Don’t get one-shotted: Use AI to test, review, merge, and deploy code
summary
AI-generated code now makes up about 46% of GitHub code, and studies warn up to half of AI outputs contain bugs or vulns. Graphite builds Diamond, an AI-native reviewer that integrates with CI, reads whole repo history, summarizes and prioritizes changes, and provides one-click suggestions to fix test failures. Deployed at enterprises, Diamond cuts review cycles and shows a 52% acceptance rate for its comments versus ~45–50% for humans. Tools mentioned include Copilot, ChatGPT, Cursor, Windsurf, VZero, and Bault.
AI-written summary · about
related talks
14mAI-Powered Code Review and Quality Assurance
19mWith great power comes great responsibility in code quality
19mThe AI Code Review Bottleneck: Escaping the Velocity Trap
26mAI Agents as Your Co-Developer
09mImproving code reviews with AI agents
15mThe reality of developers using AI | State of Code Developer Survey
33mAI assisted software delivery
31mModern Architecture for AI-Augmented Development Teams