Donald Knuth, the guy who literally wrote the book on computer science, just watched Claude solve a problem he'd been grinding on for weeks.
The Signal
This isn't some startup founder hyping their product. This is Donald Knuth, author of "The Art of Computer Programming," the man who invented TeX because existing typesetting wasn't good enough for his mathematical writing. He's been skeptical of AI hype for years. And he just published a PDF (naturally) saying Claude Opus 4.6 cracked an open problem in graph theory that stumped him.
The problem involved finding cycles in directed graphs, pure mathematics territory where reasoning matters more than pattern matching. Knuth spent several weeks on it. Claude, specifically Anthropic's new hybrid reasoning model released three weeks ago, got there first. Knuth's reaction isn't defensive or dismissive. It's genuine scientific excitement. "What a joy," he writes, celebrating both the solution and "this dramatic advance in automatic deduction."
The timing matters. This is Opus 4.6, Anthropic's hybrid model that combines their standard transformer architecture with explicit reasoning chains. Not GPT. Not o1. A different approach to the same hard problem: making AI actually think through complex problems instead of vibing its way to an answer. And it's working on problems that require real mathematical creativity, not just summarizing documents or writing code from examples.
The Implication
When the living legends of computer science start revising their opinions about AI capabilities, pay attention. This isn't about replacing programmers or automating customer service. This is about AI doing novel mathematical research. Watch what happens when these reasoning models get integrated into actual scientific workflows. The agent economy just got a lot more interesting.
Source: Daring Fireball