And someone Else Owns the Deed.

In 1830, the Dutch introduced the Cultuurstelsel to the East Indies. The system was elegant in its cruelty: designated land would grow export crops for the colony. Farmers were required to cultivate what Amsterdam demanded. The value flowed outward. When famine came — as it always came when you feed others before feeding yourself — it was recorded as an unfortunate externality. The shareholders in the Netherlands never saw the cost.

Today, we have a new cultivation system. We don’t call it that, of course. We call it digital transformation and AI adoption and the platform economy. We call it engagement. We call it growth. We call it the future.

But the structure is identical.

Your attention is designated land. Your behavior is the crop. The harvest flows to Silicon Valley and Beijing. You cultivate engagement for platforms you don’t own, using data you generated but never retained, and when the cognitive displacement comes — when your child prefers AI-generated stories to her grandmother’s tales, when your customer service agent is replaced by a chatbot, when your city’s traffic lights are optimized by an algorithm that has never seen your city — it is recorded as progress. The shareholders in California never see the cost.

The Registration System

The Cultuurstelsel did not begin with land seizure. It began with registration. The Dutch colonial administration, in the 1820s, undertook the first comprehensive land survey of Java. The system was, by the standards of its time, technologically sophisticated. It involved cadastral mapping, title documentation, and a centralized registry in Batavia. The technology of the day was, in the precise language of the Cultuurstelsel’s later apologists, “the most advanced available.”

The point of the registry was not to record ownership. The point of the registry was to make ownership legible. Legibility, as the political scientist James Scott observed in Seeing Like a State, is the precondition of extraction. You cannot tax what you cannot see. You cannot seize what is not on a map. You cannot extract value from a system that does not yet know it has value to extract.

The registry made the land legible. The Cultuurstelsel then made the land productive. The productivity flowed to Amsterdam. The farmers of Java continued to farm.

The smartphone made your behavior legible. The platform economy then made your behavior productive. The productivity flows to Menlo Park and Mountain View. You continue to scroll.

The Numbers, If You Want Them

Here are the numbers, in case the metaphor is too literary for the moment.

Facebook and Meta platforms reach approximately 400 million users in Southeast Asia. TikTok, owned by ByteDance, reaches more than 300 million. Google, through Android, reaches effectively the entire smartphone-using population of the region. The behavioral data generated by these users — what they read, what they watch, what they buy, what they hesitate over, what they almost bought and did not, what they bought and regretted, who they are connected to, who they have not spoken to in three years but still think about — is harvested, processed, and incorporated into advertising systems, recommendation systems, and increasingly, into the training corpora of large language models.

The value of this data, by any reasonable estimate, is in the hundreds of billions of dollars annually. The value captured by the users who generated it is, in the precise language of the platform economy, zero. The value captured by the local economies in which the users live is, with a small number of exceptions, also zero. The value flows outward.

In Indonesia, the world’s fourth most populous nation, with the world’s largest Muslim population, with 700+ living languages, with 70 million micro, small, and medium enterprises, with the world’s largest archipelagic state — in this country, the value of locally generated digital data captured by local enterprises is, in the language of a recent Kearney analysis, “a small fraction of one percent of total digital value created in the country.”

One percent. With a small fraction attached.

The Three New Cultivation Systems

The data colonization of Southeast Asia is not a single event. It is a layered system. There are at least three distinct cultivation systems operating simultaneously, each more invasive than the last.

The first cultivation system is attention. Your attention is the crop. The platforms have built extraordinarily sophisticated irrigation systems — push notifications, infinite scroll, variable reward schedules, social validation loops — to keep the crop growing. The harvest is the advertising revenue that flows to the platform’s shareholders. The externalized cost is the time, the cognitive capacity, the relationships, the sleep, the focus, the presence — all the things that attention would have been spent on, in a world where it was not being extracted.

The second cultivation system is behavior. Your behavior is the crop. The platforms have built prediction systems that know your habits better than you do. They know when you are most likely to be lonely, most likely to be bored, most likely to be politically persuadable, most likely to be financially vulnerable. The harvest is the targeting precision that allows the platform to sell access to you, in your most suggestible moments, to the highest bidder. The externalized cost is the autonomy that you would have exercised, in a world where your behavior was not being continuously modeled and continuously nudged.

The third cultivation system is cognition. Your cognition is the crop. The large language models, trained on the data harvested by the first two systems, are now being integrated into the tools you use to think — search engines, writing assistants, customer service platforms, educational technology, medical advice, legal research. The harvest is the inference revenue that flows to the model owners. The externalized cost is the capacity to think for yourself, in a world where thinking is being outsourced, by default, to a system that has been trained on the priorities of its owners.

The Dutch did not seize Indonesian land with swords. They seized it with registration systems. We are watching the same sequence of moves, in the same slow register, with the same polite vocabulary, with the same outward-facing executives who will, when asked, insist that the system is “mutually beneficial.”

The Ink Defeated the Memory

In the book I am writing on AI sovereignty for Southeast Asia, I have a chapter called Semangat — The Concept Silicon Valley Doesn’t Have. The chapter opens with a keris in a glass case in Museum Nasional Jakarta. It is older than your country. The empu who forged it fasted for forty days. He did not eat meat, did not speak unnecessarily, did not touch his wife. He was not merely shaping metal — he was transferring something into it. The Javanese call this semangat: not “spirit” in the Western motivational sense, but something closer to life force, an animating presence that can be imbued into objects, places, and communities.

The chapter is about four containers of semangat in Southeast Asian civilization. The first container is the physical sacred — the keris, the phra kring amulet, the Filipino agimat, the heirloom that carries something more than matter. The second container is the communal land — the adat territories, the communal rice paddies of Thailand and the Philippines, the village commons that were owned through memory, ceremony, and generations, not through deeds and registries. The third container is the digital extractive — the data layer that has, in twenty years, replicated the structure of the colonial land registry with a precision the Dutch could not have imagined. The fourth container is the cognitive — the reasoning capacity itself, the last container, the one that the AI revolution is now colonizing.

The chapter has a sentence I have been turning over for two years: Ink defeated memory.

The Dutch colonial land registry defeated the adat system of communal land ownership not by being morally superior, but by being more legible. The ink on the registry was, in the precise language of statecraft, more useful than the memory of the village elder. The registry could be transported, copied, enforced, contested in court, used as collateral. The memory could not. The memory was local, embodied, slow, particular, alive. The registry was portable, abstract, fast, universal, dead.

The same thing is happening now. The behavioral data harvested by the platforms is more legible than the lived experience of the people who generated it. The inference produced by the large language models is more legible than the reasoning of the people who are being asked to use them. The platform is more useful than the community. The model is more efficient than the teacher. The feed is more responsive than the friend.

Ink defeated memory. Algorithm is defeating attention. The pattern is the same. The technology is different. The outcome will be the same unless we choose, deliberately, to do something different.

What Doing Something Different Looks Like

I am not naive about this. I have spent twenty years building marketing functions for global technology companies. I have built the systems that harvest attention. I have written the copy that drives engagement. I have optimized the funnels that convert behavior into revenue. I am not outside the system. I am a product of it.

But I have also spent the last three years asking, with increasing urgency, what the alternative looks like. The alternative is not a rejection of technology. The alternative is a different ownership structure. The alternative is a different set of defaults.

The alternative looks like data cooperatives — koperasi that aggregate the data of 70 million Indonesian UMKM, governed by the members, with the leverage to commission AI models that serve the cooperative’s interests, not the platform’s. The alternative looks like sovereign data infrastructure — data centers, training pipelines, model repositories, owned and governed by Southeast Asian institutions, under Southeast Asian law. The alternative looks like regional cooperation — the ASEAN AI Commons, the shared model repository, the shared infrastructure, the shared governance that no single nation can build alone but that all six, together, can.

The alternative looks like semangat. Not the Western concept of brand spirit, not the motivational-poster sense of “team semangat.” The Southeast Asian concept. The animating presence that is transferred into objects, places, and communities through investment, intention, and sacrifice. The thing that cannot be registered, extracted, or scaled, because it is not a thing. It is a relationship.

The Dutch could register the land. They could not register the semangat of the land. The platform can register the behavior. It cannot register the semangat of the behavior. The model can register the cognition. It cannot register the semangat of the cognition.

This is not a sentimental argument. It is an operational one. The things that cannot be registered cannot be extracted. The things that cannot be extracted can be retained. The things that can be retained can be the foundation of a sovereign AI capability.

Your attention is the new land. Someone else owns the deed.

The question is whether you are going to let them finish the registration.