From zero to GPT.
How does a chatbot actually work? Start by counting letters in a poem, the way a Russian mathematician did in 1913. Finish by training your own small ChatGPT. Along the way you'll meet the people who figured each piece out: who they were, what they were stuck on, and how they thought their way through.
Who it's for
Middle-school classes
Written for students in grades 6–8 and a teacher who doesn't need to be an AI expert. Each lesson starts with something you can do on paper before any code.
Grown-ups learning on their own
Work at your own pace. Every idea links to the original paper and the story behind it, so you can go as deep as you like, all the way to Andrej Karpathy's nanoGPT and nanochat.
Ideas are made by people
The history matters. Breakthroughs came from people stuck on real problems. Some were ignored for decades, some argued in public, and some changed their minds. The timeline tells those stories.
The main path
Side branches
Not every important idea is on the straight road to GPT. These side branches show up on the timeline too.
The Timeline of AI
Bayes to nanochat: the ideas, the people, the winters and the booms.