Our review gaps stop growing at 90 days. Anki's default maximum is one hundred years. That single difference is the clearest illustration of who each system is built for.
At a glance
| aitutors.me | Anki (default) | |
|---|---|---|
| Maximum gap | 90 days | 100 years |
| Effect on a well-known card | Returns once a term | Effectively never returns |
| Built for | A child assessed termly | An adult who needs it for decades |
| Cost | ~4 reviews a year per mature card | Almost none |
| Risk | A little wasted effort | Material quietly vanishing before an exam |
What a ceiling actually does
Spaced repetition works by widening gaps. Left alone, the widening does not stop: multiply an interval by roughly 2.5 each time and you pass a year after seven or eight successful reviews, then a decade, then a lifetime. Anki does not consider this a problem, and for its typical user it isn't. A doctor memorising drug interactions genuinely does want a card scheduled for 2041.
We cap the widening at 90 days. On our ladder a card that keeps getting Got it goes 1 → 6 → 15 → 38 → 90 days, and then stays at 90 for as long as it lives.
Why a term is the right horizon
Three reasons, in order of importance.
A KS3 child is assessed termly. That is the real deadline the schedule is serving. The research on gap length says the optimal spacing scales with how long you need to retain something — roughly ten to twenty per cent of the retention interval. For a term-length horizon, that puts the useful maximum in the range of a few weeks to a few months. Ninety days sits at the top of that range rather than outside it.
Curriculum content comes back. Ratio in Year 8 is not finished business; it reappears in Year 9, and again at GCSE. A card scheduled four years out would surface after the exam it was meant to support.
Children's collections are small and young. The efficiency argument for enormous intervals matters when you have 10,000 cards and every unnecessary review costs real minutes. With a collection of a few hundred, the arithmetic is gentle: a card at the ceiling costs four reviews a year. A hundred mature cards is a bit over one card a day.
The honest cost
A ceiling means doing some reviews that pure retention maths would call unnecessary. We think that is the right trade for a child, but it is a trade, and it has a real edge case worth naming.
If a card genuinely is locked in — your child has known it cold for months — then a termly review is just noise. Without an exit, our ceiling would be worse than Anki's approach: their card recedes into the distance and stops bothering you, while ours returns forever.
That is precisely why a card here can be finished. Locking a card in takes it out of rotation for good while keeping its star in your child's Sky. The ceiling and the ending are two halves of one design: the ceiling guarantees nothing is silently lost, and the ending guarantees nothing is pointlessly repeated. Either one alone would be a worse product than Anki's.
What this looks like over a year
Suppose your child ends Year 8 with 200 cards, and has genuinely mastered 60 of them.
- The 60 locked-in cards cost nothing — they have left the queue.
- The remaining 140 sit at various rungs. Those at the ceiling contribute about 140 ÷ 90 ≈ 1.6 reviews a day.
- Add newly-made cards, which are the ones with short gaps and therefore most of the daily work.
The daily total stays comfortably inside the 20-card cap, which is the number chosen so that a session is always finishable. If it ever did overflow, the cap holds and the oldest-due cards go first.
Why not make the ceiling adjustable?
Anki exposes the maximum interval as a per-deck setting. We deliberately do not, for the same reason the daily cap is fixed: every setting is a decision handed to someone who has no basis for making it. A twelve-year-old asked to choose a maximum review interval will either pick the biggest number available or ignore the screen. Fixed, defensible defaults beat configurability that nobody can use well.
FAQ
What is the longest gap between reviews?
90 days. Once a card reaches that gap it stays there, so every card still in rotation is seen at least once a term.
Why not let the gaps keep growing like Anki does?
Anki's default maximum is 100 years, which suits an adult who needs something for decades. A KS3 child is assessed each term, so a card whose next appearance is four years away is not serving them.
Doesn't a 90-day cap mean more reviews than necessary?
Slightly, yes — that is the trade. A card at the ceiling costs roughly four reviews a year. In exchange nothing your child has learned silently disappears before an exam.