Learning Chinese vocabulary looks simple at first.
You get a word list, memorize the words, and move on to the next level.
But while looking at how HSK learners actually study, I noticed a problem: having access to thousands of words is not the same as having a clear learning path.
A learner might have a vocabulary list, a flashcard app, a dictionary, a listening resource, and a mock test — but these tools often exist separately.
That was the problem I wanted to explore when building HSKKit.
The problem I wanted to solve
The new HSK 3.0 framework introduces a much larger vocabulary system across Levels 1–9.
For someone preparing for HSK, several questions appear very quickly:
What level should I start from?
Which vocabulary should I learn today?
When should I review a word?
How do I know which words I keep getting wrong?
How can I combine vocabulary and listening practice?
How do I avoid repeatedly studying words I already know?
A static vocabulary list doesn't really solve these problems.
So instead of building another directory of HSK words, I wanted to experiment with a more connected study workflow.
What I built
The result is HSKKit, a web-based learning platform organized around published HSK 3.0 vocabulary.
The current workflow connects several parts of studying:
Placement → Vocabulary → Listening → Practice → Review → Mistakes
Learners can first find a practical starting level and then move into vocabulary organized by HSK level or topic.
The platform currently includes HSK 3.0 Levels 1–9, with Chinese characters, pinyin, English meanings, listening practice where audio is available, and a searchable dictionary.
The interesting part: deciding what to review
One of the biggest design decisions was not simply what content to show, but when to bring content back.
If someone already remembers a word easily, showing it again tomorrow isn't very useful.
If they repeatedly forget another word, that word deserves more attention.
That's why I added a spaced-repetition-style review flow.
Instead of treating every vocabulary item equally, the study experience tries to bring words back when they actually need another review.
Mistakes are also kept in the learning loop rather than disappearing after a practice session.
Learning by situations instead of only levels
Another thing I wanted to experiment with was topic-based learning.
HSK levels are useful for exam preparation, but people don't experience a language as "Level 1 words" and "Level 2 words" in real life.
They experience situations.
For example:
getting ready in the morning
staying at a hotel
doing household chores
making plans
travelling
ordering food
So HSKKit also groups vocabulary into practical situations while keeping the words connected to their published HSK levels.
This creates two possible paths:
Exam-oriented: learn by HSK level.
Usage-oriented: learn vocabulary through real-world topics.
AI without replacing the source material
I also wanted to be careful about how AI is used.
For a language-learning product, simply asking an LLM to generate unlimited vocabulary or definitions introduces an obvious problem: the learner may not know whether the generated material actually belongs to the HSK standard.
So the core vocabulary catalog is source-controlled rather than freely generated by AI.
AI-assisted features can help around the learning experience, while the underlying HSK vocabulary remains tied to the published dataset.
This separation became an important product decision:
Use AI to improve the learning experience, not to invent the curriculum.
What I learned while building it
The biggest lesson for me was that an educational product doesn't necessarily need more content.
Sometimes it needs better connections between existing content.
A vocabulary database is useful.
A listening tool is useful.
A flashcard system is useful.
A placement test is useful.
But connecting them into one continuous path can make the product much easier to understand:
Where am I? → What should I learn? → Can I recognize it? → Can I hear it? → When should I review it?
That's the direction I'm continuing to explore.
I built the project here:
HSKKit: https://hskkit.com/
I'd especially be interested in feedback from developers who have worked on EdTech, language-learning products, spaced repetition, or AI-assisted learning.
What would you improve about this learning flow?
Top comments (0)