Using AI to write better code more slowly

This article argues against the common perception that AI coding tools are only useful for rapidly producing low-quality code. Instead, it presents a workflow where multiple AI models (Claude, Codex, Cursor Bugbot) are used to thoroughly review pull requests, identify bugs ranked by severity, and fix them carefully. The author shares personal techniques for prioritizing critical bugs, avoiding false positives, and improving code quality without necessarily increasing velocity. The piece emphasizes using AI to write better code more deliberately, contrasting with the 'slop cannon' approach often associated with LLMs.