Four planes.
One decision.
A memory is not activated because it is nearby. It is activated only when its evidence, scope, condition, and verification state are compatible with the present.
EXPERIENCE
Observations, evidence, actions, outcomes, and context are captured as linked episodes.
JUDGMENT
Multiple decision loops condense independent checks into typed confidence and disagreement.
CONTROL
Promotion, qualification, suppression, review, and reactivation are explicit state transitions.
INTEGRITY
An antibody engine scans for drift, corruption, regressions, and unsafe reactivation.
Activation is conditional
A candidate memory must satisfy scope compatibility, temporal validity, identity, contradiction status, evidence quality, and verification history. If compatibility cannot be established, the architecture fails closed.
Suppression is reversible.
Typed suppression distinguishes hard quarantine from rethinkable states, enabling evidence-backed re-evaluation without silently restoring risk.