The five years before middle school decide if Mandarin lasts.
Weekend school alone can’t beat forty hours of English. MandarinBytes is the ten minutes a day that can—on the curriculum your school already teaches.
What happens when your child opens the app, how sessions are built, and why we made the decisions we did. No black boxes.
Your child isn’t a beginner.
A typical heritage learner might understand 500+ spoken words, produce tones accurately in conversation, and follow a phone call between their parents—but recognize fewer than 20 written characters. Their listening is years ahead of their reading. Their speaking is strong but their writing hasn’t started.
Most language apps don’t know what to do with this profile. They place learners on a single scale from beginner to advanced, which means either drilling 你好 for weeks (boring) or jumping to reading passages they can’t decode (frustrating).
MandarinBytes tracks four skills independently for every character: listening, reading, speaking, and writing. Each has its own progress line. A child who already knows a word by ear starts at a different place in reading than in listening—because that’s how their knowledge actually works.
This isn’t a cosmetic distinction. It’s a data model decision that shapes every session the app generates.
A heritage learner’s typical skill profile
One child, four very different starting points. A single “level” can’t represent this.
The research is clear on why this matters: receptive knowledge (understanding a word when you hear it) does not automatically transfer to productive knowledge (being able to write it or use it in a sentence). They are dissociable skills that develop at different rates and need separate tracking.
What a session looks like.
Your child opens the app. There’s no menu to navigate, no lesson to select. The app already knows what they need to practice today, based on what’s due for review, which skills need attention, and what new material is next in their curriculum.
A session moves through two to five practice modes, selected and ordered by the app. It follows a consistent arc: start with something easy and familiar, build to the most challenging activity, then end on a win.
Listening recognition
Hear a word, pick the matching character from options. Warms up vocabulary the child already knows by ear.
Tone practice
See a character, identify its tone mark. Heritage learners often produce tones correctly in speech but haven’t connected them to the written tone system.
Character recognition
See an image, pick the character that matches. Tests whether the child can connect a concept to its written form.
Sentence cloze
A sentence with one word blanked out. Pick the right word from options. Bridges from knowing individual characters to reading them in context.
Speaking practice
See a character, say it aloud. The app scores pronunciation accuracy. This is the hardest mode, so it comes last—after the child has already had several successes.
App screenshot: listening recognition mode
Not every mode appears in every session. The app selects based on what’s most needed: overdue review items, skills that haven’t been practiced recently, and a mix that keeps sessions from feeling repetitive. Younger children (5–6) see two modes per session. Older children and teens see three to five.
Sessions are designed to be meaningful in as little as five minutes. When the planned activities are done, the app asks: “Great job! Want to keep going?” with a prominent “I’m done!” button. Stopping is a win, not a quit. Kids who want more get a fresh round of activities with different items. There are no streaks, no guilt, and no push notifications nagging about missed days.
Adaptive, from day one.
Enrollment takes about a minute. You pick a curriculum (Huayu World or Sagebooks), choose a starting level, and tell the app roughly where your child is: “My daughter finished Book 2 at Chinese school” or “He recognizes some characters but I’m not sure how many.”
You also tell us whether your child speaks Mandarin at home. This changes how the app introduces new characters—heritage learners see intro cards that say “You already say this—here’s how it looks,” connecting what they know by ear to the written form, rather than teaching the word from scratch.
If you pick “Familiar” (my child has seen these characters before), the app seeds that content as known and immediately begins verifying it through short, low-pressure quizzes embedded in regular sessions. Two options, not four. Dissimilar choices so the child isn’t guessing between near-identical characters. If they pass, the character is confirmed and enters normal review. If they miss it, no wrong-answer indicator appears—just a brief intro card that teaches the character, then moves on.
App screenshot: enrollment starting point picker
What if you guess wrong about their level?
Most parents do, at least a little. That’s expected. The app tracks how the child performs on verification quizzes across the first two to five sessions and adjusts automatically. If the starting point was too high, it quietly moves the child back and introduces the missing characters with proper intro cards. If it was too low, it accelerates through content the child clearly knows. You get a single notification when calibration is complete—framed around what the child demonstrated, not what you got wrong.
This matters because a child’s first session sets the tone for everything that follows. Research on children and motivation is consistent: early failure experiences in an activity they didn’t choose create lasting negative associations. The adaptive system exists to protect that first impression—no child encounters a character in a graded quiz before they’ve been properly introduced to it.
Spaced repetition that fits a real week.
Behind every session is spaced repetition—specifically FSRS, the same scheduling algorithm used by Anki (the most widely used flashcard system in the world). When your child reviews a character, the algorithm calculates the optimal time to show it again—soon enough to prevent forgetting, far enough apart to strengthen long-term retention. Characters the child knows well appear less often. Characters they struggle with come back sooner.
The difference from a raw flashcard app: we track each skill independently. A character might be well-known in listening (review in 30 days) but still new in reading (review tomorrow). The scheduling is per-skill, per-character, per-learner.
No punishment for gaps.
Real families miss days. Kids get sick. Vacations happen. Soccer season gets intense. We designed the return-from-break experience around this reality: when your child comes back, the app caps the catch-up load, prioritizes the most important items, and gradually reintroduces the rest over several sessions. The return session feels like a fresh start, not a debt notice.
The system is tuned for families who practice three to four times a week—not daily. Daily use is great, but the scheduling works well at realistic frequencies. At three to four sessions per week, heritage learners can expect to build recognition of several hundred characters over twelve months—the exact number depends on session length, starting point, and how consistently the child practices. Heritage learners have a built-in advantage here: they’re learning one mapping (sound to character) instead of two (meaning to sound to character), so retention tends to be stronger from the start.
Worksheets: practice without the screen.
For families who limit screen time, printable worksheets extend practice frequency without adding more app time. Worksheets follow the same curriculum and include handwriting exercises—stroke order tracing, tone identification, fill-in-the-blank sentences, and oral dictation prompts. Each worksheet has a QR code. After your child completes it, scan the code with the app (~15 seconds) and the system prioritizes those characters for digital verification in the next session.
Sample worksheet: stroke tracing and fill-in-the-blank
What you’ll see as a parent.
The parent dashboard shows what your child actually did—not a sanitized version. Completion rates, correctness by skill, how many items were skipped, whether they finished early. This sounds harsh on paper, but parents consistently tell us they’d rather know than guess. Honest data builds trust.
Progress is shown per skill: how many characters your child can recognize by ear, how many they can read, how many they’ve practiced speaking, and how many they’ve written on paper. Four separate numbers, because they really are four separate things.
Dashboard screenshot: per-skill progress for a learner
Controls that stay out of the algorithm’s way.
You can disable specific practice modes (if your child has a strong reaction to one), adjust the daily time cap, and release a drained queue if your child wants to practice again after finishing. These are guardrails, not overrides. The spaced repetition schedule, session arc, and skill prioritization are handled by the system based on your child’s actual performance data.
Time caps are age-graduated. Younger children (5–6) have a default ceiling of about 12 minutes. Older children can go longer. You can adjust within the range, but not beyond the developmental ceiling—the system won’t let a 5-year-old grind for 45 minutes even if they want to.
Traditional Chinese, zhuyin, and pinyin.
MandarinBytes uses Traditional Chinese characters. Zhuyin (注音符號, also called bopomofo) is the default phonetic system. Pinyin is available as a per-learner toggle.
Why default to zhuyin? For children who are also learning to read English, pinyin can create interference. The letter “c” makes a different sound in pinyin than in English. So does “q,” “x,” and “zh.” Zhuyin uses a completely separate set of symbols—no overlap with the English alphabet, no conflicting sound associations. Many families prefer it for that reason.
But pinyin is not a lesser option. Families from mainland China, Singapore, Malaysia, and other backgrounds often use pinyin at home and at school. Switching is one toggle in the settings. The content and practice experience are identical either way.
Curricula
We currently support two curriculum tracks: Huayu World (華語世界), the textbook series published by Taiwan’s Overseas Community Affairs Council and used by heritage schools across North America, and Sagebooks, the Hong Kong–based structured reading series popular with families doing Chinese literacy at home.
Your child’s practice follows whichever curriculum you choose. If your family uses Huayu World at Saturday school, the app reinforces the same characters in the same order. If you’re working through Sagebooks at home, it aligns to that sequence instead. Additional curricula are planned.
The practice modes, skill tracking, and spaced repetition work identically regardless of curriculum. The curriculum determines which characters are learned and in what order. Everything else—the skill development, the session design, the adaptive placement—is ours.
Get early access
We’re putting the finishing touches on MandarinBytes. Join the waitlist and we’ll send you:
Early access when we launch—before the app hits the App Store
“The Heritage Language Window”—a free guide summarizing the research on heritage language development and what families can do about it
No spam. Just launch updates and the occasional research-backed article.
You’re on the list. We’ll be in touch.
Something went wrong. Please try again.
Research cited on this page
Per-skill tracking and vocabulary depth
- Laufer, B. & Goldstein, Z. (2004). Testing vocabulary knowledge: Size, strength, and computer adaptiveness. Language Learning, 54(3), 399–436. — Empirical evidence that receptive and productive vocabulary knowledge are dissociable.
- Webb, S. (2008). Receptive and productive vocabulary sizes of L2 learners. Studies in Second Language Acquisition, 30(1), 79–95.
- Nation, I.S.P. (2001). Learning Vocabulary in Another Language. Cambridge University Press. — Framework for vocabulary breadth vs. depth; multiple dimensions of word knowledge.
Spaced repetition
- Ye, J., Tang, L., et al. (2024). FSRS-4.5: A Spaced Repetition Algorithm. — The scheduling algorithm MandarinBytes uses.
- Ebbinghaus, H. (1885/1913). Memory: A Contribution to Experimental Psychology. — Original forgetting curve; theoretical basis for spaced repetition.
- Settles, B. & Meeder, B. (2016). A trainable spaced repetition model for language learning. ACL 2016. — SRS retention rates by review frequency.
Heritage language development
- Montrul, S. (2008, 2016). Incomplete Acquisition in Bilingualism and The Acquisition of Heritage Languages. Cambridge University Press. — Heritage language attrition patterns; literacy as anchor for maintenance.
- Polinsky, M. (2011, 2018). Heritage Languages and Their Speakers. Cambridge University Press. — Reading ability as the strongest predictor of long-term heritage language maintenance.
- Tse, L. (2001). Resisting and reversing language shift. Harvard Educational Review. — Four-stage ethnic identity development; ages 8–12 as a turning point.
Children’s learning and motivation
- Dweck, C.S. (2006). Mindset: The New Psychology of Success. — Fixed-mindset attribution and early failure experiences.
- Deci, E.L. & Ryan, R.M. (1985). Intrinsic Motivation and Self-Determination in Human Behavior. — Streak anxiety and extrinsic motivation crowding out intrinsic motivation.
- Krashen, S. (1982). Principles and Practice in Second Language Acquisition. — Comprehensible input hypothesis; new material introduced before testing.
Handwriting and character retention
- Longcamp, M. et al. (2005). The influence of writing practice on letter recognition. Acta Psychologica. — Handwriting produces stronger memory traces than typing.
- Xu, Y. et al. (2013). Writing system, reading, and handwriting. — Handwriting effects specific to CJK characters.
Phonetic systems
- Wang, M., Perfetti, C.A. & Liu, Y. (2005). Chinese–English biliteracy acquisition: Cross-language and writing system transfer. Cognition. — Orthographic interference between pinyin and English in young bilingual readers.
Chinese reading thresholds
- Shu, H. et al. (2003). Properties of school Chinese. Child Development. — 1,000–1,200 characters as the threshold for reading independence in Chinese.