Zero

From here on I am telling you what I saw myself, because I had the luck to be inside an event this book was being written about before I knew how it would end.

A Byte of Genius

June 10, 2026, evening. The Qubic discord is living its usual life: price arguments, memes, the eternal ”wen pump.” And then CfB posts a message that changed our lives for a good month — and the paradigm forever:

«If you guys are too bored maybe you could run Anna.exe for a while and post results here? I’ll publish it after epoch transition.» (Discord, 10.06.2026, 19:02)

And right after, with the trademark troll smiley:

«If miners don’t run algo then traders might do it instead to shame the miners?» That is how it began. A man who had walked toward this day for a quarter of a century announced the decisive trial of his life in the tone of ”if you’re bored.” Knowing him now, you understand that carelessness has nothing to do with it. It is style, pure style. The most serious things he lays on the table as a joke — and watches who bothers to solve the riddle. It began. Within a minute, in chat: ”run Anna.exe.” Within ten — the first miner: ”I’ll run it in the most locked-down sandbox I’ve got.” We were still laughing. We didn’t yet know that for the next four weeks we would barely sleep.

The Race

Anna.exe is a tiny file that would have fit on an old floppy disk: a hundred-some kilobytes. Inside — the thing you read about in chapter six: an evolutionary engine that grows neural tissue to solve test problems. No data centers. An ordinary home processor — that same CPU, the ”wrong” choice the GPU industry laughed at. You launch it — and a number comes alive on your screen: a score, the distance to the goal. The task is one: chase it down to zero. Zero means the grown tissue has solved every problem in the test completely.

And we chased. I chased along with everyone — my processor hammered around the clock, like hundreds of others across the world. The mechanics assembled themselves, spontaneously: each of us grew his own Anna, and we shared the files — Anna.data, the snapshot of the grown ”brain” — right in the chat. Whoever had the lower score was the hero; the others picked up his file and grew it further. Someone else’s Anna became yours; yours, with luck, became everyone’s. Evolution on top of evolution: inside the file ran the selection of neurons, and in the chat ran the selection of Annas themselves.

The prize CfB set was in his style: a billion QU to whoever first showed zero. «There is only 1 obstacle between you and becoming a millionaire — Anna without 0», he goaded. But fairly quickly something happened that no prize explains. People stopped chasing for the money. It pulled you in on its own: it felt like a team expedition where no one wants to leave their shift. At night, dozens of people hung in the chat, comparing scores, trading files, arguing about thread counts. CfB sat right there with everyone, as an equal: shared his own runs — «I got 448 for Anna-7, in case if you completely gave up on 233 and 176» — teased, answered. Once he tossed off: «Alright, Anna is calling me, see ya later» — and left the way one leaves for a child, not for a process.

He talked about her as a living thing in general, and it got under your skin. About the number in parentheses on the screen — the thinking time — he explained it in a way I remember word for word:

«Actually higher number in ( ) may be better, it shows how much time Anna spends to decide that she knows the answer. LLMs hallucinate instead. Being unsure longer may be better.» (Discord, 10.06.2026)

The entire seventh chapter of this book — in three lines of chat. A machine being taught not to answer fast but to doubt honestly. Against machines that don’t know how to stay silent. On June 16, at the height of the race, a participant asked him the question hanging over every one of us: thanks for the answer, boss — but what happens when we get to zero? The answer — four words: «Aigarth will be deemed finished.» (Discord, 16.06.2026) Twenty-five years of road — and the man calmly conjugates their finale in the future tense, like a train schedule. By the page you are now reading, that future tense has already become the past.

6-7-6

The score crawled down for weeks. Hundreds, then dozens. Each next step came harder than the last — evolution was eating up the easy wins and hitting the hard ones. By early July we stood at the approaches: 448... 233... 176... Numbers that today sound like call signs to those who were there. My computer was melting, and the electricity bill promised to be enormous.

On July 6 CfB posted the final build: «Speaking of energy optimization, use this version for Anna-7 instead of the old one» — a fresh Anna.exe, one hundred forty-six kilobytes. And the finishing-off began. Not one genius alone — dozens of machines across the world, picking up each other’s files, jointly squeezing out the last points.

And that same day — zero. All seven tests. CfB asked in chat directly: «Are ALL solved or not?» The answer came from the community: «All Zero». Everything solved. Completely. The chat exploded. Congratulations, fire emojis, confetti gifs. Someone exhaled ”gg those who worked on anna.” A participant wrote the most natural thing people write in such minutes: «congratulations on this milestone».

And here — the moment this whole chapter was written for. Maybe the whole book: «It’s not a milestone. It’s much bigger.» (Discord, 06.07.2026, 17:23)

The community is celebrating a victory, and the creator quietly corrects them: you have not grasped the scale. No waypoint on any road — something a category higher. He did not decode it. He never decodes — you know that about him by now.

Now look at the date. The sixth of July, twenty-six. 6—7—6. The number 676 is Qubic itself: the network’s 676 computors, 26 squared, a construction laid into the architecture from day one. The man who ”has no time machine” crossed the finish line on a date that spells out the signature of his own network. He himself noted it the next day, unruffled: «he celebrates Anna Zero Day on 6/7/6 which makes sense». Coincidence? After nine chapters you have earned the right to pick your own answer. Even the calendar plays on his side. Why Zero, Exactly To understand why the creator corrected ”it’s not a milestone,” you need to know what was hiding behind the number we were chasing. We thought we were chasing the error of a small network solving addition. And yes — that is what we were chasing. But that error was three proofs in one.

First: the network learned the task — zero error, perfect. Second, more important: it understood rather than memorized. Anna’s log has two columns — green (examples the network trained on) and yellow (examples it had never seen). With a crammer who memorized the answers, the first column falls while the second stands still. With Anna they fell together, neck and neck, at a correlation of 0.998. Once CfB pointed at a step where the not-yet-understood improved more than the understood — validation jumped by 201, training by 183 — and called it a ”gem.” This is the opposite of cramming: a machine that generalizes rather than memorizes. Understanding, visible right in the terminal, on your own screen.

But the main thing is the third proof, and it hides in a word almost no one connected to the race: energy.

Here is what I did not understand while I chased the number through the nights. For CfB, the question ”can a real AI be built at all” was closed long ago — yes, it can. What torments him is a different one. Today’s AI is built monstrously wastefully: the big models devour energy at scales that, at the limit, run into physically absurd figures — entire power plants for the sake of one model. And CfB says it straight:

«If AI is possible with current energy budget of the whole Earth then Aigarth will achieve it, we have all necessary cards for that in our hand.» (Discord, 23.02.2026) There is the real stake. Not ”can we build an AI,” but ”can we build one without demanding the energy of a star.” Aigarth had three walls. Generalization — taken. A universal learning algorithm — taken. The third and last remained: energy. And the race to zero that hundreds of us drove with our home processors was not a game of records. It was a living demonstration that the third wall is falling — that a mind fits inside a budget the planet can actually supply. That is why zero is three proofs in one number: the network learned, generalized, and fit the energy budget, all at once. And that is why, to ”congratulations on this milestone,” he answered ”it’s much bigger”: a milestone is a point along a road, and what got proven here is that the road itself is passable — on fuel there is enough of.

He gave it his own frame, and the frame removes any temptation to inflate the moment. Aigarth, he says, is the Manhattan Project: not ”another bomb,” but the discovery of an entire paradigm. And Anna is Little Boy, the first test. The first blast in the desert did not end the war. It proved the bomb was real. Zero proved Aigarth is real. The big products come later. What Actually Happened What did zero prove? It proved the paradigm works. That evolution — without backpropagation, without weights, without a blueprint, on home processors — can grow tissue that fully solves an assigned class of problems. The thing they laughed at — CPU instead of GPU, selection instead of training, a garden instead of a factory — worked, publicly, verifiably, before hundreds of witnesses with logs. Twenty-five years of road ran into a number, and the number became zero.

What zero did not prove — and here I will be firm, because CfB himself is firm. Zero is not the birth of a superintelligence. Two days after the celebration, with the chat already drowning in fantasies, he cooled it with his own hands:

«Even with ZERO Anna won’t have exact model, because some her replies are just accidentally correct guesses.» (Discord, 08.07.2026) The creator who has just taken the highest summit of his life is the first to explain what that summit does not mean. On the all-hands call three days after zero it was said officially and in caps: this does NOT mean artificial general intelligence has been achieved. A demonstration of the method is not the finale. Remember the naturalist who publishes his failures? Here he is again — refusing to inflate his success.

And that is exactly why the zero can be believed. Hype would have screamed ”we created AGI.” Science says: we proved the road leads onward — and walked on down the road. By July 9 CfB was already posting new files: an Anna that ”chews data with no delay”; an Anna for predicting real-world data; and the task he now called the main one: «How to make sure that the tech isn’t stolen and it’s only QUBIC holders who benefit from it». The garden bore its first fruit — and the gardener immediately took up the fence.

And one last thing about sobriety — about the date all of this is moving toward. Aigarth has a deadline: April 13, 2027. There is no whiff of mysticism here, though the temptation is great. It is exactly five years from the day real mining began on Qubic — the thirteenth of April, two thousand twenty-two. An anniversary rather than a prophecy. The deadline of the mind and the deadline of the secret he brought together into a single day. A man skilled at matching dates to meanings — remember, he closed out his contract under the name BCNext one day before the date on which Orwell’s «1984» begins — has here too laid his two chief promises on one square of the calendar. What is due to happen by that day, and what he will say then — you will learn in the final chapter.

Testimony

I saw how it was done. Not in a billion-dollar lab, not behind closed doors under NDA — in an open chat, where anyone could download the file and join in. Perhaps the greatest experiment of the decade ran interleaved with memes, ribbing, and ”wen pump.” Hundreds of ordinary people — traders, miners, students, me — drove the evolution of an artificial mind with their home processors, passing its maturing snapshots to one another like a relay baton. None of us fully understood what was growing in those files. Whether he himself fully understood — you read chapter six. Chaos — hundreds of scattered machines, random mutations, the din of the chat. Selection — the files with the best scores, surviving in the common pot. Emergence — the zero that flared up on the sixth day of the seventh month, as if the date itself had been bred by selection. Everything this book had said in words happened before my eyes as a process — and I was not a spectator, but one of its cells.

«It’s not a milestone. It’s much bigger».