Maybe it's because I'm not a professional developer and there's a good chance I'll forget things when I revisit my code months (or years) later, or maybe it's because I took Perl Best Practices too much to heart, but I prefer descriptive variables and more comments, and I write my code that way. One of the challenges I have found with using LLMs as code assistants is that they don't necessarily understand why I'm doing what I'm doing, and the comments and variable names they return aren't necessarily helpful for my purposes, and so I will rewrite them to fit my overall app better.

But that speaks to my experience with LLMs overall: if you think of them as your intern / undergraduate research assistant, they can be very helpful. They can do something you basically already know how to do quickly and save you some time, as long as you check their work carefully. Or they can do something you can't necessarily do (like, say, port a python library you want to use into perl) more quickly then you could if you didn't know python (or maybe even if you did). But they're not great at intuiting the contexts in which you're working, and they sometimes hallucinate, so you still have to check their work.