I'm more productive with AI. Why doesn't my paycheck reflect that?
In the past couple weeks, I’ve shipped 32 new Fleet-maintained apps, authored 10 new guides, and fixed 2 bugs in FleetImporter. And that’s on top of doing my actual day job; meeting with customers, filing feature requests and bug reports, running workshops, and two work trips. In the past, this would have been months of work.
I felt good about that.
Then I did the thing I always do, which is ask the question everyone else seems to have stopped asking: whose problem did I just solve, and who gets the value from solving it?
I’m salaried. My pay doesn’t move whether I merge two PRs a week or twenty. There’s no line on my paycheck for “productivity bonus, AI-assisted.” My raises, when they happen, are annual, modest, and tied to market rate - not to how many multiples faster I’m shipping. So when AI makes some part of my job three times faster, that time doesn’t become mine. It becomes output. And output that costs the same to produce but generates more value for the company doesn’t automatically split itself between the person who produced it and the company that sells it.
I’ve made a version of this argument before, about Elon Musk’s vision of a post-scarcity future - the part where he never explains how ownership of AI-driven abundance actually reaches anyone who isn’t already an owner. That post was abstract. I was picking apart a billionaire’s hypothetical. This one isn’t abstract. This is just Monday.
This isn’t new, AI just made it easier to see
The idea that productivity gains should translate into better pay for the people producing them is not a radical claim. It’s how the American economy worked for about thirty years after World War II. Then, sometime in the late 1970s, it stopped.
The Economic Policy Institute has tracked this since 1948, comparing how much the economy produces per hour of work against how much a typical worker actually gets paid for that hour. From 1948 to the late 1970s, the two lines move together. Productivity goes up, pay goes up, roughly in lockstep. Then they split. From 1979 to 2025, productivity grew by more than 90 percent. Pay for a typical worker grew by about 33 percent. Productivity has outpaced pay by nearly three to one, and the gap hasn’t closed since.
That didn’t happen by accident, and it isn’t a law of nature. EPI’s own analysis is direct about it: the split happened because specific protections - a rising minimum wage, strong union rights, tight labor markets, higher taxes at the top - were deliberately dismantled starting in the late seventies. Take away the mechanisms that force gains to be shared, and gains stop being shared. The economy doesn’t need a villain twirling a mustache for that to happen. It just needs the guardrails removed.
So where did the other fifty-seven points of growth go? EPI answers that plainly too: mostly into the salaries of highly paid executives, and into higher returns for shareholders. Not stolen, exactly. Just reallocated, over and over, one policy choice at a time, until reallocation looked like the natural order of things.
My parents grew up believing something different: work hard, be more productive, and your paycheck reflects it. That was the deal for the first thirty years after the war, and it was still the deal they told me about growing up. I have never lived inside that deal. Not for one day of my working life. By the time I got my first job, the lines had already been split apart for twenty years. This isn’t a betrayal I experienced firsthand. It’s just the only shape the world has ever had, for me.
AI isn’t the cause of this pattern. It’s just fast enough that I can watch it happen to myself in real time.
The Jevons paradox, but it’s my calendar
I keep expecting the extra time to show up as free time, and it never does.
Rajiv Pant wrote about this well, drawing on an ethnographic study out of UC Berkeley’s Haas school that followed roughly two hundred employees at a tech company using AI day to day. The finding matched exactly what I’ve felt: AI doesn’t free up time. It expands what you feel capable of taking on. One engineer in the study put it about as well as it can be put - “you don’t work less.” You just get handed more.
That’s the Jevons paradox, and it’s not a new idea either. Economists have watched it play out with tractors, with email, with every efficiency gain that arrives without a hard limit on scope attached. The tractor didn’t shorten the farmer’s day. It gave them more acres to work. Efficiency, on its own, doesn’t distribute itself as leisure. It distributes itself as expectation.
And the gains aren’t spreading evenly even among companies. PwC’s 2026 AI Performance study found that nearly three-quarters of AI’s measurable economic value is being captured by just one-fifth of companies - the ones already positioned to convert productivity into revenue, not the ones simply handing tools to their existing staff. If the value concentrates that hard at the company level, I don’t have much reason to assume it decentralizes once it gets inside the building.
The real incentive isn’t a raise, it’s not getting cut
If there’s no bigger paycheck waiting on the other side of being faster, why do I keep doing it? I don’t think the honest answer is ambition. I think it’s fear.
There’s a scene in Office Space where Peter Gibbons lays out his entire theory of workplace motivation to a pair of consultants: he doesn’t have one. He’s not chasing a raise or a promotion. He does exactly enough to avoid getting fired and nothing past that, because he has absolutely no incentive to work any harder. For a long stretch of my working life, that was an accurate description of the deal. Meet the bar. Keep the job. Nobody’s paying extra for hustle nobody asked for.
I don’t think I live in that world anymore, and I don’t think you do either. You know the old joke about outrunning a bear - you don’t have to be faster than the bear, you just have to be faster than whoever’s next to you. That used to be a joke about a hypothetical. It’s closer to a description of my actual working conditions now. I’m not trying to outrun the layoffs. I’m trying to outrun whoever else they’re coming for this quarter.
AI doesn’t just make individual workers more productive. It makes some number of workers unnecessary collectively. If a workload that used to take four people now takes two, nothing about that math requires the company to keep paying four people to feel good about it. And this isn’t hypothetical anymore. I’ve written before about GM laying off 500 to 600 IT workers the same week it had roughly 80 open positions for AI-native roles, and GM wasn’t an outlier. As of the most recent Challenger, Gray & Christmas report, AI has now led the list of reasons employers give for layoffs for five straight months, and technology-sector job cuts this year are running 67 percent ahead of the same stretch of last year. I’d bet almost everyone doing knowledge work has run this calculation in their head at some point in the last year, whether they’ve said it out loud or not.
So I hustle. Not because I believe being faster gets me ahead. Because I’m trying to make sure that when the workload gets consolidated - and I think it will, in ways that are hard to predict from inside any one team - I’m not the one whose role turns out to be redundant. That’s not a productivity strategy. That’s a survival strategy wearing a productivity strategy’s clothes, and I doubt I’m the only person at my company, or in this industry, running it.
When a team of four becomes a team of two, someone looks at who shipped the most, who seems most fluent with the new tools, who’s the least replaceable this quarter. Which means the real competition isn’t between me and some abstract notion of obsolescence. It’s between me and the people I work with every day - the ones I actually like, the ones I’d normally be the first to help. I’ve built a chunk of my identity around being the person who shares their configs, helps competitors, answers the question in Slack before anyone else gets to it. That instinct gets a lot harder to justify once I start suspecting that helping a teammate look more capable is helping them outlast me.
Nobody says any of this part out loud, which is exactly what lets it keep working. If a group of coworkers quietly decided the safest move was to out-produce each other, none of them would announce it. They’d just start doing it, and call it professionalism. A workplace full of people privately racing each other isn’t a workplace that’s in any position to ask, together, why the race exists in the first place.
And if you’re one of my coworkers reading this and you hadn’t run that math yet - I bet you are now. That’s not a threat, and it’s not really an apology either. It’s just the unfortunate reality of where we are today.
Maybe I’m wrong about this
Maybe my employer isn’t profiting from my increased productivity the way the abstract version of the story implies. Maybe the honest version of “salary” was never a promise to track my output minute to minute. It’s a fixed price for variable effort, and I agreed to that trade with my eyes open. Maybe, if AI-driven productivity keeps compounding across an entire industry, competition for people who can use these tools well eventually pulls wages up anyway. Maybe the fear I described above is overstated - automation has historically created new categories of work almost as often as it’s eliminated old ones, so four-becomes-two doesn’t have to be the permanent shape of an entire industry, even if it’s brutal for whoever’s on the wrong side of it in the meantime.
But I’m not sure I buy it. “Reinvestment” and “shareholder return” aren’t actually different categories once you zoom out - they’re both just capital, deciding what happens to the value labor created, without labor getting a vote. Fixed price for variable effort was a reasonable deal when the variance was small. It stops being reasonable once one side of the trade can 3x their output and the other side’s compensation is still anchored to how things worked before that was possible. And “wages eventually rise” has been the promise since 1979. Forty-six years isn’t “eventually.” It’s just how it works now.
I’ve written before about whether a healthy company culture can survive inside a system built to reward the opposite. My answer then was: maybe, but only if you’re honest that the contradiction is a feature, not a bug. This is that same contradiction, just wearing a different hat.
I don’t have a clean ending for this
I’m not going to stop using AI at work tomorrow. Refusing to use it doesn’t return the surplus to me - it just means I produce less while the split stays exactly the same, and I’d be the first one on the list when the workload gets consolidated. That’s not a moral stance. That’s just arithmetic.
What I can do is stop pretending the speed is neutral. I’m not opting into a fairer world by being fast. I’m opting into being fast because slow feels dangerous, and someone else is deciding what that speed is actually worth.
I’ll open Claude again tomorrow morning. Not because I think it gets me ahead. Because falling behind starts to look a lot like being the one they let go. I just don’t plan on lying to myself about which of those two things is actually happening.