Mastering Spaced Repetition: The Science of Long-Term Memory
Why reviewing the right thing at the right moment beats studying twice as long.
The problem with how most people study
You read the chapter. You highlight the notes. You feel ready. Then the test arrives two weeks later and the material has evaporated. That is not a focus problem or an intelligence problem. It is a timing problem, and cognitive science has had the answer for over a century.
In the 1880s, German psychologist Hermann Ebbinghaus mapped what he called the forgetting curve: the rate at which newly learned information leaves memory. His numbers are uncomfortable. We forget roughly 50% of new information within an hour. After one day, about 30% remains. After a week, less than 10%.
But Ebbinghaus also found the exit. Every time you successfully recall information, the curve flattens. The next time the same fact drops toward the threshold of forgetting, it takes longer to get there. Do this enough times, and the interval stretches from days to weeks to months. The information moves from fragile short-term storage to something much more durable.
Spaced repetition is the practice of reviewing information at precisely those expanding intervals. Not when you feel like it. Not when a test is three days away. When the system says you are about to forget.
Why spacing works
Retrieval practice matters more than re-reading. Actively pulling information out of memory strengthens it. The effort of the recall, not the exposure, is what builds the trace.
The spacing effect: information reviewed at intervals is retained far better than information massed into one session. Massing feels efficient because recognition is easy in the moment. The problem shows up later, when you realize you can only recognize the answer when you already know what it is.
Desirable difficulty is the counterintuitive one. Reviews should feel slightly hard. A card you answer instantly with no effort is scheduled too soon. A card you cannot remember at all was scheduled too late. That effortful middle ground is when something meaningful is happening.
The SM-2 algorithm
Cereby uses SM-2 (SuperMemo 2), one of the most studied spaced repetition algorithms, which translates those cognitive principles into a scheduling formula.
Every flashcard starts with an ease factor of 2.5 and a first review interval of one day. After each review, you rate how the recall went. The algorithm uses that rating to update two things: the ease factor and the next interval.
A simple example for a vocabulary word you encounter on Day 1:
| Day | Result | Next interval |
|---|---|---|
| Day 1 | First encounter | Review tomorrow |
| Day 2 | Remembered easily | 2.5 days (next: Day 4.5) |
| Day 5 | Remembered easily | 6.25 days (next: Day 11.25) |
| Day 11 | Remembered easily | 15.6 days (next: Day 26.6) |
| Day 27 | Remembered easily | 39 days (next: Day 66) |
Each successful recall earns a longer interval. A struggle brings the interval back down and drops the ease factor so subsequent intervals grow more slowly. Forgetting completely resets the card to Day 1 and treats it as new material.
The scheduling is just arithmetic applied consistently. Humans are bad at intuiting optimal review timing. The formula is not.
How Cereby implements it
When you create a flashcard in Cereby, question and answer are all you supply. The SM-2 schedule runs automatically.
At review time you see the question, try to recall the answer without looking, reveal it, and rate yourself: Easy (instant recall, interval stretches significantly), Good (remembered, normal growth), Hard (struggled, minimal growth), or Again (forgot, card resets).
That rating is the only input the system needs. Marking everything "Easy" to feel good pushes cards to intervals too long for your actual retention. Marking "Good" when you genuinely struggled means slower progress but honest scheduling.
Cereby AI can generate flashcards directly from your study materials. "Create flashcards from my calculus notes" or "Generate flashcards on cell biology" are valid prompts. The output goes straight into the SM-2 scheduler.
Making it stick in practice
The algorithm handles timing. You handle consistency. Those are the two jobs, and only one of them is automated.
Start with 10 to 20 cards, not 200. A backlog of overdue cards is demoralizing and harder to recover from than you expect. Build the habit with a manageable number and expand after 30 straight days without skipping.
Rate yourself honestly. Partial credit self-assessment ("I sort of got it") is the enemy of the spacing effect. If you did not clearly recall the answer before revealing it, "Again" or "Hard" is correct, not "Good."
When you build cards, make them atomic: one fact, one card. "Explain photosynthesis" is too broad. "What molecule does the Calvin cycle produce?" has a specific answer and a specific retrieval trace.
Where spaced repetition earns its reputation
Language learners use it for vocabulary because the problem is pure volume: thousands of words each needing long-term retention. Medical students rely on it for anatomy terms and drug mechanisms on licensing boards. Any domain requiring discrete fact retention over months is a domain where it outperforms alternatives.
It does not replace understanding. If you do not know why a drug inhibits an enzyme, a flashcard about the drug name will not build that knowledge. Use spaced repetition alongside explanations and practice problems, not instead of them.
The algorithm is not new. What changed is that software can now apply it automatically, at scale, to your actual study materials. Create 10 to 20 cards on something you are actively learning, review them when the system says to, rate honestly, and do that for two weeks. The difference from re-reading the same material will be noticeable.
The rest is showing up every day.