not OP but the criticism of "what's the point of a mediocre side project" does hit different when you're using LLM to do 90% of the work.

makes me think we're talking about two types of side projects. "side hustles" and work you'd enjoy doing irrespective of money.

The point is that once you throw a moderately complex problem at it, you’ll typically get something that at least subtly wrong. So now instead of having fun coding, you’re rather debugging hundreds of lines of „someone else’s” code.

Heavily disagree; with guidance, these AI tools can solve problems that otherwise require in-depth expertise. Someone ported the HD version of my game[0] to MacOS using GPT with no prior graphics knowledge.

[0]https://2009scape.org

Exactly. If you just throw things at it, it will fail. You have to break down the complex problems into sets of simple ones (for now). Then it can shine.

Sure, and at no point did the author say "look at how this thing solves complex problems for me". They wrote about how they use these tools to build neat side projects that they otherwise wouldn't have time to build. That's worthwhile.

This is the computer science equivalent of making a straight up sock puppet with googly eyes.

It really isn't though. https://jsonplayground.com/ is a good (small) project. I would be proud to put my name to it.

You must have low standards then. This is something that anyone can do. It's like making a hammer from scratch with iron you dug up from an old mine. I mean how innovative is that really?

Just as a starting point, the majority of developers likely haven't tried running jq using WebAssembly. Heck most developers probably still don't even know what jq is.

You seem to be hung up on "This is something that anyone can do" and "how innovative is that really?" - that's NOT the point of this. This isn't about building innovative things that nobody else could build, it's about being able to churn out small, useful projects more productively thanks to assistance from LLMs.