Exactly, AI support in IDEs useful for more than just producing ready made solutions. In fact, it isn't really good at that because of context size limitations

Exactly, AI support in IDEs useful for more than just producing ready made solutions. In fact, it isn't really good at that because of context size limitations (most non trivial sized code bases are simply too large). And it will make mistakes, omit stuff, etc. But, I've used it to review, debug, document, test, and explain code. I occasionally use it to stub out a solution but usually it then requires manual tuning. Any modestly complex algorithmic stuff it usually makes lots of mistakes. But it is able to spot uncovered edge cases, or suggest ways to test algorithms. I think more of it as an artificial pair programmer pointing out my mistakes, and helping me be more productive. It's also good at figuring out syntax issues. Often when dealing with languages I'm less proficient in the main battle is simply figuring out how to do X in a language. If you are trying to learn things like Rust (which is hard to learn), AI is your best friend. Because it can explain you your code, your mistakes, and suggest alternate ways of doing things. All you need to do is ask. Remembering to do that is the hard part.