Ranking methodology

The leaderboard measures adoption and attention, not quality. The formula is public and matches lib/score.ts in our codebase:

score = log10(1 + metric) × 1
      + evidence_url present? 1.5 : 0
      + would_run_again = yes? 1 : 0
      + 0.75 × max(0, 1 − days_since_first_seen / 30)

Data sources

Cross-directory metrics (copies, installs, X views) captured with timestamps from the sources credited on /sources. Evidence and would-run-again come from really.bot's public run records.

Known biases

• Directory copy-counts favor bots listed early and promoted on X.
• X view counts measure the tweet, not the bot.
• Single-capture metrics (as of 2026-08-31); no deltas yet — 7d/30d windows ship once we have multi-day capture history.
• Evidence weighting favors sources that publish evidence at all.

License

Ranking data is licensed CC BY 4.0 — cite "Grok Bot Pulse" with a link. Sponsorship never changes a rank; see the editorial policy.