Highlight
Featured this release
PR review verdicts now weigh all of a PR's tests together
The PR review agent assesses every test in a run side by side instead of one at a time, catching failures that share a root cause.
Previously, each test in a PR review was assessed on its own. Now the same agent that reviews the PR looks at all of its test results together, so it can catch cases where several tests fail for the same underlying reason – a shared login step, for instance – instead of reporting them as unrelated failures.
Tests still being assessed show as "Awaiting assessment" in the test composer while the agent works. This builds on the diff-anchored verdicts and self-repair work from the past week.