Picture this: somewhere in a Senate office building, at an hour when nothing good has ever happened, a congressman who still prints out his emails is asking the CEO of an AI company whether the robots are going to be “friendly.”
The CEO — a man who has personally helped erase entire job categories from the economy before his second coffee — smiles and says something about “responsible innovation.” Everyone nods. C-SPAN keeps rolling.
Nobody asks the only question that matters. So I will.
If artificial intelligence is about to generate trillions of dollars in new wealth while making tens of millions of people economically optional — who gets the check?
Not “will AI take your job.” That question retired years ago and is now collecting a pension it didn’t earn.
The real question — the one that should be leading every cable segment and every stump speech — is this: when the productivity miracle lands, who’s at the table, and who’s just bussing it?
01 / THE NUMBERS NOBODY WANTS TO DISCUSS
The Math Doesn’t Lie. It Just Doesn’t Care About You.
Goldman Sachs — an institution that has never once been wrong about money and never once been right about people — estimates that roughly 300 million full-time-equivalent jobs worldwide are exposed to AI automation. About a quarter of U.S. work tasks could eventually be handled by a machine that doesn’t ask for dental.
Now, before the spreadsheet economists start sharpening their pencils — yes, I know. “Exposed to automation” is not the same as “eliminated.” ATMs didn’t kill bank tellers. Excel didn’t kill accountants. The internet took its sweet time killing travel agents, and even then it mostly just relocated them to Yelp reviews.
Every prior wave of automation destroyed some jobs and, eventually, created others nobody saw coming. Nobody in 1850 was budgeting for “social media manager.”
That’s the honest version of the optimist’s case, and it deserves to be said before I take it apart: lower costs create new demand, new demand creates new industries, new industries create new jobs. It’s happened before. It might happen again.
Here’s the catch. Every previous wave of automation replaced a category of labor — manual, then clerical, then routine. This is the first wave aimed at the thing the last forty years of “safe career advice” was built on: judgment. Coding, legal research, diagnosis, writing, design, analysis, customer service — these aren’t niche fields. They’re the fields.
Exposure doesn’t guarantee elimination. The real question is whether replacement jobs arrive fast enough, in enough volume, for the people actually displaced — and whether the people doing the displacing have any incentive to make sure that happens on anyone’s timeline but their own.
And on the macro level, Goldman projects AI could add roughly $7 trillion to global GDP — about a 7% bump — over the next decade.
For a hundred years, politicians have treated GDP like the country’s report card. Number goes up, we’re winning. But GDP has never once asked the follow-up question: up for whom?
GDP is the total on the receipt. It doesn’t tell you whose name is on the card.
02 / FOLLOW THE MONEY
You Already Know Where It’s Going
The biggest financial winners of the AI era are unlikely to be electricians, teachers, line cooks, or anyone who has ever had to refresh their bank app before payday. The winners are going to be — brace yourself for the plot twist — the companies that already own everything.
NVIDIA. Microsoft. Alphabet. Amazon. OpenAI. And yes, Anthropic — the company whose AI is currently helping me write this sentence, because I’d rather get sued for accuracy than praised for cowardice.
Goldman estimates AI infrastructure spending alone could top $1.1 trillion by 2027. That’s not investment in you. That’s investment in the thing that replaces you — built by the same handful of companies, financed by the same pension funds that quietly hold stock in the firms automating away the jobs those pensions were supposed to protect.
Nobody needs a smoky back room for this. The incentives just all point the same direction, like iron filings near a magnet — and the magnet is shaped like a stock buyback.
AI is being sold to the public as a productivity revolution for humanity. Financially, it’s shaping up to be one of the largest wealth-concentration events in modern history — and possibly, when the dust settles, simply the largest, no historical comparison required.
The shareholder always wins, because the shareholder owns the part of the company that doesn’t need health insuranc
e02.5 / FOR THE RECORD
This Isn’t an Anti-AI Argument
None of this requires AI to fail.
AI may end up curing diseases, accelerating scientific discovery, slashing costs, and creating abundance on a scale previous generations could barely imagine. That part isn’t in dispute.
The question isn’t whether AI will generate wealth. It will. The question is whether the public participates in creating it, owns part of it, or merely rents access to it — same as everything else they were sold this century.
This isn’t an argument against the technology. It’s an argument against the ownership structure sitting underneath it. Those are very different things, and only one of them gets you called a Luddite on cable news.
03 / WHO FED THE MACHINE
The Heist Nobody’s Calling a Heist
Here’s a question that gets skipped past so fast in AI coverage it’s basically become furniture: what did these models actually learn from?
Not just chips and electricity. Every major AI model was trained on the accumulated output of human civilization — books, articles, forum posts, open-source code, research papers, photographs, video, decades of labor that millions of people did for free, or for a paycheck from an employer who never imagined it would end up as training data for something else entirely.
The raw material of the AI economy isn’t silicon. It’s us.
Specifically, it’s nearly everything humanity has published, uploaded, reviewed, annotated, or argued about online since roughly 1995.
So here’s the uncomfortable question underneath the uncomfortable question: if AI’s value comes from the collective intellectual output of billions of people, why does ownership of that value belong exclusively to the handful of corporations that happened to have the compute to scrape it first?
This is not a redistribution argument. Redistribution says “you have money, give some to people who don’t.” This is an ownership argument. It says: you built this thing out of our stuff — what’s our stake in it?
It’s the difference between asking your landlord for a rent reduction and asking why he gets to own the building you built. And once you frame it that way, the conversation stops being about charity and starts being about a receipt nobody’s been handed yet.
04 / BOTH PARTIES, SAME ROOM
Republicans Say the Market Will Fix It. Democrats Say They’ll Retrain You. Both Had Their Moment — and Blew It.
The Republican answer, in its most generous form, is the steelmanned version from earlier: deregulate, let capital and labor reallocate, and new industries will emerge to absorb the displaced — the way they always have.
Here’s what that looked like in practice. In July 2025, the Trump administration released its “AI Action Plan,” branded as a worker-first AI agenda. Strip away the branding and here’s what’s actually in it:
Tax-free reimbursement for AI training
Retraining programs routed through the Department of Labor
A new “AI Workforce Research Hub” whose primary job is to study how many people are getting displaced and how fast
Notice what’s missing. No income support for the gap between “your job got automated” and “your retraining finished.” No portable benefits. No mechanism — none — for workers to have any say in how AI gets deployed in their own workplace, what gets automated first, or how the resulting savings get shared.
The plan treats AI’s economic transformation like a weather event: something to prepare for, not something workers get a vote on. It’s not anti-worker. It’s just indifferent to worker power — which in practice amounts to roughly the same thing.
The Democratic answer — “we’ll retrain everybody” — sounds compassionate until you ask: retrain them for what?
To be fair, that question has a real answer this time. Skilled trades, healthcare, energy infrastructure, robotics maintenance, and AI oversight itself are all genuinely growing fields, and pretending otherwise would be its own kind of dishonesty. Retraining isn’t useless. It’s insufficient — a parachute handed out after some people have already jumped and others haven’t been told the plane is descending.
And when Democrats had a real chance to do more than hand out parachutes — to actually decide who gets a seat at the table — they spent it in a room with the people doing the displacing.
In September 2023, Senate Majority Leader Chuck Schumer convened the first of nine “AI Insight Forums”: closed-door meetings on Capitol Hill where roughly two-thirds of the Senate sat down with Elon Musk, Mark Zuckerberg, Bill Gates, and other tech executives to discuss how AI should be regulated.
The press wasn’t allowed in. The public wasn’t allowed in. Senators couldn’t even ask questions directly — they had to submit them in writing, like kids passing notes to a substitute teacher.
Even Senator Elizabeth Warren, no friend of Big Tech, called it “just plain wrong” for tech billionaires to lobby Congress behind closed doors with zero accountability. When the resulting roadmap landed eight months later, advocacy groups across the spectrum criticized it as a document that prioritized industry competitiveness over, well, anyone who isn’t industry.
For a Second Time, Trump Muses About Americans Sharing in A.I. Wealth
The party that says “we’ll retrain you” spent its biggest AI policy moment of the decade in a private room with the people doing the automating — and didn’t bring a single worker, teacher, or truck driver through the door.
WSJ: Inside the Room Where America’s Brightest Game Out How to Avoid an AI Apocalypse
05 / THE MENU OF FIXES
Choose Your Own Dystopia
Universal Basic Income. Tax AI productivity, mail everyone a check. Prevents mass poverty, keeps consumer spending alive, fits on a bumper sticker. Doesn’t restore meaning or purpose, and depending on design, risks becoming a permanent bipartisan bribe.
Is UBI freedom, or is it hush money — a check that says “here’s some cash, now stop asking who owns the robots”?
AI Ownership Dividends. The more interesting cousin of UBI. Every citizen owns a literal slice of the AI economy — Alaska’s oil dividend, but for algorithms. Not a handout. A dividend, because you own part of the machine. This is the rare idea a Wall Street Republican and a Bernie-aligned Democrat could both vote for without either needing therapy afterward.
Massive Public Works. Infrastructure, energy, housing, environmental restoration — government as employer of last resort. Builds real things. Comes with the standard government combo meal: bureaucracy, waste, and “the bridge will be done in six years, we promise.”
The Reduced Work Week. The option everyone forgot existed. If AI doubles productivity, why are we still working forty hours? Every previous productivity leap — assembly lines, computers, the internet — was supposed to buy workers more leisure. Instead it bought them more email. That’s the entire prize for a century of innovation: a Slack notification at 9:47 p.m. during your kid’s recital.
Human-Centered Capitalism. Measure mental health, family formation, trust, life expectancy, and community engagement instead of just GDP. Because if AI makes a small number of people richer than entire countries while everyone else gets lonelier and more anxious, the spreadsheet says we’re winning and reality says otherwise — and reality has a pretty good track record of eventually being right.
06 / THE RADICAL MENU
For When You’re Done Being Polite
A National AI Sovereign Wealth Fund. Norway built a trillion-dollar public fund out of oil. Imagine the same model applied to AI: every major AI company contributes a cut of profits into a fund that pays dividends to citizens. Norway did it with fossil fuels. We could do it with math.
A Robot Tax. Even Bill Gates has floated taxing automation that displaces workers, so a robot doing a job generates roughly the same public revenue a human worker did. Critics say it could slow innovation. Maybe. But “ignore the problem and hope it resolves itself” has never won a war, a marriage, or an empire.
Corporate AI Licenses. If a company’s AI deployment displaces 10,000 workers, require it to fund transition programs for those workers — a social-impact fee for industrial-scale obsolescence. Severance, but for an entire workforce, paid by the thing that replaced them.
Guardian: Bernie Sanders' AI sovereign wealth fund plan is good. But we think this is better
The Part Where I Turn This Back on You
Here’s the uncomfortable truth underneath all of it: AI isn’t creating a new problem. It’s exposing one that was already there, quietly, the whole time.
For decades, the deal in this country was simple: work hard, and you’ll earn your share.
AI doesn’t break that deal. It just stops pretending the deal was ever about you — it was always about labor, and labor was always just a temporary, inconvenient way to extract value before something cheaper came along.
You weren’t the point. You were the workaround.
And if a handful of companies end up owning the intelligence, the chips, the data centers, and the training data scraped from all of us — what we’re building isn’t a technological revolution.
It’s feudalism with a server farm. Serfs don’t get dividends. They get terms of service.
The Industrial Revolution created workers. The AI Revolution may create owners and non-owners. If that’s where we’re headed, the defining political question of the 21st century won’t be how intelligent the machines become.
It’ll be whether the people who built the future get a stake in it.
🎤💧 - R❌AN SMITH
