Spaced repetition, explained without the jargon

All posts

You've probably felt this: you cram fifty flashcards the night before a trip, feel great about it, and by the following week you can recall maybe five. That's not a willpower problem. It's how memory actually works — and it's fixable.

The forgetting curve isn't a metaphor

In the 1880s, a psychologist named Hermann Ebbinghaus ran a very tedious experiment on himself: memorize nonsense syllables, then test his own recall at intervals afterward. He found that memory doesn't decay evenly — it drops fast at first, then levels off. Most of what you forget, you forget within the first day or two.

Plot that decay and you get the "forgetting curve": steep at the start, flattening over time. The critical detail is what happens if you review the material again right before you'd have forgotten it. The curve resets, but it resets to a shallower slope — the next forgetting curve is flatter than the first. Review again at the new right moment, and it flattens further. Do this enough times and the interval between reviews stretches from a day, to a few days, to weeks, to months, while the word stays put.

Why cramming doesn't survive this

Cramming reviews everything at once, regardless of what you actually need reviewed. You spend equal time on a word you've known for a month and one you saw ten minutes ago. It feels productive because you're "doing reps," but you're not doing the one thing that actually matters: catching each word at its own edge of forgetting.

This is also why a fixed daily deck — the same 20 cards, every day, forever — quietly turns into busywork. Words you've solidified keep eating review time that a genuinely shaky word needed more of.

What a scheduler is actually doing

A spaced-repetition system tracks a strength estimate for every word you've learned, individually. Get a word right, and its next review moves further out — a day, then a handful of days, then a couple of weeks, then longer. Get it wrong, and the interval resets close to zero, because you've just learned that you don't actually know it yet, no matter how many times you'd technically "seen" it before.

The scheduling math has gotten more precise over the decades — Ming uses FSRS, a modern open algorithm that's meaningfully more accurate than the older fixed-interval formulas (Anki's default algorithm, SM-2, dates to 1987) — but the underlying idea hasn't changed: review each thing right before you'd forget it, not on a fixed schedule that ignores whether you actually remember it.

What this looks like day to day

In practice, it means your daily lesson is never the same shape twice. Some days it's mostly review, because a batch of words happen to be due. Other days there's more room for new words, because little is due yet. You don't manage this — the schedule does, which is the entire point. The two things you control are how many new words come in (by picking your daily time budget) and whether you show up.

Ming builds every lesson this way — new words paced deliberately, old words reviewed exactly when they're about to slip. See how it fits into a five-minute session.