Our mission
ARC Playground brings people together around visual reasoning: solve unfamiliar challenges, design new ones, and help grow a more human-grounded foundation for understanding—and improving—general intelligence.
Examples
See what changed.
Infer
Find the hidden rule.
Generalize
Apply it somewhere new.
The idea
An ARC task gives a solver a handful of input-and-output grids. Together they imply a rule: objects might move, colors might mark roles, or shapes might combine. The solver must infer the abstraction and use it to complete a new grid.
The format is deliberately compact, but the challenge is fundamental: can a person or a machine make sense of something new with only a few examples to learn from?
AI researcher François Chollet introduced the Abstraction and Reasoning Corpus in 2019 as a way to study this kind of flexible generalization. ARC Playground makes that question tangible—one puzzle at a time.
Read the official ARC-AGI overview ↗Why a playground?
The next useful test of reasoning will not come from repeating the same answers. It will come from encountering more ideas—and learning how humans make sense of them.
A fixed collection can hold only so many ideas. A community of creators can keep inventing new shapes, rules, and combinations—broadening the space that people and AI must reason through.
People reveal whether a challenge is intuitive, difficult, ambiguous, or elegant. That human point of reference gives progress in machine reasoning something meaningful to measure against.
A growing stream of varied, human-designed challenges can help AI builders probe where systems generalize, where they fail, and what they should build next.
The ambition
What begins as a visual puzzle can become part of a richer, human-grounded frontier for AI development.
New puzzles introduce unfamiliar rules and fresh perspectives.
Human play reveals what each challenge truly asks of a reasoner.
The growing challenge set can give AI builders new ways to investigate adaptation and failure.
ARC Playground is an independent community project inspired by the public ARC-AGI benchmark. It is not affiliated with or endorsed by the ARC Prize Foundation.
The playground needs both sides: people ready to crack unfamiliar ideas and people ready to invent them.