Answers are grounded in this paper and cite their sources.
A one-dimensional binary cellular automaton whose cells carry their own B-bit rule words, incremented by the local activation and reduced modulo the ring width N, depends on N only through the residue r = N mod 2^B, and the excitability e_1 = (r - 2^(B-2))/r of its frozen sea is the control parameter. We follow the interface between the sea and the active phase at B = 8 to 12 in runs of up to 2 x 10^8 steps. Two features of the interface are set exactly by which bit of an ignited word the lookup can read; the rest is measured. For e_1 <= 1/5 no cell whose word lies in the sea range can ever leave it, and the interface cannot advance (a theorem for r >= 2^(B-2) + 2, verified by exhaustive reachability for B = 6 to 14). Above e_1 = 1/3 the measured velocity, in units of the clock period of the sea, jumps by a factor 9 to 12 within one residue step at every word length. Between the two points the active phase can creep into the sea, by climbs that need both neighbours of a cell to be active at once, and is measured to do so over up to eight residues below r_c. The creep falls by a factor of two per residue over the first residues below r_c - 1, r_c = 3 x 2^(B-3), and more slowly and irregularly further down; at fixed distance it does not grow with the word length and, from two residues below the threshold down, is smaller at the longer word at every one of the seventeen steps measured, at eight of them by two standard errors or more; and along a line of fixed excitability it falls by a factor of 60 to 90 when the one-step channel closes and by 5 to 7 per further bit where measured, to within two standard errors of zero at B = 11, while at e_1 >= 1/3 it stays finite. On this basis we conjecture a long-word limit in which the interface is pinned below one third and moves above it, a discontinuous depinning at the exact rational point, rounded at finite word length by the creep. Below threshold the sea is absorbing and, wherever the creep is positive, metastable; random states leave long-lived localized transients, and a developed active phase keeps the ring to 10^8 steps at B = 10 down to e_1 = 0.319 apart from small holes that open and close at steady rates. Every number is produced by the scripts and run records of a public repository, which a verification suite regenerates or re-analyses.
No takes yet. Share an insight, caveat, or question.
Plamen Stoychev (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: