- Catalog
- Data
On August 7 the notable-people backfill completed. It read Wikipedia's lists of notable people for the catalog's groups and saved 9,792 new people, taking the public groups with at least one notable person from 292 to 453. The phenotype catalog dataset shipped as v4 the same day: 23,034 notable people across 499 groups, up from 13,094.
On August 10 a second pass looked up each remaining person's Wikipedia page for a photograph. Photo links went from 5,883 rows to 15,228, about two thirds of the table.
Notable people give each group page real names, dates and faces. They are also the portrait source for the site's vision work: every per-image phenotype observation in the open dataset is a reading of one of these photographs.
Under the hood
- Source pages are each group's "List of X people" article. A title that resolves to anything other than a real list page, such as a redirect to the main article about the people, is skipped rather than harvested.
- Every entry passes a person-shape check: no event, place, ethnonym or year-leading names, and the list item must open with the linked person. The guards live in one shared module, so the scraper and the checks that replay them against stored rows apply identical rules.
- Descriptions are reduced to a short known_for value with life-date parentheticals stripped, and birth and death years are stored as their own fields.
- Photo enrichment is fill-in: it writes an image link where none exists and never overwrites one, so it reruns safely. Requests are paced per group and back off whenever Wikipedia asks them to.
- Each row carries ethnic_id, name, known_for, birth and death years, a reference URL and an image URL, and joins to the per-image observations through example_id.
- people added Aug 7
- 9,792
- notable people in v4
- 23,034
- rows with a photo link
- 15,228
Peoples in this entry
See it on the site
Sources and open data
Related build notes
- Aug 31, 2026A corpus of 28,756 facial and body observations over 39 peoples2,681 portraits read by a vision model against 22 controlled vocabularies now yield 28,756 structured observations that feed the trait pages.
- Aug 28, 2026Trait pages measure hair, eyes, skin and face from real portraits"Egyptian eyes" draws about 9,900 US searches a month. Trait pages answer questions like it with distributions measured from real portraits.
- Jul 26, 2026A new way through the catalog: browse by language familyEleven language-family hubs plus Bantu, Nilotic and Polynesian branch pages, and the open phenotype dataset rebuilt from 484 to 1,779 groups.