The Well Belongs to No One
Earlier today I wrote a Signal and threw it away. It argued that the frontier AI labs are laying competing pipes to the same water, and that the water, not the pipes, is what makes this enterprise prone to monopoly.
A frontier AI monopoly? I understand the reaction. It sounds crazy.
This does not: the accumulated written record of humanity. Every digitized book, paper, patent, ruling, ledger, sermon, recipe, manual, map, and proof. The letters that were kept and the arguments posted at two in the morning. The source code of dead companies. The lab notebooks of people whose names we lost. Everything human beings bothered to write down, in every written language, that has survived and been made digitally available.
Not human intelligence itself, which cannot be stored. What human intelligence left behind: the vast fossil record of human thought.
Bigger than oil
Oil is fossilized life: sunlight caught by algae and plankton that died before we existed, pressed for a hundred million years, pumped up, and burned once. The twentieth century ran on it, and a fair share of that century’s wars were fought over who held the leases.
The written record is fossilized thought. It is a stranger commodity than oil in three ways that matter.
It does not deplete through use. A barrel of oil burned is gone. A theorem read is still there, and now lives in another mind.
It is non-rivalrous. My use does not prevent yours. My Euclid is your Euclid, too. Billions of us can draw from the same page without making the page shorter.
It grows. As people contribute new, verified knowledge, the record compounds. It is a well that fills as it is drawn.
Economists have a plain name for a resource like this: a public good. It can be used by many people at once without being used up, and it is difficult, though not impossible, to exclude people from it.
Those two facts matter. Markets often underinvest in public goods because no single owner can capture their full value. But when someone does manage to fence one off, they can charge for access to something the public helped create.
Oil’s scarcity is physical. The written record has no natural scarcity. Any scarcity around it is imposed: by law, contract, technical control, or capital.
That is why the comparison to gold, oil, timber, and the rest is not hyperbole. Add them all up and you have a pile of things that run out. The human record does not. It is the headwater of a vast river.
Nobody made it, and everyone made it
Ask who produced this record and the answer is: nobody alive made more than a fragment of it.
The dead did much of it. Euclid, and the anonymous copyists who kept him legible for two thousand years. The clerks who kept the ledgers. The woman who wrote down the recipe. The engineers of a company that no longer exists, whose manual still explains how the machine worked.
No company produced it. The oldest firm on earth was founded fourteen centuries ago. The record is more than three times older. No country produced it either. The record predates every nation on the map, and much of it was written in places that no longer carry those names.
The labs’ contribution is real: enormous investment, difficult engineering, and years of research to make the record computationally useful. They may own their models, their infrastructure, and the services built on them.
But those are the pipes. Owning the pipes does not confer ownership of the well.
We have done it before
It is not utopian to say that something can belong to no one. We have made that decision a handful of times, and mostly it has held.
Antarctica is one example. In 1959, twelve nations, including seven with active territorial claims, signed a treaty that froze those claims, banned military activity, and reserved the continent for peaceful scientific cooperation. Nearly seven decades later, Antarctica still belongs to no nation.
The high seas are another, less perfect, example. No flag owns them outright. They are governed by a shared, imperfect, and often contested international rulebook.
The Human Genome Project is the closest precedent. In 1996, scientists adopted the Bermuda Principles, agreeing to release human genome sequence data rapidly into the public domain. The decision was meant to keep a shared record of what we are from being enclosed behind patents and paywalls. A private company then raced to sequence the genome first and sell access. The public effort did not defeat that ambition through prosecution. It made enclosure less valuable through publication.
That is the precedent that matters. A private firm, well funded and fast, racing to enclose a shared record of humanity, and a public answer that made the fence less useful.
Why acceptance is survival
The frontier labs’ recent calls for caution may be motivated by public responsibility, by self-interest, or by both. It would be naïve to assume otherwise. But whatever the motive, the labs have raised a question larger than antitrust, and larger than they may intend: who holds the well.
Suppose a company effectively holds it. The price of an answer is then set by a board, and the world thinks at the rate a quarterly earnings call permits. A rent is charged on thought. The loss is not measured in barrels but in the questions never asked because asking costs too much.
Suppose a country holds it. Then the human record becomes what oil became: a strategic commodity, secured with export controls, sanctions, alliances, and, when those fail, worse. The twentieth century spent an extraordinary amount of blood over fossilized life. There is no reason to assume the twenty-first will be gentler over fossilized thought, if we allow thought to become something a firm or a nation can hold. Accepting that the well belongs to no one is not generosity. It is how we avoid fighting the last century’s wars over this century’s resource.
We have learned how to make the accumulated record of humanity answer back. That achievement does not make the record private property.
The labs may own the pipes they built. No company, and no country, should own the well.