Intent
What outcome is the system actually meant to produce?
A public architecture institution
Where intent becomes architecture.
Intentarium turns ambiguous situations into explicit systems of authority, trust, evidence, and execution—for humans and intelligent agents.
The thesis
Greater agency does not require less structure. It requires better structure.
The core method
The Agency Map
Before automating a system, make its operating logic visible. The same eight questions reveal hidden assumptions in technical platforms, organizations, creative work, and personal systems.
Actors
Which people, teams, services, and agents participate?
Authority
Who may decide, approve, execute, override, or stop?
Boundaries
What must never occur, regardless of efficiency?
Trust signals
What current evidence permits an actor to proceed?
Evidence
How will the system prove what happened and why?
Escalation
What happens when authority or confidence is insufficient?
Feedback
How does the system detect failure and adapt?
The Pattern Commons
Founding patterns
Twelve initial hypotheses about autonomy without chaos. Their maturity is visible by design: no claim becomes established merely because it sounds convincing.
01hypothesis
A mature system reduces dependence on heroics
Convert repeated individual rescue work into observable ownership, safe defaults, runbooks, automation, and learning loops.
leaderoperatorreliability
02hypothesis
A source of truth must precede meaningful automation
Establish one authoritative representation of intent before automating projections, decisions, or execution from it.
architectoperatoragent-builder
03hypothesis
Access, experience, and execution are distinct concerns
Design who may enter, what they perceive, and what effects they may cause as separate policy surfaces joined by explicit contracts.
securityproductarchitect
04hypothesis
Accountability without authority is organizational debt
Assign decision rights, access, and escalation paths wherever a person or team is held responsible for an outcome.
leaderoperatorarchitect
05hypothesis
Decisions should be versioned artifacts
Preserve consequential decisions with context, authority, evidence, alternatives, and revision history instead of leaving them in conversation.
leaderarchitectoperator
06hypothesis
Human intent must survive machine execution
Carry the approved outcome, constraints, authority, and rollback semantics into every executable representation and resulting action.
agent-builderarchitectgovernance
One canon, many interfaces
Readable by people. Usable by machines.
Every public object begins in version control and generates both human and machine representations. The site can disappear and the knowledge remains portable, inspectable, and attributable.
For people
Library and method
Browse patterns, see their maturity and relationships, understand the Agency Map, and track what changes.
For machines
Static, structured surfaces
Retrieve JSON objects, a pattern graph, the Agency Map schema, identity metadata, source commits, and the public corpus without invoking a model.