Learn AI — from the ideas to the tools to the certs.

Three concept tracks, four assistants

☺ Like you’re 10: First you learn the basics. Then, if you are curious, you can peek at how the models are built inside. And then you learn the tricks to actually ship one that works well for real people — and you can hop on whichever of four assistants you like. The basics carry over to all of them.

The course has three provider-neutral concept tracks — AI Foundations (the core ideas), AI Engineering (Applied) (the craft of shipping foundation-model apps to production), and AI Advanced (the ML science beneath and beside LLMs) — plus four assistant tracks that go deep on the tools people actually use: GitHub Copilot, Anthropic Claude, OpenAI ChatGPT, and Google Gemini. The concepts transfer to every tool, so nothing you learn is wasted.

Meet the AI Academy

☺ Like you’re 10: To make the big ideas stick, this course is taught by a team of animal friends. Each one is an expert at one job — so the moment the octopus shows up, you’ll know it’s time to talk about lots of helpers working at once.

Every concept in this course has a recurring character who is that idea — 🦉 Professor Owl teaches, 🦊 Foxy asks the questions, 🐘 Ellie remembers, 🦫 Benny builds, 🐙 Olly runs the agents, 🐢 Timmy checks the work, 🐿️ Nutty gathers the data, and 🐦 Pip carries the messages (plus a supporting cast). Watch for the “Your host for this topic” strip and the little AI Academy scenes in each lesson.

Meet the full roster and see how it works → The AI Academy

Four certification tracks — one per assistant

☺ Like you’re 10: A certification is like the swimming badge you get sewn on your towel after you pass the deep-end test — it’s an official way to show other people you really can do the thing. Now there’s a badge for each helper.

When you’re ready to prove it, aim at a credential — there’s one for each assistant. Every hub maps the exam onto this course and folds in the study aids (Self-Check & Flashcards each have a ★ filter per exam).

The one mental model that makes everything click

☺ Like you’re 10: Picture a dial that goes from “I do it and the helper just suggests” all the way to “I tell the helper to go do the whole thing.” You choose how far to turn the dial depending on how much you trust the helper with that job — and the helper is like a team captain who picks the right player for each task.

Before any features, internalize this. A modern AI assistant is not a single feature — it is a spectrum of autonomy. Every capability sits somewhere on the line above. When you’re unsure which to use, ask one question: how much autonomy do I want to hand over for this task, and how much do I want to stay in the loop? The whole field is built so you can slide along that line and delegate more as your trust grows.

The second model: these tools are increasingly an orchestration layer, not a single AI. Copilot routes your task across underlying models (GPT-class, Claude-class, Gemini-class) and agents; Claude Code and the others do the same. You bring the intent; the tool picks the engine and the workflow.

What’s inside

☺ Like you’re 10: The lessons are stacked like levels in a video game — each one uses what you learned in the level before, so you keep getting stronger instead of starting over.

The core is a progressive ladder of concepts — each topic builds on the last, taught provider-neutrally (Copilot, Claude, and GPT side by side) so the ideas transfer everywhere.

…plus interactive labs, a self-check, flashcards, a common-mistakes guide, a curated Further Reading library, and a live AI News Feed — all in the sidebar under Practice & Reference.