How this record is made

Scandalous does not claim to be a real-time news wire. It is a changelogged, versioned record of documented governance scandals. A maintainer opens a case only when it clears the bar below; from then on the file grows by appended updates, never silent rewrites. Every change is a diff you could audit.

Keeping up with a chaotic news cycle by hand is impossible, and letting an unsupervised model publish is irresponsible. So the workflow is deliberately AI-assisted but human-gated:

1. 1Humans open cases. A person decides what qualifies as a scandal. Editorial judgment about significance is not delegated to a model.
2. 2AI drafts updates. Between reviews, an assistant sweeps each open case for new filings, rulings, and high-reliability reporting, then drafts a candidate update with sources attached.
3. 3A human approves the diff. Nothing publishes until a maintainer reviews the draft and merges it. The pull request is the gate; git history is the audit trail.
4. 4The freshness is honest. Each case shows when it was last reviewed, so readers never mistake a quiet case for a current one.

A claim is only opened as a case when it reaches at least the Preliminary confidence level: multiple independent outlets reporting, or a single primary document. Anonymous, single-source claims remain Rumor and are hidden by default. Routine partisan rhetoric, policy disagreements, and ordinary process complaints are not scandals and are left out on purpose.

Every case and every appended update carries a confidence level:

Evidence is tiered by independence and accountability — not by whether a link ends in .gov. Records produced by adversarial, adjudicative, or independent-oversight processes carry the most weight. Communications issued by political actors — the White House, agency press shops under political leadership, cabinet officials, elected officials — are treated as attributable claims: evidence that something was said, not that it is true. They sit in Tier 4 regardless of the domain they are published on.

We do not cite Wikipedia or crowd-edited encyclopedias as evidence. Where a primary document is behind a bot-wall, we cite a stable archived snapshot so the link keeps resolving without losing the source's weight.

Modern politics runs on assertion, so we refuse to flatten a case into “one side says this, the other says that.” Each case file separates what is proven from what is merely claimed:

Sources are weighted by tier. Primary documents carry the most weight; opinion and commentary can never establish a fact on their own. A case's evidence score blends an editor's assessment with the tier-weighted signal of its attached sources.

A case moves to Resolved when there is a final legal or official outcome (conviction, settlement, IG finding, formal exoneration). It moves to Closed without findings when a competent review concludes with no finding of a violation — including when a prosecution ends by pardon or dismissal rather than a verdict. Cases are never deleted — the record persists.

New developments are added as dated updates — each with its own source and confidence — so readers can watch a story unfold rather than seeing it silently overwritten. When the fuller picture calls for a broader revision, we make it openly: every change is a reviewed edit with a visible history, and factual errors get a correction note. Where a specific claim awaits re-verification, it is labeled as such rather than asserted.

The same confidence and evidence thresholds apply to both spectra — Right and Left. The Statistics page publicly tracks the balance of cases across spectra so any systematic skew is visible rather than hidden. Institutional actors — courts, agencies, and officials — are filed under the party whose appointees or members are the subject of the conduct.