Agentic UX ruins the coaching

One more thing that's annoying when using AI as a coach: any amount of reasoning tokens will make me switch away. Latency actually matters here. Two seconds is okay; twenty ruins the flow.

There is also something more subtle. I don't take as much responsibility for my thoughts, am not as critical. I don't keep a mental model in my head of what is happening, and am trusting the AI to do that too much. Vibe-self-coaching. Vibe-thinking? It might be the worst part of using AI this way. If you outsource the thinking, you cannot genuinely change. Or worse, you'll change, but in some random harmful direction.

Both problems trace back to the same root: I'm running my coaching sessions in an agent UI like Claude Code. It seems like a small thing, but all my usual Claude habits, fine for everything else I use it for, now get in the way. Multitasking and doing work so you don't have to think (that's what agents are!) are key design goals for a system like that. So running a session there actively works against the success of the coaching.

A good AI coach, then, can't live inside a chat window or an agent UI. It needs to be expensive enough to take seriously, as I argued in AI coach, and an immersive experience of its own.