Interesting thread that mirrors how we see agents being built: 1. Break complicated task into a highly controllable domain specific workflow (LangGraph) 2. Test each component thoroughly (LangSmith)

→ View original post on X — @hwchase17