Am I missing something here?

What terminal / agentic workflows spawn the demonstrated dialogue boxes that require the User to click/reject the action? Aren't most such flows actually inline UIs? And finally, if the core issue is that these confirmation boxes are tied to the terminal that triggered them, which does not autofocus, what's the point of said "big arrow" that is also lurking behind without focus?

If the said "big arrow" automatically gains focus, isn't the real fix here to just make the dialogue boxes themselves gain focus automatically? Both are similarly disruptive anyway.

The first example (of HN) is the one that feels like it has the most potential to me.

"Teach me to do this" kinda stuff. "Guiding agent" rather than "doing agent".

Honestly, @franze, if you're reading this, maybe update your screenshots to show an agent in tutorial mode on some complicated app?

I think this would’ve been very handy to me a couple years ago when I was learning to use Blender and asking LLMs for help performing certain actions like “how do I display the normals of all the vertices in my mesh?” I spent a lot of time trying to figure out which button the model was talking about.