Code review is the next bottleneck people want to rebuild
GitHub put its new pull request overview page into public preview. Commenters say it misses the point. One says their pain is that first-round PR comments lose their place after later commits, so it is hard to tell if feedback was addressed. Another says they are building their own review pipeline with automated review guides and risk detection, and is surprised GitHub isn't doing more. A third groans at the "merge box, distilled" copy.
The pressure comes from AI-generated code. With agents writing more diffs, reviewing becomes the slow, human part. The Claude Code leak and the harness posts show the other half, which is how much unreviewed agent output is being committed.
The opening for startups is that teams are already writing their own tooling, which signals unmet demand. The risk is that GitHub owns the surface where reviews happen.
So what?
If you are building developer tools, review, not generation, is where users say the pain is. Focus on tracking feedback across commits and on risk triage, and expect GitHub to move toward it eventually.