Who Owns Your Future?

You have probably already trained the thing that’s coming for your job, and been paid a few dollars to do it. Right now someone is correcting an AI, and each fix is fed back into the model. Do that a few hundred thousand times, across a few hundred thousand people, and the model stops needing the corrections. It stops needing the people. Replacing them was the point of the work from the start.

I know, because I built one of these factories. I was the CTO of a data-labelling company, running the pipelines that turned human judgment into training data, per task.

You pay to build the machine in three currencies. You pay with your work, generating training data for free; with your wages, as the raise you didn’t get is redirected to AI; and with your savings, through the retirement fund that is made to buy the shares insiders sell. One flow, three taps. Few people feel all three at once, but the question in the title has a literal answer, and more and more, it isn’t you. There are shares that trade and shares that steer. You are only ever offered the first.

The work

Automation’s promise was to move us up the value chain; instead, it seated us on a new bottom rung, generating the data a model needs to do without us. You give the one input it can’t synthesize: real human judgment. And your name is on none of it. When Meta began recording its own employees’ keystrokes and screens to train AI to do their jobs, it made the trade literal, and paused only after the tool leaked their private data, never over the premise.

The wages

Since 1979, American productivity has risen about 90 percent while typical pay rose about 33 percent, gains generated broadly and captured narrowly. AI capex didn’t start that. It’s the newest, fastest turn of the same screw. A firm’s pool of money is finite, and much of it this year is going to AI, straight from compensation. When Teradata told its 5,100 employees there would be no raises, the CEO put it in writing: the company would “fund this AI investment by reallocating the budget from 2026 annual salary adjustments.” TTEC went a step further and paused its 401(k) match for US staff through the end of 2026 to pay for its AI tools, which takes from the wages and the savings in one move. And next year is already set: employers have budgeted a median 3.5 percent for 2027 raises, unchanged from this year, with half a point left over for everything else. You are financing your own replacement, on payday, with the raise you didn’t get.

The objection

Every automation wave raised the same alarm, mostly wrongly; work changed shape instead of vanishing. Concentration isn’t a flaw, it’s the mechanism. You don’t get models this good from a co-op. And you were never going to own that capital anyway: you sold your labour, you were paid, the transaction cleared, and through index funds ordinary people now hold a broader slice of the market than workers ever have. A broader slice of a narrower thing. The top 1 percent of Americans hold more stock than the bottom 90 percent combined, so the slice arrives without the say. All of it true, and none of it what changed. In every prior wave you got a way in: a new job, a retirement account that rode the market up early. What’s new is when the door opens. The great AI companies stay private longer, so the growth that once happened in public now happens privately. By the time the public is let in, the value is already divided. The index buys you late and dear, with no say.

The savings

SpaceX began trading June 12 near a $1.75 trillion valuation, which Musk framed as a way for ordinary people to finally own a piece of the future. It popped about 19 percent and traded above $200 within days, then gave it all back. When the first earnings report landed on August 4, it showed capital spending running at more than twice revenue, and the stock fell about 14 percent the next day. It has since climbed back. In late September it traded near $149, about 10 percent above the $135 the public paid at listing and roughly a third below its $225 peak.

If you bought at the listing, you are ahead, and I hope you stay ahead. The price was always the least interesting part. What you paid with was risk and silence. Your fund bought what a formula told it to, in the amount the formula chose, at a moment the formula picked, and in return you hold shares in a company whose founder keeps control through a second class of stock. You carry the downside, and you have no vote that counts. That arrangement is the same at $108 as at $149.

The formula is simple to describe. When a giant lists, the rules governing the biggest retirement funds force mechanical buying within days; SpaceX, floating only about 5 percent of its shares, was pulled in near the top, before anyone outside could read the numbers. Then the second clock started. SpaceX swapped the usual single 180-day lockup for staggered early-release windows. Starting August 6, insider shares came loose in tranches, and the tradable count grew from about 639 million shares at listing to roughly 3.7 billion. Index weights follow that count. So when the Nasdaq-100 recalculated on September 21, SpaceX’s weight more than doubled, from 1.28 to 2.82 percent, and the funds tracking it had to buy an estimated $15 billion to $22 billion more of the stock.

Read that again slowly. Each share an insider was freed to sell raised the amount your fund was required to buy. The exit and the entrance were the same event, and nobody holding those funds was asked.

The largest release is still ahead: about 1.3 billion shares unlock two trading days after the third-quarter report expected in early November. Musk’s own stake stays locked to 2027. The sequence, plainly: the rules pulled your retirement money in during June, insiders got their exits in August and September, the index made your fund buy more on the way, and the founder holds the wheel throughout. I am making no prediction about a crash, only noting who is standing where when the music slows.

And the queue is getting longer. Anthropic now aims to list in November, days before the midterms, at a reported valuation near $2 trillion. It is weighing letting existing shareholders sell into the offering, and is reportedly preparing supervoting shares for its founders, with a trust that controls the board. Its CEO has warned of wealth concentration severe enough to break society even as he pledges most of his fortune to fight it, from inside the same queue. From a proposed federal wealth fund to Senator Sanders’s bill to take half, left and right now agree the public should hold a stake. OpenAI has since offered to donate 5 percent of its equity to such a fund, a move the Financial Times reported was meant to keep the administration onside and head off political blowback, while pushing its own listing to 2027. What nobody disputes is the timing: the offer to let the public in arrives exactly as insiders line up to sell.

The answer

So who owns your future? For most people, not on the terms that matter. Asked what his fortune buys him, Musk was candid: it comes down to control of the companies. That admission is the whole argument. Shares that trade are the ones your index fund is made to buy. Shares that steer stay with the founders. It doesn’t finally matter whether AI is the value machine of the decade or an overhyped exit: either way you hold the shares that move, they hold the wheel. Good intentions don’t survive an offer with enough zeros, and the day a company goes public, the law reorders its loyalty toward the shareholders who steer.

I build one of the alternatives for a living, so discount me accordingly, and hold me to the same test as everyone else. Here is the test. Ownership has to be written into the structure, credited, paid and held in your name, so a change of heart or of control can’t quietly take it back. A promise is revocable; a structure is harder to unwind. So ask, of any company built on your work: would my share survive a large enough offer, a sale, new owners? Is it in my name, or in a policy someone can rewrite? If the honest answer is no, what you held was a story about ownership, told to the people it was built on.

Which is where both political answers miss. A public fund that buys in at the top is the same late, dear share with a flag on it. So is the version that arrives as a gift. In his August manifesto, Mark Zuckerberg points to the teachers of Richland Parish, Louisiana, who he says received a $50,000 bonus this year out of the tax revenue from the data centre going up beside them, and Meta is scaling the idea into a $1 billion fund for the towns it builds in. Take the bonus. It is real money. It is also a benefit someone else can switch off: it flows through a parish sales tax created in 1968, and next year’s depends on next year’s tax bill and next year’s decision.

The fix is to build the machine so what you make is credited and held in your name from day one, as structure instead of redistribution, decided while the rules are still being written. No one will hand you a share for asking. You can still stop treating the free thing as free, and put one question to every tool built on your work: who owns what I make here? You’re helping build the most valuable thing of this decade. Whether you own any of it is not yet settled, and for a little while longer, not settled without you.