What if anatomy isn't built first? What if an open medium acquires its own fine print and large print from the waves passing through it? Write three frequencies into a conserving sediment model. Then ask what a weak wave can read.
NEW · 20-EPISODE LEARNING EXPERIMENT — The beach's stress memory can become stale as the medium changes. Compare seven real computed controls in a separate 48-node wave-operator laboratory: frozen memory, tracking memory, teacher-free recalibration, plain field energy, and gradient/oracle controls.
Open Memory Drift lab ↗Open 8-seed trial lab ↗ This is a distinct synthetic wave experiment, not the main 288-cell sand model or a claim about actual sediment learning.
01 / The living shore · numerical wave laboratory● READY · OFFLINE CAPABLE
Write at three scalesStrong wave · bed allowed to move
Resonant mode drive excites a distributed standing-wave shape deliberately. Edge drive injects waves only at the left boundary; it is less selective and creates extra near-source structure. Try both. The waves evolve on the *current* bed; erosion and deposition act through a conservative local flux. A weak read freezes the bed.
Order is an experimentSame waves · different history
Do the same two exposures commute? A then B can change where B propagates. The difference only persists if the bed feeds back into the wave operator.
No order comparison yet. A = long, B = medium; X = common short final exposure. Both runs start with identical sand and identical wave strengths.
The shore starts flat. Choose a write pulse.
The written spatial spectrumFourier transform of the actual bed
Long: k=6Mid: k=12Short: k=24
Peaks near 6, 12 and 24 cycles/domain correspond to half-wavelength bed gratings from standing-wave stress. Other peaks and collisions are real outcomes of this toy rule; patterns are not drawn into the sand by hand.
Weak-wave address / readbackMeasured propagation, not direct sand lookup
—LONG · Δ output RMS
—MED · Δ output RMS
—SHORT · Δ output RMS
The probe drives the left boundary with a weak sinusoid through the same finite-difference wave solver, with writing OFF. Numbers show output-RMS difference from a flat reference at each frequency—not classification accuracy or a claim that the sand spectrum equals the echo spectrum.
From September's cavities to an open, writing medium.
01 / ASK BY FREQUENCY
A wavelength is an address
ObjektiYksi used frequency and ports to interrogate one conserved material. Here the wave's wavenumber selects a spatial standing-wave pattern; a later wave interrogates what remains.
02 / LET STRUCTURE EMERGE
No drawn-in compartments
The substrate starts flat. A driven mode or an incident edge source evolves with the wave equation; a slow thresholded flux moves signed height locally. Different scales share one array, with no interior walls.
03 / TRY TO FALSIFY IT
Control the feedback
Switch off bed-to-wave feedback or reverse event order. Note: a bounded sediment writer can itself be history dependent even when wave feedback is disabled. The comparison also includes a fixed-wave, unbounded control.
Research boundary. This is an original interactive 1D numerical instrument, not a faithful replication of G1/G2 held-out benchmarks or the 3112-cell floor in ComputingBeachClaude. The energy-gradient sediment mobility is a toy constitutive law, not calibrated Shields/Exner transport. Material is transported conservatively (up to floating-point roundoff) but threshold, wave amplitudes and display magnification are chosen for exploration. The mode-shaped drive deliberately selects a resonant spatial eigenmode; it does not demonstrate wavelength segregation by an unconstrained incident wave. The incident-edge option is a more difficult control. Writing stress is spatially averaged over a frequency-dependent radius (9, 4, 0 cells), which is an explicit constitutive assumption, not an emergent consequence of water dynamics. The short/long wave imprint relationship is a hypothesis to test across rules, geometries and resolution—not evidence for neural frequency organization.