A short piece of fiction, written by API Evangelist. The reader is imagined; the post, and every number and quotation from it, is real: "Extending The Business Capability Model For The API Economy," Kin Lane, November 9, 2026, https://apievangelist.com/2026/11/09/extending-the-business-capability-model-for-the-api-economy/
Dara ran developer experience at a company that made a fitness-tracking app, and she opened the post expecting another enterprise taxonomy to be polite about. She had sat through enough of those. The first thing that caught her was the list of tags the old model had left out: e-commerce, consumer, MCP, marketplace, AI agents, gaming, agentic commerce. Her own category sat in that pile. Only 4,299 of 29,502 providers had mapped to any capability at all, which meant most of the catalog had been told there was no place for it.
She read the part about fitness tracking twice. The author said he had hit the wall with a food database, with fitness tracking, and with geocoding and maps, and called them some of the most common things developers go looking for an API to do. That matched her week. Someone had asked her last Thursday where to find a nutrition API, and she had answered from memory because no catalog had a sensible heading to look under.
The care of the thing won her over a little against her will. "Upstream is untouched." She liked that the author vendored someone else's work as published and kept his own IDs in their own namespace. She liked, too, the rule that "Nothing is there because it seemed like a good idea." Every capability had to cite tag counts or a gap found in a blueprint. In her experience taxonomies grew by enthusiasm, not by receipts.
Then came the part she distrusted. The classifier had been allowed to say "nothing fits," and after the extension, 19% of 895 turned-away tags found a home. Fine. But the author graded a hundred of those himself, and 96 were right. She did not think he was lying. She thought a person who built a model and then graded its output would, without meaning to, grade kindly. Four defensible-but-thin answers and zero wrong ones is a tidy result from a judge with a stake in it. The post did not hide this. It said the model was an argument, and it asked to be told where it was wrong. That was disarming, and it also felt like homework handed to the reader.
She noticed a smaller discomfort too. Every extension capability was marked for consumer, B2B, enterprise or agent use, and her company's API served the app and also the gyms that licensed their data. She could not tell from the post whether a capability could carry more than one of those marks, or whether the audience dimension would force her product into a single drawer.
She did not decide anything about the model as a whole. It was a big claim and she had read one blog post. But the post had given her something small to do. She opened the capabilities page, found the consumer health and fitness domain, and read what was filed there against what her company's API actually did. Most of it fit. One thing sat in a neighboring domain that she would have put elsewhere.
Then she wrote the note the post asked for. It was three sentences: which capability she thought was misplaced, why, and a question about whether audience could hold two values. She read it back once, cut an adjective, and sent it.