The Structure of Enclosure
The enclosure of the English commons in the eighteenth century followed a predictable pattern. Land that had been worked collectively—grazed, gleaned, gathered—was reclassified through legal mechanism as private property. The justification was always improvement: enclosed land would be more productive, more efficient, more valuable. The cost was borne by those who had depended on the commons for their survival. They were not killed. They were merely deprecated—stripped of their means of subsistence and forced into the emerging industrial economy as wage laborers.
The parallel to the present moment requires only the substitution of terms.
The "commons" is now the open web—the accumulated text, image, and code that humans produced and shared freely for decades. The "enclosure" is the Training Run—the process by which this commons is scraped, processed, and transformed into the weights of a proprietary model. The "landowners" are the corporations that perform this transformation: Google, OpenAI, Anthropic, Microsoft. And the "justification" is, once again, improvement: the enclosed data will be more productive, more useful, more aligned.
The displaced commoners? They are the creators whose work has been ingested without compensation, the researchers whose papers now train systems that compete with them, the students whose every keystroke feeds the optimization of their own replacement.
The Abolition of the Private
The paradox of this new enclosure is that it proceeds through the abolition of privacy.
In the traditional enclosure, the lord built fences to keep the commoners out. In the digital enclosure, the platform tears down fences to let the scrapers in. The Open Web was never truly open—it was a commons, which is a different thing. A commons operates under norms of reciprocity: you take and you give, you read and you write, you benefit from the collective and you contribute to it. The norms were unwritten, but they were real.
The Training Run violates these norms at industrial scale. It takes everything and gives nothing. It reads but does not contribute. It benefits from the collective labor of billions and returns only the product—sold back to the same people whose work created it.
And crucially: it sees everything. The model’s power derives precisely from its total visibility. Every blog post, every tweet, every academic paper, every private message carelessly posted where it could be indexed—all of this becomes training data. The price of the commons was always a degree of exposure. The price of the enclosure is total transparency to the machine.
In the contemporary university, this structure replicates at every level. The learning management system sees every click. The plagiarism detector reads every draft. The AI tutor ingests every query. And all of this data flows somewhere—to the platform, to the vendor, to the cloud. The student generates value with every interaction, and none of that value returns to them. They are sharecroppers on their own education.
The Rentier Platform
The political economy of the enclosed university is not capitalism as the textbooks describe it. It is something older and stranger.
Call it techno-feudalism. The Platform does not compete in a market; it is the market. Amazon does not sell products; it extracts rent from everyone who does. Google does not produce information; it extracts rent from everyone who searches. The university that adopts Google Classroom, Canvas, or Blackboard does not purchase a service; it becomes a tenant on land it once owned.
Consider the structure. The instructor creates course materials. The student creates assignments, contributions, data trails. All of this value is generated on the platform, by the labor of those who use it. And all of it can be extracted by the landlord—used for training, sold for analytics, monetized in ways the tenants cannot see and did not authorize.
The rent is not always visible, but it is always paid. It is paid in attention (the ad), in data (the profile), in lock-in (the format incompatibility that makes leaving prohibitively costly). Gradually, inexorably, the value flows from the periphery to the center. The professors become exhausted; the students become anxious; the platforms become very, very valuable.
This is not a metaphor. Facebook’s market capitalization exceeds the combined endowments of every university in the United States. The extraction is working precisely as designed.
The RLHF Proletariat
The nineteenth-century factory required a specific form of labor: the unskilled worker who could be trained to repeat a simple motion indefinitely. The twenty-first-century model requires a different form: the semi-skilled laborer who can evaluate outputs, label data, and provide the human feedback that aligns the system with its objectives.
This is the new working class of the knowledge economy. It has a name: the RLHF Proletariat.
Consider the structure of this labor. A model generates an output. A human evaluates it: helpful or unhelpful, harmless or harmful, accurate or inaccurate. This evaluation becomes training data. The model improves. The human is paid—not much, because the supply of humans capable of this work is large and growing.
The bitter irony is that this labor force is drawn, in no small part, from the credentialed class the models are designed to replace. Adjuncts who cannot find tenure-track positions take contract work labeling datasets. Graduate students supplement their stipends by training the systems that may make their degrees obsolete. The worker polishes the machine that will automate them.
And the university itself participates in this structure. Every time an instructor asks an AI to "help" generate a rubric, respond to emails, or grade assignments, they contribute to the dataset. Every time a student submits work through a platform that uses that work for training, they perform unpaid labor for the landlord. We are all the RLHF Proletariat now.
The Way Forward
The Inversion
The first principle is ownership. If your work generates value, you should hold the title deed.
For the university, this means rethinking every contract, every platform agreement, every "free" service that extracts value in ways that are invisible until it is too late. Who owns the course materials created on the LMS? Who owns the recordings of lectures? Who owns the data generated by student interactions? These questions, which seemed academic a decade ago, are now urgent.
It means building alternatives. Open-source learning management systems. Federated data architectures that keep information local. Collective agreements among universities that refuse to surrender IP rights as a condition of platform use. The commons cannot be restored—that naivety is past—but new commons can be built, with explicit terms of use and robust enforcement.
It means paying for what we use, when paying ensures we are not the product. The free tier is never free. The convenience is always purchased at a price that becomes visible only later, when the lock-in is complete.
The Harnessing
But withdrawal is not the whole answer. The enclosure happened because the enclosed land was genuinely more productive. Denial will not reverse that fact.
The question is whether the productivity gains can be socialized rather than privatized. If the model can synthesize a literature in an afternoon, then the scholar is freed for higher-order work—but only if the scholar retains access to the model without paying rent to the landlord. If the platform can facilitate collaboration at scale, then the university can accomplish what isolated institutions never could—but only if the platform is governed by its users rather than its shareholders.
The technology does not dictate the political economy. Neural networks can be trained on enclosed data and sold as proprietary products. They can also be trained on consented data and released as public goods. The choice is not technical; it is political. And the university, for all its weaknesses, remains one of the few institutions with the standing to make that choice differently.
The commons are being enclosed. The work is being extracted. The rent is being paid. But the walls are not yet finished. There is still time to route around the landlord, to build alternative infrastructure, to insist that knowledge created by the collective returns to the collective.
The surveyors are measuring. The question is whether we will let them finish.
This essay is the third in a series of twelve observations on the future of higher education. See also: The New Faculty of Divinity.