Enriched from v1 (280,708 people, 0 relational edges).
What changed
| v1 | v2 | |
|---|---|---|
| people | 280,708 | 339,217 |
| artifact edges | 564,486 | 901,524 |
| person to person edges | 0 | 1,272,495 |
| organisations | 0 | 1,131 |
| external ids | 253,815 | 442,499 |
New in v2
A relational layer. 1,272,495 depends_on edges derived from 32.2M crates.io dependency rows (runtime only; dev and build dependencies excluded as a different claim). This answers who is load-bearing rather than who is famous: dtolnay is depended on by 45,037 distinct people, carllerche by 25,999.
A fifth domain. crates.io package maintainership: 337,038 edges, plus real install counts, a 92-day download trend, 720,772 attributed release-publishing acts and 23,288 yanks — the graph's first negative signal, since a yank is the author's own judgement that their work was broken.
Organisations as a first-class entity. 1,131 orgs with 13,533 org-to-artifact edges, so companies stop competing with humans in person-level rankings.
Admission on evidence of use. 58,509 people admitted because third-party download counts vouch for them, not because their repos cleared a 100-star floor. Among them Alice Ryhl (Tokio), Josh Triplett, and Ed Page (clap) — maintainers with billions of downloads who were previously invisible.
Wikidata identities. 3,342 new external ids: x_handle 1,663 to 2,730, plus 806 ORCIDs and 664 Google Scholar ids from zero.
Full method, receipts and negative results: docs/PEOPLE-GRAPH-ENRICHMENT-LOG.md rounds 14 to 17 in siso-foundry.