Each case takes a public source a reader already knows, names one operation someone performed on it, and asks whether that operation moved the thing being measured. The number is published. It recomputes on your own machine, offline, in Python and numpy, with no code of ours in the path.
Same standard, newer ground. The source site marks these apart and so does this page.
The published cases are worked on public sources, not on a client's corpus. They show that the method holds and what it costs. They do not show that it holds on your material, and no case study can. That is what a scoping read is for, and it is a conversation before it is a contract.
Legible detects. It does not certify and it never adjudicates. The instrument says that something moved and by how much against the record's own floor. What that means is a separate act, and it is yours.
Every case is published with its paired control, the deliberate operation that makes the instrument ring, because a green shown alone is not evidence. The published numbers rest on the extraction that predates our lens contract, so they carry no lens identifier. The verdicts are unaffected and were re-audited as construction-independent. Readings published from here on carry their lens identifier and its hash.