Linear, the project management software company, published details Sept. 21 on how it reworked its continuous integration pipeline after AI coding agents made shipping code faster than validating it, according to a post on the company's blog. "Agents have made it exponentially faster to ship code, but validating those changes hasn't quite kept up at the same rate," the company wrote.
The company said it moved off GitHub Actions to third-party runners with faster hardware, reorganized which jobs sit on the critical path that blocks a merge, cut redundant dependency installs and caching steps, and split test files for more aggressive parallelization.
Linear said the changes cut median pull request wait time from more than six minutes to just above five, roughly halved runner time per test, cut TypeScript compilation time 73% after a toolchain update and reduced linting time 55%. Module state caching saved about 17% a month, and batching changes saved roughly 87,000 runner-minutes a month, the company said, even as its test suite nearly quadrupled in size over the year.
CI has always run on its own clock, no matter how fast someone writes the code that triggers it. Linear's numbers are a rare public look at how much slack that gap can eat once an agent is writing pull requests faster than a person could, and what it costs to close it back up.