> AI thrives in messy, open-ended spaces

Isn't the exact opposite of what AI is good at? That sounds like a great way to get tons of hallucinations. Every "how to AI good" guide I read is all about providing it with ample structure and narrow instructions.

To be reductive about the example of a malleable task tracker (e.g. Notion), if the LLM correctly interprets the user's prompt, there's about 20-30 useful buttons in the interface, and there's a sequence of those buttons that achieves the user's desired outcome. It's open-ended in terms of what you make, but the set of available actions is fairly tight, and the problem is finding a sequence of those actions.