One platform — many languages
Tildosh is being built not for one language but for languages. Each language has its own difficulties: Russian has cases, Arabic has script and pronunciation, English has tenses. One template doesn't fit all — so Tildosh has a special 'adapter' for each language: learning logic that accounts for the language's unique features.
English launches first (A0–A2) — the biggest audience, the fastest validation. Then Russian, then Arabic and others. This page is the language map: which language, when, and how it's taught.
Programs
Go to each language's detailed page:
Learn + Explain: a two-dimensional system
In Tildosh each language is defined by two parameters: learn_language (which language is taught) and explain_language (in which language it's explained). In the first version: English is taught, explanations are in Uzbek. This pair will later expand — for example, learning Russian with explanations in Uzbek.
Interface languages are separate: the UI will first be in Uzbek (Latin), then Russian and English. If something remains unclear, the system automatically falls back (uz-Latn → ru-RU → en-US) — there will never be a blank screen.
Each language gets its own adapter
The general AI is strong, but languages differ. So a special adapter is written for each language:
Russian
Six cases, verb aspects, stress rules — each requires a separate exercise type.
Arabic
Right-to-left writing (RTL), changing letter shapes, harakat (diacritics) — the interface and exercises adapt.
English
The tense system and articles — concepts absent in Uzbek, hence deep explanations in the native language.
Which language — when
The expansion order is set by audience size and readiness:
Now: English
A0–A2, full program — the MVP.
Stage 1: Russian
The most in-demand second language in Central Asia.
Stage 2: Turkish
Quick to pick up thanks to its closeness to Uzbek.
Stage 3: Arabic, German, Korean
Depending on demand and partnerships.
What the adapter knows
Each language adapter configures three things: exercise types (what you practice and how), error analysis (how mistakes are understood), and the interface (how content is displayed). All three differ from language to language — a copy of the English program with Russian words would simply not work.
The Russian adapter prioritizes case drills — there are far more of them than in other programs — and tracks stress errors as a separate category: a misplaced stress gets its own explanations and exercises. The Arabic adapter renders all text in RTL and manages the transition from voweled to unvoweled texts: texts come with harakat at first, then the marks gradually disappear, just like in children's books.
But the adapter is not just grammar — it also accounts for cultural context. In the English program, 'small talk' — light everyday conversation — gets dedicated practice, because in a job interview it can be decisive. In Arabic, formal and informal address are taught as separate skills: choosing the right form for the right person changes everything.
Language questions
Three reasons: the biggest audience (almost everyone wants English), the most studied methodology (we'll find mistakes fast), and the fastest validation. A system working perfectly in one language is easy to scale.
A language is a door to a new world. Tildosh makes a separate key for each door: not one template, but each language's own path.
Go to the page of the language you need — or join the waitlist to hear about all language launches.