last30days

last30days is an open-source research Skill by Matt Van Horn. He says he built it to keep up with an AI scene that changes every day, and now also uses it before meetings, sales calls and trips to see what people have been saying over the past month. Type /last30days with a topic and its Python engine searches Reddit, Hacker News, Polymarket, GitHub and other sources in parallel; once configured it can add X, YouTube, TikTok, arXiv, Techmeme and more. It scores results by real engagement such as upvotes, comments and prediction-market bets, and the AI you use writes them up as one sourced brief. What sets it apart is that ranking follows how real people respond, with search-engine pages counted as just one signal among many. The README says Reddit, Hacker News, Polymarket and GitHub work right after install; the other sources need one run of the setup wizard.

Who it's for:people choosing topics, comparing products or preparing for meetings who want to know what's been said about something in the past month.

Before you start: in Claude Code, install it from the plugin marketplace; for Codex, Cursor and others, use npx skills add, a third-party installer that downloads files. You need Python 3.12 or later, and first-time setup may download a Python runtime and several command-line tools on its own. Your search terms go to each platform and search service, and briefs are saved in a Documents/Last30Days folder on your computer. We recommend only cookie-free searches of public information and local briefs. We don't recommend these setup options: reading browser cookies or the system keychain to sign in to X and other platforms, connecting a logged-in Xiaohongshu browser plugin, piping a remote script from curl straight into bash (the optional Grok CLI), sending watchlist results to an outside URL, or publishing reports online. Some platform APIs and paid search services charge by usage; prices in the README are the author's, so check each service's official site. MIT license.

Based on the project's public documentation. Not tested by AI Inspiration World.