The journey
Sixteen chapters and an epilogue
Part I
Foundations
From hand-written rules to machines that learn.
Chapter 1
What Is Artificial Intelligence?
Seventy years of trying to write intelligence down — and why we stopped.
Can a machine behave intelligently if we simply tell it everything it needs to know?
1 h 30 min
Chapter 2
The Math Toolkit
Six small pieces of mathematics that do almost all the work in modern AI.
Why does every AI paper look like a wall of Greek letters — and which parts actually matter?
5 h
Chapter 3
Machine Learning
Learning a function from examples — and the art of not fooling yourself.
How can a program improve at a task by looking at examples instead of following rules?
2 h 30 min
Chapter 4
Neural Networks
Stacking simple units until they learn their own features.
Hand-designed features limit what a model can learn. Can a model learn the features too?
3 h 30 min
Part II
Branches & Language
Seeing, hearing, acting — and the long road to understanding text.
Chapter 5branch
Vision, Speech & Reinforcement Learning
The wider field — and the idea of learning from consequences.
Not all data is text, and not all learning comes with correct answers attached.
2 h 30 min
Coming later
Chapter 6
Language Before Transformers
Counting words, embedding meaning, and the bottleneck that attention broke.
How do you turn words into numbers that preserve meaning — and remember what came fifty words ago?
2 h 30 min
Chapter 7
Transformers
The architecture that reshaped modern AI.
Recurrent networks read one token at a time, so distant words interact only through a long chain of steps — slow to train and hard to remember across.
4 h
Part III
Large Language Models
Prediction at scale, and how it becomes an assistant.
Chapter 8
The Rise of Large Language Models
What happened when we simply predicted the next token — at scale.
Task-specific models needed task-specific data. Could one pretrained model do many tasks?
2 h 30 min
Coming later
Chapter 9
How an LLM Is Actually Built
Data, tokens, GPUs and the laws of scale.
What would it actually take to train a modern language model?
2 h 30 min
Coming later
Chapter 10
From Base Model to Assistant
Why a pretrained model isn't yet helpful — and how post-training changes that.
A base model continues text; it doesn't answer questions or follow instructions reliably.
2 h
Coming later
Chapter 11
Inside Modern LLMs
The engineering that makes large models fast, cheap and adaptable.
The original Transformer is too slow and memory-hungry to serve at modern scale.
2 h 30 min
Coming later
Part IV
Systems
Retrieval, tools, agents, reasoning and many modalities.
Chapter 12
Embeddings, RAG & the LLM Application Stack
Giving models knowledge they weren't trained on.
A model's knowledge is frozen at training time and can't cite its sources.
2 h 30 min
Coming later
Chapter 13
Agents
Language models in a loop, with tools.
Generating text isn't the same as getting something done.
1 h 30 min
Coming later
Chapter 14
Reasoning Models
Spending more computation on thinking — at training and at inference.
Some problems can't be answered well in a single forward pass of 'intuition'.
1 h 30 min
Coming later
Chapter 15
Multimodal AI
Images, audio, video and action in a shared representation.
The world isn't made of text.
1 h 30 min
Coming later
Part V
Judgment
How we measure, trust, secure and study these systems.
Chapter 16
Evaluation, Reliability, Safety & Interpretability
How do we know what a model can do — and what it's doing inside?
Impressive demos are not evidence. How do we actually know?
3 h
Coming later
Epilogue
Reading Research
Am I ready to start reading AI papers?
Papers are written to persuade. How do you read them critically?
1 h
Coming later