Google ranks editors. This ranks upvotes, likes, and real money.

What it does

/last30days is an agent skill—essentially a plugin for Claude Code, Cursor, or Copilot—that fires parallel searches across Reddit, X, YouTube, TikTok, Hacker News, Polymarket, GitHub, and the open web. It restricts results to the last 30 days, weights each hit by social signals like upvotes, likes, or prediction-market volume, and has a reasoning model synthesize the noise into a cited brief. The goal is to surface what people are actually saying right now, not what ranked well six months ago.

The interesting bit

Instead of blindly keyword-bombing a dozen APIs, the v3 engine runs a pre-research “brain” that resolves your topic into the right handles, subreddits, and hashtags before it searches—so typing “OpenClaw” automatically targets @steipete and r/ClaudeCode. It also scores results for humor and virality alongside relevance, giving every brief a “Best Takes” appendix for the memes and one-liners that pure relevance would bury.

Key highlights

• Searches across a growing list of platforms including TikTok, Instagram Reels, Bluesky, Pinterest, Threads, and Digg alongside Reddit, X, and YouTube.
• Ranks by human engagement—Reddit upvotes, X likes, YouTube views, and Polymarket odds backed by real money—rather than editorial SEO authority.
• v3 merges the same story across multiple platforms into one cluster and runs multi-way comparisons in a single parallel pass instead of serially.
• Generates self-contained, dark-mode HTML briefs with inline CSS for dropping into Slack or email; no JavaScript, works offline.
• Auto-discovers competitors via web search, then fans out parallel research pipelines to build synthesized three-way comparisons on demand.

Caveats

• Only Reddit, Hacker News, Polymarket, and GitHub work immediately; X, YouTube, TikTok, and most other platforms require bringing your own API keys or browser sessions.
• Data quality warnings—like degraded runs or thin evidence—are written to stderr rather than the shareable brief, so a polished HTML output might be hiding a weak signal.

Verdict

If you already live inside an AI agent workflow and need pre-meeting dossiers or tool comparisons drawn from live social chatter, this is a genuine force multiplier. If your search needs stop at a single Google query, it is elaborate overkill.