Admitted Continuations Do Not Determine Their Registration: Exact Losses of Border Readings in a Discrete and a Continuous Cell
A constrained generative system admits, at each step, some continuations and excludes others. A reading of its border records some of the resulting distinctions and fuses others. We ask what such a reading keeps, what it loses and what recovers the loss, in a discrete and in a continuous cell, and we prove that the admitted continuations alone do not determine what the border registers. In the discrete cell, two presentations of the same edge shift agree on every conjugacy invariant of the admitted continuations, yet their registrations of excluded continuations grow at different rates (ρ_S^d = 1 against ρ_S^d = 2, certified for d = 2, …, 6); moves on entering arcs preserve the registration, and a registration is transported along a fixed conjugacy if and only if no distinction is created and none is lost, a criterion with four outcomes of which three have witnesses. When return lengths repeat, the fibre of a recorded word has the exact size Π μ(ℓ_j); its mean growth rate is the growth gap δ_L, and its extreme rates are explicit. The first returns to the branching piece form a basis of the first homology of the lift; the per-event reading loses exactly the kernel of a period map built from border cocycles, of rank Σ(μ − 1), strictly smaller than the loss of the aggregate time, and the class of a word in homology is necessary but not sufficient to determine its reading. A construction from primitive constraints alone splits the articulations of the border into excluded, fused and residual classes, faithfully to the definitions of oriented border recursion for return articulations under the declared reading. In the continuous cell the local reading (div, curl) loses exactly the harmonic space, of dimension b₁, which b₁ circulations recover; readings of a fixed domain at variable resolution register different numbers of holes while the area is unchanged, and two domains of equal area register one and four holes. A prediction on the inner regime of a self-similar boundary is reported as falsified. The two cells share a structure (a space of cyclic distinctions, a local reading with an exact loss, recovery by added readings limited by symmetry) but in different types, homology and cohomology; a common law is stated as an open problem. All statements are proved or cited with their hypotheses; computations are pre-registered controls.
Authors
- davide lugli
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-20
- DOI
- https://doi.org/10.5281/zenodo.24407604
- Primary Topic
- semigroups and automata theory
- Type
- preprint