[Online] AI Agents, Demystified: from ChatGPT to Grok Bot

​You've probably used ChatGPT or Claude. You've definitely heard the word "agent" more times this year than you wanted to. This is 75 minutes on what it actually means, and what it takes to build one people can rely on.

​Start with what's out there. Grok Bot launched in August: each agent gets its own cloud computer, keeps its memory, carries on working after you shut the laptop. Hermes and OpenClaw are a self-hosted version of the same idea. ChatGPT and Claude with tools switched on are related but not the same thing. We'll sort out which is which, and the handful of properties that decide it.

​The distinction turns out to be simple. In a chat window you're in the loop every turn. Nothing happens unless you're at the keyboard. An agent takes you out of the loop, and everything hard about building one comes from that.

​Second half is about building one. Everything above is something you buy, which means someone else decided what it's for, what it can reach, where it stops. Fine for a personal assistant. Less fine when colleagues depend on it, or you're the one answering for what it did. So: scope, permissions, what a run costs, how you'd know what it did.

​Then questions. Bring whatever you're stuck on.

​For product people, founders, operators and anyone who's been told to "add an agent" to something. No coding needed. If you already build with these tools, the second half is for you.


WA group: Join WhatsApp Group

LinkedIn Page: /the-builder-course