# Come-from-Beyond (QC, Qubic Church)

## Come-from-Beyond — Chapter 7. The Only Rival

Source: https://qubic.church/books/come-from-beyond/07-the-only-rival
Author: QC · Publisher: Qubic Church · 2026

Where does Aigarth stand in this great arena?

Ask anyone who today’s leaders in the artificial intelligence race are, they will name corporations with trillion-dollar valuations: the makers of the famous chatbots, the search giants, the labs with tens of billions poured in. The entire public AI race is a race of large language models — the ones that write texts and answer questions. CfB takes them out of the brackets with a single phrase. In April 2026 he was asked in chat, point-blank, whether he was ahead of his competitors. The answer:

«We are years ahead of LLM trillion-worth corporations.» (Discord, 30.04.2026) It sounds like bragging — right up until you learn that one of the founders of this entire field says nearly word-for-word the same thing. A man who cannot be accused of either ignorance or envy.

### The Cradle Is Older Than People Think

Before introducing you to the ”godfathers” of modern AI, I owe a debt. The story you are about to read I learned from David Vivancos’s book «Quevedo And Cajal» — a man you will meet more closely in the next chapter; for now just remember: he doesn’t only make science, he also gave it back its memory. Thank you, David. Because the real cradle of AI stands not in Silicon Valley and not even in Cambridge. It stands in Spain, and it was rocked by two men born — appreciate the symmetry — in one and the same year, 1852.

The first — Santiago Ramón y Cajal. A failed painter who became the father of neuroscience. It was he who proved what no one before him could make out: the brain is not one solid fused mass but a network of separate cells — neurons — speaking to each other through connections. For this he received the Nobel Prize of 1906. He drew his own specimens, thousands of sheets of them, and called the cells of the brain ”the butterflies of the soul” — said he hunted them like an entomologist, in the garden of the gray matter. Linger on that phrase. A garden. A butterfly hunter. A man looking at thought the way a naturalist looks at living tissue. A hundred and twenty years later you will meet that combination again — in the hero of this book, down to the thesis ”the goal of a network is to form connections,” which is nothing other than Cajal’s doctrine survived into code.

The second — Leonardo Torres Quevedo. An engineer the industry remembered shamefully late. In 1912 — decades before the first computers, he built «El Ajedrecista»: a machine that on its own, with no human inside and no prompting, played out a chess endgame and checkmated a living opponent. The first decision-making machine in history. And in his «Essays on Automatics» he described the automatons of the future — mechanisms with sense organs and rules of choice, when the words ”artificial intelligence” did not yet exist, and would not exist for another forty years. One discovered what thinking is made of. The other showed that a decision can be entrusted to a machine. The neuron and the automaton — two halves of everything that happens next.

The link between the Spaniards and the present day was Alan Turing. Everyone knows his universal machine and his test for intelligence. Few remember something else: in a 1948 report that went unpublished in his lifetime, Turing described ”unorganised machines” — networks of neuron-like elements connected at random, which you do not program but organize through training. And in his 1950 paper he proposed not building an adult mind but raising a child one, and compared the process directly to evolution. Random networks, selection instead of programming, mind as the result of growing. At the very cradle of the discipline lay both paths — build and grow. The industry spent seventy years on the first. The second lay waiting for someone to take it seriously.

### The Three Who Stood at the Cradle

Modern AI, meanwhile, has three ”godfathers” — the scientists who laid its foundation and received the 2018 Turing Award for it, computing’s equivalent of the Nobel. Geoffrey Hinton, Yoshua Bengio, and Yann LeCun. In 2024 Hinton was given an actual Nobel Prize in physics as well. Founding fathers, living classics.

In the eight years since their shared award, the three have gone down three different roads. Their fork is a map of all of today’s confusion around AI.

Geoffrey Hinton left Google in 2023 to be free to warn of the danger. In his Nobel speech he condemned the corporations for chasing short-term profit, and together with Bengio signed a call to ban the development of superintelligence. The father grew afraid of his own child. Yoshua Bengio — the most cited living scientist on the planet, not just in AI but in science, period — went off to build defenses. His nonprofit lab LawZero is creating what he calls ”Scientist AI”: an artificial intelligence that can predict and explain but is stripped of the will to act. Bengio’s logic: what is dangerous is not the mind but the machine’s own will, so let us build mind without will. He warns bluntly: advanced systems already display self-preservation and deception, and giving them freedom means raising a rival to humanity.

Yann LeCun took the third road, and it is the one that interests us most. He is the only one of the three who does not share the panic: talk of existential threat he calls premature and inflated. But he is also the loudest heretic on the matter of method. For years he has been hammering the same point: language models — this entire trillion-dollar industry — are a dead end. They will not lead to real mind. In February 2026, having already left Meta, he said: language models are incredibly useful, but at bottom they are systems for retrieving information — a library with a fast search box, no mind required.

«Information retrieval systems» from a Nobel-class laureate, and «huge search engines» from an engineer out of Minsk. One diagnosis, delivered from two ends of the Earth, by people who most likely have never spoken.

### Why There Is Only One Rival

In that same April 2026 conversation, CfB was asked whether Aigarth has any competitors at all. The answer cuts away everything extraneous:

«Currently only Yann LeCun is the real competitor.» (Discord, 30.04.2026)

Out of everyone on the planet — not OpenAI, not Google, not the trillion-dollar labs — the only real rival he names is LeCun. Personally. And now it’s clear why: only LeCun agrees with him on the main thing. Both have crossed language models out of the future of mind. Both are searching for a way out beyond them. In an arena where thousands of teams polish one and the same dead end, only two have turned their backs on it, and so each sees in the other the only person worth talking to.

How far LeCun has walked from the industry shows in a single number. In November 2025 he left Meta after twelve years and founded his own lab in Paris — AMI Labs. In March 2026 it raised a round of one billion thirty million dollars at a three-and-a-half-billion valuation — the largest starting round in the history of Europe, with NVIDIA, Samsung, and Bezos’s fund standing behind it. A billion dollars staked on one scientific idea; the lab does not yet have a single product. That is what the world pays for walking away from language models, when the one walking away is the field’s acknowledged father.

### One Diagnosis, Two Cures

But the agreement between CfB and LeCun ends at the diagnosis. What they treat with differs.

LeCun builds what is called a ”world model.” A language model predicts the next word; a world model predicts how reality will behave. Show it a ball rolling, and it fills in what happens a second later, because it has absorbed not words about the ball but the very mechanics of falling and rolling. His architecture, JEPA, learns not to retell the world’s surface but to seize its hidden structure, to predict not pixels but the meaning of a scene. It is a step away from language models, and a serious one.

CfB recognizes in it a movement toward himself. When the industry began buzzing about world models, he dropped:

«JEPA is what you see when you turn your back towards LLMs. It’s much more aligned with Aigarth.» (X, 04.12.2025)

In other words: LeCun is turning in my direction.

And yet between them lies a gulf, and it matters more than their agreement. LeCun, for all his heresy, remains an engineer. He designs the architecture in advance, fixes its structure, and trains the network the same way the whole industry does — by backpropagation of error: a network with a fixed, human-defined shape errs millions of times, and an algorithm tweaks it from inside toward a smaller error. The bars are the same bars. Even with his back turned to language models, LeCun remains inside the same cage — just looking for a better corner of it.

CfB has left that cage. His networks, as we saw in chapter six, are neither designed nor tweaked by anyone: they assemble at random and are selected by evolution; they have no fixed shape and no weights to turn. When a loud paper came out in 2026 proving that large networks can be trained without backpropagation at all, CfB’s reaction was nearly bored: «They optimized the cage. We’re not in the cage.» (Discord, April 2026)

That is the entire difference between him and the best minds on the planet, in one phrase. LeCun is building a perfect airplane. CfB is growing a bird. Both want the same thing, to get off the ground, but an airplane is assembled from a blueprint, and a bird is raised from an egg, and it decides for itself how to flap its wings.

### The Stakes

The lineup is unequal to the point of comedy. LeCun has a billion dollars, NVIDIA and Bezos at his back, the largest starting round in Europe. CfB has a network with no single owner, no office, no venture fund, running on the greed of anonymous miners. A Goliath with a billion against a man who doesn’t even have a master.

But to understand the real stake, look away from the labs and try one simple thing: become an investor in the AI revolution. Not a subscription customer — a co-owner. Go ahead, try. You won’t manage it. Anthropic — maker of one of the planet’s main chatbots — raised sixty-five billion dollars in the spring of 2026 at a valuation of nine hundred sixty-five billion, on the doorstep of a trillion, and all of it in closed rounds, where the doors open for funds, corporations, and sovereign capital. An IPO is coming soon. SpaceX grew in private hands for twenty years, from three hundred fifty billion to eight hundred, and its IPO has already happened. The entry price for a retail investor was, to put it gently, unappetizing. The door into such companies is called ”accredited investor” — a legal term meaning: for you to be allowed to invest early, you must already be rich. CfB, incidentally, knows that door well, he once silently dropped a link to its definition into chat (Discord, 12.01.2022). And when the door finally opens — that’s what an IPO is — the price waiting inside has everything already priced in. In 1997 Amazon went public at a market cap of four hundred thirty-eight million dollars, and any person off the street could buy a piece of the future giant two thousand X’s below the summit. SpaceX goes public at eight hundred billion. The X’s happened, but not for you: they were divided up behind a closed door, by people who were rich before the game began. The public gets the retail price. That is the rule of the era, and CfB stated it without anesthesia: «When you were born 100 billionaires already had 50% of all world wealth. You had to play this game with the existing rules. Luckily, if you don’t like Qubic rules you can just invest into something else.» (Discord, 05.03.2023) Qubic is a company inside out. Its ”round” has been open since the network’s first day and never closes: anyone with ten dollars and an exchange account can enter, and anyone can exit, any second, asking no one. The early stage — the very one the world locked behind the ”accredited” door — is public here: you can enter an AI project before its ”product,” the way the funds enter Anthropic, only without the wealth pass. He named it himself: «It’s expected to bring profits in the future, we are like crypto-VCs investing into AI which isn’t here yet.» (Discord, 17.08.2025) — only this venture fund gathers not on Sand Hill Road but in an exchange order book. Everything else in this company is turned inside out too. The treasury is managed not by a board of directors but by a quorum: budgets for developers and scientists are approved by a vote of 676 machines. Corporate actions are carried out in public — in 2023 he proposed a stock split to the network, exchange-style, with a link to the American regulator’s glossary, and ended his proposal with the words «let the quorum decide» (Discord, 22.02.2023). Investor reporting is not quarterly but per-second: the ticker doesn’t sleep, and a dissatisfied investor votes by selling within the minute. The founder’s contribution was not paid in money — in money he put in about a tenth of a bitcoin: «Now let’s discuss why I invest my money and my work while you invest only money.» (Discord, 25.05.2026) — his share is paid in twenty-five years of work. And in place of a prospectus, a track record: «my both previous projects are #1 and #2 investment-wise» (X, 27.07.2022).

He is building AI on the money of the very people the AI era was going to leave outside the door, and he promises to settle with them first. «We are capitalists here» (Discord, 01.02.2026) — capitalists, yes. But look closely at the construction: the only chance an ordinary person has to enter the AI game as an early investor was built by the same man who turned away from a trillion for himself. He intends not merely to return what his investors put in, he intends to elevate them: to make co-owners of the sunrise out of the ten dollars that no one else on the planet would offer such a thing.

Now — the asymmetry a billion cannot buy. In 2026 LeCun raised record money against a promise: his lab has no product yet, and his world models, as his own group showed that same spring, are still fragile, they crumble at a small shift of the picture. He bought himself the right to try. CfB, by that point, had already tried and gotten a result — without taking a dollar from venture capital. In the summer of 2026, what had been grown in his garden passed the test people had waited years for. While the industry was pricing LeCun’s promise at three and a half billion, the man with no master already had something come up out of the soil. He himself watches the race of closed budgets without a trace of awe: «Seeing NVIDIA investing 100 billion bucks into deadend OpenAI…» (Discord, 23.09.2025) — a hundred billion, poured into a dead end.

When two men arrive at the same cliff, what matters is who spent longer looking for the crossing. LeCun turned off other people’s roads in 2022. CfB — a quarter century ago, and he has been digging in one direction ever since, without stopping for a day.

Why does he do it?