◆ Dopamint Arena · contribution signals
Sprint Retrospective · Contribution Signals

Dopamint Arena — who built what, fairly measured

Every panel below is one facet of contribution, not a score. Leadership rotates from panel to panel — that rotation is the point. Read them together; no single number ranks a person.

Reproduce ↗ Every number here is regenerated from git + GitHub — run the exact commands yourself and re-verify.
i

How to read this. Data is filtered to real, authored code that shipped to main (= production): generated files, lockfiles, VM traces, vendored upstream and build output stripped; 19 git identities merged into 9 people; MystenLabs upstream (George, Ioannis) excluded. These are conversation starters for a blameless retro — not a public leaderboard.

Contribution facets

each panel sorted independently · bar = magnitude within that facet

What kind of work

share of each person's commits by type · conventional-commit prefix, all commits

Shapes of contribution

a fair one-line read per person — synthesised across all facets

Impact deep-dive

per person · what shipped, what they own, what they kept alive · Read = my interpretation

Method & honest limits

What "shipped / made it" means

All stats run over commits reachable from main (the production line — no release tags exist). Live code = git blame of the current tree with copy/move tracing (-w -M -C), so reverted/rewritten code drops out and moved code is credited to its original author. Origin-traced vs last-touch blame differed <2% per person — last-toucher bias was negligible.

Metric meanings

  • Live code — surviving authored lines in production now (impact footprint).
  • Lines authored — all lines ever added (credits origin, includes since-deleted).
  • Retention — live ÷ authored: how much of your code stuck.
  • Reviews given — distinct peers' PRs reviewed (self-reviews excluded).
  • Files created / subsystems — initiation & breadth.

What even this can't see

Line-volume ≠ impact: a 3-line mainnet-gas fix outweighs 2,000 lines of dead feature. Not captured: pairing, design, ops/incident response, mentoring, and the judgement in a good review. Treat low line-counts (esp. for refactor- and review-heavy people) as under-stated, never as "did little."

Window: 16 Jun – 4 Jul 2026 (~19 days) · Source: git history + GitHub PRs on CommandOSSLabs/dopamint-arena · Excluded: George & Ioannis (MystenLabs upstream). Use privately to prepare 1:1s and a blameless team discussion.