Methodology

VeraVadis is a hybrid intelligence pipeline: automated ingestion and analysis, paired with editorial judgment. This page describes the real stages, in the order data moves through them today.

1. Sourcing

We pull from a curated inventory of wire services, national and regional outlets, defense and policy analysts, human-rights organizations, official/institutional sources and monitored YouTube channels, each tagged with a tier and role. See Open sources for the tier rules. No source is treated as automatically true; a tier is a starting weight, not a verdict.

2. Analysis

Every ingested item is scored for relevance, tagged with its source tier, and checked against a deterministic five-factor confidence synthesis (direct evidence, independent sources, bias risk, temporal coherence, external corroboration), computed from counted facts in our database, not generated by a language model.

3. Convergence

When the same event is reported by more than one independent source, VeraVadis links the reports into a tracked, falsifiable question or thread, and records how many independent outlets carried it.

4. Editorial judgment and independence

Publishing itself is automated by rules our editorial team sets and monitors -- no person clicks "approve" on each individual item before it goes live. The team is not funded by, and does not report to, any party covered in our reporting; it reviews the rules, the source inventory and the output in aggregate, and can correct or retract any item after the fact.

5. Illustration, with disclosure

Every hero image is an AI-generated editorial illustration, never a real photograph of the event. Every placement carries a visible disclosure of which of two kinds it is: an editorial illustration or a deterministic data card.