What % of HN is now AI-generated?

Estimating the percentage of AI-generated content on Hacker News (HN) is a challenging task due to the lack of definitive detection methods and the evolving nature of AI tools. However, we can analyze the landscape and community dynamics to provide an informed estimate.

Key Factors and Observations

1. ·Community Moderation: HN operates on a rigorous, community-driven moderation system. Users are highly sensitive to content that exhibits the telltale signs of AI generation, such as generic summaries or overly polished language. This often leads to rapid downvoting, making it unlikely for purely AI-generated content to thrive prominently.
2. ·AI Usage Spectrum: Distinguishing between purely AI-generated and AI-assisted content is crucial. While purely AI-generated posts are rare and often caught by the community, AI-assisted content—where users draft posts and use AI for grammar checks or clarity—is likely more common. This type of content is nearly impossible to detect.
3. ·Spam and AI Wrappers: There has been a noticeable increase in spam posts, particularly 'Show HN' submissions promoting AI tools. These are often flagged quickly by moderators. Additionally, AI-generated summaries of long threads are consumed externally as utilities rather than contributions to the main discussion.

Estimating the Percentage

Given these factors, we can estimate the percentage of AI-generated content on HN.

• ·Visible Content (Front Page): Less than 5% of visible content is purely AI-generated due to the community's intolerance for genericism.
• ·Total Volume (Including flagged/sunk content): This figure could range between 10% to 20%, accounting for low-effort AI tools, spam bots, and users experimenting with LLMs.

Conclusion

HN is not overwhelmingly dominated by AI-generated content, but it is experiencing a 'signal-to-noise' degradation. The challenge lies in distinguishing between human and AI-assisted content, as the latter is increasingly blurring the lines. A more accurate estimate would require detailed content analysis, community surveys, and insights from moderators.