Can we actually filter out the AI noise on Hacker News?

The signal-to-noise ratio on Hacker News has taken a massive hit lately because every second post is some "revolutionary" AI wrapper or a generic LLM announcement. I found a tool called HN Without AI Stories that attempts to solve this by scrubbing the front page of anything tagged or categorized as AI-related, giving us back the classic software engineering and systems discussions that made the site great in the first place.

For those who aren't tired of the AI hype, this might seem counterintuitive, but for developers trying to find deep-dives into kernel optimizations or new database architectures without wading through ten "GPT-5 rumors" threads, this is a breath of fresh air. It basically acts as a specialized lens for the HN feed, prioritizing human-centric engineering over the current trend cycle.

If you want to set up a similar filter or build your own AI workflow for content curation, you can look into how these types of scrapers work. Usually, they hit the Algolia API for Hacker News and apply a negative filter on keywords like "AI," "LLM," "GPT," and "Machine Learning."

If you're interested in trying it out, the tool is live here:

https://hnnoai.readpipe.org/

From a prompt engineering perspective, the challenge with these filters is always the "false positives." For example, a post about a new C++ compiler might mention AI in the description, and a naive keyword filter would hide it. The real trick to making a "clean" feed is using an LLM agent to categorize the intent of the post rather than just searching for strings. However, for a quick and dirty way to clear the clutter, a hard keyword block is usually enough to get rid of 80% of the fluff.

I've been using it to keep my morning reading focused on actual deployment challenges and systems design. It's a reminder that while the AI boom is incredible, there's still a huge amount of non-AI innovation happening in the background that often gets buried by the algorithm. It transforms the experience from a hype-cycle feed back into a technical community.