[00]3 chapters
Prologue
We built a machine that answers in words, even though nobody can fully explain one reply. Meet the old wish behind it, the learning idea that made it possible, and the questions it leaves open.
~Contents
Everything, in the order it was meant to be read.
The numbered modules build one connected picture, each leaning on the one before it. The standalone pieces take a single subject further and can be read whenever you like.
~the course · 12 modules
The main journey
[01]2 chapters
The Machine
See how switches become bits, logic, memory, and a working computer. The machine becomes understandable from the inside out.
[02]3 chapters
Can Machines Think?
Follow machine intelligence from Turing to Dartmouth, then examine what imitation, understanding, and intelligence actually mean.
[03]3 chapters
The Long Winter
See why early AI promises collapsed, why expert systems broke outside narrow rules, and what those failures taught the field.
[04]3 chapters
Learning from Data
Build a neuron from simple parts, connect it into a network, and see how examples reshape its internal weights.
[05]4 chapters
Going Deeper
See how better training methods, parallel hardware, large labelled datasets, and one shared test turned layered networks into convincing deep learning systems, then how learning from consequences reached games and science.
[06]3 chapters
The Meaning of Words
Learn how machines break language into tokens, turn words into numbers, and place related meanings near one another.
[07]3 chapters
Paying Attention
See why older language systems lost track of distant words, then build the attention mechanism at the heart of transformers.
[08]3 chapters
The Model in the Machine
Look inside a language model: scale, unexpected abilities, next-token prediction, and what its billions of learned weights contain.
[09]3 chapters
Talking to the Model
See what the model receives, give it a clear job, and check what it gives back.
[10]3 chapters
Agents
Follow the step from generated text to real action, then meet the wider family of AI beyond language.
[11]3 chapters
Epilogue
Pull the whole thread together, then face what mechanism alone cannot settle: what stays uncertain, who chooses the objectives, who carries the risk, and how to judge the claims you meet next.
~alongside
More from Untangled
[12]2 chapters
The Sommelier
Turn the crank on a learning machine by hand, one bottle of wine at a time, until a rule nobody wrote down appears in four numbers.
[13]3 chapters
The Amnesiac Oracle
An analogous journey through an LLM conversation.
[14]1 chapter
Prompt Caching
See how AI systems avoid repeating the same work when the beginning of a prompt stays unchanged.
coming soon
[15]4 chapters
Model Context Protocol (MCP)
See why MCP exists, how its parts connect, how local and remote servers differ, and what a tiny server looks like in code.
coming soon