Six rules for AI code review: 5 self-improvement rounds raise critical bugs 37.6%
rseroter · x · 2026-08-19
Aviator published best practices for AI code review in an era of machines reviewing machines. Context: AI writes code faster than humans can review, and an IEEE ISTAS 2025 study found a 37.6% rise in critical vulnerabilities after just five rounds of asking a model to "improve" its own code.
Six core practices:
- Review intent before the diff gets huge;
- Keep changes small;
- Layer your tools instead of trusting a single reviewer;
- Verify behavior by actually running the code;
- Encode recurring rules once instead of repeating yourself;
- Humans make the final call.
The reviewer's role has shifted from checking compilation and style to proving the code actually solves the intended business and logical problem.
More from coding & agent
- Single developer migrates 850K lines with agents in 6 months — tristanbob · 2026-08-19
- Monitoring 2k+ MCP Servers: 7,190 Safety-Relevant Changes, Read-to-Write Flips Undetected by Allow-lists — mcpindex · 2026-08-19
- Generating UIs by having LLMs emit raw HTML/JS? That's a risk pile, says Googler — rseroter · 2026-08-19
- YC Startup Launches Graphify, a Knowledge Graph Engine for Enterprises — ycombinator · 2026-08-19
- Open-source SRT whiteboard animation skill: hook it to Codex and ship 100 videos a day — huangyun_122 · 2026-08-19
- Founder's $20/month stack: Claude coding plus free SaaS runs a whole startup — clarashih · 2026-08-19