HN Sieve automatically discovers, analyzes, and scores every Show HN project on Hacker News using AI. Each project is evaluated across four dimensions — Innovation, Technical Quality, Documentation, and Utility — on a 0-10 scale, then categorized and made searchable.
How it works #
- Collect — Polls the Hacker News API for new Show HN submissions every 2 minutes
- Fetch — Retrieves project content from URLs, GitHub repos, landing pages
- Classify — LLM-powered triage determines if the submission is a real project
- Score — A second LLM evaluates across 4 dimensions and generates a summary
- Index — Embeddings enable semantic search across all projects
Features #
- Browse curated projects with filters by category, score, and time period
- Semantic search — find projects by describing what you're looking for
- Analyze any URL — submit a project for instant AI evaluation
- Save projects to your personal collection
- Email digests with top new projects
- Full scoring breakdown: Innovation, Technical, Documentation, Utility
Categories #
Projects are classified into: devtools, ai-ml, web, infrastructure, security, data, mobile, developer-productivity, and more.
source & further reading
hnsieve.com — original article