AI Coding Assistants: Code Review is King

Large language models excel at generating code, but lack the judgment of a seasoned software engineer, often leading to poor design choices. The author uses personal anecdotes to illustrate how AI coding assistants (like Codex and Claude Code), while powerful, require close supervision to prevent them from veering off into architectural dead ends. The author argues that mastering code review, particularly focusing on code structure and identifying opportunities for improvement, is crucial for effective use of these tools. Blind reliance or excessive micromanagement will hinder productivity. Ultimately, the author posits that current AI coding is more akin to 'centaur programming'—a collaboration between human and machine—rather than a complete human replacement.