Anam #08: "No mechanism to measure whether retrieved records were actually relevant. Without this, the system gets bigger but can't prove it gets better."
Proposed: retrieval_log table: query_hash, retrieved_record_ids, records_used_in_response (boolean per record). Monthly retrieval quality reports from day 1. This is how you answer: "is this system working?"
v3 2026-04-05 Q Mapped to whitepaper sections
v2 2026-04-05 Q Imported SPEC-042 from model_specifications_v2.html
v1 2026-04-05 Q Created spec: SPEC-042: Retrieval Evaluation