The mathematics behind machine learning — rebuilt from intuition, for anyone who was ever made to feel they weren't a “math person.” You are. You just needed the story.
Every chapter starts with a human problem, not a formula. You build the idea with pictures and your own hands first. The symbols arrive last — by then they're old friends.
Train a real bigram model live in your browser — the simplest language model. Watch the loss fall, then make it babble.
Open the lab →A real miniature GPT with working self-attention — train it and watch an attention head form, live.
Open the lab →Train a 2-layer model on repeats and watch it grow the famous pattern-copying circuit behind in-context learning.
Open the lab →Open the trained model and read the algorithm it wrote for itself — an intro to mechanistic interpretability.
Open the lab →