About

Product

Learns · Knowledge Compiler Briefs · Context Engine Yours · Trust Layer

Surfaces

Enterprise · Soma People · Self Family / team · Self+ M2M · Axis Protocol · Nexus

IP

Patent · P202631047 MVE · Minimum Viable Exposure .aix format ↗ Research Contact hello@thegeometrical.com

01 · Knowledge Compiler

Learns.

Your company’s knowledge, compiled from the work you already do.

Decisions, roles and the reasons behind them — read from mail, meetings, documents and tickets inside your perimeter, structured into a private graph, and kept right on its own. Nobody has to feed it.

Sources14 passive source families
InputZero manual entry
UpkeepEvery 30 minutes, on its own

How it works

It structures. It maintains.

GEOMETRICAL.OS

Roles & ontology

Nobody is one thing. The graph knows which one, where.

The same person is a partner on one client, a reviewer on another project and a parent at home. Geometrical resolves every piece of content to its context and infers the role at that context — from what the content is about, not from someone filling in a form.

  • Roles are relations: role@context, never a flat label
  • Typed nodes and predicates — an ontology template per surface and vertical
  • Closed vocabulary: the extractor can't invent roles, and every edge needs evidence
GEOMETRICAL.OS

Illustrative flow of the real pipeline stages.

Self-maintaining graph

It feeds itself. It cleans itself.

Every 30 minutes the runtime scores what came in, throws away noise before spending a single token, merges duplicates, checks each new fact against the ontology, its evidence and what's already known — and only then commits it. Unclear cases wait for more evidence or go to a human.

  • Novelty × density score: noise dropped before any LLM call
  • 3-layer dedupe: exact hash → meaning → judge
  • Quality gate: ontology, evidence, contradictions, calibrated confidence

What’s inside

Four things it learns without being told.

The same compiler runs for a person, a team or a whole organization. Only the template changes: which sources it may read and which ontology it builds.

01 · Compiles

Compiles

From the work you already do

Mail, meetings, documents, tickets and chats are read inside your perimeter and turned into a private graph of who, what, when — and why. Nobody fills in a form.

Live

Sources14 passive source families
InputZero manual entry
OutputA typed, causal graph

02 · The why

The why

Causal compilation

Every decision is stored with the reason behind it and what happened next. Agents stop guessing why things are the way they are.

Live

ChainDecision → reason → consequence
RecallAsk “why” and get the chain
EvidenceEvery link points to its source

03 · Roles

Roles & ontology

Nobody is one thing

A person is a partner on one client, a reviewer on another project and a parent at home. Roles are inferred as relations to a context, inside an ontology built for your vertical.

Live

Modelrole@context
VocabularyClosed, per template
EdgesNo evidence, no edge

04 · Keeps itself right

Keeps itself right

Self-maintaining

It drops noise before spending a single token, merges duplicates, flags contradictions and stale facts, and checks every fact before committing it. Your team confirms, corrects or forgets in one tap.

Live

CycleEvery 30 minutes
DedupeHash → meaning → judge
CommitRules → model → human

Next

Briefs. →

Every AI gets exactly the context it needs. Nothing more.