Dual-Signal Retrieval for AI Agent Memory
A white paper on ranking an AI agent's memories by use without making recall worse. A two-year simulation through production code shows why use and correctness must be two separate signals, why every ranking nudge must stay smaller than the gaps it reorders, and why recall and long-term memory need different clocks.
Read →