The problem
Let F: ℂⁿ → ℂⁿ be a polynomial map whose Jacobian determinant is a nonzero constant. Must F be invertible with polynomial inverse? Keller asked for n = 2 in 1939; the case n = 3 carries the same reputation for truthfulness.
Let F: ℂⁿ → ℂⁿ be a polynomial map whose Jacobian determinant is a nonzero constant. Must F be invertible with polynomial inverse? Keller asked for n = 2 in 1939; the case n = 3 carries the same reputation for truthfulness.
On Smale's 1998 list of mathematical problems for the 21st century. Decades of partial results (injectivity criteria, reduction attempts) all pointed the same way as intuition: surely true. Almost nobody was trying to disprove it.
Disproved — Levent Alpöge (Harvard), announced 19 July 2026, working with an AI system he credits as a substantial collaborator. The counterexample fits in 216 characters: a concrete polynomial map in three variables with constant Jacobian 1 that admits no polynomial inverse. Experts (Abhishek Saha, Queen Mary) note the single-line answer was verified almost immediately across the community — the surprise is not the checking but the finding. The two-variable case survives and remains open; the result is widely described as the hardest mathematical problem yet significantly cracked by AI.