The Problem with Learning Apps
Open any learning app and you’ll meet an anxious owl begging you not to break your streak. Points, badges, leaderboards—gamification works against learning.
Extrinsic rewards undermine intrinsic motivation. The reward becomes the reason. When the reward disappears, so does the motivation. You’re not learning—you’re managing a psychological debt.
A Different Philosophy
HyggeCards takes a different approach, rooted in a simple observation: the friction should be in the learning, not the interface.
- No streaks. Miss a day? Nothing happens.
- No points. Progress is what you know, not what you’ve clicked.
- No leaderboards. Learning isn’t a competition.
- No mascots. “12 cards reviewed” instead of “Great job!”
I wanted it to feel like a nice notebook, not an app trying to be your friend. Cards develop subtle “wear” over time (visual indicators of how often they’ve been reviewed), like pages you keep returning to.
How It Actually Works
You upload your notes in Markdown. AI reads them and generates flashcards automatically—questions designed to test understanding, not just recall. The spaced repetition algorithm then schedules reviews to maximize long-term retention.
The scheduling prioritizes cards intelligently:
- Cards you struggled with recently come back within hours
- Cards semantically related to your struggles appear next (if you’re confused about recursion, you’ll also see related concepts)
- New cards are introduced gradually
- Mastered cards surface occasionally for maintenance
When you review a card, you rate your confidence: “Don’t know,” “Getting there,” or “Got it.” This confidence-based approach is more effective than binary right/wrong because learning is about strengthening memory, not performance.
The Tech (Briefly)
Under the hood, HyggeCards is a serverless application on AWS:
- AI Generation: Amazon Bedrock generates flashcards from your notes, using Titan for both text generation and semantic embeddings
- Vector Search: Cards similar to ones you’re struggling with are identified using cosine similarity on 1024-dimensional embeddings
- Spaced Repetition: A custom algorithm that considers recency, difficulty, semantic relationships, and review history
- Multi-Platform: React web app with native Android (Kotlin/Compose) and iOS (Swift/SwiftUI) apps
- Zero Servers: Lambda, DynamoDB, API Gateway—fully managed, scales to zero when not in use
Learning as a Place You Want to Be
I don’t care about daily active users. I care about whether you actually remember what you studied three weeks later.
The science behind spaced repetition is solid. A 2025 study of over 26,000 physicians found that spaced repetition produced 58% retention compared to 43% for massed practice. The effect holds across domains: medical education, language learning, professional training.
Most apps get the algorithm right and everything else wrong. HyggeCards optimizes for retrieval practice, not streaks.