.png)
One of the most impactful learning experiences I have had in my master’s program occurred outside of the classroom: when I signed up for my first hackathon.
Four designers with minimal coding experience, building a working prototype in 24 hours, in a competition full of engineers. The prompt: How can we empower digital artists to realize their creative vision fully?
As UX designers, we’ve been increasingly using AI tools like Claude, Cursor, and Figma Make to prototype interfaces. While these tools are powerful, they require designers to shift from the hands-on, visual workflows most intuitive to us to navigating chat logs, code, and fragmented outputs.
We kept running into the same tension: the faster we could create, the harder it became to manage what we created.
In tools like Figma or Adobe Creative Cloud, iteration is intuitive—we rely on version history, layers, frames, and reusable components to explore ideas without losing control. In AI tools, those fundamentals break down. Versions get buried in chat logs, changes aren’t easy to compare, and restoring work often means losing progress entirely.
We saw an opportunity to design a system that restores that sense of control—one that makes AI workflows feel visual, structured, and aligned with how designers naturally iterate.
We started by outlining a PRD, mapping our pain points to potential features. We took inspiration from design tools with established version control systems, such as Figma and Adobe Creative Cloud, which would already be intuitive to our audience.
We used Claude to generate an initial low-fidelity prototype based on the PRD. For each screen, we placed screenshots in Figma and left comments identifying where the usability and design could be improved for our next iteration.
For high-fidelity design, we worked directly in Figma for more control over the output, following a simple design system. We used the font Instrument Sans for an approachable but professional look, a minimalistic color palette, and gradients to make the design feel more dynamic. Our goal was to break away from generic AI outputs and develop a strong visual identity, leaning into our strengths as designers.
Lastly, we plugged our designs back into Claude to prototype their functionality, further iterating through prompts, and published the final product to netlify.
Iterait is a visual workflow system for AI-assisted design.
It allows designers to track iterations, compare versions side-by-side, and restore specific changes without losing progress. Instead of relying on chat logs or code, Iterait gives power back to the designer by allowing them to manage their designs through a clear, visual workflow.
Because designers use multiple AI tools, we positioned Iterait as a third-party layer that integrates across platforms like Claude, Cursor, and Lovable. This allowed iterations from different tools to live in one place.
I designed the version control experience, shaping both the functionality and design.
View the full walkthrough below & our prototype here.
In 24 hours, we turned a fragmented, text-heavy workflow into a clear, visual system. The result was a functional prototype that demonstrated how designers could:
- Compare iterations visually instead of parsing chat logs
- Reuse changes instead of rewriting prompts
- Maintain control while working across multiple AI tools
Our project placed top three in our track, and we had the opportunity to present to the hackathon and panel of judges, validating both the idea and its execution.
This project shifted how I think about AI in design.
The challenge isn’t generation—it’s control. AI can produce outputs quickly, but without structure, those outputs become difficult to manage. Designers need tools that help them understand and direct what AI generates, not just create more of it.
It also pushed me to rethink my workflow when designing with AI. While I initially prioritized pixel-perfect design in Figma, AI couldn’t accurately replicate the designs. If I were to do things differently, I would have spent less time perfecting the design in Figma and more time focusing on refining the usability and iterating within Claude.
