This has been one of the things I keep raising about LLMs and other so-called expert systems. Once everyone is reliant on them, where are the experts to train the next model? If novices can skip the trial and error part, how are they meant to learn and get experience?