Aby is not a language model. It is a structured reasoning architecture where every decision passes through multiple independent verification layers. It earns its knowledge. It catches its own mistakes. And it grows smarter from every real interaction.
Hexis
The Knowledge Map
HAL's capability ontology — every skill, its confidence score, and the real physics it's grounded in. The ground truth all other facets read from and write to.
Pseudos
The Honesty Checker
Before HAL acts, Pseudos checks whether each belief was actually earned through real verification or just assumed. Catches false confidence before it costs anything.
Phronesis
The Prioritiser
Ranks every open question and pending action by expected information gain. HAL pays attention to what matters most — not what arrived first.
Organum
The Farmer
Generates genuinely new knowledge. Forms a hypothesis, writes a verification test, runs it against real physics simulation, and promotes grounded beliefs. This is how Aby grows smarter from experience.
Leviathan
The Gap Hunter
Compares what HAL knows against what it should know. Finds structural gaps and proposes honest, discounted fills by analogy — never auto-trusted, always flagged for Organum to verify.
Aby's Dream
The Grounding Checker
Traces every proposed action back through the knowledge graph to real physics. If the chain breaks anywhere, the action is blocked. No ungrounded claims reach HAL's actuators.
Epagoge
The Rule Maker
Decides when accumulated evidence is strong enough to promote a pattern to a standing rule. One real counterexample blocks admission. Rules are earned, not assumed.
Paradeigma
The Analogy Engine
When HAL faces a new problem, Paradeigma searches for a structurally similar solved problem and proposes a transfer. Always for human review — never auto-applied.
All Together
Self-Evolving in Simulation
Every interaction in a virtual environment feeds Organum, which grows Hexis, which makes every subsequent decision more grounded. HAL gets smarter before it ever touches hardware.