- AI
- Data
- Catalog
On August 28 a first cohort of 30 peoples gained trait pages for hair, eyes and skin, plus a features overview. Each page is a set of measured distributions, one per dimension, with sample sizes and every value linked to its definition in the atlas vocabulary. The numbers come from photographs of notable people, read by a vision model.
The same day the pages grew to six families, adding nose, lips and face, and began reading observations coded directly in the site's controlled vocabularies. The first portraits were rendered from the measurements: on the Turks eyes page the most common measured values compose the image, a brown iris, an almond shape, a double eyelid crease and no epicanthic fold. Each render is captioned as an illustration of the measured distribution, and country pages link the trait pages of the listed peoples that have them.
Under the hood
- One resolver decides whether a trait page exists. A family page appears when a group has enough portraits with a readable value on that family's primary dimension, and the overview appears when two or more families exist. The route, the group page links and the country page links all ask the same resolver, so they cannot disagree.
- Values are vocabulary ids rather than prose. Labels and definitions come from the vocabulary files, each value links to its anchor in the atlas, and every page shows three catalog portraits cropped to the trait.
- Two observation sources sit behind one page contract. Every family reads the vocabulary-coded observations first; hair, eyes and skin also draw on an earlier structured analysis wherever the coded readings are thin, while nose, lips and face read the vocabulary-coded observations directly.
- Sample renders are composed from a family's modal values and stored in their own image space, touching no catalog record.
- Country composition rows resolve to their canonical twin before linking, so a country that lists a singular name still reaches the plural group that carries the observations.
- peoples in the first cohort
- 30
- trait families
- 6
- catalog portraits cropped to the trait, per page
- 3
See it on the site
Related build notes
- Aug 7, 2026The catalog's notable people grow from 13,094 to 23,034A Wikipedia backfill adds 9,792 notable people and lifts covered groups from 292 to 453. Three days later, 15,228 rows carry a photo link.
- Sep 6, 2026The authoring engine: 10,837 entities, one rubric, 253 new pagesA Wikidata harvest, a written rubric and a four-step pipeline published 253 new peoples pages in one day, from the Mosuo to the Hidatsa to the Tzotzil.
- Aug 31, 2026Measuring the ruler: 73 of 84 dimensions pass a test-retest studyA test-retest study over 196 portraits scored the head and neck dimensions, wrote the results into the vocabulary files, and made them citable.