Start here
Who the course is for, what you'll understand by the end, and how the lessons work.
Start with this short video.
Read the transcript
Welcome to How AI Works. It's for anyone who uses AI at work. You don't need a technical background.
It explains what's going on when a model answers you. It comes in four parts.
How it writes: one token at a time, picked from its odds. That's why the same request can come back different.
How it learned: what training gives it, why it sometimes makes things up, and why it tends to agree with you.
What it can see: only what the app sends it. That might be the chat so far, your documents, or an image.
And how it acts: with tools, it can look things up and take steps. And text it reads can try to give it orders.
Every recorded example in the course is a real model reply, labeled with the model and the date. Some are surprising. Asked how many r's are in strawberry, GPT-4.1 mini said two, all three times.
Each lesson goes one step at a time. You read a short part, see or try one thing, then move on. Each part ends with a short check.
First up: how a reply gets written.
16 lessons, a step at a time, with a short check after each module. No technical background needed. The recorded examples are real model replies, labeled with model and date. Module 1 is free, and the rest comes with Pro.