NoFuckery AI logo

Methods / version 1.0 / 2026-07-26

NoFuckery AI

Editorial standard

A technical-sounding sentence is not evidence. A claim earns publication only when the record lets another person inspect its support and its limits.

Accountable editor: Chris McNosky · Published: 2026-07-26 · Recheck: 2026-10-26

LinkedIn mirror: NF-001 — What counts as evidence

Evidence labels

LabelMeaningWhat it cannot imply
direct observationNoFuckery or the named operator inspected or reproduced the cited artifact or behavior.Universality, independence, or reliability beyond the observed conditions.
source reportedA named source says the claim; NoFuckery has not independently reproduced it.That the source is disinterested or that repetition elsewhere is corroboration.
independently corroboratedA materially separate evidence lineage supports the claim.That organizations sharing data, funders, authors, or privileged access are independent.
inferenceA conclusion drawn from identified facts and reasoning.That the conclusion was directly observed.
opinionAn explicit judgment or recommendation.That preference has become fact.
unknown / disputedThe record is missing, conflicting, preliminary, or contested.Permission to fill the gap with the most dramatic story.

The bare word verified is avoided because it conceals who checked what, against which record, and when.

Primary records and exact scope

Publication gate

  1. Map every material factual sentence to a precise source passage, dataset, test, commit, or artifact.
  2. Capture the best available primary record and record its date, version, publisher, access date, and incentive.
  3. Map shared funding, authors, boards, datasets, model access, and partnerships before calling sources independent.
  4. Search deliberately for the strongest evidence that would overturn or narrow the proposed verdict.
  5. Recalculate numbers and check every quotation, name, title, date, image, caption, and headline.
  6. Disclose owned projects, clients, prospects, vendors, partners, and other material interests.
  7. Give seriously criticized subjects the material findings and a meaningful opportunity to respond.
  8. Require qualified human review when legal, security, privacy, rights, or domain risk demands it.
  9. Recheck volatile facts and links within 24 hours of publication.
  10. Name the human accountable for the final record.

If the gate is not earned, the workflow status is HOLD. Softer wording does not cure missing evidence.

Required structure

Every publication opens with a specific verdict, presents the supporting evidence, includes the strongest credible countercase, states what the evidence does not prove, and ends with the operator decision the record supports.

Headlines, thumbnails, diagrams, captions, and alt text pass the same claim test as body copy.

Strongest countercase

One high-quality primary record may be stronger than several superficial accounts. The method values source quality and evidence-path independence, not logo count.

Corrections

Central errors stop distribution and receive a visible correction or retraction. Material missing context receives a timestamped clarification. Privacy, security, or legal exposure is frozen and removed immediately after the record is preserved privately. Non-substantive edits remain in the internal log.

Read the public corrections ledger or email cmcnosky@gmail.com with the exact passage, contrary evidence, and requested correction.

What this standard does not prove

A written method does not prove that every future publication will follow it perfectly, that a small founder-led practice has newsroom independence, or that a reviewed claim is true under conditions it did not test. The correction record is part of the evidence for whether the standard is actually being followed.