World Discovery Engine
Provenance before publication

Sources, provenance & machine access

Every preferred discovery must trace to an official or explicitly licensed dataset and a reproducible transformation. REAL revision evidence and CURRENT_VERIFIED indicator observations are exposed in both human-readable and machine-readable form.

Current production status: verified REAL population-revision evidence is published from compatible WDI archive vintages, and the Internet Use vertical publishes CURRENT_VERIFIED same-year observations for IT.NET.USER.ZS. Demo fixtures are not preferred discovery evidence. Real-GDP revision publishing remains blocked where release-specific methodology comparability is not established.

Primary source: World Development Indicators

The World Bank's World Development Indicators (WDI) is the first production source for the engine.

OFFICIAL

Current WDI dataset

The World Bank Data Catalog is the authoritative metadata surface for WDI and its licensing information.

Open World Bank catalog →
ARCHIVE

WDI Database Archives

World Bank DataBank exposes historical WDI releases used to reconstruct what a series said at different publication dates.

Open WDI archives →
LICENSE

Reuse & attribution

Published evidence retains source attribution and licensing metadata in the human page and machine-readable evidence payloads.

View WDI licensing metadata →

Machine-readable access

Agents, search systems and researchers can use the same verified records that power the human pages. Preferred discovery excludes DEMO and incomplete evidence.

REAL EVIDENCE

Revision evidence index

Collection-level index of provenance-gated REAL revision evidence with stable human, JSON and CSV endpoints.

Open evidence index JSON →
AI DISCOVERY

AI discovery manifest

Structured manifest joining preferred REAL revision evidence with CURRENT_VERIFIED Internet Use observations and their citation-ready endpoints.

Open AI index JSON →
Open llms.txt →

Why archive comparison is fail-closed

Matching series codes alone do not prove that two historical releases are methodologically comparable. Base years, valuation conventions, definitions and other metadata can change across releases. A revision is published only when the specific pair passes comparability checks.

GateRequired before a revision is published
Dataset identitySame official source family and known release/vintage
Series semanticsIndicator definition, unit and relevant valuation/base-year semantics compatible
GeographySame country/entity identity, not merely a matching display label
Observation periodSame year/period and frequency
ProvenanceSource URL, vintage date and import manifest retained
TransformationCalculation reproducible from preserved raw/normalized values

Evidence promotion ladder

1

Imported

A release or current observation is ingested with a stable official source reference and retained provenance.

2

Validated

Revision products pass pair-specific comparability checks; current-indicator products pass same-year, entity, unit and source checks.

3

Published

The discovery receives a stable human page plus JSON/CSV and machine-discovery metadata. Anything incomplete remains blocked from preferred discovery.