Retrieval priority
A scalar score maintained per memory that determines how prominently it surfaces in k-NN and hybrid retrieval results. Computed as a weighted combination of reinforcement signal, usage frequency, and goal relevance. Increased by Hebbian compounding on positive reinforcement; reduced by decay over time.
Retrieval priority is the primary lever the reinforcement and decay systems use to shape what the memory gateway surfaces during active cognition. It is not the same as embedding similarity - a memory that is semantically close to a query will surface via k-NN regardless of its retrieval priority, but retrieval priority modulates how it is ranked among candidates and whether it survives the final cut when the working memory budget limits how many memories can be active simultaneously. Positive reinforcement events increase retrieval priority directly; Hebbian compounding adds a logarithmic bonus for memories that are repeatedly retrieved and reinforced. Decay reduces retrieval priority multiplicatively over time for memories that are not re-consolidated. The result is a self-organizing priority landscape where frequently useful memories remain highly accessible and stale or incorrect memories gradually become less prominent.