Linear re-engineers its CI to keep up with AI code speed

AI coding agents make shipping code fast, but waiting on CI builds quickly becomes the new bottleneck.

Linear shared how a surge in AI-generated pull requests forced them to completely rework their continuous integration pipeline. Despite their test suite almost quadrupling since January, the team trimmed PR wait times from over 6 minutes to just over 5, while cutting machine time per test roughly in half.

Why it matters: AI agents let developers spit out PRs faster than ever, but every change still has to pass through CI. If validation doesn't keep up, developers sit around waiting for feedback while infrastructure bills soar.

Here's the gist of how they fixed it:

• Upgraded hardware and tools: Moving off standard GitHub Actions to third-party runners made jobs 34% faster. Switching to tsgo cut median TypeScript checks by 73%.
• Unlinked linting from typechecking: Stripping TypeScript type dependencies from ESLint rules dropped API lint time by 68%.
• Smarter git checkouts: Fetching only necessary files dropped change-detection job medians from 26 seconds to 8 seconds. Custom retries fixed network hangs on third-party runners.
• Cleared the critical path: Moving cache-marker writes out of the final merge gate shaved 42 seconds off every API pull request.

Faster code generation is useless if your build pipeline keeps you waiting in line.