Day 01 · Mon 4 JanSetting up
What is AI?
What AI can do, what it cannot do, and why it sometimes makes mistakes. The difference
between a machine that follows rules and one that learns from examples. No history lesson and no long list of
technical words.
JuniorThey play rock, paper, scissors against a bot that follows rules, then against one that learns their habits. They lose to the second one. That moment teaches more than an hour of explaining.
SeniorThey break a model on purpose, find the point where it fails and write down what happened. It stops them treating whatever the screen says as the truth.
Today they also sign the Learner AI Agreement: no full names, school, address,
medical or family details in anything they type. And if something on screen upsets them, say so straight
away. Nobody is in trouble.
They makeAn AI Can / AI Can’t card, based on their own testing
Day 02 · Tue 5 JanThe morning it clicks
Rock, Paper, Scissors
Train a simple AI model, then try to confuse it. This is how a model actually learns:
examples in, patterns out. Not magic and not a brain. Everything after this morning makes more sense because
of it.
Both groupsThey hold up their hand and take twenty photos of rock, twenty of paper and twenty of scissors on the laptop camera. The AI looks at all sixty and learns which hand shape is which. It works, and they made it. Then they try to confuse it on purpose. A hand held sideways, a shadow across the palm, scissors half closed. Now it picks the wrong shape, and they can explain exactly why. That is how AI ends up making mistakes.
Senior, going furtherThey put numbers on it. Twenty hand shapes the model has never seen, a results table in a spreadsheet, and accuracy for rock, paper and scissors separately. That is testing, taught without ever using the word.
If a model only knows what it was shown, what happens when it is shown the wrong things?
They makeA working rock, paper, scissors model, plus a written explanation of how they confused it
Day 03 · Wed 6 Jan
How to talk to AI
How to ask better questions and check the answers. There are no magic words. There is
structure: who it should be, what it needs to know, the job, the limits, an example. Then check what comes
back.
JuniorThe coach types on a shared screen and the group directs. “That is not clear enough. What do we add?” Their hands-on time uses safe tools built for children. Children under 13 never type into an open chatbot. The pattern they learn is simple: goal, details, style, check.
SeniorTheir own logins, and a game we call Prompt Golf: reach a set result in the fewest clear instructions. Then the harder version, writing instructions that work for somebody else’s task.
You gave it your name and your school. Where did that go?
They makeA Five-Prompt Playbook recording each prompt before and after, what they changed and what they learned
Day 04 · Thu 7 JanBuilding starts
Find a problem
Choose a real problem that could be solved with AI. Half the morning works out the problem,
half builds the first version. They learn what an AI app really is, as opposed to a chat window: something
goes in, a model does something, something comes out, and a real person at the end needed it.
JuniorThey pick from our list of approved projects, interview each other to find the real user, then build something that notices and reacts. A sorting game, a quiz helper, or a tool that decides whether a lunch is healthy.
SeniorThey name the person it is for, the problem, what that person does now instead, and one way to tell whether it worked. Then they build a real app without writing code.
This is the morning the program becomes real for a child. Every child leaves with something that
runs.
They makeA one-page problem plan and their first working prototype
Day 05 · Fri 8 JanWe email you tonight
Challenge Day
Work with others to solve a problem within a set time. Teams of three inside the coaching
group, one clear brief, and a demonstration at the end of the morning. The brief is narrow on purpose:
something at your school that wastes time, something that is hard for a younger child, something your
grandparents find confusing. Narrow briefs get better ideas.
Junior25 minutes on the brief, 50 minutes building with a break, 30 minutes of demonstrations.
Senior30 minutes on the problem, 60 minutes building with a break, 25 minutes of demonstrations.
This morning runs the other way round: about half an hour together, then well over an hour in
their own group, because you cannot build as a team over a broadcast. We also teach them how to give
feedback: one thing that worked, one question, one idea.
They makeA team demonstration and a feedback sheet
Day 06 · Mon 11 JanA guest drops in
AI that can take action
How AI can do more than simply answer questions. Work out the goal, the steps, the tools, the
checks, and where a person has to say yes. That last step is the whole point. A working engineer or designer
joins for twenty minutes to talk about what they did with AI yesterday, and the children ask the
questions.
JuniorThey map out a helper that plans a birthday party. What does it need to know? What can it decide by itself? What must it ask an adult first? Children are very good at this.
SeniorThey build a two-step process where the first step feeds the second. Then they break it, by giving it something it is not expecting.
Which step would you never let it do without asking you first?
They makeAn AI Helper Map, with the point marked where a person has to approve
Day 07 · Tue 12 JanThey test it on you
Build an AI Assistant
Create a useful AI helper for someone you know. A second, more complete build, using the
multi-step idea from Day 6.
The rule that makes this day workBefore the morning ends they have to test it on a real person outside the group. A parent, a brother or a sister. And they have to watch without helping. This is the part where you come in, and it takes about twenty minutes.
Your tester assumed it was always right. Whose job was it to tell them otherwise?
They makeA second prototype and user testing notes from watching someone use it
Day 08 · Wed 13 JanFamily rules come home
Can you trust AI?
How to check information, spot misleading content and use AI responsibly. By now they have
built four things and watched all four go wrong, which is why this day lands. Three parts: checking what is
true, spotting fake images and video, and AI in schoolwork.
Checking what is trueSix statements. Three are true, three were written by AI and are false, and all six sound believable. Then a habit they can use for life: who says so, can you find it twice from different places, and does the source have a reason to lie?
Fake images and video: spotting onlyChildren never make a fake. Not as a demonstration and not to show how it works. We look at real examples, the giveaways, why permission matters, and what it feels like to have your face used without it.
AI and schoolworkWhere the line sits between using AI and handing in AI. Each child writes their own rules for responsible AI use, in their own words.
They makePersonal AI rules and the Family AI rules sheet that comes home to you
Day 09 · Thu 14 JanLonger task tonight
Innovation Lab
Improve the final project and get it ready to show. Nothing new is taught. Coaching,
building, testing and tidying up, with coaches running short one-to-one sessions. Each child completes a
one-page project summary and a five-slide presentation from a GIofAI template, and practises their talk at
least once.
Junior20 minutes on how to structure a talk, 65 minutes building and testing, 25 minutes practising. Their talks run two minutes.
Senior25 minutes on structure, 70 minutes building and testing, 25 minutes practising. Their talks run three to four minutes.
Tonight’s activity runs up to an hour. It is the one evening in the fortnight worth
keeping free.
They makeA project one-pager and a five-slide presentation, practised
Day 10 · Fri 15 JanYou are invited · runs 2h30
Final Showcase
Present the project and explain how it works, what it does well and what it cannot do. Your
child presents to their own group: their coach, the other children in the group, and their families. Not a
hall of two hundred strangers. That afternoon, every group sends its best projects to the Grand Showcase,
shown to guest mentors in front of the whole program.
How the morning runsToday is mostly small group, because you cannot present to a room over a broadcast. Welcome and house rules (15 minutes), the presentations (75 minutes), a break (15 minutes), feedback from guest mentors that is specific, encouraging and never ranked (30 minutes), then awards, certificates and a close (30 minutes). The Grand Showcase follows in the afternoon.
Six main awards, and a named mention for every single child. A handful of
prizes on their own would send most of the room home feeling like they lost, so we do not run it that way.
They take homeA certificate, their project one-pager and final presentation, a link to their work, their AI rules and the family rules