Code is read far more often than it’s written. Clean code optimises for the AI and the human seldom looks at it.
Unlike Robert C. Martin’s 2008 book Clean Code: A Handbook of Agile Software Craftsmanship which popularised the term and a set of concrete practices around it, meaningful names, small functions, minimal duplication as a discipline is fast colluding with machine written “code that works.” That diminishing distinction has since drawn real empirical attention: controlled studies of code-readability testing find that most programmers can measurably improve their ability to write readable code within a handful of guided sessions, and that unreadable code correlates with slower maintenance and more introduced bugs.
The definition of clean code however is shifting from a solely human-focused discipline to one optimised for AI-driven development and machine execution. While grounded in traditional practices like those in Robert C. Martin’s 2008 text, modern clean code focuses on semantic predictability to aid LLM context windows, supported by empirical evidence linking readability to lower maintenance costs.
Source: Sedano, T. “Code Readability Testing, an Empirical Study,” IEEE CSEE&T, 2016. Foundational: Martin, R.C. Clean Code, 2008.