GitHub New pull request overview page in public preview
GitHub’s new pull request overview page preview draws mixed reactions, with many welcoming small layout tweaks like a more prominent merge box but criticizing the change as largely cosmetic. Commenters argue that the real pain points are the growing volume and complexity of PRs—especially with AI-generated code—along with slow, unreliable PR pages, broken merge queues, and weak support for large-scale review workflows. Several point to alternative tools and ideas such as semantic diffs, better comment threading, and agent-assisted triage as the kinds of deeper changes needed to make code review sustainable.
Overall reaction to the new PR overview page
- Many see the change as minor layout tweaks, not a fundamental improvement.
- Some appreciate having status checks and the merge/approve controls more visible at the top.
- Others find it adds clicks and still forces jumping between tabs; they want a single, dense page with code, comments, and checks together.
- Several argue it does not address emerging AI-driven workflows or the core pain of modern code review.
Persistent PR UX and reliability issues
- Strong frustration about performance: PR pages load slowly, especially with many comments; comment submission can lag several seconds.
- Reported bugs: “merge when ready” often failing silently; unreliable merge queue; delayed or missing workflow scheduling; sidebar file links not scrolling to the right file.
- Conversation views are described as cluttered and hard to navigate, with excessive “load more,” outdated bot reports, hidden comment threads, and no good way to track whether feedback was addressed.
- The notification inbox is described as inaccurate and slow.
- Several feel GitHub prioritizes new UI over fixing regressions and reliability.
Code review in a high-volume / AI world
- Many describe code review as the main bottleneck, especially with agents producing large or frequent PRs.
- Opinions split on PRs themselves: some say the PR-centric lifecycle is broken and unsustainable at current volumes; others argue PRs are fine if changes are small, pre-discussed, and well-described.
- LLM-authored code is seen as changing review dynamics: fewer obvious human mistakes, more subtle or “alien” errors, breaking reviewers’ old heuristics.
- Some admit people often don’t fully read AI-generated diffs anymore and instead lean heavily on CI and bots.
Proposed improvements and alternative tools
- Desired features include: semantic diffs (e.g., recognizing renames/moves), “hunk checkpoints” or line-level review progress, better diff layout control, interdiffs, and strong keyboard navigation.
- Several third-party tools are mentioned as offering better PR UX, stack awareness, or merge queues.
- Some teams are experimenting with agents to perform initial reviews, generate diagrams and code walks, or even handle merges, with humans only focusing on risky changes.
Cultural and philosophical undercurrents
- Some enjoy reviewing and see difficulty as a symptom of poor upstream process (lack of design review, giant surprise diffs).
- Others think human-centric, line-by-line review is becoming infeasible, pushing towards system-level oversight and invariant-based tooling.
- Marketing-style copy in GitHub’s announcement is criticized as vacuous and possibly AI-written.