iOS

Low-latency on-device captions in Swift with SpeechAnalyzer

Transcribing a 50-minute audiobook chapter before showing the first caption is unusable. How I get the first line in ~1–2s: a playhead-windowed AsyncStream feeding iOS 26's SpeechAnalyzer, with the feeder and results consumer racing on separate tasks — all on-device.

Vendoring FSRS spaced repetition as a pure Swift function

My audiobook app turns highlights into flashcards scheduled by FSRS-5 — the algorithm Anki uses. Why I vendored the math as a pure, dependency-free, unit-testable Swift function instead of adding a package, and how the scheduler actually works.