DesignArena began as an attempt by college friends to build an AI game engine a few weeks before graduation in 2025. The model-made games worked but were not fun, which led co-founder Grace Li and her team to ask how anyone could reliably judge fun without human input. Their answer became DesignArena, a tool now used by 5.3 million people, where users rank AI-generated design outputs through A vs. B choices.

The user-facing side resembles a sophisticated model router: a ChatGPT-style prompt window with dropdowns for websites, images, and a dozen other visual formats. Users submit a request, choose format and style, then rank the resulting outputs from best to worst. For non-enterprise users, that ranking is a useful service. For enterprises, it is a source of continuous, scalable human feedback for media-generating models.

Because users are mostly indifferent about which model they are ranking and simply want the best output, their preferences provide direct signals on what people actually want in AI-generated designs. Li says this was the missing bottleneck for many AI models trying to improve in the design space. The company behind DesignArena, called Intelligence, closed its first major deal with a frontier lab about a week after pivoting, and the site is now generating $60 million in annual recurring revenue.

The platform requires users to log in to receive outputs, which lets Intelligence track how tastes change across continents and over time. Li notes, for example, that web dashboards in Asia tend toward a more maximalist design style. These human evaluations are positioned as a complement to automated benchmarks, which can operate at larger scale but are vulnerable to gaming and manipulation, as demonstrated by the Hugging Face breach.

Still, the viability of crowdsourced human feedback as a market is not guaranteed. Yupp, a similar startup, shut down earlier this year after raising $33 million from a16z crypto’s Chris Dixon, despite having frontier model customers and over 1.3 million users. In contrast, LM Arena, which applies a similar approach to text-based responses, raised $150 million in a Series A in January, only four months after launching its paid product. Intelligence‘s seed round of $7.9 million, led by Index Ventures with participation from Conviction, A*, Valkyrie, and others, reflects continued investor appetite for human-led evaluation data.