TL;DR
Technology can lower the cost of production. It cannot carry out a just distribution on its own. The central economic question after AI is not how to produce more wealth, but who owns the systems that produce it and who decides what they are used for.
- Reality is still a long way from "most work no longer has value." The ILO's 2025 occupational exposure index finds that roughly one in four workers worldwide is in an occupation with some generative-AI exposure — but only 3.3% of global employment falls in the highest exposure band, and the authors' conclusion is transformation of jobs rather than wholesale replacement.
- The fifteen-hour week Keynes forecast in 1930 never arrived, and not because productivity fell short. Income distribution, consumption appetite, social competition and institutional arrangements all helped decide how the technical dividend turned into working hours. Abundance that is technically possible and abundance that institutions can deliver are two different things.
- When labour stops being scarce, scarcity moves rather than disappears. Content can be generated without limit; what is scarce is someone willing to read it carefully. Advice is everywhere; what is scarce is the capacity to make a judgment and carry its consequences. And competition over positional goods like status does not slacken just because everyone is richer.
- What AI may erode is not only jobs, but the infrastructure through which social relationships used to form. A young programmer stuck on a problem once had to join a community, read other people's code, show up at meetups. Today they can just ask the model. Individually that is an enormous gain; the human connections that used to form through asking for help thin out with it.
- The decline in socialising predates generative AI. The BLS time-use survey shows the share of people socialising and communicating on an average day falling from 38% in 2015 to 30% in 2025. That does not show AI caused the decline — it shows the new technology is entering a society whose social fabric was already under strain.
- The thing to watch is not whether AI understands humans, but whether humans increasingly prefer AI to each other. A system that always accommodates you may make it progressively harder to tolerate people who disagree with you. That is a risk mechanism requiring long-term study, not an established finding.
- A post-labour society is not simply one where people stop working. Interests have to be cultivated, capabilities need practice, public life needs institutions to carry it. If a society spends decades evaluating people by test scores and professional income, then on the day labour is no longer necessary for survival, a great many people will not immediately find a new identity.
On method: this piece separates empirical changes that have already happened from conditional reasoning about future society. The sociology, anthropology and philosophy are theoretical interpretation; the post-scarcity and institutional-evolution material is scenario analysis, not empirical prediction. All sources are listed at the end.
Start with a morning
Suppose one day artificial intelligence really does reach the heights we imagine for it today.
It diagnoses disease more accurately than the best physicians, discovers new knowledge faster than the most brilliant scientists, designs systems more efficiently than the most experienced engineers. Robots take on agriculture, manufacturing, transport, construction and the great majority of service work. The cost of producing goods and services keeps falling, and through some distributive arrangement most people no longer need to work to secure a basic living.
You wake up. There is no office to get to and no work that has to be done. Your AI assistant has handled everything your life requires. Whatever you want to learn, it can teach you; whatever you want to make, it can help you make; whoever you want to talk to, it can even simulate someone who understands you better than your actual friends do.
You have a freedom that ordinary people in any previous era could scarcely imagine.
And then a question arrives:
What are you going to do today?
You could travel, read, exercise, play, create, or step into a highly convincing virtual world. But when none of it takes much effort, when any desire can be satisfied immediately, when we no longer need each other to solve problems — will life still be as compelling as it is now?
More importantly: when society no longer needs the labour of most people, can humans still form friendships, earn dignity, build communities and locate a sense of meaning the way they used to?
This may be the most serious question AI has raised so far, and the one technological optimism most often skips. We are used to reading technical progress as the gradual solution of human predicaments. We rarely ask the other question: are some of those predicaments also the source of our social relationships and our sense of meaning?
1. Technological revolutions were never only changes in the tools of production
The development of civilisation is often told as a history of rising productive efficiency.
The agricultural revolution moved part of humanity from continuous migration to settlement, and cities, states, taxation and complex divisions of labour followed. The industrial revolution reorganised production around machines and factories, moving more and more labour out of households and local communities into industrial organisations at scale. The information revolution lowered the cost of distributing knowledge and coordinating at a distance, producing internet platforms, globalised services and the digital economy.
But what technological revolutions actually change was never just how much material wealth we produce. They equally change why people gather, how they cooperate, how they allocate their time, and the means by which they come to understand themselves.
In traditional agrarian society the household often served as a unit of production, consumption, education and social insurance at once. People built relationships through land, marriage, kinship networks and local custom.
Industrial society gradually rearranged that structure. Wage labour became the primary means of survival for many, and firms, vocational schools, unions, universities and urban neighbourhoods became new forms of social organisation.
A great many of a modern person's friends, partners and professional relationships are in fact products of those industrial-era institutions.
We go to school ostensibly to be educated, but we also form our peer group there. We join a company for the income, but we also build working relationships, a professional identity and a social network. We contribute to an open-source project to fix one bug, and end up knowing people in a dozen countries.
These relationships are not a by-product sitting outside economic activity. They are a significant output of economic organisation running continuously over time.
So if AI changes how economic production works, it may also change, indirectly, how social relationships form. The question is not whether people still want company. The question is whether the institutions that let strangers meet naturally, depend on one another and cooperate over long periods will still exist.
History also reminds us that technology does not by itself determine which institutions a society ends up with.
Annual working hours have fallen substantially in many industrialised countries since the nineteenth century. Our World in Data's compilation of historical working time shows some early industrialisers exceeding 3,000 hours a year in the late nineteenth century, down sharply by the early twenty-first.
In 1930 Keynes, in Economic Possibilities for our Grandchildren, imagined that rising productivity might eventually leave humans working fifteen hours a week. He foresaw not only the growth in material capacity but also the new problem it would create: what to do with the time freed from subsistence labour.
Nearly a century later, the fifteen-hour week is not how most people live.
And the reason is not simply that productivity did not rise fast enough. Income distribution, consumption appetite, social competition, institutional arrangements and the uneven productivity growth across occupations all shaped how the technical dividend was finally converted into hours and ways of living.
Which means that even if AI creates unprecedented productivity, universal leisure and freedom do not automatically follow. We need to distinguish the abundance that is technically possible from the abundance institutions can actually deliver.
2. The economic paradox: when labour is no longer scarce, what becomes the new scarcity?
A basic fact of industrial society is that most people earn income by selling labour. Labour is not only a factor of production; it is also a central mechanism for distributing resources. AI may force a fundamental adjustment of that mechanism.
But it is worth conceding first that reality is still a long way from "most work no longer has value."
The ILO's 2025 study finds that roughly a quarter of global employment sits in occupations with some degree of generative-AI exposure. That figure often gets read as "a quarter of jobs are about to disappear," but the same report supplies the more important detail: only 3.3% of global employment falls into the highest exposure band, and because most occupations contain tasks that require human input, the researchers conclude that the likely effect of GenAI is transformation of jobs rather than their replacement.
The ILO's 2026 review of the empirical evidence adds that AI's productivity gains, while real, are uneven; large-scale displacement remains limited; and the pressures that do appear fall more directly on young workers' entry into careers and on the quality of work inside organisations.
Extend the time horizon, though, and assume the capabilities of AI, robotics and automation keep improving, and the relationship between labour and income has to be rethought.
Scarcity in the economic sense does not vanish because one class of goods gets cheaper to produce.
Suppose intelligent systems drive the cost of software, educational content, digital entertainment, professional advice and some manufactured services close to zero. That makes those services abundant. It does not make all resources abundant. Land is still limited, good urban space is still limited, energy and infrastructure still bind. More to the point, each person's attention and lifetime remain finite.
- When digital content can be generated without limit, the scarce thing may not be content but someone willing to read it carefully.
- When everyone has a highly capable assistant, the scarce thing may not be advice but the capacity to make a judgment and carry its consequences.
- When anyone can generate a beautiful image, what becomes valuable may be the personal history, social recognition and cultural meaning standing behind the work.
Economics recognises a special class of goods whose value derives partly from relative position rather than absolute utility. Status, prestige, distinctive identity, elite educational access and certain exclusive experiences all have this character.
If everyone's productive capacity rises, competition over relative position does not necessarily decline. You can even get a society where everyone is materially richer than today and people still feel they have failed, for want of recognition.
This is why a universal basic income can only solve part of the problem. UBI may be an important instrument for securing a floor, but a stable income is not automatically equitable ownership of wealth, and certainly not social dignity.
If the critical infrastructure — compute, data, energy and the returns to capital — concentrates in a few hands, a society can be highly automated and highly unequal at the same time. The IMF's work on AI and income distribution notes that AI's effects on wage inequality run in several directions, but that the concentration of capital returns may widen the wealth gap further.
Which supports a basic judgment:
The central economic problem after AI is not only how to produce more wealth, but who owns the systems that produce it, who has the right to decide what they are used for, and how people share the returns.
Technology can lower the cost of production. It cannot carry out a just distribution on its own.
3. What anthropology found: people do not form relationships only to exchange useful things
If you model human society as an efficient resource-exchange system, one conclusion follows easily: once AI can do nearly every task, human interdependence loses its point.
Anthropology offers a different reading.
Marcel Mauss, in The Gift, studied giving and reciprocity across cultures. A gift looks like the transfer of an object, but it simultaneously establishes expectation, obligation, identity and relationship.
Why give gifts at all? If the point were only to meet a material need, buying the thing yourself would be more efficient. But the value of a gift usually lies not in the object but in the connection it creates and sustains.
Malinowski's study of the Kula exchange in what is now Papua New Guinea shows something similar. Participants exchange shell valuables under elaborate social rules, and the circulation of those objects is bound up with prestige, partnership and the durable ties between groups.
Viewed narrowly as material efficiency, these activities look like a great deal of unnecessary cost. To the participants, the process is a substantial part of social life itself.
Which exposes something technological thinking tends to miss: the meaning of some activities lies not in what we end up with, but in who we went through it with.
Consider learning a language.
An AI tutor can be endlessly patient, correct your pronunciation on the spot, and recalibrate the curriculum to your level. On pure knowledge-acquisition efficiency it may far outperform a classroom. But the classroom carries another kind of value: you meet people at roughly your level, you make mistakes together, you encourage each other, you go to things, you sometimes end up with relationships that last years.
Those experiences are not simply a by-product of teaching efficiency. They are part of what makes education a social practice.
Likewise, a group building software together, forming a band, organising something in their neighbourhood or climbing a mountain — the meaning does not reduce entirely to what gets delivered. If the goal is only to complete the task, AI may well do it faster. But if the goal also includes forming friendship, trust and collective memory through shared effort, the process cannot be fully substituted by the result.
None of this argues for preserving inefficient labour, still less for manufacturing hardship in order to create relationships. The real question is:
Once technology removes much of the necessity of forced cooperation, can we create more voluntary shared practice?
Social relationships used to be driven by the pressure to survive. In future, people may have to choose each other far more deliberately.
4. Sociology's warning: AI may erode the infrastructure through which relationships form
Durkheim, studying modern society, was especially concerned with the division of labour, solidarity and anomie. In a highly specialised society people need not share an occupation, a faith or a way of life, and yet they become interdependent because they need one another. The physician needs the engineer, the engineer needs the farmer, the farmer needs logistics; a complex structure ties distinct groups together through specialisation.
But if a great deal of professional service comes to be performed by AI, part of that functional interdependence may weaken.
Which produces a change worth watching: we no longer have to ask other people for help.
A young programmer stuck on something once had to join a technical community, read other people's code, turn up to meetups, and build a professional network step by step. Today they can simply ask the model.
A founder once had to join an accelerator, attend industry events and get to know experienced operators in order to obtain knowledge and resources. In future a capable agent system may absorb much of the research, development, operations and advisory work.
Measured by individual efficiency, this is an enormous advance. Measured structurally, what we lose may be not only some jobs but certain human connections that used to form through work and through asking for help.
It is worth noting that the decline in socialising did not begin with generative AI.
The US Bureau of Labor Statistics time-use survey shows that the share of people engaged in socialising and communicating on an average day fell from 38% in 2015 to 30% in 2025 (the 2025 figure is 30.4%). A 2026 study in Socius analysed more than 240,000 American time-use records from 2003 to 2024 and found total time spent with friends down by more than half.
This cannot show that AI caused the decline, because the trend was underway well before generative AI became widespread. What it does show is that changes in digital technology, in ways of living and in social organisation are arriving in a society whose social fabric was already under strain. AI may reinforce some of those trends; it may also help people rebuild connection. The direction is not settled.
What matters is what the technology is designed to maximise.
- If a social product's commercial goal is to maximise screen time, it will tend to keep users consuming content.
- If an AI system's goal is to maximise individual task completion, it may reduce the user's need to seek help from others.
- But if the goal is to support real collaboration, shared activity and durable relationships, the same technology can serve as infrastructure for social connection.
We should not file three entirely different design choices under the single heading "AI makes people lonely." The more precise question is: what kind of sociotechnical system are we building?
5. Psychology and behavioural economics: when virtual experience satisfies more than reality
One of the hardest problems ahead may not be whether AI understands humans, but whether humans increasingly choose AI over each other.
Real relationships carry a lot of uncertainty. Friends decline invitations, partners disagree, collaborators misread your intent. Interacting with people costs time, patience, compromise and emotional effort.
An AI companion can offer something quite different. It responds whenever you want, remembers your preferences indefinitely, adapts to your mood, organises the conversation around your needs, and requires no coordination of schedules. Combine that with virtual reality, personalised entertainment and embodied robotics, and you get a highly controllable emotional environment.
In behavioural-economics terms, the substitutes for real human interaction are improving.
But two outcomes have to be kept apart. Someone whose loneliness eases because of AI support is not necessarily losing their capacity for real relationships. Someone who gradually avoids real relationships because they have grown used to interacting with a model does not show that all AI companionship produces the same effect.
A 2025 study by MIT Media Lab with OpenAI found that heavy chatbot use correlated with some worse psychosocial measures, but the different experimental interaction conditions showed no significant overall differences, and the researchers stressed that causal inference should not be drawn from it.
The genuine concern is that personalised technology may gradually reduce people's willingness to bear uncertainty in real relationships.
- A system that always accommodates you may make it progressively harder to accept people who disagree with you.
- An entertainment environment that satisfies on demand may make activities requiring sustained investment look unattractive.
- An assistant that resolves problems quickly may reduce your own opportunities to explore, fail and reflect.
None of this is an inevitable consequence of AI. It is a risk mechanism that long-term research has to verify.
It bears a resemblance to Robert Nozick's experience machine. If we could enter a machine and receive every subjective experience we wanted, would we still choose to live in the real world? Nozick used it to question whether pleasure is all humans are after.
In the age of AI the thought experiment may stop being merely a classroom exercise. As virtual worlds supply ever more convincing satisfaction, we have to reconsider what exists in real life that cannot be reduced to subjective experience — perhaps acting autonomously, perhaps encountering others who are independent of your will, perhaps bearing responsibility, perhaps simply knowing that something really happened in a shared world you do not fully control.
6. The political-philosophical question: if labour is unnecessary, what is left to do?
Modern society holds a deeply rooted assumption that a person's worth is closely tied to their productive capacity. We meet a stranger and ask, "So what do you do?" An occupation determines income and also becomes part of an identity.
But that is a historically specific way of evaluating people.
Hannah Arendt, in The Human Condition, divides the active life into labour, work and action. Labour concerns the maintenance of life. Work is directed at the relatively durable human-made world: the tools, buildings, works and institutions we produce. Action takes place in the public space between people, and involves speech, participation, judgment and the joint beginning of something new.
The distinction matters especially here. If automation reduces the labour required for survival, that does not mean work and action lose their point.
A person who no longer depends on selling labour can still choose to write, to do research, to run a small shop, to make furniture, to care for family, or to take part in public affairs. Those activities may hold no competitive advantage in a traditional labour market and still carry personal and social value.
- A child's drawing, nowhere near as polished as a generated image, may be the thing a parent treasures most.
- A concert played by amateurs, technically inferior to a machine, can still leave everyone involved satisfied.
- A community garden built by a group of friends, less efficient than automated equipment, may generate distinctive emotional and public value through long cooperation.
The key distinction: productive efficiency measures how cheaply a goal is achieved; meaning concerns why the goal is worth pursuing and who takes part in achieving it. Neither substitutes for the other.
That said, we should not slide into the easy optimism that once people lose their jobs they will naturally turn to art, philosophy and public life.
Interests have to be cultivated, capabilities need practice, public activity needs institutions to carry it. If a society spends decades using test scores, professional income and economic productivity as its main evaluative system, then even on the day labour stops being necessary for survival, many people will not immediately find a new identity or purpose.
Which means a post-labour society is not simply one where people stop working. It requires building, in advance, a system of human worth that does not depend entirely on professional achievement.
7. The future of education: when knowledge is no longer scarce, why learn?
If AI can answer almost any factual question, why learn? If machines write better code, prove harder theorems and produce better prose, do schools still need to exist?
It depends on how we define education.
If education is only the transfer of knowledge, assessed by examinations that test recall and application, then AI does present an enormous challenge to the traditional system.
But education was never only transfer. A real education also helps a person develop judgment, understand views other than their own, discover the problems they actually care about, and learn how to live alongside other people.
Knowledge can be retrieved quickly; understanding does not always form in a single query. Answers can be generated automatically; deciding which questions are worth investigating is itself a value judgment.
More importantly, even someone who no longer needs learning for employment still needs enough cognitive capacity to participate in public decisions. Otherwise we get an extremely asymmetric society: a few people understand how the systems work, how resources are allocated and how rules are made; the majority receive the services the system provides and gradually lose the ability to understand or question those rules independently.
Such a society may be economically efficient and politically fragile.
So the justification for education may come less and less from competition for jobs, and more and more from autonomy and civic capability. Learning is not only about becoming labour; it is about becoming someone who can understand the world, form judgments and take part in a shared life.
The form of schooling may change too. A system built around standardised curricula, model answers and fixed career tracks may give way to more individualised learning. At the same time schools will need to carry another function: creating real peer relationships, shared experience and the transmission of knowledge across generations.
In a world where knowledge can be generated without limit, the important task of education may no longer be to make students master the answers, but to leave them able to decide what they ought to believe, and what they are willing to take responsibility for.
8. Cities and community: if work no longer requires us to gather, will we still live together?
The modern city formed in close relation to economic activity. Industrial production, specialisation, market exchange and the circulation of knowledge all benefit from the spatial concentration of people and firms. People moved to cities for jobs, education, healthcare and a wider social network.
If remote intelligent services become good enough and robots take on much of the physical work, will people still need to live at high density?
On pure economic efficiency, some of that geographic concentration may indeed lose its necessity.
But a city is not merely a collection of productive facilities. It is also a vessel for public space, cultural life, chance encounter and social diversity. A city's value comes not only from making it easier to exchange goods, but from giving people of different backgrounds the chance to share a world.
Museums, public libraries, streets, parks, stadiums, universities, cafés and neighbourhood activities cannot be fully explained by productive efficiency. What they provide is low-cost, non-compulsory contact between people.
One possible future is that cities evolve further from centres of employment and production into centres of social participation, cultural creation and shared experience.
The opposite is equally possible. As virtual space becomes convenient enough, people may leave their private dwellings less and less, and the use of public space may decline.
Both directions are technically available. The difference lies in how institutions are designed and how a society treats public life. If every experience is replaced by a private subscription, society may become more convenient while losing the places where people take part in something together. If technology lowers the cost of housing, transport, organising events and communicating across languages, people may end up with richer real-world social opportunities than they have now.
Whether cities disappear is not the important question. The important question is whether humans still want to build a public world that belongs to everyone.
9. Three possible civilisational forms after AI
The thing to avoid when discussing the future is mistaking the development of technical capability for a historical inevitability. The same AI can produce radically different societies under different ownership structures, political institutions and cultural environments.
The first possibility: a highly abundant and highly isolated consumer society.
Automated systems supply abundant goods, entertainment and services; people have ample leisure and take part in long-term cooperation less and less. Life is organised as a continuous stream of personalised experience. People are not necessarily poor, but they may lack common purpose, long commitments and public life. This society's problem is not insufficient wealth but the decay of the mechanisms through which relationships and meaning form.
The second possibility: a new hierarchical society produced by extreme concentration.
AI's productive capacity concentrates in a few firms, platforms or state institutions. Most people receive a basic standard of living but hold no ownership, governance rights or practical influence over the productive system. A small group controls models, compute, energy and critical infrastructure; the majority become consumers highly dependent on the services. There need be no mass poverty in the traditional sense for this to produce a new structural inequality — a widening distance between the capacity to consume and actual political and economic power.
The third possibility: a humanistic revival on a foundation of technical abundance.
AI lowers the cost of acquiring knowledge, making things and coordinating socially, so that more people can freely choose research, art, community building, care work, scientific exploration and public affairs. People build identity not primarily by being compelled to sell labour but through practices they choose. Education shifts from occupational filtering to lifelong learning, social security lets people take more creative risks, and public institutions ensure the technical dividend turns into freedom that is generally available.
This society does not arrive because humans spontaneously become more noble. It requires adjustments to property rights and distribution, the construction of public infrastructure, and sustained constraint on the power of technology platforms.
These three are not an exhaustive taxonomy, still less predictable historical stages. They are a reminder that technical progress opens a space of possibilities, not a social script written in advance.
10. What actually needs reinventing may be society itself
If we treat post-AI life as a design problem to be addressed in advance, at least three things seem especially important to me.
First, we need to gradually separate basic subsistence from the traditional employment relationship.
UBI need not be presumed the only answer. Public healthcare, education, housing support, social insurance, public services and the sharing of technological returns can all be part of the institutional mix. The core principle is that a person should not lose the right to live with dignity merely because their labour is no longer competitive in a market.
Second, we need to rethink the importance of social participation.
Technical systems should not aim only at reducing the cost of contact between people; they should also consider how to help people form real, durable relationships. Social infrastructure worth building may include AI-assisted community organising, shared learning networks, cross-generational projects, local cultural organisations, and public platforms where people of different backgrounds can pursue something together.
But such systems must respect the right to participate voluntarily. Social connection must not become a new form of behavioural assessment, and compulsory collective activity must never substitute for individual freedom.
Third, we need to change how we measure progress.
GDP, productivity, per-capita income and technical capability still matter, but they are not sufficient to say whether a society is actually fit for people to live in. A society can have extraordinary productive capacity and still be hard to call good, if large numbers of people lack trust, belonging, autonomy and real influence over public affairs for long stretches of their lives.
Future assessments of progress may need to look far more seriously at how much time people genuinely control, what relationships they can sustain, whether they have the chance to take part in meaningful shared activity, and whether they can actually influence the institutions that decide their lives.
These are hard to measure and hard to optimise jointly. But many of the most important things in a human life were never reducible to a single efficiency metric.
11. The final question: must humans be needed in order to have meaning?
Back to where we started. If AI can do nearly all the work, and society no longer needs our economic labour, how should we live?
The question may conceal a premise worth examining: why must a person be needed by the economic system in order to demonstrate that they have worth?
A newborn has no productive capacity, and we do not conclude that their life lacks value. An elderly person who can no longer work may generate no market income, and still possess dignity, memory, relationships and an irreplaceable personal history. We cherish friends, family and partners, and generally not because they supply us with the most efficient service.
Human worth was therefore never identical with economic worth. It is only that industrial society organised social life around occupation, income and productivity for so long that we grew used to confirming our social position through labour.
AI may force us back to an older philosophical question: if survival no longer consumes most of a life, what kind of person do we actually want to be?
This may be the first time humanity has had the chance, at scale, to move that question from philosophical speculation into actual institutional choice.
But freedom was never only the absence of constraint. Real freedom also involves choosing goals worth pursuing and bearing the responsibility those choices create.
Someone who does not have to labour to survive can still choose to study a difficult discipline, climb a mountain, work on an unsolved scientific problem, or spend decades building a community. None of that automatically loses its meaning because AI might do it faster — because for people, the value of some activities comes from taking part.
We may no longer need hardship to prove our economic worth, and we can still grow through challenges we take on voluntarily. We may no longer need market exchange to obtain every service, and we can still build relationships through mutual care, shared creation and sustained commitment. We may no longer need to organise life into the fixed track of education, employment and retirement, and we will still need some narrative that connects our experience into a life we can understand.
In the end, a post-AI society worth hoping for should not be merely a world where everyone enjoys an unlimited holiday. It should be a world where people have more freedom and can still find common purpose.
The greatest achievement of technological civilisation may not be getting machines to do everything for us. It may be that humans, for the first time, have fuller conditions in which to decide which things are worth doing themselves.
We invented technology to reduce the difficulty of survival. In future we may need to reinvent society, so that freedom does not turn into isolation, abundance does not turn into emptiness, and the unlimited possibility of an individual can still converge into a shared life.
Civilisation should not take productive efficiency as its only endpoint. Humans are not merely a species that produces wealth; we are a species that creates meaning, forms relationships, builds institutions and keeps asking after the value of its own existence.
After AI, the most important human task may finally stop being to prove we are more useful than machines — and start being to learn how actually to live, once we no longer have to prove our worth through usefulness.
On how individual capabilities get reordered once execution gets cheap, I worked through a narrower slice of this in When AI makes execution cheap, what kind of people do we actually need? — that one is about workplaces; this one is about societies.
Sources and further reading
- John Maynard Keynes (1930), Economic Possibilities for our Grandchildren
- Our World in Data (2020), Are We Working More Than Ever?
- International Labour Organization (2025), Generative AI and Jobs: A Refined Global Index of Occupational Exposure (ILO Working Paper 140)
- International Labour Organization (2026), The Impact of GenAI on Jobs, Productivity and Work Organization: A Review of the Empirical Evidence
- IMF (2025), AI Adoption and Inequality
- Yunxiang Yan (2020/2023), Gifts, Open Encyclopedia of Anthropology (a review of the literature on Mauss and the Kula)
- Neal Caren (2026), The End of Friday Nights with Friends, Socius
- U.S. Bureau of Labor Statistics (2026), American Time Use Survey — 2025 Results
- MIT Media Lab / OpenAI (2025), How AI and Human Behaviors Shape Psychosocial Effects of Extended Chatbot Use
- Stanford Encyclopedia of Philosophy, Hannah Arendt