I am, despite my distrust of code generating LLMs, using one to write a piece of software that I have intended to write for a while, but haven't, thanks to lazi

I am, despite my distrust of code generating LLMs, using one to write a piece of software that I have intended to write for a while, but haven't, thanks to laziness and inertia. It's interesting, but I'm finding I'm spending a lot of time just figuring out how to describe something to the LLM. A bit less than writing it myself, really, except for the Qt-related stuff, which I've read up on and practiced a bit, but I wouldn't say I'm competent, yet. The generated code is OK and it works the first time, exactly the way I describe it to the LLM. I believe you can easily spot the problem there. It will take a while to learn how to integrate AI into my workflow, and I can't say it's exactly an enjoyable experience, but I feel like it's something I need to do. I do feel it's a crutch, though. I will say, there are two areas where it shines so far: writing tests and making me practice my code reviewing. We'll see how it goes. It's going to take amazing results to ever get me to pay a monthly fee for something running remotely that seems like it should run locally.