You forget on a schedule. A card that comes back just as the word is slipping puts it back for longer than a card that comes back while you still know it cold.That trade is the whole design. Review here is meant to feel slightly too late.
Spaced practice beats cramming in second-language learning, and the advantage widens on delayed tests. Equal and expanding gap schedules come out equivalent, so what carries the effect is the gap itself together with the act of recall.
A card tests you
Producing an answer from memory beats re-reading the same material. So nothing in the queue simply re-shows you a word. Every card asks first.
Why rules get mixed together
Once a rule stops being new, the queue starts mixing it with the rules you confuse it with. Interleaving beats practising one rule in a block when the test comes a week later. It is harder in the moment, so new material is blocked first and interleaved afterwards.
The scheduler is SM-2
Infinite Stories schedules with SM-2. Keeping it in this release was a scope decision rather than an evidence claim. The spacing research says the shape of the schedule is not where the effect lives, so a classic scheduler was enough to ship.
FSRS with default parameters is the measured next step. We will judge it against your own check-up re-take results rather than against published benchmarks.
Getting the most out of it
Clear your due cards before opening a new course section. The queue is ordered by what is closest to being forgotten.
Let a card run late rather than doing it early. The delay is the ingredient.
Say or type the answer before revealing it. A card you only recognise is a card you have not tested.
Cards you keep missing shorten their own steps. There is nothing to configure.
Under the hood
Each card carries an interval, an ease factor and a lapse count. A correct answer multiplies the interval by the ease factor, with a fuzz of plus or minus 15 percent so cards learned on one day do not all fall due on one future day.Four lapses flag a card as a trouble word, which sends it back through relearning steps of 10 minutes and 1 day. The daily queue is capped at 20 cards, so a long absence produces a session instead of a wall.
Sources
Kim & Webb (2022) The effects of spacing on second language learning: a meta-analysis. Language Learning.