David Marete

⚡️The pause will fail, AI development will accelerate, and the political system will eventually stop trying to prevent intelligence abundance and start fighting over who owns it. Sanders is early evidence of that transition. No serious great power can voluntarily freeze frontier AI once it believes machine intelligence translates into military capability, cyber capability, scientific discovery, intelligence analysis, industrial productivity, and economic dominance. The first government that genuinely believes its rival might reach a decisive capability advantage has overwhelming incentive to continue. That makes a durable global pause structurally unstable. Even if OpenAI, Anthropic and Meta stopped tomorrow, Google, xAI, Chinese labs, open-source developers, governments, intelligence agencies, defense contractors, and eventually countless smaller actors would remain. Knowledge already discovered cannot be undiscovered. Compute can be regulated, chips can be controlled, and large training runs can be monitored, but the underlying capability frontier has escaped the laboratory. So development continues. And the deeper political rupture arrives when AI starts winning too visibly. Right now politicians can frame the issue around safety because labor displacement remains diffuse. Once companies begin reporting rising revenue and profits alongside shrinking headcount, disappearing junior positions, collapsing demand for routine professional work, and enormous productivity gains, the political ontology changes. People will understand what happened. Capital learned how to manufacture cognition. That sentence contains the entire coming conflict. Historically, capital needed labor because machines could not think. Workers possessed something owners could not manufacture independently: cognition, judgment, communication, coordination, analysis, and knowledge. AI allows capital to buy those capabilities as infrastructure. Once that scales, the bargaining relationship between capital and labor changes permanently. A company that once needed 10,000 educated workers may eventually produce more with 2,000 humans and enormous quantities of machine intelligence. Those remaining humans can become extraordinarily productive and highly paid while aggregate labor demand still falls relative to output. That produces the political contradiction that cannot survive indefinitely: Society distributes purchasing power largely through labor while production becomes progressively less dependent on labor. That is where this ends. Governments will first try familiar responses. Regulation. Worker protections. AI disclosure. Restrictions on automated decisions. Retraining programs. Hiring incentives. Severance mandates. Perhaps taxes on automation or compute. Those measures will slow pieces of the transition and fail to reverse the underlying economics. Then politics moves toward the real question. Ownership. If machine capital produces an increasing fraction of national income, access to machine capital becomes the mechanism through which people need exposure to that income. That means some combination of broader equity ownership, sovereign AI funds, retirement-system ownership of AI infrastructure, social dividends, taxation of machine-generated rents, universal capital accounts, or direct transfers becomes increasingly likely. The precise institutional form is uncertain. The direction is not. Labor income gradually loses its monopoly as the mechanism connecting ordinary people to production. And this is why the AI companies are eventually going to become political objects on a scale comparable to banks, oil companies, railroads, utilities, and monopolies combined. They will control something deeper than a product. They will control productive intelligence. If five or ten corporations own most frontier models, compute infrastructure, agent platforms, and distribution while those systems substitute for tens of millions of units of human cognitive work, governments will never treat those corporations as ordinary private businesses indefinitely. The productive core becomes too politically important. That is where Sanders is actually early. His proposed solution, stopping development, loses to geopolitics. His underlying recognition, that AI is becoming powerful enough to force political intervention, is correct. And the intervention becomes stronger precisely because the technology succeeds. The better AI gets, the more valuable machine capital becomes. The more valuable machine capital becomes, the more wealth moves toward its owners. The more labor it substitutes for, the weaker labor’s claim on national income becomes. The weaker that claim becomes, the larger the constituency demanding access to machine-generated wealth. Eventually democracy collides with concentrated ownership of intelligence. Democracy wins some portion of that collision because the alternative becomes politically impossible: an economy capable of producing extraordinary abundance while a majority of citizens lack sufficient claims on the production. That is the destination. AI does not get stopped. AI gets built faster. Human cognitive labor gets repriced downward. Capital captures the first enormous wave of gains. Inequality rises. Politics radicalizes around the distribution of those gains. Then the ownership architecture changes. The real battle of the 2030s will therefore be over who owns productive intelligence. Sanders is firing one of the first shots before most people even understand what the war will be about.

In April I met this random American guy in Berlin who'd come to Germany to learn how to bring manufacturing back to the US 🇺🇸. He spent the next 2 months visiting Mittelstand companies scattered across Germany 🇩🇪, working in factories, even living with some of the factory owners. This person was Kieran Velasquez - one of the most hard-working and impressive people I know, who also became a great friend. Last week he released all he learned as a Substack article, and it is one of the greatest things I've read this year. check it out - it made me update on what's actually worth working on"