Design Arena maker Intelligence raises $7.9M seed at $60M ARR

A seed round is usually the moment a company shows a pitch deck and a demo. Intelligence, the startup behind Design Arena, is closing one with a reported $60 million in annual recurring revenue already on the books, which is the number worth staring at before anything else.

TechCrunch reports that the company raised $7.9 million led by Index Ventures, with Conviction (Sarah Guo and Mike Vernal), A* and Valkyrie participating. The product is deceptively simple: 5.3 million users globally show up to pick between two AI-generated outputs, whether that's a website, an image, or another piece of visual media. Those A-versus-B votes are then packaged and sold to frontier AI labs as training feedback for their media-generation models. Co-founder Grace Li describes taste as "the missing bottleneck for a lot of these models to make improvements in the design space," and says the company closed a major frontier lab deal within a week of launching.

The strategic read is that human preference data has quietly graduated from a research input to a priced piece of infrastructure. Text has already been through this cycle. LM Arena, the text-based leaderboard, raised $150 million in a Series A in January 2026. What Intelligence is arguing is that design and media need their own version, because functional output and desirable output are not the same problem, and only humans can bridge that gap at scale.

The honest caveat is that ARR at seed is self-reported, and the enterprise side almost certainly leans on a small cluster of frontier lab contracts. This category has already produced one cautionary tale: Yupp, another human-evaluation platform, shut down earlier in 2026 despite raising $33 million. What the reporting doesn't give us is the shape of Intelligence's contracts, retention on those 5.3 million voters, or how it defends against the same labs eventually building their own taste pipelines the way they absorbed earlier RLHF work.

Still, if the numbers hold, this is a signal about where the next layer of AI value is settling. Compute is commoditizing, models are converging, and the differentiator for anything creative is starting to look like whoever owns the human judgment stream. That is a very different bet than the one most infrastructure investors have been making for the last two years.