LLMs reward expertise, not prompting tricks

Sean Goedecke’s “LLMs reward expertise” makes a case worth sitting with: the ceiling on what you get out of an LLM is set by your own domain expertise, not your prompting technique. His example is mathematician Terence Tao getting real value out of ChatGPT on a hard problem — not because he prompts specially, but because he can instantly tell a promising lead from a dead end, something a novice can’t do. The Hacker News discussion pushes back a little around the edges, but the core point holds up: the model isn’t the bottleneck, the human’s judgment is.