●builderYou can test whether your coding agents can build interactive, physically-correct environments rather than just static visuals.
●researcherThis establishes a new standard for evaluating the causal and mechanical reasoning capabilities of AI agents.