Agentic AI code review at scale (Sashiko, Qodo, Crafting.dev, CodeRabbit, OpenCode, Claude Channels, R2C2)
Key Questions
What is Sashiko in agentic AI code review?
Sashiko is an AI code review tool that garnered 93 Hacker News points for scalable agentic reviews. It fits into ecosystems with closed-loop systems like Crafting.dev. Benchmarks compare it favorably against Devin, Cursor 3, and others.
How does CodeRabbit perform AI code reviews?
CodeRabbit automates pull request workflows with tutorials on fixes and effective reviews. It generates PRs faster than teams can review, addressing slop mitigation. Videos demonstrate practical implementation for real-world use.
What is Qodo and its significance?
Qodo raised $70M for agentic AI code review at scale, integrating with tools like OpenCode and Windsurf. It features in benchmarks alongside Claude and Gemini CLI. It supports closed-loop reviews and PR automation.
How do benchmarks compare AI code review tools?
Benchmarks evaluate Devin, Cursor 3, R2C2, Qodo, Claude, Gemini CLI, and Windsurf SWE-1.5 on speed and accuracy. Windsurf claims 14x faster performance than Claude. They highlight slop mitigation and real-world efficacy.
Can AI coding assistants discover vulnerabilities unprompted?
An AI coding assistant found a real Linux kernel vulnerability without being asked, showcasing proactive capabilities. This occurred during routine tasks, raising implications for security. It underscores potential in agentic reviews like Claude Code kernel hunts.
Sashiko (93 HN); Crafting.dev closed-loop; CodeRabbit tuts/fixes; Qodo $70M; OpenCode/MiMo/Windsurf; Continue/Claude Channels/PR/Auto Mode; R2C2; Claude Code kernel vuln discovery; slop mitigation; benchmarks (Devin/Cursor 3/R2C2/Qodo/Claude/Gemini CLI/Windsurf SWE-1.5).