How This Brief Is Made

The method behind Noise Removed: America — Clear Signal v7.0 and the Audio Edition

Why we’re showing our work

Noise Removed: America — the daily brief and its fifteen-minute spoken edition — is produced by artificial intelligence under human direction. We could be vague about that. We’re not going to be. A reader’s trust in any news product depends on knowing what produced it: who chose the stories, who wrote the words, who checked the facts, who decides what publishes, and who answers when it’s wrong. This page says all of that.

If what you read here doesn’t meet your standard for trustworthy news, the right answer is to read and listen skeptically, or not at all. If it does, the trust is earned rather than assumed.

Original journalism is essential

We do not replace original journalism. We could not exist without it. Every story in the brief draws on reporting done by working journalists at named outlets; every story’s Sources line names them. Their work is our raw material.

Journalists provide the signal. We remove the noise. The brief’s job is to take reporting from across the political spectrum and strip the framing, the spin and the partisan packaging that drowns it out — and then show you the framings, named, beside the record, so you can see what was removed. The signal is theirs. The discipline of removing the noise is ours. Both are required; neither alone produces what you get.

Who does what

Produced by AI. Directed by humans. That credit closes every episode and every brief, and it is literal.

Directed by Chad Kaul. Chad designed the method and directs the operation. He wrote the editorial mission — equipping citizens as sovereign voters in a constitutional republic — the criteria stories are scored against, the structure every story is built on, the disciplines that govern verification and tone, and the rules for what the AI may and may not do. He makes the calls the rules don’t cover, he listens to every episode before it publishes, and he answers for every word. He is the person behind The Republic Desk.

The Republic Desk is the newsroom’s name and its editorial voice: short declarative sentences, concrete diction, attribution over adjectives, the framings laid out and the record laid beside them. It is not a person. It is the register everything here is written and spoken in, and it is the desk that answers for the product: republicdesk@signalmedianews.com.

Produced by Claude. The daily work — the research, the fact-checking, the scoring, the drafting, the self-review, the script — is done by Anthropic’s Claude. The production model is named in every byline (Claude Opus 5.5 as of this writing); when it changes, the byline changes, because different models produce different work. The method itself is developed in a standing workshop between Chad and Claude, where every rule on this page was argued over before it was written down.

Hosted by James Calder. The episode is read by James Calder, an AI host designed by The Republic Desk. He is not a person. His voice is synthesized with ElevenLabs; the words he speaks are the day’s brief, rendered into a script — he adds nothing to it, holds no opinions about the news, and reads no endorsements. His page says what he is and what he does. A person listens to every episode before it goes out.

The method

The method is called Clear Signal. It is written down, version-stamped in every byline, and it changes only through the workshop, on the record. It is audited — a fact-check audit, a bias audit, a voice audit — when the workshop judges one is needed, and the findings become spec changes; the audits are the method’s tuning instrument, not a daily guarantee. The daily guarantee is the receipt. The parts of it that are editorial intellectual property — the scoring, the framework, the discipline roster — are not published. What is published is what it commits to, and the receipt it leaves behind.

Every morning it:

Reads the day’s reporting from across the political spectrum — sixty to eighty articles, in full, left, center and right — including a directed pass through right-of-center and center-right outlets, so that the pool of candidate stories is built from every side’s front page, not one.
Selects stories by what they mean for you as a voter, scored against a fixed set of civic criteria, with a diversity rule so no single topic dominates the day.
Separates the record from the framing. For every story: what happened, why it matters, the background, who is affected, what is disputed — then each side’s framing, named and attributed, then the factual record with a plain confidence rating. We do not manufacture symmetry: where the evidence is one-sided, we say so.
Measures the spin. Every story carries a Spin Index, a 0–100 gauge of how much framing it arrived wrapped in. The three most spun become the stories told in full on the air.
Traces every fact. No name, number, date or outcome comes from the AI’s memory. Each is traced to a cited source or a primary document — or flagged. What can’t be traced doesn’t publish.
Checks itself for bias as it writes, under a roster of named disciplines — loaded language, source-voice independence, framing symmetry, antecedent framing, and others — and counts every application.
Writes the brief, then the script. The brief, about 12,000 words, by 6:40 Eastern. The spoken edition, about 2,500 words, by eight.
Shows the work. Five numbers in every masthead and four in every spoken sign-off, each counted by a tool from a written ledger, never asserted.

The Audio Edition

Noise Removed: America is fifteen minutes, weekday mornings and Sundays, by eight Eastern. Six segments, the same every day: The Day in a Minute · The Top Ten · The Next Five · The Lead and the Loudest — the lead story and the two most spun, told in full with every side’s framing read and named · The Questions — four things the record can’t answer yet, addressed to the people who could · Worth Discussing Today — three facts to take with you. Then the receipt, spoken: how many articles were read, from how many sources, how many facts checked, how many biases unspun. Then the credit.

A material error is corrected out loud at the top of the next episode, in a fixed slot, and in the notes of the episode that carried it. No episode is replaced or quietly edited.

The Methodology Pledge

Every brief publishes five figures in its colophon; the sign-off speaks four of them. Each is a count, never an estimate, and each is counted by a tool from a ledger the run writes as it works:

Articles Researched — distinct articles the run read and used, at any depth, for a purpose. Pages considered and not opened never count.
Sources — distinct outlets in the brief’s Sources lines, counted from the published text.
Citations — outlet citations across all Sources lines, with repeats.
Facts Checked — fact-points in the published stories confirmed against a cited source, once per story, from a written ledger of each fact and its source.
Bias Unspun — every application of the anti-bias disciplines across the brief, one per application. It is the number behind the word unspun.

Each has an expected range. A reading outside it opens a review of the instrument — never a number to hit. The figures are recomputable from the ledgers and the Run Metrics behind each brief.

The editorial commitments

Fact-anchored. Wire services and public broadcasters anchor verification; primary documents are preferred over secondary reporting; cable news and partisan sites appear only as attributed reactions, never as the record.

Non-partisan, not bland. A partisan reader whose side doesn’t have the facts on a given story should feel disappointed at the record — never angry at the brief. We aim at the truth-anchor, not the comfortable middle.

Citizen-first. Every story is translated into what it means for people — voters, families, communities — and what to watch next.

Anti-spin, shown. You see the framings, not just the absence of them. The point is that you recognize the spin when it reaches your feed later in the day — including from your own side.

Disciplined prose. Short declarative sentences. Concrete diction. Active verbs. Adjectives and adverbs stripped to a minimum.

Corrected out loud. In the brief, on the corrections page, and on the air.

What we don’t do

No opinion columns, no editorializing. The brief and the host report; neither has a take.
No anonymous sources. Every claim traces to a named outlet or a named primary document.
No AI-generated images of real people or places, ever. The brand’s images are instruments — a spin meter, a chart of the record — never a synthetic photograph.
No manufactured symmetry. A well-documented claim is never presented as the equal of a thin one for the sake of balance.
No endorsements. James Calder endorses nothing — not a product, a candidate or a cause. If a sponsor is ever read on the air, it is read as a sponsor’s message and labeled as one; political buyers are not accepted.
No undisclosed AI. Every brief and every episode carries the credit and the model’s name.

How to verify any claim

Every story closes with a Sources line. Find the story, read its Sources line, look up the cited outlets’ coverage, compare. If a claim doesn’t match its sources, that is an error, and we want to know: republicdesk@signalmedianews.com.

The byline, explained

Directed by: Chad Kaul · Voice: The Republic Desk (Clear Signal v7.0 + Claude Opus 5.5) · Published as drafted; review on demand

Directed by Chad Kaul names the person responsible for the method, the standards, the mission, and every morning’s rulings — the person behind the Desk. Directed, not designed: the design is a thing he did; the direction is a thing he does.
Voice: The Republic Desk (Clear Signal v7.0 + Claude Opus 5.5) names the editorial voice, the method and its version, and the AI model that did that day’s work. Either updates in the byline the day it changes.
Published as drafted; review on demand is the honest description of the loop for the written brief: it publishes as the method produced it, and a person reviews any story a reader flags. The spoken edition goes further: a person listens to every episode before it publishes.

This page describes the editorial practice in public terms. The scoring systems, framework, discipline roster and production sequence that implement it are the proprietary intellectual property of Signal Media and are not published.