Donald Knuth Publishes Paper on Claude AI Solving an Open Graph Theory Problem

Donald Knuth Publishes Paper on Claude AI Solving an Open Graph Theory Problem

In early March 2026, Donald Knuth — widely regarded as one of the founding figures of computer science and the author of the seminal series The Art of Computer Programming — published a paper titled "Claude's Cycles" documenting a striking encounter with Anthropic's Claude Opus 4.6.

Knuth described his reaction as "Shock!" upon discovering that the AI had solved a complex, open problem in graph theory: constructing Hamiltonian cycles in a specific class of 3D directed graphs. Knuth had been working on the same problem himself while preparing new material for his book series.

Why This Matters

Hamiltonian cycle problems — which ask whether a path exists that visits every vertex in a graph exactly once and returns to the starting point — are computationally hard in the general case and belong to the class of NP-complete problems. Knuth's specific graph structure presented a non-trivial instance that had resisted his own attempts at a solution.

The paper represents one of the most credible independent assessments of frontier AI mathematical reasoning to date. Knuth has historically been a careful and skeptical observer of AI claims, making his documented surprise significant. The publication adds to a growing body of evidence that large language models are beginning to make genuine contributions to hard problems in mathematics and computer science.