CSS has long been a source of frustration for developers due to its complex layout rules. Centering elements like a <div> remains a notorious challenge, and the author reflects on the convoluted methods once needed to achieve layouts like the "Holy Grail" design. While CSS is deterministic—meaning it produces the same output if the same code is run twice—its predictability suffers because the rules are intricate and often counterintuitive. The author, who has a strong background in CSS specifications, points out that even seemingly simple properties like `text-align: center` can lead to unexpected results due to conditions that may not align with user intentions. One proposed alternative to traditional CSS is a constraint system, which allows developers to set specific relationships between elements, such as centering an object within its parent. This method theoretically improves predictability because it aims to satisfy defined constraints. However, real-world applications reveal significant challenges. Layouts can become under- or over-determined, meaning there may be too few or too many constraints to create a stable design. This leads to situations where the browser might disregard certain constraints, further complicating predictability. The author highlights that while constraint systems can streamline layout processes, they introduce their own set of problems. As complexity increases—either through the number of constraints or the rules governing conflict resolution—predictability diminishes. Testing layouts on different screen sizes further complicates matters, as a constraint that works well on one device might fail on another. Real-world examples, like the experiences with LaTeX and iOS's constraint layout, show that even established systems struggle with maintaining predictable outcomes, leaving developers in a state of uncertainty.