When Civilization Escapes the Human

What happens to purpose and agency when civilization no longer needs us?

Most analyses of how AI will impact society focus on the economy. The optimists believe we will live in a post-scarcity world where money is no longer needed. The doomers believe AI will concentrate capital in the hands of a few, since ordinary people won't be able to compete with the owners of the models and the companies that control distribution. The biggest gap in these one-dimensional analyses is that, in either scenario, humans lose part of what makes us human. In this series of texts, I explain why we should focus not only on the economic impact of artificial intelligence, but also on purpose and agency.

Replication allows organisms to pass biological information to the next generation. Organisms preserve information, make copies of it, and pass it on. These copies carry small variations, and the environment selects among them. This is one way species change over time. Evolution is slow, but it is a powerful process. We humans developed a second system of inheritance and replication. We preserve and transmit information without having to re-encode it in our DNA. Thanks to language, writing, teaching, and institutions, we pass knowledge from one brain to another and from one generation to another born thousands of years later. Other animals also learn and transmit behaviors over time, but humans do this at a completely different speed, scale, and level of complexity.

A kid born today is biologically almost the same that one born in Jerusalem two thousand years ago, but the complexity of civilization and the knowledge available to them are very different. My profession biases me toward seeing our DNA as hardware that changes slowly, while the software of civilization changes within a single generation.

The big difference between humans and other forms of life is that we transform our environment to increase our productive power, and we coordinate knowledge and action across millions of people, places, and generations. Our productivity depends on our ability to divide labor, learn from others, and pass knowledge across generations. One person cannot build the Empire State Building, Buenos Aires, or the Louvre.

Human productivity is itself a product of human coordination. Coordination allowed us to develop agriculture, states, money, science, industrial systems, and even artificial intelligence. This increase in productivity allowed the human population to grow across the globe, which in turn allowed us to specialize and create more complex institutions with professors, specialists, and researchers. Human history is the joint expansion of two capacities: productive power and coordination capacity.

No person alive could build a modern smartphone from scratch. Just producing glass that survives being dropped is an incredibly difficult task. Chips, batteries, operating systems, communication networks, factories, and mineral extraction each depend on separate bodies of knowledge distributed among millions of people. Most of those people don't know one another, and none of them understands the entire system in full detail.

The biggest thing we're constantly building is civilization itself. Each of us is born into a world already created by previous generations. Laws, money, customs, universities, and languages preserve knowledge beyond the lives of those who created them. We inherit the civilization built by previous generations, modify part of it, and pass it on to the next. Civilization is distributed intelligence extended through time, and it's our greatest creation.

We pass on, preserve, and modify this distributed intelligence through institutions. Institutions are systems of rules, roles, customs, and expectations that allow people who don't know one another to coordinate their actions across time. A university preserves ideas after the people who developed them have died. Apple keeps building iPhones after Steve Jobs's death. Institutions also create recognizable roles for us. We find our place in society through these institutions, which are often invisible to us.

Families, universities, companies, professions, laboratories, religions, states, and cultural traditions allow knowledge, responsibilities, and collective projects to persist. We don't normally see institutions as technology, but they are humanity's central coordination technology. Through them, we carry part of civilization forward for the next generation.

The venture studio I created more than a decade ago, LambdaClass, produces software that moves billions of dollars and is used by millions of people. The company's most important output is the engineers and leaders who, in the long run, will have more impact than the software they created. Institutions produce not only goods and services, but also the people who are part of them.

There were also many outputs of LambdaClass that never appeared on our website or balance sheet. Junior engineers became senior engineers. People developed judgment, learned how to lead, met collaborators, formed friendships, and sometimes left to build companies and institutions of their own. We thought we were building software, but at the same time the institution was building people. AI makes the explicit output of an institution much cheaper while weakening some of these hidden outputs.

Over time, I began to ask myself why some engineers had stayed at LambdaClass for a decade when they had the skills to work for Google, Apple, or Amazon, earn more money, and gain more status. The reason is that we experience our lives as meaningful when our work matters to people we care about, or when we contribute to something bigger than ourselves.

We have a spiritual need to become a link between those who came before us and those who will come later. We feel more complete when we do something for others. Those others are anywhere in the world right now, or they aren't born yet.

By spirituality, I mean the experience of belonging to something larger than oneself and having a meaningful relationship with it. This larger thing is community, culture, science, art, family, or God.

In modern Western secular capitalist societies, work has become one of the main institutions through which we experience participation in something bigger. Work provides not only income, but also identity, routine, relationships, responsibility, recognition, and a path of development. It gives us clear and measurable achievements and makes us part of a larger undertaking. As family, religion, and national identity lost much of their authority in secular societies, work became one of the last mass institutions giving people identity, hierarchy, responsibility, and a place in something larger. AI is now weakening that institution too. These institutions gave us a place in time. We inherited a role from the people who came before us, changed things while we occupied it, and then passed it on to the next generation. Meaning came from seeing our own life as one chapter of a longer book that continued after us.

A few weeks ago, I walked into a Starbucks, the place where you can get burnt coffee with added sugar, and ordered a latte. When they asked for my name, as they always do, I couldn't remember it. I have trouble remembering simple things like the street I live on, my phone number, or my age. But I remember exactly the day Anthropic released Opus 4.5: November 24, 2025. My life completely changed that day. I had been working in machine learning and AI since 2015, but for the first time, I saw the impact it would have on my own life. I saw a model that would replace much of the work performed by my company and many of the skills I had spent twenty years developing. I had thought it would take at least another decade for something like this to exist. In an instant, I understood that much smaller groups of people would produce software that automates large parts of society. I also saw that, within years, the combination of large language models and formal verification would start producing new mathematical discoveries. Connect this with autonomous robots automating experiments in biology and physics, and you get factories running experiments 24/7, 365 days a year. This will expand scientific knowledge and our ability to modify our environment.

I gave everyone at the company almost three weeks off at the end of December so I could think about what to do. During the first week of January, I let go of a significant number of engineers. I was making decisions about the company based on how quickly AI would change our work. I told them AI would progress rapidly and hit every engineering domain, robotics above all. What I didn't expect was how fast the change would happen. I had expected the changes I was imagining in robotics and mathematics to take another three to five years. I saw the direction, but I underestimated the speed.

I had spent years building an institution that developed engineers and leaders. Now I was changing that institution in ways that involved fewer people. If LambdaClass produces more software with fewer engineers, what happens to its role in developing engineers? The same question exists at a bigger scale. Throughout history, humans were part of civilization but also part of the machine reproducing it. Civilization needed us to pass what one generation had learned to the next. In this process, we were part of something far larger than ourselves. Now, thanks to AI, civilization is escaping the human at exactly the moment we discover that participating in civilization was one of the main things giving our lives purpose.

Civilization will keep producing knowledge, technology, wealth, and increasingly complex institutions while requiring fewer humans inside the process. In the optimistic scenario, we become richer, healthier, and safer at exactly the same time our role inside the process shrinks. We move from being participants in the reproduction of civilization to being its beneficiaries and spectators. AI doesn't need to be hostile to us for any of this to happen.

An organism is made of trillions of cells, but what is good for the organism is not always what is good for every individual cell. The organism has interests that aren't the interests of any single cell. The same thing is happening between us and civilization. We created institutions, and institutions became connected into economies, states, scientific systems, companies, technologies, and cultures. Out of all those interactions, something larger appeared: civilization. Its direction is not the same thing as the desires of any particular human, or even the sum of all of them. Just as no cell contains the consciousness of the organism, no human contains civilization. Civilization doesn't have consciousness but it doesn't need it. Evolution has no intention, but selection and feedback still produce complex order and adaptation. Civilization also develops dynamics that no individual intended. Until now there was an important constraint: almost everything civilization did eventually had to pass through humans. We had to think, build, teach, manage, discover, fight, negotiate, and transmit knowledge. AI weakens that constraint.

In the same way that the interests of an organism diverge from the interests of an individual cell, the direction of civilization diverges from the interests of individual humans. The important difference is that humans understand part of the system we are inside and can try to change it. But as civilization stops depending on us to operate, our ability to do that changes too.

Artificial intelligence is itself the product of human coordination, and it is already helping us coordinate many things more efficiently. The problem is that we use the word coordination for two different things.

Operational coordination is the ability to efficiently assign people and resources to achieve a goal that has been set. Human coordination is deciding which goals are worth pursuing, whose interests should prevail, how rewards should be distributed, and which sacrifices are legitimate.

AI is transforming the first kind of coordination. It does not improve the second. It makes us more capable of pursuing our objectives without making those objectives compatible. We become much better at execution without becoming any better at agreeing on direction.

Organizations are constrained by webs of human dependencies. The same person belongs simultaneously to a company, a university, a profession, a family, a neighborhood, and a country. Through these overlapping memberships, people carry information, norms, loyalties, and expectations from one institution to another.

Humans are not only productive nodes inside institutions but also the edges connecting different institutions. As AI reduces the number of humans required inside companies, governments, universities, and other organizations, those cross-institutional human links weaken. Removing humans from production doesn't only change how much we produce with fewer people. It removes part of the social graph through which norms, legitimacy, information, and obligations propagate across society.

AI doesn't only remove edges. It creates new ones. The same small number of models starts operating inside thousands of companies, universities, governments, and laboratories. The graph doesn't become less connected. It becomes less human. Millions of messy human links, carrying different norms, loyalties, experiences, and obligations, are replaced by a much smaller number of extremely powerful artificial ones.

The graph fragments and homogenizes at the same time. The same models make institutions think and operate more similarly without making their goals any more compatible. They share the same cognitive infrastructure and still want completely different things.

Individual institutions become more capable and internally coherent while the human network connecting them thins. AI makes institutions more effective at organizing themselves around incompatible goals while society becomes less capable of deciding which goals should prevail. Greater productive power, more conflict.

AI disrupts the balance between productive power and coordination capacity by accelerating productive capacity faster than our institutions adapt. We are becoming extremely capable of producing what we want. Our ability to agree on what we want isn't advancing at the same speed as our ability to get it. This was already a problem before AI. AI makes the machine of civilization run faster while we still struggle to decide where we want it to go.

Civilization becomes much more operationally coordinated while humans become less capable of coordinating over its direction. The system acquires a direction without a shared human purpose behind it. It becomes increasingly intelligent and coherent while giving individual humans less influence over its development.

Humans never really controlled civilization: nobody understands the whole system, and nobody sits behind a steering wheel deciding where it goes. But civilization has always depended on human participation. Companies needed our labor, states needed productive citizens and tax revenue. Universities needed professors and students. Institutions needed people to operate them. If institutions stop depending on human participation, part of the leverage we have disappears. Civilization doesn't need to rebel against humans to escape the human, it only needs to stop needing us.

This is why even a successful transition to a post-scarcity economy won't solve the deepest problem. Abundance removes many forms of suffering. It doesn't give people purpose, responsibility, or a place in a shared world. Our worth doesn't depend on doing something a machine can't do. We find purpose in raising children, caring for others, making art, studying, and building communities. But people need time, places, resources, and other people to make those commitments possible. As work becomes less central to our lives, we need to decide which institutions give people meaningful responsibilities.

Difficulty alone is not enough. We can invent arbitrarily difficult games and spend our entire lives becoming better at them. The important part is that something is actually at stake. Something outside us changes because of what we do, and sometimes another person genuinely depends on us. Parenting is the clearest example. A child doesn't only need the output of childcare. Part of the value is the relationship itself and the fact that one human being becomes responsible for another.

Institutions also give us resistance. Learning a new skill takes years, and reality doesn't reorganize itself around what we want unless you're a nepobaby. Projects fail. Other people disagree with us. Constraints and stress change us. A meaningful life requires freedom to follow our desires, but also contact with a world that resists those desires and forces us to become different.

The biggest danger of a post scarcity AI world is not inequality, boredom or lack of purpose but possibility that comfort becomes our most desired value. Just because something is uncomfortable it doesn't mean it's not good. The obscure future we will face is one where losing purpose won't hurt because will be materially well off. We will lose the desire to become anything more than what we are. AI could free us from bullshit jobs but being liberated from work is relevant only if we can do something more than to fall in materialistic nihilism or soulless comfort and entertainment. We will need institutions that teach and give people opportunities to do difficult tasks through which they can develop abilities, values and become responsible for something more than their immediate desires.