The dataset starts from 975 web pages saved as plain text: company career pages, individual job postings, job search result pages, and articles and discussions about remote technical work. They were collected through October 2026 with a focus on remote technology roles, and many search pages were scoped to a country or city, most often the United States, followed by Canada, Singapore, Ireland and the United Kingdom.
Composition by page type
Classifying the sources by page title gives an approximate breakdown: about a quarter company career pages, about a quarter single postings, about a fifth job search results, about a tenth articles and discussions, and the rest pages with little or no hiring content. Page titles are an imperfect guide, so treat these shares as rough.
Raw text, not structured feeds
Pages were saved as raw text, including navigation menus, cookie banners and footers. Nothing about the source format was trusted: the extraction step decided what counted as a position. This is why the pipeline can use career pages and search results that offer no structured data.
An earlier, smaller run
An earlier run in July 2026 processed 667 of these pages. Its results are used on this site only as a consistency check. See two runs compared.