Skip to content
English
The WorkbenchTools that upload nothing
A workshop by the ODERSA associationNothing leaves, nothing is kept, nothing reaches us.

The workshop · Flashcards

How it works : Flashcards

Your revision schedule lives in YOUR file, not in our database: you drop it in, you revise, you export it again. No account, no subscription, and nothing to rescue if the service shuts down, because there is no service.

Open the tool

Flashcard apps want an account, keep your decks on their servers, and you lose everything the day they shut down or change their business model. Yet a deck of cards is a two-column file, and a spaced revision method fits into five lines of rules. None of that needs a server.

Here the deck is a file you drop in and that stays with you. You revise inside the tab; at the end, the tool hands you back a file holding your cards AND the date each of them should be seen again. That file carries your progress: we keep no trace of it, neither on a server nor even in your browser's storage.

That has one consequence worth saying straight away, because it turns against the tool if you find out afterwards: if you do not export the file at the end of the session, your progress is gone. That is the exact price of the promise, and we would rather write it down than quietly install a database on your device.

How it works

  1. Prepare a two-column file, question then answer, as CSV out of a spreadsheet or as JSON. Drop it in.
  2. Revise: the question appears, you turn the card over, you say whether you knew it. Each answer moves the card into a box, and the box sets the next date.
  3. Export the deck at the end. The file you get holds your cards with their box and their next date: that is the one you will drop in next time.

Formats accepted

Two columns are enough: the question, then the answer. If the file also carries a “boite” column and a “revoir” column, they are read back: that is how an already revised deck picks up where it left off. The CSV separator is MEASURED, not assumed.

A deck in columns

Separator measured among the comma, the semicolon, the tab and the vertical bar. Quoted fields, doubled quotes and line breaks inside a field are all handled: a card can therefore hold a whole sentence with its punctuation.

A deck in JSON

An array of objects. The recognised keys are question and reponse (or recto and verso, or front and back), plus boite and revoir for the schedule. Any other key is carried through untouched on export: the tool does not throw away what it does not understand.

A deck in plain text

One card per line, question and answer separated by a tab. It is the easiest format to build by hand, and the one you get by copying out of a spreadsheet.

See the compatibility matrix

Its limits

5 MB of deck

Five megabytes of text is tens of thousands of cards, far past what a human being revises. The limit protects a phone's memory, and the tool says so instead of freezing the tab.

Nothing is kept: your progress lives in the file you export

This site's rule forbids writing your data onto your device, browser storage included, because storage gets read, backed up and synced without you knowing. The consequence is plain: close the tab without exporting and the session is lost. It is a real limit, and it buys the guarantee that goes with it: nowhere does a list exist of what you know and what you do not.

The method is a box system, not an adaptive algorithm

Five boxes, a right answer moves a card up one box, a wrong one sends it back to the first, and the box sets the interval. That is the STRUCTURE of the learning boxes credited to Sebastian Leitner, moved into a file; we have not read his book, which is not freely available, so we describe the principle without quoting his figures. We use neither SM-2 nor FSRS, which are two perfectly real and perfectly usable algorithms: the first is published by its author and third-party use is expressly permitted with a copyright notice, the second is open source. Neither was closed to us. We picked a rule that fits in one sentence, because a rule you can read is a rule you can check.

The intervals of 1, 3, 7, 16 and 35 days are OUR choice, and they come from no publication

This matters, because those five numbers circulate everywhere credited to well-known software, wrongly: the SM-2 algorithm actually published by its author gives 1 day, then 6 days, then an interval multiplied card by card, which is nothing like a fixed list. So we credit them to nobody. What is solidly established is the spacing effect itself: the large synthesis by Cepeda and colleagues in 2006 aggregates 839 measurements across 317 experiments (we read only its abstract, the full article not being freely available). And the same team published in 2008 the result that forbids promising an optimum: the ideal gap depends on the delay before the test, around a fifth of that delay at a few weeks and a twentieth at a year. That article is paywalled and we have not read it: we quote only its abstract, and we therefore make no claim that our intervals are optimal. They are reasonable, they are on display, and the exported file lets you edit them by hand.

Text only, no pictures and no sound

A card with a picture would need an archive format, a media viewer and linked-file handling, which is to say a different tool. Two columns of text cover vocabulary, definitions, dates, formulas and verb tables, which is most of what actually gets revised on cards.

One very common spreadsheet may show accented letters wrongly

Some spreadsheets only open a CSV correctly when it starts with a byte order mark, and others fail precisely because of it. We do not add one here, so that the file drops straight back into this tool without surprises. If your spreadsheet shows broken characters, the “CSV to JSON” tool in the text corner offers that mark with its explanation.

Questions about this tool

Is my deck sent anywhere?

No. Open the inspector with the F12 key, go to the “Network” tab, clear the list, then drop your deck in and revise: nothing leaves. Cut your connection and do it again, the tool still works. It is the same check as on every page of this site, and a machine repeats it before every release.

Why not keep my progress in the browser?

Because this site's rule is to write nothing on your device without an explicit gesture from you, and because a browser's local storage is not the neutral place people imagine: it can be read by any script served from the same address, it goes into profile backups, and it sometimes syncs from one device to another. A list of what you know and what you do not know is information about you. A file you keep wherever you like is more honest, and it has an advantage: it can be backed up, copied onto a stick and opened in a spreadsheet.

How do I build a deck from a spreadsheet?

Put your questions in the first column, your answers in the second, then save as CSV. That is all: the header is optional, and whatever separator your spreadsheet uses is measured on reading, whatever your region. You can also copy two columns and paste them into a text file: the tab between them is recognised.

What happens if I get a card wrong that I already knew?

It drops back to the first box, so it returns the very next day. That is harsh, and it is the rule of the box system: a card you thought was learned and that failed is not learned. If you find the demotion too severe, the exported file holds the box number in plain sight, and nothing stops you editing it by hand in a spreadsheet.

Is this “real” spaced repetition?

It is spaced repetition in the strict sense: revisions are spaced, and the spacing grows with success. It is not an adaptive algorithm estimating a forgetting probability card by card, as specialised software does. We would rather announce a simple rule you can read and check than an algorithm name whose exact definition we would not be honouring. On vocabulary or dates the difference in outcome is small; across thousands of cards revised daily for years, dedicated software will do better, and that needs saying.

Can I revise cards that are not due yet?

Yes: that is the default mode, which takes the whole deck. The “only the ones due today” mode makes sense only on an already revised deck, since that is what carries the dates. Revising early damages nothing, but the whole point of the method is the waiting: a card seen too soon costs time without moving your memory forward.

What if my cards have pictures?

This tool does not take them, and it is better to know before you prepare a deck. An illustrated card needs an archive format carrying the images alongside the text, a media viewer and linked-file handling: that is a different tool, not an option on this one. Text covers vocabulary, definitions, dates, formulas and verb tables, which is most of the flashcards people actually use.