Spaced repetition, explained without the hype

The forgetting curve, in one minute

Hermann Ebbinghaus measured something obvious and stubborn: a new fact fades fast, then fades more slowly. What he also found is that each successful recall flattens the curve. Twenty minutes after learning a name you might remember half of it; after a successful recall tomorrow, the half-life stretches to days; after a retrieval a fortnight later, it stretches to months.

Spaced repetition is not a trick for remembering without effort. It is a way of spending retrieval effort at the moment it buys the most: shortly before you would have forgotten. That is the entire idea, and everything else in a scheduler is bookkeeping around it.

The practical consequence is that intervals should grow. Reviewing a fact five times in one afternoon adds almost nothing after the first success, because the memory never had a chance to fade. Waiting is not laziness; it is the part of the method that does the work.

Recall is the workout

Rereading feels productive and is mostly not. The page is in front of you; familiarity is not retrieval. Trying to produce an answer before seeing it — even failing, then seeing the answer — is what creates a durable memory.

This is why a card should ask a question that has an answer you can check. “Explain photosynthesis” is not a card. “Which organelle performs photosynthesis?” is. If you cannot tell whether your answer was right, the review cannot teach the scheduler anything, and it cannot teach you much either.

The effortful part is the useful part. A card you answer in half a second is not a problem to be solved, but a card you never once struggle with is a card you probably did not need to write down.

What a good card looks like

One card, one fact. Long lists buried in a single card feel efficient and review badly, because you remember three of five items and mark the card as known. Split it: five cards, each answerable in a few seconds.

Add context when the term alone is ambiguous. A bare year is a lottery ticket; “Which empire fell in 1453?” is a question. Keep the wording close to how you need to recall it — recognition on the page is easier than production from memory, and only one of those is the skill you want.

A useful test before saving: could a stranger mark your answer right or wrong without asking you what you meant? If not, the card is under-specified, and it will be under-specified in six weeks too.

A habit small enough to keep

A queue of 300 cards is not a habit; it is a debt. Start with one deck and a handful of new cards a day. Ten honest minutes beats an hour of scrolling through cards you are not really reading.

Anchor the session to something that already happens: the first coffee, the commute seat, the moment a laptop opens. When you miss a day, do not double the next one. Resume at the normal size.

The size of the queue is a design decision you can revisit weekly. If it grows faster than you clear it, the problem is the intake, not your discipline, and the fix is fewer new cards rather than longer sessions.

Where Miyu fits

Miyu keeps the mechanics out of the way: FSRS picks the dates, the reviewer offers Again and Good with the next interval visible, and your notes stay in a folder you own. The catalog gives you something worth remembering on day one — a course you can read and turn into cards as you go.

The plan is not a streak. It is a queue that stays small enough to finish, and cards boring enough to answer honestly. Everything else — the algorithm, the settings, the sync — exists to keep that queue in front of you and out of your way.