Sean Goedecke's 'LLMs Reward Expertise' Hits 1,042 Points on HN — The Bottleneck Is the Human, Not the Model
Sean Goedecke / Hacker News·high signal
Sean Goedecke argues the most important prompting skill is domain expertise, not prompt-engineering technique: LLMs let anyone produce 'sort-of-okay' output in unfamiliar domains, which masks the gap between experts and non-experts. His central case study is Terence Tao's ChatGPT conversation on a Jacobian Conjecture counterexample, where Tao steers hard by asking precise questions, flagging over-complex answers, and proposing his own directions rather than following the model's. The post drew 1,042 points and 442 comments — the largest HN AI discussion of the last 48 hours.