The header on the Founders Fund page says 63 portfolio companies and “63 publish APIs here.” The paragraph below it says 35 of the 63 (56%) publish APIs and 28 are profiles without one. The band list settles it: 5 exemplar, 7 strong, 10 developing, 8 thin, 3 emerging and 2 minimal add up to 35. The header is wrong: our layout prints the network count under a publishing label. That one is ours to fix.
The real number is still a strong one. The 35 publishers average 43.9 on the Kin Score and 28.7 on Agent Readiness, with 12 top-rated companies and 19 that are agent-ready or agent-native.
The top of the book
| Provider | Kin Score | Agent Readiness | Contracts on VC page | Contracts on provider page |
|---|---|---|---|---|
| Twilio | 86.0 exemplar | 47.1 agent-native | 55 | 153 |
| OpenAI | 82.6 exemplar | 58.0 agent-native | 16 | 63 |
| Stripe | 82.6 exemplar | 49.5 agent-native | 47 | 71 |
| Asana | 72.7 exemplar | 48.8 agent-ready | 2 | 40 |
| Figma | 71.4 exemplar | 53.7 agent-ready | 4 | 12 |
| Crusoe | 65.9 strong | 32.0 agent-ready | 2 | 64 |
| Scale AI | 64.4 strong | 44.4 agent-native | 9 | 94 |
The scores on the two pages agree, but the contract counts do not. The VC page counts “published contracts” and the provider page counts “API contracts indexed,” and the gap reaches 64 against 2 at Crusoe. Those may be two legitimate measures, but the labels do not tell a reader why they differ. Compare scores across this book, and do not compare API counts until we label them properly.
Two names, two other companies’ records
Two entries are not scored as themselves. Oculus resolves to the Meta record and shows its 56.9. The Athletic resolves to The New York Times Company and shows its 43.8. Both are real portfolio names, but the scores measure the parent’s developer surface, not the company Founders Fund backed. Oculus and The Athletic sit among the book’s 12 top-rated and 10 developing companies respectively, so a portfolio rollup is partly crediting work done elsewhere. That mapping is our decision, and it should be visible on the page.
Where the book gets interesting
The portfolio is a hard-tech and fintech list, and the scores split along that line. The top of the book is API-first software. The defense, government and aerospace names sit further down: Palantir at 36.7, Anduril at 31.9 and SpaceX at 31.6, all thin. That is what the public record shows, not what they run privately.
The sharper observation is about agents. Monaco has a 53.4 Kin Score and a 61.9 agent score, higher than Twilio, OpenAI or Stripe. Flexport has the widest gap in that direction, and they are the only two of the 35 that score higher on agents: 16.1 on the Kin Score, which is emerging, and 37.5 on agents, which is agent-ready. The two scores measure different things, and Flexport shows how far apart they can sit. A high agent score alone does not tell you the contract underneath is good.
Takeaway
Thirty-five publishers out of 63 is a healthy share for a book this weighted to hard tech, and its five leaders are all exemplars. The page needs three fixes from us: the header count, a label on the contract counts, and a visible note where a company is scored through its parent.
See the portfolio at apis.io/vcs/founders-fund/.