The core problem: your brain drops thoughts within ~30 seconds. Notes apps don't solve this because they require you to organize as you capture. ThoughtTrack separates capture from organization entirely.
How it works:
- Tap one button to voice-capture a thought. It transcribes on-device using Apple Speech.
-The "Nebula" view shows all your thoughts as a drifting particle system — no folders, no hierarchy. Connected ideas glow brighter and drift closer.
- A composite scoring algorithm finds "Thought Bridges" between entries using NLEmbedding (semantic similarity, 0.55 weight), NLTagger for named entity overlap (0.25), and tag Jaccard similarity (0.20). Threshold of 0.40 to form a connection.
- A context-aware resurfacing engine uses activity state, focus mode, time, location, and audio routing to decide when to bring a thought back — not just which one. It adapts its weights based on whether you engage with or dismiss notifications.
Stack: Pure SwiftUI + SwiftData + Observation framework. All NLP runs on-device via Apple's NaturalLanguage framework — nothing leaves the phone. Also has a watchOS companion for wrist capture and a widget for the daily "Spark."
The architecture challenge I'm most proud of solving: decomposing a monolithic AppState into separate ThoughtRepository, NavigationCoordinator, and NotificationCoordinator while keeping SwiftData's CloudKit sync working with conflict resolution.
Free 7-day trial, then 10 thoughts/month on the free tier. Would love feedback on the resurfacing algorithm — I'm still tuning the context weights and curious if anyone has worked on similar "right moment" notification systems.