Now that coding agents can run with minimal human effort, what should the human do during that time?

Do we just become “watchers” who wait for results? Should we be monitoring them closely, or is it better to step away and do something else—like reading, watching a lecture, or even playing games—while the agents run?

For people who use coding agents regularly: what’s your workflow? How do you stay productive without hovering, and how do you decide when to intervene?