Show HN: I made Google Trends for Hacker News by indexing 18 years of comments

Original Article ↗ Hacker News Discussion ↗

An independent project has indexed 18 years of Hacker News posts and comments to create a “Google Trends–style” tool that charts how often specific terms appear over time, revealing shifts in interest across programming languages, companies, technologies, and memes. Users praise its design and exploratory value, while frequently requesting normalization by total site activity, sentiment analysis, better handling of ambiguous terms, and filters for things like “Who is hiring?” or Show HN posts. Heavy traffic temporarily overwhelmed the backend, and some commenters also debate data licensing and how closely such a tool can or should mirror Google Trends versus n‑gram–style text analysis.

Overall reception

  • Many commenters find the tool “very cool,” useful, and fun to explore.
  • Several say they had been wanting exactly this for HN, or compare it favorably to Google Trends / Ngram–style tools.
  • A few find it underwhelming while search is disabled or data is incomplete and say they’ll revisit later.

Data sources & alternatives

  • The HN archive is described as highly valuable; links shared to the official Firebase API, a Hugging Face dataset, and a public ClickHouse database mirroring HN.
  • Some note the project effectively acts as a specialized search/insight engine over HN content.

Technical issues and performance

  • The site repeatedly suffers from “hug of death”: timeouts, 502/504 errors, and rate-limits from the backing Upstash Redis Search.
  • At times, data after 2019 disappears, “Who is Hiring?” graphs show no data, and certain queries fail or misbehave.

Feature requests & UX feedback

  • Common requests:
    • Normalize trends by total HN volume / comment count over time.
    • Full-height / fullscreen charts and consistent color mapping between thumbnails and main graphs.
    • Click-through from a spike to the corresponding HN front page date.
    • Filters for “Show HN,” “Who is Hiring?”, companies, or categories.
    • Better handling of multi-word queries and newly indexed words.
  • Some report broken scrolling and confusing or missing data for specific examples.

Ambiguity, search behavior & AI ideas

  • Ambiguous terms (e.g., “atom,” “go,” “grunt,” “Fastly,” “C#”) can mislead trends.
  • There’s debate over automatic synonym grouping vs. strict literal search; some want AI-powered categories, others strongly prefer explicit operators and opt-outs.
  • Suggestions include:
    • Sentiment analysis over people/companies and trends in positive vs. negative sentiment.
    • Embedding-based indexing to disambiguate topics.
    • Automatic discovery of emerging trends without specifying keywords.

Legal and licensing debate

  • A side discussion questions whether republishing HN data (e.g., via public datasets) is allowed under HN’s terms.
  • Commenters dispute whether the API / dataset licensing covers the content itself; outcome remains unresolved in the thread.