Hide and Seek
A simple game. An unexpected arms race of strategies.
The project
OpenAI studied agents playing hide and seek in a simulated environment with movable objects. Across training, hiders and seekers developed strategies involving barriers, ramps, and tool use.
Where AI comes in
Multi-agent reinforcement learning and self-play produced increasingly complex behavior without individually scripting each strategy. The environment and competition shaped what the agents learned.
For game makers
A useful reference for emergent gameplay and sandbox design. The game rules and available objects can create surprising strategic depth, even when the starting task is simple.
Go to the source
Emergent tool use · OpenAICode & environments · GitHubResearch paper · arXivThe official page restricts automated access. See the source notes for the supporting reference.
Keep exploring
Explore
Relight: Lantern Expanse
Fly between worlds, explore their surfaces, and relight the way home.
Hollow Crown
A browser-playable pixel adventure, with AI helping build and demonstrate the game.
Wildbrush
A family-made adventure with paintbrush combat and floating islands.