A new AI “harms” article from Harvard basically proves a simple equation. Boredom is a signal to stop, and stopping is what lets you recover. Even more to the point, boredom is what makes us happy, like saying a vacation on a beach with nothing to do is a great way to become more productive again.

AI Doesn’t Reduce Work—It Intensifies It

AI eliminates the friction that used to trigger that signal. No large blank page, no unknown starting point, no waiting. Every moment is being curated by an army of “attention seeking” engineers to get you to sit longer and “just send one more prompt.” The stimulation never breaks long enough for boredom that would allow you to recover and regain happiness.

What’s particularly sharp in the article findings is the mechanism of denying happiness by replacing it with simulation. A chat doesn’t feel like work, yet it drains people as much if not more than work. Prompting feels like chatting, with someone who drains you. You end up training the thing that was supposed to show up trained. So the usual internal alarms like “I’m overworked, I should stop” somehow never fire. You’re wasted by the labor, but also you’re wasted by the absence of boredom. The recovery periods just evaporate.

The “AI practice” recommendations at the conclusion feel forced and formulaic. They read like telling someone to schedule boredom, which ends up being a contradiction that tells you how deep the problem goes. You can’t intentionally recreate the thing that only works when it happens unintentionally. I mean you could force people to take mandatory leave, but you aren’t achieving the purest form of boredom where it just shows up because there’s a natural barrier instead of a compliant one.

“Protected intervals to assess alignment” is corporate for “stare out the window for five minutes,” except now you’ve turned staring out the window into a compliance task to stress about. And if you get into a chat about when and how best to stare out the window… that’s the cycle.