The process of building Bugbot began with qualitative assessments and gradually evolved into a more systematic approach, using a custom AI-driven metric to hill-climb on quality.

Cursor

Cursor's Bugbot is an AI-powered code review agent that analyzes pull requests for bugs, performance issues, and security vulnerabilities. The team evolved it from qualitative assessments to a systematic approach using a custom AI-driven metric called resolution rate. Through 40 major experiments, they increased the resolution rate from 52% to over 70% and more than doubled resolved bugs per PR. Key improvements included parallel bug-finding passes with majority voting, transitioning to a fully agentic architecture with dynamic context discovery, and building supporting infrastructure for Git integration and custom rules. The system now reviews over 2 million PRs monthly for major companies, with upcoming features including automatic bug fixes and continuous codebase scanning.

Building a better Bugbot