Awesome HN: The sites that mine its comments for picks (techaddo.com)

A significant wave of innovation has emerged from the Hacker News community, with the launch of several platforms designed to aggregate and curate content based on discussions from the site. Initiated by Andrej Karpathy’s HackerTube—a YouTube-like site showcasing videos frequently recommended by commenters—this trend has rapidly expanded to include HackerBooks, HackerPapers, and other spin-offs, covering a broad range of media such as books, podcasts, and academic papers. These platforms utilize a public archive of Hacker News comments and advanced large language models (LLMs) to filter content, ensuring only items with genuine recommendations from multiple commenters are included. This development is noteworthy for the AI/ML community not only because it highlights the practical application of LLMs in content curation but also due to the challenges it reiterates regarding data freshness and community relevance. Despite positive initial feedback, the platforms face critiques over stale picks and matching errors. The technical processes behind these sites involve extracting links, labeling the sentiment of comments, and generating metadata from various sources, all structured to present a static site without user accounts. As more services emerge, the community's response and the evolution of these tools may reshape how content is discovered and consumed online, illustrating both the potential and the complications of automated content aggregation.