How we work

Editorial standards

Our deal with you: label what kind of claim you are reading, show the limits of the evidence, and keep humans responsible for every published word.

Sources we actually inspect

Factual claims should trace to material we inspected: a study, dataset, official statistic, or original document. Our working notes distinguish full text from an abstract or bibliographic record. A search result is not a source.

Our articles are not peer-reviewed publications. We use peer-reviewed research where it fits and say when the evidence is thin, mixed, inaccessible, or absent.

Four evidence lanes

  • Research finding: tied to a source, sample, and limitations.
  • Interpretation: a reasoned reading that may be disputed.
  • Experience: something a founder lived, not proof about anyone else.
  • Opinion: a position offered openly as a position.

Women, men, and brains

Average group differences are overlapping distributions from particular samples, not destinies or diagnoses. We do not turn “this study found an average difference” into “men are” or “women are.” Counterevidence and women’s original scholarship belong in the substance of the work.

Two founders, separate voices

Founder experience is labeled and chosen by the person telling it. Neither founder speaks for the other. His, hers, and joint pieces stay visibly distinct, and genuine disagreements remain visible rather than being edited into false consensus.

AI works through a disclosed editorial system

Our locally run AI system can research, commission, draft, audit, and publish scheduled editorial articles. It follows the publication’s maintained research, thinkers, editorial, fact-check, and discovery instructions; retains its source and generation records; and labels metadata-only research honestly. Founders can correct, archive, or stop the automation. AI-assisted publication is disclosed rather than presented as unaided human authorship.

Commerce without purchased conclusions

Reviews disclose how an item was obtained and whether a commercial relationship exists. Affiliate eligibility never decides the verdict.

Companies can buy our attention. They cannot buy our opinion.

Corrections

Material factual errors should be corrected with a dated note explaining what changed. A thesis-level mistake should be named, not silently rewritten.

Boundaries

We do not diagnose readers, invent rarity statistics, treat feelings as proof of another person’s motives, or turn responsibility into blame for abuse, coercion, deception, or violence.