Hacker News
Martin Fowler Writes That He Dislikes LLMs on Grounds That Have Nothing to Do With Whether They Work
Fowler published "I Don't Like LLMs" on September 17, and it drew 235 points and 264 comments. He concedes the productivity case and objects on three other axes: the "grating LLM-voice" combined with false confidence that mixes useful answers with fabricated ones, the fact that the systems carry the values of their Silicon Valley creators, and a consistency argument from his own life, that he has spent decades choosing to spend time with pleasant capable people and sees no reason to suspend that for software. He is explicit that this is not an argument against using them. Coming from the person who spent two years writing carefully about agentic workflows, the shift in register is the story.
↳ Follow the thread