Skip to content

The DAW

systemBlue edited this page Jun 11, 2026 · 1 revision

The DAW

DeadCat is building a DAW for iPhone, iPad, and Mac. A digital audio workstation written in Swift on Apple's frameworks, with an AI assistant built in.

The app is built one milestone at a time. Each milestone ships something usable on real devices; the next milestone does not begin until the previous one passes.

Who it serves

Two audiences: musicians and producers recording takes and mixing tracks, and audio journalists and podcast producers cutting interviews and narrated speech, where the output has a loudness specification. The milestones are chosen so each release serves both.

Principles

  • Milestones. Small increments, each one usable, each one tested on real devices before the next begins.
  • The simplest thing that works. No speculative architecture. The engine grows features when a milestone needs them.
  • Native everything. SwiftUI for the interface, AVAudioEngine and Core Audio for sound, SwiftData for the document, Foundation Models and Core AI for intelligence. No third-party dependencies.
  • The assistant. The assistant proposes, explains, and operates; creative and editorial decisions stay with you.
  • Data storage. Audio and sessions stay on the device. On-device models are the default for intelligence; anything that would leave the device is a deliberate, visible opt-in.

The path

Milestone What you get Status
Scaffold The app exists on iPhone, iPad, and Mac with the design system's foundations. In progress
Sketchpad Record one take, play it back, name it, and keep it. Planned
Tracks A few parallel tracks with volume, pan, mute, and solo. Planned
Arrange Move, trim, and loop regions on a timeline. Planned
Transcribe Spoken takes get on-device transcripts, so an interview can be searched and navigated by what was said. Planned
Assist The on-device assistant arrives: name takes, summarize a session or an interview, pull a quote, operate the transport and mixer in plain language. Planned
Deliver Export with loudness measured against podcast and broadcast targets, so the exported file meets the spec. Planned
Beyond Effects, automation, and more. Later

Milestones are tracked as issues; the changelog records what ships.

How it gets tested

Every increment runs on real devices continuously, not at the end. Automated tests cover the engine and the document model; manual testing uses real recording and editing sessions. A milestone that cannot survive that is not done.

What it is not

  • Not porting an existing engine, and not wrapping one. The engine is built here, on Apple's audio frameworks.
  • Not shipping a plugin host, surround, or video scoring in the first year.
  • Not building for other platforms. This is an Apple-native app.

Clone this wiki locally