How the record is built and checked.

From source to page.

1. Read the official record. The ISO page gives the stage, the dates, the edition and the page count. The OBP gives the public scope, references, terms and bibliography. The BSI draft cover gives the national ballot window.
2. Read the held text. The 2020 requirement set gives the clause records and the annex rows. The 2022 set gives the current clause requirement for each annex page. The 19011:2026 text gives the change the DIS aligns to.
3. Write the record. One entry per unit: code, title, citation, binding party, requirement in plain English, evidence, gaps, probe, preparation, cross-references, dates and a source pointer on every field. Where the text is silent, the field says "not stated in the text".
4. Paraphrase and test. Every authored field is tested against the held sources with a strict seven-consecutive-word test. Titles, defined terms and data element names are the only verbatim strings. A hit is rewritten, not waived.
5. Generate every output from the record. The pages, the free check, the kit, the CSV, the XLSX and the plain-text files read the same record. A fix is not done until they agree.
6. Refresh and diff. scripts/refresh re-downloads the public text, rebuilds the record and fails loudly on a change. That diff is what the paid watch sends.

What is interpretation

The DIS body is not public in full. The clause-level treatment is our reading of the public structure and the 19011:2026 change. Each page says where the public record is silent. The annex note is our reading of the 2022 clause record and the DIS foreword. The planning date is an estimate from the stage, not an ISO date.

What is not done

No page is generated by a model at request time. The free check does arithmetic in the browser. The kit is a file generated at build time. The watch is a diff of a record, not a summary written by an engine.