Agentic software engineering builds digital music instruments 3x faster
AI agents re-create iconic musical software as native plugins using C++ and JUCE.
Matthew John Yee-King’s new paper on arXiv demonstrates how agentic software engineering (ASE) can dramatically accelerate the creation of digital music instruments. Using the C++ language and the JUCE framework, the author conducted three distinct case studies: re-implementing Laurie Spiegel’s iconic 'Music Mouse' as a native plugin, translating François Pachet’s 'Continuator' system from Python into a native plugin, and developing a novel 3D user interface for a traditional tracker sequencer using OpenGL. The agentic software agents handled boilerplate code, library integration, and cross-language translation, allowing the human developer to focus on creative decisions and debugging.
Beyond speed, the paper highlights how ASE can solve long-standing challenges in the music software community: longevity (native plugins outlive scripting environments), interoperability (JUCE plugins run on any modern DAW), and accessibility (non-programmers can potentially direct agents to build instruments). The autoethnographic analysis of prompt logs and software snapshots reveals best practices for prompt engineering in this domain, such as breaking tasks into atomic steps and providing reference implementations. The work suggests that with further refinement, ASE could empower musicians without coding skills to design their own custom digital instruments.
- Three case studies using ASE with C++/JUCE: Music Mouse re-implementation, Continuator Python-to-native translation, and 3D OpenGL tracker sequencer UI.
- ASE reduced development time by automating boilerplate, library linking, and cross-language translation.
- Autoethnographic analysis of prompt logs identifies best practices for agentic software engineering in audio creation.
Why It Matters
ASE could democratize digital instrument design, letting musicians without programming skills build native plugins efficiently.