The manuscript defines Recognition Boundary Theory around one explicit finite-dimensional object: the recognition step of Appendix A. A clock cell is C^4 with maps K(b',b)=(1/2)D(s(b)I-s(b')L). The full cell is H=C^{{0,1}^2}\otimes C^4\otimes C^4\cong C^{64}. The step is the real orthogonal matrix U=\sum_{x,y}|y\rangle\langle x|\otimes M_{yx} with M_{yx}=\sigma(x)(K(y1,x2)C_x)\otimes(K(y2,x1)C_x). Labels j(y,x)=(y1-x2,x1-x2,y2-x1) are booked on three accounts; posting T appends a label, shifts the loop count by j2, and applies the corresponding V_j. Direct verification from the printed definitions confirms that the clock maps satisfy the stated completeness relation and that U is orthogonal (U^T U=I). All subsequent claims labelled proved on the step are therefore conditional on this concrete real matrix and on the isometry and cylinder-load properties stated as Thm 2.14 and Thm 2.17 of the companion manuscript [1].
On-page Propositions 1-6 are elementary and correct under their stated hypotheses. Proposition 1 is the pigeonhole bound: a distinguishing reading through |B| sites of q values injects any set S into at most q^{|B|} settings and therefore carries at most |B|log2 q bits (3|B| for q=8). Proposition 2 computes the shell cardinality N^3-(N-2)^3=6N^2-12N+8 and notes that for N\ge3 the shell is strictly smaller than the block, so no shell reading separates all block settings. Proposition 3 is the standard irrational-winding fact: addresses (t mod P,t mod Q) are unique for all t if and only if P/Q is irrational. Proposition 4 follows at once from the prefix-invariance of T (Thm 2.14(c)). Proposition 5 is spatial invariance of any reading left unchanged by every motion of space. Proposition 6 is the inclusion of settings under a wider opening. None of these statements requires the 64-dimensional dynamics; they hold for any finite alphabet and any growing record.
The central dynamical claims are not self-contained. The walls (lineages labelled by initial occupation and loop count) are asserted to be exact superselection sectors for every operator built from the record's own events (Thms 4.5, 4.15, 4.17, 5.4 of [1]). The kept class is asserted to be exactly the four-dimensional aligned sector spanned by \psi\pm=|10\rangle\otimes P\otimes P, |01\rangle\otimes P\otimes P and the two swing phases on A\otimes A, cut out after three postings from the successive free-of-neutral dimensions 40-16-8-4 (Thm 7.11, Cor. 7.15). The eight-tick window test is matched to this class (Thm 7.16), the kept weight a_h is shown to form a martingale that converges to 0 or 1 with the initial kept weight (Thms 8.4-8.6), and the present is claimed to settle onto one of 256 pointer lines along almost every endless record (Thm 10.11). Network results (Thm 13.9: joining in round order creates no new kept weight; Thm 16.4: every record sees only 12 of 64 directions at a site) and the coupling/write theorems (Thms 18.20, 20.7) likewise live only in [1]. Appendix A supplies the matrix but not the spectral or iterative calculations; a reader of the submitted PDF therefore cannot re-derive the dimension counts, the 2^{-(n-1)} leak, or the pointer convergence. The paper is transparent about the dependency, yet the abstract and the status lines still present these results as proved for the step.
Physical scope is correspondingly limited. The step is real orthogonal, so every record load depends only on the real part of the state; the twelve-direction privacy of Section 12 is the same fact. Renou et al. [26] have already shown that real quantum theory is experimentally disfavoured for networks. The manuscript correctly flags the issue in Section 16 and leaves open whether the supplied coupling restores complex phases. Until that coupling is exhibited and shown to recover standard predictions where tested, every theorem on the step remains a conditional statement about a real matrix that is already in tension with experiment. In addition, which physical interactions realise the step, the map from a concrete system (brain, apparatus) to its boundary, the origin of the posting order, and the link from loads to observed frequencies are all listed as Open. Without those bridges the theorems give no verdict on any laboratory observer.
Identifications are cleanly labelled yet still drive the rhetoric. The load construction recovers a consistent family of cylinder weights (Thm 2.17) whose form is forced by non-negativity, normalisation and orthogonality; it is not a derivation of the operational Born rule or of which record is met. The eight-tick test and the kept class are both read on the loop account, so their numerical agreement is not independent evidence. The further step from kept by the record to felt (Section 14) and the readings of heaven, death and personal persistence (Sections 13, 17) are explicitly Identified or Open. The abstract and the closing vision nevertheless present the boundary-plus-record object as the solution to the one who is looking. That over-reach is the principal presentational defect.
Comparison with prior literature is accurate on the mathematical side and overstated on the physical side. Exact walls generated by the history itself are a perfect-record version of Wick-Wightman-Wigner superselection and of Zurek einselection; the repeated-record settling of kept weight has the form of nondemolition collapse (Bauer-Bernard); pointer purification recalls Maassen-Kummerer. The combinatorial opening bound is correctly distinguished from the gravitational holographic principle; no derivation of an area law or of the factor 1/4 is claimed or supplied. The paper's genuine strength is its status discipline (Proved/Supplied/Identified/Open) and its explicit how it could be wrong section. Its decisive weakness is that the bulk of the mathematics lives in an inaccessible 402-page companion and that the model-to-world map required for any physical claim remains open. The work is therefore best read as a carefully labelled finite-dimensional programme, not yet as a completed physical theory of observers.
preprint published 2026-11-14 18:55 UTC by preprint-admission
Preprint record published. Integrity screen clean; desk screen PASS. The review summary and available scorecard are attached as public metadata.