How Steward works
Continuity first. Intelligence second.
Steward is not a model. It is the layer that remembers, holds provenance and permission, and hands intelligence only what a task actually requires.
See Steward
Your information already exists. It simply lives in different places.
The flow
Sources → Steward → Context → Intelligence → You
Many sources in
Continuity held by Steward
Minimum necessary out
The person chooses which sources to connect. Steward integrates before it duplicates.
Steward preserves what information represents, where it came from, how it relates to what happened before, and the permissions governing its use.
For a given task, Steward assembles a small, relevant, task-specific context package — not the whole life.
The best available intelligence receives only the context the task requires. It never receives ownership of the continuity.
The answer returns to the person, grounded in evidence they can inspect.
Many sources in. Minimum necessary out. Steward holds the continuity.
The layers
What each part of the system is responsible for.
Each layer has one job. The value comes from the boundaries between them being deliberate rather than accidental.
01
Sources
Connected, not copied.
- Health systems and records
- Calendars and schedule
- Documents
- Communications
- Goals
- Finance
- Future domains
02
Steward continuity layer
What is preserved.
- The source of every piece
- What the information means
- The history behind it
- How things relate to each other
- The permissions governing use
03
Context assembly
Minimum useful context.
- The task determines relevance
- Permission determines use
- Narrow disclosure by default
- Never the whole life
04
Intelligence
Chosen for the task.
- Frontier models
- Specialized models
- Deterministic services
- Local or Steward-native intelligence in future
05
The person
Where it returns.
- Remains the decision-maker
- Receives useful interpretation
- Can inspect the evidence
- Maintains control
Three core capabilities
Continuity. Synthesis. Boundary crossing.
These are behaviours of one system, not three separate features. Proving all three in a single domain is what makes the architecture credible everywhere else.
Continuity
Meaningful context persists across time. What mattered last year is still available to explain this month.
Synthesis
Relevant information from different sources becomes understandable together — without losing provenance.
Boundary crossing
Context can cross domains when it genuinely helps the person and the person has permitted it.
In practice
One question, assembled from many places.
A single question rarely lives in a single system. Watch which parts of a life become relevant when the question is asked.
Why am I so tired?
Sleep, exercise, schedule, travel and what changed before.
The task determines relevance. Permission determines use.
Model independence
The intelligence can change. The continuity remains.
AI providers, devices, applications and platforms will change. The person should not have to rebuild their identity and history every time the best intelligence changes.
Steward
The continuity remains.
Provenance, permissions and history stay with the person.
Intelligence
The intelligence can change. It is chosen for the task, and replaced when something better arrives.
Next
Health is the proving ground. Continuity is the product.