How to read itSIG-001 to SIG-057 are Report 01, observed 30 August 2026. SIG-058 to SIG-105 are the Report 02 six-market sample, observed 31 August 2026. SIG-106 and SIG-107 were coded on 9 September 2026 during re-verification, when the market matrix was found to rest on two observations that could not support their rows.

Each signal cites exactly one source. Repeat citations are why 57 signals resolve to 48 distinct sources.

Re-check and ruling Grade A records the kind of source a signal cites, not whether the passage was on the page today. Re-check records what the 9 September 2026 retrieval found. Ruling records whether the observation is published, published as context, or held. Held observations are not asserted anywhere in the reports.

Withheld observations Six rows print their status instead of their observation, because the cited source will not retrieve. Every other observation was read against its source page by hand and carries a quoted passage. They stay in the register because it is an index of what is coded, and the count would be wrong without them. They are not published as findings and nothing in either report rests on them. The full log, with the basis for each ruling and the retrieval method and date, is published as the publication-ruling log (CSV). For each of the 67 retrieved sources, the source evidence record (CSV) publishes the URL, the retrieval date, the character count, a SHA-256 of the retrieved text and each quoted passage with its offset, so the check can be repeated. The retrieved page text is not republished: the pages carry their own copyright.

Every sourceAll 71 entries, including the two that no signal cites. They are shown rather than removed so the register reconciles against the published counts.

Every source is a page published by the institution itself.