Three buttons, each showing the real interval it will produce. That's the whole review interface, and the honesty of those three taps is what decides whether spaced repetition works for you or just occupies you.

What the room looks like

One card, centred, with a generous shadow — the card is the only thing on screen. A thin progress rule across the top tells you how far through the session you are. A small label in the corner says what kind of card it is (Cloze, and so on).

The card face is set in a serif, which is the app's way of marking a tutor's voice throughout. Tap anywhere to flip.

At the bottom, once you've flipped: Again · Good · Easy, each with its resulting interval underneath in small mono type. And a footer line that matters more than it looks:

Cards never use a Pi turn.

You can review for an hour and it costs you nothing. That split is deliberate.

Try before you flip — this is the whole mechanism

The single most common way to waste a review session is to read the front, think yes, I know that, flip, and see that you were right.

You've just practised recognition, and recognition isn't what an exam asks for. What works — and this is one of the better-established findings in learning research — is retrieval: actually producing the answer from nothing, and feeling the effort of it.

So the instruction is dull and it's the only one that matters. Read the front. Try to answer, out loud or in your head, all the way to a real answer. Then flip.

The three grades

Again — you didn't have it, or you guessed. The card comes back soon.

Good — you got it, with effort. The interval stretches by the standard step. This should be most of your cards; effortful and correct is exactly the state spaced repetition is built around.

Easy — it was genuinely instant. The interval stretches further.

Each button shows the actual date it produces, so grading isn't abstract. You can see that Easy means eighteen days before you commit to it.

Why not four buttons, or five?

Because a fourth grade adds a decision without adding information.

The precision people imagine they're getting from a "Hard" between Again and Good doesn't survive contact with honest self-assessment — nobody can reliably distinguish their own 60% from their own 75% recall. What a fourth button reliably does add is a moment spent thinking about the buttons instead of the card, twenty times a session.

Grade honestly, and specifically don't grade kindly

The temptation is Good on a card you half-had, because Again feels like a failure and makes the session longer.

What that actually does is push the card three weeks out, where you'll have forgotten it completely, and it'll return as a stranger the day before an assessment. Again on a card you half-knew is not a bad mark. It's the system doing its job — the pile is a schedule, not a scorecard, and nobody is looking at it.

The reverse mistake is Again on everything out of caution, which turns your daily twenty into the same twenty cards forever. If you produced the answer, even slowly, that's Good.

Twenty a day, and ninety days maximum

Two fixed numbers, both different from tools built for adults.

Twenty cards a day. Not configurable. Enough to matter, small enough that a tired fourteen-year-old finishes it. An unclearable backlog is how flashcard habits end.

Ninety days maximum interval. However well you know a card, it comes back within a term. Anki's default cap is a hundred years — correct for a lifelong learner, wrong for a child with an assessment in April.

When a card won't go in

Four Agains on the same card is not a scheduling problem. It's a teaching problem, and the review room can't fix it.

That's what the button on the back of a graded card is for: take that exact card to the professor who owns the subject, with the question already asked. It's the most important gesture in the app, and it lives exactly where the frustration is.

Updated 10 Sep 2026. Written by Duke Harewood, who builds the tutors at aitutors.me.