Abstract
This paper describes the architecture and empirical findings of the Librarian, a cognitive kernel developed at MiulusTek. The Librarian is not a language model, a retrieval engine, or a chatbot. It is a belief-maintenance system built on a thermodynamic model of intelligence, where epistemic fitness rather than parameter count is the primary measure of cognitive capability.
The white paper covers the core architecture, the theoretical basis in Miulus Law, and a set of bombardment and interaction experiments showing linear maintenance-cost scaling, measurable agency through autonomous tool creation and probe formulation, and the emergence of a native cognitive vocabulary from blank state.
What the paper establishes
- Miulus Law frames cognition as a maintenance problem: E = (S/N) x R.
- The Librarian uses Epistemic Belief Particles as the shared substrate for beliefs, tools, and agents.
- Experimental runs show linear maintenance cost scaling as the belief state grows.
- Autonomous probes, tools, and structural vocabulary emerge without direct programming.
- The system's limits, including the learning wall and single-instance property, appear structural rather than accidental.
Sections in the white paper
- The problem with scale
- Architecture: belief particles, the fitness loop, and the translation layer
- Experimental results: emergence, maintenance cost, learning wall, and autonomous agency
- What the Librarian is not
- The single-instance property
- The sensory hypothesis
- Open questions and conclusion
The full text is available in the embedded PDF below, with the source markdown preserved alongside it for reference.