// R&D

Artificial Knowledge

An AI that distinguishes what is true from what is believed

research prototype

// What it is

A new generation of systems that, unlike generalist models, maintains the distinction between facts, opinions, norms and hypotheses — designed for historical archives, cultural heritage, scientific research and the legal domain, where these differences are essential.

// Where we are

Q2 2026
F0 · formalization of the 5 SEG levels · internal paper
Q3-Q4 2026
F1 prototype · EAL Engine + tripartite ontology on a test domain
Q1 2027
Cultural or regulatory archive pilot with a design partner
// Technical details · for industry insiders +

Technical problem solved

LLMs treat every token equally. Crucial distinctions (what is true, what is believed, what is normative, what is hypothesis, what is verified) collapse into a single probability distribution. For historical archives, cultural heritage, scientific knowledge and law, an infrastructure is needed that preserves these distinctions.

Technical positioning

Epistemic infrastructure · stratified graph + formal verification

Architecture / approach

  • 01 Level 1 — Empirical Manifold · typed primary observations
  • 02 Level 2 — Dynamic Ontology · classes that evolve with the domain
  • 03 Level 3 — Neurosymbolic Bridge · LLM ↔ symbolic anchoring
  • 04 Level 4 — Epistemic Judgment · EAL Engine for modal judgment
  • 05 Level 5 — Architectonic · meta-level of global coherence
  • 06 Wittgensteinian tripartite ontology — separated fact/value/grammar

Technology stack

PostgreSQL Neo4j Qdrant Python LangChain RDF/OWL FastAPI

TRL and development status

Prototipo · TRL 3/6
Current TRL
3
Target TRL
6
Phase
Prototipo

Reference standards

RDF/OWL SKOS PROV-O CIDOC CRM

// Related services

The expertise behind Artificial Knowledge directly powers some of our project-based services:

Approach to intellectual property

Why we patent and what it means for clients

// Other R&D projects