In practice, retrieval is a system with its own failure modes, its own latency budget and its own quality requirements.

This isn't about rejecting large models; it's about having the engineering discipline to use smaller, specialized models ...