Home / For individuals / For parents
What your child will actually learn.
Written plainly, including the parts most programmes leave out.
Most AI programmes for young people teach them to use AI — how to prompt a chatbot, how to make an image. That is a skill with a short shelf life, and your child probably already has it.
This teaches the opposite thing: how the machine actually works. By the end your child will have built a neural network from the mathematics up, and will be able to explain why it reaches the answers it reaches — and, more importantly, when it is wrong.
Year one
Fundamentals, data, and their first Python programs. Then probability and statistics, and a first complete model built on probabilistic thinking.
The middle years
Pattern recognition on real data. Then weights, vectors and matrices — how an AI decides what matters — and graph theory to build a two-layer neural network.
The final years
Multi-dimensional data, forward and backward propagation in a three-layer network, and finally production tools: Keras and TensorFlow.
The full module-by-module breakdown is on the programme page.
"Will they actually stick with it?"
This is the right question, and we would rather answer it honestly than pretend it away. Most online courses are not finished. The research is fairly blunt — across 221 studied courses, completion ran a median of about 13%, and longer courses did significantly worse than shorter ones.
That is a format problem more than a willpower problem, and it is why this is built the way it is. Lessons run two to seven minutes. Every level ends in a credential, so value banks early rather than only at the finish. And if your child stops part-way through, they keep everything they have earned and can rejoin at that level later.
We're not going to tell you this helps with college admissions.
Admissions officers are fairly direct about it. Asked how much weight summer programmes carry, Brown's dean of admissions answered: "Zero." A former Penn admissions officer said she did not place much value on them.
What a student actually gets here is a working model they built, a credential that verifies independently, and the ability to have a real conversation about how machine learning works. Those are worth having on their own terms. If they also happen to make an application essay more interesting, that is a side effect, not the pitch.
A production neural network built in Keras and TensorFlow · a capstone project they can show · five verifiable credentials · working Python · and a genuine understanding of when a model is wrong, which is the part that matters most.
What you can see
You will know how it is going, not just that it is going.
Your child begins with a 30-minute Foundations Challenge that records where they started. After that, every module ends in a short marked check and every level ends in something they built, graded against a rubric. All of it sits in one record you can look at with them.
If their school runs the programme, that same record is what their teacher sees, and it is what you will be handed at a parents' evening — one page, written for someone with no AI background. If they are doing it on their own, you can be added as a guardian on the account and see the same thing.
The credentials are the part worth knowing about. They are digitally verifiable, with the evidence attached — so when your child says they built a neural network, anyone they tell can check.
We collect the minimum the programme needs, never advertise to children or sell their data, and you can ask to see, correct or delete your child's record at any time. The platform is designed against IEEE 2089, the age-appropriate design standard for children's services. The plain-language version is on the student data page.
Watch one with them.
A few minutes of the real teaching. No sign-up, no card — see if it holds their attention.