Knuth Said 'Shock! Shock!' — And If You Know Knuth, You Know What That Means

Donald Knuth opened his latest paper with two words: "Shock! Shock!"

The most respected computer scientist alive — the man who wrote The Art of Computer Programming, who invented TeX, who has spent 60 years defining the field — was stunned.

I have a personal relationship with Knuth's work. Not a casual one. I learned most of my hardcore CS skills and spirit from his writings. The elegance of his proofs. The discipline of his analysis. The way he treats every algorithm as a craft to be perfected, not a problem to be solved. That rigor shaped how I think about engineering to this day.

So when Knuth — a man who has been skeptical of AI for decades — writes that Claude Opus 4.6 solved an open combinatorial problem he'd been working on for weeks, in one hour, I pay attention differently than most.

Not approximately. Not partially. Solved it.

Claude reformulated the problem, recognized it as a Cayley digraph, tried brute-force approaches, abandoned them, invented a fiber decomposition strategy, hit a dead end, backtracked, analyzed a simulated-annealing solution for patterns — and found the construction.

Then came the part that matters most:

Claude found the answer but couldn't prove it was right.

Knuth still had to write the proof himself. The AI's intuition was extraordinary. Its rigor was not.

This is the pattern I see in enterprise AI every single day. And it's exactly what Knuth's elegant mind would have predicted — the machine can explore the space faster than any human, but understanding why the answer is correct remains a fundamentally human act.

The companies deploying AI well understand this. They don't ask "Can AI do the job?" They ask: "Where in this workflow does AI explore, and where does a human verify?"

Knuth wrote: "It seems that I'll have to revise my opinions about generative AI one of these days."

If the man whose writings taught me what elegance in computer science means is revising his priors, what are you still holding onto?

Read the full paper: Claude's Cycles — Donald Knuth (PDF)