How We Track Builder Activity
Transparent methodology. No black boxes. Every score is reproducible.
What We Track
Four primary signals, weighted by reliability:
- Public repo commits — raw commit count across all project repositories, deduplicated by author
- Deployment frequency — production deploys detected via Vercel, Cloudflare, and on-chain program upgrades
- Feature releases — net-new user-facing functionality confirmed live on production
- On-site verification — manual confirmation that shipped features function as described
$ devwatch scan --chain solana --project $HANK
scanning public repos...
found: 3 repos, 62 commits (30d), 1 contributor
deploy events: 28 (vercel), 0 (on-chain upgrade)
score: velocity=62 | deploys=28 | features=9
How We Score
Each project receives a composite velocity score from three weighted inputs:
velocity = commits_30d * 0.4
+ deploy_freq * 0.35
+ features_shipped * 0.25
adjusted = velocity / contributors
active = velocity > 30
slowing = velocity 10-30 OR month-over-month decline > 25%
dead = velocity < 10 AND no deploy in 30d
Rankings use raw velocity for ordering. The per-capita adjustment surfaces outliers like solo-dev projects that ship at team-level rates.
What We Don't Track
- Token price — irrelevant to builder activity. Never referenced in scoring.
- Social metrics — follower counts, Discord size, and tweet frequency are noise.
- Private repos — we only count what we can independently verify. If code isn't public, it doesn't exist in our system.
- TVL / trading volume — financial metrics tell you about users, not builders.
- Announcements — roadmap promises, "coming soon" posts, and partnership tweets receive zero weight.
The signal is code. Deployed, running, verifiable code.