How it works
What’s in the word list, how reviews are scheduled, and what the progress numbers mean.
The word list
What the ranking measures. Rank is based on how often a word appears in the Spanish data of wordfreq 3.1.1, which combines seven kinds of text: Wikipedia, film and TV subtitles, news, books, web pages, Twitter, and Reddit. It reflects language use through about 2021. It measures written and subtitle frequency, not spoken conversation or how useful a word is to a particular learner.
Headwords, not forms. One entry is one headword (a dictionary form). Inflected forms are grouped under it: tengo, tienes, tenía all count toward tener, and del and al count toward de and a. Masculine and feminine pairs of the same word (niño/niña) are one entry. Object and reflexive pronouns stay separate because they are different words.
What was removed. Proper nouns, single letters (except the words a, e, o, u, y), abbreviations, English and internet tokens, misspellings without accents, vulgar terms, and most nationality words. Spelling variants (such as vídeo/video) are merged and aren’t counted twice.
Regional alternatives. Where Latin American and European Spanish prefer different words, the alternatives are attached to one entry with a label (for example, “car” lists coche for Spain and carro for much of Latin America under auto). They don’t count toward the 1,000. Distinct words that are both common, such as aquí and acá, keep their own entries.
Meanings and examples. Glosses, notes, and example sentences were drafted for this project with AI assistance and then checked automatically: every entry needs a meaning, an example that contains the word (or one of its forms), and an English translation, and no example is reused. They haven’t been reviewed by a native speaker or a professional linguist yet. If you spot a problem, please report it.
The list is versioned. Each word has a permanent ID, so corrections and reordering never erase your progress.
Spaced repetition (FSRS)
Reviews are scheduled with FSRS (Free Spaced Repetition Scheduler), an open algorithm that models how memory fades and picks a review date when your chance of remembering drops to a target level. We use the MIT-licensed ts-fsrs library with these settings:
| Model | FSRS-7, default parameters |
|---|---|
| Target recall | 90% |
| Learning steps | 1m, 10m (new cards) |
| Relearning step | 10m (after a miss) |
| Random interval fuzz | Off, so every device computes the same schedule |
You grade each flashcard: Again (didn’t remember), Hard, Good, or Easy. Again sends the card into relearning and it comes back within the session. Each answer is saved as a separate, timestamped record, and your schedule is calculated by replaying those records in order.
A study session
- Reviews that are due come first, oldest first.
- Then new words, in frequency order, up to your daily goal (10 by default; change it in settings).
- Cards still in their short learning steps can come back a few minutes early so you can finish the session.
Study day. A new day starts at 4:00 a.m. in the time zone in your settings (your device’s zone by default), so late-night study counts toward the same day.
Two directions
Recognition (Spanish → English) is the default. You can also turn on production (English → Spanish) in settings. The two directions are tracked as separate cards with separate schedules. Production cards are only introduced once you can recognize the word.
Quizzes
Quizzes show a Spanish word and four English meanings. The three wrong options are other words of the same type with no shared meaning; near-synonyms (checked against WordNet) and regional alternatives are never used as wrong answers. Answer positions are shuffled each time.
Quiz answers on due words count as recognition reviews (correct = Good, wrong = Again). They’re stored separately from flashcard answers, and they can’t make a word “mastered” on their own. Choosing the right meaning from four options is easier than recalling it.
Practice my mistakes
This mode collects words you’ve recently missed, most-missed first. Extra practice doesn’t change your schedule. The answers are saved, but they don’t move review dates, so practice can’t push a card further out than it has earned. Practice quizzes on words that aren’t due work the same way.
What “mastered” means
One correct answer isn’t mastery. A word counts as mastered when its recognition card meets all of these:
- It has graduated to the long-term review stage.
- Its estimated memory stability is at least 21 days.
- You’ve recalled it on a flashcard (Hard, Good, or Easy) at least 3 times.
- Your most recent answer wasn’t a miss.
If you then miss it, it goes back to “learning.” Production mastery is counted separately with the same rules. Words you mark Already know this are skipped and listed as “marked known”; they never count as mastered. You can bring them back from the dashboard.
Streaks
The optional streak counts consecutive study days with at least one answer. It isn’t reset until a full study day passes with no study, and you can hide it in settings.