PrecognitaGet demo

THREE LENSES · ONE ONTOLOGY

Three lenses, one ontology.

Precognita indexes a company once and renders it three ways. An agent connected to it reads all three against the same question, which is why its answer holds up: where the lenses overlap is where a claim can be checked twice.

The knowledge graph as a disc: entities joined by the relationships between them.Connections
The source tree as a disc: connectors at the centre, every indexed document on the outer ring.Hierarchy
The embedding space as a sphere: every indexed chunk placed by what it means.Recall

Connections

7,556 entities · 22,057 relationships

Entities and the relations between them, extracted from the material and then corroborated. Nothing is entered on a single source's word: two independent sources have to assert it, and the agent can return both.

Hierarchy

646 sources · 6 levels deep

The shape the material already has — workspaces, drives and folders, down to every indexed document on the outer ring. Nothing here is inferred and nothing is rearranged.

Recall

233,703 chunks · 20 topics

Every passage reduced from 1,024 dimensions and packed into a ball, so that position carries similarity and nothing else. Two documents filed a continent apart sit together if they say the same thing.

One corpus · 3 connectors · 646 sources · 233,703 passages. The three plates are the same index asked three different questions, drawn from live data rather than mocked, and none of them carries a filename, a label or a line of anyone’s content.