Research / publication
Gergely Vámossy
Work on AI/LLM governance, epistemic infrastructure, mathematical ontology, and machine-checkable methods that let a model reason without letting it rule. This site is a research publication platform: each artifact has a permanent URL, crawlable HTML, and a link to its canonical download where one exists.
Research focus
- AI/LLM governance
- Epistemic infrastructure
- Mathematical ontology
- Machine-checkable governance
- AI-assisted reasoning
- Human authority and validation
The method published here governs authority, not correctness. It keeps adoption separate from validation, reports unresolved dependencies in the open, and treats human signature slots as load-bearing. Where a source marks a claim as a candidate or a recommendation, this site keeps that status.
Featured publications
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.
Reference · August 2026
The Mathematics Ontology Bible
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.
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.
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.