I agree that prompting is a much shallower skill than engineering, and I really hate the term “prompt engineering” since I feel it puts something that (by design) is easy to learn on the same level as something relatively hard to learn.

There is admittedly a bit of skill to use these tools, but they’re really not hard to use. For the most part Claude can decipher my typo’d-to-hell prompt and figure out what I want, and even if it gets it wrong the first time I can usually correct it. The hardest part is determining some objective test for Claude Code to test against and “anchor” it, but I still think that’s easier than writing a bunch of unit and integration tests.

I like these tools; I think Claude Code is very useful and even if I didn’t like it I would still use it because it’s clear that’s where the world is headed and I want to stay competitive.

For personal projects I actually care about, I still write code by hand. I do think I write better code than Claude but I would still do it by hand even if I didn’t, just because I realized I was losing my intuition about software because I was outsourcing to Claude for everything.

There are plenty of things that I am honestly fine losing my intuition with. I don’t really derive much enjoyment from reading dmesg logs, and I am more than happy to let a robot have its fun with that, but it can take the mathey shit from my cold dead hands now.