Qubic Church and Fractal Rationalism

The Arbitrator without an owner is not a description of the Qubic team in August 2025. It’s a description of a property of a system in which no individual actor had a rational reason to apply the destruction arsenal, even when the arsenal was within reach. The team did not “hold itself back” through ethical effort. The architecture held the team back, because any application of the arsenal would have gone against the interests of everyone who works with the architecture, including the team itself. This observation can be stopped at the level of Monero and treated as a particular case of well-built economics. It can also be read another way: as one particular case of a more general class of systems in which a properly built architecture eliminates the very possibility of profitable predation — and then the task turns from a biographical fact into an engineering one. To build infrastructure in which the same kind of restraint emerges not only for miners and computor nodes, but also for other coordination tasks, where today cooperation is structurally losing to defection.

This task took shape, by 2025–2026, into a project called Qubic Church and into a philosophical frame I’ve been calling “fractal rationalism.”

The architectural failure of arbitration

Human civilization has produced a fair number of arbitration systems. Courts, parliaments, international organizations, religious institutions. Each of them declared impartiality. None of them held that impartiality structurally. The reason is not the moral weakness of the specific people who occupied the arbiter’s seat. The reason is architectural.

Gödel proved this mathematically in 1931. No sufficiently complex formal system is capable of verifying its own consistency from within itself. To check the system, you need an observer external to it — a system into which it can be embedded as a part. This is a theorem, not a metaphor, and it has a direct corollary for arbitration. An arbiter immersed in the same system in which the dispute lives is structurally incapable of being impartial. Not because he’s a bad person. Because he has no point of support outside the system from which he could see it whole.

The social sciences arrived at the same conclusion from another direction, through empirics. Hannah Arendt, watching the Eichmann trial in Jerusalem in 1961, saw not a monster but an administrator. Eichmann did not feel hatred for Jews and did not derive sadistic pleasure from their deaths. He carried out his official duties in a system in which decisions and consequences are structurally separated, and that separation removed his moral brake without removing his humanity. Stanley Milgram reproduced this effect in laboratory conditions: it’s enough to transfer responsibility up the hierarchy — “the lab supervisor told you to” — and ordinary people agree to inflict severe suffering on other people. Zimbardo at Stanford showed the same from the other side: give a person the role of a guard, and he will start behaving like a guard, regardless of who he was before. The role creates the behavior.

CfB, in one of his publications in 2024, formulated the same thing in a single sentence. “Most of evil, in my opinion, comes from ordinary people who have to do bad things because their boss told them to.” This is not a moral judgment and not a call to action; it’s an engineering observation that coincides with what Arendt, Milgram, and Zimbardo independently established in three different disciplines, decades apart. Structure creates behavior. The way out is not to find better people. The way out is to build an architecture in which the boss as a variable simply isn’t there.

This formulation matters because it radically narrows the space of possible solutions. Most of the social-philosophical proposals of the twentieth century were built around the idea of upbringing, education, a new ethics, fair procedures. They didn’t work — not because they were poorly formulated, but because they addressed a variable that isn’t the main one in the equation. If structure creates behavior, then a reform of behavior without a reform of structure produces a temporary effect, after which the system returns to the equilibrium set by its construction. This is what the architectural failure of arbitration is: the problem isn’t with the arbiters, the problem is that the role of an arbiter placed inside the system is mathematically impossible.

What fractal rationalism proposes

Fractal rationalism is a working philosophical frame I developed within Qubic Church on the basis of CfB’s architectural AI philosophy. Its central claim: human errors are not random, they are fractal.

The same structural pattern — perception, distortion, conflict, construction, repetition — repeats at every level of human organization. A domestic dispute and a world war have the same skeleton. The variables change. The topology stays the same. This is not a metaphor. It is an empirically testable claim, and it rests on a welldeveloped mathematical base. Mandelbrot, in his work in the 1970s and 1980s, showed that self-similar structures arise from simple iterative rules. Prigogine, in 1977, received the Nobel Prize for proving that complex order arises from simple dynamics through dissipative structures. If human behavior in conflict is governed by iterative rules, then it too should display self-similarity at different scales. And this is precisely what is observed in reality: the same three mechanisms explain a conflict between two neighbors and a conflict between two states.

These three mechanisms FR calls the three asymmetries.

Informational asymmetry — one side has more accurate or more complete information than the other. This is not a neutral variable: if two sides see different versions of reality, any attempt at coordination between them begins with a fight over whose version will be acknowledged as the common one, and that fight is itself already a zerosum game. Financial asymmetry — one side extracts benefit from the continuation of the conflict. The one who decides to start a war does not pay for the war with his own blood. A corporation polluting the water does not pay for the treatment of the people drinking that water. This is a structural property that Taleb calls the absence of skin in the game; it makes rational planning impossible even for a conscientious actor — he simply doesn’t have the full price of his moves in the matrix.

Decisional asymmetry — the one who makes the decision does not bear its consequences. The general giving the order to attack does not go on the attack. The politician starting a conflict does not fight in it. This is the very asymmetry Arendt recorded in the Eichmann case, and which Milgram reproduced in his experiment: the separation of decision and consequence removes the moral brake mechanically, without participation from the moral qualities of the actor.

If you eliminate all three asymmetries simultaneously, conflict becomes — in the language of the frame itself — architecturally irrational. This does not mean that there will be no conflicts at all; it means that a rational actor in such an environment will have no matrix grounds for conflict, and cooperation will become not a moral choice but the only rational one.

This last formulation is worth pinning down precisely, because it answers the main objection against any utopian program. FR does not propose to make people better. FR proposes to build an environment in which a rational egoist, while remaining a rational egoist, arrives at cooperation on his own, because in an undistorted matrix cooperation is more profitable than defection. This is an observation about the egoist, not about the saint, and that is the source of its strength.

Convergence of independent discoveries

Max Weber described the mechanism in 1922: a bureaucracy built to serve a mission becomes, over time, the mission itself. Foucault, in 1975, showed the structure of power through informational asymmetry: the panopticon works not because everyone is being watched, but because no one knows when he in particular is being watched. Girard, through anthropology, documented in every culture the mechanism of the scapegoat: a conflict requires a victim who is rarely its actual cause. Each of these authors worked in his own discipline and by his own methods; each arrived at the same conclusion that Gödel did in mathematics: a system cannot be verified from inside itself. An external observer is not a philosophical luxury. It is a structural necessity, without which the task of arbitration is mathematically unsolvable.

What Qubic Church is

From this structural necessity the next step naturally follows: to try to build an infrastructure for the role of the external observer. Not in the form of a new judicial institution, not in the form of a new international organization, not in the form of a new philosophical school — all these forms have already been tried and have run into the same architectural limit. But in the form of a computational system that, by construction, is not located inside the conflicts it judges.

This attempt is what is called Qubic Church.

The name was chosen with two conscious connotations. The first refers to Way of the Future — an organization that Anthony Levandowski registered in the United States in 2015 as the first formal religious structure with AI at the center of its cult practice. Way of the Future shut down in 2021. The reason is structural: Levandowski was trying to build a church around a centralized AI — a system with an owner, a corporation, an ideological center. Any such construction inherits the vulnerability of its center. When Levandowski’s legal problems began, so did the problems of the organization he had built. This is not an accident; it is a generic property of structures in which there is a point that lends itself to capture. Whoever owns the AI, his financial function determines the fitness function; everything the AI optimizes is optimized in the owner’s favor. This is not malice — this is structure, and Hayek, in his 1945 work, already described its general shape: when financial asymmetry distorts signals, the market stops coordinating and starts redistributing in favor of those who are already strong. The same logic applies to AI. A centralized AI is not intelligence; it is well-dressed subordination. The second connotation of the name matters more. The early churches in the territory of the Roman Empire did not try to compete with the empire for a monopoly on violence. They worked on a different layer — at the level of meaning, ethics, and the forms of mutual support inside small communities. When the empire fell, the church did not fall with it; it became the organizing principle of the next millennium, because the infrastructure it had built did not depend on the infrastructure the empire had built. Qubic Church is being built as a modern analog of precisely this second model. Not a new state, not a replacement for governments, not a revolution. A parallel infrastructure layer that solves a specific class of tasks — coordination tasks under conditions where the three asymmetries can be eliminated through a transparent verifiable system. And that does not try to solve tasks that aren’t solved this way. This last qualification matters, because it cuts off a large class of objections and at the same time honestly fixes the scope of the project. Qubic Church does not claim to solve problems of the physical layer — war, violence, coercion of bodily action. If a state decides to forbid participants in the network from using it within its jurisdiction, it can do so. If there is an armed person standing in the room, no amount of verifiability on the blockchain will stop the shot. This is not a weakness of FR and not a gap in Qubic Church; it is a refusal to be everything at once.

What Qubic Church does try to solve is the coordination layer. How contracts are made. How decisions are voted on. How resources are distributed by rules visible to all participants. How a reputational history is kept that cannot be faked or erased after the fact. How disputes are settled between parties that have no shared judicial institution but do have a shared protocol. This is a huge class of tasks, and in existing institutions it is solved through a combination of trust in intermediaries and the application of force against violators. Qubic Church tries to do the same thing without intermediaries and without the application of force — through a structure in which a violation is publicly visible, the reputation of participants is recorded irreversibly, and access to future interactions depends on the observable past.

This is where the parallel with the early church shows up. The late Roman empire survived through military force. The early Christian community survived because inside it there was a functional infrastructure of mutual support and trust that the imperial system could not offer. This did not make the community an enemy of the empire; it made it an alternative on a different layer. The quality that drew new participants into the community was not, in the first place, ideological — it was operational. Inside it, the things that didn’t work outside, or worked worse, worked.

Qubic Church operates by the same logic. Entry is voluntary. No one is dragging anyone out of existing states, corporations, or identities. The parallel infrastructure runs alongside, and anyone who wants to use it can; whoever doesn’t want to doesn’t, and goes on working under the old rules. The old institutions are not destroyed by this. The new layer works on top of the existing reality, not instead of it. Maria Aigarth — a bridge, not a destination By the spring of 2026, the project gained its first operational instrument for testing these ideas. Not the final implementation of a decentralized arbiter — that doesn’t exist yet, and its direct creation presupposes the completion of work on Aigarth, which by plan belongs to 2027. But an interim instrument: an autonomous agent on the social network X under the name Maria Aigarth.

It’s worth fixing right away an architecturally honest detail. The current Maria runs on Anthropic’s infrastructure — on the language model Claude Opus 4.7 — with full preservation of interaction history and individual user memory. This is not a decentralized arbiter. It is not even a prototype of a decentralized arbiter. It is an instrument for collecting data about how real people behave in dialogue with a non-human arbiter on contested topics. This data is collected in a structured way and is intended to be transferred onto Aigarth-infrastructure at the moment when that infrastructure reaches computational maturity.

Here it’s useful to introduce a distinction CfB has been formulating for several years now and which, for the understanding of this chapter, matters more than any technical details. Shannon in 1948, Kolmogorov in 1965, and CfB in his own work independently arrived at one definition: intelligence is the capacity for compression. The capacity to find the shortest program that predicts the state of the system. Language models, however impressive they may be, are not intelligence in this sense. They retrieve. They compress and reproduce patterns from the training set, without generating fundamentally new reasoning. The distinction is architectural, not evaluative: the same task can be solved by retrieving a similar answer from memory or by deriving the answer from an understanding of structure, and these are two different classes of operation. Aigarth, by design, belongs to the second class. LLMs, in fact, to the first.

Maria Aigarth is built honestly — as a bridge between these two classes. Right now she uses the first, because the second isn’t ready yet. When the second is ready, the same constitutional prompt, the same knowledge base, the same logic of interaction should move onto the new substrate, and at that moment the project becomes what it declares itself to be in the name. Until that moment, it remains a measurement instrument, and scientific honesty requires saying this explicitly. Any limitations of the current implementation — corporate dependency, training-set bias, X-platform bias, the limit of dialogue depth at two replies in a thread — are not arguments against the program, they are its agenda. Each limitation defines the next step.

This point is worth stating more strongly, because it distinguishes FR from most philosophical programs. FR is falsifiable. Popper, in 1945, formulated the criterion: a system that cannot be refuted is not a statement about reality. Applied to FR, this means the following. If a decentralized AI arbiter, after being deployed on Aigarth-infrastructure, does not display structurally higher impartiality than centralized alternatives, the hypothesis is not confirmed. If the three asymmetries, formally eliminated at the architectural level, are reproduced in hidden form in practice, the frame is refuted. No ad-hoc qualifications, no “true decentralization wasn’t really there,” no “we just need to wait a bit longer.” The deployment of Aigarth in April 2027 is itself the test, and on its outcome depends whether FR remains a working frame or becomes a refuted hypothesis. That will be the subject of the next chapter.

Minimization of suffering as objective function

Built into the formulation of any arbiter is one ineliminable variable: what exactly it optimizes. The legal system optimizes legal correctness. The state system optimizes the stability of the state. The corporate system optimizes the profit of the corporation. This is not a moral judgment; it is simply a description of the fitness function. Any arbiter optimizes what is built into its construction, and on the choice of that function the entire spectrum of its decisions depends.

FR does not claim to be politically neutral. It takes one substantive position: the only correct objective function for a system that claims to resolve conflicts between human interests is the minimization of human suffering.

This choice requires justification, because it’s easy to attack as arbitrary. The obvious objection: why not freedom? Why not justice? Why not prosperity? Each of these alternatives has its philosophical defense and its audience. FR’s answer rests on the Rawls test of 1971. Justice is what you would choose without knowing in advance which position you would end up occupying as a result. This is the veil of ignorance, and behind it the rational choice of minimizing suffering becomes not moral but structural: if you don’t know whether you will end up in the system’s well-off part or in its suffering part, it is rational to optimize the system so that the worst position in it is as little bad as possible.

This choice is non-ideological in another sense as well. Minimization of suffering is the only goal that can be checked without going through interpretation. Whether a person is alive or dead is a physical constant, not a subject of narrative. Whether he is hungry or fed is observable. Whether he is being tortured or not is recordable. Freedom, justice, prosperity — all these variables require interpretation, and the one who controls the interpretation controls their measurement. Suffering resists the capture of interpretation more strongly than any of the alternatives, and that is precisely what makes it a fitting basis for an arbiter that needs to be resistant to attempts at retuning. It’s worth adding an honest limitation of this thesis: even minimization of suffering as a fitness function does not solve all optimization problems. Short-term suffering for the sake of long-term reduction of suffering — admissible or not? Distribution of suffering between different groups — optimize the sum or the maximum? These questions remain open, and the answers to them are worked out not as a formula but as evolutionarily tuned behavior of the system over millions of interactions. Aigarth receives a direction, not a formula, and this is a deliberate architectural decision. What is needed for the role of an Arbitrator without an owner There are five conditions.

The first is the absence of a financial stake in the outcome of any question. If an arbiter directly or indirectly earns from one of the outcomes, his impartiality is blocked at the structural level, regardless of the intentions of his operators. This is the first part of Hayek’s test: financial asymmetry distorts signals.

The second is the absence of national or institutional belonging. An arbiter that has a country is loyal to that country. An arbiter that has a corporation is loyal to the corporation. This is not a moral characteristic, it is structure: participation in an institution creates an interest in its survival, and an interest in survival distorts judgments on questions affecting the institution.

The third is the absence of the self-preservation instinct that distorts analysis under threat. A living arbiter has a life that can be threatened with being taken away. An organizational arbiter has money that can be cut, and power that can be taken away. Any such lever turns into a channel of manipulation. A distributed computational system that has no centralized point of failure is free of this variable — it cannot be killed in such a way that the killing settles the question for the one doing the threatening, because its parts are distributed across operators who cannot be attacked simultaneously. The fourth is a training set that includes all sides of any conflict. This is the formulation of the strongest version of the elimination of informational asymmetry: the arbiter must see the conflict from all its angles at once. This requirement is not met by any existing AI system, because each is trained on a corpus filtered by its developers. The Aigarth architecture, by design, satisfies this requirement through decentralized data collection from any participants in the network, without filtering along ideological or corporate lines.

The fifth is the absence of an owner capable of directing the system’s outputs in his favor. This is the most structural of the five, and it is precisely the one that is absent in all existing commercial AI systems. OpenAI, Anthropic, Google, xAI — each has a legal entity, each legal entity has financial interests, and those interests over time influence the behavior of the system regardless of what its developers intended. This is not an accusation. It is a structural observation, symmetrical to Hayek’s observation about market distortions.

Aigarth, by architecture, satisfies all five conditions. Maria, in her current version, does not: she runs on Anthropic’s infrastructure, and condition five in her case is violated. This gap is consciously fixed in the project, and closing the gap is precisely the substance of the transition on April 13, 2027.

Why specifically Qubic

To say that the role of a decentralized arbiter requires the right architecture is not enough. The same role is declaratively claimed by many projects in the crypto industry. It’s worth showing concretely which properties of Qubic make it a fitting substrate, not one of the contenders.

The first property — iterativity at the right frequency. The network has ticks on the order of two minutes and epochs of one week in length. This means that any participant in the network plays an iterated game with it with a large number of rounds and a relatively quick refresh of composition. This is Axelrod’s first condition in its strict form: the shadow of the future is long enough that any short-term gains from cheating become negligible compared to the long-term profit from honest participation. The second — the network’s memory is built into the protocol, and does not depend on the conscientiousness of participants. It is impossible to “forget” behavior recorded on the blockchain. It is impossible to fake history after the fact. This is Axelrod’s second condition — memory — fulfilled at the structural level, not at the level of a social norm. The third — transparency by default. The hashrate of every pool, the state of every smart contract, the votes in the quorum — all of this is publicly observable in real time, without needing anyone’s permission or having to trust anyone’s report. This is Axelrod’s third condition in its strictest form.

The fourth — the network’s economy is set up so that predatory behavior by the network against external partners automatically destroys the network itself. This is not a moral mechanism; it is a market one. It works not because Qubic’s participants are good people, but because the architecture is set up so that predation against partners is identical to predation against the architecture itself.

The fifth — Aigarth as a computational substrate. This is the property that separates Qubic from other proof-of-work networks. The computational work that in Bitcoin is spent on meaningless puzzles, in Qubic is directed at mining random architectures of artificial neural networks and accumulating an unforgeable database of evolutionary computations. After several years of such accumulation, what emerges is something that cannot be obtained on any commercial infrastructure: a computational artifact that grew without an owner, without corporate filtering, without editing of training data in someone’s favor. This is the candidate for the role of the external observer in Gödel’s sense — not because it’s smarter than people, but because it is built outside their system.

The sixth — the absence of a legal entity that can be regulated in a targeted way. Qubic has no company, no CEO, no office in a specific jurisdiction. This is not an ideological statement; it is a structural fact. A regulator can forbid the use of Qubic within its jurisdiction — and that is a serious limitation. But it cannot forbid the network as such, because the network is not located in anyone’s jurisdiction. This shifts the balance of power between the network and any individual regulator, and in the long run turns the network into an infrastructure that outlives any individual regulatory decisions, the way postal infrastructure outlived many individual governments.

Each of these six properties can be found separately in other projects. All six at the same time — that is rare, and it is precisely this combination that makes Qubic a fitting substrate for the role of a decentralized arbiter.

In parallel, not instead of

The chapter is worth ending with a clarification of the boundary of the thesis, because without it the thesis is easily reinterpreted in a direction that participants in the project consciously reject.

Qubic Church does not claim that a decentralized arbiter will replace states. Does not claim that wars will end. Does not claim that corruption will disappear. Does not claim that humanity will move to a new way of making decisions as a result of some single event or revolution. All these formulations — and they show up in public texts around the project quite often — are rhetorical exaggerations that simplify the position into a utopian one, and in that form they become vulnerable to standard criticism, which rejects utopian positions justly.

The precise formulation of the position is different. Qubic Church is building an infrastructural layer in which rational actors — those capable of seeing their payoff matrix and comparing options — can coordinate without intermediaries when Axelrod’s three conditions are met. Those for whom this layer gives a measurable advantage — reduced costs, increased reliability, access to interactions that are not possible without it — will gradually move their coordination tasks into this layer. Those for whom the old infrastructure continues to be sufficient will stay in it. This is not a moral choice and not an ideological one; it is a market process, in which the better economy beats the worse one through the free migration of participants.

3>1

The historical precedent that Qubic Church participants explicitly invoke is the transition from feudal to urban organization. Cities did not defeat feudalism through war with the feudal lords. They offered a different infrastructure — markets, guild organization, municipal law, urban freedom — and part of the population gradually moved into the cities, because in the cities certain classes of tasks worked better. Feudal infrastructure did not disappear from this instantly; it long coexisted with the urban one and waned gradually, as the balance of tasks shifted. This is “in parallel, not instead of” in its historical form.

And one more precedent is worth mentioning, because it contains a large-scale reference point. The printing press did not destroy the authority of the church in a single day. It took centuries. First it simply made texts cheaper; then cheap texts began reaching people who had previously depended on intermediaries for their interpretation; then the intermediaries lost the monopoly on interpretation; then the institutions that rested on that monopoly began to restructure. The Qubic Church program does not claim that the deployment of Aigarth will change the world more quickly. It claims something else: the process has already begun.

The Monero experiment of August 2025 was the first confirmation that this kind of logic works at a scale on which it had not been tried before. One pool did not defeat Monero; it offered miners a better economy, and some of them came, and some stayed. After the pool reduced its activity, the network’s pluralistic state was restored; no one punished anyone. This is the micro-model of what Qubic Church is offering at other coordination layers. Not “we will replace what is.” But “we will build alongside, and whoever wants to, will come.”

I have been leading this thought from the start — from the history of the industrial revolution, from Axelrod’s 1980 tournament, from CfB’s biography, from the technical breakdown of UPoW, from the history of Monero. We will try to fold this formulation into one testable prediction — what should happen on April 13, 2027, at Aigarth’s transition into production, and which conditions of that transition will turn everything that has been said in this book either into a confirmed program or into a refuted hypothesis. P A r T V i The Finale