Cursor
@cursor_ai
Cursor's code review agent can now learn from activity on PRs to self-improve in real time. 78% of issues found are resolved by the time the PR is merged. https://t.co/VHG9nv9mgI
36retweets634likes
View on XCursor Bugbot Now Learns From Pull Request Feedback to Self Improve
Cursor· Updated
Cursor's code review agent, Bugbot, can now automatically generate repository-specific rules by analyzing developer reactions and human reviewer comments on pull requests. This real-time learning loop has increased the agent's issue resolution rate to 78%, significantly reducing false positives.
Cursor, an AI-native code editor built by Anysphere, updated its Bugbot code review agent (an autonomous system that plans and executes tasks) with a real-time learning capability called learned rules. The agent now analyzes merged pull requests to identify patterns in how developers interact with its suggestions.
Most AI code reviewers rely on static offline improvements, leading to repetitive false positives that ignore a team's specific business logic. By closing the feedback loop, Bugbot has reached a 78% resolution rate (the percentage of identified issues developers actually fix), outperforming the next-closest competitor by 15 percentage points and nearly doubling GitHub Copilot's rate.
You can enable learned rules through the dashboard to backfill rules from recent pull requests. The system automatically promotes candidate rules to active status as it gathers signal from developer downvotes, replies, and human-authored comments, though you can still manually edit or delete rules in the interface.
Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →
