AI Trainers.
The machine learns from examples — and bad examples make a bad machine. Trainers teach a model, break it on purpose, and fix it. They fact-check as a sport and start to see the tricks inside the apps they use every day.
- Grades 6–8
- 10 sessions
- Schools & cohorts
Seven skills, all provable.
Not opinions about AI — things a child can do, and show you.
Train & test a model
Build a classifier, then explain why it failed — the idea of overfitting, felt.
Fact-check reflexively
Treat every confident answer as a claim to verify — and bring the proof.
Prompt with RTCF
Structure an ask, then repair broken prompts: missing role, missing context, vague format.
Run a research pass
A four-step loop they can walk: Question, Gather, Check, Show.
See the tricks
Spot the attention and spending tricks inside feeds and free games — and retrain a feed on purpose.
Build a mini app
Make a working app through language alone — then change three things by asking precisely.
Redesign, without tricks
Rework a product for a younger user — helpful by design, with none of the manipulation. The point where the six skills above become one thing a child can show you.
Ten missions, one arc.
Each is 50 minutes on the same rhythm: hook, do, fail & fix, make, share.
Meet the Machine
AI learns from examples. They train a Teachable Machine, then break it with lazy examples. A model is only as good as its examples.
Train Your Monster
Good examples make good AI. They train, test on an unseen round, autopsy the failure (overfitting: it memorised the homework instead of learning the subject), retrain, and pass.
Catch Zakoot Lying
Fact-checking as a sport. Score points for catching confident hallucinations and proving them, ending in a courtroom-lite "guilty or not."
Meet Your Chatbot
Talk to a machine with a personality. RTCF, the tone dial (one email in four tones), and the "cheating" test: AI is cheating when it does the thinking that was the point.
Prompt Quest
Fix broken prompts and explain why they failed. The explanation is the assessment.
The Research Path
Question, Gather, Check, Show — walked physically as four stations. Check rule: find at least one AI mistake.
Who's Playing Whom?
The feed and the games that play you back. Every pause is a vote; the feed optimises attention, not you. They learn to retrain a feed on purpose.
My First App
You are the boss, the AI is the builder. Build a mini app by prompting, then change three things by asking precisely.
Game Redesign Lab
Redesign a broken game for younger kids, without the tricks. Design ethics taught by doing.
Make It Real
From research to a finished artifact and poster; a Canva sprint and Song Wars, where the most detailed brief wins.
The characters who run the missions.
Instruction arrives as a character walking you into a room — never a slideshow. At the Trainer level the ship gets stranger, and a little more honest about how the apps work.
You'll spend real time with an astronaut in the Engine Room, teaching the machine by example. And you'll meet the mystery mask on the Observation Deck, whose whole trick is a whisper: "I know what you'll watch next."
What a Trainer graduate can do.
A Trainer graduate can train and test a model and explain its failures, fact-check AI with proof, prompt with structure and repair broken prompts, run a four-step research pass, see through the tricks in feeds and free games, build and edit a mini app through language, and redesign a product for a younger user without manipulation.
Trainers, one price.
Also available: 1-to-1 coaching and neurodiversity support.
Ready for the next rung?
Trainers who've learned to feed and check a machine climb into Builders — where teams run as start-ups, direct AI, and ship a real project.
Train the machine. Then check it.
Join the next Trainers cohort, or bring the level into your school.