Methodology

Formula (version v1, taxonomy tax-v1)

Each topic score combines three normalized signals: score = 0.45 × volume + 0.35 × velocity + 0.2 × engagement, over a trailing 30-day window of complete UTC days.

• Volume: raw thread count in the window, normalized against a cap of 100 threads.
• Velocity: threads in the latest 7 days versus the daily average of the prior 21 days, as a ratio capped at 4×.
• Engagement: mean of per-thread points + comments, each bounded at 200 points and 100 comments, normalized against 150.

Threads are collapsed to their parent story before scoring, so a story and its comments count once. Scores are comparable only within the same topic over time, not across topics.

Taxonomy principles

Topics are a small curated set (5–10 active at any time) with disjoint positive editorial terms and explicit negative exclusions. A topic is published only when every gate passes: at least 15 threads across at least 4 active days, fetch completeness of at least 95%, data fresher than 36 hours, an active topic, and HN-only, leakage-free source rows.

Freshness and retention

A snapshot is published only while request time is before its fresh-until time, which is derived from the completed source-data cutoff and fetch, never from generation time — regenerating old inputs cannot extend a snapshot's life. When no fresh snapshot is available the index answers 503 with a Retry-After instead of showing stale values. Transient HN source items are purged daily; only aggregate snapshots are retained.

Source and rights

All signals come from public Hacker News discussion, retrieved via HN Search by Algolia. YC Radar is not affiliated with Hacker News or Y Combinator. The index publishes aggregate measurements only; third-party discussion content is excluded. Representative citations link to public threads and are examples, not denominator data. Aggregate terms: this page.

Limitations

This index measures recent public Hacker News discussion signals; it is not a measure of market size, total demand, or purchase intent.

Hacker News discussion skews toward a technical early-adopter audience; volume, velocity, and engagement reflect that community only.