Rewrite audit
Four passes were repeated until the document had a clear product hierarchy, traceable requirements, an honest trust model, and one accessible screen/print source.
Round 1
Finding. The source mixed product definition, implementation detail, future architecture, and governance without a short reading path.
Change. Added a new cover, executive orientation, direct vertical-slice path, six document parts, and one-sentence purpose summaries for all 37 sections.
Gate. A reader can identify the first build without reading the optional browser-native or network targets.
Round 2
Finding. Normative language existed, but requirements lacked stable identifiers and machine-readable extraction.
Change. Indexed every normative paragraph and list item, generated exact hashes, source ranges, status labels, and downloadable requirement and source-map manifests.
Gate. Every MUST, SHOULD, and MAY can be linked, tested, compared, and audited without changing its original force.
Round 3
Finding. Permission, participant trust, assurance, verification, and acceptance were easy to read as one security score.
Change. Made the four independent decisions explicit in the front matter and preserved the complete participant-trust, receipt, dispute, falsification, and acceptance architecture.
Gate. The document never treats a signature, sandbox, model response, attestation, or majority vote as semantic truth by itself.
Round 4
Finding. The source was complete but visually flat, difficult to navigate, and not designed for paged media.
Change. Added semantic landmarks, keyboard focus, searchable navigation, readable and compact modes, section status, print contents, A4 rules, running strings, page counters, widows/orphans, compact code, and reference columns.
Gate. The same HTML remains accessible on screen and produces a compact, legible print edition without a separate content fork.
c49684e0b01487206118be9eb4f790f54f91b378e62cc59664ffb8a2f5fb739f