Skip to the content

// the connections

The knowledge graph

A knowledge graph names the things worth keeping, the relationships between them and the context needed to read those relationships.

// three parts

The connections carry meaning.

Nodes: things worth naming
A topic, project, lesson, source or grouping page. A canonical identity keeps a subject recognizable when its display name changes.
Edges: named relationships
One page links to another, belongs to a group or is part of a broader theme. An explicitly supported semantic relationship can say more.
Metadata: context around a record
Type, summary, aliases, labels, source, freshness and trust help a reader decide what to open and how to interpret it.

arrows = the named relation from one record to another · lit = an example lesson

  1. Group The invented topic Retry policy and lesson Bound retry work each belong to the Reliable systems grouping page.
  2. Cross-link The lesson links sideways to the topic; their shared group supplies another route between them.
  3. Inspect evidence The lesson has an illustrative supported_by relation to an invented work note. Its evidence and trust labels still need to be read.
  4. Enter anywhere Start at a group, topic, lesson or source. Follow outgoing relations, backlinks or inverse membership to discover wider context.
Illustrative records only. Two pages can share a group and link sideways; an explicit supported_by relation provides a route to evidence. Following a link exposes a source to inspect, not a guarantee that its claim is true.

Download the knowledge graph SVG ↓

// progressive discovery

A hierarchy inside a web.

Grouping pages provide broad-to-narrow routes. Cross-links connect useful neighbors. A topic can belong to several groups, and a reader can enter at any page.

Depth belongs to the path. The same lesson might sit near the surface of a project overview and deeper in a broad engineering theme. There is no single reading level permanently assigned to every page.

The group hierarchy has no parent cycles. Ordinary cross-links can form loops, so navigation uses deliberate routes and bounded exploration.

// entity resolution

Resolve identity before joining things.

A canonical page is the chosen home for a subject. Known aliases point to that identity, and explicit rename records preserve it when a path changes. A similar-looking name alone does not justify merging records.

When a name could mean several things, the candidates stay visible for review. Missing or ambiguous references retain their original query and a route into a review record. That route makes the gap inspectable; it does not invent the missing knowledge.

// closed relationship vocabulary

Say which connection you mean.

links_to
A page has an ordinary link to another page.
member_of and part_of
A record belongs to a group, or a group/page sits under a broader structural parent.
applies_to, supported_by and used_in
Declared semantic connections can identify relevant context, supporting evidence or a place an idea was used.
contradicted_by, supersedes and inspired_by
Other declared connections can preserve contrary evidence, replacement or an originating influence.

Having a relationship type available does not assert that relationship. Semantic edges need declared support. Communities discovered from graph structure remain proposals until a separate decision adopts them.

Follow a question through the search process →