mvanhorn/last30days-skill is a Python AI agent skill for researching recent topics across social, video, market, and web sources, then turning those signals into a grounded summary ranked by real engagement.
This repository provides an AI agent skill called /last30days that helps answer questions about a topic or person using recent signals from sources such as Reddit, X, YouTube, Hacker News, Polymarket, GitHub, and the web. The README says the current documentation refers to a v3 pipeline, while the runtime behavior is defined in the linked skill spec. It is positioned as a search-and-synthesis tool rather than a conventional search engine.
The project addresses the gap between fast-moving public discussions and slower, editor-curated or SEO-shaped search results. Its aim is to help users see what people are actually saying and engaging with over the last 30 days, especially when standard search or general-purpose AI tools miss platform-specific context.
Conceptually, the skill gathers recent information from multiple platforms in parallel, then uses engagement signals such as upvotes, likes, comments, views, and Polymarket odds to weight what seems most relevant. It then asks an AI judge to synthesize the collected evidence into a brief, grounded summary. The README also indicates that some sources work immediately, while others are unlocked through a setup wizard or additional keys and browser sessions; beyond that, the exact implementation details are not stated here.
It is gaining attention because the README frames it around a very current need: tracking rapidly changing AI and internet discourse from the places where those conversations happen first. The repository also shows strong momentum in the metadata, with a large star count and a significant weekly increase, which suggests that many users are finding the workflow timely and useful. Its appeal is amplified by the promise of combining several high-signal sources into one grounded answer.
The README explicitly contrasts it with Google-style search, general-purpose chatbots such as ChatGPT, Gemini, and Claude, and editor-curated discovery. It also implies overlap with platform-native search on Reddit, X, YouTube, or GitHub, as well as broader web search and research tools. Based on the README, the differentiator is not a single source, but the ability to bridge many sources and rank them by real engagement.
AI-explained · grounded in each repo's README