Three posts landed in September that I think engineering leaders should read together.
Anthropic’s engineering team wrote that their continuous integration (CI) job volume grew 25x in six months. Their engineers now ship about 8x as much code per quarter as they did from 2021 to 2025. The fix they shipped was test impact analysis: run only the tests a change could affect.
A week later, Linear published a post titled AI coding has made CI a bottleneck. Their test suite has nearly quadrupled since January, and agents now write most of their tests. They reworked the pipeline end to end to keep up.
Then Depot’s CEO wrote that CI is changing, and that the future is “giving agents a way to validate code and maintain trust as they work.”
Anthropic and Linear do not sell CI tooling. They are reporting what happened to their own pipelines, which is what makes it worth paying attention to. Here is my read: all three are right about the problem, and two of them are fixing the wrong layer.
For twenty years, CI was sized for human output. A developer opened a few pull requests (PRs) a week, and the pipeline ran after each one. If it took 20 minutes, nobody cared much, because the developer was already on the next task.
Agents broke that arithmetic in two ways. First, volume. When one engineer runs several agents in parallel, PR count goes up by a multiple, not a percentage. Anthropic’s 25x number is not an outlier.
Source link







