As part of my “learn Python” project, I’m building a desktop RSS reader. This could be a pretty straight-forward little dingus, but I want to use it to explore structuring code in a way that can be extended and tested later, so I’m trying a class-based approach. It has been like a lot of the other creative things I’ve done over the years, in the sense that you work on it for a bit, create something bad, then smoosh it down and start over with what you’ve learned.
My big take-away so far is that (SHOCKING!!) ChatGPT is not as useful as it looks at first. OK, so it definitely IS useful. You can ask it things like “how would I build a basic GUI using Python?” and it gives you Tkinter syntax examples to get the ball rolling. But it reminds me of doing translation (my day job) using a machine translation engine: 1) you get something that looks good; 2) great! you set out to tweak it and clean it up; 3) you find out it needs a fundamental re-write; 4) you’ve Ship of Thesus’d the thing, and it took longer than just doing it from scratch.
Still, it’s been useful when I get stuck, and it’s a nice place to get quick answers to easy questions. What this experience is reinforcing for me is that humans are still best at the big-brained strategic/creative planning. While something like Code Pilot can help you build the pieces, you’re the one who has to know how to put the pieces together.
Leave a Reply