Claude's Cycles: Claude Opus 4.6 solves a problem posed by Don Knuth [pdf] (www-cs-faculty.stanford.edu)

Claude Opus 4.6, a hybrid reasoning model from Anthropic, has made headlines by solving a complex problem posed by renowned computer scientist Don Knuth regarding directed Hamiltonian cycles. This problem focuses on decomposing arcs in a specific directed graph into three Hamiltonian cycles, a challenge that had stumped researchers for some time. Knuth's excitement indicates a shift in perspective on generative AI, now recognized for its capabilities in automatic deduction and creative problem-solving. The significance of this achievement lies in its implications for both theoretical computer science and practical applications in AI/ML. Claude’s approach involved a combination of rigorous exploration techniques, including depth-first search and a novel "serpentine pattern" analysis, leading to the formulation of a Python program that effectively produced valid decompositions for various odd values of the parameter. The results not only provide a new solution methodology for graph theory problems but also highlight the growing role of AI in tackling deep mathematical inquiries, potentially reshaping future research efforts in this domain.