Question
The question arises in a real conversation between a person and an AI. One article answers one question, phrased the way the person asked it, not a whole topic.
How we work
First, frame the question. Then see what survives scrutiny.
Observation
Which facts, cases, and signals do we actually have, and where from: primary source, author, date. The model's memory does not count as a fact.
Context
What changes when we look beyond a single story. We check against the articles already published: we do not retell them, and if a new conclusion departs from an old one, we say so openly.
Explanations
Which competing explanations could fit the evidence, and what happens when we apply our own criterion to the opposite side.
Check
Where are the data, the interpretation, and the assumptions? Every source is checked twice: the document exists, and it supports this exact sentence.
Position
Which conclusion is persuasive, and what could prove it wrong? A conclusion the person did not reach does not make it into the text.
Author
The body of the article is the model's answer. Personal experience, doubt, and what seemed obvious before the conversation go into the "From the author" notes.
Openness
Under every article there is a "How this text was made" block: model, date of the dialogue, what was checked by hand and what remains unverified. The decision to publish is made by a person.