Unscarcity
Sign in for free: Preamble (PDF, ebook & audiobook) + Forum access + Direct purchases Sign In

Unscarcity Research

Human Reserved: Can You Fence Off Work From the Machines?

Bill Gates wants to wall off caregiving and jury duty from AI, funded by a robot tax. Pressed by Axios, he could only protect 40% of jobs.

12 min read 2770 words Updated August 2026 /a/human-reserved-occupations

Note: This is a research note supplementing the book Unscarcity, now available for purchase. These notes expand on concepts from the main text. Start here or get the book.

Human Reserved: Can You Fence Off Work From the Machines?

Bill Gates’s father died of Alzheimer’s in 2020. In the final stretch, he was cared for around the clock by paid professionals who understood him even when the disease had taken his words. “He couldn’t always tell them when he was hungry,” Gates wrote in a Gates Notes essay published August 26, 2026, “but they always knew.” Then the sentence that gives this idea its name: “Something in the care they gave my dad was irreplaceably human. No robot could or should have done it.”

That’s the emotional core of a policy proposal Gates calls Human Reserved: a domain of work that society deliberately leaves to people, the way a nature reserve leaves land undeveloped. “It makes me think of nature reserves,” he writes, “places where we could put buildings and roads, but we choose not to because the loss would be too great.” Pressed by Axios on how far the idea could actually stretch, Gates gave an answer sharper than the essay itself: in the most extreme version he could design, he could protect roughly 40% of jobs — “but that’s as high as I can get.” Getting there, he admitted, was harder than he expected.

That number is the whole story. Not the 40% he saved. The 60% he couldn’t.

What Gates Actually Proposed

Strip away the news cycle and the essay makes a narrower, more specific argument than the headlines suggest. Gates isn’t proposing that jury service and childcare get legally shielded — those examples came out in the Axios interview around the essay’s launch, where he called them “clearly” suited to humans. The essay itself is narrower and, frankly, more interesting: it treats Human Reserved as a category with two different justifications that get bundled together under one label.

The first justification is economic. Gates uses a construction worker as his example: “You can’t tell a 55-year-old who has worked in construction their whole career that they need to go work at an elder care facility and expect them to find it fulfilling.” Some jobs, in this reading, get reserved not because a machine can’t do them, but because yanking the humans out too fast breaks people who have no plausible next chapter.

The second justification is about what the work signals, not what it produces. “Imagine a robot giving you the awful news that you have an incurable disease,” Gates writes. “There’s no technical reason why it couldn’t. Yet it shouldn’t.” Here the point isn’t capability, or even quality — a well-tuned model can probably deliver a diagnosis more gently and more consistently than an exhausted resident at 2 a.m. The point is that some information needs a moral witness attached to the messenger, someone who can be held, blamed, thanked, or sued. That is the same insight this site has already mapped as the liability gap: a license, or in this case simple humanness, sells accountability, not accuracy.

Gates is explicit that the boundary moves by country. “Some countries might insist on having humans take care of the elderly,” he writes. “But a country like Japan, which has a shrinking workforce and not enough young people to care for the old, may welcome a caregiving robot.” That’s not a footnote — it’s an admission that Human Reserved isn’t a fact about the work, it’s a policy choice a society makes for its own reasons, and different societies will draw the line in different places for the same job.

And the funding mechanism isn’t new. Gates has been pitching a version of a robot tax since 2017, and economists laughed at it then for the same reason they’re skeptical now: it functions as a tax on productivity dressed up as a tax on progress. The 2026 version targets AI tokens as well as robots. His argument is a simple asymmetry: “If you’re an employer and you hire someone, you pay payroll taxes on their earnings. But if you buy a robot, you can usually write it off right away as a business expense. The tax system nudges you toward replacing people with machines.” He wants that asymmetry closed, with the proceeds funding retraining and a stronger safety net for the people the reserve doesn’t cover — which, per his own arithmetic, is most people.

The Ceiling Is the Argument

Here’s what should stop you: Gates isn’t a pessimist stumbling into an uncomfortable number. He’s a technologist who has spent decades betting on capability curves, doing the friendliest possible version of the math for his own proposal, and still landing at 40% as a hard ceiling — “in a very extreme form of it.” Extreme, in this context, means Gates letting his imagination run generous: wide reservations, slow phase-ins, broad definitions of “mix” roles like education and mental health where a human stays formally in charge. Push the concept as far as it can plausibly go and it still only shelters two out of every five jobs.

And this is a man who is not hedging on the underlying trend. “I’d love to be convinced that I’m wrong about the job market, but I’m not,” he told Axios, adding that he’d stake his reputation on it being “very bad for the job market.” In the essay, he’s specific about why the usual comfort blanket doesn’t apply here: past transitions like agriculture to office work “proceeded over several generations and created new jobs where human cognition was required.” This one substitutes for the cognition itself, and it’s arriving “over the course of a decade rather than a few generations.” The Substitution Threshold — the moment the cheapest reliable provider of a task stops being a person — doesn’t cross the whole economy on one Tuesday, but Gates is describing it crossing job category after job category on a decade’s clock, which is functionally the same shock delivered in slow motion.

So run the numbers the way Gates himself would: even his best-case, most aggressive, most imaginative version of walling off work from machines saves 40 points of the labor market and cedes 60. That is not a rounding error in a transition plan. That is the transition plan admitting, in the mouth of one of its own authors, that it cannot cover most of the territory it needs to cover. A policy whose author calls its own ceiling “as high as I can get” is telling you, plainly, that it is a partial patch and not a destination.

The Jobs Are Already Reserved — For Poverty Wages

Here is the detail that should stop anyone reading the essay uncritically: the labor market has already run Gates’s experiment. We already treat elder care, home health, and childcare as work that requires a human — not by law, but by practice — and the market’s verdict on that work has been to pay it almost nothing.

The national median wage for home health and personal care aides — precisely the workers Gates describes caring for his father — is $35,800 a year, about $17.21 an hour, per BLS occupational data. Childcare workers, another category Gates named to Axios as clearly human-suited, sit lower still: roughly $27,500 a year, about $13.20 an hour. A sizable share of home care aides earn little enough that they qualify for Medicaid or food assistance while providing it. And that’s the paid tier. Below it sits an invisible workforce that never shows up on a payroll at all: American family caregivers logged 49.5 billion hours of unpaid elder and disability care in 2024, work AARP values at $1.01 trillion a year if it had been paid at the going rate — and the going rate they used, $20.41 an hour, is itself barely above what a professional aide earns for the identical task.

Jury service tells the same joke in a different key. Federal jurors are compensated $50 a day — roughly $6 an hour if you work an eight-hour trial day — rising to $60 only after ten straight days on the same case. Some states do worse: Mississippi and New Jersey pay $5 a day, Missouri and Texas pay $6, and Illinois and South Carolina don’t compensate jurors at all beyond travel. This is, by Gates’s own framing, one of the two clearest examples of work “clearly” suited to humans and nothing else. Society has already decided it’s irreplaceable. Society just declined to pay for it.

This is the joke the whole proposal is standing on without noticing it. Gates wants to fence certain work off from machines because the value of the work is the humanness itself — and he’s right that this value is real. But market wages were never pricing that value in the first place. A robot tax that funnels a little more money toward these occupations is better than nothing, and Gates deserves credit for at least aiming the mechanism at the right target. But “reserved for humans” and “paid like it matters” are two entirely separate policy fights, and the essay only wins the first one. If the plan is to keep humans doing this work at $17.21 an hour while calling it “irreplaceable,” the reserve isn’t dignifying the work. It’s just keeping the same underpaid arrangement in place and adding a fence around it so a robot can’t underbid the humans further.

Who Pays, and What That Pattern Rhymes With

Give Gates this much: funding the reserve out of a tax on the very technology it excludes is structurally the right move, even if the tax rate and mechanics are contested. It’s the same logic this site has argued for at a much larger scale: when value migrates away from a displaced input, the money doesn’t vanish, it moves to whoever holds the next scarce thing — and the fair claim on that migrated value belongs to the people the migration displaced, not just whoever happens to own the new bottleneck. Gates is applying a smaller version of the same idea the EXIT Protocol applies to billionaire wealth: convert the gains from the old order into funding for people’s transition into the new one, rather than pretending the gains and the disruption are unrelated events.

But Gates’s version stops at redistribution. He never asks the deeper question this site keeps returning to: why does surviving have to be earned by wage labor at all, reserved or otherwise? His own essay names the trap without escaping it: “In a capitalist society, employment is the way most people get the money they need to pay for the basics of life as well as being a key source of dignity and social connection.” He wants to preserve the employment half of that sentence for 40% of people. The book’s answer is to sever the survival half from employment entirely, for everyone, and then rebuild the dignity half on a currency built for exactly the work Gates is trying to protect.

What the Book Would Do With the Other 60%

This is where Human Reserved runs into the wall the book was written to name. The Foundation doesn’t try to keep 40% of people employed in protected jobs while the other 60% scramble for whatever’s left of the old economy. It removes survival from the wage question entirely — food, shelter, healthcare, energy, and compute, guaranteed, so that no one’s next meal depends on which side of Gates’s fence their occupation landed on.

And once survival stops being the thing labor buys, the caregiving Gates is trying to protect doesn’t need a market wage to be valued — it needs a better instrument than a wage. That’s Impact: a currency built specifically to register exactly the kind of contribution the market has always underpriced. The book’s own caregiver, Elara, spends forty years on the same work Gates is describing — dementia care, presence, the kind of attention a robot could technically fake but shouldn’t — and the system doesn’t pay her $17.21 an hour for it. Diverse councils rate her contribution “Transformative” year after year, she never sees it decay because she keeps showing up, and by 75 that verified record is what gets her into a scarce life-extension program that money literally cannot buy a spot in. That’s what it looks like to actually value irreplaceably human work, instead of just forbidding a robot from doing it at the same depressed wage.

Civic Service closes the other gap Gates leaves open: elder care as one of its standard placements, not because a law reserved the job, but because completing it is how you earn the right to vote and govern in the first place. Nobody has to be told a robot can’t take that job. Nobody wants a robot to take it, because doing the work is the point, not a cost to be minimized.

The Honest Verdict

Give the essay full credit for the parts it gets right. Gates correctly identifies that some value is inseparable from the fact that a human is the one doing it — that’s a real category, not sentiment. He correctly identifies that letting the tax code keep subsidizing the machine over the person is backwards. And he’s honest, more honest than most people making this kind of proposal in public, about exactly where his own idea runs out: 40%, “as high as I can get,” admitted on the record to a reporter who pushed him on it.

That honesty is the most useful thing in the whole essay. A reserve is a fence, and a fence, by definition, only protects what’s inside it. Gates spent his credibility building the strongest version of the fence he could design and told us himself how much land it covers. The other 60% isn’t a footnote to be patched later. It’s the actual size of the problem — the one the Labor Cliff describes and the one Human Reserved was never built to solve.

The next essay Gates has promised, on rebalancing labor and capital taxes, is worth reading when it lands. But the shape of the answer is already visible from the outside: you don’t solve a labor cliff by fencing off a shrinking meadow at the top of it. You build a floor at the bottom.

  • The Substitution Threshold — the mechanism Human Reserved is trying to hold back, one job at a time
  • The Liability Gap — why accountability, not accuracy, is what a human (or a license) is really selling
  • Impact — the currency built to value exactly the caregiving work a wage never priced correctly, including Elara’s forty years of dementia care
  • Civic Service — how elder care becomes the path to citizenship instead of a fenced-off job category
  • The Foundation — why survival shouldn’t depend on which side of the fence your occupation lands on
  • The EXIT Protocol — the larger-scale version of funding a transition out of the wealth the transition itself displaces
  • Value Migration — where the money goes once an input, in this case labor, gets cheap
  • The 2025-2030 Labor Cliff — the scale of the problem the other 60% represents

References

Share this article:

Share Share Share