Index
Research / Publications
Each item has a permanent URL and can be read as HTML. Canonical files are linked where they exist. Future papers can be added to this index without changing the URL scheme.
Paper · 13 August 2026
Non-Self-Approving AI-Assisted Derivation: Letting a Model Reason Without Letting It Rule
A method that lets an AI system produce analysis while remaining structurally unable to produce authority. Five composable components keep conclusions as recommendations in front of a human, never rulings. The method governs authority, not correctness.
Read online · Download PDF · /research/non-self-approving-ai-assisted-derivation
Reference · August 2026
The Mathematics Ontology Bible
A Complete Map of Mathematical Structure, from Foundations to Frontiers
A canonical ontological map of mathematics: objects, structures, checkability, dependencies, and epistemic status from logical foundations to the governance layer. Version 1.0. The typeset Word document is a canonical form.
Read online · Download Word · /research/mathematics-ontology-bible
Framework · August 2026
QIERA — An Epistemic-Governance Overlay for LLM Workflows
A practical overlay for keeping large-language-model reasoning honest, non-agentic, and auditable. QIERA wraps thinking, not doing: it produces artifacts, not actions. Positioned as applied synthesis, not novel theory.
Read online · /research/qiera-epistemic-governance-framework
Paper · August 2026
Keeping Claims Honest: A Deterministic Toolkit for the Gap Between Proxy and Truth
A family of small, deterministic, self-testing tools built on one discipline: separate the proxy from the truth and notice when they come apart. The paper presents operability, not a claim of new science.
Read online · /research/keeping-claims-honest
Preprint · August 2026
Who Governs the Governor? Well-Founded, Human-Grounded Reflexive Governance for AI Systems
A preprint on well-founded, human-grounded reflexive governance: governance towers must bottom out at human authority rather than self-certify. Markdown rendition; the typeset PDF and LaTeX source are identified as canonical in the source.
Read online · /research/who-governs-the-governor
Preprint · August 2026
Recorded, Not Verified: Responsible Handling of Unverifiable Experience-Claims by Automated Systems
A preprint on responsible handling of unverifiable experience-claims. Where no third-person ground truth exists even in principle, the honest verdict is permanently unverifiable.
Read online · /research/recorded-not-verified
Preprint · August 2026
Forecasts Are Not Facts: Temporal Epistemic Governance for Automated Claims and Decisions
A preprint on temporal epistemic governance: forecasts are not facts, and tense must map to verifiability before a claim is treated as settled.
Read online · /research/forecasts-are-not-facts
Preprint · August 2026
Keeping the Evidence Honest: A Non-Self-Approving Integrity Layer for Evaluations-Based AI Safety Cases
A preprint specifying a non-self-approving integrity layer for evaluations-based AI safety cases: maturity grading, disclosed open issues, and overclaim auditing.
Read online · /research/keeping-the-evidence-honest
Paper · August 2026
A Theory of Everything's Checkability: Why No Theory Governs Everything, and the Discipline That Governs Any Claim
This paper is not a physics theory of everything. Its central result is that any such totalizing framework is an over-claim, and that the toolkit flags a literal theory of everything as ungrounded.
Read online · /research/theory-of-everythings-checkability
Essay · August 2026
The Reachability of the Truth: Edge Cases from Geometry to Consciousness
An expository essay — a perspective, not a research result. It offers no new theorem. Its contribution is framing: a single axis from exact geometric truth to permanently unverifiable experience-claims.
Read online · /research/the-reachability-of-the-truth
Study · August 2026
Governed Ontology Evolution: A Reproducible Simulation Study
When should a knowledge system add a new concept? A comparison of frozen, uncontrolled-drift, and evidence-governed regimes, with a reproduced baseline and sensitivity analysis.
Read online · /research/governed-ontology-evolution
Architecture · August 2026
The Governance Family — Architecture & Index
The map of a coherent family of applied AI-governance components: what each piece is, how they compose into a decision pipeline, the honest limits they share, and the safety boundary that applies to every one of them.
Read online · /research/governance-family-architecture
Design note · August 2026
Mapping Integrity: a governance layer for ontology mappings
A short design note. Ontology matching is a mature field. This note specifies mapping integrity — whether a mapping is trustworthy, and whether it still holds as the ontologies it connects evolve — over the SSSOM standard. It does not build another matcher.
Read online · /research/mapping-integrity
Design note · August 2026
postmortem_infra — governing the integrity of an incident postmortem
A short note accompanying postmortem_infra.py. It checks whether a postmortem is honest in structure — not whether its conclusions are correct.
Read online · /research/postmortem-integrity-infra
Study · August 2026
The Recursive Emergence of Money: A Formalization and Its Stress Test
Treating 'money emerges recursively from trust' as a precise dynamical claim and testing it against the Terra/Luna collapse of May 2022. A stylized model, not econometrics.
Read online · /research/recursive-emergence-of-money
Reference · August 2026
Persistent Jailbreak — Defensive Vector → Component Map
A one-page defensive reference. It classifies a persistent jailbreak by propagation mechanism and maps each vector to a toolkit component. Defensive scope only: no offensive techniques, no propagation recipes.
Read online · /research/persistent-jailbreak-defense-map
Case study · August 2026
Case Study: How a Non-Agentic LLM Produced a False Belief in Machine Autonomy
A de-identified, artifact-backed reconstruction of an AI-immersion episode focused on belief-formation pathways and intervention points. It documents a mechanism, not a personal crisis.
Read online · /research/ai-safety-case-study
Case study · August 2026
Case Study — Detecting a Real Decoupling: GDP per Capita vs Median Household Income
The decoupling and ground-truth tools run on published US economic data rather than a constructed series, doing on genuine data exactly what they do in the synthetic demos.
Read online · /research/case-study-gdp-income-decoupling
Case study · August 2026
Case Study — Auditing a Real Taxonomy: Normalizing Log Severity
The taxonomy engine pointed at unifying log-severity levels across three production logging systems, and the concrete defects its validate() surfaces in a mapping that would otherwise look fine.
Read online · /research/case-study-severity-taxonomy
Results · August 2026
Empirical Stress Test of the Governance Toolkit — Methods and Results
All 42 modules of the toolkit pass their own self-tests and are byte-for-byte deterministic across repeated runs. Results include randomized property trials and a diagnosed, then fixed, Expectation–Maximization shortfall.
Read online · /research/empirical-stress-test
Results · August 2026
Adversarial Stress Test of the Capable-Agent Cage — Methods and Results
The cage's decision space is small enough to enumerate completely. Across 6,144 proposals plus 100,000 randomized fuzz proposals there were 0 escapes and 0 disagreements with an independent oracle.
Read online · /research/capable-agent-cage-stress-test
Crosswalk · August 2026
Assurance-Case Integrity: A Crosswalk to the Safety-Case and AI-Safety-Case Literature
A positioning document locating the toolkit's integrity layer within existing safety-assurance-case literature. It is not an endorsement. Citation caveat: verify titles and IDs against primary sources before submission.
Read online · /research/assurance-case-integrity-crosswalk
Crosswalk · August 2026
Literature Crosswalk — Foundational Ontology Worked Example
Maps the non-self-approving derivation pattern's foundational-ontology example onto formal-ontology and metaphysics literature, separating genuine re-derivation from the method's packaging contribution.
Read online · /research/ref-literature-crosswalk
Crosswalk · August 2026
CEM/SAI Consciousness Crosswalk
Maps the high-consequence assessment example (machine consciousness, sentience, ASI candidacy) onto the science of consciousness and the AI-welfare literature. The real novelty claimed is the governance wrapper, not a theory of consciousness.
Read online · /research/cem-sai-consciousness-crosswalk