Linear optimized CI infrastructure to speed up AI code processing

The team at Linear conducted a major overhaul of the continuous integration (CI) system to eliminate delays caused by the growth of code volumes generated by AI tools. Due to the high frequency of commits, standard testing processes became a "bottleneck" in the development cycle.

To solve the problem, engineers updated the hardware infrastructure and switched to using the native TypeScript compiler. The key change was the abandonment of full type checking during linting — the rules were rewritten to exclude resource-intensive operations. The fetch depth when cloning repositories was also limited, which significantly reduced the load on the network and disk subsystem.

The result of the optimization was a reduction in PR waiting time from 6+ minutes to 5+ minutes. At the same time, the company achieved fourfold increase in test coverage, which allowed maintaining the stability of the codebase with increased development intensity. The new configuration allows the system to handle the flow of changes generated by modern AI assistants.

This modernization highlights the growing need to adapt DevOps tools to the realities of AI-driven development. As developers delegate code writing to neural networks, the load on CI/CD systems increases manifold, requiring a transition from classic testing methods to more lightweight and parallel architectures.