Ask HN: What do you do when your AI coding agent is doing its job?

I am into my early 50s and recently diagnosed with adult ADHD and maladaptive daydreaming.

I realized that I have struggled with it since I was 16.

I always wanted to work in programming, but it never worked out for me. I am now in a coding-adjacent job, as a sysadmin (Windows).

I have recently started doing coding - some side/hobby projects - with the help of AI. But I realize that when the Codex is doing its work, I kind of zone out or get busy with something else, and it is difficult to return to the task. So, I try to keep the coding sessions more interactive. I always ask Codex to go one step at a time. It is not very efficient, and it doesn't always work - also impractical sometimes.

I would like to know what other HN users generally do when their coding agents are doing their job.

The zoning out got better for me when I stopped treating the run as waiting time and started using it as the slot where I write the check, new prompt, new tasks. While it works, I review code, check new GUI, communicate on socials about my product. Mostly code review, prompting, because I run 3-4 agents in parallel on different coding tasks. They finish at a different time, so no free time for a rest.