CONNECTORS → ONTOLOGY
Connect the tools. Watch the graph appear.
Ten systems come online one at a time and everything in them is read into a single index. What settles out is this: 7,668 entities and 22,395 relations, drawn from a real export rather than an illustration of one.
Slack
Google Drive
Confluence
Jira
GitHub
Linear
Notion
Salesforce
Gong
DiscordThis is the export, not an impression of one
7,668 ENTITIES
Every dot is at the coordinate the extractor gave it and every line is a relation it actually found. The file is on disk in the repository; the page unpacks it and draws it. There is no stand-in layout and no decorative filler between the real points.
Structure first, detail after
ORDERED BY DEGREE, THEN RADIUS
The order things appear in is read off the graph rather than imposed on it: the entities the most sources agree on are laid down first, and the front then moves outward to the rim. What you watch is the skeleton arriving before the tissue.
Then it stops
22,395 RELATIONS
It builds once and holds. The finished plate is the point, not the motion, so when the last relation lands the animation frame is cancelled and the component costs nothing for the rest of the session.
Source: precognita-graph-2026-08-07.svg · 7,668 entities · 22,395 relations · 10 connectors. The 568 KB export is delta-packed to about 110 KB and loaded after first paint. Relations keep the export’s own colours by type and entities keep theirs, except that mentions, which is over half of all relations, is held back to a noise floor so the load-bearing kinds read. The settled graph is stroked once onto a canvas that is never redrawn; the packets sit on a second canvas erased by the boxes they drew the frame before. When the last relation lands the animation frame is cancelled. Requesting reduced motion draws the finished graph immediately and never starts the loop at all.