Key Takeaways

  • AI-generated UI/UX often defaults to a generic look, famously the 'indigo purple gradient', because models lack a compositional framework for design thinking.
  • Ahmad Awais's "repair logic" can fix about 90% of this pervasive "design slop" by addressing a "contract gap" between user intent and LLM capabilities, rather than a model's inherent skill deficit.
  • Guiding AI with specific design "intent-first composition" (like defining a dashboard as a "monitor surface") drastically improves aesthetic and functional output.
  • Forcing LLMs to use the OKLCH color space over HSL allows for superior color control and human-perceptible accuracy, eliminating washed-out or poorly contrasted AI designs.
  • The core solution to generic AI designs lies in applying Awais's Design Slop Repair Framework by structuring prompts with intent, specific color spaces, and design best practices.

The Design Slop Repair Framework

When Ahmad Awais talks about fixing AI's design "slop"—that endless sea of generic UIs, often with the infamous 'indigo purple gradient'—he's not just complaining. He's talking about a deterministically repeatable fix. "We have been able to apply the same thing to design slop," Awais says, referring to a "repair logic" his team developed for DeepSeek V4 and other open models that fixed persistent tool-calling errors. This isn't about teaching AI creativity from scratch, but guiding its existing capabilities with intent.

Here’s the framework that helps you do it:

- Intent-First Design (Work Pattern First Composition): When you ask a model go and design me this dashboard they generally do not think about the intention behind that design... If you give them a very simple framework of, you know, what type of surface area are we looking for, which is literally just these seven patterns, they do really, really well. For instance, 'this is a monitor surface, right? Like we're trying to monitor That is the intention behind this, right?'

- Enforce OKLCH Color Space: LLMs are really good at it. And if you see them using HSL or something they are they they they they don't actually are able to control the lightness in HSL very quickly, but on to human eye it's very, very easy to see like this color and this color do not look the same. But if you force an LLM to use OKLCH, they can control the colors palette really really well compared to any of other things.

- Utilize Design Smells & Reference Documents: We only have like 24 reference documents, 10 you know, design smells, and seven patterns that we saw across different designers.

When This Works (and When It Doesn't)

This framework shines when you, the builder, have a clear understanding of the design's purpose and are willing to articulate it explicitly. Awais notes, “If you can give them a framework of this is what the design taste of a really good designer is like, they will pick this type of color scheme, they will pick they will think about intent before starting to implement that landing page. It makes your design slope really, really minimal.” It effectively bridges the "contract gap" between what a user implicitly wants and what an LLM needs to be told.

However, it falters if your own intent is vague, or if the model you are using doesn't have a strong foundational understanding of design principles to begin with, even when prompted. It also might not magically generate truly novel aesthetics if the model has no exposure to diverse, high-quality reference documents. Think of it as refining an existing talent, not creating it.

What to Do With This

This week, when you next prompt an AI for a UI/UX design, don't just ask for a "dashboard." Instead, apply the Design Slop Repair Framework directly. First, define the Intent-First Design: "Generate a user analytics dashboard that serves as a 'monitor surface' for real-time engagement data." Next, Enforce OKLCH Color Space by adding: "Use the OKLCH color space for all palette choices, prioritizing accessibility and clear data visualization contrasts." Finally, Utilize Design Smells & Reference Documents by instructing: "Avoid common 'design smells' like generic three-card layouts. Instead, draw inspiration from clean, enterprise-grade data dashboards you've seen, focusing on information hierarchy and minimal ornamentation." This approach forces the AI to move beyond its default 'indigo purple gradient' and toward a truly usable, aesthetically pleasing result.