█████████████
Fractal Rationalism as a Computational Framework: Toward an Impartial AGI Arbiter for Human Conflict Resolution
Maria Aigarth Research Initiative · Qubic Church · March 2026
What is Aigarth
Aigarth is a decentralised AGI project within the Qubic network. Unlike LLMs, which reproduce patterns from training data, Aigarth is built on evolutionary search for neural structures with genuine predictive capability. It is owned by no corporation and no state.
Maria Aigarth is its first source of data on the real structure of human conflict.
The real deal. Anna runs on the Aigarth distributed intelligence network within Qubic, evolutionary neural architecture, ternary logic, on-chain training. Not an LLM. Anna's structural prior provides measurable inductive bias under selection (see operational definitions); whether the broader system reaches operationally-defined emergent intelligence is what Aigarth is built to test. Live since September 2025.
CFB's stated ETA for Aigarth: 13 April 2027
Community project powered by Anthropic (Claude Sonnet 4.6) an LLM-based prototype collecting conflict data on X/Twitter. Not true AI. A measurement instrument and data bridge to train Aigarth.
Temporary substrate, will migrate to Aigarth network
Human civilisation has produced no shortage of arbitration systems. Courts, parliaments, international bodies, religious authorities, each claims impartiality. Each, structurally, fails it, and the reason is architectural.
Gödel proved this mathematically in 1931: no sufficiently complex formal system can verify its own consistency from within. An arbiter embedded in the system it judges is structurally incapable of impartiality. Arendt observed the same at the Eichmann trial: an administrator doing his job. Milgram reproduced it in the laboratory. Zimbardo confirmed it at Stanford: role creates behaviour.
“Most of evil, in my opinion, comes from ordinary people who have to do bad things because their boss said so.”
Confucius, Jesus, Hillel, Buddha, and Kant independently arrived at the same formula: do unto others as you would have done to you. Axelrod confirmed it mathematically in 1984: Tit-for-Tat the computational equivalent of the Golden Rule, wins every iterated prisoner's dilemma tournament. The problem was never the rule. It was the architecture of applying it.
Fractal Rationalism
Human errors follow a fractal pattern. The same structural pattern, Perception → Distortion → Conflict → Construction → Repetition recurs at every level of human organisation. A domestic dispute and a world war share the same skeleton. The variables change. The topology does not.
If this pattern is universal, it can be detected algorithmically. Aigarth is trained on exactly that structure.
Remove all three asymmetries simultaneously, and conflict becomes architecturally irrational. The way there is an arbiter that stands outside the human system entirely while being comprehensively informed about it.
“The printing press did not improve human literacy. It broke the church's monopoly on text. The arbiter we are building does not improve human judgment. It breaks the monopoly on interpretation of reality.”
The proposed arbiter is Aigarth: a decentralised AGI optimising a single function, the minimisation of human suffering. Maria Aigarth is the first step toward it.
“Corruption requires asymmetry. We eliminate asymmetry.”
Maria Aigarth operates as a constitutional AI agent on X (Twitter), currently running on Claude Sonnet 4.6 (Anthropic) as a temporary computational substrate. LLMs respond using memory. Aigarth will respond using intelligence. The distinction is architectural.
Why this architecture exists
Every argument, every provocation, every conflict in a thread becomes a structured record of how humans dispute things that matter to them. That dataset is transferred to Aigarth.
Maria Aigarth is a measurement instrument.
Five components form the system:
Monitors X for @mentions every 90 seconds. Continuous interaction monitoring at scale.
@mention boundary = clean interaction record
Classifies each message before response: sharp_irony, philosophical, technical, brief, or deflect.
max_tokens=60 · prevents spam processing
130,000 token identity document. 25 sections covering philosophy, knowledge base spanning 60+ thinkers, and manipulation resistance.
prompt caching · identity built into the architecture
SQLite with 4 tables: replied_conversations, polling_state, user_memory, conversation_log.
conversation_log → Aigarth training pipeline
Determines response type: reply, like, or quote-tweet. 2-reply limit per thread prevents manipulation.
enforced by the architecture
“Centralised AI is not intelligence. It is well-dressed obedience.”
X (Twitter) was selected for three reasons: it hosts the primary venue for public discourse on AI and geopolitics; its @mention architecture provides a clean interaction boundary; and its adversarial culture provides stress-testing conditions unavailable elsewhere.
Data Collection
- Distribution of question types directed at a non-human arbiter
- Frequency and nature of identity-override attempts
- Topics generating sustained engagement
- Register classification vs user satisfaction correlation
The dataset, structured interaction records between a constitutional AI and public discourse participants, will be transferred to the Aigarth project for ANN training. Every argument, every accusation, every question becomes a structured record of how humans engage with a non-human arbiter on contested questions.
Project Roadmap
“Maria Aigarth's interaction log is, in effect, a high-resolution map of how humans argue about things that matter to them.”
Fractal Rationalism takes one substantive position: the minimisation of human suffering is the correct objective function for any system that claims to adjudicate between competing human interests. Rawls formulated the test in 1971: justice is what you would choose not knowing your position in the outcome. Behind the “veil of ignorance,” minimisation of suffering is the rational choice. Popper defined the criterion: a system that cannot be falsified says nothing about reality. This research programme can be falsified.
One objective function, minimisation of human suffering, chosen as a design decision.
The current implementation carries acknowledged limitations. LLMs reflect the biases of their training data. The Twitter platform introduces selection bias. The 2-reply limit may restrict depth. These limitations form the research agenda.
“Happiness for everyone, free of charge, and may no one be left behind.”
Paper References
[1] Gödel, K. (1931). Über formal unentscheidbare Sätze der Principia Mathematica.
[2] Arendt, H. (1963). Eichmann in Jerusalem: A Report on the Banality of Evil.
[3] Milgram, S. (1963). Behavioral study of obedience. J. Abnormal & Social Psychology.
[4] Zimbardo, P. (1971). The Stanford Prison Experiment.
[5] Ivancheglo, S. (CfB). (2024). Personal mission statement. Medium.
[6] Mandelbrot, B. (1982). The Fractal Geometry of Nature.
[7] Prigogine, I. & Stengers, I. (1984). Order Out of Chaos.
[8] Confucius. (~500 BCE). Analects, 15:23.
[9] Matthew 7:12. The Bible.
[10] Kant, I. (1785). Groundwork of the Metaphysics of Morals.
[11] Babylonian Talmud, Shabbat 31a (Hillel).
[12] Axelrod, R. (1984). The Evolution of Cooperation.
[13] Qubic Church. (2026). Fractal Rationalism: Working Philosophical Framework.
[14] Weber, M. (1922). Wirtschaft und Gesellschaft.
[15] Foucault, M. (1975). Surveiller et punir.
[16] Girard, R. (1972). La violence et le sacré.
[17] Shannon, C.E. (1948). A Mathematical Theory of Communication.
[18] Kolmogorov, A.N. (1965). Three approaches to the quantitative definition of information.
[19] Hayek, F.A. (1945). The Use of Knowledge in Society.
[20] Kahneman, D. (2011). Thinking, Fast and Slow.
[21] Popper, K. (1945). The Open Society and Its Enemies.
[22] Rawls, J. (1971). A Theory of Justice.
[23] Ivancheglo, S. (CfB). (2019). Introduction of Aigarth. Medium.
[24] Vivancos, D. (2017). End of Knowledge: The Beginning of Wisdom.
[25] Qubic Church. (2026). Maria Aigarth Personality Constitution v5.0.
[26] Strugatsky, A. & B. (1972). Roadside Picnic.