ONE INDEX · EIGHT CONNECTORS · 7 AUGUST 2026
A company, as a graph, as it arrives.
Not an illustration. Every point and line here was read out of a knowledge graph Precognita exported from one company’s live index: 7,668 entities and 22,395 relations, in their real positions and coloured by what the extractor decided each one was. The eight systems below fed it.
0 entities0 relations
RELATIONS
Slack
Google Drive
Confluence
Jira
Linear
GitHub
Notion
SalesforceWHAT IS REAL
Positions, sizes, entity types and relation types all come straight out of precognita-graph-2026-08-07.svg, along with the colours, which are the export’s own. Every primitive in that file is on the plate: all 7,668 entities and 22,395 relations, including the 21 entities the extractor never managed to type, which are drawn grey and counted as unclassified rather than quietly left out. The export carries no per-entity provenance, so none is claimed: a parcel crossing from a chip says material arrives from that system, not that those particular entities did.
WHY IT ARRIVES THIS WAY
A relation cannot be drawn until both of its ends exist, so the graph is cut into 64 equal parcels — sixteen radial shells, each quartered by angle — and delivered from the rim inward. The wavefront contracting on the core is the order the picture is genuinely available in, and it puts the crescendo where the graph actually is: 15,000 of the 22,395 relations live in the last ten parcels.
WHAT IT COSTS
The graph is stamped into an offscreen canvas as it lands and never redrawn, so a settled frame is two composited images plus a handful of moving dots — the same cost whether there are two hundred relations on screen or twenty-two thousand. Isolating a relation type restrokes one extra layer, once, on the hover. Under reduced motion the graph is drawn in a single frame and nothing moves.