Anthropic Scales Test Impact Analysis to Handle 25x CI Job Growth from Agentic Coding

Anthropic’s CI job volume increased 25x over six months as AI agents accelerated code generation and review. Engineers now ship 8x more code per quarter compared to 2021-2022, with Claude authoring 80% of that code and playing a significant role in PR reviews. This surge threatened to overload the company's test impact analysis service, leading to a redesign after three failed patches (lasting 70 days, 29 days, and less than a day, respectively). The final solution moved from a single-process listener to a horizontally scalable architecture using an in-memory data store, allowing listener processes to scale independently. This change was guided by long-running sessions with Claude Tag, which helped identify the need for horizontal scaling and generated code for the new architecture.