Maru on building Stage+, Universal Music's classical streaming platform, and running a multi-agent workflow across Opencode, Codex, and Claude Code.
"My workflow starts with an AI agent to define scope and a plan. Then coding agents handle implementation, validation, and first-pass review."
I’m Maru, a Senior Backend Engineer at Universal Music. I work on Stage+, a classical music streaming platform.
Editor: Zed.
Coding agents (work): Opencode, Pi, Codex.
Coding agent (personal projects): Claude Code.
I use AI for brainstorming, planning, code generation, code reviews, bug fixes, investigations, and occasionally as a study companion.
My workflow typically begins by discussing a task with an AI agent to define the scope and create an implementation plan. Once the plan is finalized, I delegate it to coding agents for implementation, validation, and an initial code review.
27-inch external monitor, Glove80 split keyboard, trackpad.
I’d like a custom orchestration harness that manages my entire agent workflow: spawning agents, provisioning workspaces, gathering context, executing tasks, and seamlessly passing artifacts between workflow stages.
The goal is to automate the coordination work so I can focus on higher-value thinking and decision-making.
