That was a big concern when TPMs first started appearing. I’m glad that that didn’t go the way it might’ve. Yet.

I’ve had similar experiences. Try to do something semi-difficult and it fails, sometimes in an entertainingly shit way at least. Try something simple where I already know the answer? Good chance there’s at least one fundamental issue with the output.

So what are people who use this tech actually getting out of it? Do they just make it regurgitate things from StackOverflow? Do they have a larger tolerance for cleaning up trash? Or do they just not check the output?