I'd be curious about the quality of the output. In my experience so far, even when programming with AI, I have to stay constantly engaged to give instructions, review, discuss, revise, etc. Otherwise things would go in a completely different direction from where I want them to go. In other words, it remains a full-time job for me. It's hard to imagine that, in its current form, AI could independently produce something I'd really want without this constant engagement and supervision from me. But then I haven't seriously tried to use agents. So I'm curious as to the quality of the output they are capable of producing if left to their own devices.