Guide
How to make good flashcards
Most flashcard decks fail not because of the algorithm, but because the cards are bad. These are the principles that actually matter.
The atomic rule
One card, one fact.
Every card should test exactly one thing. If a card covers three facts and you remember two, the algorithm gets a noisy signal — it either over-schedules the forgotten fact or resets the ones you already know. When in doubt, split it.
Write specific questions
Avoid "What is X?" — start with "What does X do?" or "How does X differ from Y?"
Yes/No questions are weak — they give the algorithm a 50% baseline and teach you nothing useful. Add context to remove ambiguity: "In the Krebs cycle, what does..." instead of "What does..." Context triggers the right retrieval path.
Understand before you card
Read the chapter before you make the cards.
Trying to memorize something you don't fully understand creates fragile memories that break under slightly different questions. Get the big picture of a section first, then extract cards from it. This order matters.
Card types
Match the card type to the knowledge type.
Best for definitions, single facts, and anything with a clean right/wrong answer.
For vocabulary or bidirectional associations — term→definition and definition→term. Builds stronger, more flexible recall.
One card per step, plus one card for the overall sequence. Don't cram an entire pathway onto one card.
One card with the formula. One card per variable's meaning. Keep them separate.
Subject-specific
Different subjects call for different card shapes.
Card language syntax, keywords, library functions, and edge cases. Use code samples on the back. Associate abstract logic with a visual — picturing a "camera viewfinder" for a sliding window algorithm makes it stick.
Use three templates: Definition, Element (a single element of a test, e.g. "breach" in negligence), and Case Principle. For multi-element tests, one cloze deletion per element. Practice IRAC on short fact patterns.
Anchor anatomy to spatial memory whenever possible — describe where something is, not just what it does. Physiological pathways work well as step-by-step process cards.
Memory hooks
A vivid, weird image beats a clever acronym.
For hard vocabulary, find a word that sounds similar (an acoustic link) and build a bizarre image connecting the sound to the meaning. Humor and slightly absurd visuals trigger stronger consolidation than clean, forgettable ones. Images on cards work the same way — visual memory often outlasts verbal memory.
Grading honestly
The algorithm only works if you're honest.
FSRS schedules cards based on your grades. If you use "Hard" when you actually forgot something, you'll end up with unreasonably long intervals and cards you keep blanking on. The rule: use "Again" if you forgot it, "Hard" only if you recalled with real hesitation, "Good" for solid recall. Never use "Hard" to soften a failing grade.
AI-generated cards
AI drafts fast. You still have to edit.
AI can turn a lecture PDF into 30 draft cards in seconds — but it often misses the atomic rule, bundles facts, or generates generic questions. Always do a human pass: delete the filler, split the bundles, and rewrite questions that could mean multiple things. The AI gets you 80% of the way. The last 20% is yours.
Ready to put it into practice?
Start studying