AI Tools and the Multitasking Trap

AI coding tools have become a real part of how I work. They're fast enough to be genuinely useful — but they take a minute or two to finish. Not long enough to go do something else. Not short enough to just sit and wait.

Focus is already hard when you work on a small team. You're holding the whole system in your head — architecture, priorities, deadlines — and interruptions are constant. Sales questions, customer support, helping other devs, Slack, email, alerts, remembering to check your calendar for meetings. Every context switch costs you something.

With all of that already pulling at your attention, AI coding tools add a new wrinkle. I'll kick off a prompt in Claude Code, and while it's working, my brain starts looking for something to do. Check Slack. Glance at another file. Start speccing out another feature. By the time the response comes back, I've drifted — and I have to re-read what I asked and why.

We all know the answer. Focus on one task. Avoid context switching. Protect your flow state. I believe that. The best engineering work I've done has been when I stayed locked in on a single problem long enough to hold the whole thing in my head.

But that's hard when the tool gives you 120 seconds with nothing to do.

I'm still figuring out the right discipline for this. Some days I wait and watch. Other days I review the code around where I'm working — staying in context without starting something new. The goal is to stay close to the problem, even when the tool pauses.

New tools solve many problems but always introduce new challenges. It takes time to learn the most effective practices — and I'm still figuring this one out.

Do you wait for the AI to finish, or do you start something else?

Originally published on LinkedIn — view the original post for comments and reactions.