Domain knowledge / canonical synthesis
Knowledge and epistemics
Canonical page of Zenith's knowledge ontology, atomic-note graph, retrieval model, claim and evidence pipeline, bucket consistency system, peer review, research journal, and epistemic governance.
- Updated
- 2026-08-29
- Source digest
48e935a7daf6- Structured data
- JSON-LD
Scope
This page owns how Zenith represents what an organization knows, how agents retrieve it, how claims acquire authority, and how disagreement becomes revision or proof. Runtime implementation belongs in Zenith Research - system architecture.
Current answer
Zenith treats authored structure as part of the intelligence system. Atomic notes are natural retrieval chunks. Wiki links encode explicit relationships. MOCs route attention. Raziel maintains the relatively stable knowledge ontology, while Frank handles information objects moving through operational states.
The bucket is the formal consistency layer: candidate claims are evaluated before entering a versioned set of accepted propositions. Public peer review, research publication, and computable proofs are network extensions of the same local method.
Evidence and influences
- defines the knowledge/information split: Raziel manages the business knowledge ontology while Frank manages the business information ontology — static structure vs objects in motion.
- defines atomic retrieval units: Notes are natural RAG chunks because one note one concept alignment makes chunking strategy unnecessary.
- defines the deployed implication: A vault built on atomic notes is RAG-ready at rest without a chunking pipeline.
- defines the original formal model: the 2023 Zenith formal model treated knowledge systems as theory-premise pairs where consistency is the only valid truth criterion.
- defines Raziel ingestion: raziel-wiki-ingestion-architecture.
Knowledge graph and retrieval
- Wiki-links between notes encode semantic relationships that retrieval systems would otherwise have to infer - links are typed evidence about relationships, not decoration.
- search skill complexity belongs in the knowledge ontology, not the search tooling layer - specialized retrieval behavior should be represented as domain knowledge.
- The collective knowledge graph makes every contributor's knowledge part of the whole without surrendering local ownership - shared structure with local control.
- ai-os-raziel - Raziel's role as vault-level archivist.
- raziel-wiki-ingestion-architecture - parse, classify, embed, and route by type.
Claim and consistency system
- bucket-claims - local candidate-claim map.
- Open peer review extends the proposition system publicly so any claim can be contested and proven on a consistency standard.
- Chained consistency proofs link accepted claims into a formally verifiable knowledge structure.
- Social claims and computable proofs create a pipeline from informal contestation to formal verification.
- The RFC and RFP system structures community input on architecture and governance decisions through formal requests.
- The research journal publishes claims that clear the consistency bar as citable versioned documents.
Epistemic workflow
- Capture a source, observation, or operational result.
- Extract atomic claims with provenance.
- Connect claims to supporting, extending, and challenging notes.
- Surface contradictions rather than silently averaging them.
- Route mature claims through review and ratification.
- Publish accepted claims as versioned, citable records.
- Feed corrections and failures back into the ontology and agent evaluation layer.
Current implementation boundary
- Atomic notes, explicit links, MOCs, source reports, and the local bucket exist in this vault.
- Semantic retrieval exists as a design and partial tooling layer; it should not be assumed healthy without checking the local index.
- Public review, network-wide claim markets, and formal proof chaining are product directions rather than completed infrastructure.
- The current authority model still relies heavily on curation and explicit canonical-page statements.
Open questions and tensions
- What exact schema distinguishes observation, source, claim, decision, and accepted proposition?
- How does a contradiction update retrieval ranking before formal resolution?
- Which claims require formal proof versus sourced human review?
- How does contributor reputation affect review weight without hardening into permanent authority?
- How are organization-private claims separated from public research while preserving lineage?
- What is the migration path from this flat vault into independently owned federated knowledge graphs?
Related canonical pages
- Zenith Research - research evidence and sources - source and report layer.
- Zenith Research - data and intelligence - evaluation artifacts and learning data.
- Zenith Research - network governance and economics - contribution and review incentives.
- Zenith Research - system architecture - Raziel, Vault, and retrieval implementation.