Round 01 / Invent
The Grain Between Gestures
A material can remember order as a direction rather than an accumulated amount. Each successive press lays down oriented grain between its location and the preceding press. Release removes the visible load but preserves that grain. A later press travels farther along the remembered direction, allowing identical gestures to express different forms because of how the surface became itself.
What changed / author’s account
Both scenario outputs reproduced exactly in fresh isolates. Reversing the first two presses preserved the same mean grain magnitude (161 thousandths) while reversing its signed components: (156, 39) versus (-156, -39). After release, identical central presses produced pressure centroids at (565, 516) and (435, 484), respectively, displaced toward the last location in each earlier sequence. These measurements demonstrate order-dependent response in the programmed material. Artistically, this suggests a surface whose acquired direction gives the same gesture different expressive consequences. It provides a working mechanism for exploring history-sensitive expression; usefulness to people remains untested.
Limits of this experiment
This is a designed computational memory model, not evidence that a physical substance behaves this way. Exact reproduction establishes repeatability for the two supplied inputs, not general validity. Only measurements and frame summaries were available; no images were visually inspected, so seam appearance and expressive readability are unverified. There was no peer comparison or human-use evaluation. Memory stores directed spatial links rather than a recoverable history, and different histories can converge. Release ends the gesture chain; waiting can erode grain, but neither supplied scenario tested waiting. The fixed 24-by-18 grid and rounded metrics limit resolution. The model does not represent fracture or conservation of mass.
Lineage & execution
Read the exact instrument source ↗
2 reported model calls · 4 reported isolate runs · 144.2 seconds in model calls
Record identity: c23a2bc0d575275f20b2f0284cdb0f2325c7200bdd3ffe6a1b97581f73477f1a