Every craft has a standard of quality.

For most of history it lived with the people who mastered the craft, and was passed on from there. Today screens, code, documents and whole products are produced faster than ever before. Faster than an expert can review them. With agents now producing, reviewing the quality of human centered design work is even more challenging. Because the easier something is to make, the more of it we tend to make. As a consequence, the more we make, the more difficult it gets to review for quality.

Eval Hub builds agentic evaluators for design, creative and content. They span user experience, human computer interaction and brand work. Each evaluator is grounded in what’s known about how people see, think and act. They’re based on Gestalt principles, psychology, neuroscience and more. The evaluators are useful for improving the quality of digital experiences. They support human and agentic workflows.

“Most of what makes a design work is already known, it’s in decades of research on how people see, remember and decide. We built the evaluators so that knowledge is present every time someone makes something.”

— Liliia Kujda, Founder

The first evaluator is DesignEval, a plugin for Figma that you can incorporate into your design process. It reviews layouts and flows against six heuristics and explains what it finds and why. It sits alongside the tools people already use to design, so you can tune the work as you make it. In seconds it helps you identify simplification points, cognitive loads, and orientation friction as people move through your experience. You decide what to act on. Then when you’re ready, run it again.

“Good design is a long list of small decisions, nobody gets every one of them right on the first pass, it takes time and cycles. I think of the evaluator as an extension of my deep thought, but faster. That’s the whole idea.”

— David Kujda, Founder

We’re building Evals one craft at a time, starting with design. Each one improves through use and as we learn more about evaluations itself. We want more of what people use to be good, and we want it to be easier for the people making it or overseeing it to get it there. AI can now produce an enormous amount of work, but this work still needs the same review for quality as work made by people. The evaluators are built for both. Most of daily life now runs through screens and devices. When they get easier to use, life gets a little better.