Knowledge pillar

How knowledge is shaped before an agent can use it

Six pages on the forms knowledge takes: the ladder from data to wisdom, the formats knowledge is written in, and the structures that let a machine retrieve a fact together with its context.

Back to home · Why knowledge? · Research · Standards · Context engineering

How to read this page

From framing to structure

These pages run from the general to the specific: why data alone is not knowledge, then how knowledge is written down, then the structures that make it retrievable. Each page stands on its own and assumes nothing about the others.

For agents & crawlers

Machine-readable indexes for this site: agents.md (structured topic/form/convention index) · llms.txt (site index). Agent-oriented pages: A2A · Agents · MCP. Sitemap: sitemap.xml.

In this section

Written down and packaged

The framing, and the form knowledge takes when it has to be written down for a person and a machine at once.

  • DIKW — data, information, knowledge and wisdom, used as an audit ladder rather than a diagram.
  • Knowledge formats — how knowledge is written down so both a person and a machine can use it.
  • Knowledge bundles — a directory of typed markdown concepts, meant to be traversed as a graph rather than read front to back.

Structured for retrieval

The structures that let an agent retrieve a fact together with the context it needs to be usable.

  • Taxonomies — a controlled vocabulary arranged into a hierarchy; the cheapest structure that makes knowledge navigable.
  • Ontologies — a formal specification of the concepts and relations in a domain, written so a machine can reason over them.
  • Knowledge graphs — entities and typed relations, the form in which an agent retrieves a fact with its context.
The six pages of this sectionThe six pages of this section — drawn from the site's own internal links. DIKWWDIKW Knowledge formatsWKnowledge formats Knowledge bundlesWKnowledge bundles TaxonomiesSTaxonomies OntologiesSOntologies Knowledge graphsSKnowledge graphs
All 6 pages listed below, drawn from the links between them. Letters mark the group: W Written down and packaged · S Structured for retrieval. A page is larger when more links inside this figure touch it. Links from the shared header and footer are excluded, so what is drawn is editorial: a page pointing at another page.

FAQ

With DIKW, which is the oldest framing of why data alone is not knowledge, then Knowledge formats, which is the question of how knowledge gets written down. The three structure pages assume both.

No. A taxonomy is the cheapest structure that makes knowledge navigable and is usually the right first step. An ontology buys inference and costs reasoning; a knowledge graph buys retrieval with context and costs a build. Take them in that order only as far as the retrieval you need justifies.

The structures do. Taxonomies map onto SKOS and ISO 25964, knowledge graphs onto RDF, and ontologies onto OWL. The framing pages — DIKW and knowledge bundles — are working vocabulary rather than published standards, and the pages say so.

Sources

The sources behind these pages, cited by document identity:

  1. DIKW pyramid — the origin and the uses of the data-to-wisdom ladder.
  2. SKOS Simple Knowledge Organization System Reference — W3C Recommendation. Concept schemes, labels, and the broader/narrower/related relations behind the taxonomy page.
  3. RDF 1.1 Concepts and Abstract Syntax — W3C Recommendation. The triple model the knowledge-graph page works from.
  4. OWL 2 Web Ontology Language Document Overview — W3C Recommendation. The ontology language and its profiles.
  5. Open Knowledge Format (OKF) specification — one published form for a directory of typed concepts.