Spaced-repetition and study apps that actually help#

There is a particular kind of disappointment that comes from a beautiful study app. You install it, tap through a slick onboarding, and a week later you’ve learned almost nothing — but you feel productive. Most study tools are like this. A small handful are genuinely transformative, and the difference comes down to one question: does the tool do the remembering for you, or does it force you to do the remembering, on the best possible schedule? This page is about the second kind — what they are, how to use them, and the one mistake that turns even a great tool into busywork.

The one tool worth your time: spaced-repetition software#

If you only adopt one piece of learning software, make it a spaced-repetition system (SRS) — and the standard-bearer is Anki. It’s free, it’s open source, it runs on everything, and it quietly automates the two most powerful moves in all of learning: spacing your reviews and forcing you to retrieve.

Here’s the whole idea. You make digital flashcards. Each day the app hands you exactly the cards that are due — the ones it predicts you’re about to forget — and nothing else. You try to recall the answer, you flip the card, and you rate how it went. Get it easily and the app pushes the next review further out: a day becomes a week, a week becomes a month, a month becomes a year. Stumble, and it brings the card back soon and rebuilds from there. Easy cards drift to the back; stubborn cards get the attention. You spend your minutes only where memory is actually wearing thin.

That scheduling isn’t a gimmick — it’s the most reliable finding in the science of learning made automatic. When Cepeda and colleagues pooled over a hundred experiments on distributed practice, the verdict was unambiguous: spacing your reviews beats cramming them, and the gap widens the longer you need to remember (Cepeda et al., 2006). Anki just does the spacing arithmetic for hundreds of facts at once — exactly the part no human keeps up by hand. I cover the underlying effect on the spacing page , and the retrieval half of the magic under retrieval practice .

Anki’s current scheduler is FSRS (the Free Spaced Repetition Scheduler). It goes beyond the old fixed-interval logic: it fits a model of your forgetting to your review history and aims each card at a retention target you choose — say, a 90% chance you’ll still recall it when it next comes due (Anki/FSRS documentation, 2024). Turn it on, pick a desired retention around 90%, do your reps honestly, and otherwise leave the knobs alone.

When to trust auto-generated cards — and when to write your own#

The newest wrinkle is AI card generation: paste in a chapter, a PDF, a set of lecture notes, and a tool spits out a deck of flashcards in seconds. It’s seductive, and it’s where most people quietly sabotage themselves.

Here’s the problem. The hard, slow work of turning a messy chapter into a few crisp question-and-answer pairs — deciding what actually matters, phrasing it as a question, putting the answer in your own words — is itself a huge part of the learning. Hand that work to a machine and you skip the very step that builds understanding. You end up with a fat deck of cards for material you’ve never really wrestled with.

So use AI generation, but use it in its place:

• Great for raw material, lousy as a finished product. Let the tool draft a first pass, then edit every card before it enters your deck. Treat its output as a rough quarry, not a building.
• Best when you already understand the topic. Auto-generation shines for bulk, well-structured facts you’ve met before — vocabulary, anatomy, drug names, dates — where you just need coverage fast. It’s weakest for the conceptual heart of something you’re still trying to grasp; write those cards yourself.
• Watch for the two classic failures. AI cards love to test trivia the source merely mentioned, and they will occasionally state something confidently wrong. Read each card and ask: is this worth remembering, and is it actually true?
• Reject the bloat. A generator will happily hand you three hundred cards from one chapter. Most are noise. A dozen good cards you’ll review for years beats three hundred you’ll abandon by Friday.

The honest test is simple. If editing the auto-generated cards makes you think — no, the real point is this; that fact doesn’t matter — the tool is helping. If you’re tapping “accept” without engaging, you’ve automated away the learning and kept the chore.

Make good cards, and keep the effort in#

Whether you write cards by hand or polish ones a machine drafted, the same craft rules decide whether the deck works. The scheduler is clever, but garbage in, garbage out — FSRS can’t rescue a vague or overloaded card.

• One idea per card. If your answer is a paragraph, you’ve packed a lecture into a flashcard. Split it. The front should ask one thing; the back should answer that one thing.
• In your own words. Rephrasing forces you to process the meaning rather than memorise a sentence. A card copied verbatim from the text teaches you to parrot the text.
• Make the front demand real recall. A good card can’t be answered by recognising a familiar phrase; it makes you produce the answer. If you can guess it from the wording, it’s testing fluency, not memory.
• Add a little context or a why. Cards that connect to something — a reason, an example, an image — stick far better than bare associations.

And then the rule that ties everything together, because it’s the one the tools are forever tempting you to break: keep the effort in. A flashcard only counts if you genuinely try to produce the answer before you flip it; glancing at both sides is just rereading in disguise. Rate your recall honestly, not generously — the whole schedule is built on truthful grades. The mild discomfort of straining to remember isn’t the system failing; it’s the system working. Every tool here is, at bottom, a machine for making you do that slightly-uncomfortable thing on the perfect day. The moment it starts doing the remembering for you, it has stopped helping.

The takeaway#

Pick one spaced-repetition tool — Anki with FSRS is the safe default — and actually use it: a short session every day, cards added as you learn, desired retention around 90%. Let AI draft cards from dense material if you like, but edit every one and ruthlessly cut the bloat; write the conceptual cards yourself. Keep cards small, in your own words, and answer before you flip. Do that and the software handles the calendar while you do the only part that ever mattered — the remembering.