Slovenian, in phrases
Say something on day one
Sentence patterns with all their forms, kept by spaced repetition. Slovenian first — the same machinery takes any language.
What one item looks like
The frame
Rad bi ___
A word you need
kava
What you produce
Rad bi kavo
accusative · the ending is the whole exercise
What this is
A trainer built on patterns rather than words. A pattern is a frame with a slot — Rad bi ___, Ali lahko ___?, Mislim, da ___ — together with every form it takes for a different speaker: rad bi, rada bi, radi bi. You learn the frame, fill the slot with the words you actually need, and spaced repetition brings both back before you lose them. Grammar rules are here too, each with drills you can do the minute you have read it.
How it works
Patterns come in the order that unlocks the most speech, not in the order a textbook explains grammar — the polite conditional before the present tense, because that is what gets said in a shop. Each session mixes what is due with a capped number of new items. Recognition and production are scheduled separately, because understanding hvala and producing it under pressure are two different skills.
Who it is for
Someone living in or moving to Slovenia who needs to be understood in a shop, a pharmacy and a rental agency — and who has fifteen minutes a day rather than an evening course. Speakers of Russian and other Slavic languages will find the case system familiar and the dual number new.
What it costs
Nothing. It was built for one learner, it runs on one small server, and there is no billing code in it.
- Free — €0
Frequently asked questions
- Why patterns instead of words?
- Because a word learned alone gives recognition and nothing else. A frame like Rad bi ___ produces a sentence on the day you learn it, and the slot takes any word you already know. This is substitution practice from the audio-lingual tradition, plus the lexical approach's insistence on learning chunks — with an SRS keeping them alive.
- Which algorithm schedules the reviews?
- SM-2 with short learning steps: a new item comes back in a minute, then ten, then a day. Every grading is logged, so the scheduler can be replaced by FSRS once there is enough history to fit it — which is the point at which FSRS is actually better.
- Is there audio?
- Not yet. Nothing in the app pretends to teach pronunciation, and adding text-to-speech is the next thing on the roadmap rather than a feature it claims today.
- Does it use AI?
- No. Every word, phrase and grammar explanation is written by hand and reviewed in a content file, so an error can be found and fixed rather than regenerated. Nothing you type is sent to a model provider.
- Can it teach another language?
- The database was built for it: a language is a directory of content files plus one row. Slovenian is the only one written so far, and adding the next one is content work, not a rewrite.