A family of independent, non-commercial YouTube archives — every full session of a public figure, podcast host, educator or homeschool channel on its own subdomain, six languages, discovered automatically and refreshed nightly, listed on a directory with a request board. Built from one generic template: a new archive is a persona file, not a codebase. Directory at cosmicflow.app.
YouTube is built to keep you watching, not to help you finish a body of work. Its feed surfaces the newest clip, the most-clicked short, the algorithmic neighbour — never “here is every full session this person has given, in order, with nothing missing.” For a podcast with hundreds of episodes, a teacher with a multi-year curriculum, or a public figure whose long-form talks are scattered across dozens of uploaders, the complete sequence simply does not exist as a page anywhere.
An archive fixes that with the dullest possible design: one site per person or channel, every full session, chronological, filterable, in the reader's language. No recommendations, no autoplay into something else, no ads. The videos stay on YouTube (embedded, never re-hosted); the archive is the missing table of contents.
Every archive is the same application. What differs is a persona file: who the archive is about, which channels and search terms find their sessions, which of the three modes applies, the accent colour, the six languages' worth of copy. Adding an archive means writing that file and pointing a subdomain at it — no new codebase, no fork, no copy-paste of last month's fixes.
That constraint pays off in both directions. A fix to the video grid, the language switcher or the structured data lands on every archive at the next deploy. And a request on the directory's board (“please archive this podcast”) is a small, bounded task rather than a new project.
Not every body of work is shaped the same way, so the template has three modes, chosen in the persona file.
For a public figure whose sessions come from many uploaders: interviews, speeches, panels, performances. Grouped by theme and year; the archive's job is completeness and de-duplication.
For a host with a numbered run of episodes: one canonical sequence, episode numbers preserved, guests and topics searchable, the whole run readable start to finish.
For an educator or homeschool channel: sessions ordered as a course, by subject and level, so a family can follow a year of lessons in the order they were meant to be taught.
Each archive keeps its own watermark — the newest session it already knows about. Every night it asks YouTube only for what is newer than that, per source, de-duplicates against what it has, and republishes if anything changed. Archives that gained nothing stay untouched; archives with a new session are current by morning. No archive depends on another's schedule, and a source that goes quiet costs nothing.
An archive that only a browser can read is invisible to the people looking for it. So every archive renders its lists — the sessions, the themes, the languages — as plain HTML on the server, before any script runs. Each page carries a canonical URL, hreflang links for its six languages, Open Graph and Twitter cards, and a JSON-LD graph that says which person or channel the page is about and which site it belongs to. The robots file names the search and answer-engine crawlers explicitly, a sitemap is generated from the data, and a plain-text summary of the site exists for language models. The directory does the same for the family as a whole.
Browse the family, or ask for one: cosmicflow.app. The first archive and the story of how it was built: the Sai Pallavi Archive.