Documents

Every message and article we collected, cleaned: deduplicated, language-detected, audio and video transcribed; deletions and edits recorded; nothing rewritten. Text is shown in its original language. Web articles appear as an excerpt only (the scraper does not keep titles yet). The AI labels are not on this page; they are under Methods.

Message texts are not published yet: this table shows each message's metadata and a link to the original post. See the changes log.

Download the whole layer

Monthly files (JSON Lines, gzip), the source roster, the source graph and the weekly counts. The first line of every file is a header with the field list, the version, the licence and the citation. Each file is described on its dataset page: Telegram messages, web articles, all datasets.