On November 30, 2022, something that would reshape the cognitive landscape of 675 million people was released into the world. A student in Jakarta used it to write her essay. A government official in Bangkok used it to draft a policy speech. A call center agent in Manila read about it on her break and wondered — with that particular stillness that comes before fear — if her job would exist in five years.

None of them were consulted. None of them received a letter, a notification, a town hall. The most significant transfer of cognitive power since the invention of the printing press happened on a Tuesday, and nobody in this part of the world was invited to the ceremony.

This Is Not Progress. This Is Migration.

In every transfer of power throughout Southeast Asian history, some moved forward while others were left behind — holding a letter they never received, realizing too late that something had been taken from them in a language they didn’t know was being spoken.

The 1830 Cultuurstelsel worked the same way. The Dutch did not arrive one morning and announce: we are taking your land. They arrived with registration systems, then incentives, then requirements. The farmers of Java continued to farm. The rice continued to grow. The value simply flowed elsewhere. The structure was elegant in its cruelty: designate the land, designate the crop, designate the destination. The farmer never received a memo. He just noticed, a few seasons in, that his children were thinner.

ChatGPT is the registration system. The land being designated is your attention. The crop is your behavior. The destination is the inference layer of someone else’s product. The farmer is still farming. The harvest is just leaving.

The Three Vignettes

I have spent the last three years sitting with three of those November 30 users.

The student in Jakarta is now twenty-two. She graduated with the help of AI assistance her professors never detected. She took a job at a digital agency. She uses the same tools there. She is competent, fast, employed. She is also, in a way she cannot articulate, slightly less herself than she was at nineteen. The version of her essay that her professors graded was not written by her. The version of her thinking that produced that essay is also not entirely hers. The transaction was clean. The exchange rate was terrible.

The official in Bangkok commissioned a national AI strategy six months later. It was well-written. It referenced the right frameworks. It had the right number of stakeholder consultations. It was, in the precise language of policy work, performative. The document described a future that had already been built by someone else, in someone else’s datacenter, governed by someone else’s terms of service. The official was not lazy. He was outmanned. The strategy needed to be written in three months. The technology was advancing in three weeks.

The call center agent in Manila is still working, for now. She is twenty-nine. She has moved from customer service to “escalation specialist” — the calls the chatbot cannot handle. The pay is the same. The job security is worse. She told me, over a meal in Quezon City, that she does not blame the chatbot. “It is what it is,” she said, in the Filipino phrase that carries more weight than its syllables. She is, by any honest accounting, training her own replacement every time she takes a call. The data from her escalation becomes the training set for the next model. The harvesting is invisible. The memo never came.

What the Memo Would Have Said

If we had been consulted — if a town hall had been called, if a notification had been sent, if the cognitive revolution had arrived the way other revolutions once arrived, with banners and opposition — we would have asked three questions.

First: who owns the inference? When a student in Jakarta writes an essay with the help of a model trained on data she never consented to, who is the author? When a Bangkok official drafts a policy with the help of a system governed by terms of service in a language she did not read, who is accountable? Ownership is not a legal question. It is a power question. We did not get to ask it.

Second: who trains the model? The training data is the new tanah. It is the new oil, the new rice, the new lithium. The economies of Indonesia, the Philippines, Vietnam, Thailand, Malaysia, and Singapore are the largest underutilized training corpora in the world. Our languages, our behaviors, our transactions, our disagreements, our recipes, our lullabies, our complaints about traffic — all of it is being scraped, cleaned, and incorporated into models that will be sold back to us at premium prices. We are the feedstock. We are not the customer.

Third: who gets the productivity gains? The World Bank has noted, in language carefully chosen to avoid offending anyone, that AI benefits are “mainly captured by wealthy nations and major tech firms.” This is the polite translation of an impolite fact: the rising tide does not lift all boats. It lifts the yachts. The rest of us get better-organized statistics about how fast we are sinking.

The Transfer Is Not Over

Here is what makes this moment different from the Cultuurstelsel, different from the printing press, different from the steam engine: the transfer is not finished. The model is still being trained. The inference layer is still being designed. The governance frameworks are still being written. The window for asking the three questions — and getting an answer that is not predetermined — is still open.

It will not be open for long.

Southeast Asia is home to 675 million people. We are the fifth-largest economy in the world by nominal GDP, the third by purchasing power. We have 70 million MSMEs. We have 1,200 living languages. We have the youngest digital population on Earth. We have, in the precise words of every AI strategy document ever written, “untapped potential.” What we do not have is sovereignty over the system that is being built on top of that potential.

The memo did not come. We will have to write our own.

This is the first post in a series about what that memo should say. It will draw from a book I have been writing on AI sovereignty for Southeast Asia, called Inteligensia. It will not be polite. It will not be neutral. It will, I hope, be useful.

The day everything changed was a Tuesday. We missed the ceremony. We can still build the cathedral.