Generative AI is reshaping how software engineers collaborate, but it's creating specific inefficiencies. The author, a software engineer, highlights that while AI tools like coding agents can assist in writing code, they often produce lengthy and contextually lacking pull request (PR) descriptions. These descriptions tend to overwhelm the reader with unnecessary details and fail to convey critical information about the changes made. The lack of context means engineers spend more time sifting through redundant information rather than focusing on what truly matters in the changes being proposed. Another significant issue is the introduction of bugs that differ from those typically made by human coders. The author shares an example where a coding agent incorrectly implemented a command-line argument, leading to a security vulnerability. Such errors are not immediately recognizable because coding agents don't think or question design choices like humans do. This shift means engineers must adopt a more vigilant approach to code review. Every line of code demands scrutiny, making the review process more time-consuming and stressful. The problem extends to authorship and institutional knowledge. In traditional settings, knowing who authored significant code changes helped identify who to consult for questions. Now, when many contributions come from AI-generated code, it's unclear who truly understands the system. This ambiguity complicates communication and slows down problem-solving, as engineers must spend extra time determining who actually has the relevant knowledge about the codebase. Overall, while generative AI offers potential benefits, it also introduces challenges that can hinder team efficiency and collaboration.