Persistent companion-organism architecture
Can an autonomous companion maintain an internal life instead of performing one?
UMBRA-CORE is a reusable organism and brain core for believable autonomous companions. Its apparent aliveness is meant to come from persistent causal organization and lived history, not a chatbot wrapped in an avatar.
Identity, physiology, perception, self-model, world model, arbitration, governance, development, selective memory, relationships, individuality, embodiment, and habitat all participate in the action loop. The same individual is meant to continue across restarts, migrations, bodies, renderers, interfaces, and relationships.
UMBRA is not an LLM-driven agent loop, scripted virtual pet, authored personality, scalar affection score, direct command dispatcher, renderer, or digital protocell. Behavior must arise from internal condition, uncertain perception, learned expectations, verified consequences, habits, history, and environmental opportunities.
D-010 temporal continuity is functionally implemented but fails the frozen performance boundary. An integrated D-012 P0 rerun also failed after one authorized remediation. The qualified D-009 foundation remains intact, but the complete project goal is not yet earned.
UMBRA-CORE is a companion-organism foundation. It does not claim consciousness or biological life.
Read the source record in Notion ↗