What do you do while your AI agent is working?

I used to work with the Pomodoro technique: 25 minutes of focused work followed by a 5-minute break. It worked surprisingly well before AI-code-agents-era because most of my day was spent writing code. Last year (especially last 6 months!), I’ve noticed that this rhythm no longer fits the way I work. When I’m working with coding agents, I often find myself in a strange situation. I give the agent a task, it starts working, and I have two, three, maybe five minutes of free time. That’s not enough time to switch to something else without losing context, but it’s also too much time to just sit there waiting. And once the agent is done, I still need to review the code, test it, think through edge cases, and decide what the next step should be. The challenge isn’t that there’s nothing to do, is that there are plenty of things to do, but most of them require a context switch that costs more than the few minutes the agent is working. What’s interesting is that the bottleneck seems to be shifting. A few years ago, most of my time was spent turning ideas into code. Now, more and more of my time is spent deciding what should be built, explaining it clearly, and evaluating the result. The actual coding is gradually becoming a smaller part of the process. I’m still not sure what the best use of those small gaps is. Sometimes I prepare the next prompt. Sometimes I think through the architecture. Sometimes I just stand up, grab some water, and let my brain continue working on the problem in the background. But it does feel like we’re developing a completely new way of working, and I’m curious whether others have run into the same thing.