The easy version of this app asks a random question from whatever topic you picked, every time you try to open something blocked. It would work, technically. It would also waste most of its unlocks — some questions land on things you already know cold, others hit something you saw once, three weeks ago, and never again. Neither teaches you much. The scheduler exists to stop that from happening.
The forgetting curve is not a metaphor
Hermann Ebbinghaus measured it on himself in the 1880s: memory doesn't fade at a steady rate, it falls off fast right after you learn something and then levels out. Review a fact right before you'd have forgotten it, and the memory gets stronger than either reviewing it too soon — while it's still fresh, which teaches you nothing — or too late, after it's gone and you're relearning it from scratch.
We don't get to choose when you'll try to open a blocked app. That part is entirely yours — a habit, a boring meeting, a bad moment at 11pm. What the scheduler controls is which topic gets put in front of you when you do. Answer a question right and it moves further out; miss it and it comes back sooner. The unlock you didn't ask for is also, quietly, a review sitting exactly where the curve says it should.
Why streaks, not scores
A running score rewards being right, which sounds correct until you notice what it optimises for: sticking to what you already know so the number never drops. A streak rewards showing up regardless of the answer, which is the actual bottleneck for spaced repetition — the technique only works if you're still opening the app in three weeks, not if you got every question right this week.
The risk on the other side is real. A streak that only counts opens, not effort, turns into a chore you tap through without reading. So a streak here tracks days you answered something, not days you unlocked something — the honeypot for gaming your own habit is smaller if the thing being counted is the harder one.
What "the next question" has to balance
Four things pull against each other on every unlock:
- Timing. Anything overdue for review outranks anything fresh — that's the whole point of the curve.
- Difficulty. A topic you've been missing needs an easier question inside it before a harder one, or every unlock becomes a small failure.
- Coverage. Left alone, the scheduler would happily loop the three topics you're worst at forever. It has to force breadth across the curriculum you actually asked to be quizzed on, not just its favourite weak spots.
- Variety. The same question twice in a row, even if it's technically due, reads as broken rather than smart.
None of those four wins outright. The scheduler is a set of trade-offs, not a formula with one right answer.
What we're still unsure about
The forgetting curve needs a history to work from, and a brand-new account doesn't have one. The first few days for anyone starting a topic are closer to a guess than a schedule — we don't yet know how badly that shows, or how quickly it corrects itself once there's real answer data to work with. If your early unlocks feel more random than the ones a month in, that's expected right now, not a bug we've decided to ignore.
If there's a subject you'd want scheduled properly, hand in a request slip — the more topics people actually ask for, the less guessing the coverage rule has to do.