RBF-Attention – Trading dot-products for Euclidean distance
Replaces dot-product attention with Euclidean distance to stop vector magnitude bullying.

Fixes treadmill interval tracking where Strava and Nike Run Club don't.
Treadmill runners seeking structured interval workouts
Strava · Nike Run Club · Peloton
I kept losing count of reps during treadmill intervals. Tried paper notes, tried the big running apps – none of them are really built for treadmill use.
Kuta guides you through workouts with large visual cues (designed for sweaty hands on a bouncing treadmill), calculates distance from pace × time instead of the treadmill sensor, and syncs to Strava via Apple Health.
SwiftUI, iOS 17+. 22 free preset workouts. Custom workouts are a one-time $2.90 unlock. No subscription.
This is my first app. Got rejected by Apple twice before approval. Happy to talk about the build or the process.
Replaces dot-product attention with Euclidean distance to stop vector magnitude bullying.
Clean, polished breathing app—but exists in a crowded wellness tooling space.
Variable singing bowl recordings prevent meditation timer monotony.
Front-camera distance calibration beats guessing with a tape measure.
This is a practical, no-nonsense play: someone trained YOLOX from scratch, released MIT-licensed weights, and packaged a path toward running it on iOS. The value is procedural — dataset curation, training recipe, and an export/convert-for-iOS pipeline — but it's not a conceptual breakthrough; I'd like to see clear mAP numbers, model size and on-device latency benchmarks before recommending it for production.
Zero-ad NYC subway tracker loads faster than Citymapper at 2.5MB.