Purrx teaches a few concepts at a time, then hands you real starter code with the interesting parts missing. Write it, run it, get checked, level up — from your first API call to an agent you built yourself.
Free · Most lessons read without an account · Bring your own free Gemini API key
A playground is not a place to flounder. You are taught the idea first — the exercise is where it becomes yours.
A short, focused explanation — two or three new ideas, not a textbook chapter. Worked examples included.
Every lesson hands you real starter code with TODOs. You write the part that matters, in a real editor.
Your code runs for real. Checks tell you exactly what passed, what didn't and why — then you level up.
One complete course, 19 lessons, from "what is a token?" to a support assistant you built yourself.
Go from 'what is an LLM?' to building a working AI agent — writing real code against the Gemini API in your browser. Concepts first, then a playground that checks your work.
Build an AI slide generator that writes real, editable .pptx files, step by step: a layout language, text fitting, themes and components, Gemini and LangChain, then retries, fallbacks, streaming and tests.
The same format, applied to the rest of the stack — including the testing tools most learning platforms skip entirely.
Courses aren't split by audience. Each one starts from first principles and climbs to advanced material, so you enter wherever your gap actually is.
You call the actual Gemini API with your own free key. Nothing is faked, which means nothing you learn has to be unlearned later.
Model output changes every run, so we check what your code did — which call, which parameters, what shape came back. The way you'd test it in production.
Your Gemini key is encrypted before it's stored and only ever used to run your exercises. Playground code runs in a sandboxed worker that talks straight to Google.
XP, levels, badges and streaks, saved to your account so they're there on every device. Export it all as JSON any time — it's yours.
Every lesson ends somewhere useful: a working call, a parsed schema, a tool the model can use, an agent loop you wrote yourself.
Ready to write your first line?
Lesson one takes about six minutes, including getting your key.
