June 29, 2026

We Built the Library. Someone Else Started Charging Admission.

There is a particular kind of exhaustion that comes from being told you are obsolete by a machine that learned everything it knows from you. It is the exhaustion of the unthanked. A developer named Suraj Sharma put words to it recently, and the words traveled fast because they named something millions of people had felt but had not yet said out loud. The internet spent twenty years building a library. We filled it with tutorials, with open source projects, with blog posts written at midnight, with answers given freely to strangers we would never meet. Then a handful of companies walked into that library, copied everything on the shelves, and built products worth billions. And now the very people who wrote the books are being told they are replaceable.

I want to sit with that feeling, because it is not just an economic complaint. It is a moral one, and it touches something old in us, something the ancients understood better than our quarterly earnings reports do.

The shape of the gift

Begin with what actually happened, because the facts matter more than the outrage.

The large language models that now write code, draft essays, and answer questions did not arrive from nowhere. They were trained on us. When OpenAI released GPT-3 in 2020, the researchers behind it described a model trained on roughly three hundred billion tokens of text, drawn from Common Crawl, from WebText, from digitized books, and from Wikipedia. That figure comes from the original paper, Brown and colleagues, “Language Models are Few-Shot Learners.” Three hundred billion tokens is a number too large to feel. So translate it. Every token is a fragment of something a human being once wrote and chose to put into the world, often for free, often for no reason other than the quiet hope that it might help someone.

The Stack Overflow answer you wrote at one in the morning, solving a bug for a person you would never meet, that became a token. The README you polished on a Sunday, the obscure forum post where you explained a workaround nobody else had documented, the Wikipedia paragraph you edited because the existing one was wrong, all of it became training data. The model did not invent its competence. It inherited it from a commons that millions of people built without compensation, often without even leaving their names.

What the anthropologists saw

Here psychology and anthropology arrive to make the discomfort legible.

In 1925 the French sociologist Marcel Mauss published a small book called The Gift, and it remains one of the most quietly radical things ever written about how human societies hold together. Mauss studied gift exchange across many cultures, and he noticed that the gift is never truly free. A gift, he argued, carries three obligations woven into it, the obligation to give, the obligation to receive, and the obligation to repay. When you accept a gift, you do not simply gain a thing, you enter a relationship, you take on a debt that the social fabric expects you to honor.

The internet, for all its chaos, ran on something close to a gift economy. People answered questions because someone had once answered theirs. They released code openly because openness had been extended to them. The whole structure depended on an unspoken third obligation, the expectation that what you take, you eventually give back, that the cycle continues, that the library stays open because everyone who borrows also lends.

What unsettles people about the current moment is that the third obligation has been quietly severed. The AI companies received the gift, the entire accumulated gift of human knowledge offered freely online, and they built extraordinary products on it. The receiving happened at planetary scale. The repaying did not. The cycle that Mauss described, the loop that turns a gift into a relationship rather than a theft, was broken at the exact point where the value was extracted. We gave. They received. And then, instead of repayment, the recipients turned to the givers and informed them they were no longer needed.

The philosophers of the commons

The ancients would not have been surprised, and this is where philosophy deepens the wound rather than soothing it.

Aristotle, writing in the Politics more than two thousand years ago, made an observation so durable that economists rediscovered it in the twentieth century and called it the tragedy of the commons. That which is common to the greatest number has the least care bestowed upon it, he wrote. Every man thinks chiefly of his own, hardly at all of the common interest. Aristotle was warning that shared resources tend to be neglected, because no single person feels responsible for them, because everyone assumes someone else will tend the garden.

The internet was a strange and beautiful counterexample to Aristotle, for a while. For about two decades, the commons was not neglected, it was lovingly maintained by millions of people who tended it precisely because it belonged to everyone. The miracle of the open internet was that it briefly defeated the tragedy Aristotle predicted. People cared for what was common, they watered the garden, they kept the library shelves full.

The tragedy returned, just in a form Aristotle did not anticipate. The commons was not destroyed by neglect, it was enclosed. Historians use that word, enclosure, to describe the period in England when land that had been shared by villagers was fenced off and made private, when the common pasture became someone’s property and the people who had grazed their animals there for generations were turned away. The open internet is being enclosed in the same manner, the shared intellectual pasture fenced and monetized, and the villagers who built it are now told the fence was always inevitable.

Why it feels like betrayal, and not just loss

There is a difference between losing something and being betrayed, and psychology has measured that difference with some precision.

Research on reciprocity, going back decades, shows that the human sense of fairness is not abstract, it is visceral and it is wired deep. Studies using the ultimatum game, where one person proposes how to split a sum of money and the other can accept or reject the offer, consistently find that people will reject unfair splits even when rejecting means they get nothing. We would rather walk away with empty hands than accept an arrangement that insults our sense of fairness. The behavior makes no sense to a purely economic model of human beings, and it makes perfect sense to anyone who has ever felt used.

The current arrangement with AI triggers exactly this circuitry. People are not only worried about their jobs, though they are right to worry. They are responding to a violation of reciprocity so large it is almost hard to perceive, the way it is hard to perceive a building that is too big to fit in your field of vision.

We are not angry because the machine is good. We are angry because the machine was built from our generosity and then handed to someone else to sell back to us.

Seneca, the Roman Stoic, understood the moral weight of receiving. In his work On Benefits, he wrote that he who receives a benefit with gratitude repays the first installment of it. Notice the structure of that sentence, because it is doing quiet philosophical work. To receive a gift gratefully is itself the beginning of repayment, the first installment on a debt that gratitude keeps alive. Ingratitude, in Seneca’s framework, was among the worst of vices, because it broke the social bond at its root, it took the benefit and denied the relationship that the benefit created.

Cicero went further. Gratitude, he wrote, is not only the greatest of virtues, but the parent of all the others. Read that as a diagnosis of our moment rather than a greeting card. If gratitude is the parent of the virtues, then a system built on ingratitude at scale is not merely unfair, it is a system that has cut itself off from the source of every other virtue. A technology that received the largest gift in the history of human knowledge and responded with no acknowledgment, no repayment, no relationship, is a technology whose moral foundation is hollow, however impressive its outputs.

The machine that eats its own soil

Here is where the story turns from grievance into warning, and where linguistics and economics meet in an uncomfortable place.

When researchers at Oxford studied what happened to Stack Overflow after ChatGPT was released, they found something that should alarm even the people building these systems. In a study published in 2024, del Rio-Chanona and colleagues found that within six months of ChatGPT’s release, activity on Stack Overflow had fallen by about twenty five percent, measured against comparable platforms in Russia and China where access to ChatGPT was restricted. The drop was not a vague vibe, it was a measurable collapse in the very behavior that produces the knowledge these models depend on.

Think about what that means as a loop. The model was trained on human questions and human answers. The model became good enough that people stopped asking and answering, because it was easier to ask the machine. And so the well that fed the machine began, quietly, to run dry. The commons that produced the training data is being drained by the thing the training data produced. This is not a metaphor, it is a documented decline in human knowledge sharing on the platforms that taught the machine to be useful.

There is a word for an organism that consumes the resource it depends on faster than the resource can regenerate. We call it unsustainable. The AI economy, as currently arranged, is eating the soil it grows in. If the people who write the tutorials stop writing them, because writing them no longer pays and no longer earns even the thin currency of recognition, then the next generation of models inherits a thinner and staler commons, increasingly trained on the output of earlier models rather than on fresh human thought. The library does not just get fenced off. It slowly stops acquiring new books.

The linguistics of “replaceable”

Pay attention to the language, because the words we use to describe this shift are doing more than describing it, they are shaping how we feel about it.

The word replaceable is a remarkable piece of rhetorical machinery. To call a person replaceable is to perform a small act of philosophical violence, it is to reduce a human being to a function, to say that what you are is exhausted by what you produce, and that what you produce can be reproduced by other means. The word treats a person the way you would treat a part in a machine, interchangeable, defined entirely by its output.

This is the same linguistic move that lets a company describe the wholesale copying of human work as training, a gentle word borrowed from the gym and the classroom, rather than extraction, which is what it more honestly is. Language launders the moral content out of the transaction. When we say a model was trained on the internet, we picture a diligent student. When we say a company extracted billions in value from uncompensated human labor, we picture something closer to the truth. The choice between those two descriptions is not neutral, and the people who profit from the arrangement have a strong interest in which words win.

The philosopher Ludwig Wittgenstein observed that the limits of our language are the limits of our world. If the only available words for what is happening are the industry’s words, training, alignment, scaling, disruption, then we will struggle to even think the thought that something was taken. We need older words, the words Mauss and Seneca used, gift and debt and gratitude and obligation, because those words carry the moral information that the technical vocabulary deliberately strips out.

What we are actually owed

I want to be careful here, because it would be easy to slide into a kind of romantic despair, and despair is useless.

The point is not that the technology should not exist, the technology is extraordinary and it will not be un-invented. The point is narrower and more solvable. A gift economy that has lost its third obligation can have it restored. Mauss did not describe an iron law of physics, he described a social arrangement, and social arrangements can be rebuilt by the same societies that broke them.

What would repayment actually look like. It would look like compensation flowing back to the people and platforms whose work trained these systems, the way licensing and royalties flow in every other creative industry that took a century to civilize itself. It would look like attribution, the simple act of acknowledging that the answer the machine gave you began as someone’s late night generosity. It would look like the companies that enclosed the commons reinvesting in keeping it alive, funding the platforms and the people who still produce genuinely new knowledge, because their own future supply depends on it. It would look, in Seneca’s terms, like receiving the benefit with gratitude, which is the first installment of repayment, rather than receiving it with the silence of ingratitude that Cicero called the death of every other virtue.

None of this is utopian. It is simply the ancient logic of the gift, applied to a new and enormous case. The villagers did not ask to own the whole world, they asked only that the people who fenced the common pasture remember whose hands had cleared it.

The part the machine cannot inherit

There is a final consolation, and it is not a small one.

The model inherited the library, but it did not inherit the impulse that built it. It can reproduce ten thousand tutorials, and it cannot feel the specific human satisfaction of solving a stranger’s problem at one in the morning for nothing in return. That satisfaction, the quiet joy of the unpaid gift, is precisely the thing the training data cannot capture, because it lives in the giver and not in the gift. The machine has the answers. It does not have the reason anyone ever bothered to write them down.

That reason is still ours. The instinct to teach, to share, to leave the campsite better than we found it, to answer the question because someone once answered ours, that instinct is older than the internet and it will outlast this arrangement of it. The companies enclosed the commons, and they cannot enclose the thing that made human beings want to build a commons in the first place.

So yes, we built the library. Yes, someone else started charging admission, and that is a genuine injustice that deserves genuine remedy, in law and in money and in plain acknowledgment. We should be honest that it stings, and honest about why. The sting is the sting of broken reciprocity, the most ancient of human wounds, named by Mauss, condemned by Seneca, and felt today by every person who ever gave their work away freely and watched it sold back to them.

Yet the deeper truth is the one the machine can never copy. The library was never really the books. It was the people who could not stop themselves from writing them. And those people, the ones who give without being asked, are not replaceable. They are the only renewable resource the whole system was ever built on.

Originally published on Substack. ← Back to all articles