Ask HN: How do you deal with fragmented attention?

Since coding agents have become more capable, I often have them working on 2-5 disparate tasks simultaneously during my work day. The tasks are often in separate or even unrelated repos. Or worse, sometimes the tasks are similar enough that they are easily mixed-up.

Much of my attention goes to switching: Checking on results, restarting stopped tasks, fixing errors or providing hints, etc.

The context-switching takes a toll in the form of distraction and inefficiency.

What do you do that helps with this?