Features

Spaced repetition that adapts to you

Miyu schedules reviews with FSRS, and shows you what each grade does to the next date before you answer.

What FSRS does

Miyu does not march through a fixed pile of due cards. Every card carries a memory state that FSRS — the Free Spaced Repetition Scheduler — updates after each review: how stable the memory is, and how difficult the material has proven for you. From those two numbers the scheduler picks the next date that should land near your target retention, 90% by default.

Because the estimate is per card, a phrase you have answered correctly ten times can drop to a yearly review while a new character stays close. The queue adapts to the card in front of you rather than to a single global interval.

Miyu uses FSRS through ts-fsrs, the TypeScript implementation of the algorithm. The card state is ordinary collection data: it travels with your notes, so a review on your phone is the same review the scheduler sees on your desktop.

Two honest grades

The reviewer offers Again and Good. Each button shows the interval it would schedule, so the next due date is visible before you answer. Two grades make the choice honest: either you recalled it or you did not. Longer grade scales invite self-flattery, and flattery inflates intervals until cards come back forgotten.

Because the preview is attached to the button, the choice stays concrete. Again might offer ten minutes; Good might offer four days. If those numbers look wrong for the material, that is a preset question rather than a grading one, and it has an answer in settings instead of in a guess on the card.

A schedule you can inspect

Presets are editable. You can change the target retention, the maximum interval, and the learning steps a new card passes through before it graduates. Interval fuzz is on by default, so cards learned on the same day do not all return in one unmanageable block.

Presets belong to decks, so a vocabulary deck can be gentler than a deck of dates without touching anything global. The defaults are deliberately conservative: 90% retention and a maximum interval measured in decades, which a card only reaches after years of successful reviews.

The numbers in settings are the numbers the scheduler uses. Nothing hidden is deciding your week.

Why less can be more

A review that feels easy is not proof of learning. Miyu optimises for the session you will actually do tomorrow: a short queue, a clear next date, and no streak counter to punish a missed day. When you return, FSRS recalculates from what you actually recalled.

Consistency compounds. A small queue reviewed daily produces more durable memories than a heroic weekend session, because the scheduler is measuring retrieval and retrieval only happens when you sit down. Miyu tries to make sitting down the easy part.

Keep reading

Two more decisions behind Miyu, plus something to study.