Editorial Policy — How MYRIAD Content Is Made

MYRIAD is a data publication. Everything we publish begins with a number from a named primary source — the UN World Population Prospects, the World Bank, the Maddison Project, NCD-RisC, Barro-Lee, or Our World in Data — and every page tells you exactly which one.

How our pages are produced

Our data pages are generated from the datasets themselves: the tables, charts and rankings you see are direct renderings of the underlying series, compiled and normalized by our engine. The narrative text that accompanies them is drafted with the assistance of language models and then passed through a strict verification gate we built for exactly this purpose: any draft containing a number that does not appear verbatim in the page's own dataset is automatically rejected. No invented figures, no "approximately" from a model's memory, no unsourced claims. Drafts also carry a quality self-assessment, and anything below our threshold is discarded rather than published.

Flagship essays and all trust pages (including this one) are written and edited by a human. Where we estimate — pre-1950 historical values, or projections beyond the present — the number is visibly labeled estimate or projection on the page, and our methodology explains how it was derived.

What we will not do

We do not publish content about events we cannot source. We do not present projections as facts. We do not accept payment for coverage, and advertising on MYRIAD is always visually distinct from content. Ads never influence what the data says.

Corrections

When a reader or our own monitoring finds an error, we fix the page, purge caches so the fix is visible immediately, and keep an internal version history of every change (every page on this site is versioned). Material corrections are noted on the affected page. To report an error, use the contact page — data corrections are our highest-priority inbox.

Who is responsible

MYRIAD is owned and edited by its founding team; see the team page. The editor reviews flagship content before publication and audits samples of machine-assisted pages weekly.