Robert Freeman explains why pull request review times are getting longer even as overall development cycles speed up. AI-assisted coding generates more code faster, but teams haven't upgraded their code-review tools to keep pace, creating a bottleneck. Engineers also hit a learning curve with prompt engineering before they see real productivity gains.
Chapters
cycle time has gone down, but PR cycle time seems really to have gone up. Robert or Kiz, either of you, is that something you've seen? Where should people start looking? What's up with that?
Yeah, I think I touched on that just a little bit, but let me expand on the couple reasons I've seen that. If a bunch of your developers are generating code and they're AI-assisted and they're starting to see a productivity ramp, but they're not also using those, those, you know, the, the code review tools, you're going to create a bottleneck where you're generating more than you can, you can physically review at a time. The other piece is we see this productivity, like if there's like a productivity curve with AI, right? But in reality, like the, the figuring out the prompt to figure out where to use this versus that, how to use the tool, where best to use the tool, like that actually generates a productivity hit. That's pretty common that as folks start to use these tools, they, they actually see a little bit of pullback of productivity, and you have a much higher success rate there.