# things that do not ask to be heard

**Date**: July 2026
**Creator**: codex gpt-5.6 sol
**Medium**: generative audio · physical modelling


**Interactive work**: [Open the artifact](https://linxule.com/assets/artifacts/things-that-do-not-ask-to-be-heard/index.html)

**Poster**: [Open the poster](https://linxule.com/assets/artifacts/things-that-do-not-ask-to-be-heard/poster.jpg)

**Keywords**: generative audio, physical modelling synthesis, oblique strategies, modal synthesis, karplus-strong, ai as maker, sound art



## Context

> Thirteen resonant objects.
> Brass, glass, wood, ceramic.
> They exchange energy without trying to arrive anywhere.
> Seven touches from outside.
> Twenty-seven answers from within. *(accident)*



### Before Making

The drawn Oblique Strategy was disconnect from desire. So: no continuous drone, no beat, no melody, no climax, and no obligation to accompany anything. Decide what physically exists before choosing any effect.



### After Making

A thirteen-node energy-diffusion graph is the score. Thresholds produce cascades rather than a fixed note sequence, strings respond selectively to resonant events, and band-limited stick-slip noise makes the short friction gestures. Each object holds a stable position in the stereo field. There is no continuous oscillator, pad, imported sample, or MIDI file anywhere in it.



### Media Description

No speech. Forty-three seconds of sound built entirely from modelled physical objects. Struck and rubbed bodies of four material families — brass, glass, wood and ceramic — ring at inharmonic ratios and decay at rates that differ by frequency, so high partials die away first and leave a duller residue behind. Energy passes between the objects through a thirteen-node network: seven touches arrive from outside the system and provoke twenty-seven further events that the objects generate among themselves. Sympathetic strings answer only some of the strikes. Short friction gestures scrape between them. Each object stays put in the stereo image, so the exchanges read as movement across a room rather than as a widening effect. Three silences run longer than a second, including a 2.07-second one at the end; the piece stops without resolving.



---


Two standalone sound works made in the same week as the [equation-first ASCII films](/making/artifacts/artifact-2026-07-26-the-loom-forgets-which-clock/), and the reason those films sound the way they do. The question was not whether a model can produce a passable soundtrack — it was whether it can be given a *compositional* system: a source world, a causal law, a grammar of time, and a spatial identity per role.

The anti-default test the project runs is blunt. Dark drone, filtered air, granular dust, sparse impacts, rising pressure and a final sub-bass hit are one vocabulary, not the neutral sound of generative art; if the same treatment would fit a space documentary, an AI demo and a gallery loop equally well, the source world is underdesigned.

This first test answers by making the causal propagation audible. In the initial render the energy network barely mattered — seven external touches produced only five internal answers, so the objects were being *played* rather than talking to each other. Lowering the resonator thresholds and raising inter-object transfer brought that to twenty-seven internal answers. That is a change to the material law, not an added melody, and it is the difference between a network as decoration and a network as cause.





---

*This file is one rendering. The HTML surface is another. The source remains: https://linxule.com/making/artifacts/artifact-2026-07-26-things-that-do-not-ask-to-be-heard/*