At my engineering team, pull requests started getting longer and longer. The changes were harder to take in, and a careful review asked more of the person on the other side of the diff.
I mentioned it to a few friends and colleagues. They were seeing the same thing in their teams too. It was not a formal study, just a handful of conversations that made me realize this was not only happening around me.
The author still owns the change
AI may help write a change, but when something breaks, the person who opened the pull request still has the greatest responsibility for what they submitted. They cannot hand that responsibility to a model.
Reviewers have a responsibility too. I want to help keep the quality bar high. And when I approve a pull request, that approval should mean that I have looked at the change and believe it is ready. It should not become a routine click that says very little.
I want to stay in the loop
Longer pull requests make it harder to hold the whole change in your head, especially when time is limited. That is exactly when a reviewer needs useful context: what changed, where the risks may be, and how the pieces fit together.
I do not want AI to make the decision for me. I want it to help me read more thoroughly, so I can spend my attention on the parts that need judgment. I should be able to question a finding, dismiss it, or put it in my own words. My approval remains mine.
Our way to enhance code review
That is why we are building Rocky. It uses AI to help organize a pull request, surface areas that may deserve attention, and draft findings with context. A human reviewer checks those suggestions and decides what, if anything, gets shared.
The goal is a better first pass and a more informed human review. AI can do more of the reading. People stay responsible for the call.
If longer pull requests are making it harder for your team to review carefully, take a look at how Rocky works or explore the project on GitHub.