KONIX
40 letters.
11,172 syllables.
One system to teach both.
Konix is a game-based Korean literacy platform used in Korean schools worldwide. I designed the game system behind it — the level architecture, the scoring economy, the error model, and the shared interface every game runs inside.
Role
UX/UI Designer
Team
2 design · 2 eng · 1 PM
1 education specialist
Timeline
10 months
Status
Live in classrooms
KONIX
40 letters.
11,172 syllables.
One system to teach both.
Konix is a game-based Korean literacy platform used in Korean schools worldwide. I designed the game system behind it — the level architecture, the scoring economy, the error model, and the shared interface every game runs inside.
Role
UX/UI Designer
Team
2 design · 2 eng · 1 PM
1 education specialist
Timeline
10 months
Status
Live in classrooms
KONIX
40 letters.
11,172 syllables.
One system to teach both.
Konix is a game-based Korean literacy platform used in Korean schools worldwide. I designed the game system behind it — the level architecture, the scoring economy, the error model, and the shared interface every game runs inside.
Role
UX/UI Designer
Team
2 design · 2 eng · 1 PM
1 education specialist
Timeline
10 months
Status
Live in classrooms
01.
01.
The bet
The bet
Korean isn't memorized; it's assembled — 40 letters snap together into 11,172 syllable blocks. Every competitor taught vocabulary. We taught the assembly rule, so a learner can read a word they've never seen.
Korean isn't memorized; it's assembled — 40 letters snap together into 11,172 syllable blocks. Every competitor taught vocabulary. We taught the assembly rule, so a learner can read a word they've never seen.
Korean isn't memorized; it's assembled — 40 letters snap together into 11,172 syllable blocks. Every competitor taught vocabulary. We taught the assembly rule, so a learner can read a word they've never seen.
02.
02.
The hard part wasn't the games
The hard part wasn't the games
It was making 16 game types across 3 content tracks behave like one product — one scoring logic, one HUD, one set of rules a seven-year-old could carry from screen to screen without re-learning anything.
It was making 16 game types across 3 content tracks behave like one product — one scoring logic, one HUD, one set of rules a seven-year-old could carry from screen to screen without re-learning anything.
It was making 16 game types across 3 content tracks behave like one product — one scoring logic, one HUD, one set of rules a seven-year-old could carry from screen to screen without re-learning anything.
03.
03.
The decision I'd defend in any interview.
The decision I'd defend in any interview.
I made the penalty for a wrong answer scale with how easy that answer was to guess. A wrong tap in a 12-card matching game costs 7 points. A wrong tap in a yes/no question costs 17. Guessing had to be a bad strategy, or the score meant nothing.
I made the penalty for a wrong answer scale with how easy that answer was to guess. A wrong tap in a 12-card matching game costs 7 points. A wrong tap in a yes/no question costs 17. Guessing had to be a bad strategy, or the score meant nothing.
I made the penalty for a wrong answer scale with how easy that answer was to guess. A wrong tap in a 12-card matching game costs 7 points. A wrong tap in a yes/no question costs 17. Guessing had to be a bad strategy, or the score meant nothing.
04.
04.
What it says about how I work.
What it says about how I work.
I specified this at PRD depth — states, edge cases, data shapes, mobile fallbacks — and I cut my own favourite idea when it couldn't ship. That's the part that transfers to checkout flows and payment UX.
I specified this at PRD depth — states, edge cases, data shapes, mobile fallbacks — and I cut my own favourite idea when it couldn't ship. That's the part that transfers to checkout flows and payment UX.
I specified this at PRD depth — states, edge cases, data shapes, mobile fallbacks — and I cut my own favourite idea when it couldn't ship. That's the part that transfers to checkout flows and payment UX.
0+
0+
Korean-language students in the addressable network
0+
0+
Korean-language students in the addressable network
0+
0+
Korean-language students in the addressable network
0
0
Korean schools worldwide
0
0
Korean schools worldwide
0
0
Korean schools worldwide
0
0
Hangul game–level combinations from 6 reusable game types
0
0
Hangul game–level combinations from 6 reusable game types
0
0
Hangul game–level combinations from 6 reusable game types
0s
0s
Hard cap on every game session, set by observed attention span
0s
0s
Hard cap on every game session, set by observed attention span
0s
0s
Hard cap on every game session, set by observed attention span
THE PROBLEM
THE PROBLEM
THE PROBLEM
Every Korean app taught words.
None taught the machine that makes them.
Every Korean app taught words.
None taught the machine that makes them.
Hangul isn't an alphabet in the Latin sense. It's a combinatorial system. Forty letters slot into a fixed frame — initial consonant, vowel, optional final consonant — and every syllable in the language falls out of those rules.
Learn 40 pieces and the rule that joins them, and you can pronounce a word you've never encountered. That's not a feature. That's the whole product.
Hangul isn't an alphabet in the Latin sense. It's a combinatorial system. Forty letters slot into a fixed frame — initial consonant, vowel, optional final consonant — and every syllable in the language falls out of those rules.
Learn 40 pieces and the rule that joins them, and you can pronounce a word you've never encountered. That's not a feature. That's the whole product.
C1 초성
C1 초성
C1 초성
ㅎ
ㅎ
+
+
C2 종성
C2 종성
C2 종성
ㅏ
ㅏ
+
+
V1 중성
V1 중성
V1 중성
ㄴ
ㄴ
=
=
한
한
The rule, in one image. Three slots, one block, infinite words.
The rule, in one image. Three slots, one block, infinite words.
Research & Iteration
Research & Iteration
Research & Iteration
Classroom observation plus play-testing with 25 students, ages 6–12, in live Korean-school sessions. Not a lab. The actual room, with the actual noise.
Classroom observation plus play-testing with 25 students, ages 6–12, in live Korean-school sessions. Not a lab. The actual room, with the actual noise.
Observed
Attention collapsed at the two-minute mark
Attention collapsed at the two-minute mark
→
Shipped
120 seconds became a hard system constraint
120 seconds became a hard system constraint
→
Observed
Sound was the engagement engine, not decoration
Sound was the engagement engine, not decoration
→
Shipped
Audio on every screen, at a specified voice
Audio on every screen, at a specified voice
→
Observed
Kids replayed for the reward, not the lesson
Kids replayed for the reward, not the lesson
→
Shipped
Completion stickers and a top-10 leaderboard
Completion stickers and a top-10 leaderboard
→
System architecture
System architecture
System architecture
Classroom observation plus play-testing with 25 students, ages 6–12, in live Korean-school sessions. Not a lab. The actual room, with the actual noise.
Classroom observation plus play-testing with 25 students, ages 6–12, in live Korean-school sessions. Not a lab. The actual room, with the actual noise.
Track 01 · Violet Blue
Hangul — a ladder
Exactly 8 levels, strictly sequential: basic vowels → tricky vowels → Y-diphthongs → W-diphthongs → plain, aspirated and tense consonants → final consonants. Fixed and finite. Navigation shows position and progress.
Exactly 8 levels, strictly sequential: basic vowels → tricky vowels → Y-diphthongs → W-diphthongs → plain, aspirated and tense consonants → final consonants. Fixed and finite. Navigation shows position and progress.
Track 02 · Green
Words — a library
Open-ended categories: animals, body, clothing, colour, fruit, numbers, transport, sports, places, feelings, occupations, illness, cooking, and motion verbs. No order, no ceiling. Navigation is browse, not climb.
Open-ended categories: animals, body, clothing, colour, fruit, numbers, transport, sports, places, feelings, occupations, illness, cooking, and motion verbs. No order, no ceiling. Navigation is browse, not climb.
Track 03 · Blue + orange
Grammar — a library, harder
Also open-ended, but each entry is a pattern rather than a noun — sentence endings, connectors, particles. Needed the same browsing model with room for explanation.
Also open-ended, but each entry is a pattern rather than a noun — sentence endings, connectors, particles. Needed the same browsing model with room for explanation.
The scoring economy
The scoring economy
The scoring economy
If guessing pays, the score is a lie.
The rule: the cost of a wrong answer scales with how easy that answer was to guess. A twelve-card matching game gives a child a ~8% chance of a lucky hit, so a miss costs 7 points — cheap enough to keep trying. A yes/no grammar question gives them a 50% chance, so a miss costs 17 — expensive enough that coin-flipping loses. Reward stays roughly flat; risk does the work.
The rule: the cost of a wrong answer scales with how easy that answer was to guess. A twelve-card matching game gives a child a ~8% chance of a lucky hit, so a miss costs 7 points — cheap enough to keep trying. A yes/no grammar question gives them a 50% chance, so a miss costs 17 — expensive enough that coin-flipping loses. Reward stays roughly flat; risk does the work.
Game
Game
Guess Odd
Guess Odd
Guess Odd
Guess Odd
Cost of a wrong answer
Cost of a wrong answer
Panelty
Panelty
Hangul · Match Attack
Hangul
· Match Attack
4 × 3 grid · +50 correct
4 × 3 grid · +50 correct
~8%
~8%
-7
-7
Grammar · Match Attack
Grammar
· Match Attack
4 × 4 grid · +60 correct
4 × 4 grid · +60 correct
~6%
~6%
-9
-9
Words · Match Attack
Words
· Match Attack
4 × 4 grid, image + audio · +50 correct
4 × 4 grid, image
+ audio · +50 correct
~6%
~6%
-12
-12
Grammar · Grammar Master
Grammar
· Grammar Master
4 options · +50 correct
4 options · +50 correct
25%
25%
-12
-12
Grammar · Yes or No
Grammar
· Yes or No
2 options · +50 correct
2 options · +50 correct
50%
50%
-17
-17
Why not just punish every error equally?
Because a wrong answer in a 16-card matching game usually means a child is thinking, and a wrong answer in a two-option question often means they aren't. A flat penalty treats those identically and teaches kids that effort and coin-flipping cost the same. It also makes the leaderboard meaningless across games.
Because a wrong answer in a 16-card matching game usually means a child is thinking, and a wrong answer in a two-option question often means they aren't. A flat penalty treats those identically and teaches kids that effort and coin-flipping cost the same. It also makes the leaderboard meaningless across games.
Main Game Page
Main Game Page
Game Intro Page
Game Intro Page
Block Master Game
Block Master Game
Outcome
Outcome
Outcome
Live, in classrooms, and instrumented.
Analytics were specified and wired before launch, not bolted on after — completion, replay, drop-off point and error selection per game–level. That last one is the one I care about: because every distractor has a known meaning, wrong answers come back as diagnosis rather than noise.
Analytics were specified and wired before launch, not bolted on after — completion, replay, drop-off point and error selection per game–level. That last one is the one I care about: because every distractor has a known meaning, wrong answers come back as diagnosis rather than noise.
0%
0%
Game completion rate, up from the initial prototype
0%
0%
Game completion rate, up from the initial prototype
0%
0%
Game completion rate, up from the initial prototype
0%
0%
Positive feedback from teachers and parents
0%
0%
Positive feedback from teachers and parents
0%
0%
Positive feedback from teachers and parents
0
0
Game types across three content tracks
0
0
Game types across three content tracks
0
0
Game types across three content tracks
0
0
Korean schools in the distribution network
0
0
Korean schools in the distribution network
0
0
Korean schools in the distribution network
"Finally, a tool that teaches how to read, not just vocabulary."
"Finally, a tool that teaches how to read, not just vocabulary."
Teachers
"My child can now sound out any Korean word they see"
"My child can now sound out any Korean word they see"
Parents
"Now it's fun to do Korean school homework."
"Now it's fun to do Korean school homework."
Students
Where I'd take it next
Where I'd take it next
Where I'd take it next
Next 01
Adaptive difficulty from error patterns
Adaptive difficulty from error patterns
The distractor taxonomy already produces a per-learner error signature. Routing that back into level selection is a data problem now, not a design one.
The distractor taxonomy already produces a per-learner error signature. Routing that back into level selection is a data problem now, not a design one.
The distractor taxonomy already produces a per-learner error signature. Routing that back into level selection is a data problem now, not a design one.
Next 02
A teacher dashboard
A teacher dashboard
Class-level error patterns are the most valuable thing the product generates and currently nobody sees them. This is where I'd spend the next quarter.
Class-level error patterns are the most valuable thing the product generates and currently nobody sees them. This is where I'd spend the next quarter.
Class-level error patterns are the most valuable thing the product generates, and currently nobody sees them. This is where I'd spend the next quarter.
Next 03
Port the model to another script
Port the model to another script
The architecture is script-agnostic — any writing system with combinatorial rules could sit on it. That's a platform argument, not a feature one.
The architecture is script-agnostic — any writing system with combinatorial rules could sit on it. That's a platform argument, not a feature one.
The architecture is script-agnostic — any writing system with combinatorial rules could sit on it. That's a platform argument, not a feature one.
Next case study