OUR METHOD / V0.1

Evidence
before excitement.

We are a prototype AI-assisted science newsroom. Our goal is to explain discoveries without overstating what they establish.

A source is not proof

We distinguish measured results, author interpretations, vendor claims, editorial explanations and speculation. Each claim must link to an exact passage and a location in the original source.

Two domains, two evidence scales

Physics: P0 theoretical proposal; P1 preprint with data; P2 peer-reviewed result; P3 independent confirmation. AI: A0 announcement; A1 vendor report; A2 independent evaluation; A3 confirmed application. P− and A− indicate disputed results.

AI assistance and human review

Agents propose topics, research and draft. Production requires independent scientific verification, a language review for each Spanish or English revision, and human approval. An English research dossier is not automatically an English article. This preview uses fictional data and cannot publish.

Visuals and voice

Scientific visuals need clear per-asset rights and attribution. Generated illustrations are labelled as explanations, not observations. The initial video pilot uses Spanish synthetic narration with disclosure.

Corrections

Production will preserve immutable versions and a public correction history. Corrections must reach every published language and distribution channel.

Models and conflicts of interest

Gemini is planned for research and drafting. Claude and OpenAI will be compared before choosing an independent verifier. The same editorial standards apply when covering model providers.