World model
A persistent, self-updated internal model of the system's operating environment and recent state, maintained by the world-loop to ground planning and self-improvement in real signals rather than assumptions.
grounded in: world_loop.py / world_loop.sh present in the harness scripts, and MEMORY.md project_knowledge_loops noting 'world-loop feed the self-improve/idea engines'
Connected concepts
Agent memory, Context engineering, Context ontology, Planning & reflection, Self-improvement daemon, Idea engine
Explore it live in the knowledge graph →