Terms & Conditions

In July of 1934, with San Francisco shut down by the General Strike, my grandfather walked streets lined with small shops — and recorded what their owners had taped in the windows. Hand-lettered signs, written by people whose registers had gone quiet by choice: "CLOSED TILL THE BOYS WIN." "WE'RE WITH YOU FELLOWS. STICK IT OUT."¹⁵

No accountant would have approved those signs. Every one of them was a daily loss, entered voluntarily, by merchants who had decided the ledger wasn't the only book that mattered. They were measuring something else: whether their neighbors could live on what their work paid. By that measure, closing was profitable.

This essay is about that other ledger — and about the deals that decide which one a society keeps.

The first two essays in this series established the ground. The machine's gains follow power, not merit: productivity nearly doubled since 1979 while the typical paycheck crawled,⁵ and the AI build-out is already running the same play — your bills up, their taxes down. None of that is a law of nature. All of it is a deal — written in contracts, ordinances, rate cases, and tax codes, mostly while nobody was watching.

Here is the thing about this moment: the deals for the next economy are being drafted right now. The contracts that govern AI in your workplace, the agreements that decide who pays for the grid, the codes that determine whether a data center is a neighbor or a landlord — they're all open documents for a few short years, until the concrete sets and the precedents harden. Whoever shows up to the drafting writes the terms. Whoever doesn't, accepts them.

Before the bargaining points, the measure — because you can't negotiate well if you're keeping score wrong. The country has spent fifty years measuring success in one column: profits, share price, GDP. By that scorecard, the last half-century was a triumph. By any scorecard that includes the people doing the work, it was a heist.

One picture tells it. Across a century of American data, the share of income captured by the richest tenth moves like a mirror image of one number: how many workers belong to a union. When a third of workers carried a union card, the top tenth's share fell to a third. As membership was ground down to one in ten, the top tenth climbed back to nearly half of everything.¹ Economists attribute roughly a third of the rise in inequality since the late seventies to union decline alone — and where unions held on, ordinary workers kept measurably more of the productivity they created.²

And the union difference isn't abstract. In 2025, the median union member earned $1,404 a week; the median nonunion worker, $1,174. That's $230 a week — roughly twelve thousand dollars a year — and the premium runs largest for the workers with the least, spilling over to lift nonunion wages wherever density is high.^3,2

So here is the scorecard this series proposes — the one to carry into every negotiation, council vote, and rate hearing in the AI era. Profit appears nowhere on it, because profit will take care of itself. It always has.

Now the language. Not vibes, not principles — clauses. Each one already exists somewhere in a signed agreement or a working model; none of this is utopian. The job is to make the exception the standard.

01The gains-sharing clause.

This is the clause that attacks inequality at the root, because it re-couples what 1979 decoupled: output and pay. The evidence is on labor's side — the landmark field study found AI lifted the newest, lowest-paid workers most, with novice productivity up 34 percent.⁴ When the company's own dashboards prove the gain, the raise stops being a request and becomes an invoice.

02Time is a raise.

Equality isn't only measured in dollars. If the machine does in four days what took five, the fifth day belongs on the table — and for millions of workers, an hour returned to family, health, or a second dream is worth more than its cash equivalent. A century ago the labor movement bought the weekend with the productivity of the assembly line. The AI dividend can buy the next one.

03No algorithmic bosses.

The autoworkers wrote a version of this in 2019.⁸ It matters because management-by-algorithm is how the gains get quietly clawed back — the "surveillance pay" trap from earlier in this series. A raise means nothing if the machine speeds the line until you can't keep it.

04Notice and a seat.

This exists. Microsoft's contract with CWA workers at ZeniMax requires the company to notify the union when AI could affect workers, and to negotiate in good faith on request — language whose stated goal is that the tools "benefit rather than harm workers."¹⁰ Healthcare locals have gone further, winning participation in the selection, design, and building of new systems.⁹ The earlier the seat, the cheaper the fight.

05Training on the clock.

"Learn new skills" is an insult when it means nights and weekends at your own expense while your old job evaporates. A just transition is paid, scheduled, and lands somewhere specific. Unions and labor federations have made retraining, advance notice, and priority placement the centerpiece of their AI principles for a reason: it's the difference between a transition and a layoff with homework.¹²

06You own your double.

The actors proved this winnable: striking performers won consent and compensation rights over digital replicas of themselves.¹¹ The principle scales to everyone. The machine learns from your work — your calls, your fixes, your judgment built over twenty years. That's not exhaust. That's an asset, and assets have owners.

The contract covers the shop floor. Equality is decided in bigger rooms too — and civic responsibility means treating those rooms as bargaining tables, because that's what they are.

07Neutrality as the price of partnership.

The template exists at the largest scale: Microsoft signed a neutrality framework with the AFL-CIO — sixty unions, twelve and a half million members — committing to respect organizing rights, share information on AI trends, and take worker input into the technology's development.¹³ One federation president noted the contrast plainly: every other company fights. Make the Microsoft posture the market standard, and the seventh line of the scorecard — the voice — starts moving for the first time in fifty years.

08Permission has a price.

This was the whole of essay No. 02, compressed to a sentence. The AI build-out needs what only communities can grant. Loudoun County funds a two-billion-dollar school system on untaxed-away data center revenue; Oregon makes the industry carry its own grid costs.¹⁴ Every council vote is a contract negotiation whether the council knows it or not.

09The public's cut.

Profit is a flow; equality requires a stock. A county that banks a tenth of its data center revenue, a community foundation holding a negotiated endowment, a town that leases its land instead of selling it — these are how a boom becomes a balance sheet that belongs to everyone, including the people not born yet.¹⁴

10Show up. Then keep showing up.

This is the civic responsibility part, and it is not optional. The money moves in meetings nobody attends. The proof that attendance works is recent: public anger over data-center-driven utility bills helped swing two governors' races in a single November.¹⁴ Half of American adults fear AI will cost them or their families their jobs¹² — that's not despair, that's a coalition waiting for a meeting time. In 1934 the shopkeepers stood with the longshoremen because they understood their fates were one ledger. The grocer and the gig worker, the ratepayer and the pipefitter, the parish and the local — same ledger, still.

The honest paragraph, as always. Paper promises are paper. A neutrality framework is only as strong as the organizing that tests it; a gains-sharing clause is only as real as the steward who grieves it; a community benefits agreement without enforcement is a press release with signatures. None of these terms self-execute. They are claims we have to keep staking, in rooms we have to keep filling, against counterparties with more lawyers and more patience than any of us has alone.

And that's precisely why the scorecard can't be profit. Profit is their metric, and on their metric they will always win — they own the denominator. Our metrics are the split, the bills, the hours, the ladder, the access, the reserve, the voice. On those, fifty years of evidence says one thing: we win exactly as much as we organize for, and not a dollar, an hour, or a kid's apprenticeship more.

The shopkeepers of 1934 understood something we've been talked out of: that an economy is not a scoreboard for capital — it's an arrangement among neighbors, and neighbors can change the arrangement. They posted their terms in the window in their own handwriting. "WE'RE WITH YOU FELLOWS. STICK IT OUT."¹⁵

The machine age will be negotiated, one clause at a time, whether we attend the negotiation or not. These are our terms. Post them in the window.

One more thing before I let you go, because this framework reaches past the workplace and the county line. Consider the deal you've already signed — the one nobody reads because it arrives as a button. Social media. The platforms. The feed.

There, the arrangement is stranger than anything else in this series: you are the consumer and the product at the same time. Your attention is the inventory — packaged, auctioned, and sold by the hour. Your words, your photos, your late-night searches are the raw material training the machines. The most valuable companies in history run on an input they get for free: us. Now run that deal through the scorecard. The split? You're paid nothing. The hours? Taken — engineered, in fact, to take more. The voice? A terms-of-service agreement you can't negotiate, only accept. By every line on the card, it is the most lopsided contract in the economy — and we re-sign it every morning before coffee.

It doesn't have to stay that way, and the framework in this essay is the way out. Clause 06 — consent and compensation for what's built from you — scales from the actor's digital double to every person's data. Collective bargaining scales from the union hall to the data union: millions of users negotiating as one counterparty instead of clicking alone. And policy can do for the feed what it once did for the factory floor — the eight-hour day, the weekend, the wage floor — by treating the value people create as theirs: data dividends, attention compensated like the labor it is, platforms structured and taxed to serve the people who power them. Proper policy doesn't just referee the pie. It hands out forks.

That's the next essay in this series. For now, hold the thought in a single sentence, and carry it into every room this series has named: it's time to stop giving our attention away for free.