An ongoing project. New lesson every few months.
Visual, interactive lessons
about how AI actually works.
No hype. No prophecy. No "this changes everything." Just careful, honest explanations of the ideas inside the machines, with diagrams you can poke at, drag, and break.
Interactive lesson preview
Why I am making this.
There is a strange gap right now. Hundreds of millions of people use AI every week, and almost none of them know what is happening when they hit enter. That is not anyone's fault; the field grew up faster than its explanations did.
This site is my attempt to close that gap, one idea at a time. Each lesson starts from zero and builds up slowly, with diagrams you can play with instead of just look at. If you have ever wanted to actually understand what a model is doing, I hope these help.
Written by one person, in their evenings, with a lot of coffee.
The lessons.
Six published · three planned
A real-tokenizer interactive lesson about how AI models read text by breaking words into tokens before turning them into numbers.
An interactive lesson on how AI models turn token IDs into positions in a high-dimensional space — where meaning becomes geography.
An interactive lesson on how transformer models decide which words to look at — the mechanism that made modern AI possible.
An interactive lesson on the transformer architecture — the assembly of tokens, embeddings, and attention that powers every modern LLM.
An interactive lesson on how temperature controls randomness in language model outputs — the single knob that shapes everything from code to creative writing.
An interactive lesson on why language models produce confident, fluent, and sometimes completely wrong answers — and what to do about it.
What the model can and cannot see at any moment, and why long conversations sometimes lose the thread.
How a general-purpose model learns to specialise — and what you give up when you teach it new tricks.
Giving the model a library: how retrieval-augmented generation grounds answers in real documents.