The question goes in. The answer comes out.
They paste the homework, copy what comes back, and never read the middle. The part where someone works something out has been removed, and nothing marks that it is missing.
Conundrum Curriculum
When answers are free, we raise the children who ask.
When your child uses AI for homework — is the AI thinking for them, or are they directing the AI? That difference gets set between 8 and 11. We train the second kind.
- Ages 8–11
- 8 weeks a season
- 20–40 minutes on the main day
- A script for every activity
You have probably already seen this.
“It said so.”
Ask how they know, and the reason is that a machine told them. That is a citation, not a reason — and a child who cannot tell those apart will believe the next confident thing just as easily.
The question never gets sharper.
“Is this a good bridge?” has no answer. Good at what? Nobody taught them to ask — and the machine will answer the vague question anyway, fluently, which is the part that makes it stick.
None of this is a discipline problem, and none of it is unusual. It is a habit forming — in the four years habits like this form.
Knowing got cheap. Judging, asking and owning the result did not.
This does not replace what school teaches, and it does not compete with it. Knowledge is what makes error detection possible: you can only catch a confident machine getting something wrong in a subject you have real material in. School supplies the material. We train what a child does with it.
Judging
Commit to an answer, out loud, with a reason — before the reveal. The answer is not reachable until they do; that is a rule in the software, not a suggestion in a script. A guess kept in your head quietly becomes whatever the answer turned out to be.
Asking
Turn “is this any good?” into a question the world can settle. Find the number hiding inside it — good becomes how far, strong becomes how much weight. Some questions have no number in them, and knowing which is which is half the skill.
Owning it
Build a real object, out of things already in the house, that either works or does not. A hallway does not give partial credit and it does not care who is watching — which makes it the most honest feedback in a child’s week.
A week is a list, not a lesson plan.
Everything for the week sits on one page, each item marked with what it is, how hard it is, and how long it takes. You pick by how much of the evening is left. Nothing is locked in order — where a concept has to come first, the card says so in words.
This week — 6 items
- ★ The one thing —Two folds. Which one flies further?ConundrumWarm-up5 min
- Say it before you see itSessionWarm-up5 min
- How sure are you?SessionWarm-up5 min
- How far can a paper aeroplane fly?Do “How sure are you?” firstConundrumMain30 min
- A guess you can checkSessionMain5 min
- The four-throw notebookProjectMain40 min
★ The one thing
Every week has exactly one starred card, and it is always five minutes. On the week that falls apart, do that one. It is written to carry the week’s whole idea on its own, so a family who only ever does the starred card is still doing the thing. Skipping beats stopping.
Three kinds of thing, one rhythm.
5-minute session
Three to five a week
Installs one thinking concept — value against cost, estimating against measuring, right against kind. It is the glue, not the moat, and it is always installed before the week needs it.
Conundrum
Two a week — one big, one small
The main event. Four kinds: a binary that has a right answer, an open dilemma that has none, a research question they investigate, and a ranking of real things where the child has to state the criteria they used. The last one is the only activity here that asks a child to produce a standard rather than an answer.
Physical project
Three a season, a step each week
It has to end up as an object. The steps, the difficulty tiers and a sample outcome all come with it — and two of the three have a step designed to collapse. A plan that never fails teaches nothing about fixing one. The child sets the success criterion at the start and checks it against the thing at the end.
We are not anti-AI. We are anti-outsourcing.
Every activity declares which rung it sits on, and the script says it out loud — framed as design rather than morality. The point is never that the machine is forbidden. The point is who is doing the thinking this hour.
- 1
No AI
Reading, arithmetic, first drafts, and sitting with a hard question. The line in the script is “this hour is for the difficult version”, not “that would be cheating”.
- 2
Ask it, then justify it
They may ask. Then they answer one question: why do you believe it? This is the default rung for 8–11.
- 3
Collaborate
The machine researches; the child keeps the goal, the judgement and the trade-offs. Late in the age band only.
- 4
Direct it
Designing and checking a workflow. Past this age band — it is written down so you know where the ladder goes, not because it is in here.
You will not have to improvise.
You are the one running this, so the hard part of the design is not the child’s side. It is making the parent’s side survive a bad evening.
The actual words
Not “guide your child to reflect”. The lines, in order, including where to stop talking: “Why do you think so?” Then be quiet for ten seconds. Count them. Most of what a child says arrives in the gap after the first answer, and the gap only exists if you leave it.
You are not marking it
On an open dilemma the script permits one response, and it is “Interesting — why?”. No score, no correction, no better answer supplied afterwards. Several scripts are written specifically to stop you helping.
Nothing to collect
No streaks, no badges, no leaderboards, no points, no comparison with other families. That is a rule we hold even where it would obviously improve retention, because a child who is doing this for a token has stopped doing the thing it trains.
A floor, not a treadmill
Twenty to forty minutes on the main day, five minutes on the days around it, one project step a week. On the worst week: five minutes, once.
Week 0 is free, and it is the whole shape.
Not a sample chapter and not a trailer — a complete week, built exactly like a paid one: four sessions, two conundrums, one project, every parent script included. No account, no card, no email address. The only thing a paid week adds is the record — the uploads, the follow-up questions, the portfolio at the end.
Everything it needs is already in your house: paper, tape, string, a tape measure.
This one is about paper aeroplanes — and about saying what you think before you find out.
Season 0
One product, two ways to pay, and no monthly plan. A season is a one-time purchase that lines up with the eight weeks. A year is a subscription, and it says on the tin how many seasons that is.
One season
US$49.50US$99 after the first 100
USD · one-time, per season
8 weeks of material. Buy the next one when it opens, or don’t.
Join the waitlist
Seats open Monday 7 September 2026.
One year
US$195US$390 after the first 100
USD · billed yearly
5 seasons, 40 weeks of material, with breaks between seasons. 21% less than buying five seasons one at a time. Renews yearly; cancel any time.
Join the waitlist
Seats open Monday 7 September 2026.
The first 100 families pay half price, for as long as they keep going. In return we ask for two pieces of feedback: one mid-season, one at the end — that is the entry condition, not an ongoing one. Season 0’s pilot families take the first twenty seats.
- Seats open
- Monday 7 September 2026
- First week unlocks
- Monday 21 September 2026
- Season ends
- 15 November 2026 (8 weeks)
- What you get, when
- Week 0 and any completed past seasons the moment you pay; the new season’s weeks unlock one a week from the start date.
- Refunds
- 30 days from the day your season starts, in full, no reason needed
Writing to us puts you on the list; there is no form, no endpoint, and nothing stored until seats open on Monday 7 September 2026.
Questions parents actually ask
My child already uses AI for homework. Is this going to be a fight?
No, and nothing here asks you to take anything away. Rung 2 is the default: they may ask the machine, and then they answer one question about the answer. Most of the change is that the question now gets asked out loud.
Do I need to know the subject?
No. Your job is to read the lines and hold the tape measure. Several of the scripts exist specifically to stop you helping, because the useful part is watching a child work out something you have not handed them.
Will this help with school, or with applications later?
Not with tests, and we will not claim it does. This is not exam preparation and it is not competing with it.
What it builds does turn up later. An interview, a personal statement, a TOK essay or an extended essay, a scholarship panel — all of them reward a child who can define a question and defend a judgement. That is a side effect of the thing, not the thing.
How much time is this, really?
Twenty to forty minutes on the main day, about five minutes on the days around it, and one project step a week. Everything is on one page with its time printed on it, so you choose by what is left of the evening.
The worst week is five minutes, once. That is deliberate: the alternative we are competing with is skipping the week entirely.
What happens to my child’s data?
There are no child accounts and no fields for a child’s personal information — the only login is a parent email address, and a child is referred to by whatever name you type.
Uploads stay private to your family. We ask you to photograph the work rather than the child. Audio is transcribed and the recording is deleted within seven days, so what is kept long-term is text.
What if it turns out not to be for us?
30 days from the day your season starts, in full, no reason needed. The window runs from the day your season starts rather than the day you paid, because seats open two weeks before the first week unlocks and you should not lose part of it by buying early.
It comes to four Saturdays — long enough to see whether this fits your family, and deliberately longer than the product’s own cycle.