Two complete courses. 25 units. ~69,000 words. Written for adults and teenagers with no technical background — no maths, no code, no prerequisites.
Fork it, translate it, teach from it. Free for any use including commercial, with attribution.
Read it here in plain markdown, or take the interactive version with audio narration and AI-graded exercises at thinkmodel.ai.
How AI actually works, why it fails, and how to get good at using it. Read the course → · 12 units · ~2,700 words each
| Unit | ||
|---|---|---|
| 01 | Pattern Machines | Your feed, your autocomplete, your recommendations: all the same trick |
| 02 | Nobody Programmed This | Nobody wrote the rules for recognising a cat. Where the rules actually come from |
| 03 | Where the Knowledge Comes From | The knowledge comes from data, and the data has holes |
| 04 | It Doesn't Know Anything | No understanding, no database — just the most likely next chunk of text |
| 05 | This Stuff Isn't Free | Warehouses of chips, small-town electricity bills, and the cost of a three-second answer |
| 06 | Words Are the New Code | The same question, asked two ways, gets two different answers |
| 07 | Context Is Everything | Why an identical prompt lands differently depending on what surrounds it |
| 08 | Be the Judge, Not the Audience | AI is equally confident when it's right and when it's wrong |
| 09 | AI That Does Things | From answering to doing: what changes when the model can act |
| 10 | Build Something Real | Ship a working thing you can show someone, without writing code |
| 11 | Your AI Toolkit | Hundreds of tools, a new one every week. How to pick and how to keep picking |
| 12 | What's Yours | If it can write the essay, what's the point of you? |
Stop chatting and start operating: configure workspaces, choose models deliberately, delegate to agents with guardrails, and run systems that keep working when you are not watching. Read the course → · 13 units · ~2,800 words each
| Unit | ||
|---|---|---|
| 00 | Start Here | What changes when you stop using AI and start operating it |
| 01 | From Chats to Systems | A workspace that briefs the model for you, instead of retyping context |
| 02 | The Right Brain for the Job | Choosing a model deliberately, and what you trade away either direction |
| 03 | Research Like a Team of Ten | Delegate the reading. Never delegate the believing |
| 04 | The Terminal Isn't Scary | When your own files become the thing the agent works on |
| 05 | Inside the Harness | Why the same model behaves differently in two different tools |
| 06 | Let It Run | When it's safe to let an agent run alone, and where the checkpoint belongs |
| 07 | Give It Your Keys (Carefully) | Private data, untrusted input and outbound actions must never meet |
| 08 | While You Sleep | Turning something you keep redoing by hand into something that runs without you |
| 09 | You're the Art Director Now | When AI makes the visuals, your job becomes direction rather than production |
| 10 | Your Own Private AI | Running a model on your own machine versus renting a frontier one |
| 11 | Break Your Own Stuff | An untested AI system is already broken somewhere. Go find it |
| 12 | Stay Frontier | Staying current by watching instruments instead of following influencers |
Think about how you recognize a friend across a crowded room. You don't consciously analyze their height, hair color, posture, and walking style. You just... know. Your brain has seen them so many times that it built an internal model — a pattern — and now it matches that pattern instantly, even from behind, even in bad lighting.
AI does the exact same thing. But instead of using eyesight and human experience, it uses data. Lots of data. An almost incomprehensible amount of data.
— Unit 01, Pattern Machines
Every unit opens with something you already know and builds to the idea. Nothing is named before it has been experienced. Each one ends with a glossary of the terms it introduced, and quizzes keep their answers behind a fold so you still get to think first.
If you want to learn — open 101 unit 01 and read in order. About 8–13 minutes of reading per unit — the same estimate the site publishes. 102 assumes you have the ideas from 101, but re-explains what it needs, so you can start there if you already use AI daily and want to build with it.
If you want to teach it — there are facilitator guides for each course, session by session, with what to protect time for and where discussions go wrong: 101 · 102. The licence covers classroom and commercial teaching; you do not need to ask.
If you want to translate it — this is the most useful thing you can make, and
the units are built for it: self-contained prose, no build step, no tooling.
Work from the plain markdown in 101-fundamentals/ and 102-operator/. Two
things to keep: the unit numbering, because units cross-reference each other by
number, and the marked notes where a unit points at an interactive exercise that
only exists in the online reader — the surrounding material stands on its own.
CC BY 4.0 — copy, translate, adapt and teach from this, including commercially. Credit it and say if you changed it. That is the whole deal. Full text in LICENSE.
Credit it like this:
ThinkModel — thinkmodel.ai
The videos are CC BY 4.0 too. Each unit links one, on YouTube under the same terms, so a translation can subtitle or re-host them rather than sending readers back here.
The converter and tests in tools/ are MIT.
What the licence cannot cover: several units quote studies and link work by other people — CGP Grey, Andrej Karpathy, Joy Buolamwini among them. That belongs to its creators and stays under their terms. Quoting and linking is fine; relicensing it is not mine to offer. The ThinkModel name and logo are likewise not licensed for use in a way that implies your adaptation is the original.
AI facts date fast, and these units carry specific numbers, studies and model names. If something is wrong or has gone stale, open an issue — corrections flow back into the units every learner reads, on the site and here.
source/ holds the canonical text in the directive dialect the interactive
reader renders (:::quiz, :::socratic, {{term:name|definition}}), documented
in AUTHORING.md. The plain courses above are
generated from it, so fix things in source/ and regenerate:
node tools/build-plain.mjs source/course-101 101-fundamentals
node tools/build-plain.mjs source/course-102 102-operator
node --test tools/build-plain.test.mjsTranslating? Ignore all of this and work from the plain markdown.