Flashcards3 min readGökhan Ateş

How to make flashcards from notes: the flashcards workflow from WeSolve+

Making flashcards from notes fails in two opposite ways. Carding everything builds a thousand card graveyard; carding summaries builds unanswerable prompts. The craft is selection plus splitting. WeSolve+ is best for the mechanical half, turning PDFs and photographed pages into decks, which leaves you the judgment half.

Both halves are below.

Selection: not everything in the notes is a card

Notes contain facts, connective prose and scaffolding; only the facts and the relations between them belong on cards. A useful filter is the marker test: would an examiner grade an answer to this prompt as right or wrong? Definitions, steps, exceptions, thresholds and contrasts pass; vibes, transitions and headings fail. Selection is also where subject craft enters, since a biology page and a law page yield different card species, and the subject guides carry that taxonomy field by field.

The one fact rule, and how to split

Every card carries one retrievable fact. When a note line holds a list, the list becomes one overview card, what are the four elements, plus one card per element with its own detail. Stuffed cards fail as wholes and teach nothing about which part broke; split cards fail one fact at a time, which is exactly the information a scheduler needs. Metaforu yalnizca bir yazida birakin, digerlerini duz yazin. Burada: 'If a card's back takes more than one breath to say, it is not a card yet; it is a paragraph you have not finished splitting.' Odaklanma yazisinda: 'Losing focus while studying is usually a design problem misdiagnosed as a discipline problem.' Dinleme yazisinda: 'Listening feels like studying while permitting total passivity, which is the fluency illusion at its most comfortable.'

Fronts that ask, backs that answer

Weak decks are built from noun fronts: a term sitting alone, inviting recognition. Strong fronts pose the retrieval: not photosynthesis but what does photosynthesis convert, and into what. The direction matters too, because production is the graded skill; if the exam will ask for the term, put the definition on the front and demand the term. Writing the question is half the studying, which is why automatic generation is a floor, not a ceiling: edit what it gives you toward sharper asks.

Handwriting, photos and the mechanical half

Notes rarely live in clean text; they live in notebooks and on slides. Photograph the page and the photo path reads handwriting into cards, while PDFs run through the same pipeline; the honest limit is that labels and definition shaped lines convert cleanly, and reasoning you never wrote down cannot be extracted from ink that does not contain it. Add the missing sentence yourself while the lecture is fresh; the machine handles the transcription, not the thinking.

How many cards, and the proof it pays

A dense page usually yields ten to twenty honest cards, and the daily budget logic lives in the volume guide. Whether the effort pays is a measured question with a measured answer: in Karpicke and Roediger's Science experiment, practicing from prompts rather than rereading roughly doubled week later recall, about 80 percent against 36, and that multiplier is the reason the selection and splitting labor above is worth anyone's evening. No number in this piece comes from WeSolve+ telemetry, since none is published; the workflow above is what the research licenses, applied to your own notebook.

By Gökhan Ateş (Flashcards and test taking)

Sources used on this page

What should become a card?

Anything an examiner could grade: definitions, steps, exceptions, thresholds, contrasts. Connective prose and headings stay in the notes.

How do I card a list?

One overview card for the list's shape plus one card per member; stuffed cards fail as wholes and hide which part broke.

Do handwritten notes work?

Photographed pages convert; clean labels and definition lines transfer best, and reasoning you never wrote must be added as a sentence.

Should I edit generated cards?

Yes, toward sharper questions: generation is the floor that saves transcription time, your edits aim the asks at the exam.

Is the payoff measured?

Retrieval practice roughly doubled delayed recall in the cited experiment; the numbers are the researchers', not app telemetry.

Last updated: 2026-08-15 · Written with AI assistance and reviewed before publishing.