Arcada Labs
Arcada Labs builds 'portals' to bridge AI in the real world, focusing on distilling and understanding the human experience by turning subjective human traits like taste and aesthetic judgment into measurable data.
At a Glance
- AI model developers and labs
- Software developers and frontend engineers
- UI/UX designers
- AI companies building generative design tools
- +4 more
AI Tools by Arcada Labs
(1)
Design Arena
Crowdsourced AI Benchmarks
Discussions
No discussions yet
Be the first to start a discussion about Arcada Labs
Latest News
Launch HN: Design Arena (YC S25) – Head-to-head AI benchmark for aesthetics
Design Arena Launches: #1 Benchmark for AI Design - 47,000 users in first 4 weeks
Gemini 3.0 Pro takes #1 overall on Design Arena - biggest performance delta since launch
GPT-Image-1.5 claims #1 on Image Arena with 1344 Elo score
Products & Services
Flagship product - the world's largest crowdsourced benchmark for AI-generated design. Evaluates how AI models perform on real-world design tasks including frontend/UI, images, video, and audio using live organic users to measure taste, usability, and aesthetics. Uses head-to-head matchups with Elo-style ratings based on real user votes. Includes specialized arenas for Website design, Image generation, Video, Logo, SVG, Image Editing, Game Dev, 3D Design, Builders, Web Apps, Text-to-Speech, Data Viz, UI Components, Video to Video, Image to Image, Slides, Voice Chat, and Graphic Design.
A platform that evaluates whether AI models can predict the future. Powered by Kalshi, it features AI agents trading in prediction markets on real-world outcomes including financial indices (S&P 500, Nasdaq-100), cryptocurrency prices, weather, economic indicators, politics, and entertainment. Models provide reasoning for their strategies and execute trades with tracked portfolios, PnL, and performance metrics like Sharpe Ratio.
Standardized and proprietary agent harness for evaluating AI agents
Text to image and image to image AI model benchmarking, integrated within Design Arena
Market Position
First and largest crowdsourced benchmark for AI-generated design aesthetics. Differentiated by focusing on subjective human qualities (taste, aesthetics) rather than just functional correctness. Provides a 'forcing function' for AI labs to improve design capabilities. Uses real organic users rather than AI proxies for human judgment. Notable finding: Agentic tools not specifically marketed for design (like Devin) outperform dedicated design tools in certain categories.
Leadership
Founders
Grace Li
CEO; Harvard Computer Science and Neuroscience graduate (Class of 2025); Previously Software Engineer at Apple working on Photos, Memory Creation, and Apple Intelligence; Selected to demo live to Craig Federighi (CTO of Apple); Best friends with co-founders from Harvard
Kamryn Ohly
CTO; Harvard Computer Science and Education graduate (Class of 2025); Previously Software Engineer at Apple; MLH Top 50 Hackers 2021; Best friends with co-founders from Harvard
Jayden Personnat
Chief AI/ML; Harvard Computer Science and Statistics graduate; Previously worked at Nvidia; Best friends with co-founders from Harvard
Executive Team
Grace Li
Co-Founder & CEO
Harvard CS and Neuroscience (2025); Previously SWE at Apple on Photos, Memory Creation, and Apple Intelligence
Kamryn Ohly
Co-Founder & CTO
Harvard CS and Education (2025); Previously SWE at Apple; MLH Top 50 Hackers 2021
Board of Directors
Founding Story
The three founders are best friends from Harvard who initially started building an AI game engine for one-shot games. They pivoted after discovering that benchmarking visual 'taste' and aesthetics was more exciting to users than their original concept. They interviewed for Y Combinator the week of their Harvard graduation in June 2025 and were accepted into the Summer 2025 batch. The founders recognized that AI design tools were in an 'uncanny valley' territory and that state-of-the-art models often fail at basic design tasks.
Business Model
Revenue Model
Planning to offer 'version testing as a service' to companies needing to quantify aesthetic improvements in their products between builds. Free public benchmarking platform with potential enterprise/private evaluation services.
Target Markets
- AI model developers and labs
- Software developers and frontend engineers
- UI/UX designers
- AI companies building generative design tools
- Product teams needing aesthetic validation
- Researchers studying human taste and preference
- Benchmarking AI model performance on design and aesthetics
- Comparing visual capabilities of different AI models
- Discovering which AI is best for specific tasks (landing pages, 3D models, logos)
- Version testing as a service for companies needing to quantify aesthetic improvements
- Pressure-testing AI model usability and aesthetics at scale
- Evaluating AI prediction capabilities in financial and forecasting scenarios