I don't think this is an issue of experience here. I don't know that anyone has claimed that LLMs can create truly novel solutions to complex problems given that the technology is token prediction.