Reviewing material at increasing time intervals, each review timed to catch you just before you'd forget.

Instead of cramming everything at once, you revisit each item after a gap — then after a longer gap, then a longer one still. Every review that succeeds on time resets the clock and makes forgetting slower next time, so the intervals can safely keep growing.

The pay-off is efficiency: you spend the fewest possible reviews to keep a fact retrievable, because you only revisit it right when it's about to slip away rather than while it's still fresh.

Because a slightly effortful retrieval strengthens the memory far more than an easy one — you learn most when the memory has decayed a little but not fully.

Recall this from the memory model: a memory has a storage strength (how well-learned it is) and a retrieval strength (how accessible it is right now, which fades with time). Reviewing while retrieval strength is still high is nearly wasted effort. Letting it drop a little, then successfully pulling the answer, produces the largest boost to durable storage strength.

The catch is balance: wait too long and you fail the retrieval entirely, gaining nothing and having to relearn. Spaced-repetition algorithms exist to estimate that sweet spot for you.