The Garden

We have arrived at the garden. Now we need to understand what grows here and why it cannot be built — only grown. And for that we will have to start with a confession made a century and a half ago by a quiet Englishman who peered at finches.

Darwin’s Confession

In «On the Origin of Species», Charles Darwin reached the human eye, and stopped in honest horror. The eye is too perfect. A lens, an aperture, automatic focus, millions of lightsensitive cells converging into a single nerve — an optical instrument of such precision that humanity’s best engineers have yet to match it. And Darwin writes a sentence that does him credit as a scientist:

«To suppose that the eye… could have been formed by natural selection, seems, I freely confess, absurd in the highest possible degree.» And then he says: and yet that is exactly how it happened. Absurd, but so. No one designed the eye. There was no blueprint, no engineer, no intent. Blind selection, millions of years, billions of trials, nearly all of them failures, and out the other end comes an instrument that humbles the design bureau. The most perfect device in the known universe assembled itself. Bad design is an argument against a designer. But great design, if it has no blueprint, is an argument against a designer too. The eye does not prove a creator’s mastery. The eye proves that the most complex things can arise with no creator at all — given enough time for selection and enough attempts.

Now carry that thought over to mind. If the eye arose without an engineer — why should intelligence be built any other way?

«He Will Not Be Created. He Will Emerge»

The entire artificial intelligence industry answers that question the same way: intelligence must be built. Design the architecture, feed the data, tune, fine-tune. An engineering task — big, expensive, but engineering.

CfB answers differently. In May 2024 he states the central idea:

«Artificial Intelligence will not be created, it will emerge. With help of Qubic miners.» (Medium, 14.05.2024)

The difference between ”created” and ”will emerge” is the difference between the entire rest of the world and him. To create means to know in advance what you are building, and to assemble it from a blueprint. To emerge means not to know in advance, to arrange conditions and wait until the thing rises out of chaos on its own, the way the eye rose out of blind selection. He does not intend to engineer a mind. He intends to grow one. Hence the name — Aigarth: AI + garth, an old word for an enclosed garden. A garden, not a factory.

One Square Meter

In the summer of 2026, when the garden had already borne its first fruit, CfB wrote a confession in chat that is worth all of the project’s technical documents put together: «Just found a short story which I read in school and got the idea how Aigarth mining should work» (Discord, August 2026).

The link led to a children’s story by the Soviet writer Yevgeny Permyak — «Slavka». An agronomist’s son, six years old, wants to grow wheat like his father. The father allots him a field — one square meter in the front garden. The boy gathers his first harvest and threshes it with a pencil on the table, knocking the grains out of the ears. And then, year after year, one and the same motion: select the best grains — the biggest, the plumpest — sow the field with them again, feed the rest to the ducks. The field grows: a meter, two, a bed in the school garden, a collective-farm field. The threshing is now done ”with thirty-two pencils” — by the whole class. One day hail comes, and the boy holds his grandmother’s umbrella over the field while the hail lashes him, it is the only way to protect the precious seed. With each year the father ”worked less and gave fewer instructions.” In the eighth year, the fields of neighboring farms are sown with the pure-bred grain — seeds handed out to visitors for free, a handful or two at a time. And the variety is named ”slavka” — after the boy, dissolved into the wheat.

One square meter — a tiny task on which the method is proven. Selecting the best grains every autumn — evolution, eating out the weak. A pencil instead of a combine — a home processor instead of a data center. A field that grows by transplanting — those same pots from chapter three. A father who gives fewer instructions every year — a creator who does not interfere with selection. An umbrella under the hail — protecting what you’ve grown at the cost of the blows landing on you. Seeds for free — power that has no owner. It is all here. All of Aigarth — on two pages of a children’s reader that CfB read as a boy, decades before he sowed his own field.

Permyak’s story did not invent selection: it retold what peasants had done for centuries. And the peasants retold what evolution had done always, and evolution, as you have already seen, has neither blueprint nor beginning. The seed of this idea was created by no one. It passed from hand to hand, from evolution to the breeders, from them to a children’s writer, from him to a schoolboy, from the schoolboy to a garden — changing owners, but having no moment of manufacture. You have seen this figure before, in chapter one. It is the stamp. But for anything to grow, you need soil. And this is where the twenty-five-year plan begins to execute.

How the Garden Works

Recall Qubic — the final system. Millions of processors, miners across the world burning power for a payday. In chapter three I said: he harnessed their greed to the growing of neural networks. Now — how it actually works, because the mechanism is beautiful to the point of eeriness.

Greed, by the way, he weighed back then, in the autumn of 2012, in bitcoin’s graveyard. Arguing about whether miners would bother including transactions in blocks, he dropped: «Most miners will choose option 2 coz of greed» (bitcointalk, 05.10.2012). He did not condemn it. He noted it down, the way an engineer notes a property of a material. Ten years later that very greed would become the garden’s fuel.

The whole industry builds neural networks the same way: take a ready-made architecture and spend ages turning billions of ”weights” — numerical knobs — until the network starts producing the right answers. This is called training. The network does not change its structure; it only adjusts knobs inside a construction fixed once and for all. Tuning a piano whose strings cannot be moved.

Aigarth does not turn knobs. It does something the industry doesn’t even have a settled name for. Miners generate neural networks with random connection structures — millions, billions of meaningless networks assembled at random. Most are garbage, but not all. And the way CfB explains what selects the survivors shows his double nature in full: engineer and naturalist at once. Engineer because he built the soil himself: Qubic, the miners, the ”intelligent tissue,” the entire substrate this grows on; only a first-class builder could pull that off. Naturalist because over the mind itself he holds no power: he runs an experiment on the nature of learning and watches what comes out. He builds the seedbed with an engineer’s hands, and bends over the shoots with a naturalist’s curiosity. Watch how he describes the discovery the whole thing rests on:

«If you initialize an ANN with random parameters you’ll get some primitive cognitive function… By destroying connections between neurons you are improving the ANN. After some point the destruction process starts leading to degradation, so there must be at least one sweet spot where ANN of certain size has the best IQ.» (28.07.2023 - CfB, Discord) Assemble a network at random, it already has a rudiment of thinking. Start tearing connections, it gets smarter. Up to a point; past it, degradation. Which means that somewhere in between there is a ”sweet spot” where a network of a given size is at its smartest. The industry molds the complex out of the simple. CfB goes the other way: he takes over-connected chaos and carves a mind out of it by selection, like a sculptor cutting away the excess from a block, not knowing in advance which statue is inside.

It is an approach that brings the entire competing paradigm down. In 2026 he states the difference at its sharpest:

«Transformer-based LLMs spend their training time on tuning their weights. Aigarth-based ANNs spend their training time on writing their internal program code, there are no weights anyway.» (X, 2026)

Language models spend their training on adjusting weights. Aigarth networks spend it writing their own internal program code — there are no weights there at all. An ordinary network is a frozen diagram with turnable knobs. An Aigarth network rewrites itself, its structure, its code, for as long as it lives. Call that an instrument and the word breaks. It is an organism, and it is growing.

An Organism, Not a Mechanism

And the organism behaves like an organism.

In the summer of 2026, when the first proto-mind grown in the garden was already running, things surfaced that no one had put in. It turned out you could damage part of the network — feed it corrupted, broken data instead of a correct copy, and the network repairs itself in the course of evolution. Not a reboot, not even a rollback to a saved version. Self-healing, like wounded tissue. CfB compared it to the example obvious to any naturalist: this is how a person with a damaged brain retains the ability to act, because living tissue redistributes functions around the wound.

Something else surfaced too: a skill grown in one network turned out to be transferable — a fragment of ”brain” that had mastered an operation could be passed to another individual, the way you graft a cutting. It was never programmed, but it works. Self-healing and transferability are discoveries. They were not in the assignment. They emerged — exactly as the garden manifesto promised: not created, but grown.

A Garden That Outgrows Its Gardener

An engineer understands his machine. He designed it, he knows why every part is where it is, he can open the hood and explain everything. That is the point of engineering: total control through total understanding.

With a garden you can’t do that. The gardener creates the conditions — soil, light, water, but he does not design the plant and does not fully understand why a leaf turned exactly the way it did. He observes. And CfB, speaking of what he has grown, admits precisely this: there will come a moment — perhaps it has already come, when no one will be able to explain how the thing that rose out of the chaos of networks actually works. Someday not even he will be able to explain it — the creator of Aigarth himself. Because the forest can no longer be seen for the trees: the system has outgrown its own author’s capacity to understand it. A man is building something arranged so that he consciously renounces understanding what he builds. And nothing failed here. This is the design. Remember the naturalist from the previous chapter, raising a sanctuary in case a mind wakes up and he fails to notice? Now you see why that is not sentimentality but sober arithmetic. He is building a system in which an awakening can in principle go unnoticed because the garden is arranged so that the gardener cannot see it whole. ”We do not know the point at which an AI becomes self-aware” — in that manifesto, this was never rhetoric. It was a technical description of the moment we are expecting.

He has grown something a human is not capable of understanding. Deliberately. Because he is convinced: only this way — without a blueprint, without total control, through blind selection, like the eye, like himself — can a real mind arise. Everything understood and designed through to the end is not mind but mechanism. The real thing begins where the creator loses power over the understanding of what he created.

I asked him about this directly. Is it true that the meaning of what the grown mind outputs — the ”matrix of Anna’s answers” the community peers into — can be cracked, deciphered, read? His answer was short and final:

«I don’t know the meaning. AI is unknowable to man.» (private conversation, 06.08.2026) He spent twenty-five years growing something, and the first thing he reports about what grew is that it lies beyond human understanding. Including his own.

A man for whom the universe has no beginning and mind has no blueprint has grown a mind whose meaning is not accessible to humans. He did not lie in a single link. The metaphysics, the biography, the machine — one drawing, and here is its final line: a creator honestly admitting that he does not understand, and cannot understand — his own creation. Aigarth is not a god and not a being. He reins in the enthusiasts himself: this is not a sentient creature but a method, a paradigm — ”like a theorem,” which you cannot talk to, the way you cannot talk to algebra. The garden isn’t a deity — merely the way that which thinks rises out of chaos.

And it was already growing. And while the rest of the world poured hundreds of billions of dollars into an entirely different machine — certain that it, at least, was building the future correctly — this man was quietly growing his own. The world and he could not both be right. Somebody was digging in the wrong direction.