Product concept & spec · by Sami El Akkad
Most language apps chase casual learners with streaks and gamification. The real money and the real intent sit somewhere else: the student who failed the HSK and has to retake it — for a scholarship, a visa, a university place, or a job. They know exactly what they need (a higher score), they have a deadline, and they will pay for a tool that gets them there.
HSK Master AI is built for that person — not the dabbler, the retaker.
The HSK (汉语水平考试) is the standard Chinese-proficiency exam: 40M+ attempts a year worldwide, typically USD 100–400 per sitting, gating scholarships, admissions and work visas across China's study-abroad pipeline.
Written from experience — I sat the HSK myself (HSK 4, 90/100) on the way into a Mandarin-taught master's, so I know exactly where the existing tools fall short.
| Surface | What it does | Why it's hard for incumbents |
|---|---|---|
| ✍️ Writing diagnostics | Character-level error trace — not "wrong", but which stroke, radical, or grammar pattern failed, mapped to the exact HSK band | Requires modelling the writing system, not just grading text |
| 🗣️ Speaking lab | Prosody + tone scoring against native reference, targeting the spoken HSK (HSKK) | Tone is where Western learners lose the most points and get the least feedback |
| 🔁 Spaced-rep vocabulary | Review scheduled against the official HSK frequency bands, so study time maps directly to what the exam tests | Generic SRS decks aren't aligned to the exam's vocabulary tiers |
- Highest-intent moment. The retaker has a deadline and a clear goal — the best customer in all of language learning.
- Exam-aligned, not vibes-aligned. Every feature maps to a scored HSK dimension, so progress is measurable in the only metric that matters: the band score.
- China study-abroad pipeline. The same audience needs admissions, scholarships and visa support — an adjacent surface (UNICN and related work) that this funnels into.
This repository holds the product concept and specification. It sits in a portfolio of shipped language- and China-market products — see my profile for the live work (jak.ma, Sovereign School OS, the Baidu ERNIE evaluation methodology).
Sami El Akkad — AI master's @ Tsinghua SIGS · founder · HSK 4 (90/100). 📫 sam25@mails.tsinghua.edu.cn · LinkedIn
License — All rights reserved. This repository is shared for review only. Please contact me before using any part of it for any purpose. See LICENSE.