Starting working on a problem, even if this exact problem has already been solved thousands of time, it's a given that most of the code you'll write will be thrown out.

And this is because writing code is part of understanding the problem. You need to modelize and try stuff to start understanding how big your problem is. And being able to call on the experience of previously solved problems, play in how fast you can solve harder ones, if even solving them at all.

This is what AI takes away from you. If your problem is hard, your prompt will always be bad, because you don't understand yet what exactly you're trying to solve. When the AI outputs you code, it doesn't help you understand your problem better, because the AI gives you exactly what you asked for. On the other hand if your problems are simple, you're slowly losing your ability to solve them, and thus, to solve harder problems.

AI takes also away your ability to communicate, because AI is very good at understanding you, even fuzzy requests will give you a good result. Before AI you had to search for the right terms. And those terms aquired on the way of solving a problem, gave you ability to communicate with others and state problems efficiently. (eg. "Programming video game AI", "Shortest path to goal", "A* Algorithm", congratz you learned something)

AI might makes things faster, it might be useful for some redundant task on large scale projects. Maybe. But overall, it's been taking away things from me more than it gave.

Stopped using AI for programming for a week. Probably will continue for rest of my life.