Agentic AI and Code Reviews: Patterns and Case Studies
summary
AI-generated code rises 19% to 57% in three quarters, and top teams now move 9x faster than the median. The presentation lays out concrete patterns: small incremental changes, design for verification and spec-driven development, context engineering, deterministic checks, and adversarial and multi-agent review. DryRun Security fixes a missed multi-tenant isolation by adding product-specific context and spec checks; Cloudflare shifts reviews into CI with an orchestrator to dedupe noisy feedback; Block pilots local AI reviews then scales to merge-ready automated fixes. These practices remove review as the bottleneck.
AI-written summary · about
