Context Engine.
A local cognitive engine designed to continuously build and maintain a model of your world.
Under development. This page describes the architecture we are exploring, not a generally available product.
Understanding shouldn’t
start from zero.
The information required to understand our lives is scattered across documents, conversations, projects, meetings, messages and applications.
Today’s AI systems repeatedly reconstruct that context every time they need it. We are exploring a different architecture.
Turn information into persistent, structured local state.
Context,
not documents.
Documents are evidence.
The useful representation is the world they describe.
Context Engine is being designed to connect observations to people, projects, relationships, decisions and commitments — with their sources intact.
Temporal
by default.
Facts change. Projects change. People change.
The intended architecture preserves both what is true now and what was true before. A new observation should update understanding without erasing its history.
What was true before.
What is true now.
Evidence
before belief.
Keep the reason alongside the conclusion.
The system is designed to preserve where information came from, when it was observed and how confident an interpretation should be.
Architecture
before brute force.
Context Engine is designed around a pipeline where deterministic processing and small specialist models perform routine work.
More expensive generative reasoning is reserved for ambiguity.
specialist model
state engine
reasoning — only when needed
Local from
the beginning.
We are developing around private local computation, offline operation, low background resource usage and native device capabilities.
Portable canonical state is a core goal. Encrypted synchronisation is intended as an optional layer rather than a prerequisite.
These are design goals. Availability, supported platforms and implemented security properties will be documented as the product develops.
The principles behind the work