Agents

How an AI model goes from answering to acting: working toward a goal step by step, and how to tell when it has gone wrong.

1. 01What makes something an agentWhat separates an agent from a chat reply: a goal, and the model choosing its own next step.about 6 minutes
2. 02How a model asks for a tool to be runThe model never runs anything itself. It writes a request, and separate software decides whether to act on it.about 7 minutes
3. 03The loop it runs, and how it stopsHow an agent decides, acts, looks at what came back and goes again, and how it knows when to stop.about 8 minutes
4. 04Giving an agent a goal it can finishTurning a vague request into a goal with a finish line, limits, and things the agent may not do.about 7 minutes
5. 05What an agent remembers between stepsWhat an agent holds while it works, what it writes down, and what it loses along the way.about 7 minutes
6. 06When you need an agent, and when you don'tMany tasks run better as fixed steps. How to tell which ones need an agent, and where a person should sign off.about 7 minutes
7. 07When one agent hands work to anotherWhat changes when one agent coordinates others, and why a second agent often adds failure points rather than capability.about 7 minutes
8. 08When an agent goes wrong, and what to changeReading the record an agent keeps of its steps to find where a run went wrong, and deciding what to change.about 9 minutes

Opening a lesson needs membership. During the preview it's free for invited members. See plans