Quoting Kenton Varda

Kenton Varda has implemented a ban on the use of AI-generated text for change descriptions, including pull requests, commit messages, and issue tickets within his team. This decision stems from his experience reviewing code where AI-written summaries proved to be counterproductive.

The primary issue identified is that AI tools tend to focus on superficial details that are already obvious from reading the code itself. By highlighting these trivial aspects, the AI fails to provide the necessary high-level context or intent behind the changes. This lack of framing makes it difficult for reviewers to understand the broader purpose or reasoning behind the modifications.

This shift highlights a growing tension in software development as teams integrate generative AI into their workflows. While AI can automate documentation tasks, it often struggles to capture the human perspective required for effective code review. By requiring human-written descriptions, Varda aims to ensure that developers communicate the essential context that automated tools currently miss, ultimately improving the clarity and efficiency of the review process.