aī
Interactive course

How AI actually works.

The fundamentals of modern AI, explained with hands-on demos. Tokens, prompts, embeddings, agents, RAG, reasoning, all the things you keep hearing about, but with something real to poke at on every page.

Start the course → See all lessons

16 lessons · 13 interactive demos · ≈ 95 minutes · free, no sign-up

The course

Progress saves in this browser. No account.

Module 01

What is AI?

  1. What even is AI?
    A friendly tour of what "AI" actually means, and what it doesn't.
  2. What is a Large Language Model?
    Meet the "really big autocomplete" that powers ChatGPT, Claude, and friends.
  3. How does an LLM learn?
    A peek inside "training", how billions of guesses become a knowing-ish AI.
Module 02

How LLMs work

  1. Tokens: the puzzle pieces of language
    AIs don't see letters or whole words. They see "tokens." Let's break some.
  2. Prompts: how to talk to an AI
    The way you ask is half the answer. Try it with a real model.
  3. Temperature: the AI's dice
    Why the same question can get different answers. Turn the dial yourself.
  4. Embeddings: turning meaning into numbers
    How an AI knows that "dog" and "puppy" are friends, but "dog" and "spaceship" are not.
Module 03

Strengths & limits

  1. Hallucinations: when AI sounds confident but is wrong
    The most important AI failure mode, explained. Spot one yourself.
  2. Bias: AI learns from us, including our mistakes
    Why a model trained on human text can repeat very human unfairness.
  3. Using AI well: when to trust it, when to double-check
    A short, practical guide to making AI your teammate, not your boss.
Module 04

AI that does things

  1. Tool use: giving the AI hands
    An LLM by itself can only talk. Tool use is what lets it do things like check the weather or do real math.
  2. Agents: AI that works for you
    An agent is an LLM that doesn't just answer, it plans, acts, and keeps going until the job is done.
  3. Oversight: keeping agents on a leash
    How to let an AI agent help without letting it cause real damage. The pilot-and-autopilot rule.
Module 05

Beyond text

  1. Multimodal AI: more than just words
    Modern AI doesn't just read and write. It sees, hears, and creates pictures, sounds, and video.
  2. RAG: looking it up first
    How AI gives accurate, sourced answers, by reading the right page before it speaks.
  3. Reasoning models: AI that thinks before it answers
    Why some 2026 models pause, scribble in a notebook, and then answer. And why that's a big deal.