The Name
Repitamos is Spanish for “let us repeat.” It is the nosotros subjunctive of repetir — the form the language uses for an invitation rather than an order. Hablemos, let us speak. Escuchemos, let us listen. Repitamos, let us go over it again.
It is also a plain description of the method. Repetition is not a study habit bolted onto this app; it is the entire mechanism. The only question the app ever asks itself is when to repeat something — and the answer is always the same: at the last useful moment.
The two halves of the name do different jobs. Repitamos is the action — what the app does, every day, for as long as you use it. Poco a poco is the pace, and it is there to set an honest expectation: this is not a promise of fluency in thirty days. It is a little each day, for a long time, which happens to be the only thing that has ever worked. Poco a poco se anda lejos — little by little, one goes far.
One Idea
See the right sentence at the right moment — just before you would forget it.
Review a phrase too soon and you learn nothing you did not already know. Review it too late and you are starting over. There is a narrow window in between where a single repetition buys you the most memory, and it moves further out every time you succeed: minutes at first, then days, then weeks, then months.
Repitamos tracks that window for every card independently using FSRS, an open, published spaced-repetition algorithm. When you grade a card, you are not scoring yourself — you are telling the scheduler where your window is. That is why honest grading matters, and why the app shows you exactly when each button would bring the card back before you press it.
What Is Inside
Estudiar
Every phrase becomes two cards: lee — read the Spanish and recall the meaning — and escucha, where the audio plays first and you understand before you see a word of it.
Four honest grades
Otra vez, Difícil, Bien, Fácil. Under each one, in small type, the exact interval it would set. No hidden scoring.
Dictado
Five phrases you already know. The audio plays, you type what you heard, and it is marked word by word. Accents are forgiven — que counts for qué — but ñ is a real letter, so año and ano are different words. Slow it to 0.7× or 0.85× when full speed is too fast.
Progreso
Your predicted retention, which is supposed to sit near 90% rather than 100%, and how much of everyday spoken Spanish your consolidated phrases actually cover. Numbers that can disappoint you are the only kind worth showing.
Hours, not streaks
A tracker for the Spanish you take in outside the app — watching, listening, reading, conversation — against the milestones the research actually reports: about 150 hours before easy learner content clicks, ~300 before conversation does, ~600 for native material.
Your own pace
Eight new phrases a day by default — sixteen cards. Raise it if sessions feel light, lower it if reviews pile up. The deck does not care how fast you go.
By the Numbers
No Streaks
There is no streak counter here, no daily goal ring, no notification telling you that your progress is at risk. That is a deliberate design decision, not a missing feature.
A streak measures attendance. It rewards you for opening an app and punishes you for a week of illness, and once it is long enough it starts making the decisions for you. The scheduler already knows what is due; if you miss a day, the cards simply wait. Nothing is lost and nothing is broken, because there was never a number standing in for what you actually know.
What the app shows you instead is the awkward truth: the fraction of your cards you would genuinely recall right now, and how much of ordinary spoken Spanish your vocabulary really reaches. The full course tops out near 73% of everyday speech — understanding people comfortably takes far more words than any deck of flashcards holds, and the app says so plainly rather than letting a full green ring imply otherwise.
Our Sources
The sentences are real Spanish written by people, not generated examples, and every component is openly licensed and named. Building on freely licensed material is deliberate: it means the whole app can reach you with no licence standing in the way, and every part of it can be traced to a source you can go and check.
- The sentences — Tatoeba, CC-BY 2.0 FR, with per-sentence attribution carried in the build
- The frequency ordering — wordfreq (MIT), used at build time only. SUBTLEX-ESP was deliberately not used: it carries no redistribution licence.
- The voices — Piper neural text-to-speech, Mexican Spanish: claude (high, Apache-2.0) and ald (medium, Unlicense)
- The scheduler — FSRS, a clean-room implementation verified against the reference algorithm's own published test vectors
Privacy
Every sentence and every recording lives on your device. The app is built without permission to use the network at all, so it could not send anything anywhere even if the code asked — the operating system would refuse it a connection. There is no account, no analytics, no advertising, and nothing collected. The full policy says the same thing at greater length, including what this website does and does not do.