Editorial Policy

This page explains how WhenMachines reports on AI and machine intelligence, how we check our work, and how to hold us to it.

Research standard

We write for readers who are smart but busy: people who need to understand what a model, tool, or technique actually does before deciding whether it matters to them. Every explainer starts with primary sources: model cards, technical documentation, research papers, release notes, public evaluations, and first-hand testing when it is possible. We do not treat press releases as evidence.

When we describe how something works, we try to separate what is demonstrated from what is claimed by a vendor. Where a capability is unproven or contested, we say so plainly instead of rounding up.

Review process

Daniel Reyes, our founding editor, reviews each piece before publication. That review checks the technical account against the cited source material: what a paper actually measured, what documentation actually promises, and whether the practical conclusion follows. For topics with material technical uncertainty, the reporting is checked again before it is published. The goal is accuracy, not the appearance of certainty.

Sourcing and citations

We link outward to the primary sources behind material claims wherever we can, so you can inspect the papers, documentation, model cards, datasets, and official technical records yourself. Secondary reporting can add context, but it does not substitute for a citation to the underlying evidence. When the evidence is limited, unavailable, or based on hands-on testing, we label that clearly.

Updates and corrections

AI changes quickly, and we make mistakes. An article's “Updated on” date changes when we make a substantive revision: a correction, a material new development, or a meaningful improvement to the explanation. Small copy edits do not reset that date. For substantial corrections, we note what changed in the article.

If you spot an error, email editorial@whenmachines.com or use our Contact page. Include the page, the specific claim, and the source that supports the correction. We check reports against the original evidence and correct confirmed errors; corrections take priority in that inbox.

Independence and AI use

We are not paid to cover, promote, or favor a product. Sponsored placement, advertising, affiliate relationships, and editorial coverage are kept separate; any material conflict of interest will be disclosed on the relevant page.

We may use AI tools for drafting, organization, or research support. They are not a source of record. A human editor remains responsible for every published claim and reviews the reporting before it goes live.