Intelligence nets $7.9M for Design Arena, a 5.3M-user feedback engine already pulling $60M ARR
Curated by the Inblix editorial team
Grace Li and her co-founders stumbled into a massive business by failing to make a fun video game. Weeks before graduating in 2025, their AI game engine spit out functional but boring titles, forcing a hard question: how do you even measure fun at scale? The answer became Design Arena, a platform now used by 5.3 million people that lets AI labs tap into endless human preference data. The company behind it, Intelligence, just closed a $7.9 million seed round led by Index Ventures, with Conviction, A*, and Valkyrie also participating.
For the average user, Design Arena works like a sophisticated model router with a ChatGPT-style prompt window. You request a website mockup, an image, or one of a dozen other visual formats, then rank a series of A-vs-B outputs until you’ve sorted them best to worst. It’s useful, but the real money is on the enterprise side. Frontier AI labs pay for access to this river of human judgment because the users don’t care which model wins — they just want the best result. That indifference is exactly what makes the rankings so valuable for training and fine-tuning media-generation models. Li says the site is currently generating $60 million in annual recurring revenue.
Crucially, the platform requires a login, so Intelligence can track how aesthetic preferences shift across continents and over time. Li points to a concrete example: web dashboards in Asia tend toward a more maximalist design style. This kind of granular, demographic-sliced feedback is something automated benchmarks simply cannot replicate. The need for human evaluation data has only grown sharper after last week’s Hugging Face breach demonstrated just how easily automated tests can be gamed or manipulated.
Not every startup in this space has figured out the economics. Yupp, which raised $33 million from a16z crypto’s Chris Dixon, shut down less than a year after launch despite claiming over 1.3 million users and some frontier lab customers. The comparison stings, but Intelligence’s $60 million ARR figure suggests a very different trajectory. On the text side, LM Arena raised $150 million in January just four months after launching its paid product. The market for human-led evaluation appears to be consolidating around a few players who can actually convert usage into revenue — and for now, Intelligence looks like one of them.
💡 Key Takeaways
- Intelligence's Design Arena has scaled to 5.3 million users and $60 million in ARR by selling human preference data back to the AI labs that need it most.
- Requiring a login lets the company track how aesthetic tastes vary by geography and over time, offering a demographic lens that automated benchmarks completely miss.
- The recent Hugging Face breach underscored the fragility of automated evals, making diverse, hard-to-game human feedback more valuable to frontier labs.
- Yupp's failure after a $33M raise proves traffic alone doesn't guarantee a business, but Intelligence's revenue suggests it has solved the monetization puzzle that killed its competitor.
Keep reading: See related articles below for more coverage on this topic.
Get smarter about AI
The sharpest AI news, curated daily. Delivered free to your inbox.