code-review

A developer compares running Qwen3.8-Flash-Next locally (UD-Q4_K_XL on llama-server with ROCm) against Claude Opus in Claude Code for reviewing a real Swift commit: Qwen produced 8 findings with zero false positives and caught 2 issues Claude missed, but missed 5 others, making it a solid private first-pass reviewer rather than a final authority.

The author shares a retrospective of using their own subagent-team skill to lead a Claude/Codex/etc. AI team through a major monorepo overhaul: cross-vendor model reviews offer the best ROI, rework volume depends on task risk rather than model tier, and they summarize a division of labor where low-tier models act as coordinators, strong models handle risky implementation, and top-tier models handle reviews.

A blog post arguing that strict small-PR 'annealing' policies hinder rapid change, and that AI-driven coding produces big, interconnected changes which sometimes can't be cleanly decomposed into small reviewable steps.

The author agrees with poteto, criticizing adversarial code review between sub-agents as a waste of tokens, pointing out its three fatal flaws — meaningless nitpicking, over-engineering, and mismatched compute — and noting that top teams are shifting toward real testing, formal verification, and actual red-teaming to find issues.

An episode of Jane Street's official podcast Signals and Threads, hosted by Ron Minsky, featuring internal tools engineer John Crepezzi sharing hands-on experience rolling out AI coding tools at the company. Topics include the in-house JaneGPT/AIDE platform, dedicated AI identities and permissions, Slack-triggered tasks, whitelist-based sandboxes, guarding against reward hacking, and the importance of validation in code review.

A developer is building 'lineup', an open-source CLI that summarizes what coding agents changed and flags risky edits (deleted tests, config/env changes, auth code, new dependencies) before they become a PR. The post also asks the community what edits they'd most want flagged and how they currently review agent changes.

Kent Beck argues that trust, not code generation speed, is the binding constraint on software delivery in the AI era, and that XP practices like testing and pairing manufacture the confidence needed to release; he also promotes his upcoming paid class on trust-optimized AI-assisted development.

PR Cockpit 是一款面向 macOS 和 Linux 的开源桌面应用，帮助开发者在一个工作区中高效审查 GitHub 拉取请求，提供优先级队列、键盘导航、聚焦差异审查和 CLI 支持。

A developer shares a cautionary tale about an AI agent opening a 40-file PR that silently broke a nightly export by touching a shared util, and how enabling CodeRabbit's blast-radius analysis helped catch downstream dependencies on later large agent PRs. The post highlights the unresolved problem of setting PR size limits for AI-generated changes.

Kwant is a product launch featuring AI bots that cross-check each other's work before marking tasks as done, aimed at improving reliability of automated AI workflows.

A developer is seeking feedback on an open-source framework for coding agents that catches AI-generated 'slop' before it reaches the repo by having a second agent automatically review every change and send findings back, along with a session manager, ledger, and guards against repeated model mistakes.

A Google paper shows that Gemini AI agents are far less likely to wrongly undo correct work after a user's unfounded complaint when they first log a 'Thinking-in-Bets' checklist of verified claims and confidence levels, cutting erroneous rework from 12.2% to 0.02% in patch review.

Phorge is an open-source, self-hosted code review and software development management platform, a community fork of Phabricator. The page content here is just the site's bot-protection (Anubis Proof-of-Work) landing screen rather than an actual announcement.

diff-tour launches to help developers read and understand AI-written code like a senior engineer, now featured on Product Hunt.

GitHub Copilot Automations can now learn from real outcomes on GitHub—such as label changes, comment acceptance, and merged PRs—to propose prompt improvements, with a human-in-the-loop review before changes are applied. The feature is available across Copilot Pro, Pro+, Max, Business, and Enterprise plans.

GitHub faces infrastructure bottlenecks in the AI era: monthly pushes grew 4.9x in a year, and the 5-server sync architecture per repository has become the bottleneck. The article analyzes GitHub's rebuild plans, moves by new players (Entire, Cursor, etc.), and argues that the real revolution isn't replacing GitHub as a platform, but AI code review replacing human line-by-line review.

CodeRabbit launched CodeRabbit Secure, whose Deep Scan traces vulnerabilities across linked repos and replays exploits in a throwaway sandbox to confirm real risk before flagging, while tracking which merges introduced or fixed each bug.

Entire 发布了首个专为 agent-human 协作打造的 Git forge，核心包括 Trails 代码评审体验、分布式 Git 托管（比 GitHub 快 89 倍）以及 Continuous Memory 三大原语，面向 AI-native 团队和编码代理开放使用。

Greptile is leveraging GPT-6 Astra's computer use capabilities to automate application testing beyond static code review, alongside its Daybreak security product and Codex-integrated Knowledge Base, in a push to reduce bugs and rework for developers.

CodeRabbit is using OpenAI's GPT-6.1 Sol as the primary model in its security agent to strengthen code security across complex codebases, enabling cross-file risk analysis, dependency investigation, exploit verification in sandboxes, and automated fix pull requests.