What this is
A free, open-source course that teaches product managers how AI actually works — enough to spec, evaluate, and ship AI features with a real mental model instead of vibes. No math, no framework, no login. Every concept is one plain-English explanation, one interactive widget, the product-judgment takeaway, and an interview-ready line.
It's built as a resource for the PM community, not a personal portfolio piece. Fork it, adapt it for your own domain, and share it.
Who it's for
Any PM, anywhere, whose product touches a recommendation, a chatbot, a search box, or anything that generates text — which, in 2026, is most of them. The examples are grocery-delivery flavoured so the concepts attach to something concrete; swap in your own domain when you present.
How it stays good as AI changes
The 25 concepts are foundational — tokenisation, attention, RAG, evals — and stay relevant for years. They're treated like a textbook: updated only when there's a genuinely clearer explanation, not because something new shipped this week. New ideas are vetted through discussion first (the "Suggest a concept" issue template) so the curation bar holds. The course gets better, not just bigger.
Contributing
Contributions are welcome and the bar is deliberately high — the audience is busy PMs who deserve a resource that respects their time.
- Typo / clarity fixes — open a pull request (edit the concept in
assets/js/concepts.js). - A clearer example or widget — open a PR, keeping the four-part pattern intact.
- A new core concept — open an issue first using the "Suggest a concept" template, so we align before any code.
See CONTRIBUTING.md for the full guide, including the note on writing good quiz questions.
Author & licence
Created by Puneet, distilled from a guided AI/ML learning series and built for the PM community.
Everything here — code and course content — is released under the MIT License. Fork it, adapt it for your own domain, and use it however you like; just keep the copyright notice. Attribution is appreciated but not required. Full text in the repo's LICENSE file.