Day 32: Why agent behavior is a loop, not a single prompt

Strip away the mystique and an agent is a loop. observe -> plan -> act -> check -> decide ^ | |-------------------| One LLM call is a single shot: prompt i