AI: Economy & Learning

Work should buy a life. Learning should enlarge it. AI should return time to and compound capability for human people. Agentic administrators can take care of paperwork so we can get work done. Inferential geniuses can provide the contextual sidecar to solving our biggest problems.
We can't push the river, but we can channel it.

That makes AI a constitutional problem as well as an economic one: who owns the infrastructure, who sets the terms, who watches the rest of us, and who claims the gains?

The Iron Law of Wages treated subsistence as fate. Taylorism showed how organization could multiply production, and how quickly a person could become just a cog in the machine. If you'd like a Senator who knows how they got us in the past and has personally burned billions of tokens trying to understand how these systems work so they can't get us in the future, read on.

AI, human time, and ownership

AI’s gains should not belong only to the companies building it. The public should own a stake. Workers should not carry the transition alone. Creators (like me, Jeff!) should be able to refuse, bargain, license, inspect the records, and get paid.

The choice in front of us

The numbers cannot tell us how many jobs will disappear. They do make one thing impossible to deny: AI is reaching across the economy. OpenAI estimated that approximately 80% of U.S. workers could see at least 10% of their tasks affected. That is potential exposure, not observed use. Anthropic reports product-specific Claude use across every state and the occupations it tracks. Google mapped 15 million Gemini interactions to more than 800 occupations representing just above 88% of U.S. employment. Neither company’s data says every worker uses AI. Use is not automation, and none of these figures is a job-loss count.

The danger is not today’s displacement count. It is cheap swarms of agents running on machine time—milliseconds—while Betty Smith still pays rent, buys groceries, and lives on wages in human time. If owners keep the speed and pocket the gains while workers inherit the risk, the wage bargain breaks.

What I will fight for

Bernie Sanders has supplied the right starting point: the public should own a stake in the largest AI companies and share in the gains. I would build from his American AI Sovereign Wealth Fund proposal with worker power, creator rights, computing power reserved for the public good, and American open models. Chips, power, water, data, and models cannot become private choke points built on public science and human work.

Workers should have a seat at the table before large employers use AI to eliminate jobs or radically redesign them. Employers should fund wage and benefit continuity. When displacement spikes, support should flow automatically to the communities taking the hit. Bargaining power belongs in the deal before the losses arrive, not in the cleanup afterward.

Privacy, human review, and freedom from workplace surveillance are rights, not perks. Schools should use AI to enlarge judgment—not sort people into those who direct machines and those directed by them.

I am a historian and a writer. I can imagine my own work being sucked into a model without my knowledge, stripped of my name, and sold back for pennies while competing with the work I still have to make. That makes me furious. I want a real market: creators can refuse, bargain together, license directly, see what was used, and audit what they are paid. Current law does not guarantee every creator a payment. That is the problem to solve, not a reason to surrender!

My best guide comes from the fight after Napster: EFF’s voluntary collective-licensing model. Let creators license together without surrendering direct deals—but add the records, opt-outs, audits, and anti-capture rules an AI market needs. It is a guide, not enacted AI law. We can’t have another Spotify, but for AI: a giant pool where the strongest catalogs dominate and everyone else gets pocket change.

This is where I am going: the public shares the wealth; workers have power before displacement; creators decide how their work is used and get paid; and people—not a handful of companies—set the rules.

What democracy still has to decide—before it is decided for us.

• How big a public stake should we take, how directly should people share in the dividend, and how much computing power must remain available for public use?
• When displacement hits, what triggers wage and benefit continuity—and automatic support for the communities taking the blow?
• Which uses require creators’ individual permission, which can be handled through collective licenses with a real opt-out, and what do we do about work already ingested?
• Who audits the public funds and creator institutions, who gets to challenge them, and how do we stop a new gatekeeper from replacing the old one?

Who owns the gains?

Four tests for whether AI concentrates capital and control—or returns power to people.

Work Should Buy a Life

AI can multiply returns to capital while weakening labor’s claim on the gains. Government can work like a clutch—protecting people and bringing productivity gains into sync with pay, bargaining power, and time as we shift into a different future.

Learning Should Enlarge It

Education cannot become a sorting system between people who direct machines and people directed by them. Give teachers and students tools that expand judgment, authorship, and room to experiment.

Compute Is Capital

Chips, energy, data, and models are productive infrastructure. Public-interest access—including American open models—can widen who builds, scrutinizes, and benefits.

Shared Gains, Not Smaller People

Automation should return time, not turn work and learning into surveillance. The ordinary-life test is whether people gain capability and choice—or concentrated institutions gain new leverage over them.