Practice Labs

Reading about AI is not the same as building with it. These labs make you write the thing, then check whether it actually works.

💻

Code labs

Write real Python in the browser and have it graded against hidden tests. CPython runs on your own machine through WebAssembly - nothing is uploaded.

✍

Prompt lab

Practise prompting against a real model using your own API key, and get your attempt scored against a rubric. Your key stays in your browser.

🎬

Decision scenarios

Branching situations from AI governance, security and rollout work. Choose, see the consequence, and find out what a practitioner would have done.

Code labs

Each lab is a single function with a handful of tests. They are drawn from the parts of real AI engineering that people get wrong first: chunking, cost maths, retries, redaction and retrieval.

Why in-browser. A hosted code sandbox means accounts, containers, queues and a bill that grows with traffic. Pyodide moves all of that to the visitor's own machine: the interpreter is downloaded once (about 10 MB) and cached, your code never leaves the browser, and the labs keep working offline and cost nothing to run. The trade-off is that packages with C extensions are limited to what Pyodide ships, which is why these labs use the standard library only.