Design Arena's creators raise $7.9M to bring taste to AI models

As co-founder Grace Li tells it, her company started a few weeks before graduation in 2025, with a handful of college friends trying to make their AI game engine work. The models could make functional games, but none of the games were fun — which raised the question of how you tell if a game will be fun. There was no substitute for human judgment, they decided, and soon they were brainstorming ways to get honest human feedback at scale. The result became Design Arena, an AI tool now used by 5.3 million people around the world. As it turned out, many AI companies were looking for scalable user feedback — and willing to pay for it.

“It was the missing bottleneck for a lot of these models to make improvements in the design space,” Li says. “About a week later, we closed our first major deal with a frontier lab, and the rest is kind of history.”

On Monday, the company behind Design Arena — dubbed Intelligence — announced a $7.9 million seed round led by Index Ventures, with participation from Conviction (Sarah Guo and Mike Vernal), A*, Valkyrie, and others.

For non-enterprise users, Design Arena works like a sophisticated model router: a ChatGPT-style prompt window with separate dropdowns for websites, images, and a dozen other visual formats. After submitting a request, format, and style, users see a series of “A vs. B” choices until they’ve ranked a handful of outputs from best to worst. The real value, though, comes from the enterprise side, where participating models treat the platform as a source of instant feedback for their media-generating models. Because users tend to be indifferent to which models they’re ranking, their rankings give critical input on what users really want — a service frontier labs pay for, Li says, with the site currently generating $60 million in ARR. Because users must log in to see results, Intelligence can also track how tastes change across continents and over time; Li notes that web dashboards in Asia tend toward a more maximalist design style. These measures complement automated benchmarks, which operate at greater scale but can be gamed or manipulated, as the Hugging Face breach demonstrated last week.

Crowdsourced human feedback isn’t an automatic winning market, though: Yupp shuttered less than a year after launching, after raising $33 million from a16z crypto’s Chris Dixon and signing on frontier-model customers with over 1.3 million users. Other human-evaluation startups are thriving, though — LM Arena, which takes a similar approach for text-based responses, raised $150 million in a Series A in January, four months after launching its paid product.

Why it matters

As automated AI benchmarks prove easier to game, paid human-preference data is becoming a distinct revenue line for frontier labs’ suppliers — though Yupp’s shutdown shows the category doesn’t guarantee a sustainable business.