I agree with the larger point you’re making, but an LLM absolutely can ask the right questions to determine scope and features for a basic application. It can even turn those into decent user stories. It’s the edge cases and little details that will be messed up.