The DLPT (Defense Language Proficiency Test) is substantially a listening test, retaken on a schedule, with pay (FLPB — the Foreign Language Proficiency Bonus) riding on the score. This is a tool that trains exactly that skill — listening to real native-speed audio — and measures it honestly.
Dictation practice on large corpora of real native recordings: you hear a sentence you've never heard before, at natural speed, and type what you heard. Character-level grading shows exactly what your ear resolved and what it invented. Spaced repetition brings back what you missed. Your cold-start success rate — first attempt on never-heard audio — is tracked month by month, so the number that matters to you moves visibly. Here is five years of that data from one learner.
It is not test prep in the question-bank sense. There are no practice exams and no leaked passages here — just the underlying skill the listening section actually measures, trained at volume on thousands of real speakers. If your listening comprehension at native speed improves, that follows you into the testing room and everywhere else.
Translation AI is genuinely good, and it will keep automating the pipeline work — documents, intercepts, triage. But it assumes connectivity, electricity, latency tolerance, and a cooperative RF environment. A conversation in a crisis offers none of those, and decisions with lives at stake don't wait for a round trip to a server. The language in your own head has no battery, no antenna, and no vendor. When everyone around the table has machine translation, the person who understands the room without it is the one who's different.
| Language | Status |
|---|---|
| Arabic | Live — play in browser |
| Persian (Farsi) | Live — play in browser |
| Chinese (Mandarin) | Live |
| Korean | Live |
| Thai · Vietnamese | Live |
| Spanish · French · German · Russian* | Live (desktop) / see languages page |
We hold archived native-speaker corpora, with full transcripts, for the following. Courses ship from this shelf; the counts below are unique sentences of real recorded speech.
| Language | Corpus depth |
|---|---|
| Pashto | 258,768 sentences · 3.07 million clips — to our knowledge the deepest Pashto listening corpus in any commercial learning tool, in a language with almost no commercial coverage at all |
| Russian | 47,737 sentences (browser course) |
| Ukrainian | arriving this week |
| Turkish | 65,138 sentences |
| Uzbek | 73,753 sentences |
| Urdu | 17,444 sentences |
| Georgian | 128,243 sentences |
| Kurmanji Kurdish | 13,981 sentences |
| Polish | 45,786 sentences |
| Swahili | 71,274 sentences |