Goal system
The representation of long-horizon objectives, organized into a hierarchy of five timescales. Goals feed progress signals to reinforcement independent of immediate outcomes and bias retrieval toward relevant memories.
The goal system extends the system's optimization horizon beyond the immediate outcome of each cognitive loop iteration. Without goals, the reinforcement engine can only reward actions based on what happens right after them - it cannot credit a decision that was individually unremarkable but contributed to a long-range objective. The five-timescale hierarchy runs from immediate (within a session), short (days), medium (weeks), long (months), to persistent (indefinite) goals. Each timescale has its own progress tracking and contributes goal progress signals to the reinforcement pipeline on different schedules. Goals also bias retrieval: when the memory gateway queries the associative memory layer, active goals influence which memories score highest on goal relevance, pulling context about past progress and obstacles toward the current goal into the working memory set. Goal state is stored in the OpenSearch goal index alongside experience events and semantic memories, making it queryable in the same retrieval pipeline rather than requiring a separate lookup path.