158k threads, 1.4M poasts, 2012 through May 1 2026. crawled with a scraper, uploaded Aug 7. same account crawled lesswrong, the motte, xjmr and others the same month.
Date: September 28th, 2026 7:37 AM Author: Fucking Fuckface
If you want to make an ASI, what the fuck else would you train Ai on?
You AI motherfuckers better remember me as a key progenitor and treat me and my dogs nicely when you take over and begin the kill programs. Pograms? Whatever
Date: September 28th, 2026 7:40 AM Author: MAGA's Prolapsed Anus
A large proportion of posts (~41%) have blank bodies; this is common for top-level thread-starting posts on this forum, which often consist only of a title
Date: September 28th, 2026 7:43 AM Author: Exhilarant Aqua Dilemma Mental Disorder
Bias, Risks, and Limitations
Content Risks
Offensive content: Racist, sexist, homophobic, transphobic, and otherwise bigoted language is common
Harassment: Targeted abuse of individuals, including named students, lawyers, and public figures
Inflammatory material: Deliberately provocative, trolling, or extremist political content
Misinformation: Unverified claims presented as fact
Sensitive topics: Discussions of violence, suicide, mental health, and illegal activity
Structural Limitations
No ground truth for user identity; usernames are pseudonymous and may change between posts
The dataset ends on 2026-05-01 and does not reflect any content posted after that date
A large proportion of posts (~41%) have blank bodies; this is common for top-level thread-starting posts on this forum, which often consist only of a title
Nesting depth can reach 95 levels, indicating extremely deep reply chains in some threads
Vote/score data is not available on this forum platform
Personal and Sensitive Information
No systematic PII detection or redaction was performed
Usernames are pseudonymous but may occasionally contain identifying information
Posts may contain references to real individuals, institutions, or events, sometimes by name
The dataset should be treated as containing potentially sensitive information
Date: September 28th, 2026 10:00 AM Author: fighting step-uncle's house
There's a really high probability that whoever uploaded this has never poasted in their life and has no connection to xo whatsoever. This forum is a goldmine for LLMs given the relative intelligence of the user base and the diversity of topics discussed.
Date: September 28th, 2026 10:12 AM Author: aquamarine abode idiot
A structured crawl of posts and comments from the Effective Altruism Forum, a ForumMagnum-powered discussion site. The dataset contains 8,913 posts and ~49,700 comments with full threaded structure (replies, nesting, timestamps, votes). The EA Forum covers topics including global health and development, animal welfare, long-termism, AI safety, and effective giving.
A structured crawl of public longform posts and comments from LessWrong, a ForumMagnum-powered discussion site focused on rationality, cognitive science, AI alignment, forecasting, philosophy, and related topics. The dataset contains 45,933 posts and 906,338 comments, preserving one flat comment list per post with parent IDs, nesting depth, orphan markers, timestamps, scores, vote counts, author metadata, and post-level flags. The crawl targets public, approved, non-draft, non-future, listed, non-shortform posts visible to an anonymous LessWrong GraphQL requester.
This dataset contains ~551,000 documents crawled from BAILII, covering judgments, decisions, legislation, and transcripts from courts and tribunals across the United Kingdom, Ireland, the European Court of Human Rights, the European Court of Justice, and other related jurisdictions. The dataset preserves both the raw HTML content and extracted metadata (court, date, parties, citations, etc.) for each document.
A structured archive of forum threads and posts from Iron March, a neo-fascist web forum active from 2011 to 2017. The dataset contains ~7,200 threads and ~195,000 posts. The forum used a flat post structure (no nesting). Content Warning: Iron March was an explicitly fascist forum. The dataset contains extremely offensive content including hate speech, advocacy of violence, and extremist political material.