Trust score
A per-memory scalar that evolves over time based on reinforcement outcomes. Rises when the memory contributes to correct, helpful results; falls when it leads to corrections or poor outcomes. The primary quality signal used by tiered storage assignment, retrieval priority weighting, and the forgetting system's suppression decisions.
Trust score is the memory system's answer to the question: which memories are reliable? It is not set at creation time and never changes - it evolves continuously based on the memory's track record. Each time a memory contributes to a reasoning outcome, the reinforcement engine evaluates that outcome and updates the memory's trust score accordingly. Positive outcomes increase it; explicit corrections or negative feedback decrease it. Over time, memories that consistently contribute to good answers accumulate high trust scores and rise in retrieval priority; memories that mislead or prove incorrect accumulate critique events and low trust scores, eventually triggering suppression by the forgetting system. Trust score also governs tiered storage assignment: high-trust memories stay in the hot tier where retrieval is fastest; low-trust memories are candidates for demotion to warm or cold storage. The result is a self-organizing quality gradient that requires no manual curation.