Fresh Context Coffee: The Human-in-the-Loop UX Problem
About Event
Human-in-the-loop made sense when the AI suggested one thing and a human said yes or no.
The promise of agents is to break that. And now agents are finally breaking this. They take fifty actions to complete a task, and the human reviewer is staring at a queue that grew faster than they can read.
So what does review actually mean now?
Sampling? Reviewing the trajectory instead of the steps?
Trusting the agent on the easy stuff and pulling the human in only when something looks off — and figuring out what "looks off" even means?
Founding teams the best agentic companies working through this. Chatham House rules.
Hosted by Exosphere
Rooms curated in the past: