AI Keeps Building the Same Purple Gradient Website (prg.sh)

In August 2025, Adam Wathan addressed the design community over X, apologizing for the dominance of a specific purple gradient in AI-generated interfaces, a result of his choice for Tailwind UI five years prior. This phenomenon, termed “AI slop,” has led to a homogenization of web design aesthetics, characterized by predictable elements like the Inter font, three-box layouts, and the excessive use of purple gradients. This issue showcases how training data deeply influences the outputs of large language models (LLMs), culminating in a style defined more by statistical averages than creative design. The significance of this announcement lies in its implications for the AI/ML community, particularly for the development and training of LLMs. The reliance on generic design prompts leads to superficially appealing but fundamentally flawed interactions, often lacking necessary features like user validation in forms. To address this, the Anthropic team proposed specific strategies, emphasizing the importance of explicit design constraints in prompts. Their cookbook suggests guiding LLMs with precise design dimensions, referencing distinct inspirations, and explicitly detailing what to avoid. This approach not only enhances the creativity of AI-generated designs but also improves their functionality, fostering a move towards more personalized and effective user experiences in web design.