#code-review

The thread’s useful tension was not whether AI can write code fast, but whether slower review loops produce code teams can actually trust.

HN latched onto a pain every heavy coding-tool user knows: the bug is tiny, but the diff balloons anyway. A new write-up turns that annoyance into a measurable benchmark and argues that better prompting and RL can make models edit with more restraint.

HN’s interest in Stage centered less on the chapter UI itself and more on the harder question: how humans stay responsible for code that agents helped create.

Hacker News upvoted this because it hits a pain every large repo has: giant PRs that are hard to review and harder to land. GitHub’s Stacked PRs preview turns a big change into an ordered chain of smaller PRs, then gives that chain first-class UI, merge behavior, and a dedicated gh stack CLI.

In an April 8, 2026 X post, Cursor said its code review agent can learn from pull-request activity in real time. The company also claimed that 78% of the issues the agent finds are resolved by the time the PR is merged.

A widely discussed Hacker News thread surfaced a Rust community summary that sees AI as useful for search, review assistance, and tedious semi-structured work, but risky for learning, subtle defects, ethics, power use, and vendor concentration.

GitHub said on March 20, 2026 that Copilot code review has surpassed 60 million reviews. The company’s March 5 blog says usage is up 10x since launch, now covers more than one in five code reviews on GitHub, and relies on an agentic architecture tuned for higher-signal feedback.

A high-traffic Ask HN thread shows a polarized view of AI coding tools: developers report clear gains on small scoped tasks, but many say autogenerated specs and cleanup work are eroding team velocity.

Hacker News reacted to reports that Amazon is requiring stronger senior sign-off on AI-assisted code changes after recent outages, reading it as a sign that verification and accountability are becoming the real bottlenecks in AI coding adoption.

Claude said Claude Code now includes Code Review, a feature that dispatches multiple agents on every pull request. Anthropic says the feature is in research preview for Team and Enterprise, with depth-first reviews rather than lightweight skims.

A LocalLLaMA post pointed to a new Hugging Face dataset of human-written code reviews, pairing before-and-after code changes with inline reviewer comments and negative examples across 37 languages.

GitHub said on March 5, 2026 that Copilot code review now runs on an agentic tool-calling architecture and is generally available for Copilot Pro, Pro+, Business, and Enterprise. The update is designed to pull wider repository context into reviews so comments are higher-signal and less noisy.