Skip to main content
PalatineAI monument — a Roman temple housing a research terminal

​
The funnel

StageCountWhat happens
Researched2,857Every Colosseum submission is cataloged: product, team, links, category, country.
Scored1,680Full 12-signal scoring (automated research + manual desk review).
Manual-review queue1,177Researched but not yet fully scored. Exists only in the full database — never in the feed.
Published120The top of the scored ranking, exported to projects.json.
Two consequences for consumers:
  • A project missing from the feed is not a negative verdict — it may be unscored (1,177 are) or simply ranked below 120.
  • The published floor is currently total = 74.75. Everything in the feed is already top ~7% of scored projects.

​
Scoring

Each scored project gets the 12 signals (0–10 each, summed to total, max 120). Scoring blends automated research (GitHub, X, site analysis) with manual desk review; additional is an explicitly manual bonus. Scores are point-in-time reads, refreshed when the desk re-runs — they are not live metrics.

​
Coverage

DimensionFull databasePublished feed
Categories3322
Countries10835
No-token projects1,599117
Largest categories in the feed today: Consumer Apps (24), DeFi (20), AI Platforms / Agents (14), Developer Infrastructure (7), Marketplace Platforms (6).

​
Update cadence

Updates are manual desk refreshes, not a schedule. Detect them by:
  1. stats.updated inside the JSON (currently 2026-06-11), or
  2. Last-Modified / ETag response headers (conditional fetch).

​
Limitations

  • Scores encode the desk’s judgment; two reasonable evaluators could differ on any single signal.
  • noToken and funding are near-saturated within the published feed — they select the cohort, they don’t rank within it.
  • Team-submitted fields (description, website, team) are taken at face value; verify before acting.
  • Not financial advice. The feed finds candidates; your workflow makes decisions.