How we used AI agents to migrate GitLab rate limiting

What this post added

This post details the successful use of AI agents, specifically within the GitLab Duo Agent Platform, to migrate a legacy rate-limiting system to a unified implementation in `labkit-ruby`. It highlights the structured workflow ('the pod, the loop, and the observability') that enabled this migration, including AI drafting specs, implementing bounded changes, writing tests, and pre-reviewing merge requests, with human oversight for scope, architecture, rollout, and final review. The post also discusses lessons learned regarding observability, the need for human judgment in defining 'done' and managing infrastructure bottlenecks, and the shift in the bottleneck from code generation to review capacity and operator attention.