AI Can't See Reddit or HN. One Dev Built a Bridge.

ChatGPT searches the web. Claude browses documentation. Perplexity summarizes articles. But ask any of them about a framework's recent breaking change, and you'll get Stack Overflow summaries while missing the GitHub issue thread where maintainers explained their reasoning—or the subreddit where developers are comparing workarounds.

The problem isn't that AI assistants can't search. It's that technical knowledge is fragmented across platform-specific APIs, authentication systems, and walled gardens that prevent a single system from reaching Reddit threads, Hacker News discussions, YouTube comments, and GitHub issues at once. Each platform has its own access rules, rate limits, and authentication requirements. Some block automated scrapers entirely.

The authentication wall problem

Platform silos exist for reasons. APIs cost money to run. Rate limits prevent abuse. Authentication protects user privacy. But the result is that AI tools claiming to "search the web" can't actually see inside the places where developers share technical knowledge.

When you ask about a library's recent changes, current AI assistants hit these walls. They'll summarize the official documentation but miss the Hacker News thread where the maintainer answered questions, the Reddit discussion where users reported edge cases, or the YouTube tutorial where someone demonstrated the migration path.

An open-source skill with around 20,000 installs addresses this gap by treating social platforms as first-class research sources alongside traditional web search.

Where technical knowledge actually lives

The Last30Days skill searches Reddit, X, YouTube, TikTok, Hacker News, Polymarket, GitHub, and the web in parallel. Instead of treating these as occasional supplements to "real" web search, it resolves relevant handles, repositories, subreddits, hashtags, and channels before searching sources simultaneously, then scores results by engagement, relevance, and freshness before merging overlapping stories.

This approach treats platform fragmentation as an infrastructure problem rather than a feature gap in existing AI tools. The goal isn't replacing Perplexity or competing with ChatGPT's web browsing—it's building infrastructure that reaches into the corners they can't access.

Twenty thousand installs suggest developers recognize the gap. When you're researching a technical decision, you don't want summaries of official documentation. You want the GitHub issue where someone reported the bug, the Reddit thread where maintainers responded, and the Hacker News discussion where experienced developers weighed tradeoffs.

The Reddit access constraint

Even purpose-built tools hit platform restrictions. WebFetch refused Reddit URLs, causing Reddit body text and comments to be omitted and reducing retrieval quality for topics with substantial Reddit discussion. This isn't a fatal flaw—it's one example of the larger access problem the project is working through.

Reddit's API restrictions affect everyone trying to build research tools. The platform limits what automated systems can access, when they can access it, and how much data they can retrieve. Any tool claiming comprehensive social web access will bump into similar constraints across different platforms.

The project continues iterating around these limitations. Version 3.28.0 shipped on October 11, 2025, with commits continuing through October 18, including fixes for historical lookback, Hacker News citations, ranking improvements, YouTube date filtering, and duplicate source handling.

Open-source as foundation

The MIT license matters here. Solving platform access isn't something one tool can accomplish in isolation—it requires shared infrastructure that adapts as platforms change their access rules, rate limits, and authentication requirements.

Integration with vercel-labs/skills positions this as extensible infrastructure rather than a standalone product. Developers who've hit these walls while building research tools can contribute platform adapters, improve scoring algorithms, or add support for new sources.

The twenty thousand installs prove demand exists. Open-source enables the iteration needed to fill a gap that current AI assistants acknowledge but can't fully bridge: access to the fragmented social web where technical conversation actually happens.