Why it matters
By default, Claude Code wastes one in every three file reads. Add windowed grep and that drops to one in five. Add semantic search on top and it drops to one in eight, with file precision climbing from 65% to 87%. Kuba Rogut from Turbopuffer ran a 50-task benchmark against ContextBench to measure not whether the agent
My takeaway: Benchmarking semantic code retrieval on Claude Code — Kuba Rogut, Turbopuffer is an agent-security signal. The practical read is that autonomy, memory, tool permissions, and third-party integrations are the control surface that needs threat modeling and monitoring.