When you learn something new, you forget it on a predictable curve. Reviewing right before the forgetting point resets the curve to a shallower slope. Each successful review pushes the next review further into the future — minutes → hours → days → weeks → months.
The two dominant algorithms are SM-2 (the one Anki and QuickCards use) and FSRS (the newer, machine-learning-based successor). Both adjust card intervals based on how you self-rate your recall after each review.
In practice: 15–20 minutes of spaced repetition daily produces dramatically better long-term retention than 2 hours of cramming once a week.