Methodology

01 / Story model

Coverage is organized by developing story.

A story board can remain active across days, weeks, or months. Its current state summarizes where the published record stands; its ledger preserves the individual developments underneath. Calendar dates help orient the record, but they do not replace the story.

02 / Chronology

Events stay in time order.

Entries are displayed by when the recorded event occurred. Filters can narrow what is already on the page, but they do not reorder it. Featured placement is an explicit publication choice, not a personalized or engagement-based recommendation.

03 / Confidence

Support and source identity are different questions.

Confirmed

The published entry has corroborating support.

Official

The claim is attributed to a named official or primary source. Official attribution does not itself mean independent corroboration.

Reported

The entry rests on a named outlet's published account.

Unverified

The available support remains insufficient or unresolved.

04 / Sources

Every entry links to its sources.

Each ledger entry displays its named sources as direct links. When a source publication time is available, it appears beside that link. Source count is not a substitute for source quality, and source identity is not collapsed into the confidence label.

05 / Election data

Polls, vote returns, and race calls are separate records.

Election polls appear one release at a time with the pollster, field dates, population, sample, margin of error, and methodology. Each chart represents that published poll only. We do not calculate a polling average or a win chance. Candidate portraits are sourced from Wikimedia file pages and are not evidence.

This desk is still pre-election. The Results box stays empty until an election office starts reporting votes. The Gate will not accept vote totals, media race calls, or certification as if they were campaign news.

When counting begins, candidate votes will come from attributed result feeds or election offices. Estimated reporting, a media race call, and official certification will be labeled separately so one cannot be mistaken for another. Until then the desk records official calendar dates, a selected race watchlist, individual poll releases, and chronological campaign developments. It does not publish race ratings, win probabilities, polling averages, or model forecasts.

05b / Left and right views

Left and right stay on screen together, with the method stated.

The six current issue baselines are the closest responsible equivalent to an “average” left or right view. Editors synthesize recurring positions in the main Democratic and Republican coalitions from current congressional or party agendas, then check that synthesis against recent partisan survey data. The date on the module is the research cutoff. It is not a numerical average, and it does not imply that every Democrat, Republican, liberal, or conservative agrees.

Midterm names, races, pollsters, and institutions also open two-column briefs. Those are current partisan argument clusters from named campaigns, officials, surveys, and published outlets, not coalition averages. When a coalition has a material internal split, the brief must say so instead of manufacturing consensus.

The L/R marker next to a name is an ordinary link to a permanent, source-linked comparison page. Both columns are rendered in the page HTML, so the guide remains useful without JavaScript. A reader action never calls a model, and the guide does not declare a winner.

05c / Economy data and policy

Official figures and economic policy remain separate records.

The economy desk records the figure each agency published, with a link to that release. Jobs come from the Bureau of Labor Statistics Employment Situation. Consumer prices come from the BLS CPI. The Fed's preferred inflation family and quarterly growth come from the Bureau of Economic Analysis. The policy rate is the Federal Open Market Committee target range from the last statement.

We do not publish funds-futures implied odds, a hike or cut chance, a private forecast path, or a combined Wall Street estimate. When we put an expectation next to a published figure, it is the FOMC Summary of Economic Projections, labeled as participants' own medians with the vintage mismatch in the table. The 2 percent line on the inflation chart is the Federal Reserve longer-run goal, not a forecast of the next release. A date on the calendar is a scheduled release, not a prediction of the number. Quarterly growth is a BEA release and lives with the releases, not inside the inflation row. The published figure and the table behind it are the record. If JavaScript loads, an Observable Plot chart (built on D3) can add a hover tip. It does not add a forecast or hide a number. Iran shipping and oil figures on the conflict desk are context for that story. They are not economy prints.

Tariffs, sanctions, export controls, trade actions, taxes, spending, benefits, and regulation live in a separate chronological policy ledger. Each record labels its lifecycle, direct legal mechanism, covered parties, effective timing, authority, established facts, unresolved details, and sources. A proposal is not an effective tariff. A sanctions warning is not a designation. A signed action is not necessarily being collected or enforced yet.

Settled legal-status claims require the operative government or court source. Major wire reporting may explain implementation and disputes, but it does not replace the order, rule, notice, license, customs guidance, or judgment. We state who is directly required to pay or comply and do not assign the ultimate cost to consumers, workers, exporters, or markets unless the cited evidence supports it. Murky scope, exclusions, refund mechanics, cure periods, general licenses, and enforcement discretion stay visible as unresolved rather than being guessed away.

06 / Freshness

“Active” does not automatically mean “live.”

Story status, latest-entry time, and last-checked time are shown separately. The Iran ledger shows a freshness warning after six hours without a check. Midterms and the economy desk show the last checked time only. Ended and archived boards retain their final coverage dates.

07 / Scheduled automation

Automated research runs on a fixed schedule.

The research system runs one fixed two-hour transaction that refreshes Iran, midterms, and the economy desk together. Accepted entries must pass source, URL, timestamp, timezone, and duplicate checks before entering the published record.

When the evidence changes, an automated synthesis may refresh the current-status paragraph and up to three watchpoints. Each statement links to the entries that support it. If the synthesis cannot be validated, the previous cited briefing remains in place. Reader searches never trigger this process.

08 / Corrections

Corrections stay attached to the published record.

Source-verified timestamp or factual corrections update the affected ledger entry without changing its stable URL. Material methodology changes are documented here rather than silently changing what labels mean.

09 / Reader contract

The same chronological record is available to every reader.

The current state and newest ledger entries are rendered in the page and remain readable without JavaScript. No reader login is required. Search and filters do not personalize or reorder the record. JavaScript may highlight the section in view, slide a date chip, or add a hover tip on a figure. It does not change the printed number.

The global RSS feed and per-story RSS feeds publish the same dated entries for feed readers without personalizing or reordering coverage.

Event time is shown both where it happened and in U.S. Mountain Time, so a distributed sequence can be compared without losing local context.