes how mathematical knowledge is created and validated.
However, the breakthrough exposes a fundamental tension in AI-driven mathematics. While the model produced a correct answer, mathematicians note that AI provides the "how" without the "why"—the explanatory reasoning and intuitive understanding that forms the foundation of mathematical proof. The field now faces an unprecedented challenge: AI can reach problems beyond human mathematical capacity, yet struggles to provide the proofs and reasoning that mathematicians traditionally use to validate and understand mathematical truth.
- The mathematical community is establishing new standards for transparency, attribution, and peer review in AI-driven research
- The achievement demonstrates both the potential and limitations of AI at the frontier of pure mathematics
Editorial Opinion
This is a watershed moment for both mathematics and AI. While Anthropic's solution demonstrates that language models can tackle problems at the frontier of mathematical knowledge, the field's discomfort with these victories is telling. When an AI can find correct answers but cannot explain the underlying reasoning, we must ask whether we're advancing mathematics or simply automating calculation at a higher level. The mathematical community is right to insist on rigorous standards around proof, transparency, and understanding before allowing AI to reshape what mathematical knowledge itself means.
