Methodology

How we verify and publish

Every public claim should carry source provenance, time context, and a confidence decision. Facts, themes, and analysis stay labeled separately. Sources are provenance — not the marketing pitch.

Pipeline

Collect permitted signals → extract facts/themes → normalize → verify → detect patterns → generate original opportunity cards & analysis → human publish.

Verification rules

  • Every claim keeps source URL, published date, and collected date.
  • Non-trivial claims need two sources or one primary source.
  • Primary documents outrank secondary summaries; outlet themes are synthesized, never republished as our story.
  • Data pages and cards carry an “as of” date.

Confidence tiers

  • High — eligible after light human edit.
  • Medium — review queue required.
  • Low — stored only; never published on the public site.

Politics, sanctions, crime, legal interpretation, investment advice, breaking accusations, and Turkmenistan country narratives always require human judgment regardless of score.

AI disclosure

We use AI (including Gemini Flash) to help extract structured facts and draft text. A human reviews and publishes. Creative suggestions for other products are labeled separately and are not reported news.

Corrections

Material errors get a visible update or corrections note. Contact [email protected].