Description
ARC Arcade was conceived as a next step for the ARC Prize: a gamified platform designed to explore the boundary where problems remain intuitive for humans but challenging for AI systems.
The ambition was to move beyond static benchmarks and create a living environment where users could solve ARC-AGI tasks, build personal histories, and contribute to a growing ecosystem of difficulty and performance data. At its core, the platform aimed to make abstract research engaging, measurable, and comparable over time.
The MVP focused on laying strong foundations. We designed a task-solving environment with user accounts, basic profiles, and detailed tracking of both task and user performance. Each solved task contributed to an evolving difficulty profile, while users built up histories based on completion time, success rate, and consistency.
The experience was intentionally simple and functional, prioritizing clarity, speed, and focus. This allowed the system to scale toward more advanced features such as task creation, ELO-style rankings, leaderboards, and eventually automated AI submissions and analytics.
Unfortunately, the project did not go live. A change in direction within the ARC Foundation led to a shift in priorities, and ARC Arcade was paused before public launch.
Credits
Creative Direction
Product Design
Motion Design
3D Design
Sound Design
Engineering
Amine Diboune, Julián David
Case Study Design
The content of this page reflects the state of the project during Tinloof's engagement.