A Deck of 300 Cards You Never Review Is Worth Nothing

Making flashcards used to be the bottleneck. Typing out three hundred cards before you could start reviewing was hours of work that felt like studying and was mostly typing.

That bottleneck is gone. You can generate three hundred cards from a chapter in under a minute.

Which has quietly moved the problem somewhere else: now anyone can have a large deck, and most large decks do not work. The constraint was never the typing. It was knowing what belongs on a card — and generation does not decide that for you.

The three ways a deck fails

Almost every abandoned deck fails in one of these ways, generated or hand-made.

It has more than one idea per card. The back says: definition, three examples, and the exception. You will learn the definition, glance at the rest, and mark it correct. Four facts became one, and the three you skipped are the ones the exam asks about. One card, one retrievable thing.

The question gives away the answer. "What is the primary function of the mitochondria in cellular respiration?" contains enough of its own answer that you can produce it without knowing anything. Generated cards do this constantly, because restating the source sentence as a question is the easiest transformation available. If you can answer a card from its phrasing alone, it is testing reading.

It is too big to finish. A three-hundred-card deck from one chapter means a review session you cannot complete, and a session you cannot complete is one you stop starting. Decks do not usually die from being wrong. They die from being long.

The rule: one card, one retrieval, no clues

Three questions to run a card through:

1. Is there exactly one thing to recall? If the answer contains "and", suspect it.
2. Could I answer this without knowing the material? Read the question alone. If the phrasing hands you the answer, rewrite it shorter and blunter.
3. Would I care if I got this wrong in the exam? If not, delete it. Most generated decks are a third trivia by volume, and the trivia is what makes the deck too long to review.

That third question does most of the work. The instinct is to keep every card because deleting feels like losing work — but you did not do the work, the generator did, and a deck you finish daily beats a complete one you avoid.

What generation is actually for

It removes the typing, which was never the valuable part, and leaves you the editing, which always was.

A workflow that produces decks people keep using:

Generate per topic, not per document. A chapter produces an unreviewable pile. A topic produces fifteen to twenty-five cards you can get through in a sitting and actually finish.

Delete first, before you review anything. Go through once and remove trivia, duplicates and anything you already know cold. This takes five minutes and typically cuts a third. It is the highest-value five minutes in the whole process.

Then fix the survivors. Split any card with two ideas. Rewrite any question that leaks its answer. Add your own mnemonic where one helps — a card in your own words is recalled better than one in the source's.

Generate from your own material. Cards built from your documents inherit your course's vocabulary, which matters because the exam uses it too. This is what generating flashcards from your own notes does, and it is the difference between a card about the subject and a card about your syllabus.

Then leave the scheduling alone. Once the deck is good, the reviewing is a solved problem: increasing intervals, handled by the tool. The reason spacing fails is almost never the algorithm — it is the deck being too big to face.

Four card types worth knowing

Most decks contain one kind of card — term on the front, definition on the back — and that kind only tests vocabulary. A deck that mixes formats tests more of what an exam actually asks.

Definition cards. Term → meaning. The default, and fine in moderation. Keep them for vocabulary you genuinely need to produce, not for every bolded word in the chapter.

Reverse cards. Meaning → term. Harder and more useful, because exams ask you to name the thing from its description far more often than they ask you to define a term you have been handed. If a term matters, have it in both directions.

Discrimination cards. What separates X from Y? This is where most marks are actually lost — two concepts you can each define and cannot tell apart under pressure. Any pair of similar ideas in your material deserves one of these, and generated decks almost never produce them, because the distinction is rarely stated in one place in the source.

Application cards. Given this situation, which principle applies? A short fact pattern on the front, the rule on the back. These are the closest a card gets to an exam question, and they are the ones worth writing yourself.

A useful mix for a dense topic: a handful of definitions, the same terms reversed, three or four discriminations, and two or three applications. Twenty cards, each testing something different, beats sixty that all test the same reflex.

What flashcards cannot do

They test whether you can retrieve a fact. That is all, and it is less than an exam asks for.

They will not tell you whether you can apply a rule to a new set of facts, structure an argument, or connect two topics that were taught in different weeks. For those you need questions with more surface area — practice exams generated from the same material — and the deck is the foundation underneath, not the whole building.

The clearest symptom of relying on cards alone: you know every definition and the exam question still stops you, because it asked you to use one.

Where this works, and where it does not

Cards earn their place in subjects built on terminology and discrete relationships: biology, chemistry, anatomy, pharmacology, law (cases and principles), economics (models and definitions), history (dates, causes, consequences).

They are a poor fit for pure mathematics, where the thing being learned is a procedure and the only way to learn it is to work problems; for creative writing, where there is no single retrievable answer; and for conversational language, where cards build vocabulary but not fluency. For a long syllabus that has to stay alive for months, the deck is necessary but the review schedule is what does the work.

The test

Open the deck you made last month. If you have reviewed it in the last week, it works. If you have not, it is too long or too trivial, and no amount of better generation will fix that — only deleting will.

Thirty good cards you review every day beat three hundred you generated once and never opened.

Build the deck from your own material and then cut it: generate flashcards from your notes, and pair them with practice questions for the part cards cannot test. The science underneath the intervals is in flashcards and spaced repetition.