CONCEPT
Artisanal Coding
Joel Spolsky’s term for software written by a human engineer who has thought through not just what the code does but why it is structured the way it is—the practice of owning the design decisions that AI-generated code makes invisibly, and the standard against which AI output must be evaluated to prevent the most dangerous class of abstraction leaks.
Artisanal coding is
Joel Spolsky’s name for the practice that the
Law of Leaky Abstractions implies is non-negotiable in any era of powerful abstraction tools. It is not nostalgia for hand-written code; it is a quality standard. Code that has been artisanally produced is code whose author can explain why it is the way it is rather than some other way—which structural decisions are load-bearing, which are arbitrary, how the system will behave at scale, how it will fail, and how it will need to evolve when requirements change. The contrast class is code whose authorship is statistical: AI-generated output that matches the patterns of its training data without encoding any of the design reasoning that would allow a future maintainer to evaluate, modify, or debug it confidently. Artisanal coding in the AI era does not mean