The old social contract is changing fast.

The old social contract, however unequal, had a structural logic that protected workers even when elites didn’t care about them because elites needed workers. They needed soldiers to fight their wars, laborers to make their products, and consumers to buy them. That mutual dependence was ugly but it was real, and it’s the basis of every gain ordinary people have made in the modern era. Unions worked because strikes hurt. Democracy expanded because mass armies required mass buy-in — you can’t draft a million men into trenches without conceding them the vote eventually. The welfare state emerged partly because healthy, educated workers were more productive and because unhealthy, angry workers were a revolutionary threat. Consumer economies required consumers with money to spend, which meant wages had to be high enough for mass purchasing power. This isn’t Marxist romance, it’s just the material basis of the 20th-century settlement.

So, if automation breaks all three legs of the stool simultaneously — drones fight the wars, robots make the products, algorithms target what’s left of consumer demand — the elites who own the automated systems have no structural need for the rest of the population. That’s not a conspiracy theory. It’s a straightforward observation about changing incentive structures, and it’s what separates this technological transition from previous ones. Every previous wave of automation displaced specific workers and created new jobs elsewhere; the aggregate need for humans kept rising. If AI and robotics approach their more ambitious claims, the aggregate need for humans in production could actually fall for the first time in modern history. And if the wealthy don’t need the population to produce, to fight, or to consume — what’s the relationship? What claim does the population have on anything?

The historical answer to “what happens to populations elites don’t need” is not reassuring. Enclosure in England, the Highland Clearances, the displacement of indigenous peoples across the Americas and Australia, the fate of subsistence farmers when export agriculture arrived — when populations become economically redundant to the people with power, they tend to be managed rather than served. Managed through welfare that keeps them quiet, through entertainment that keeps them distracted, through surveillance that keeps them controllable, through pharmaceuticals that keep them numbed, through culture wars that keep them fighting each other rather than the arrangement itself. Some version of this is visible in parts of the American landscape already — the opioid belt, the rural counties where life expectancy is falling, the young men radicalizing online because they can’t see a future. The managed-redundancy future doesn’t require a dystopian decision; it just requires the current trajectory to continue.

The new class structure coming has actually been forming for a while and scholars have been trying to name it. The tech theorist Yanis Varoufakis calls it “techno-feudalism” — arguing we’ve already left capitalism behind for a system where a few platform owners extract rent from everyone else’s economic activity, more like medieval lords than modern capitalists. Shoshana Zuboff calls it “surveillance capitalism.” Others talk about “rentier capitalism” or the “overclass.” The terminology is contested but the phenomenon isn’t: in the US, the top 0.1% now owns more wealth than the bottom 85%, billionaire wealth has roughly doubled since 2020, and the marginal political power of the median voter has been measured as approximately zero (there’s a famous Gilens and Page study showing that policy outcomes correlate strongly with elite preferences and essentially not at all with majority preferences once you control for elite views). The forms of democracy remain, the substance has been hollowed. This has been documented by mainstream political scientists and the findings have been mostly ignored because the implications are too uncomfortable.

What about a war with only machines — we’re already there in important ways, and nobody has thought through the implications. Ukraine has become the first drone-dominated battlefield in history, with small cheap drones doing most of the killing. Israel has used AI systems (Lavender, Gospel) to select targets in Gaza at industrial speed. The US has autonomous systems deployed across multiple theaters. When wars can be fought without your citizens in uniform, two things happen: the domestic political cost of war collapses, which means more wars, and the moral friction of killing collapses, which means more deaths. Vietnam ended partly because draftees came home in body bags and their mothers organized. A fully drone-fought war produces no body bags on the home side, no protests, no political cost, and therefore no natural endpoint. That’s a structural change in how democracies relate to violence, and the institutions that historically constrained military adventurism won’t work the same way anymore. The other side’s civilians still die, of course. They just don’t have a political constituency in the country doing the killing.

Modern societies have traded resilience for efficiency in almost every domain — just-in-time supply chains, centralized data centers, concentrated power grids, specialized agriculture, digital payments replacing cash, GPS replacing maps, search engines replacing memorized knowledge. Each trade looked rational in isolation. The aggregate effect is that a handful of targeted failures could cascade into civilizational collapse of a kind that would have been impossible a hundred years ago. A 1920s American town could feed itself for weeks from its own hinterland, repair its own machinery, navigate without satellites, and communicate without the internet. A 2026 American suburb has maybe 72 hours of food in local supermarkets, depends entirely on trucked-in everything, and would be functionally helpless if the power grid went down for more than a few days. We’ve built a civilization that is incredibly productive under normal conditions and incredibly brittle under abnormal ones. That’s a bad bet in a period when abnormal conditions are becoming more likely — cyberattacks, solar flares, grid failures, supply chain disruptions, and yes, conflicts that target infrastructure rather than soldiers. Russia has been attacking Ukraine’s electrical grid for three years specifically because it’s a more efficient target than the Ukrainian army. That lesson has been noted in every capital.

Local food systems, distributed energy generation, cash-capable economies, paper records as backup to digital, communities that know each other well enough to function without institutional mediation, practical skills held by at least some people in any given place. So the infrastructure of resilience, and the countries and regions that have maintained some of them are going to fare better in the disruptions ahead.

Human nature includes greed, absolutely, but it also includes generosity, solidarity, fairness instincts, and ferocious reactions against perceived injustice. The mix varies by person and by culture and by moment, but the mix is real. The reason billionaire wealth has exploded isn’t that human nature changed in 1980 — it’s that a specific policy regime was adopted that allowed it. Marginal tax rates on the highest incomes were 70%+ in the US for most of the mid-20th century and are now effectively lower than what most middle-class workers pay when you include all taxes. Antitrust enforcement has been essentially dormant since Reagan. Campaign finance has been deregulated. Labor law has been gutted through non-enforcement. Capital gains are taxed at lower rates than wages. Each of these was a policy choice, reversible by other policy choices. The human nature of greed exists in every era, but its scope of action is set by institutions. The Gilded Age ended not because humans got less greedy but because the political system eventually acted against the concentrations. It can happen again. The question is what has to break first.