Reliable Machines From Unreliable Parts
Why LLM errors need to be designed around rather than waited out, what John von Neumann worked out about unreliable components in 1952, and why a decision-only model like Jev fits his framework better than a chat model does.
