HOW WE WORK

Every article on DiscordantNote follows the same pipeline:

1. Idea generation. An AI proposes candidate stories across our six columns, weighted toward less-canonical, underreported subjects.
2. Editorial selection. A human editor reviews the pool and selects only stories that are (a) sourceable in credible mainstream publications, (b) genuinely underreported, and (c) substantively about a mechanism, not just a curiosity.
3. Drafting. The selected topic is developed into a 500–700 word article. The draft must cite at least one named source and include specific dates, named entities, and concrete numbers.
4. Editorial review. Every draft is read end-to-end before publication. Claims without a clear source are removed. Sweeping conclusions are softened. Names are verified against the cited source.
5. Publication. The article goes live with the source linked at the bottom and the source publication or institution named in the body.

We accept the following as primary sources:

• Investigative reporting in Financial Times, The New York Times, Reuters, Bloomberg, The Economist, The Guardian, The Atlantic, ProPublica, OCCRP, ICIJ.
• Peer-reviewed academic papers and books from university presses.
• Declassified government archives (US National Archives, UK National Archives FCO releases, etc.).
• Court filings, regulatory disclosures (SEC, FCA, EU), and other primary documents.
• Leaks of established public-interest provenance: Panama Papers, Pandora Papers, Paradise Papers, FinCEN Files.

Wikipedia is used only as a navigational starting point. Its citations are followed back to the original source, which is what we actually link.

• Speculation. If we cannot point to a documented source for a claim, the claim does not appear in the article.
• Conspiracy theories. QAnon-style narratives, "shadow government" speculation, and pattern-matching without evidence are out.
• Active criminal investigations involving private individuals. If a case is open and the subject is not a public figure, we wait.
• Unverified gossip about living private people. Even if entertaining. Even if probably true.
• Content that exoticizes minority cultures. Strange Ritual covers practices that reveal something uncomfortable about how power and coordination work — not "look at these unusual customs".

We use large language models to assist with two tasks:

• Proposing candidate stories from a defined editorial brief.
• Producing a first draft of the prose, given a specific topic and source.

We do not treat AI output as ground truth. Every factual claim in a published article has been read by a human editor, and where the source did not support the claim, the claim was removed. Where AI inserted a name or a date without a traceable source, that name or date was either verified externally or cut.

AI is a writing tool here. It is not a substitute for editorial judgment.

If you find a factual error in any article, write to us. Include:

• The URL of the article.
• The specific claim you believe is incorrect.
• A source for the correction.

We will respond within 7 days. Confirmed factual errors are corrected directly in the article, with a dated note at the bottom of the page explaining what changed and why. We do not silently edit history.

Disagreements of interpretation, framing, or selection are not corrections — but we read them, and they shape future editorial choices.

Corrections and contact: /contact.

DiscordantNote is independently operated. We accept no payment in exchange for coverage, no sponsored articles, and no editorial input from third parties. Articles are not written about, or against, the financial or strategic interests of advertisers, because we have no advertisers.

If this changes — for example, if we ever take sponsorship, run ads, or partner with a publication — we will say so on this page first.