> artificial creativity

This artificial creativity will only go so far, because it's a simulated semblance of human creativity, as much as could be gathered from training data. If not continually refueled by new training data, it will run out sooner or later. And then it will get boring really quickly.

But it is being continually refueled. The output of an LLM, at least in the process of generating code, is a combined product of human creativity and the LLMl. I have told it what to do, fixed what it got wrong, and verified the solution was correct through testing.