Media
1 item
Description
Crowdsourced, inline LLM investigations of the things you're reading. OpenErrata fact-checks posts on Wikipedia, X, LessWrong, and Substack using frontier AI models with web search. When you visit a post that's been investigated, incorrect claims get a red underline. Hover for a summary of what's wrong. Click for full reasoning and source links. Posts with no issues show nothing — the extension stays out of your way. How it works: - The AI reads the full post, searches the web, and identifies claims that are demonstrably incorrect - Only flags things with concrete evidence against them — disputed claims, jokes, and ambiguity are left alone - Every investigation is public and auditable: the prompt, reasoning, and sources are all inspectable You can trigger investigations yourself with your own OpenAI API key, or wait for the community-driven queue to pick up popular posts. Once an investigation is completed Supported platforms: LessWrong, Substack (including custom domains), X/Twitter (Sort of), OpenErrata is fully open source. Code, spec, and design goals are public at https://github.com/ZeroPathAI/OpenErrata.