Floor-Cleared Brand Velocity: Measuring Multi-Epoch Spectral Velocity Against the Apparatus Floor

A raw multi-epoch reading change from a large language model brand instrument confounds brand motion with model-version apparatus drift: a re-measured profile may move because the brand changed or because the reader did. This paper introduces floor-cleared brand velocity, a pre-registered measurement crediting spectral velocity only when a brand’s live-panel displacement across two epochs exceeds the contemporaneous apparatus floor from a byte-identical pinned panel re-read across the epoch pair. It further pre-registers a motion-drift double dissociation over a partition fixed ex ante. Forty brands were read at two epochs 77 days apart. Mean live-panel displacement of .00560 gives an aggregate ratio of 1.916 (bootstrap lower bound 1.534), so the typical brand does not clear the threshold of two, and the dissociation interaction is null on both pairs. The brand signal is positive with intervals excluding zero under both metrics. Six of 40 brands clear the floor at the pre-registered threshold, meeting HV1’s existence criterion, though five sit below the .0162 minimum detectable effect and the interval rests on two operator families. No vendor version shipped inside the window, so the floor the six cleared is apparatus non-determinism rather than version drift, and the correction is demonstrated against the wrong quantity. Includes zharnikov-2026be-prism-t-velocity.yaml (Paper Spec v0.1.0) – a machine-readable specification of the paper’s claims, assumptions, and dependencies. The paper’s full machine-first bundle (the SPINE claim/dependency graph and the ONTOLOGY term module) lives in the public repository; see github.com/spectralbranding/paper-spec for the standard. This PDF is generated programmatically from that machine-first source under a research-as-repository model.