Notes.
Building Native Apple Solutions. Documenting the process of building Mac and iOS software with on-device AI and Apple Foundation Models. Short-form architecture snapshots from active product development.
Apple PCC summarizes 80K transcripts without keeping session data
When meeting transcripts outgrow on-device context, Private Cloud Compute runs encrypted summary work in isolated enclaves, then clears the session. CoreRecord keeps raw audio on the Mac.
Read note →AFM 3 on-device speech comprehension
Apple AFM 3 Core Advanced adds an active-listener multimodal layer for local dictation, sparse activation, and clean meeting notes without cloud audio.
Read note →Building an On-Device Ecosystem · Part 4 of 4
How to align preview visuals and landing copy with the same on-device story engineering built into the app.
Read note →Building an On-Device Ecosystem · Part 3 of 4
How to guarantee 100% uptime and true portability for professional creator tools.
Read note →Building an On-Device Ecosystem · Part 2 of 4
How to turn an on-device privacy boundary into a clear App Store positioning hook for professional creators.
Read note →Building an On-Device Ecosystem · Part 1 of 4
How to keep a teleprompter dictation loop in sync with the creator's pace during on-device capture.
Read note →Songen — MIDI as the foundation
Songen generates editable multi-track MIDI on the device so you can change sounds, edit the piano roll, and export to your DAW without rebuilding the sequence.
Read note →On-Device AI — Private, Offline, Instant
How Apple Foundation Models on the Neural Engine keep inference private, offline-capable, and instant on iOS.
Read note →On-Device Foundation Models
How Kaari, LocalMemo, LocalPlan, and SpeechBeat run Apple Foundation Models on the Neural Engine, offline and on your phone.
Read note →
How Jim Collins' Hedgehog Concept shaped RollSearch, one job, filename search, 100% on-device.
Read note →
