← Back to Work
Solo Full-Stack Developer & Designer · 8-week end-to-end development

8 user interviews · 12 TestFlight feedback points · 2 major usability improvements

Smart Study App · Mobile Learning · Self-funded · Bootstrapped

MemQ: Turn AI Conversations into Lasting Knowledge

People ask AI the same questions over and over — not because they're lazy, but because nothing helps them actually retain the answer. MemQ closes that loop: ask, get knowledge cards, learn through spaced repetition, own the knowledge forever.

01Ask AI
02Generate Cards
03Spaced Repetition
04Master It
Live on the App Store
The Problem

People ask AI the same question five times — not from laziness, but because there's no bridge between “I got an answer” and “I actually know this.” The knowledge stays in the chat window, never in their head.

AI as a Crutch

Users ask AI the same questions repeatedly because the answer never gets internalized — it stays in the chat window, not in their head.

Capture Friction

Manually creating study cards from AI answers is too tedious — users never bridge the gap between getting an answer and learning it.

Knowledge Fades

Even when users save notes, there's no system to resurface them at the right time — knowledge decays without spaced repetition.

Design Process

From Observation to Shipped Product

8 weeks, one person, two major rounds of testing. Here's how the product evolved from a single observation into a shipped app.

01
Research
User Interviews

8 interviews across students, professionals, and language learners. Key observation: they keep asking AI the same questions.

2 weeks
02
Synthesis
Core Insight

The problem isn't AI quality — it's that nothing bridges "got an answer" and "actually know this." People need a retention layer.

3 days
03
Build
MVP Development

Designed and shipped the core loop: AI card generation + spaced repetition engine. Cut everything else to validate the fundamentals.

3 weeks
04
Test
TestFlight Round 1

6 sessions. Found: card creation was the drop-off point. AI generation redesigned as the primary path. Manual entry became secondary.

1 week
05
Iterate
Platform Expansion

Explore, smarter lesson creation, and Assistant upgrades — Quiz me with lesson pickers plus one-tap "new lesson" from generated terms. Feature questions replaced basic how-tos.

2 weeks
What We Heard

I ask ChatGPT the same questions about Python syntax at least once a week. I just can't seem to make it stick.

Working professional, software adjacent

I take notes when I'm studying but I never go back to them. I just search for it again next time I need it.

University student, exam prep

I use Anki but making the cards takes longer than actually learning. So I just don't make them.

Language learner, Japanese
The Solution

The Learning Loop

Every part of MemQ serves one workflow. Only the knowledge you personally asked about ever enters it.

01

Ask AI

Type a question, topic, or concept you want to learn. No templates — just ask naturally.

02

Generate Cards

AI creates structured knowledge cards — vocabulary gets definitions, concepts get "Why" and "How" questions.

03

Spaced Repetition

The memory curve resurfaces cards at optimal intervals. Review sessions take minutes, not hours.

04

Master It

Knowledge moves from short-term recall to long-term retention. You stop asking AI — because you already know.

User Journey · From Dependency to Ownership
🤔
Frustrated
Discovery

Realizes AI answers never stick

"I've asked this five times already."

😮
Surprised
First Card

AI generates cards from their question

"That took 3 seconds?"

😌
Engaged
First Review

Spaced repetition surfaces a card at the right moment

"Oh — I actually remember this."

💪
Committed
Building Habit

Daily review becomes automatic, not effortful

"2 minutes a day is enough."

🎯
Confident
Mastery

Knowledge is internalized, AI dependency broken

"I stopped needing to ask."

Phase 01 · MVP

Ship the Core Loop

The MVP was built to validate one thing: can we make the jump from “I asked AI” to “I actually learned this” feel effortless? Instead of making users build flashcard decks manually, MemQ lets you type a question or topic, and the AI generates structured knowledge cards ready to study.

The spaced repetition engine does the rest — surfacing cards at the right intervals so knowledge moves from short-term recall to long-term retention. Every card in your library is something you wanted to learn.

AI Card GenerationType a topic or question, get structured knowledge cards instantly.

Context-aware FormattingVocabulary gets definitions and examples; concepts get "Why" and "How" questions.

Spaced Repetition EngineThe memory curve resurfaces cards at optimal intervals for long-term retention.

Phase 02 · Iteration

Explore Platform, Assistant & Smarter Creation

After validating the core loop, Phase 02 layered on a curated Explore experience, a redesigned lesson-creation path, and a stronger AI Assistant — including a new Quiz me mode where learners pick any lesson to practice, plus a direct Create new lesson action when assistant-generated terms deserve their own deck.

Redesigned Lesson Creation

Before
Manual terms, topic→AI, or PDF/Doc upload — but no clean way to fix just one term
Lesson shell first
Structure before content felt right
Manual terms
Create & edit by hand
Topic → AI
Generate from a topic
Upload PDF / Doc
AI analyzes & drafts cards
One off card?
Rework by hand or redo the batch
All three creation paths existed, yet naming and scaffolding still ran ahead of the material — and a single weak term meant painful rework.
After
Same three ways in — plus regenerate AI content for a single term, and name the lesson when it's ready
Manual terms
Create & edit by hand
Topic → AI
Generate from a topic
Upload PDF / Doc
AI analyzes & drafts cards
Regenerate term
One card, new AI pass
Name lesson
Title after content lands
Matches how people think: pick the material first, label the lesson when you're done — not the other way around.

AI Assistant & Quiz me

The assistant stayed grounded in personal decks, but got clearer structure and faster study loops — so asking, saving, and drilling read as one product instead of disconnected experiments.

Assistant IA, prompts, and save flows tuned for quicker daily Q&A
New Quiz me tab: AI-driven practice on real lesson cards
Lesson picker inside Quiz me — choose which deck to drill against
Generated terms: add Create new lesson to spin strong suggestions into their own lesson
MemQ AI Assistant screenshot 1
MemQ AI Assistant screenshot 2

Explore Platform

A curated library of high-quality lessons users can add to their library with one tap. Power users can publish their own lessons for the community — turning MemQ into a shared knowledge network, not just a personal study tool.

Curated lessons across popular learning topics
One-tap add to personal library
Publish your own lessons for others to learn from
MemQ Explore platform screenshot 1
MemQ Explore platform screenshot 2
MemQ Explore platform screenshot 3
Visual Design · Iteration

Editorial

A quiet, typography-first system for spaced-repetition learning. Borrowed from editorial print — generous hierarchy, hairline rules, restrained colour. The screen treated as a page. One accent. One surface tone. One radius scale. Decisions, not options.

Editorial Behaviour
i.
Type before chrome

Hierarchy by weight, scale, and white space — not boxes, shadows, or accent fills. A heading does the work a card would do elsewhere.

ii.
One semantic accent

Teal #1A8A72 reserved for state: due cards, active tabs, primary actions, mastery progress. It earns attention because it is rare.

iii.
Hairlines, not shadows

1px borders on warm white separate regions. Shadows not used. Surfaces are flat and meet at clean edges.

iv.
Numbers are display

Streak counts, queue totals, mastery percentages set in the heaviest grotesk weight at large sizes. Data is the headline.

v.
Cards in a stream

Distinct rounded cards with 8px gaps — never flush list-rows. Each card is its own object on the page canvas.

vi.
One elevated action

Create is the one moment shadow appears: a circular FAB raised above the tab bar. Every other surface is flat.

Colour Tokens
Accent#1A8A72
Due · Active · Primary · Mastery
Canvas
#F9F8F5
Surface
#FFFFFF
Muted
rgba(26,26,24,0.05)
Ink
#1A1A18
Ink Dim
45% opacity
Hairline
10% opacity
Type Scale
MemQ34px/800
Your Lessons22px/600
Due in 2 days17px/400
12 CARDS DUE12px/500
iv. Numbers are Display
127
day streak
85%
mastery
32
due today
14d
next review
Key Screens
In-app captures
MemQ home
Home
Today's queue
MemQ library
Library
Lessons, cards & review counts
MemQ lesson detail
Lesson detail
Mastery, terms & start review
MemQ multiple-choice study
Study
Multiple-choice question
MemQ answer feedback
Study
Right vs wrong, try again
MemQ explore
Explore
Featured & trending lessons
MemQ recall mode
Recall
Type an answer, see the model
MemQ profile
Profile
Streak, stats & focus queue
MemQ lesson from Explore
Lesson preview
From Explore — add to library
MemQ AI Assistant Q&A
Assistant
Q&A grounded in your decks
MemQ Assistant save card
Assistant
Save the answer as a card
MemQ Assistant pick lesson
Assistant
Pick or create a lesson
Impact

The Loop Held Up in Testing

Testers stopped asking “how do I add a card” and started asking “can I import more file formats” — the shift confirmed they'd moved from onboarding friction into active daily use.

78%
preferred AI generation over manual entry
85%
wanted content from their own materials, not pre-made decks
0
critical drop-offs in the final validation round

Preferred method of card creation

AI / Auto-generation78%
Manual Typing15%
Copy & Pasting7%

AI generation is the primary action — manual entry is a fallback, not the default.

Importance of personalized study content

Custom content (my own materials)85%
Generic decks (pre-made lists)15%

Every card in MemQ was generated for you, from what you personally wanted to learn.

“This is the most seamless study experience I've seen. I would feel 100% confident ditching my old messy notes for this.”

— Round 2 usability testing participant

What's Next

Future Directions

LLM & Browser Integration

A browser extension that intercepts AI chat sessions and lets users save any answer as a knowledge card in one tap — making the learn-from-AI workflow completely seamless.

Richer Multi-modal Recall

Add image support — starting with camera capture — to reinforce memory at both the term and question level, tying abstract concepts to concrete visual cues.

Vocabulary-first Card Depth

Optimize vocabulary cards as a dedicated surface: pronunciation (audio and IPA where it helps), richer example sentences, and explicit synonym/antonym links — so language decks feel as deep as dedicated apps without breaking the MemQ review loop.

Social Learning Loops

Build on the Explore platform with learning cohorts — where users can follow each other's progress, remix published lessons, and keep each other accountable.

AVAILABLE NOW

Try MemQ on iPhone

Turn any AI conversation into lasting knowledge. Free to download.

Download on the
App Store
iOS 17+iPhoneFree
← Back to Work

Let's work together

© 2025 Mei Chai. Designed with intention.