Source Map
Primary Source
Official records: legislation, votes, court opinions, agency data, and government statements — the ground truth this app is built on.
News Reporting
Reporting from journalists describing what happened, with attribution and dates.
Expert Analysis
Nonpartisan research, academic, and think-tank analysis — labeled by its institution, not treated as neutral fact.
Political Statement
Statements from politicians, campaigns, or government officials — shown as what they said, not verified fact.
Trust Commitments
Always
- Prefer primary sources
- Link to original sources
- Show publication dates
- Distinguish facts from claims
- Explain uncertainty
- Correct errors transparently
Never
- Invent information, quotes, or sources
- Hide contradictory evidence
- Present opinions as facts
- Tell you who to vote for
- Promote a political ideology
Where The Data Comes From
- News — nine outlets' own public RSS feeds: NPR, The Hill, Fox News, ABC News, CBS News, NBC News, The New York Times, The Washington Post and MS NOW. Only the past seven days. Every headline links back to the outlet that wrote it.
- Members of Congress, bills and committees — the official Congress.gov API.
- Roll call votes — from the chambers themselves: the Office of the Clerk of the House via the Congress.gov API, and the Senate's own published vote records. Including how every individual member voted.
- Presidential actions — the Federal Register, the official daily journal of the U.S. government. Executive orders, proclamations, memoranda and determinations, reproduced with their own titles and dates.
- Commentators — each person's own public YouTube, podcast and newsletter feeds.
The banner at the top of every page shows when this was last refreshed.
How Major Stories Are Assembled
A Story gathers what several outlets published about the same event. It is assembled, never authored.
Articles are grouped by how much distinctive wording they share — rare words count for more than common ones, so "headquarters" identifies an event and "federal" does not. An event has to be covered by at least three different outlets to become a Story: one newsroom filing several follow-ups is a series, three newsrooms covering the same thing independently is a signal.
The Story's headline and opening description are not written by The Record. They are reproduced word for word from the single article most central to the group, credited to the outlet that wrote them, and that credit appears on the page. Everything else on a Story page is the other outlets' coverage in their own words, with links.
How Quotations and the Official Record Get Filled In
These are extracted, never written. Nothing is composed, summarized, paraphrased or characterized by The Record or by any AI.
What Both Sides Are Saying, on a story page, is built only from passages that were already inside quotation marks in the original reporting and already attributed there to a named speaker. The words are copied character for character and linked back to the article so you can check them in context. Where the speaker is a sitting member of Congress, their party comes from the official member record, and that is what decides which column a quote appears in. Where the speaker can't be identified confidently the quote still runs, just without a party or profile link. Anything ambiguous is dropped rather than guessed at.
From the Article reproduces the opening paragraphs of the original reporting, bounded at roughly three paragraphs and cut only at a sentence boundary. That bound is a copyright decision, not a design one: the outlet did the reporting and should get the reader, so every excerpt is attributed and links onward.
What We Know is not our summary. It is the official action history of whatever legislation a story formally cites, copied word for word from Congress.gov. Only formal citations ("H.R. 1234") are followed — a story that describes a bill in prose is never linked to one, because picking which bill was meant is a judgment call.
About the Voices Page
Elected officials aren't the only people shaping how a story lands, and most commentary gets read one side at a time. Voices pulls what commentators across the spectrum are publishing right now, from their own public feeds, and puts it side by side. The page is laid out as columns by side, so the comparison is the layout rather than something you have to switch on.
Each item is the title, date and description the commentator's own feed published, plus a link to the thing itself. The Record does not transcribe, summarize, rate, rank or fact-check any of it. The left/right grouping is a shelving label, not a score.
The Record does not claim a commentator has spoken about a particular story. It tried: posts were matched to individual stories by shared wording, and the result attached a podcast episode about celebrity gossip to a story about a federal court ruling. The obstacle is not fixable by tuning — a single video covers several unrelated segments, and its description is largely sponsor copy, so there is no reliable signal about what was actually discussed. Where commentary appears alongside a Story it is labelled as a topic match and nothing more.
X, TikTok and Instagram are linked but never pulled from. None offers a free public feed and all three block automated readers, so filling that space would mean scraped guesses or invented posts.
The Tools, and What Each One Is Built On
Four things sit under More. None of them is analysis; each is a different way of reading the same public record.
Where They Stand answers what has actually been done about an issue. It counts two things, and only two: putting your name on a bill, and how you voted when one reached the floor. Both are formal published acts. It deliberately does not collect what anyone said about an issue — sorting commentary by topic was measured here at roughly 45% wrong, and a page called "where they stand" built on that would be the most confidently wrong page in the app. A bill belongs to an issue because Congress.gov assigned it that policy area, not because we read the title.
Elections shows which seats are on your ballot and who holds them now, with their record. Which Senate seats are up is worked out from each senator's own term dates rather than a list typed in by hand — a hardcoded list gets appointments and special elections wrong. Where a seat can't be worked out from the official record it is left out and the page says so, because omitting a race is a smaller error than naming the wrong senator.
Compare Outlets puts two newsrooms' coverage of the same week side by side, leading with the events only one of them ran. Fact Checks reproduces what other organisations found, in their wording.
Nothing in any of them rates a person, scores a party, predicts a result, or says a position is right.
Following Things, Without an Account
Bills are the only thing you can follow, and that is deliberate. A bill moves stage, gets voted on, becomes law — each a real event with a date, detectable by comparing one refresh to the next. Following used to be offered on issues, politicians, races and agencies too, and none of it ever worked: nothing here detects a change to a person or a race, so those buttons promised notifications that could never arrive. You can still save an article, a Major Story or a fact check to read later. Updates shows a followed bill moving stage or being voted on, a saved story gaining coverage, and events several outlets covered at once.
There is no account, and no server knows what you follow. Every page carries all the changes, and your own browser decides which apply to you — that filtering happens after the page arrives. The list of what you follow is stored on your device and is never sent anywhere.
A saved or shared link keeps working after The Record stops tracking the thing it points at. Only a rolling window is live — seven days of news, the most recently acted-on bills — and pages used to simply stop existing. They are kept now: the page stays at the same address, showing what it showed when it was last checked, with a note saying so and a link to congress.gov or the outlet for the current version. Bills, Major Stories and fact checks are kept indefinitely. Individual news articles are kept for three weeks from publication and then removed — the site has to stay small enough to publish, and old single articles are the part that grows without limit. Before one goes you can download the page to keep it, and every saved item also carries the outlet's own link, which never expires; the page you land on afterwards offers it.
Anything with its own page can be shared — a bill, an article, a story, a fact check, a politician, a category. There is nothing to sign up for and the link carries no identifier; it is simply the address of the page.
The cost of doing it this way is that every page carries a little data most readers never see. That is the trade for not having an account and not tracking anyone.
What This Site Counts About You
The Record counts page views, so the person running it can tell which parts get used. Saying so here is not a formality: this site argues that people should be told what is being recorded about them, and it would be a poor argument if it quietly made an exception for itself.
What is counted: which page was opened, when, whether the window was phone-sized, the name of the site you arrived from if you came from one, and a random string this browser invented for itself so that "somebody came back" can be told apart from "somebody new arrived".
What is not: your IP address, which browser you use, where you are, your name, your email, anything you type into a search box, and anything at all about what you follow or save. That last one is structural rather than a promise — what you follow never leaves your device, so there is nothing for a page view to be joined to.
If your browser sends Do Not Track or Global Privacy Control, nothing is recorded at all. Those are you saying plainly that you would rather not be counted, and the whole premise here is that stated positions get taken at their word.
The counts go to a small service on the same account that publishes this site. Nothing is shared with an advertiser, an analytics company, or anyone else, because nothing else is involved.
What Is Deliberately Empty
Some sections of this app show nothing. That is a choice, not a gap waiting to be filled with something plausible.
- Congressional hearings, cross-branch activity and official statements — no free, clean, non-fabricated source is wired up for these yet, so they are empty rather than invented.
- Who is running for governor, or for any state or local office — the Federal Election Commission publishes candidate filings for President, Senate and House only, and every source that covers state races is commercial. Rather than compile a roster by hand that would quietly go out of date, those races link to the state's own election office. An attempt to build one produced confident, wrong answers, which settled it.
- An excerpt from every article — some outlets put their articles behind a paywall or decline automated readers. Where they publish the text in their own feed it is used; where they don't, the story page says so instead of leaving a gap that looks like a fault here.
- Fact Check shows checks other organisations published, never one of ours. Where a checker publishes machine-readable ClaimReview data the exact claim and their own verdict wording are reproduced; where they don't, the check still appears with no rating rather than a rating we inferred. A missing verdict is shown as missing.
- Most story pages carry no quotations, because most reporting doesn't contain a passage that is both in quotation marks and attributed to a named speaker. Under-filling is the intended failure mode.
A thinner page costs you a little context. Putting words in someone's mouth would cost the thing this app is for.
No AI Writes Here
There is no part of The Record where a machine writes prose. Every word on this site was either written by the outlet, the official, or the institution it is credited to, or is a count of something you can check yourself.
There used to be one exception — an AI assistant page, clearly labelled as the only place answers were generated. It was removed. A single generated corner quietly weakens the claim every other page depends on, and the exception was not worth what it cost.