Worked example · from Wikidata

Typed graphs, by example.

One item from Wikidata, eight typed edges to other items. Every box has a type. Every arrow has a name.

P800 notable_work P106 occupation P27 citizenship P19 place_of_birth P569 date_of_birth P69 educated_at P26 spouse P1412 languages HUMAN Q42 Douglas Adams 1952–2001 · British author NOVEL · BOOK SERIES Q25169 Hitchhiker's Guide to the Galaxy first published 1979 OCCUPATION Q36180 writer + Q245068 comedian COUNTRY Q145 United Kingdom also → sub-edges CITY Q350 Cambridge located_in: England LITERAL VALUE · DATE 1952-03-11 COLLEGE Q691283 St John's, Cambridge part_of: U. of Cambridge HUMAN Q3304605 Jane Belson married 1991 LANGUAGE Q1860 English + written + spoken READ AS: Douglas Adams is a human · born 1952-03-11 in Cambridge · citizen of UK · educated at St John's · occupation writer · spouse Jane Belson · spoke English · notable work Hitchhiker's Guide.

The alphabet that powers it

Every node has a type from this list. Every edge has a name from this list. Wikidata has ~10,000 of each — here's a tiny slice.

Node types — sample

"instance of" — what kind of thing this node represents

human country city book film painting song species organization university building chemical element gene disease language mountain programming language software mathematical theorem fictional character

Edge types — sample

"property" — what kind of relationship the arrow carries

instance_of subclass_of place_of_birth date_of_birth citizenship occupation member_of educated_at spouse child author director composer published_in genre located_in capital_of currency population coordinates ISBN discovered_by

What the typing makes possible

Four real features. None of them work on untyped data.

01
Auto-generated Wikipedia infoboxes
Douglas Adams BORN 11 March 1952, Cambridge DIED 11 May 2001, California OCCUPATION Writer, comedian SPOUSE Jane Belson populated from typed edges above
Every Wikipedia infobox is auto-built by querying the typed properties of one item. Change language: same data, translated labels.
02
Structured queries (SPARQL)
SELECT ?writer WHERE { ?writer instance_of human ; citizenship France ; occupation mathematician ; date_of_birth < 1900 . } → 287 results in 0.3s precise because typed
"French women mathematicians born before 1900" → 287 results in 300ms. Only possible because every field has a known type.
03
Voice-assistant factual answers
"Who directed Inception?" Q25188 → P57 (director) → ? "Christopher Nolan." one typed lookup, no NLP guessing
Siri, Alexa, Google all run typed-property lookups against graphs like Wikidata. "Who directed X?" → one edge traversal.
04
Automatic consistency checks
Q12345 · HUMAN P569 date_of_birth: 1990 P570 date_of_death: 1985 ⚠ CONSTRAINT VIOLATED date_of_death must be ≥ date_of_birth
The system catches that a death date precedes a birth date — because both fields are typed as temporal and the constraint is declarable.
The same shape

Your estate system, smaller alphabet.

Same mechanic. Different domain. Wikidata models the world; your system models one family's plan. The math of "typed nodes, typed edges, structured queries, rule predicates" works at both scales.

Wikidata
~10,000 node types
~10,000 edge types
~100M items
All of human knowledge, structured.
Legacy.
~11 node types
~11 edge types
~thousands of nodes
One family's plan, fully audited.