# Unexecuted next step: spatial construction and dynamics

Compare one direct spatial builder with one contextual growth rule using the same
affine/nonlinear and gated-accumulator primitives, public ports/coordinates,
parameter/state inventory, training labels and historical search budget. Freeze
both before testing; keep generated topology, geometry-dependent transmission
delays, intrinsic cell time constants and observed temporal responses as separate
variables. Coordinate metadata alone does not establish a 3D simulation.

One timing intervention: permute edge delays within matched delay/count bins while
keeping operators, parameters, connectivity and initial state fixed. Predict a
changed impulse-response latency/decay profile if those delays participate in the
computation; assess a nonperiodic last-write-and-route stream with irregular
inter-query gaps and overwrite bursts. An intervention-specific performance loss
would motivate a controlled timing hypothesis, not a general explanation of
intelligence by frequencies. Three dimensions remain optional and untested.

As a null check, identical final operators, parameters, topology, initial states
and schedules should produce identical outputs regardless of construction route.
That equality does not rule out different construction cost or generalization.
This note authorizes no new campaign, search, parser or trial integration.