Paper 62 stratified the Boltzmann-brain comparison on the acquisition axis and typed the practice's four standard cosmological measure specifications — the proper-time cutoff, the scale-factor cutoff, the causal patch, and the causal diamond — as uniformly provenance-blind, as specified: no count variable anywhere factors through warrant strata. It selected nothing, and it left exactly one successor row open, add-only and welcomed: "provenance-respecting specifications (the add-only row). " While that cell stays empty, two misreadings prosper. Read one way, the classification looks like an impossibility result — as if nothing could respect provenance. Read the other way, any future occupant arrives as a recommendation — the entrance of the Global Selector, the specification-layer ranking/selection authority for whom a provenance-respecting specification would be "the better measure. " This paper fills the cell and starves the Selector: the row is taken constructively, and nothing is selected. The instrument is the PRSpec normal form, declared in two layers: "PRSpec is defined generally for any base specification admitted by 62's add-only extension route. This paper constructs and audits PRSpec instances only for the four reference specifications classified in Paper 62. " Its stratum dependence is a finite declared table f over FibStrat's cells — one domain, the partition induced by WGrade's declared finite codomain — together with a declared-support list, so the respecting-test is decidable as three finite-table checks: the table exists as a declared field; f is non-constant on the declared-support strata; f's inputs are WGrade values only. A wᵖrev-type prevalence input fails the third check and is typed to the claim layer — exactly where Paper 62 sent it. Intuitively, a PRSpec is a measure specification whose counting rule is allowed to see a member's acquisition/export grade, but not how prevalent that grade is. The constructions occupy the row: for each reference specification, the graded-weight augmentation Kᶠₛpec = Kₛpec (f o WGrade), lifted by a displayed case distinction (counted occurrences re-weighted through the eligible members they already are; off-fiber occurrences keep weight 1), normalized by f (intact) =1, bounded, support-preserving, ontology-fixing, Prev-independent, n/a-neutral, and subject to the non-exclusion condition f>0 — Paper 10's rule in kernel form: no stratum is excluded by weight zero. Each row's declared support is carried by a certificate — source-provided where the source row itself displays the comparison (the scale-factor cutoff), conditional on the displayed benchmark instance (H-sup) elsewhere, the grade displayed per row and never averaged. The boundary theorem comes in two halves with matched quantifiers: every grade-weight factor f o WGrade is invariant under all of 62's deformation class W; the Prev-conditioned candidate moves under 62's exhibited mover w'=1+1S — one mover suffices for failure. The classification is extended add-only: Table A re-displays 62's four rows character-identically; Table B adds the four augmentation rows; the negative-example check reproduces 62's blind typing (the live K1 test). The ceiling is strict, and it is the paper's registered slogan: "To construct is not to select; a type is not a ranking. " The selection seat stays empty on every branch: no clause consumes a function from PRSpec to selected / rejected / better / worse (the no-selection invariant, proven clause-exhaustively) ; provenance-respecting is a type label, not a virtue label; and "the original provenance-blind specifications remain admissible-as-typed; they are not demoted, invalidated, or corrected by the new rows. " Measure selection, Boltzmann-brain frequencies, anthropic and doomsday-style credence, and the physics rows stay open at (i5), where 62's (z5) left them; no datum is scored on any branch. SpecStatus (provenance) = scoped at typing/model grade on the construction domain — the support grade displayed, never silently upgraded.
Tomoyuki Uchida (Mon,) studied this question.