thear

You watch AI work all day. I wanted to hear it.

Working with AI coding agents is an almost entirely visual experience — you watch tokens appear. Over long stretches that gets dissociative: you lose the thread of what's actually happening. Sound solves this in a way another dashboard can't, because hearing doesn't demand your eyes.

thear gives agentic work a soundtrack. It's a small Python daemon that listens to Claude Code's hook events — tool calls, file writes, completions — and emits OSC messages in real time. Ableton Live picks them up through a Max for Live device and turns them into whatever sounds you've designed.

Events become gestures, sessions become music

• Tool calls

  Short percussive hits — the tick of an agent actually doing things.

• File writes

  Filter cutoff modulated by file size — big changes literally sound bigger.

• Completions

  A long reverb tail released — the audible exhale of a finished task.

• Anywhere OSC goes

  Ableton is the reference target, but the output is plain OSC — route it to any synth, visual engine, or hardware that listens.

The surprise is how musical it is. File writes cluster rhythmically; long reasoning steps produce slow ambient swells. A busy session has a groove, and a stuck one goes quiet — which turns out to be exactly the signal you want.

Peripheral awareness, restored

thear is released and open source. It sits in the same family as my other AI tooling — instruments for living with AI rather than just consuming it — and it's the one that best captures the point: the machine's work becomes ambient, legible, and a little bit beautiful.

I stopped staring at the terminal to know what my agents were doing. Now the studio just tells me — in rhythm, swells, and the occasional very satisfying reverb tail.