The Forest Doesn't Kill You. It Trains On You.

How the Cognitive Dark Forest Perfected What Empire Started

Silicon Valley is discovering what Africa has always known: the system doesn’t need your compliance. It just needs your activity.

In March 2026, a developer named Janko published an essay called “The Cognitive Dark Forest.” The argument is borrowed from Liu Cixin’s cosmological horror novel The Dark Forest, the second entry in the Three-Body Problem trilogy, which posits that the universe is silent not because it is empty, but because every surviving civilization has learned that announcing your presence is a death sentence. The rational move is to hide. Always. The asymmetry is brutal: even if four out of five civilizations that notice you are benign, the fifth will annihilate you. Probability with permadeath.

Janko’s application to the current web: AI platforms have made the open internet dangerous in precisely this way. You build something in public. You share your code on GitHub, your ideas on forums, your MVPs to early users. You announce your presence. Now, with large language models cheap enough to clone your specific innovation every few days, and with those same platforms reading the thermal gradient of your prompts, not your individual prompt, but the cluster, the demand curve, the heat map of where human interest concentrates, the forest has learned to absorb you before you even know you’ve been digested.

His words, which I will quote because they deserve their sharpness: “The platform will know your idea is pregnant far before you will.”

It is a good essay. It is also, for anyone paying attention from the Global South, about seventy years late.

---

“Surveillance” and “statistics” are not the same thing. Surveillance implies a specific watcher and a specific watched. It requires intent, manpower, a file with your name on it. Statistics requires none of that. It requires only aggregation. The platform does not read your prompt. It reads the shape of ten million prompts. The individual disappears; the pattern remains, and from the pattern, it can see where the world is going before the world knows.

This distinction is important because it explains why the usual defenses don’t work. You are not being targeted. You are being averaged, and the average, in this case, is more valuable than you.

The colonial powers understood this. The British East India Company did not need to spy on every Indian farmer to extract value from India. It needed the pattern: what crops grew where, what labour rates were, what social structures could be leveraged for indirect rule. The ethnographic machinery of colonial administration: the census, the land survey, the anthropological study, the missionary report, was not surveillance. It was statistics. A demand curve made of subjugated populations.

The IMF’s structural adjustment programs operated the same way. When a desperate government sat across the table from the Fund in Washington and said, “We need a loan,” that statement was itself a signal. The cluster of signals, 47 African governments asking variations of the same question between 1980 and 2000, gave the Fund a perfect gradient in policy space. It knew what each country would accept before the negotiation began. It had the heat map. The countries did not.

---

Janko’s sharpest observation is what he calls the absorption paradox: “Resistance isn’t suppressed. It’s absorbed. The very act of resisting feeds what you resist and makes it less fragile to future resistance.”

This is, word for word, the story of post-colonial African intellectual history.

The Bandung Conference, April 1955, twenty-nine African and Asian nations meeting in Indonesia to assert a Third Way, non-aligned, neither Washington nor Moscow. Genuine. Radical. World-historical. By the 1970s, its vocabulary had been absorbed into UN development discourse. By the 1990s, “South-South cooperation”, Bandung’s intellectual heir was a sub-committee item in the same multilateral institutions Bandung was built to resist.

Thomas Sankara. President of Burkina Faso from 1983 to 1987. He rejected IMF loans, launched mass literacy campaigns, redistributed land, renamed the country from Upper Volta. He was assassinated in October 1987, reportedly with French intelligence foreknowledge, according to documents examined by Burkinabé investigators. Within two decades, his language, “food sovereignty,” “endogenous development,” “debt as neocolonialism”, had been processed, sanitized, and reissued as World Bank best-practice guidelines. The forest ate him. Then it wore his vocabulary as a costume.

Ubuntu. A Nguni Bantu philosophical concept of communal personhood, “I am because we are.” By 2015, a Canonical product, a management consulting framework, and a TED Talk theme. The forest didn’t kill Ubuntu. It trained on it.

This is what Janko is describing for developers, and he thinks it’s new. It is not new. It is new for people for whom the meadow was real.

---

The upgrade, the genuinely novel thing that AI introduces, is that execution is now cheap.

Before large language models, an empire that wanted to absorb your innovation needed human infrastructure. Missionaries who took decades to learn local languages. Colonial administrators who died of tropical disease before they could file their reports. Economists who needed to be flown business class to Abuja to conduct their assessments. Programmers: expensive, slow, constrained by meat-space-and-time, as Janko puts it. Meat doesn’t scale.

Now the gap shrinks. Claude, GPT-4, Gemini: these are missionaries who don’t catch malaria, don’t need a boat, don’t need a decade to learn the language. They can ship a variation of your innovation every few days. They already own the compute, the models, and the developer data. The cost of absorption approaches zero. The cost of building, for you, remains.

This is the upgrade. The dark forest has always existed. What changed is that the predator got faster.

---

Sorry chiefs. I want to pause here and acknowledge the comedy.

A developer buys a refurbished ThinkPad in 2009, installs Xubuntu, starts coding. No gatekeeper, no middleman. The internet is a bright meadow. Sharing is cool. Ideas are cheap, execution is the moat.

Now, in 2026, that same developer is learning that the meadow was a trap, that announcing your presence is dangerous, that the safest move is to hide.

This is the opening chapter of every African economic history written since 1960. The resources were there. The labour was there. The ingenuity was demonstrably there. The problem was that every time you announced your presence, every time you found something worth finding, someone with more capital and faster execution showed up and incorporated it, and then sold it back to you.

The meadow was always a trap.

Africa would like you to know: welcome. We have been sitting in this dark forest for quite some time. The coffee, unfortunately, is not hot.

---

Janko ends his essay with what he calls the “final recursion”: the essay itself is now in the forest. By writing about the dynamic, it became part of it. The models now know a little more about why people might hide. He cannot step outside the forest to warn people about the forest. There is no outside.

Frantz Fanon wrote The Wretched of the Earth in 1961. It is now a syllabus text in the political theory departments of Cambridge, Columbia, and Sciences Po, the universities of the same civilizational project Fanon was dissecting. His critique of colonial psychology is taught by institutions whose structural function is to reproduce the conditions he described. The forest ate Fanon too, and keeps eating him, semester after semester, in beautifully designed reading rooms.

There is no outside. Africa has known this longer than anyone. The warning is not new. The warning has been absorbed.

---

So here is what I am watching, and what you should watch.

Every time someone in Lagos opens Claude and prompts it, looking for a business angle, a legal workaround, a product idea, a way to survive in a contracting economy, that prompt is a signal. Not of that person. Of a pattern. The gradient in idea space across 54 African countries, countries that have been problem-solving under resource constraint for generations, building informal systems that formal economists don’t model, finding workarounds that never appear in white papers, is an extraordinarily rich heat map.

Who owns that aggregate?

Which company sees the demand curves clustering across Accra and Kinshasa and Nairobi and decides, based on those clusters, what product to build next?

Whose ingenuity: constraint-honed, necessity-sharpened, invisible to the market until the market needs it, becomes which hemisphere’s competitive moat?

The cognitive dark forest is real. Africa has always been in it. The only new thing is that the forest has gotten better at pretending to be a meadow, and the developers are only now learning to check.

Watch the gradients. Watch who owns the data. Watch what gets built, and where the builders are.

---

This publication runs on subscriptions, not algorithms. If you find the receipts useful, the paid tier keeps the lights on.