Almost every conversation I have about AI lands on the same question.
Will it take my job? Will it take yours? People argue about the timing, fast or slow, five years or fifty, but the question underneath is always that one. This whole thing is about AI, work, and who ends up holding the money. And I’ll be honest, for a long time my answer was the obvious one. Get faster. Get better at the work. Build a lean stack of AI tools and run marketing with AI better than anyone. Stay one step ahead of the machine.
Then I started reading the economists who study this for a living. And they’re not even asking that question anymore.
They’ve moved on to a sharper one: when the machine can do the work, who owns the machine?
The bread is going free
Picture the work you actually sell. The copy. The campaign. The analysis. The deck you stayed up finishing. Call it the bread, the thing your labour produces.
That bread is getting cheaper to make. Fast, and for everyone.
“I’ll just bake faster with AI” feels like a plan. The problem is it’s a plan literally everyone has. Every freelancer, every agency, every in-house team is reaching for the same tools to make the same bread quicker. When everyone can produce the work cheaply, the thing you were charging for, the producing, stops being rare. And rare is the only thing people pay a premium for.
Even the optimistic reads agree the work itself is exposed. Goldman Sachs put a number on it back in 2023: about a quarter of US work hours are automatable, with a base case where the work gets complemented rather than erased, plus a big productivity jump (Goldman Sachs). Take the gentle version of that and it still lands in the same place. The hours you sell are exactly the hours the machine is reaching for.
So if your whole edge is that you make the bread, and now the bread bakes itself, the edge thins on its own. Not because you got worse. Because everyone else got faster at the same time you did.
What the economists are actually arguing about
The economists stopped arguing about jobs a while ago. What they moved to sits one layer underneath the whole question.
There’s a paper by Anton Korinek and Donghyun Suh that does the cleanest job I’ve found of laying out the two futures (NBER w32255). It hinges on one question, and the question is weirdly specific: is the set of tasks only humans can do bounded, or unbounded? Is there always some new corner of work the machine can’t reach, or do we run out of corners?
If we never run out, in their words, “wages may rise forever.” There’s always higher work to climb to, so human labour keeps its value. But if the things only humans can do turn out to be a finite list, and the machine eventually gets all of them, then, again their words, “wages collapse.” Toward subsistence. Toward almost nothing.
The part that pulled me up short: the squeeze doesn’t wait for the machine to finish the job.
They describe a middle case where, in their phrasing, “large-scale automation outpaces capital accumulation and makes labor too abundant.” Read that plainly: there’s still plenty of human work to go around, but there are too many humans able to do it, so the price of doing it falls anyway. Wages can slide while you still have a job. That’s the part that makes this feel near instead of sci-fi.
Now name the thing underneath. When the machine catches most of the tasks, your income stops coming from working and starts coming from owning. From owning a slice of the machine, or of something that earns the way the machine does.
That’s the whole game in one line. The return to your labour is the variable that gets squeezed. The return to what you own is the one that doesn’t.
The bread, versus the bakery.
I want to be straight that this is a live argument, not a settled prophecy. Korinek has a second paper whose abstract literally frames it as open, asking “what roles may remain for labor post-AGI, and which production factors will grow in importance” (NBER w32980). Nobody has the answer yet. But notice that even the open version of the question has already left “will it take your job” behind. The people doing the real work moved the goalposts to ownership, and most of the internet hasn’t caught up.
It was never about affordability, it’s about ownership
There’s a comforting story going around that goes like this: the machines will make so much money that there’ll be plenty to share, AI profits just pay everyone a basic income, no new taxes, everybody’s fine.
Half of that holds up. The other half is wishful, and the gap between them is the whole point.
The half that holds: the pie can plausibly get big enough. A recent paper by Maryam Nayebi runs the threshold and finds that AI systems would need to reach only “5-7 times today’s automation productivity to fund an 11%-of-GDP UBI” (arXiv 2505.18687). That’s a real, checkable feasibility number. So “we simply can’t afford to share it” is probably not the thing standing in the way.
The half that doesn’t hold: that the money routes itself to people on its own, painlessly, with nobody having to decide anything. Being able to afford something in principle is not the same as having a mechanism that actually puts it in your account. Affordable and automatic are different words.
So the hard part was never the arithmetic. It’s the ownership. Who holds the machines, and what, if anything, gets shared. That’s a power question, not a math question.
You know it’s the real fight because the people at the very top are already having it, out loud, right now.
Sam Altman, who used to push universal basic income, has pivoted. In his words he’s “much more interested in ways where we think about kind of collective ownership that could be in compute or in equities or something else” (Fortune). A flat cash handout, he says, doesn’t get at what’s needed as “the balance between labor and capital shifts.” The man running one of the labs is telling you the thing that matters is the balance between labour and capital. Between baking and owning.
Come at it from the opposite end of politics and you land in the same place. Bernie Sanders just put forward the American AI Sovereign Wealth Fund Act, which would take a one-time 50% stake in the largest AI firms and hand it to a public fund, with the upside paid out to people (Common Dreams). His line for it: “No longer would the future of AI be dictated by a handful of Big Tech oligarchs.”
Sit with that for a second. Altman and Sanders agree on almost nothing. And they are both, right now, in mid-2026, arguing about the exact same thing. Not whether AI takes your job. Who owns the bakery.
So what does owning the bakery actually look like
I want to bring this down to ground, but I’m not going to hand you a four-step framework, because I don’t think this is a four-step kind of thing.
Owning the bakery means turning the hours you work into something that keeps earning after you stop working them. The bread feeds you this month. The bakery feeds you next year. One is a thing you sell once. The other is a thing you keep.
The contrast that made it click for me: the same AI that lets you bake ten times faster lets the next person bake ten times faster too. So baking faster is a treadmill. You run hard, you stay in place, and next month you re-earn the exact same hours from scratch. Owning the oven is not a treadmill. The oven keeps making bread whether you’re standing there or not.
Below is the same idea, just sorted. Left is the work you do and get paid for once. Right is the work that turns into something you hold.
| The bread (your labour) | The bakery (what you own) |
|---|---|
| A campaign you ran for a client | A product that runs without you |
| A deck you wrote for one meeting | An audience that’s yours, not rented from a platform |
| This month’s content, posted and gone | A tool other people pay to use |
| Hours billed, then billed again next month | Equity, or a body of work that compounds |
One quick warning, because the hype crowd muddies this. “Own the bakery” does not mean “own an AI model.” Owning the model, the weights, the infrastructure, that’s for almost nobody, and the people telling you to go do it are usually selling something. The bakery I mean is the output of your own work, turned into something you keep. A product, an audience, a tool, equity, a body of work. That’s available to anyone selling hours right now, today.
I should say plainly why I’m even writing this. I’m not an economist. I’m a marketer who got pulled deep into AI and now follows where it’s going pretty closely, probably more closely than is healthy. At some point it felt strange to keep all of this rattling around in my head, so I’m writing it down. Building this kind of leverage into how a team works happens to be what I do for a living (here), but honestly that’s not why I wrote this. If it lands with a few people who’ve been feeling the same shift, that’s a win.
Bread feeds you today, but the bakery is the thing worth wanting
Let me concede the obvious pushback, because it’s a good one.
A lot of serious researchers don’t think we’re anywhere near AI doing everything, and they might be right. Plenty of them think the big leap is far off, and the markets aren’t pricing in a near-term jolt either. So if you want to call all of this premature, that’s fair.
But the move survives even if the doubters are right. The thesis doesn’t need a date. The squeeze starts before the machine finishes, as Korinek and Suh’s own middle case shows. And “turn your work into things you own while the work still pays” is just plain good sense, even if AI flatlines tomorrow. It’s leverage. It was leverage before any of this. You’d want it either way.
Two honest hedges before I close. Nobody knows the timing, including the people closest to it. The best measurement I’ve seen, METR’s, shows AI handling longer and longer tasks, doubling roughly every seven months, but it’s mostly coding, at coin-flip reliability, with error bars wide enough to drive a truck through (METR), so anyone selling you a date is guessing. And the second hedge: you can’t out-vote whoever ends up owning the labs. That fight happens above your pay grade and mine.
What’s left is the one thing that is yours to decide. Whether your own work, this year, becomes hours you rent out, or things you own.
The bread still has to feed you. I’m not telling you to quit and go build, that’s a privilege not everyone has, and the work pays the rent today. I’m saying point some of it sideways. Every month, route a little of what you make toward something you keep.
Because the bread feeds you today. The bakery is the thing actually worth wanting.
I might have this wrong. If you think I do, I’d genuinely like to hear where.