📌 This is a trimmed extract of the master prep guide. The full version, every domain expanded with practice sets and resources, ships with the certification. We keep this cut public so anyone weighing the cert can see what it actually measures.
The AI-Native Developer Certification shows you can work with AI on real software. That means giving AI clear direction, questioning what it comes back with, catching its mistakes, and taking responsibility for whatever ends up shipping.
This guide covers what the exam is looking for, and what to practise before you take it.
You won't be asked to recall many facts. The exam is built around the things you deal with every day: a vague request, a plan an agent has proposed, or a failure in code you didn't write.
Whatever form a question takes, it's your judgement being assessed: what you noticed, what you'd check before acting, and what would change your mind. Underneath most of it is the question you face on every real piece of work. Is this ready, does something need fixing first, or is the task itself the problem?
You should be able to explain any call you make in a few plain sentences: what you decided, what you checked to get there, and what you're still not sure about.
There are six scored domains. The sections below cover what you need to be able to do in each one, plus a few ways to practise.
This isn't a framework exam, so there's nothing to memorise. Most of your prep should happen in real codebases, ideally inside the work you're already doing.
As you go, keep a simple log of the calls you got wrong or weren't sure about, and look back over it now and then. You'll learn more from the patterns in your own mistakes than from drilling what you're already good at.
1. Spec & Intent. Turning a vague request into something an agent could actually build, and catching what was left unsaid before it turns into a problem.
2. Architecture & Design. Judging a plan before anyone builds it. Will it hold at ten times the data, what happens when part of it fails, and does it really need all those moving parts.
3. Agent Direction & Context. Giving an agent enough context to do useful work without burying the important bits, and setting constraints you can check afterwards.
4. Review & Verification. Reviewing what an agent produces with a clear eye, without assuming it must be broken, or that green tests mean it's right.
5. Security & Risk. Finding the concrete security problems in what an agent produces, and proving they're real.
6. Debugging & Diagnosis. Finding the root cause in code you didn't write, using evidence to narrow it down instead of changing things at random.
Most of this will look familiar, and that's expected. The exam is built on judgement you already use. The prep is mostly about naming it, and sharpening the few spots you're less sure of.