Key takeaways
- Write one atomic fact per card so the review system scores a single retrieval instead of masking forgotten facts.
- Use a fixed interval ladder of 1 day, 3 days, 1 week, 2 weeks, and monthly instead of calculating complex schedules.
- Rewrite or suspend cards you fail repeatedly, as leeches usually indicate a poorly written card rather than poor memory.
- Split 100-page PDFs into 10-20 page chunks before generating cards to prevent models from dropping material.
- Expect recall of recent material in the first two weeks, with compounding retention showing around the one-month mark.
Flashcards spaced repetition works when you run it as a process: write cards that test one atomic fact, review them on expanding intervals, and rewrite or suspend anything you keep failing. This guide gives you that operating procedure, from card-writing rules to interval tuning to picking between free apps, paid apps, and Notion workarounds.
What is spaced repetition and why do flashcards work with it?
Spaced repetition is scheduling: you review a fact at the moment you're about to forget it, and each successful recall pushes the next review further out. Retrieval practice is the other half, you answer from memory instead of rereading. Flashcards are the natural format because a card forces exactly one retrieval with immediate feedback.
This combination beats rereading a 100-page PDF. Rereading feels productive but mostly builds familiarity, not recall; Dunlosky et al.'s review of ten learning techniques found practice testing and distributed practice the most effective across ages and subjects (Spacing effect). That's the same problem behind the "loses a LOT of information" complaint with feeding slides to a chatbot: summarizing is exposure, not retrieval.
For cold-call classrooms, the payoff is practical. A deck built from each session means you walk in having retrieved the material several times, not having skimmed it once. If your sessions are recorded, the recording policy covers consent and storage before you turn audio into cards.
How do you write flashcards that spaced repetition actually improves?
Spaced repetition amplifies whatever you feed it. A vague card reviewed twenty times stays vague; a precise card becomes permanent. The writing rules below matter more than the app or the interval algorithm.
One fact per card, or the card fails quietly
A card covering three facts feels fine when you write it. At review, you recall one fact, hit "good," and never notice the other two slipped. That failure is invisible because the review system only scores the card, not the facts inside it. Split anything with an "and" or a list into separate cards. One card, one retrievable fact, one honest score.
Turn slide fluff into cloze cards
Dense lecture PDFs and slide decks are mostly filler around a few load-bearing facts. Cloze deletion handles this well: copy the exact sentence from the slide, blank the one term that matters, and delete the rest of the deck. "The Fed's dual mandate is ___ and ___" beats a paragraph-length question about monetary policy. Question-first phrasing helps when the fact has no natural sentence to hide inside, but cloze is faster to write and faster to review, which compounds across a full course load.
Cards that die: definitions you can look up
The worst cards ask you to recite something a search box answers in two seconds: a term's definition, a formula's name, a date you can find in the syllabus. They pass review, teach nothing, and eat intervals you could spend on cards that build actual recall, like applying a framework to a case fact or deriving the formula's logic. If a card's answer is "something I could look up in five seconds under cold-call pressure," it is not a card, it is a bookmark. Cut it or rewrite it so the answer requires retrieval, not recognition.
This is also why "telling the model to be more detailed" produces bad cards. Detail and card quality are different axes: a model told to be exhaustive generates long, multi-fact cards that fail the one-fact rule and bury the answer under context. A generated practice set is a draft, not a deck. Every card still needs the one-fact test applied by a human who sat in the room. Tools like Praesto, NotebookLM, or Otter.ai can draft from a transcript, but the editing pass is where the deck earns its keep, and the terms of service for whatever tool you use (terms of use) are worth a skim before you feed it recorded coursework.
How do you set and adjust review intervals without a formula?
The default ladder that works for a course load
Skip the math. Use a fixed ladder: new card gets reviewed the next day, then 3 days, then 1 week, then 2 weeks, then monthly. That is roughly what most schedulers converge on anyway; Anki's default intervals follow a similar progression, and its docs describe the reasoning without requiring you to compute anything (Anki manual). If you would rather not trust defaults, this ladder is easy to run by hand in any app that lets you set due dates.
Tuning by feel: easy, hard, and blanked-out answers
Rate each recall with three grades, not four. Easy: push to the next rung and skip one (3 days becomes 2 weeks). Hard: repeat the current interval once before advancing. Blanked out: reset the card to 1 day. If a card sits at "hard" three reviews in a row, the problem is the card, not the interval. Rewrite it.
What to do when you miss three days
Do not reset everything. Cards you have seen five or more times keep their interval; just review them when you can. Only brand-new cards restart at 1 day. Cold-call prep for a specific class (a case discussion for your health law seminar, say) is a separate problem from retention, and cramming the night before is fine for that. Spaced repetition handles the long game; the night-before pass handles Tuesday.
How do you choose between free and paid spaced repetition flashcard apps?
Five criteria that matter for a full course load
Judge any app on five things: how fast you get cards out of lecture material (transcript, PDF, recording), how the scheduler handles a 100-card day when you skip a week, whether it works offline on iPad with Apple Pencil, what it costs for one person running a full course load, and whether the review material actually matches what your professor said instead of a lossy chatbot summary.
Free vs paid at a glance
| Criterion | Free tier apps (Otter.ai, NotebookLM-style tools) | Paid study apps (Jamworks, Studley AI, Playbacks) | Praesto |
|---|---|---|---|
| Card authoring from lectures | Transcript or summary first; you write cards yourself | Auto-generates quizzes and flashcards from recordings | Turns the session into a practice set, topic map, and spaced-repetition flashcards |
| Scheduling | Basic or none | Varies by product | Interval tuning built for a course load |
| iPad + Pencil | Partial | Varies | Native on Apple devices |
| Cost | Free, with limits on length or volume | Subscription | Priced for one person, a full course load |
| Cold-call classroom fit | Poor: you transcribe instead of listening | Better: record and review after | Built and used by Wharton MBA students for exactly this |
Where Praesto fits, and where it doesn't
Praesto wins if your bottleneck is presence in a cold-call classroom and you want reviewable material from the session itself. Its AI terms and limits are in the AI disclaimer. It falls short if you need a meeting bot joining calls for a work record, or fine-grained manual card editing on par with dedicated flashcard tools. If you already hand-write good cards and only need scheduling, a free option may be enough.
Can you do spaced repetition flashcards in Notion, and is it worth it?
The database workaround, step by step
Build a cards database with a Next Review date field and a Last Interval number. After each review, set Next Review to today plus the interval (use a formula or just type the date) and double the interval if you recalled it, reset it to 1 if you didn't. Notion's databases and formulas docs cover the property types you need. Expect one to two hours of setup and a manual date edit on every single card, every single review.
What Notion can't do for you
Notion won't reschedule anything automatically, won't bury a card until you touch it, and shows no due queue, so overdue cards silently pile up. You also lose typing latency and handwriting, which matters if you review on iPad with an Apple Pencil between classes.
When the workaround is enough
If you have under 50 cards for one course and you already live in Notion, it's fine. Past that, the manual rescheduling tax eats the time the method saves. Dedicated tools handle scheduling for you; our own Praesto does too, but it is built around class sessions, not a lightweight Notion replacement.
When should you rewrite, suspend, or delete a flashcard?
A card gets one of three actions: rewrite it if you keep forgetting it but the material matters, suspend it if you might need it later, delete it if it fails both tests.
The leech rule
If you've reviewed a card eight or more times and still miss it, it's a leech. Anki marks these automatically; other apps don't, so track your gut count. A leech almost always means the card is bad, not that you're bad: too vague, testing two facts at once, or written in language you didn't understand when you made it. Rewrite it into a smaller, concrete question or drop it.
Suspend vs delete
Suspend when the card is fine but the timing is wrong, like cold-call prep for a class that already happened. Delete when the card was really notes in disguise: "Explain the three drivers of WACC" isn't a card, it's a lecture outline. Split it into one-fact cards or throw it out.
End-of-term cleanup
After finals, delete anything the next course won't build on and suspend the rest. Suspended cards cost nothing to keep and can be unsuspended if a later course loops back to the material.
How do you get cards out of a 100-page PDF or a recorded lecture?
Chunk long documents before generating cards
A 100-page deck exceeds what most models hold at once, so they summarize and drop material. Split the PDF into 10-20 page chunks by topic, generate cards per chunk, then merge the sets. This is the same context-window problem behind "loses a LOT of information" complaints; chunking sidesteps it without a bigger model.
From transcript to practice set
For recorded lectures, export the transcript and ask for cards keyed to timestamps, so a wrong answer links back to the moment it was taught. Praesto is built for this flow: capture the session, get a practice set from it, and stay present in the room instead of transcribing. Its pricing targets one person carrying a full course load, so check the current plans on its site before assuming a tier fits. General-purpose recorders like Otter.ai give you the transcript but stop there; the card-writing and scheduling work stays yours, per the Anki manual.
Checking AI cards for dropped content
Spot-check three random chunks against the source: if a slide's key term is missing from the generated set, the chunking or the prompt dropped it. Regenerate that chunk, don't patch the whole run.
How long until spaced repetition with flashcards shows results?
In the first two weeks, expect recall of recent material, not mastery. Cards from week one will feel shaky when they resurface; that is the interval working, not failing. The compounding shows up around the one-month mark, when older cards return and most of them stick, so review sessions get faster instead of longer.
Before exams, the payoff is different: you already know what you forgot. Your queue surfaces weak cards, so the last week is targeted repair rather than a full re-read of a 100-page PDF.
The failure mode is starting the night before. Cramming gets you through tomorrow's cold call and nothing else; without intervals, there is no schedule doing the retention work for you.
FAQ
What is the best free spaced repetition flashcards app?
Anki is the strongest free option because its scheduler runs expanding intervals automatically and marks leeches for you. NotebookLM and Otter.ai are free for drafting cards from transcripts, but neither schedules reviews. If you only need scheduling and already write good cards, free is enough.
How many flashcards should I review per day?
There is no fixed number; the queue decides. A full course load can produce a 100-card day, so budget 20-30 minutes rather than a card count. Cap new card introductions at 10-20 per day. Missed days matter less than you think because mature cards keep their intervals, keeping the backlog manageable.
Can I use Notion for spaced repetition flashcards?
Yes, with a database workaround using a Next Review date field and a Last Interval number, updated manually after every review. It works under 50 cards for one course if you already live in Notion. Past that, the manual rescheduling tax outweighs the method's savings, so use dedicated tools.
Do spaced repetition flashcards actually work for MBA or law school material?
Yes, because cold-call classrooms reward retrieval, not familiarity. Practice testing ranked as the most effective learning technique in Dunlosky et al.'s review. Build one card per atomic fact from each session, cut look-up-able definitions, and rewrite leeches. The night-before cram handles tomorrow's case; intervals handle the long game.
How long should intervals be for flashcards?
Use a fixed ladder: 1 day, 3 days, 1 week, 2 weeks, then monthly. Grade each recall with three ratings: easy skips a rung, hard repeats the current interval, and blanked-out resets to 1 day. No formula is needed. If a card sits at hard three reviews running, rewrite the card instead of touching the interval.