About this series

Who it's for

Business professionals and nonprofit employees — from your twenties to your eighties — who use or are about to use AI at work, and who want conceptual confidence rather than programming skills. No technical background is assumed; every term is defined before it's relied on.

How it teaches

Each lesson introduces exactly one idea and follows the same shape: a prediction you commit to before the reveal, one concrete analogy (with its limits stated honestly), a hands-on interaction that lets you cause the effect yourself, a realistic workplace example, a knowledge check that tests understanding rather than vocabulary, and one habit you can start the same day. Optional "Go deeper" sections carry the technical detail for those who want it.

You'll meet four recurring professionals throughout: Maya, an AI experience designer at a financial advice firm; Luka, a nonprofit fundraiser and grants consultant; Zane, an education and student-intake consultant; and Kerry, an allied health practitioner. Their situations are composites, built to be realistic.

Standards we hold ourselves to

Honesty about machines. We never describe an AI model as thinking, knowing or understanding the way a person does, and we never call it a database, search engine or brain. When an analogy helps, we also tell you exactly where it breaks.

Honesty about claims. Durable concepts are taught as concepts; perishable facts (product features, prices, regulations) are flagged and dated, verified before publication, and scheduled for review. Where models and providers differ, we say so.

Honesty about limits. The demonstrations on this site are simulations running in your browser, labelled as illustrative. We use simulated behaviour wherever reproducibility matters, and would only use a live AI where variability itself is the lesson — with clear notice first.

Balance. No doom, no hype. Capabilities, limitations and consequences, in proportion.

A model-neutral course

This series is model neutral. It does not speak for, favour or promote any one frontier model or any one open-weight or open-source model — it teaches the concepts that hold true across providers, not the habits of a single product.

Multiple models were used to help build it. That wasn't strictly necessary, but it was an important learning exercise in its own right. Now the meta part — how it was built:

Standing on the shoulders of giants

The course follows part of my own learning journey, and draws on the work of these people and pairs:

Best sources by your AI stack

If a particular layer of the stack interests you, here is who to read next.

Best sources by AI stack layer
Your lesson layerBest people to consult
Training dataAndrew Ng, Jeremy Howard, Sebastian Raschka
Tokens and generationAndrej Karpathy, Jay Alammar, Sebastian Raschka
Neural networks and attentionKarpathy, Alammar, Raschka, Howard
Context and promptingLilian Weng, Ethan Mollick, Simon Willison
MultimodalityJay Alammar, Chip Huyen
Search and retrievalSimon Willison, Eugene Yan, Jason Liu
Tools and actionsSimon Willison, Chip Huyen, Hamel Husain
Memory and agentsLilian Weng, Nathan Lambert, Simon Willison
Failure and safetyNarayanan, Kapoor, Mitchell, Willison
Intelligence and the futureChollet, Mitchell, Marcus, Lambert
Human judgement and workMollick, Ng, Narayanan and Kapoor

Where this is going

All 21 lessons are live — the series is complete. Optional accounts (synchronised progress, a completion certificate and a capability profile) are still in production, and will never be required for learning. Content follows Australian privacy, consumer protection and accessibility expectations while staying globally relevant.

This series is general information, not legal, financial or clinical advice.