Dozens of small PRs or one big one? Test suite runtime is the crux for AI-generated optimizations
ducktapeprogra1 · x · 2026-10-30
On PR granularity for AI-assisted optimizations: is it always better to drop dozens of small PRs versus a collection of improvements that can be mined? The challenge is the test suite takes hours to run — should it run for every small PR? This question set up the follow-up "stacked logical commits" practice in the same thread.