Skip to content
Road to Intelligence

An interactive guide to AI, from rules to reasoning

The Road
to Intelligence

From machines that followed rules to systems that learn, reason, retrieve, create and act.

Not a list of a hundred topics, and not a prediction of where AI is going — a map of the road it has actually travelled, one idea at a time, each told through the problem it was invented to solve. Built for engineers who use AI every day and want the theory underneath to finally click.

Start the core pathExplore the AI timelineOpen the Lab →

Follow one thread

From a list of numbers to the mechanism inside every LLM.

Every idea here depends on a few before it. Follow this chain and self-attention stops being an equation and becomes something you could have invented.

  1. 01VectorsModels can only do arithmetic. A sentence, a photo or a customer is not a number.Must knowKnow well
  2. 02Dot ProductOnce things are vectors, we need a cheap way to ask: how similar — how aligned — are these two?Must knowImplement
  3. 03EmbeddingsOne-hot vectors give every word its own dimension, so 'cat' and 'kitten' are exactly as different as 'cat' and 'carburettor'.Must knowKnow well
  4. 04AttentionAn encoder–decoder RNN squeezes a whole sentence into one fixed-size vector before translating it, so long sentences lose information.Must knowKnow well
  5. 05Self-AttentionIn a recurrent network, information between distant words must travel step by step through every word in between — slow to compute and easy to lose.Must knowImplement

You do not need to memorize everything

Every idea tells you how much it deserves.

The core path covers the must-know ideas in about 40 hours. Deeper material is one click away when you want it — and stays out of your way when you don't.

How to read the labels →
Must know
Understand properly before graduate study. Example: self-attention.
Should know
Improves research literacy. Example: grouped-query attention.
Frontier
Know it exists and why. Example: test-time compute scaling.
Understand
Know what it is and why it exists.
Know well
Explain the mechanism and its trade-offs.
Implement
Code a small version yourself.

The journey

Five parts, sixteen chapters.

  1. Part I

    Foundations

    From hand-written rules to machines that learn.

    • 1What Is Artificial Intelligence?
    • 2The Math Toolkit
    • 3Machine Learning
    • 4Neural Networks

  2. Part II

    Branches & Language

    Seeing, hearing, acting — and the long road to understanding text.

    • 5Vision, Speech & Reinforcement Learning
    • 6Language Before Transformers
    • 7Transformers

  3. Part III

    Large Language Models

    Prediction at scale, and how it becomes an assistant.

    • 8The Rise of Large Language Models
    • 9How an LLM Is Actually Built
    • 10From Base Model to Assistant
    • 11Inside Modern LLMs

  4. Part IV

    Systems

    Retrieval, tools, agents, reasoning and many modalities.

    • 12Embeddings, RAG & the LLM Application Stack
    • 13Agents
    • 14Reasoning Models
    • 15Multimodal AI

  5. Part V

    Judgment

    How we measure, trust, secure and study these systems.

    • 16Evaluation, Reliability, Safety & Interpretability
    • EReading Research
See every chapter →