🔗 Using AI to write better code more slowly

A lot of people think that AI coding is about churning out mediocre code as quickly as possible. However, Nolan Lawson argues that large language models (LLMs) can actually be leveraged to produce high-quality code, but at a more deliberate pace. Instead of mere slop, LLMs can unearth a plethora of bugs in existing codebases, allowing developers to enhance code quality over speed.

Lawson suggests using multiple models, like a Claude agent alongside Codex and Cursor Bugbot, for thorough code reviews. This method reveals critical bugs and teaches developers about their codebase's intricacies. Although this approach might not accelerate coding speed in terms of raw output, it leads to a healthier codebase and deeper understanding.

The post encourages developers to embrace this slower, more methodical style of coding with agents, calling it a super-powered version of careful programming.