Screen time you work for, not rot for.

The questions get harder as the round goes on, and that is the whole trick. Every question carries a difficulty from one to five, and the portal builds each round easy to hard, adapting the mix to how that child has been going lately. Two things hold whatever the mix says: a round always opens with the easiest questions it has, and a round is always passable. A kid who gets the first one right leans in. A kid whose first question is the hardest in the bank shuts the tablet and goes to find a screen you do not control. Confidence first, difficulty second, which is roughly how good teaching has always worked.

Every question carries an explanation, and that is the rule that shaped the whole thing. Not "correct: Africa", but a sentence written for the kid who got it wrong. Madagascar sits in the Indian Ocean off the south-east coast of Africa. That explanation is not decoration. It is the study material, it is what a child reads on the Read up page, and a bank without them teaches nothing. All 1,408 of them have one, and the checker refuses a bank that is missing any.

The road code bank exists because a teenager was learning to drive. That is the honest origin of most of them, and it is also the invitation. A bank on the thing a child in your house is actually into beats another worthy maths bank, and the project would rather have thirty niche banks than five respectable ones.

Every bank has a study page. It lists every question in the bank, the right answer, and the explanation, laid out to be read rather than tested. It is one tap from the quiz, and it is there before the first attempt as well as after a bad one, so a kid who wants to prepare can.

There is a badge for reading the whole study page and then passing a round of that bank, which is the one behaviour in the entire product most worth encouraging: read the material, then have a go.

If a child can only answer what they already knew, learn to earn pays for recall and not for learning. So a set of reference sites survives a total cut: Wikipedia and Simple English Wikipedia, Wiktionary, Britannica, Te Ara and the National Library of New Zealand, the Department of Conservation, NASA and the European Space Agency, the Natural History Museum, the Science Learning Hub, Maths Is Fun, NZ History, Te Papa, NZQA and the Ministry of Education's curriculum site, the Australian Curriculum, CSIRO and the Australian Museum, the British Library, the National Archives, Cambridge's NRICH maths problems, the Library of Congress, the Smithsonian Learning Lab, NOAA and the US Geological Survey.

A child out of time can go and genuinely learn something, then come back and cash it in. That is a very different bargain to "answer what you already knew".

A household is not a matched field. A seven year old and a fifteen year old are not playing the same game, whatever the scoreboard says. Rank children on total minutes earned or total quizzes passed and the youngest loses every day, forever, by construction. They have had fewer years to accumulate a total, and no quiz bank changes that. A feature meant to make learning feel good would become a daily reminder of a race they were entered into without being able to win.

Streaks were never built either, on purpose. A chain you must not break is one of the best-worn tools of the attention economy this whole project exists to push back against. Importing it to make badges sticky would be copying the opponent's playbook into the one part of Genkan meant to feel like the opposite of a phone that will not leave you alone. A streak that punishes a child for being sick, or busy, or just not in the mood, is not a reward.

And handicaps were rejected as well. "You get a bonus because you are younger" is technically fair and resented anyway, because it puts a formula between a child and an honest answer to "how am I doing". A clever kid reverse-engineers it, and then a parent has to defend it.

Three of them can be earned again for a different bank, so a curious kid who tries eight subjects ends up with eight small badges rather than one that stopped mattering after the first. And "bank mastered" was nearly a "pass this ten times" counter, which would have paid for repetition. Instead it checks whether a child has answered every question in the bank correctly at least once, ever, across all their rounds. What it certifies is that you have seen the whole bank and got it right, not that you clicked through it a lot.

The board never ranks. There is no second place and no ordinal language anywhere. Each category names whoever is leading it right now, then lists everyone else's own number beside their own name. A category with nothing useful to say is left off entirely rather than crowned on a coin flip, and a one-child household gets an honest placeholder instead of a staged board with an empty field. Every window rolls over a fortnight or a month, so this month's spotlight is not whoever got there first two years ago and has been coasting since.

And your AI can write banks too. Ask whichever assistant your family already uses to write a bank on ancient Egypt, netball rules, the periodic table or how a jet engine works, drop the file in, and it is on the menu that evening. The project ships a runbook telling your agent exactly what a good bank looks like: the age band, the ramp from level one to level five, the marking, the house rules on tone. There is also a briefing command that gathers what one child passes, avoids, gets wrong and has been reading about, and prints it with a prompt on the end for you to paste.

Recall is cheap and learning is not, and right now they pay the same. A child who already knows their times tables can pass rounds quickly. A child who spent forty minutes reading about the solar system and then passed a hard astronomy round earns exactly the same. Time spent reading is not measured, and it should be worth more.

There are obvious fixes, and none of them is built yet: pay more for the first pass of a bank than the tenth, pay more for a bank a child has been reading about, pay more for questions they previously got wrong, price the harder half of a round above the easier half. The risk to hold on to is that every one of those turns learning into an optimisation problem, and a clever child will optimise it. A system that feels fair gets used honestly. A system that feels like a puzzle gets treated like one, which is why none of them has been rushed.

There is a quiet argument in all of this. The feeds your kids use are built by people who are very good at making the next twenty minutes feel free. Learn to earn puts a small, fair price on those twenty minutes, paid in something the kid keeps. It is not a punishment and it is not a lecture. It is just the first place in their day where attention costs what it is worth.