THE RECORDEvidence · context · accountability
Updated October 2, 2026
The Record’s value depends on keeping those categories visible, preserving corrections, and refusing to turn a large archive into an unreviewable claim of completeness.
THE FACTS records the sourced event. SIGNIFICANCE states the editor’s contextual analysis. GOALPOST / RESPONSE records the strongest relevant defense, explanation, or shifting rhetorical frame without treating it as established fact.
Primary legal documents, official roll calls, bill text, and agency records are preferred for what an institution formally did. Independent reporting is used to corroborate, supply context, and identify disputes. Official press releases establish the speaker’s position—not the truth of every claim inside the release.
Every release has a visible checked time. Relative words such as “today” are avoided inside durable entries. A current brief never implies that the entire historical archive was revalidated in the same pass.
Send the entry ID, disputed wording, and supporting documentation to the project address. Corrections identify what changed and when. A materially relevant response from an affected office or candidate is attached to the relevant entry rather than hidden in a separate page.
The canonical file stores 4,742 legacy rows: 4,731 active records and 11 retained duplicate tombstones. The national bridge adds 333, producing 5,064 active records at runtime. Tombstones redirect old IDs but never enter search, source totals, or the event count. The same build counts 7,694 source references and 6,300 distinct stored URLs.
Current research organization, drafting support, code, and adversarial review use ChatGPT 5.6 Sol Max and Claude Fable 5 Max (Cowork). Phillip Linstrum is the acceptance authority for canonical publication. AI output is not a source.
For auditability, the public federation JSON may expose substantive AI-assisted research drafts before review. Public accessibility does not make them published findings: staged drafts are labeled unreviewed, unauthorized for canonical publication, excluded from canonical totals, and not rendered as accepted analysis. The five April 2026 AI Opinion essays retain their original attribution to Claude (Anthropic, Opus 4). This October 2, 2026 pass updates the surrounding maintenance disclosure; it does not silently reassign or rewrite those essays’ authorship. See the provenance record.
This release contains 337 dated current-layer entries, including 4 national records added or materially refreshed in the backfill beginning August 17 published in release 10.114.0. It does not certify the complete legacy archive, independently validate every historical source, or resolve all open IN-6 claims. The canonical legacy and current bridge layers overlap; the generated metric detects no inter-layer continuity gap. See the generated quality dashboard.
NO SOURCE → NO FACT CLAIM OFFICIAL STATEMENT → WHAT THE INSTITUTION SAYS INDEPENDENT REPORTING → CORROBORATION AND CONTEXT ANALYSIS → LABELED EXACT COUNT → GENERATED FROM CANONICAL DATA CORRECTION → TIMESTAMPED AND PRESERVED