Working with AI Agents
Why people feel more burnt out with AI Agents instead of reclaiming their time back
The Mental Stack
Your brain can hold about 4-7 things at once. That’s it. Every unfinished task, every deferred decision, every open loop competes for that space.
Engineers call this a stack overflow. HBS researchers now call it brainfry.
AI agents made it worse. You’re now managing your own work plus a dozen parallel threads you’ve delegated: a Claude writing a draft, a Cursor fixing a bug, an agent researching a market. Each one needs a context switch to check in. Each switch costs stack space. The productivity gain is real, but so is the overhead of orchestrating it.
Most people respond by trying to hold more. Better apps, more tabs, tighter schedules. But the problem isn’t your system. It’s the volume.
Every possession you own, every commitment you haven’t resolved, every agent thread you’re half-tracking is running in the background. The more you pile on, the less you can do with any of it.
A few things actually help:
Write things down before you delegate. Externalizing a task clears it from your stack. If you’re spinning up an AI agent, write the brief, then let it go. Don’t half-hold it.
Batch your check-ins. Treat agent outputs like email. Check them at set times, not continuously. Constant monitoring is just a new form of multitasking.
Limit active threads. Three parallel workstreams is about the limit before you’re spending more energy tracking than doing. More agents doesn’t mean more output.
Resolve open loops daily. Anything you’re “meaning to get back to” is occupying space. Decide, delegate, or delete it before it becomes background noise.
The stack is limited. AI gives you more horsepower, but the cockpit is still the same size. The people who benefit most aren’t the ones running the most agents. They’re the ones who stay clear on what they’re actually trying to get done.