Most modern music is written and produced with “digital audio workstations” or DAWs, which are GUI applications for recording, mixing and producing audio. Like many other graphical applications, DAWs suffer from several problems, including hidden state, complex menus and low discoverability, limited extensibility, lack of composability, and difficulty in versioning and tracking changes.

A less widespread tradition of music making on the computer comes from live coding, in which artists write and edit programs live, with the purpose of making music, usually generative in nature. However, live coding, like the rest of programming, suffers from its non-visual nature, which is arguably critical in this context, and accounts at least in part for the limited success of live coding in the music world.

Here I present an attempt at bridging this gap. This prototype, named “PAW” for “Programmable dAW”, is an exploration of applying concepts from bidirectional and live programming to the context of music making. In PAW, the author writes a program which produces the piece they’re working on, and a live GUI is generated from code analysis of that program. Using this GUI, the author can make immediate changes to both the audio output and the program that generated it.