Tip: Track AI Coding Productivity Net of Rework — Merged PRs Alone Are Misleading

paddo.dev·high signal

The paddo.dev '43% Denominator' analysis reveals most teams measure AI productivity by visible outputs (merged PRs, tasks closed) while ignoring the denominator: post-merge bug-fix commits, average redeploy cycles per AI fix (Lightrun says 2-3 cycles minimum, with 11% needing 4-6), and production debugging rates. Track three specific metrics: bug-fix commits within 48 hours of merge, ratio of AI-authored PRs that spawn follow-up fixes, and time-to-stable for AI vs human PRs. Without the denominator, you're measuring velocity to nowhere.