From questioning AI to building with it, in five levels.

The Northmind Academy curriculum. A step-by-step path for teens 12–17, based on UNESCO's AI Competency Framework for Students and teaching methods tested in real classrooms.

5
levels
12
classes each
2–3
teens per class
12–17
years old

Built on

  • UNESCOAI Competency Framework for Students, 2024
  • Australian GovernmentFramework for Generative AI in Schools, 2023
  • 45 sourcesstudies, reviews and reports behind every design choice

Every level ends with something they made.

Each level is 12 weekly classes. Levels 1 to 3 each stand on their own. Levels 4 and 5 are for teens who want to go further.

  1. No code

    Level 1: AI Foundations: Think with AI

    How AI learns, why it gets things wrong, and how to stay safe using it.

    Ends with

    a game they draw and control with their hands, and their own rules for using AI

    Start here
  2. No code

    Level 2: Build without Code

    Build and test their own AI tools, and write clear instructions for AI.

    Ends with

    their art set against an AI's, or a story character they build two ways

  3. Python

    Level 3: Code with AI: Python Foundations

    Write Python programs, and check AI-written code before using it.

    Ends with

    their own Python game or generative art, explained line by line without AI

  4. Python + data

    Level 4: Machine Learning Lab

    Train models on real data, measure them honestly, and check for fairness.

    Ends with

    a model trained on their own doodles, sounds or music, with a report on its limits

  5. Python + AI

    Level 5: Build Real AI Apps

    Design, build and test an AI-powered app that solves a real problem.

    Ends with

    a working creative or useful AI app, presented on demo day

The PATH method. Pause, attempt, talk, hold.

How every class works

  1. Pause

    Decide whether AI is worth using here. A reasoned “no” earns credit.

  2. Attempt

    Work out your own answer first.

  3. Talk

    Use AI for hints, not answers, where the task allows it.

  4. Hold

    Check the output and explain it in your own words. Hold it to account.

Inside a 60-minute class

  1. Recall5 min
  2. Mini-lesson10 min
  3. Build · Pause, Attempt, Talk, Hold30 min · Pause, Attempt, Talk, Hold
  4. Show & explain10 min
  5. Reflect5 min

What makes it different

We test them with the AI switched off.

It's the only honest way to see what your teen has learned, so every level has AI-off checks.

  • finds the AI's mistake
  • explains the fix
  • solves it alone

The AI gives hints. Never the answer.

Every AI tool we use in class is set up to guide your teen, not to do the work.

Teen: Just tell me the answer.

AI: Here's a hint instead: what do the wrong ones have in common?

Their idea. Their art. Their project.

AI is something they build, train or test, never the author. A maker's log shows how the work got done, and we never use AI detectors.

Room to disagree.

Level 1 opens with their own verdict on AI. We talk honestly about cheating, who's behind AI, what it costs and what it means for artists, and every teen writes their own rules for using it.

Two or three teens, one tutor who knows them.

Groups are split by age, 12–14 and 15–17, and the tutor stays the same for the whole level.

Safe from the first class.

Clear rules for AI, for contact and for recordings, written down and shared with you.

  • Parent consent, every student
  • Under-13s use AI only in class
  • No private chats
  • Recordings deleted when the level ends
  • Materials made by people

We teach the way the evidence says works.

The research behind it

  • 17%

    lower exam scores for students who practiced with an answer-giving chatbot. Those with a hint-only tutor didn't fall behind.

    Bastani et al., PNAS 2025 · about 1,000 students

  • 2 : 1

    Online tutoring with two students per tutor raised test scores and maths grades for 12–15-year-olds.

    Gortazar et al., 2024 · randomised trial

  • 435

    studies, reviewed together, show that specific feedback (what's wrong, what to try next) works far better than points and badges. Our reports work that way.

    Wisniewski et al., 2020 · meta-analysis

  • Alike

    AI ideas made each person's story rate as more creative, but made everyone's stories more alike. So the ideas come before the AI.

    Doshi & Hauser, Science Advances 2024

No course exactly like ours has been studied yet. That's why we measure every student's progress without AI and share it with you each month.

See it for yourself

Book a free 30-minute lesson. Your teen tells us what they think of AI, tries a real task, and you see exactly how we teach.

Free, online, no sales call.

Book a free 30-minute trial