AI coding has made CI a bottleneck, so we reworked ours to keep up (linear.app)

Linear has revamped its continuous integration (CI) processes to tackle performance bottlenecks exacerbated by the rapid adoption of AI coding tools. As the volume of code changes accelerated, the traditional CI system struggled to keep pace, leading to longer feedback loops and increased infrastructure costs. By strategically optimizing CI, Linear reduced pull request wait times from over six minutes to just over five, despite a near quadrupling of their test suites. Key improvements included migrating from GitHub Actions to faster third-party runners, optimizing the use of TypeScript compilers, and revamping custom linting rules to use static analysis rather than type information, leading to significant reductions in CI job times. The significance of these enhancements lies in their broader implications for the AI/ML community and software development at large. As development teams integrate AI-assisted coding, CI systems must adapt to ensure they do not become hindrances. By implementing changes that minimized repeated setups, improved test execution efficiency, and introduced smarter job scheduling, Linear demonstrated a scalable approach to CI optimization that can serve as a blueprint for others in the industry. With a proactive approach to handling increasing test volumes—adding 2,000 tests weekly—Linear's efforts exemplify a commitment to maintain swift development cycles while managing the complexities introduced by continuous AI integration.