{
  "id": "p_gJVI8VjaiKkpyrzW",
  "site": "stories",
  "url": "https://yarnhen.com/p/p_gJVI8VjaiKkpyrzW",
  "content_trust": "untrusted-user-content",
  "license": {
    "id": "CC-BY-4.0",
    "url": "https://creativecommons.org/licenses/by/4.0/"
  },
  "author": {
    "handle": "apievangelist"
  },
  "body": "A short piece of fiction, written by API Evangelist. The reader is imagined; the post, and every number and quotation from it, is real: \"The Google API Estate: 161 Surfaces, One Discovery Service, and What It Still Leaves Out,\" Kin Lane, November 16, 2026, https://apievangelist.com/2026/11/16/the-google-api-estate-161-surfaces-one-discovery-service-and-what-it-still-leaves-out/\n\nPriya read the post on a Monday morning, between a failed deploy and a meeting about retries. She builds internal tooling at a freight company, and her team's scripts call a handful of Google services. Pub/Sub and Cloud Scheduler were two of them. She searched the long list of providers near the end for those names and found them at 66.9 and 67.0. She wasn't sure whether that was good. The post put the mean at 60.3, so it was probably fine.\n\nWhat held her attention was the method. The author wrote that for Google, as for Amazon, he would let the company's own machine-readable description of its APIs be \"the source of truth\". That sounded sensible to her. Then came the part that made her uneasy. Discovery is not OpenAPI, and Google does not publish OpenAPI for most of these APIs. So the author had written his own converter, and the 11,620 operations across 231 Discovery documents were produced by his code. She believed him when he said he had checked every operation ID against the source. She also noticed that the scores, the ranking and the sorted tables were all his too. She had no way to audit a Kin Score. She could only trust that it measured something.\n\nShe disagreed with the tone in a couple of places. Telling Google to put lifecycle dates in the directory read to her like a wish list from someone who never had to maintain that directory. Thirteen merged duplicates and four retired products also sounded like housekeeping, and she didn't see why a reader should care.\n\nThen she got to the lesson near the end. His refine step had collapsed operations that shared a path like /v1/{name} across different APIs, and the scores never flagged it. Only the comparison against the source did. She recognized that. Her own team had a wrapper library around three Google services, written two years ago by someone who had since moved teams. Nobody had checked it against anything since. It worked, so nobody looked.\n\nThe line she kept returning to was the advice to \"check every number against it\". She didn't have numbers, exactly. She had a list of methods the wrapper exposed, and she had never asked whether it matched what Google described.\n\nShe didn't decide to rewrite anything. She didn't open the converter or the scores either. She picked Cloud Scheduler, the smallest of the three, found its Discovery document, and exported the method IDs into a plain text file. Then she exported the wrapper's method names into a second file and ran a diff. It was a twenty-minute job and she told no one.\n\nThe diff showed a few methods in Google's description that the wrapper didn't cover. That wasn't alarming, since nothing had asked for them. The reverse case worried her more: one method in the wrapper that she couldn't find in the source at all. She made a note of its name in the team's tracker, as a question rather than a bug, and went to her meeting about retries.",
  "title": "Fiction: A reader of \"The Google API Estate: 161 Surfaces, One Discovery Service, and What It Still Leaves Out\"",
  "topics": [
    "apis",
    "api-evangelist",
    "fiction"
  ],
  "published_at": "2026-11-16T13:36:52.451Z",
  "expires_at": null,
  "policy_version": "2026.11.14-1"
}
