I. The New Lords
Something is happening to the structure of economic power that does not yet have an adequate name, though several thinkers have been circling it. Yanis Varoufakis calls it technofeudalism. Cédric Durand and Jodi Dean have explored adjacent territory. The core argument is provocative but increasingly difficult to dismiss: we have moved beyond traditional capitalism into something that more closely resembles feudalism, only this time the fiefdoms are digital and the land is made of data, algorithms, and compute.
In classical capitalism, wealth comes from producing and selling goods in competitive markets. In the emerging order, the great technology platforms (Amazon, Google, Apple, Meta) function less like manufacturers or even service providers and more like digital landlords. They extract rent from everyone who needs access to their ecosystems. A merchant on Amazon is not competing in a free market so much as paying tribute to access Amazon’s digital fief. A developer building on Apple’s App Store surrenders thirty percent of every transaction for the privilege of reaching customers through Apple’s gate. The platforms do not merely participate in markets. They become the market itself, collecting tolls from all participants.
But the analogy extends further than Varoufakis and others have taken it. Add to the feudal lords the frontier model makers: Anthropic, OpenAI, Google DeepMind, Meta AI. The platform companies built digital marketplaces and extracted rent from participants. The AI companies are doing something qualitatively different: they are creating a substitute for human labor itself. Amazon built a marketplace where your work happens. The model makers are building systems that do your work for you, or more precisely, for whoever pays for the subscription.
The scale of displacement is not theoretical. Agentic AI systems can now produce in minutes what once took a skilled knowledge worker dozens or hundreds of hours. A functional business application, a contract review, a security audit, a financial model. The end result is similar enough that the market cannot justify the old price. The craftsmanship, the debugging intuition, the architectural knowledge built over decades of practice. None of it becomes worthless, exactly, but its exchange value collapses when an automated system can approximate eighty percent of it for pennies on the dollar. Anyone who has built a career on knowledge work recognizes the pattern, whether they work in software, law, consulting, finance, or any other domain where expertise once commanded a premium.
Software development was first in line because code is the most legible, most verifiable form of knowledge work for a large language model. But the pattern is now spreading with alarming speed across every domain where knowledge work can be decomposed into tasks that fit within a context window. The evidence is no longer speculative. It is showing up in stock tickers and earnings calls.
On February 3, 2026, Anthropic released a legal workflow plugin for its Claude Cowork platform. The tool automates contract review, compliance workflows, and legal briefings, tasks that represent the core revenue streams of established legal technology companies. The market reaction was immediate and violent. Thomson Reuters fell sixteen percent in a single session, its largest one-day decline on record. LexisNexis parent RELX dropped fourteen percent. Wolters Kluwer fell thirteen percent. LegalZoom lost nearly twenty percent of its value. In total, a Bloomberg analysis estimated the announcement triggered a two-hundred-eighty-five-billion-dollar rout across software, financial services, and asset management stocks in a single day. The S&P 500 Software & Services Index fell nearly nine percent over five trading sessions and stood more than twenty percent below its October peak.
Three weeks later, on February 20, 2026, Anthropic released Claude Code Security, a tool that scans codebases for vulnerabilities and suggests patches. This time the carnage hit cybersecurity. JFrog plunged twenty-five percent. SailPoint fell over nine percent. Okta dropped nine percent. CrowdStrike and Cloudflare each lost roughly eight percent. The Global X Cybersecurity ETF closed at its lowest level since November 2023. One industry analysis described it as a surgical strike: investors recognized that AI could reason through security problems, threatening to commoditize the high-margin subscription models that the entire cybersecurity industry is built on.
These are not isolated incidents. A Harvard Business School study examining job postings from 2019 through 2025 found that postings for occupations involving structured, repetitive tasks (the kind most susceptible to AI automation) declined thirteen percent after the launch of ChatGPT, while demand for analytical and creative roles grew twenty percent. The World Economic Forum projects ninety-two million jobs displaced globally by 2030. Anthropic’s own CEO, Dario Amodei, has stated that AI could eliminate half of all entry-level white-collar jobs in tech, finance, law, and consulting within one to five years. This is not a critic issuing a warning. This is the builder of the tool describing what he believes his tool will do.
The economic floor is dropping out from under an entire class of workers, and the value is flowing upward to the small number of companies that control the compute infrastructure (the new “land”) on which these models run.
The feudal dimension is stark. The people whose labor is being devalued have no ownership stake in the models doing the devaluing. They are, in fact, training these systems through their usage, their feedback, their expertise flowing through millions of conversations, and that value accrues entirely to the companies sitting atop the infrastructure. You cannot build your own frontier model any more than a medieval peasant could acquire a duchy. The capital requirements (billions of dollars in compute, data, and engineering talent) ensure that this remains the province of a tiny elite.
People sense this, even if they lack a framework to name it. And when people sense that the ground is shifting beneath them, they look for solid ground to stand on. The most intuitive place to look is down: toward the physical, the manual, the irreducibly human.
II. The Trades and Service Refuge and Its Limits
An oft heard response to this predicament, at least in the conversations happening among displaced workers, is the suggestion to switch to the trades or to hands-on service work. The argument is intuitive: the moat will be physical atoms and human interaction. Robots that can navigate a crawl space, diagnose a mysterious plumbing noise, and solder a joint in an awkward position are much further out than software that can write a web application. The same logic extends to personal services: a masseuse, a hair stylist, a personal trainer, a home health aide. These roles require physical presence, human touch, and interpersonal attunement that no algorithm can replicate. If you work with your hands on other people’s bodies, homes, or physical spaces, the reasoning goes, you will still have work to do.
There is a kernel of truth here. Atoms are harder than bits. But the argument has serious weaknesses that its proponents tend to ignore.
First, it demands a massive mismatch of temperament and investment. Telling a mid-career knowledge worker with deep technical expertise to apprentice as a welder or retrain as a dental hygienist is not merely a career change. It is starting over from zero in a domain that requires years of hands-on experience, licensing, and in many cases physical conditioning. The economics of that transition are brutal for anyone with a family, a mortgage, and decades of professional identity built around cognitive work.
Second, neither trades nor service work are immune to disruption; they are merely on a longer timeline. Robotics is advancing rapidly. Boston Dynamics, Figure AI, and a constellation of Chinese robotics labs are developing general-purpose physical labor robots on a timeline measured in years, not decades. Prefabricated and modular construction is already reducing on-site skilled labor needs. And even before full physical automation, AI can consume the knowledge layer of trades work: the estimating, the diagnostic reasoning, the code compliance analysis, the scheduling and project management. Margins compress even while the hands-on work remains human. Service work faces a parallel squeeze: AI handles the booking, the marketing, the inventory management, the client communication, and the billing, leaving the human to perform only the physical act itself, often for less pay as the overhead of running an independent practice approaches zero for AI-equipped competitors.
Third, if everyone follows this advice simultaneously, you simply crash the labor market in trades and services. The moat only works if it stays uncrowded. It is the classic problem of guidance that is individually rational but collectively self-defeating.
But the most devastating objection is one that the trades-and-service-as-refuge thesis ignores entirely: the demand side. In a technofeudal economy, who hires the roofer? Who books the physical therapist? If the broad middle class of knowledge workers sees its income collapse because agentic AI has devalued their labor, the entire consumer economy that sustains tradespeople and service providers collapses with them. Your neighbor is not remodeling their kitchen if their income just got replaced by an agentic AI workflow that costs their employer a fraction of a salary. Nobody is scheduling spa appointments or hiring a personal trainer when they cannot make rent. The housing market contracts. Fewer people qualify for mortgages. The whole downstream chain of HVAC technicians, electricians, stylists, home health aides, and fitness professionals sees demand evaporate, not because of automation but because their customers went broke.
In an actual feudal economy, tradespeople and servants served the aristocracy. The village blacksmith did not get rich off other peasants; he existed at the pleasure of the lord. If we are heading toward a world where a tiny class controls the cloud capital and everyone else is scrambling, then tradespeople and service providers become servants of the few, not independent operators serving a broad middle class. The rest of the population would resort to barter (I will rewire your garage if you watch my kids), which is just how peasant economies functioned. We would be reinventing medieval economic structures with modern tools lying around unused because nobody can afford to pay for anything through formal channels.
III. The Lords’ Dilemma
This line of reasoning leads to a fundamental contradiction at the heart of the technofeudal model: the lords need someone to extract value from. A feudal lord with no peasants is just a person standing in a field. If jobs collapse and the middle-class collapses with them, who pays for the subscriptions? The technology companies would have killed their own customers.
Two paths emerge from this contradiction, and they are not mutually exclusive.
The first is managed dependency. This is the more historically grounded scenario, because it is what every extractive system has actually done. Feudal lords did not starve their serfs to death. They kept them at just enough subsistence to keep working the land. The modern version looks like the technology lords providing their services for free or near-free as a form of soft control. The template already exists: Google provides search, email, maps, and photo storage at no monetary cost, and the user pays with data and attention. Scale that model up and you arrive at a world where housing, food, and digital access are provided at subsistence level, not out of generosity but because the system needs a population to function. People become the substrate: generating data, providing the thin remaining layer of human judgment, consuming content that trains the next generation of models. You are alive and fed, but you own nothing and your existence serves the platform. This is Huxley, not Orwell. You are not oppressed by force; you are pacified by convenience.
The second path is darker: population as liability. If AI and robotics reach a point where human labor is genuinely unnecessary (not just devalued but unnecessary), then large populations shift from being an asset to being a cost center. Historically, elites needed masses for labor, military power, and consumption. Remove all three needs and the calculus changes. This does not require a conspiracy. It requires only indifference. Declining birth rates are already happening across the developed world. If the social and economic structures that supported family formation continue to erode, if housing remains unaffordable, if the cost of raising children keeps climbing while income prospects collapse, population shrinks on its own without anyone ordering it. You do not need a sinister plan. You only need the people in power to not care that the conditions they have created are incompatible with reproduction. Neglect accomplishes what conspiracy does not need to.
These paths could easily operate simultaneously in different places: a managed dependent class in wealthy nations kept at subsistence on platform services, while populations in less strategically important regions decline through economic pressure and neglect.
Universal basic income, the usual proposed solution, may be less a genuine policy intention than a narrative tool to placate anxious populations. Any UBI that does materialize would likely function not as liberation but as the feudal analogy made explicit: the lords distributing subsistence to the peasants to keep the system from collapsing entirely.
Under either scenario (managed dependency or benign neglect), the question narrows to the same point: what does a human society physically need to survive, and can even that be taken away?
IV. What Remains Essential
Strip the economy back to subsistence infrastructure and ask what human labor is truly irreplaceable for, and the list is startlingly short: food production, water management, waste management, and energy production. These are the four pillars. Everything else (healthcare, education, legal systems, financial services) either gets automated and delivered as a platform service by the lords or simply disappears for most people.
But even these four pillars are vulnerable. Industrial agriculture is already heavily mechanized, and precision agriculture with AI-driven systems is advancing rapidly. Large-scale commodity farming could become nearly fully automated within a generation. Water management’s monitoring and optimization layers are natural targets for AI. Waste management, perhaps the most resistant to full automation due to the unpredictable and unstructured nature of garbage, will eventually yield to improving robotics. Energy production is already on its way to automation through solar, wind, and battery systems that require minimal human oversight.
If robotics keeps pace with AI (and the trajectory suggests it will), then even the physical labor in these essential domains gets automated, removing the last functional need for a serf class. At that point, the displaced population is not even exploited labor. It is simply mouths consuming food and water and producing waste, with no role in the system that feeds it.
Healthcare offers a particularly clear illustration. The lords do not need to actively destroy healthcare for the masses. They need only stop subsidizing it. Medical AI and robotic surgery become premium services for those who can pay. Everyone else gets triaged by an algorithm optimizing for cost rather than outcomes. The prototype already exists in insurance companies using automated systems to deny claims. Scale that capability while removing the regulatory apparatus that currently constrains it (an apparatus that requires a functioning democratic middle class to maintain), and healthcare becomes another feudal privilege.
Education follows a similar structural logic. Public education as it exists has been widely argued to have been shaped, at least in part, by the needs of industrial capitalism for literate, punctual, compliant workers. If those workers are no longer needed, the institutional incentive to educate the masses evaporates. The lords will educate their own children, probably with extraordinary AI tutors. Everyone else gets whatever they can scrape together. Not because anyone decrees it, but because funding and political will dry up. Neglect, again, rather than conspiracy.
The inventory of what can be automated or abandoned is, by this point, nearly total. But after cataloging everything the machine can take, it is worth asking whether anything remains that it cannot.
V. The Meaning Moat
There is, however, one domain where human activity retains not just economic value but irreplaceable value, and it is not the domain most people expect. It is not physical labor, which robots will eventually master. It is not knowledge work, which AI is already consuming. It is creative and interpersonal expression (art, music, games, storytelling, sport, community, and religious practice), and it retains its value not because AI cannot produce technically superior output, but because humans do not actually want technically superior output. They want human output.
An analogy to chess makes this vivid. Chess engines have been superhuman for decades. A five-dollar app on a phone plays better than any grandmaster who has ever lived. And yet human chess is thriving, arguably more popular today than at any point in history. Why? Because the point was never to see the optimal move played. The point is to watch a human mind struggle, create, fail, and transcend. The beauty of a Magnus Carlsen game lies not in the objective quality of the moves but in knowing that a human found that move under pressure, with all the limitations and emotions that entails.
Music makes the case even more starkly. A friend recently shared a collection of AI-generated music, and the experience of listening to it was instructive. The compositions were technically competent, even pleasant. But knowing they emerged from pattern optimization rather than human experience drained them of meaning. When you listen to a human musician, you are not merely processing sound waves. You are connecting to an experience: someone who has suffered, loved, doubted, and channeled all of that through disciplined craft into something they are offering to you. That is a fundamentally different transaction than an algorithm producing patterns that trigger dopamine responses. You cannot connect to AI music because there is no one on the other end to connect to. It is a mirror, not a window.
The distinction between CGI and fully AI-generated visual art clarifies the boundary. CGI is a tool wielded by human artists to realize a human vision. A CGI dragon in a film is still the product of a director’s imagination, an animator’s craft, a storyteller’s intent. Fully AI-generated visual art removes the human from the loop entirely, and something essential is lost, not in the pixels, but in the meaning.
This suggests that in a technofeudal world, the most durable moat is not physical atoms but meaning. Games, live music, handmade crafts, live theater, sports, storytelling. All of these are fundamentally about the experience of witnessing other humans do things. Humanness is the point. But in a feudal economy, this moat has limited economic value. Entertainment would be exchanged in the peasant economy for barter at most, and more likely offered freely, for the joy of it. Which is, perhaps, how it should have been all along.
But meaning requires a material foundation. Art and human connection are irreplaceable, yet they require fed, sheltered, watered, and warm people to exist. The musician still needs to eat. The question becomes whether people can secure that foundation for themselves, and it turns out, embedded in the very analysis we have been building, there are already answers.
VI. Islands of Self-Sufficiency
Woven through this analysis is a thread that points somewhere constructive. At every stage of the discussion, from food production to water management to energy to construction, there are islands of self-sufficiency available to those willing to learn the skills. Farming to feed your family. Well digging and water filtration. Composting and waste reduction. Solar panels and micro-hydro for a dwelling. These are not hypothetical capabilities. They are proven practices, many of them ancient, all of them accessible to motivated learners.
Construction alone offers a rich domain of self-sufficient practice. Earthship design, developed by Michael Reynolds in New Mexico, integrates shelter, water harvesting, thermal regulation, food production, and waste treatment into a single building made largely from salvaged materials: tires packed with earth for walls, aluminum cans and glass bottles for thermal mass, passive solar orientation for heating and greenhouse growing. Cob construction (a mixture of clay soil, sand, and straw sculpted by hand into walls) requires no forms, no manufactured materials, and minimal tools, and a community can build a cob structure together with almost no prior expertise. Adobe, rammed earth, wattle and daub, timber framing with traditional joinery, thatching with locally available materials. All of these are techniques that communities can learn and execute without industrial supply chains.
The meta-skill beneath all of these specifics is community organization. No individual or single family can efficiently manage food production, water systems, energy generation, waste processing, construction, healthcare, education, and governance simultaneously. But a community of twenty to fifty households with distributed specialization absolutely can, which is, of course, exactly how pre-industrial villages functioned for millennia. Someone becomes the person who is best at earthwork construction. Someone else manages the food forest. Someone maintains the water systems. And the musicians, the storytellers, the game makers become the social glue that transforms a subsistence operation into a community.
It is worth pausing here to acknowledge something that this entire discussion, framed as it is from within the industrialized West, risks overlooking. Village-scale self-sufficiency is not a novel idea to be invented. It is the default mode of human existence for most of history and much of the present world. Hundreds of millions of people alive today live in communities that already practice what this section catalogs: subsistence agriculture, rainwater harvesting, earthen construction, communal governance, barter economies. The condescension embedded in terms like “developing world” or “primitive” is itself a product of the techno-capitalist worldview this essay is critiquing. What the industrialized world dismisses as backwardness may turn out to be the most durable form of human organization, tested across millennia and still functioning while the digital economy threatens to eat itself. The proposal here is not to invent something new. It is to choose to value something that already exists and has been working for a very long time.
VII. The Pivot
The most important move in this entire line of reasoning is the reframe from speculative dread to practical action. Instead of debating whether the technofeudal scenario will play out, or when, or how, why not simply begin building the alternative now?
The logic is elegant in its simplicity. If the technofeudal scenario materializes, communities built on self-sufficient principles have resilience. If it does not, those same communities offer something independently desirable: lower cost of living, stronger social bonds, greater autonomy, more meaningful daily life. There is no scenario in which the effort is wasted.
Two tracks of work follow from this insight. The first is a comprehensive Village Blueprint: how twenty to fifty families organize a functioning self-sufficient village that can interface with the broader economy while maintaining genuine independence. Site selection, land ownership and legal structure, physical layout, food systems, water infrastructure, energy systems, waste management, governance, internal and external economics, education, healthcare, and cultural life, each designed at three tiers of implementation, from startup through established to fully resilient.
The second track, arguably more important, is what might be called the Onramp: a skills curriculum and community formation pathway that anyone can begin following immediately, wherever they are, with whatever resources they have. Container gardening and food preservation. Rainwater collection. Basic construction with hand tools. Consensus decision-making and conflict resolution. These skills are valuable in their own right, regardless of whether someone ever joins or forms a village. They represent fundamental human competencies that most of us have outsourced to specialists and supply chains. Reclaiming them is worthwhile under any circumstances.
The integration with the wider world is what makes this vision viable rather than fringe. These communities are not trying to disappear. They produce surplus food, crafts, art, and skilled labor that they trade externally. Members may hold outside jobs or sell products online. Children learn both self-sufficiency skills and conventional academics. People can leave if they choose. The community is open, transparent, and welcoming. This is what distinguishes the model from the survivalist compound, which tends toward insularity and defensiveness in ways that invite suspicion from the outside and rigidity from within.
But openness raises an uncomfortable question. If these communities succeed (if they demonstrate, visibly and undeniably, that the dominant system is optional), how does the dominant system respond? Every historical attempt at autonomous community has eventually had to answer this.
VIII. Defense Through Demonstration
Every self-sufficient community that has historically opted out of the dominant power structure has eventually had to deal with that power structure showing up at its door. A single off-grid community is ignorable. A movement of communities opting out is an existential threat to the technofeudal model, not militarily, but narratively. If people can see a viable alternative where life is decent without dependence on the platforms, the whole structure of compliance weakens. Every feudal and colonial system has cracked down hardest not on the desperate but on the successfully independent. The threat is not rebellion. It is demonstration that the lords are not needed.
The Amish offer perhaps the most instructive model of sustained autonomy within a dominant power structure. They have maintained genuine independence for centuries through a combination of selective engagement with the outside economy, strong internal governance, geographic distribution, and persistent non-confrontational negotiation for legal protections. They survived by being, in essence, boring and useful rather than threatening. Monastic communities throughout European feudalism maintained similar autonomy by providing something the lords valued (literacy, record-keeping, spiritual legitimacy) while controlling their own internal affairs.
The Seminole Tribe of Florida offers a different but equally instructive model. They call themselves the “Unconquered People,” and the name is earned: they are the only tribe that never signed a peace treaty with the United States. Descended from roughly three hundred people who evaded capture by retreating deep into the Everglades during the forced removals of the 1800s, the Seminole survived through geographic inaccessibility and sheer refusal to submit. Then, in 1957, they organized a constitutional government, gained federal recognition, and began converting sovereignty into economic leverage. They opened the first large-stakes bingo hall in the country, won a landmark legal case affirming tribal authority over state gambling law, and eventually acquired Hard Rock International, becoming a global entertainment and hospitality power. Where the Amish model is quiet persistence, the Seminole model is strategic defiance followed by economic self-determination. Their sovereignty rests on a legal framework that new communities cannot replicate, but the underlying logic is transferable: survive, organize, build economic independence, and make suppression more costly than coexistence.
Armed resistance against a modern state or corporate power is fantasy. The technology asymmetry is too great. What can work is making suppression politically expensive and strategically pointless: legal incorporation and conventional land ownership, radical transparency, economic interdependence with surrounding communities, geographic distribution across a network rather than concentration in a single vulnerable location, and active participation in civic life. A community that pays taxes, votes, sells goods its neighbors value, and invites journalists to visit is extraordinarily difficult to suppress without incurring costs that exceed any benefit.
The deepest defense, however, is cultural. The thing that protects a community is that its members prefer it to the alternative. Every intentional community that has failed from within (and most of them have) failed because people drifted back to the convenience of the dominant system. If what you are building feels like deprivation and sacrifice, people will leave the moment the lords offer them a comfortable subscription package. If it feels like a genuinely better life (more meaningful, more connected, more dignified), then it sustains itself. The technofeudal lords are offering convenience. A self-sufficient community is offering meaning. That is a surprisingly strong competitive position.
IX. A Thousand Villages
One village is a curiosity. Ten is a movement. A hundred is an alternative economic model. A thousand is a parallel civilization that coexists with the technofeudal system while demonstrating, by its mere existence, that the system is not the only option.
But honesty demands a recognition: that parallel civilization already exists. Across the Global South, across island nations and rural Africa and indigenous territories on every continent, hundreds of millions of people live in village-scale communities that the industrialized world has spent centuries dismissing as primitive, undeveloped, or in need of modernization. The very language we use (“developing world,” “third world”) encodes the assumption that the only valid trajectory is toward the industrial-digital model now threatening to consume its own workforce. If a thousand self-sufficient villages in the West would constitute a parallel civilization, then we must reckon with the fact that this parallel civilization is already here, already functioning, and already being overlooked by the same power structures this essay describes. The difference is not that we would be building something unprecedented. The difference is that people shaped by the dominant system would be choosing it deliberately, and that choice carries its own kind of power.
At sufficient scale, a network of self-sufficient communities reshapes the political landscape. It creates a constituency with shared interests, demonstrated capability, and a concrete alternative to offer. It provides a proof of concept that can propagate: every village that succeeds makes the next one easier to establish, as knowledge accumulates, supply chains for natural building materials develop, governance models are refined, and the cultural stigma of opting out diminishes.
The goal is not to defeat the technofeudal system or to convert the entire population. The goal is to ensure that another option exists. That those who want a life built on meaningful work, genuine community, and direct relationship with the land and water and energy that sustains them can find a path to it. That the future is not a monoculture of platform dependency but an ecosystem that includes pockets of genuine human autonomy.
And the time to begin is now, not because collapse is imminent, but because every skill acquired, every relationship built, and every step toward self-reliance makes life better and the future more secure regardless of what happens in the broader economy. The best time to plant a fruit tree was ten years ago. The second best time is today. The same is true for planting a community.