What We Learned From 12 Million Nostr Events

Patterns, anomalies, and surprises from indexing over 12M events across 50+ relays.

The Dataset

bigNOSTRals indexes the public relay network into an open BigQuery warehouse. Today that means a dataset of hundreds of millions of events across all major kinds — metadata (kind:0), text notes (kind:1), recommend servers (kind:2), contacts (kind:3), encrypted DMs (kind:4), and — of course — zaps (kind:9734/9735) — spanning 1.6M+ authors. Clean, documented, and queryable by anyone.

Key Patterns

Growth is organic but noisy. Event volume follows a weekly cycle with spikes during Bitcoin conferences and protocol upgrades. Bots account for roughly 15% of text notes but are easy to isolate via tag clustering.

Zaps are the most underexplored signal. Despite being a relatively new kind, zaps encode both attention and economic value. Their network graph reveals tight-knit creator communities across multiple relays.

Anomalies

We found relays that silently drop kind:3 contact lists (possibly to save storage) and relay pairs that share 90%+ of their event IDs despite claiming autonomy. The network is more interdependent than it looks.