Ars Technica reports on research using data from engineering-analytics vendor Jellyfish that found AI coding agents increase how much code is written without increasing the amount of finished software, because the average pull request review time balloons 49 percent after agents are introduced. The study found the share of pull requests with changes requested nearly doubled and comments per pull request rose 35 percent, while the share of workers doing code reviews grew 14 percent and the researchers said they could not attribute significant employment changes to AI. AI review tooling had only a marginal effect so far: 80 percent of measured firms used some form of AI code review by March 2026, but agents produced just 23.3 percent of review comments and 10.8 percent of pull requests, leaving the vast majority of review work to humans.
Study links AI coding agents to 49 percent longer pull request reviews
Ars Technica reports on research using data from engineering-analytics vendor Jellyfish that found AI coding agents increase how much code is written without increasing the amount of finished software, because the average pull request review time balloons 49 percent after agents are introduced. The study found the share of pull requests with changes requested nearly doubled and comments per pull request rose 35 percent, while the share of workers doing code reviews grew 14 percent and the researchers said they could not attribute significant employment changes to AI. AI review tooling had only a marginal effect so far: 80 percent of measured firms used some form of AI code review by March 2026, but agents produced just 23.3 percent of review comments and 10.8 percent of pull requests, leaving the vast majority of review work to humans.
Source: Arstechnica