Last30Days (mvanhorn/last30days-skill) is a multi-source research agent that has rocketed to 36,800+ stars with a +3,177/day velocity, holding the #1 spot on GitHub Trending as of June 9, 2026. It runs as a Claude Code skill that aggregates content from a rolling 30-day window across GitHub, Twitter/X, Reddit, Hacker News, arxiv, Polymarket, YouTube, and general web — then synthesizes the scattered signals into structured, grounded summaries. Rather than returning raw link lists, Last30Days cross-references signals across platforms: a trending GitHub repo discussed on HN with a critical Reddit thread gets surfaced as one coherent story. Built by mvanhorn, the skill requires no separate infrastructure — it installs as a standard SKILL.md file and runs inside any compatible agent harness.
Last30Days is the fastest-growing research agent on GitHub, hitting 36,800+ stars with a sustained velocity of +3,177 stars/day as of June 9, 2026. The premise is straightforward: instead of manually checking a dozen platforms to understand what happened recently in any domain, you invoke the skill and get a synthesized report that cross-references signals across sources.
The agent aggregates content from GitHub trending repos, Twitter/X discussions, Reddit threads, Hacker News front-page posts and comment threads, arxiv papers, Polymarket prediction markets, YouTube videos, and general web sources. The 30-day rolling window keeps results current without drowning in historical noise. The key differentiator is synthesis over aggregation — related signals across platforms are grouped into coherent narratives rather than presented as isolated link lists.
Multi-source pipeline: Each source has a dedicated fetcher that handles platform-specific APIs and rate limits. Results are normalized into a common schema before synthesis. Source selection is configurable per query — you can focus on just GitHub and HN for a technical topic, or cast wide across all platforms for broader trend detection.
Cross-platform deduplication: When the same story appears on HN, Reddit, and X simultaneously, Last30Days identifies the overlap and presents it as one item with cross-platform context — HN point count, Reddit upvotes, X engagement metrics — rather than three separate entries.
Polymarket integration: For topics with active prediction markets, the agent pulls current market prices as a quantitative sentiment signal. This is particularly useful for crypto, AI regulation, and tech industry developments where prediction markets provide a different signal than social discussion.
Grounded output: Every claim in the synthesized report links back to specific sources with timestamps. The skill flags when signals conflict across platforms — e.g., bullish sentiment on X but skeptical top comments on HN.
Weekly research briefs for engineering teams tracking a specific technology or competitive landscape. Trend detection for content teams identifying emerging tools, techniques, or conversations before they peak. Due diligence for investors and analysts who need a fast scan of recent discourse around a company or technology. Onboarding context for anyone joining a project or domain who needs to quickly understand the last month of developments.
The agent’s quality depends on platform API access and rate limits. X/Twitter data requires API credentials. Polymarket integration is most useful for topics with active prediction markets — niche technical topics may not have relevant markets. The 30-day window is a deliberate constraint; for longer historical analysis, other tools are better suited.
CLI orchestration tool that installs and configures platform connectors for 16+ social and web sources — Twitter/X, Reddit, YouTube, GitHub, Bilibili, and more. Zero API fees.
AI agents that autonomously conduct ML research experiments on single-GPU training, running a keep-or-revert loop overnight.
An AI-powered academic search engine that finds and synthesizes evidence-based answers from peer-reviewed scientific research.