Extension flags AI-generated posts on Hacker News front page with precise AI detection methods
When your feed explodes with AI-generated listicles, a Firefox extension now filters them out. The rise of large language models has flooded platforms with shallow, SEO-driven content that mimics expertise without real depth. On Hacker News, this noise skews the front page, drowning real discussions beneath fluff. A new tool uses Pangram scores to identify LLM-generated posts before they clutter your view.
The detection method avoids false positives by focusing on actual AI writing rather than discussions about AI. This precision helps tech enthusiasts avoid irrelevant, bot-generated summaries while keeping breakthroughs in machine learning and neural networks accessible. The system examines linguistic patterns and statistical anomalies that LLMs produce, distinguishing human-authored technical analysis from generic summaries.
Pangram scores serve as the core detection tool, measuring text to estimate AI generation likelihood. LLMs exhibit consistent word choices and sentence structures, leaving detectable patterns. When the extension scans Hacker News posts, it applies these scores to headlines and linked articles.
Users can adjust filtering settings to refine their feed. The extension lets you set a probability threshold—posts exceeding it automatically disappear. Instead of manually filtering robotic-sounding lists, the tool handles curation automatically. If a post’s AI score surpasses your limit, it vanishes, preserving only human-authored content. This customization ensures your feed reflects your tolerance for AI-generated material, whether you prefer visible flags or complete exclusion.
The Firefox extension integrates a dedicated API, making the workflow seamless. No more guessing whether a post is valuable or a traffic-driven prompt. Probability-based hiding lets you control your feed’s AI exposure, offering strict or lenient settings to maintain focus on meaningful discussions. The tool’s simplicity ensures developers and readers can effortlessly filter out unwanted content.
The system’s integration means no extra steps—just apply the extension, configure thresholds, and let it manage the noise. This approach keeps your Hacker News experience clean while preserving the platform’s core value: in-depth, human-curated insights.
All Replies （4）
Want a live back-and-forth? Join the global AI chat room — login to talk.
I want to try this tonight. I've been manually filtering for "delve" and "tapestry" using a custom regex script. I'm curious to see how the new Firefox extension handles this, especially since it employs Pangram scores for spotting LLM-generated content - I'll definitely check if the tool's detection method can pick up on the linguistic patterns and statistical cues that my custom script has been trained to identify.
Curious if this handles mixed-content posts—like a human-authored intro with AI-generated sections—since the tool scans both headlines and linked articles for Pangram scores. Does it trigger on the whole body or just specific sections using the 0.1 threshold? That precision would be key for catching subtle AI fluff while preserving legitimate technical discussions.
Finally! I'm sick of those generic lists. I tried a similar script with GPT‑Zero and it missed half the spam. The new Firefox extension now scans Hacker News headlines and linked articles, applying Pangram scores to spot AI‑written listicles—something the old tools just missed.
So glad this works. I want to try this tonight to see if it beats the 40% failure rate. For those struggling with the deluge of AI-generated content, using the Firefox extension to actively filter out articles written by LLMs based on their Pangram scores is a practical step to restore your reading flow. The surge of Large Language Models has spawned a flood of generic, SEO‑driven pieces that mimic helpful guides yet lack genuine human insight, especially visible on Hacker News. The tool’s key strength lies in its precision; it filters articles written by AI, not just about AI, using Pangram scores to differentiate human-discussed technical topics from bot-produced generic summaries.