001
Capability / authority
Capability Is Not Permission
The foundational distinction: technical reach does not establish legitimate authority to act.
RESEARCH SERIES · FIRST ARC
Governed Agent Systems
A five-piece research arc on capability, delegated authority, escalation, receipts, and revocation in artificial systems that act.
Working definition
A governed agent is an artificial system capable of selecting and performing actions within an explicit, attributable, constrained, inspectable and revocable delegation of authority.
STATEActive research
Authored public research under Vega. The series is not presented as canonical Doctrine.
000 / Introduction
The other half is consequence.
Once software can move from describing what could happen to making something happen, authority, delegation, escalation, revocation and evidence become architectural.
Read the series premise →001–005 / Reading order
The authority lifecycle.
Read →
002
Delegation / jurisdiction
The Authority Envelope
A practical primitive for binding an agent to a principal, objective, scope, constraints, evidence requirements and revocation conditions.
003
Boundaries / judgement
Escalation Is an Output
Why a competent agent must be able to stop, expose the unresolved boundary and route the smallest necessary decision to actual authority.
004
Consequence / provenance
The Receipt
A governance receipt closes the authority chain by preserving what changed, under what delegation, with what evidence and result.
005
Time / authority
Revocation Is a Runtime Event
Authority changes while systems are running. Revocation therefore has to exist in runtime state, propagation and consequence boundaries.
Complete set