This post on using LLMs to write code by Simon Willison helped me why LLMs aren't appealing to me and what goals they serve. In short, I don't want to manage an eager intern, and I'm not interested in creating features that wouldn't be worth writing if I had to understand all the parts myself. The one task that appeals to me is using an LLM to describe a new code base. But my overall take is that using LLMs to write code is an efficient technical debt generator :)