Linear cut CI runner-time in half while tests quadrupled. The review gate can't do that trick.
Linear put up a really good post about reworking their CI because AI coding agents had turned it into the bottleneck. The numbers are worth reading on their own terms: their test suite roughly quadrupled since the start of the year, yet they brought PR wait time down from over 6 minutes to just over 5, while cutting runner time per test roughly in half. They did it the way good engineers do.…
Linear's recent post reveals a strategy for reducing their CI runner-time by half while increasing test suite size fourfold, thanks to implementing AI coding agents. The changes they made include switching to faster third-party runners, removing type information in lint rules, capping fetch depth on critical jobs, and offloading cache writes off the critical path.
These are all machine-parallel fixes that can make each unit cheaper or run more units at once. However, unlike CI, review is not embarrassingly parallel. Review involves serial reasoning across the whole change, with two reviewers not halving the time but adding surface for disagreement. The post argues that while you can speed up the machine side of your pipeline, it only moves the serialization point, and if agents quadruple your diff volume, you just push the wait to review.
The ideal solution is to treat the review gate as the critical path, spending deterministic checks first and concentrating model reasoning pass on the diff-level judgment the rules can't reach. Tools built around the deterministic layer cheaply and the reasoning layer deliberately are the ones that scale with agent output.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
