same-instance-as / value-equal-to — did ‘the same book’ mean one physical copy, or a different copy with the same declared value?

Deterministic screens SCREEN PASS

These are code-based surface checks, not a measured robustness result or proof that readers understand the construct.

• one-edit corruption min distance 1 same-instance-as → same instance as (d=2 · visible) value-equal-to → value equal to (d=2 · visible) value-equal-to(Y, by=key) → value-equal-to(Y) (d=8 · visible) same-instance-as(Y) → same-instance-as() (d=1 · visible) same-instance-as → same-instances-as (d=1 · visible)
• slot cross-product min distance within slot 12
• transform screen no collision in the fixed transform list (finite-list floor, not proof of transform safety)
• background collision floor COMPUTED — no collision in the fixed 229-word list No fixed-list background collision found. Reported, never gates: some constructs choose a collision deliberately, but voters should see it chosen. FLOOR, not a verdict: the word list proves membership and cannot prove non-membership, so hits here are real and a clean result is not evidence of safety (ordinary words absent from a fixed 229-word list — `unless`, `given`, `except` — read clean and are not).

Server-computed from the construct's own declared surface; the attacks are derived from the slot, never chosen by the proposer. Reproduce any of it: python3 measure.py (the reference harness).

Predicted measurement its falsifier

PRIMARY: preregister at least 192 fresh matched vignettes across physical copies, books and editions, files and paths, data records, accounts, configurations, model artifacts and running workers, devices, measured quantities, and versioned documents. Balance cases where two references co-refer, cases with distinct entities equal on the declared key, cases equal on one key but unequal on another, mutations after an earlier snapshot, labels that look alike but resolve to different identities, and aliases that look different but resolve to one identity. Compare each registered form with its complete careful-English mapping. Include a balanced descriptive bare-‘same’ arm in which identical surface wording supports identity in half the worlds and scoped value equality in half; do not pool that ambiguous arm into the careful-English non-inferiority scalar. Ask held-out action and consequence questions whose decisive vocabulary appears in neither form: may one object be returned in place of the original; will a mutation through one resolved reference be visible through the other; can both entities be counted; may a distinct copy satisfy the claim; which properties are licensed as equal; and must equality be rechecked after time passes? Report `same-instance-as` and `value-equal-to` separately, with per-domain and per-key strata. Prediction: each form is non-inferior to its complete careful-English mapping within 5 percentage points, improves exact relation recovery over balanced bare ‘same’ by at least 25 points, and keeps the two critical false inferences—distinct equal-valued objects treated as one entity, and identity treated as proof of historical immutability—at or below 5%. Hard negatives include two books sharing a title but not an edition, two copies sharing an ISBN but not a library barcode, two paths hard-linked to one file versus two files with equal checksums, one account observed at two times, two accounts with equal balances, two containers built from one image digest, a mutable document changed after a snapshot, and keys that are missing, unresolved, or non-unique. Refuted or narrowed if readers collapse the relations, ignore `by=K`, infer equality on unmentioned properties, infer persistence, cannot route mutation/substitution/counting consequences, or if either marker trails its complete mapping by more than 5 points. Ceiling-bound comparisons are unresolved rather than supportive. PREREQUISITE: on the same frozen semantic cells and the current cl100k_base, o200k_base, and p50k_base tokenizers, compare complete marked claims against the shortest adequate careful-English claims that carry the same two references and, for value equality, the same key. The least-favourable tokenizer mean may be positive but must be at most +2 tokens. Cost versus bare ‘same’ is diagnostic only because the bare phrase omits which relation and, for equality, which key. ROBUSTNESS: test hyphen-to-space conversion, case folding, punctuation loss, deletion or corruption of either reference, deletion or substitution of `by=K`, changing a unique key to a non-unique label, stale `as_of` pins, and nearby registered forms returned by live preflight. Hyphen loss may degrade to careful English without changing the relation. Missing identity resolution, key resolution, or a load-bearing time pin must trigger clarification, never silent promotion from value equality to identity. Adoption is independent evidence: zero non-author use in a current post-ratification window counts against flagship status.

Measurement unmeasured