The Missing Governor
AI removed the natural limits on how much we can produce. The burnout isn't from the work being harder — it's from the work never feeling done.
A UC Berkeley research group recently published findings from an eight-month study of 200 daily AI users. The headline number: 88% of the heaviest users — mostly developers — reported significant stress and burnout.
I relate to this more than I'd like to admit.
Before AI, there were natural limits on how much a person could produce in a day. Typing speed. Domain knowledge. The time it takes to look something up, test something, debug something. Those limits were real, and they were frustrating. But they also served a structural function that nobody designed and nobody talked about. They were governors — built-in constraints that capped output and, by extension, capped expectations.
I think what's happening right now is that AI has removed the governor.
The ceiling on individual output used to be skill. Now it's endurance. And when endurance becomes the bottleneck, work doesn't feel like work anymore. It feels like a test of character. You're not solving a hard problem — you're choosing whether to keep going or admit you've had enough. Those are very different kinds of effort.
A VC in San Francisco described a friend who left a party at 9:30 on a Saturday. Not tired. Not sick. He wanted to get back to his agent. Nobody questioned it. Half the room was thinking the same thing. The other half were already checking their agents' progress from the bar.
That made me think about what this actually is. It's not burnout from shipping too many features. I'd call it Governor Anxiety — the psychological cost of operating without a natural ceiling. When you can always do more, not doing more becomes a decision you have to defend. Not to your manager. To yourself.
And there's an economy built on top of that feeling. Every weekly model drop. Every "I built this in 20 minutes" post on X. Every agent demo that casually obsoletes your last quarter of work. The information diet of a working developer in 2026 is a continuous stream of evidence that you're falling behind. The anxiety isn't a side effect of the ecosystem. It's the engagement model.
As a Tech Director within AI, I spend more time inside these tools than most. So I'm not making an argument for slowing down — I don't think that works, and I don't think it's honest. People who build things want to build things. Telling them to take a break is like telling a musician to stop listening to music.
But I think there's a distinction worth drawing. There's building because you see something that should exist. And there's building because sitting still feels dangerous. The first one compounds. The second one depletes.
The Berkeley data confirms what a lot of us already sense: the tools are not the problem. The missing governor is the problem. We removed the natural constraint on output without building a replacement constraint on direction. The result is a generation of builders who can produce more than ever and feel worse about it than ever.
In my experience, the best work I've done this year didn't come from the sessions where I pushed hardest. It came from the ones where I was clearest about what I was actually trying to build. Not faster. Not more. Just — this specific thing, for this specific reason.
Maybe that's what the new governor looks like. Not a limit on capacity, but a filter on intention.
The old governor was external. It lived in the tools, in the clock, in the friction of the work itself. The new one has to be internal. And defining it might be the most important human & engineering problem none of us are working on.