Skip to content

Intellectual Foundations — Cited Bibliography

The papers, books, prior art, and tooling that shape untool.ai's ontological approach. Each entry states what idea was taken from it and where it surfaces in the platform.

How this page relates to the rest of Research & References

The other pages in this section (Ontology Foundations, Ontology Stack, Persistence & Time, LLM & Agent Stack, Coordination & VFS, Generative Pipeline, Graph Visualization & UI) are analytical — they make trade-space arguments and lay out why we chose what we chose. This page is bibliographic — it is the citation table the analytical pages draw from, with direct links to the canonical source for each work. When you want the argument, read the analytical page. When you want the source, come here.

See also: the comprehensive Standards Index for normative specs (W3C, ISO, RFCs), which is normative-document focused rather than scholarly-work focused.


Foundational Ontologies

Guizzardi et al. — gUFO (2024)

A Lightweight Implementation of the Unified Foundational Ontology as an OWL Ontology Giancarlo Guizzardi et al. (NEMO-UFES) · arXiv:2603.20948 https://arxiv.org/abs/2603.20948

The practical OWL 2 DL implementation of UFO. A single Turtle file (CC BY 4.0), no runtime dependencies, native to the description-logic toolchain. This is the primary authoring and reasoning profile — every candidate fragment is classified against gUFO stereotypes before BFO grounding. The stereotypes (Kind, Role, Relator, Phase, Mode, Quality, Event…) give the ontology pipeline its type vocabulary.


Guizzardi et al. — UFO (Applied Ontology 2022)

Unified Foundational Ontology Giancarlo Guizzardi et al. · Applied Ontology journal https://journals.sagepub.com/doi/abs/10.3233/AO-210256

The modern statement of UFO as a descriptive, conceptualization-oriented foundational ontology — contrasted with BFO's prescriptive realism. The distinction between "how we model the world for systems" (UFO) and "how the world actually is" (BFO) is the core of ARC-ADR-039.


Guizzardi — Relations in Ontology-Driven Conceptual Modeling (2019)

Relations in Ontology-Driven Conceptual Modeling Giancarlo Guizzardi et al. https://www.inf.ufes.br/~gguizzardi/Relations_in_Ontology-Driven_Conceptual.pdf

Establishes relators as first-class endurants that mediate n-ary relations via typed roles. The relator is the ontological substrate for the hyperedge-as-vertex pattern (ARC-ADR-016): a relation with its own identity, lifecycle, and the ability to play a role in a further relator — a holon.


Smith — Beyond Concepts: Ontology as Reality Representation (FOIS 2004)

Barry Smith · Formal Ontology in Information Systems 2004

The canonical statement of BFO's realist commitments. Where UFO describes how we conceptualise the world, BFO carves reality at its joints — prescriptive, cross-domain interoperability for scientific and regulatory contexts. Cited in ARC-ADR-039 as the reason BFO is carried as the parallel profile rather than the primary authoring target.


Arp, Smith, Spear — Building Ontologies with Basic Formal Ontology (MIT Press, 2015)

Robert Arp, Barry Smith, Andrew D. Spear · MIT Press

The definitive methodology guide for BFO 2020 (ISO/IEC 21838-2). Continuant vs Occurrent classification, how to author definitions that are scientifically defensible, and why the realist discipline matters for long-term ontology maintenance. The standard reference behind every BFO design decision in the sift-sort validation ladder.


Meta-Cognitive Framing

Cabrera — DSRP Theory

Distinctions, Systems, Relationships, Perspectives Derek Cabrera · Cabrera Research Lab https://cabreraresearch.org/

An empirically grounded meta-cognitive framework whose Perspective primitive gives a rigorous basis for treating UFO and BFO as sovereign, complementary lenses rather than competing ontologies. Used in ARC-ADR-039 to frame the "Foundations as Perspectives" decision — both profiles are maintained in parallel because each illuminates things the other cannot.


Category Theory & Formal Bridging

Spivak — Functorial Data Migration (arXiv:1009.1166)

David I. Spivak https://arxiv.org/abs/1009.1166

Structure-preserving transformations between ontological schemas via functors. The formal basis for the UFO ↔ BFO projection: every concept can be mapped across foundations without lossy hand-mapping, provided the mapping respects the categorical structure. Cited in ARC-ADR-039 alongside the DOL / Ontohub institution-theoretic bridging line.


Leinster — Higher Operads, Higher Categories

Tom Leinster · Cambridge University Press

Operadic composition: the capability of a whole derives from the capabilities of its parts via a structured composition law. Cited in the Holonic Skill Acquisition vision note and ARC-ADR-033 for the type-theoretic claim that capability composition is a functor, not an emergent side-effect.


Reification & Hypergraph Modeling

W3C — Defining N-ary Relations on the Semantic Web (2006)

W3C Semantic Web Best Practices Working Group · W3C Note https://www.w3.org/TR/swbp-n-aryRelations/

The canonical justification for reifying n-ary relations as classes with binary role properties. This is precisely the hyperedge-as-vertex pattern: a relation becomes a node with its own identity and role edges. Direct reference for the ArcadeDB relator design.


Halpin — Object-Role Modeling Overview

Terry Halpin · ORM whitepaper https://www.orm.net/pdf/ORMwhitePaper.pdf

The n-ary role uniqueness constraint test: a valid fact type must require at least (n−1) roles to be uniquely identified. Used as a design-time validator when deciding whether to reify a relation — if the constraint doesn't hold, the reification is over-engineered.


Fowler — Bitemporal History

Martin Fowler · martinfowler.com https://martinfowler.com/articles/bitemporal-history.html

Two-axis temporal model: valid time (when the fact was true in the world) and transaction time (when it was recorded). Applied to reified relators so every edge in the hypergraph carries full audit provenance. Informs the PROV-O lineage design — "snapped, not plausible" is only meaningful if the snap event is timestamped on both axes.


HyperGraphDB — Iordanov (Springer 2010)

HyperGraphDB: A Generalized Graph Database Marko A. Iordanov · Transactions on Large-Scale Data- and Knowledge-Centered Systems III https://link.springer.com/chapter/10.1007/978-3-642-16720-1_3

Foundational academic work on hypergraph representation in database systems. Precedent for the relator-as-vertex structure and typed incidence maps used in the ArcadeDB schema.


TypeDB — The Case for a Structured Hypergraph

TypeDB Engineering https://typedb.com/blog/the-case-for-a-structured-hypergraph

Peer design demonstrating typed roles and nested reification in a production graph system. The closest prior art to the platform's "relation as a participant in a further relation" pattern — studied for both what to adopt and what to avoid (TypeDB's query language couples the schema too tightly to the storage engine).


Holonic Systems

Koestler — The Ghost in the Machine (1967)

Arthur Koestler

Introduces the holon — a whole that is simultaneously a part of a larger whole. The conceptual foundation for holonic skill acquisition: a reified relator has its own identity (a whole) and can play a role in a further relator (a part). Capabilities modeled on holons can be dynamically acquired at query time rather than loaded from a static pack.


Swarm Intelligence & Multi-Agent Consensus

Rosenberg — Collective Swarm Intelligence

Louis Rosenberg PhD (Unanimous AI / UNU / ENSO) https://www.unanimousai.com/

Empirical research on conviction-weighted swarm-aggregation in real-time multi-agent decision-making. Cited in ARC-ADR-044 as the empirical grounding for the Ontology-Orchestrated Swarm Intelligence vision — the idea that a purpose-driven swarm of ontologically-grounded agents can produce collectively superior judgements to any single model.


Prior Art — Platform Design

LinkML

https://linkml.io/

The closest match to the "ontology → projections compiler" vision: YAML-canonical model → generates JSON-Schema, Python dataclasses, SHACL, and documentation. Studied for the build-vs-invent decision. The generator model influenced the forge pipeline; the YAML-first authoring approach was ultimately too flat for the OntoUML relator requirements.

Palantir Foundry Ontology

Objects + Links + Actions as enterprise digital twin. Actions ≈ platform scenarios; evidence ≈ audit trail. The commercial prior art closest to the UDA/Object Model vision — studied to understand what Palantir got right (schema-enforced typed objects, ontology-driven UI) and wrong (proprietary lock-in, no formal foundation).

EnterpriseWeb

Model-as-runtime, hypergraph, dynamic composition, late binding. The strongest conceptual precedent for "one model, many projections" and runtime-derived behavior. Cited in the Codegen vs Interpreted vision note.

dbt (data build tool)

The "compiled, tested, documented artifact from a model" discipline. Generated SQL is disposable; the model is the asset. The direct inspiration for the *.g.cs / DO NOT EDIT discipline and the CI drift-gate pattern.


Active Research — Object Model Hill-Climb

2026-06-06 · Source: feat/collapse-backend-core branch

Microsoft — MAI-Thinking-1: Building a Hill-Climbing Machine (2026-06-02)

Microsoft AI Research

Frames iterative, evidence-gated improvement loops as a hill-climbing machine — each step produces measurable, reproducible evidence of improvement before the next step is permitted. Applied to the ontology authoring loop in backend-core: each harvest/sift/mapping run emits quality, dedupe, eval, quarantine, repair, and provenance evidence that gates the next run. The hill-climb framing makes "snapped, not plausible" a convergence guarantee, not just a one-shot gate.


Tools & Infrastructure

The concrete tools the ontology pipeline runs on. Each entry states what it does and where it sits in the validation ladder or runtime.


Ontology Files

Tool License Source Role
gUFO CC BY 4.0 https://github.com/nemo-ufes/gufo Primary OWL profile — authoring target for all fragment classification
BFO 2020 CC BY 4.0 https://github.com/BFO-ontology/BFO-2020 Parallel profile — scientific / regulatory interoperability projection
CCO (Common Core Ontologies) CC BY 4.0 https://github.com/CommonCoreOntology/CommonCoreOntologies BFO-grounded mid-level vocabulary used in the dual-grounding validation pass

Validation & Reasoning

Tool Language Source Role
rdflib + owlrl Python https://owlrl.readthedocs.io/ L3 seed reasoner — OWL 2 RL forward chaining; lowest-friction first proof target (ARC-ADR-019 spike)
pyshacl Python https://github.com/RDFLib/pySHACL L4 SHACL gate — relator under-mediation and shape rules; sh:conforms true required for canonical promotion
HermiT / Pellet Java HermiT · Pellet Escalation from rdflib+owlrl for full OWL 2 DL classification when needed
z3-solver Python (MIT) https://github.com/Z3Prover/z3 Theorem prover for BFO CLIF axioms that exceed OWL 2 DL expressivity

Graph & Semantic Stores

Tool Language Source Role
Apache Jena Fuseki Java https://jena.apache.org/documentation/fuseki2/ Canonical RDF store — only proven, PROV-O-threaded triples live here
ArcadeDB Java https://arcadedb.com/ Holographic LPG — staging graph where candidates carry full judgment context
Oxigraph Rust https://github.com/oxigraph/oxigraph Embedded SPARQL / RDF store in the runner image for fast local graph queries

LLM & Embedding Services

Service Model Role
Cerebras AI zai-glm-4.7 (primary) · gpt-oss-120b (fallback) Fast structured-output LLM proposer — generates candidate fragments inside the proof box; does not decide acceptance
Cohere Embed v4 embed-v-4-0 · 1536-d (Azure Foundry) Embedding cosine for concept snapping — same vector space as the RAG index
Tavily Search + Extract Web grounding at step 2 of the sift loop — finds authoritative external definitions for proposed concepts

Provenance & Standards

Standard URL Role
PROV-O https://www.w3.org/TR/prov-o/ Every canonical triple traces to source span, proposing activity, and validation reports
OWL 2 https://www.w3.org/OWL/ Ontology language for both gUFO and BFO 2020; DL profile for classification, RL for forward chaining
SHACL https://www.w3.org/TR/shacl/ Shape constraint language for L4 validation gate
RDF / Turtle https://www.w3.org/RDF/ Canonical graph serialisation in Fuseki
SPARQL 1.1 https://www.w3.org/TR/sparql11-query/ Query language for the canonical graph and Oxigraph runtime

To add a reference: append an entry with date, authors, title, URL, and a paragraph on what idea it contributed and where it surfaces in the platform.