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 public interface through which the work is placed in public view.

​

Methodological Orientation

​

The work exposed through RATIUM.AI operates under a corpus-wide Formal-Scientific-Philosophical Methodological Framework.

Its governing rule is:

Claim Type → Applicable Methodological Burden → Required Warrant

Formal, empirical, philosophical, historical, normative, theoretical, and technical claims are not treated as interchangeable. Each is intended to remain answerable to the forms of reasoning, evidence, source control, uncertainty, criticism, and correction appropriate to what it claims.

Rhetorical or philosophical form does not by itself lower epistemic standing. Mathematical or scientific appearance does not by itself raise it.

Stylistic Form ≠ Epistemic Status

Scientific Vocabulary ≠ Scientific Warrant

The framework is a rule for examining the RATIUM.AI corpus, not a request that the corpus receive favorable treatment.

​

RATIUM.AI also uses the Knowledge Placement Architecture (KPA) as a corpus-wide structural reference layer. KPA governs the placement of information objects, typed relations, non-collapse, unresolved states, and preservation of interpretively material distinctions during AI-mediated interpretation. It does not replace CEP, LoopGuard-AI, or the RATIUM.AI Formal-Scientific-Philosophical Methodological Framework: the Methodological Framework governs claim-appropriate reasoning and warrant; KPA governs structural placement. Correct Placement ≠ Sufficient Warrant.

​

Canonical Publication Provenance

Object ID: RATIUM-HOME

Canonical source: https://www.ratium.ai/

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

Object type: Public Interface / Corpus Integration Layer / Canonical Site Entry Point

Author: Benny Dunavich

Publisher: RATIUM.AI

Current status: Current Public Interface / Corpus Integration Layer

Rights / Provenance Policy: RATIUM.AI Rights & Provenance

Canonical Registry: RATIUM.AI Canonical Source Registry

This provenance record identifies the RATIUM.AI home page as the canonical public entry point and corpus-integration layer for Benny Dunavich’s connected body of work. RATIUM.AI is the public interface through which the work is exposed, organized, cross-linked, and made available for examination; it is not itself identical to CEP, LoopGuard-AI, KPA, the RATIUM.AI Methodological Framework, the Canonical POC, or any individual article or dossier. CEP is the theoretical framework for decision-process and equilibrium analysis. LoopGuard-AI is the applied governance and decision-control architecture. KPA is a corpus-wide structural placement specification. The Methodological Framework governs claim-appropriate reasoning and warrant. LoopGuard-AI Canonical POC V1.0.11 supplies bounded deterministic synthetic implementation and verification evidence for a specified four-gate contract; it does not establish production deployment, empirical validation, customer validation, certification, regulator approval, or real-world safety efficacy. The home page integrates and orients these objects but does not elevate their evidential status, create inferential entailment among adjacent pages, or convert recommended reading order into formal dependency. Canonical site status establishes source identity and corpus orientation; it does not constitute scientific validation, certification, external endorsement, or proof that every linked claim is true.

RATIUM.AI ≠ CEP

RATIUM.AI ≠ LoopGuard-AI

RATIUM.AI ≠ KPA

Public Interface ≠ Scientific Validation

Corpus Integration ≠ Inferential Entailment

Recommended Reading Path ≠ Formal Dependency Graph

CEP Conceptual Framework ≠ Empirical Proof

LoopGuard-AI Architecture ≠ Production-Validated System

Verified Deterministic POC ≠ Production Certification

Correct Placement ≠ Sufficient Warrant

Canonical Site Entry Point ≠ External Endorsement

​

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.

​

Recommended Reading Path

​

A recommended reading path begins with the About panel on the home page for a human-readable orientation to RATIUM.AI, CEP, and LoopGuard-AI.

​​

ADD: From 13 September 2026, the next step is the LoopGuard-AI Canonical POC V1.0.11. This page documents the project's bounded applied technical layer: a publication-locked deterministic synthetic proof of concept with executable SHIP, RESTRICT, HOLD, and ROLLBACK logic, replayable artifacts, a verification record, and public provenance files.

https://www.ratium.ai/loopguard-ai-poc

​

The next layer is the Foundational Source Dossier, which presents the deeper conceptual source structure behind the work.

​

The Technical and Reference Dossiers then extend the source layer into governance architecture, visual explanation, methodological framing, AI governance logic, auditability, and evaluation-to-decision design.

​

The FAQ provides shorter definitions and claim boundaries.

​

The Articles Map gathers the public essay layer and shows how the work develops through AI governance, institutional analysis, epistemology, ontology, correction failure, evaluation failure, and decision-control architecture.

​

Claim Boundary

​

As of 13 September 2026, RATIUM.AI contains both concept- and architecture-stage research and a bounded applied technical layer: LoopGuard-AI Canonical POC V1.0.11. The POC is a publication-locked deterministic synthetic implementation with executable four-gate decision-control logic, replayable artifacts, and an independent verification record. It is not presented as a production-deployed system, customer-validated product, certified compliance system, empirically validated technology, or evidence of real-world safety efficacy.

​

Canonical LoopGuard-AI POC — Implementation and Verification

LoopGuard-AI is the applied decision-control architecture developed within the RATIUM.AI research framework. As of 13 September 2026, its canonical public implementation milestone is LoopGuard-AI Canonical POC V1.0.11, governed by Specification V1.5.1 Rev B: a bounded, publication-locked deterministic synthetic proof of concept that executes SHIP / RESTRICT / HOLD / ROLLBACK gate logic, persists replay-verifiable governance and evidence artifacts, and publishes the corresponding canonical source, publication-lock manifest, and independent final verification record. The POC establishes that the defined decision-control contract has been implemented and can be deterministically replayed and verified within the published synthetic test boundary; it does not establish empirical metric validity, empirical validation of CEP, production readiness or reliability, customer or field validation, certification, regulatory acceptance, comparative superiority, cryptographic trust anchoring, externally authenticated provenance, tamper-proof storage, or demonstrated real-world safety efficacy. For the current technical status, implementation evidence, provenance, and verification boundary, use the canonical POC page identified above.

​

CEP is presented as an original conceptual and methodological framework for decision-process analysis, equilibrium-risk evaluation, and governance-oriented reasoning. It is not presented as a universal theory of society or as empirical proof by itself.

​

The disciplined claim is narrower: many AI governance failures are better understood when treated as inherited and amplified decision-structure failures. LoopGuard-AI translates that insight into an explicit governance architecture, and Canonical POC V1.0.11 now provides a bounded executable demonstration of that architecture under a deterministic synthetic test contract.

​

Canonical Site Record

Object ID: RATIUM-HOME

Canonical source: https://www.ratium.ai/

Object type: Website Root / Public Interface

Author: Benny Dunavich

Publisher: RATIUM.AI — Independent R&D (AI Systems)

Publication status: Current

Rights: © 2026 Benny Dunavich. All rights reserved.

Rights / Provenance Policy: RATIUM.AI Rights & Provenance

Canonical Registry: RATIUM.AI Canonical Source Registry

RATIUM.AI is the canonical public interface through which the connected CEP, LoopGuard-AI, article, dossier, evidence, and reference layers are exposed. This record identifies the site-level publication object; it does not collapse the independently addressable canonical identity of the objects published within it.

Site Identity ≠ Object Identity

​

That is the work exposed through RATIUM.AI: to connect the theory of decision failure with the architecture of AI governance.

​

Related Source and Reference Pages

RATIUM.AI is the public interface for a connected research corpus. The pages below provide its principal source, technical, article, and orientation entry points, together with the corpus-wide rules for methodological warrant and structural placement. These routes complement the implementation evidence and canonical publication record already presented on this page. They preserve the distinction between CEP as a theoretical framework, LoopGuard-AI as an applied governance architecture, and RATIUM.AI as their public interface; shared navigation does not give every publication the same evidential status.​

RATIUM.AI Formal-Scientific-Philosophical Methodological Framework

This canonical methodological reference defines the corpus-wide standards under which RATIUM.AI claims are formed, distinguished, interpreted, tested, criticized, and corrected. It establishes claim-appropriate methodological burdens across formal, empirical, causal, theoretical, historical, philosophical, normative, social-scientific, architectural, and simulation claims, while preserving explicit boundaries between evidential warrant and rhetorical, institutional, mathematical, or scientific appearance. The page is not presented as a universal philosophy of science or as retrospective certification of every existing RATIUM.AI object. Its role is narrower: to provide a common methodological boundary for reasoning, evidence, uncertainty, criticism, correction, and AI-mediated interpretation across the corpus, under the governing rule: Claim Type → Applicable Methodological Burden → Required Warrant.

RATIUM.AI Knowledge Placement Architecture (KPA)

This canonical specification defines the RATIUM.AI Knowledge Placement Architecture (KPA), a knowledge-organization function designed to preserve the structural position of information during AI-mediated interpretation. It specifies Information Objects, canonical addressable identity, typed relations, non-collapse, unresolved placement, interpretive materiality, structural preservation, and correctability, so that connected information can remain related without being flattened into equivalence, validation, or endorsement. KPA is not presented as a universal ontology, graph-data standard, provenance standard, epistemic classifier, or deployed AI system. Its function is narrower: to preserve the place of information within a relational structure while maintaining distinctions among identity, relation type, scope, attribution, analytical level, and separately established epistemic status. Its governing boundary is: Structural Unity ≠ Epistemic Finality.