Performance = how well you do right now; learning = durable change you can still show later — and they can move in opposite directions.

Educational scientists (notably Nicholas Soderstrom and Robert Bjork) stress that what you observe during study — smooth, error-free performance — is a poor and sometimes misleading index of learning. Conditions that boost immediate performance (cramming, re-reading, having the answer in view) often produce little lasting memory, while conditions that depress immediate performance (spacing, self-testing, making errors) tend to produce more. The tip: distrust fluency as a gauge of learning; ask instead "could I still do this cold in a week?"

This very app schedules your cards with exactly this FSRS logic — showing each card right as you're about to forget it.

Every flashcard and duel you answer feeds a per-card memory estimate; the scheduler then surfaces each item near the moment its recall probability dips toward the target. That's the most efficient possible use of review time: no wasted reviews on cards you'd ace, no forgotten cards left too long.

So the cards teaching you the algorithm are themselves being scheduled by it — learning how to learn and watching the machinery run, at the same time.