Veteran devs unveil SpiraCSS to make AI fix its own CSS disasters—because who has time for human code reviews anymore?

The evolution of CSS architecture is being reevaluated with the emergence of AI coding agents, such as Claude Code, Cursor, and Windsurf, which are becoming increasingly prevalent in UI implementation. Traditional naming conventions like BEM and SMACSS, as well as utility-first design philosophies, were designed with human authors in mind. However, these approaches have limitations, including the need to memorize conventions and subjective decision-making. To address these challenges, a new approach is being proposed, focusing on feedback systems rather than rules-based designs. Inspired by TypeScript's type system, this approach involves the use of lint output to provide feedback and guide corrections, enabling both human and AI developers to produce consistent and high-quality code. SpiraCSS, a CSS architecture methodology, is being explored as a concrete example of this approach, utilizing Stylelint to verify component structure and property placement, and providing error messages with fix instructions to facilitate self-correction.