- Pronunciation is fully system-based (AVSpeech). - There’s a cross-folder review mode («Words you often forget») that aggregates weak cards across the entire library based on actual performance history, not just per-deck repetition. - Folders are portable files. You can export acardsmy file and send it via AirDrop, and it just imports on the other side. No account needed. - CloudKit handles full sync, including images. - AI examples of word usage. AI generation of cards and folders.
No reviews yetBe the first to leave a review for Cards.my
Maker
📌
I wanted to share something I’ve been building, Cards.my. It's a vocabulary flashcard app for iOS - I started it in early 2022 and have been chipping at it on weekends since.
At its core, it’s still just flashcards:
* folders (topic / level / whatever structure you want)
* word pairs (your language ↔ foreign language, optional images)
* two study modes:
* Learn: self-paced, reveal/hide, free navigation, shuffle anytime
* Test: shuffled, typed answers, case-insensitive + trimmed input, instant feedback, no backtracking, final score + retry if performance is bad
Nothing novel there - that part is intentionally boring.
What ended up making it interesting for me were the small «product correctness» details and constraints that stack up over time:
- Pronunciation is fully system-based (AVSpeech). The speaker only appears when iOS actually supports a voice for that language, and the app auto-detects the dominant language per folder so you don’t end up with broken audio buttons everywhere.
- There’s a cross-folder review mode («Words you often forget») that aggregates weak cards across the entire library based on actual performance history, not just per-deck repetition.
- Folders are portable files. You can export a .cardsmy file and send it via AirDrop / Messages / email, and it just imports on the other side. No account needed.
- CloudKit handles full sync, including images.
Stats are intentionally not «streak-based gamification». Instead:
* 7-day practice strip
* mastered / weak word lists
* revisit queue based on real errors
* reminders that adapt to activity instead of nagging for streak preservation
There’s also AI-assisted example generation per card:
* tap “Examples” → generates a few usage sentences for the target word
* strict guardrails (rate limits, quotas, refusal rules)
* explicit “AI may be wrong” labeling everywhere it appears
On the engineering side, a lot of effort went into boring reliability work:
* debounce search everywhere
* undo window for folder deletion (5 days)
* stable sort persistence
* exam hint system
* deep links
And then the AI layer, which turned out to be more prompt engineering + validation than UI work:
* strict JSON schema enforcement
* deduping via normalization (case, whitespace, diacritics)
* rejection of invalid scopes («10k words», multi-folder requests, etc.)
* language-specific grammar rules where needed
Tech stack:
* 100% native SwiftUI
* CloudKit sync
* OpenAI for example sentences (with strict quotas + rate limiting to avoid cost blowups)
Free tier: 4 folders, 50 cards each. Subscription unlocks everything with a 7-day trial. No ads, no tracking beyond crash + product analytics.
If anyone tries it, I’d genuinely appreciate feedback - especially on:
* AI sentence quality edge cases
* where the study flow feels unintuitive or too «manual»
Happy to share a promo code with anyone who wants to try it and give feedback.