May Walter

This talk addresses the challenge of silently accumulating performance issues in mature codebases, where pausing feature development for investigation is difficult. It presents a case study on integrating runtime intelligence into coding agents to enable continuous performance optimization in production. The approach involves agents analyzing real production data to identify high-return-on-investment fixes, prioritized by complexity and impact.