Design Arena raises $7.9M to train AI with human taste
A scrappy AI startup just raised $7.9M to help models learn what humans actually like...
Intelligence, the company behind Design Arena, has raised a $7.9 million seed round led by Index Ventures to expand its platform for human-led AI evaluation. Design Arena started as a student project by Grace Li and her team, who noticed that AI-generated games lacked fun—a quality only humans could judge at scale. Today, the platform hosts 5.3 million users who rank AI outputs through a ChatGPT-style interface, providing instant feedback on everything from game designs to web dashboards.
The real value lies in its enterprise tier, where AI labs pay for access to this human feedback data. Design Arena is now generating $60 million in annual recurring revenue (ARR), positioning itself as a key player in the AI evaluation market. Unlike automated benchmarks, which can be gamed (as seen in the recent Hugging Face breach), Design Arena’s crowdsourced rankings offer a more reliable measure of user preferences—even tracking regional tastes, such as maximalist design trends in Asia.
- Design Arena by Intelligence raised $7.9M to scale its human feedback platform for AI models, now used by 5.3M people.
- The tool generates $60M ARR by letting users rank AI outputs (e.g., games, images) to train models on real human preferences.
- Enterprise clients include frontier labs paying for instant feedback data, which is harder to manipulate than automated benchmarks.
Why It Matters
AI models finally have a scalable way to learn human taste—bridging the gap between functional and delightful outputs.