The problem most builders face
When you’re building with AI tools, things move extremely fast. New prompting techniques, better models, and useful tricks appear every week on Reddit, X, YouTube, Hacker News, and other platforms. By the time you’ve checked a few places, the conversation has already moved on, and you risk using outdated advice.
What last30days actually does
last30days is a Claude Code skill that automatically researches any topic across the last 30 days. It searches Reddit, X (Twitter), Bluesky, YouTube, TikTok, Instagram, Hacker News, Polymarket, and the open web. Then it scores the results, removes duplicates, and creates a concise, well-organized briefing with real citations.
Instead of spending hours scrolling, you get a clean summary that shows what the community is actually talking about right now — including what people are upvoting, betting on, and sharing in videos.
How it works under the hood
It uses a two-phase search process. First, it does a broad search across all sources. Then it looks for important details like usernames, subreddit names, or market tags and runs extra targeted searches to fill in the gaps.
The results are scored based on factors like recency, engagement (likes, upvotes, comments), and how much different platforms agree on the same idea. This helps surface the most relevant and trustworthy information.
What’s new in the latest version (v2.9.5)
The recent update added support for Bluesky (you can opt-in with your Bluesky handle and app password). It also introduced a “comparative mode” — you can ask something like “Cursor vs Windsurf” and get a side-by-side comparison.
Other improvements include better Reddit searching through ScrapeCreators (one API key now covers Reddit, TikTok, and Instagram) and the ability to store configuration files next to your project for easier setup.
Practical examples for builders
Let’s say you’re a vibe coder working on image generation prompts. You type `/last30days best Midjourney prompts March 2026` and get a summary of what people are successfully using right now, including real examples that are getting good results.
Or if you’re choosing between AI coding tools, you can ask `/last30days Cursor vs Windsurf` and receive a clear comparison table based on recent community feedback instead of old reviews.
Trade-offs you should know
Because it checks so many sources and does careful scoring, a full research run usually takes 2 to 8 minutes depending on how popular or niche the topic is. There’s a `–quick` mode that makes it faster by checking fewer sources, but you get less depth and fewer cross-platform insights.
For quick questions, the fast mode works well. For important decisions (like choosing a new tool or updating your prompts), the full mode gives much better context.
Who it’s good for
If you’re building AI-powered products, maintaining prompt libraries, or just want to stay current without burning hours on social media, last30days can save a lot of time. It’s especially useful for vibe coders who want real-world usage patterns rather than marketing claims.
Technical users appreciate the scoring system and entity-based searches that find relevant posts even when the exact keywords aren’t used.
Official resources
- GitHub Repository: mvanhorn/last30days-skill
- Installation Guide: Available in the repo README (works with Claude Code / Cursor)
- Latest Updates: Check the releases tab on GitHub for v2.9.5 changes
What you should do next
Try installing the skill in your Claude Code or Cursor setup (it’s open source and free to start). Pick one topic you’re currently working on or curious about, run a full research query, and see how useful the summary feels. If you like the results, add it to your regular workflow — many builders now run it before starting new prompts or choosing tools. Start small, experiment with both normal and quick modes, and decide what works best for your building style.