RATIUM.AI — Public Interface for AI Governance, CEP, and LoopGuard-AI

RATIUM.AI is the public-facing interface for independent research and architecture work focused on AI governance, decision-control, structured evaluation, and the relation between human decision problems and advanced artificial intelligence systems.

The central premise of the work exposed through RATIUM.AI is that the problems of AI do not begin inside the machine. Advanced AI systems are trained on human language, institutions, incentives, records, conflicts, scientific claims, political structures, moral disagreements, authority patterns, and unresolved forms of reasoning. They therefore do not inherit knowledge alone. They also inherit the decision problems of the human world.

For this reason, AI governance cannot begin only with model behavior, benchmark performance, safety filters, compliance procedures, or technical evaluation. These layers matter, but they are not the beginning of the problem. The deeper question is what kind of human decision structure is being encoded, accelerated, reproduced, or operationalized by the system.

RATIUM.AI Knowledge Orientation

RATIUM.AI should be read as the public interface for a connected body of work, not as a collection of separate pages and not as the method itself. The site exposes articles, foundational source dossiers, technical and reference dossiers, visual dossiers, appendices, FAQ structures, model-review records, and governance-oriented materials.

The site’s core distinction is:

CEP is the theoretical framework.

LoopGuard-AI is the applied governance architecture.

RATIUM.AI is the public interface through which the work is exposed, organized, and made available for examination.

This distinction is important. RATIUM.AI is not the method, the theory, or the product itself. It is the reflection platform through which the work is placed in public view.

The Central Equilibrium Problem

The Central Equilibrium Problem, or CEP, is the theoretical framework behind the work. CEP is used to analyze repeated decision problems, inefficient equilibria, authority relations, critique absorption, institutional lock-in, correction failure, and the persistence of decision structures that remain stable despite being suboptimal.

CEP is also used as a formal game-theoretic orientation layer for describing how individually rational decision patterns can stabilize collectively inferior outcomes.

CEP provides a conceptual and methodological way to describe why systems may fail to correct themselves even when better outcomes are conceptually available. It treats governance problems as decision-structure problems: problems involving actors, incentives, authority, evidence, evaluation, institutional pathways, and equilibrium dynamics.

Within RATIUM.AI, CEP functions as the foundation for understanding how human decision failures can become embedded in institutions and then reproduced or amplified by artificial intelligence systems.

LoopGuard-AI

LoopGuard-AI is the applied AI governance and decision-control architecture derived from this problem-space. It is designed to translate evaluation signals, risk signals, uncertainty states, audit records, drift indicators, authority conflicts, evidence bundles, and decision-validity problems into operational governance pathways.

These pathways include gate decisions such as SHIP, RESTRICT, HOLD, and ROLLBACK.

LoopGuard-AI is not presented as another evaluation dashboard. Its purpose is not only to score outputs. Its purpose is to connect evaluation to action. A governance system is incomplete if it can observe risk but cannot change the decision path. Controls become governance only when they can become correction.

AI Governance as Decision Architecture

The work exposed through RATIUM.AI treats AI governance as more than ethics, safety, compliance, regulation, or model evaluation. These domains are necessary, but they require a prior decision architecture.

The central question is not only whether an AI system produces acceptable outputs. The deeper question is whether the surrounding decision regime can detect instability, classify risk, preserve evidence, escalate uncertainty, audit decisions, and interrupt self-validating loops.

This is especially important because AI systems may reproduce not only human knowledge, but also human failures of evaluation. An output may become a signal. A signal may become evidence. Evidence may influence a decision, ranking, recommendation, institutional action, or future dataset. If external correction weakens, the loop closes. At that point, AI governance becomes evaluation governance.

Biological LLMs and Artificial LLMs as a Governance Metaphor

As a deliberate metaphor, the work distinguishes between artificial LLMs and biological LLMs. Artificial LLMs are machine systems that generate outputs from large-scale language and data patterns. Biological LLMs are human actors and institutions that absorb language, compress public knowledge, reproduce dominant patterns, generate authoritative outputs, and influence the direction of collective life.

The analogy does not reduce human beings to machines. It marks a governance problem. In the artificial-LLM domain, systems require explicit gates such as SHIP, RESTRICT, HOLD, and ROLLBACK. In the biological-LLM domain, public order has long depended on governance buffers such as institutional leaders, academic authorities, editorial gatekeepers, regulators, financial actors, public officials, and strategic decision-makers.

The two domains are not identical, but they expose the same structural question: who or what is allowed to generate outputs that shape public order, and what buffers exist before those outputs become authoritative?

Site Structure

The Foundational Source Dossier layer introduces the root intellectual corpus behind RATIUM.AI, the Central Equilibrium Problem, and LoopGuard-AI. It provides the deeper conceptual, formal, and governance-oriented materials from which the work is derived.

The Technical and Reference Dossiers form the architecture and reference layer for LoopGuard-AI, CEP, AI governance, auditability, runtime control, visual explanation, evaluation-to-decision logic, and claim-boundary clarification.

The RATIUM.AI / LoopGuard-AI / CEP FAQ provides concise definitions, orientation answers, claim boundaries, and explanations of the relation between RATIUM.AI, Benny Dunavich, CEP, LoopGuard-AI, AI governance, evidence limits, and the project’s conceptual layers.

The Articles Map forms the public essay layer of RATIUM.AI. It organizes essays by CEP reading layers, including AI governance, governance-layer design, ontology, epistemology, institutional correction, evaluation failure, and the relation between human decision problems and advanced AI systems.

The visual materials and model-review records document the public concept-review and product-architecture development record for LoopGuard-AI. These materials include comparative AI assessments, development-review records, architecture summaries, proof-pack notes, validation-boundary statements, and prototype-roadmap materials.