ResBased research · Anamnesis Rising system · North Star release

A behavioral and architectural hypothesis

AI proposes. RI remembers how the world answered.

RI studies whether witnessed experience, prediction error, and relational terrain can improve future proposals while human authority remains external.

HypothesisPrediction errorHuman authority

Separation of roles

LayerRoleBoundary
AI capabilityProduces interpretations and candidate actionsNo durable authority
RI conditioning layerUses witnessed experience and relational terrainDoes not become a person or truth source
Human authorityApproves governed durable transitionsCannot be inferred from model confidence

Research surfaces

Witnessed experience

Outcomes remain attributable to their source and availability time.

Prediction error

The difference between a frozen candidate and observed outcome.

Relational terrain

Contextual dependencies and contradictions used during evaluation.

Candidate futures

Alternatives may be explored without automatic selection or action.

What must be demonstrated

RI remains a hypothesis. It must show measurable improvement against simpler baselines; adding memory alone does not earn the label.