Hacker News is becoming an impossible noise floor because of AI

The signal-to-noise ratio on Hacker News has plummeted over the last few years, and it feels like the AI boom is the primary culprit. It used to be the gold standard for finding the "hidden gems" of the internet—deep technical dives, obscure whitepapers, and genuine engineering debates. Now, the front page is often cluttered with AI-generated summaries, low-effort "top 10 tools" lists, and marketing fluff that looks like a technical post but reads like a press release.

The volume of content is just too high to scroll through manually without missing the actual breakthroughs. If you aren't refreshing every ten minutes, the truly high-value threads get buried under a mountain of generic AI hype. It’s frustrating because the community is still there, but the discovery mechanism is breaking.

To fight this, I've had to change how I consume the feed. Instead of the main page, I've been relying on curated aggregators or specific search queries to find the "long-form" content that actually matters. I'm looking for specific markers of quality—like a high comment-to-upvote ratio—which usually indicates a real debate is happening rather than just a viral link.

For those trying to build a sustainable AI workflow for information gathering, I've found that setting up a custom LLM agent to filter RSS feeds can help, though it's ironic to use AI to filter out the noise created by AI. A basic setup involves pulling the HN API and running a prompt that asks the model to categorize posts by "technical depth" versus "marketing hype."

If you're trying to build a similar filter from scratch, you can use a simple Python script to hit the API and filter for specific keywords that usually signal high-quality engineering posts (like "implementation," "latency," "benchmark," or "kernel") while ignoring generic terms like "revolutionary" or "game-changing."

import requests

def get_top_stories():
    top_stories_url = "https://hacker-news.firebaseio.com/v0/topstories.json"
    stories = requests.get(top_stories_url).json()
    
    # Filter for high-signal keywords to avoid AI fluff
    high_signal_keywords = ['benchmark', 'implementation', 'deep dive', 'rfc', 'architecture']
    filtered_stories = []
    
    for story_id in stories[:50]: # Check top 50
        item = requests.get(f"https://hacker-news.firebaseio.com/v0/item/{story_id}.json").json()
        if any(word in item.get('title', '').lower() for word in high_signal_keywords):
            filtered_stories.append(item)
            
    return filtered_stories

The real challenge is that the "interesting stuff" is becoming a needle in a haystack. We're moving toward a world where the only way to find quality human insight is to actively filter out the synthetic noise. It makes me wonder if the era of the general-purpose aggregator is ending and we're moving back toward small, invite-only curated mailing lists.