1. Plain English First: What This Thing Does

Give it a keyword. It sweeps X, Reddit, YouTube, TikTok, Hacker News, GitHub and more for the last 30 days of real discussion, scores everything by actual engagement (likes, comments, upvotes), and hands you a ranked briefing with links to the original posts. You take that briefing and use it to pick content topics, pick products, or sell it to clients as a research report. That’s the whole play.

Why do you need it? Google serves you articles from three months ago. Trending lists go cold in 24 hours. Pinterest and Lemon8 keep feeding you more of what you already saw. By the time you notice a term is exploding on X, Reddit has already argued about it for hundreds of comments and TikTok’s traffic wave is over. Every platform is a walled island too: Google doesn’t index Reddit comments, ChatGPT can’t see X or TikTok, Claude connects to zero native social sources. Doing it by hand means tab-hopping and copy-pasting until the moment has passed.

2. What last30days Actually Does

There’s an open-source project on GitHub called last30days-skill (repo: https://github.com/mvanhorn/last30days-skill, with 62,826 stars, 5,463 forks, MIT license). It does one thing with total focus: an AI agent scans a dozen-plus platforms in parallel, scores content from the last 30 days with real engagement data, dedupes, ranks, and compiles a briefing with citation links.

Its tagline is “searches people, not editors.” Translation: Google aggregates pages that editors and SEO teams filtered. This aggregates signals that real users voted on with their likes and their wallets. You give it a keyword, it tells you where that keyword caught fire in the last 30 days, why it took off, and who’s driving the conversation.

3. Installation: Let’s Be Honest About the Bar

There’s a real bar here. You need to be comfortable with a terminal-based AI client like Claude Code or Cursor. If you’ve never touched a terminal, stop reading here. Don’t let anyone tell you this takes five minutes for everyone.

If you qualify, the main path is one line. In Claude Code, run:

1
2

/plugin marketplace add mvanhorn/last30days-skill
/plugin install last30days

For other clients (Cursor, Codex, Copilot, Gemini CLI, etc.):

1

npx skills add mvanhorn/last30days-skill -g

-g installs it globally so every project can call it. Reddit, Hacker News, Polymarket, and GitHub work out of the box, with zero config and no API key. After one run of the setup wizard, it walks you through connecting X, YouTube, TikTok and other sources. Heads up: sources like X and TikTok require your own login credentials or API keys. It’s not “instant connect.” Check the official docs for the actual setup cost.

Once installed, verify with:

1

/last30days Cursor IDE

Within a few minutes you should get a briefing: what Cursor shipped recently, which Reddit thread has the most discussion, which developers are talking about it on X, and whether Hacker News is complaining. Every citation is linked and timestamped. (That’s the effect described in the official README; your actual output is what counts.)

Advanced (skippable): Want it running daily on autopilot? Set up a cron job on a VPS and push results to Slack or email. Don’t know VPS? Skip it. Manual runs are plenty.

4. The Main Play: Selling Research as an AI Service

This is the most concrete monetization path I’ve found. Full chain, step by step.

Step one: find clients. Search Upwork, Fiverr, or LinkedIn for “market research” and “competitor analysis” gigs, or post a “custom AI industry briefing” offer on Lemon8 or your Substack. The buyers are specific: small business owners who want into AI but don’t know which niche, marketers who need industry reports, and sales reps doing homework before a pitch.

Step two: run the keyword. Client says “I want to build a business around AI agents.” You run /last30days AI agent workflow and look at who’s repeatedly begging for solutions on Reddit, which niche gets the densest complaints on X, and which automation plays are pulling the most engagement on TikTok.

Step three: deliver. Three things in the briefing count as deliverables: platform source, original post links, and engagement data with timestamps. Organize those into a table sorted by engagement, and add your one-paragraph interpretation. That’s what the client pays for.

Step four: price it. Bill for your actual human hours, not “the plugin ran for two minutes.” A cleaned-up report with analysis takes 3-5 hours. Price against market rates for research services, roughly $70-450 per report depending on depth. Go cheap early to build case studies and reviews, then raise prices once you have samples. Orders won’t flood in day one. Client acquisition is slow work. I’m telling you straight.

Other ways to use it (two sentences each, no deep dive):

• Product research: Run /last30days desktop air purifier and read only the Reddit/X/TikTok user chatter: which model keeps getting named, and where the complaints cluster. Then search the same keywords on eBay or Mercari to check price bands and seller angles before you commit.
• Content topics: Run 10 candidate topics, rank by two metrics: engagement density (how fast likes and comments pile up) and platform spread (one platform arguing, or several at once). Write the top 3. Beats guessing.
• Developer angle: Polymarket odds and GitHub star velocity show where attention and bets are flowing on AI tools. But that’s a weather vane, not a sales forecast. Don’t pick products off it.

5. Why This Is Worth Doing Now

In the AI era, information decays by the day: Reddit posts sink in hours, takes on X get buried in a day, and TikTok videos finish their traffic run in three. By the time Google surfaces it, most of the money value is gone. last30days flips the window from “wait passively” to “grab actively.” You end up among the first people in your circle who know what’s happening right now.

It’s MIT-licensed, fully open source, and four core sources work with zero config. Tools like this used to cost a monthly subscription. Now it’s one plugin install away.

Open your terminal, run those two commands, and fire off /last30days your main niche first. Feel the gap between “30 days of real web-wide data” and “stale Google articles.” If you install it and feel nothing, come find me.

Let me double-check: