Read this first
Artifact: the decision to keep reading, or close the tab
1. What this course is
This is a course on the engineering patterns you need once your AI demo has to become a system.
Most resources on "building with AI" stop at the first call: how to set up an SDK, how to write a basic prompt, how to handle streaming. That's where the real work begins, not where it ends. The moment your demo has to run for ten thousand users, handle errors gracefully, stay under a cost budget, and produce trustworthy outputs — that's when patterns matter.
This course is the catalog of those patterns. Fourteen of them, grouped by the problem they solve.
2. The thing this course is explicit about
Classical software design patterns — Strategy, Adapter, Factory, Observer, all of it — still apply to AI code. Your surrounding code still wants those, and we're not going to retread them. If those aren't fluent for you yet, two excellent free references:
- Refactoring.Guru for plain-language explanations: refactoring.guru/design-patterns
- Gang of Four (the original book) if you want the canonical text: Design Patterns: Elements of Reusable Object-Oriented Software
This course is about the patterns layered on top of classical software — the ones that only exist because there's now a model in your system.
3. Who it is for
You, if you have ever:
- Built a working AI demo that fell apart the moment a real user tried it.
- Written a prompt that returned valid JSON 90% of the time and shipped it anyway.
- Spent three days debugging why your "agent" got stuck in a loop.
- Looked at your AI feature's bill and realized cost-per-request was 50x what you budgeted.
- Had to explain to a stakeholder why the model "just" got something wrong, with no way to prevent it next time.
If none of that applies, you already know the patterns this course names. Read it for vocabulary, or close the tab.
4. What it is not
This is not a tutorial on a specific SDK or model. The code examples are TypeScript (using the Vercel AI SDK syntax, because it's the cleanest current standard) but every pattern applies to any SDK in any language. The Python equivalents are one library swap away.
This is also not a survey of every AI library on the internet. It's the patterns, with one canonical example each.
5. The next page
The next page is the map: which problem each pattern solves, in what order to read them, and how they compose.