Amplitude tripled PR volume in six months by fixing CI, not agents

mobileraj · x · 2026-09-15

Amplitude tripled its PR volume in six months and cut PR cycle time from 5.2 hours to 44 minutes—mostly by fixing unglamorous infrastructure rather than improving agents. Key moves: swapping the JavaScript toolchain for native equivalents (tsc→tsgo, ESLint→oxlint, Prettier→oxfmt) plus affected-only Turborepo builds and parallel stages cut CI from 30 minutes to 3; a devbox-based custom CLI starts every local service in 28 seconds instead of 3 minutes; and a PR label triggers validation with zero local setup. The takeaway: when agents flood you with PRs, the bottleneck shifts to infra—speed up the human/AI verification loop, not just the model.

Original post →

More from coding & agent

coding & agent channel →