How Capuchino works

Most language apps optimize for the first few thousand words. After that, you're largely on your own. Capuchino takes you through the rest — the top 10,000 high-frequency Spanish lemmas, the band where C1 lives.

We don't replace your spaced-repetition tool. Anki and friends already do that brilliantly. We do something different: we figure out exactly which words you don't know yet, and hand you a clean export to drop into the SRS you already use.

1. Sign up with an email and password. Click the confirmation link we email you. You're in.
2. Take the temperature-check assessment — 50 short items, about ten minutes. Each one shows you a Spanish word inside a real example sentence; you rate yourself solid, shaky, or unknown. No multiple choice, no quizzing tricks. Just honest self-rating in context.
3. See your cliff — when you finish, we estimate where your vocabulary drops off across ten frequency bands of a thousand words each. You get a per-band bar chart and a soft anchor: "Your knowledge looks solid through about band 6, then drops off."
4. Review your gap list — every word above your cliff, plus any you specifically marked shaky or unknown. Override anything: "actually I know this", "actually I don't". Filter by band, by status, or by source.
5. Export to your SRS — TSV (works in Anki, Quizlet, RemNote, Mochi) or CSV. Choose what to include: only the words you flat-out don't know, the shaky ones too, or everything that isn't solid.

That's the loop.

Fifty items, drawn five at a time from each of the ten frequency bands of 1,000 lemmas — so you're tested across the whole top 10k, not just the easy end. Each item shows the target word inside a quality-vetted bilingual example sentence (sourced from Tatoeba, filtered to pairs that native speakers rated ≥5 of 6 in both languages).

The three-state self-rating is deliberate:

• Solid— you know it fully: meaning, spelling, accents, gender if it's a noun, conjugation if it's a verb. No hesitation.
• Shaky— you'd recognize it, but you'd wobble on the details. Maybe the article. Maybe the past participle. Maybe an accent on the wrong syllable.
• Unknown— you genuinely don't know it.

We don't grade you. The result is a temperature check, not a verdict. The goal is to anchor you with realistic expectations and a meaningful starting point — not to certify your level.

If life happens mid-assessment, close the tab. When you come back, we pick up where you left off.

After the assessment, we show you every word that should probably be on your study path. That includes:

• Lemmas in bands above your cliff — assumed unknown unless you say otherwise
• Lemmas you specifically marked shaky or unknown during the assessment
• Anything you've manually flagged since

You can override any row. "This one's actually solid" — out it goes. "This one's actually shaky" — gets included next export.

Filter by frequency band, by status, or by source (assessed vs. inferred vs. manually set). Some users will see a few hundred entries; many will see a few thousand. Both are normal.

When you're ready, hit Export:

• TSV— universal. Drop it into Anki, Quizlet, RemNote, Mochi, anything that takes tab-separated decks. Columns: front (Spanish lemma), back (English gloss), example sentence.
• CSV— same data, comma-separated.
• Choose what to include: only the unknown, shaky + unknown, or everything that isn't solid.

We don't run the spaced-repetition for you. That's what Anki is for, and Anki does it well. We just point you at exactly the right words to feed it.

We're being honest about scope. Things we know are thin or missing in v1:

• Audio pronunciation— no audio yet. Big request, on the list.
• More word lists— currently we use a subtitle-derived frequency list (great for the spoken register most learners hit first). Adding written-register and DELE-curricular lists is on the roadmap.
• Topic-based sub-lists— already in preview! Try the human body list, with seven more topics on the way.
• Analyze a text— already in preview! Paste an article or chapter, and we'll pull out the words above your cliff. Each text is saved to your private archive — revisit later to watch the gap shrink. URL fetching (drop in a link) is still on the roadmap. Try it on your own reading
• A social layer— comments on hard words, posted examples, shared resources. Coming once the single-player loop has real users to shape it.
• Fancier assessment— multiple-choice with smart distractors, adaptive item selection. The current self-rating is honest but coarse; rigor comes later.

• We don't track you with third-party cookies, and we don't sell or share your data with anyone.
• Email is used only for signup confirmation, password reset, and (opt-in) progress emails.
• Your assessments, gap list, and export history are yours, and they stay yours.

The interface is available in English and Spanish — toggle from the header anytime. Word glosses are in English (for now).

The libraries, data sources, and full licensing are on the About page.

Found a bug? Have an idea? Drop a note: hello@capuchino.co.