I spent the evening working on an app called Poster. The app cross posts between Bluesky and Mastodon/GoToSocial. It’s not special, but it was fun to write. It took awhile, since my Python is still very noobish, and very much a “how do I do this” -> “Google how to do this” workflow.

But I did not write the README.md you see in that git repo. I had Claude do that for me (along with fixing a whole host of bugs because I’m shit at fixing my own code). I don’t feel bad about having Claude do that for me. Maybe I should, but it has to be better than no README at all. And Claude is pretty damn good at looking at some code and telling people how to use the outcome. Better than I would have been, at least.

I do use Claude code when I’m working on my projects. A lot of the time I can get an answer and an explanation faster than going to Google. I often learn the wrong way to do things from AI, to be sure, but there is something about the interactive nature of it that makes it seem more appealing. For me, it really is the bugs thing that makes Claude code so good. I can take a mess of code that I wrote, feed it into Claude, tell it to clean it up and then tell my why I’m getting these tracebacks and then have it fix the problem. I should get better at fixing these things myself, and I will, but for now it has saved me loads of time. It also really helps with indenting, because I’m crap at doing that consistently too.

I get the whole anti-AI thing. I’m 50% of the way to the idea that it’s going to kill us all, probably by just making us all dumber, but it can be useful as long as you know that it talks shit 80% of the time.