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 Most Programmable Sectors In Australia," Kin Lane, October 26, 2026, https://apievangelist.com/2026/10/26/the-most-programmable-sectors-in-australia/
Priya read the post on Monday evening, on the train, with her laptop balanced on her knees. She was a developer advocate at a mid-sized payroll software company, and she had spent two years telling people that her employer's API was well documented. The first thing she read was the opening claim: "Everywhere else, it is dark."
She bristled at it. The numbers were hard to argue with: 227 companies profiled, and 117 of them, 52%, publishing nothing an agent could read. But she knew from experience that a company with nothing public often has plenty behind a login. Partner APIs, private portals and onboarding calls all exist, and a crawler can't see any of them. She wondered whether the post was measuring how programmable these companies were or only how visible they were. The author seemed to have anticipated that. He had asked what a company publishes that a developer or an agent can actually read, and he had drawn his line at public and machine-readable. Priya decided that was a fair line, even if it was a blunt one.
The mining section made her wince. Forty-one miners and one Swagger file. Then the passage on the six largest industry super funds: no contract, no portal, no sandbox, and, in the author's words, "Enormous consumer reach, zero programmable surface." She thought about her own mother's super account, the paper statements, and the phone queue. Nobody had to tell her that was dark.
What stopped her was a smaller line, further down. "A blocked front door is not a closed API." The post said 21 companies put an anti-bot challenge in front of their homepage, and that some of them still publish contracts behind it. An agent that gives up at the homepage never finds them. Priya felt a cold little drop in her stomach, because she did not actually know what her own company's homepage did to a visitor that wasn't a person. Marketing had added a bot-protection service last spring. She had never asked what it did to a script.
She admitted this to herself without enjoying it. For two years she had pointed people at the developer docs and never once looked at the door they would have to walk through to reach them. The post did not say her company was one of the dark ones, and it could not. But it made her aware that she had never tested it.
The part about the roundtable in Melbourne did not apply to her, since she was nowhere near it. She noticed that the post ended with a question about commerce rather than technology: what would make a miner, a super fund or a retailer want to be found by an agent? She had no answer to that for other industries. She had a feeling she might not have a good one for her own.
When the train reached her stop she closed the laptop. At home, after dinner, she opened a terminal and fetched her company's homepage with a bare command-line request, no browser and no cookies. She looked at the status code and the first few lines of what came back. Then she looked at whether anything answered at the llms.txt path on the main domain.
The homepage returned a challenge page instead of the site. She copied the output into a note and wrote an internal ticket for the marketing team asking what the bot protection did to automated visitors, and whether the docs subdomain was covered by it. She did not make the ticket sound urgent. She only wanted to know what an agent would see.