Lutz’s Law.

Less waiting, even in sequence.

One large pull request can take far longer to land than a series of smaller ones, even when the smaller changes add up to more code.

The first law is about whether a change lands. The second is about how long it takes to get there.

Every extra idea in a pull request gives a reviewer more to understand, more interactions to consider, and more reasons to ask for another revision. Smaller, coherent changes let the team make progress without first agreeing on everything.

These are rules of thumb born of open-source experience, not measured probabilities or guarantees. Some changes really do belong together. The point is to make the unit of review understandable.

Sometimes, the biggest PR gets the least review.

Inside a company, I've also seen the opposite pattern: the larger a PR gets, the less likely it is to receive a proper review, and the more likely it is to be rubber-stamped.

In that setting, one enormous PR can merge much faster than the same work split across several smaller ones. Each small PR gets real scrutiny; the giant one gets an approval.

That is not an argument for enormous PRs. It is a warning about review quality: a fast merge is not the same as a proper review.

Present-day commentary · Updated November 6, 2026. These reflections are separate from the original 2021 posts.

More review capacity. Still a need for judgment.

With AI doing much of the review work, the original laws apply less strongly. Large PRs can now receive detailed automated review rather than simply overwhelming the available human attention. Size alone becomes a less reliable predictor of whether a PR will merge or how long that will take.

That does not make every large PR a good idea. Reviewers may still push back, for good reason. Scope, architectural context, and whether the whole change can be understood and safely landed still matter.

Small enough to understand. Complete enough to merge.

Split by idea, not by file count. Each change should tell one reviewable story. A small PR that cannot be understood without three unmerged siblings has not necessarily made the review smaller.

Keep the tests with the behavior. Smaller changes still need to be correct, useful, and safe to merge. Moving their tests into a later PR is not the shortcut.

Find the piece that unblocks someone else. A self-contained foundation can land while the rest of the feature is still being worked out.