Photo by Charlotte May on Pexels

A Google Gemini user has reported significant challenges while using the AI model for code review, citing instances of cascading context leakage and subsequent output failures. According to the user’s report, supplemented by screenshots, the issues stem from Gemini’s difficulty in maintaining context throughout the review process. This revelation has sparked a discussion within the AI community regarding the limitations of large language models when handling complex, context-dependent tasks like code analysis. The original discussion and supporting evidence can be found on Reddit: [https://old.reddit.com/r/artificial/comments/1mcyay8/cascading_context_failure_and_leakage/]