LLMs make basic work easier, but expert users get better results by steering models with domain knowledge

The article argues that while LLMs let nearly anyone perform generalist tasks like writing passable code or prose, effective prompting still depends heavily on expertise. It uses Terence Tao’s ChatGPT exchange about advanced mathematics to show how a domain expert can guide, challenge, and extract better output from the model.

As LLMs commoditize baseline skills, the advantage shifts toward people who can evaluate and steer model output in specialized domains. This suggests domain expertise may become more valuable, not less, because experts can use AI as a collaborator rather than a black-box answer machine.