Hacker News Sans AI: Filtering AI Content from Hacker News

The Rise of AI Fatigue on Hacker News

Developers and technical enthusiasts are increasingly seeking ways to filter AI-related content from Hacker News due to a perceived saturation of AI topics, often referred to as "AI fatigue." The project "Hacker News, Sans AI" aims to provide a curated experience that removes AI-centric stories and discussions, allowing users to return to a broader range of technical topics.

Community-Driven Filtering Solutions

Several developers have implemented various methods to scrub AI content from the Hacker News feed, ranging from simple keyword filters to more complex API-driven tools.

Existing Filter Implementations

  • Hcker.news: A robust filter that offers both inclusive and exclusive AI filtering, moving beyond simple keyword matching.
  • Simon Willison's Filter: A static HTML version running against the Hacker News API, hosted on GitHub Pages to eliminate the need for a dedicated server.
  • RSS Feed Filtering: Some users employ feed readers like FreshRSS to create custom bookmarks that isolate "AI nonsense" across multiple sources, allowing them to bulk-mark AI-adjacent terms as read.

AI-Powered Filtering Paradox

There is a recurring debate within the community regarding the irony of using Large Language Models (LLMs) to categorize and filter out AI content. Some users have noted that the "Hacker News, Sans AI" project may use AI to solve the very problem of AI saturation, leading to questions about the purity of the filtering process.

Technical Challenges and Limitations

Filtering AI content is technically difficult because AI is often integrated into projects that are not primarily "about" AI. This leads to several common failure points in current filtering tools:

  • Contextual Leakage: Filters often miss stories where AI is mentioned in the comments or as a secondary tool used to build the project (e.g., projects built with Claude) rather than being the primary subject.
  • Keyword Limitations: Simple keyword filters may miss nuanced mentions of "agents" or specific model names, while over-aggressive filters may hide relevant non-AI technical content.
  • Performance Issues: Users have reported slower load times for some filtered proxies compared to the native Hacker News experience.

Psychological Impact of Content Framing

Beyond simple removal, some developers are experimenting with how the presentation of AI content affects user engagement. One developer noted that using AI to tag stories with labels such as "lacking evidence," "unsupported by research," or "personal anecdotes only" reduces the emotional impulse to engage or argue with the content.

"A comment that might've made you feel the need to reply before has less emotional weight if it's highlighted in red and a diminished font... It drives down the feeling of needing to engage. Which is horrible for site engagement, but good for mental health."

User Requests for Expanded Filtering

The demand for filtered views extends beyond AI. Users have expressed a desire for additional filters to remove "clickbait" and subscription-only paywalls from sites like The Economist or The New York Times to improve the overall quality of the browsing experience.

Sources