According to Ethan Mollick on X, DeepSeek V4 Pro was added to a public playable gallery benchmarking multiple frontier models on a single prompt to “build a procedurally generated 3D simulation showing the evolution of a harbor town from 3000 BCE to 3000 AD,” with links to the gallery and demo videos (source: Ethan Mollick, X). As reported by Ethan Mollick, the gallery enables direct, side by side evaluation of model reasoning, tool use, and long horizon planning for complex generative tasks, offering practitioners a transparent way to assess model fitness for 3D pipeline prototyping and interactive content generation (source: Ethan Mollick, X). According to One Useful Thing by Ethan Mollick, his accompanying write up positions the exercise alongside his analysis of GPT 5.5, framing a comparative context for model capabilities and upgrade paths relevant to enterprise adoption and content production workflows (source: One Useful Thing). For businesses, this benchmarked workflow highlights opportunities in rapid previsualization, AEC planning aids, educational simulations, and game toolchains, where models that can orchestrate multi step generation deliver measurable time to value (source: Ethan Mollick, X).

According to Ethan Mollick on X, multiple foundation models were prompted to “build a procedurally generated 3D simulation showing the evolution of a harbor town from 3000 BCE to 3000 AD,” with an interactive gallery published at hg-20f7d1a3ce.netlify.app and a detailed write-up on GPT-5.5 on One Useful Thing. According to One Useful Thing, the test highlights differences in long-horizon tool use, multi-step code generation, and spatial reasoning required to synthesize geometry, materials, and time-based events into a single runnable experience. As reported by Ethan Mollick, single-prompt performance exposes practical strengths in code reliability, asset orchestration, and runtime debugging—key business factors for teams shipping generative 3D content and simulations. According to the linked gallery, the comparison provides concrete evidence of which models better handle procedural generation pipelines end to end, informing buyers on model selection for game prototyping, digital twins, and historical visualizations. According to One Useful Thing, GPT-5.5 is analyzed for its improved reasoning and tool-use consistency, suggesting reduced engineering overhead for production workflows in 3D generation, though results vary by task and environment.

According to Google DeepMind, SIMA 2 was evaluated within 3D virtual worlds generated by the Genie 3 world model, demonstrating unprecedented adaptability in navigating complex digital environments and taking strategic actions toward defined objectives (source: Google DeepMind Twitter, Nov 13, 2025). This advancement highlights significant progress in reinforcement learning, environment simulation, and real-world AI application potential. Such capabilities present new business opportunities for industries seeking adaptive AI agents for simulation, training, and autonomous virtual interactions.

According to Google DeepMind, SIMA 2 was rigorously tested in simulated 3D environments created by their world model Genie 3. The AI agent demonstrated unprecedented adaptability by efficiently navigating complex surroundings and taking meaningful steps toward achieving its objectives. This breakthrough highlights significant advancements in AI-driven simulation, paving the way for practical applications in robotics, digital twins, and autonomous systems across industries. The results underscore the growing potential for AI to handle real-world tasks in dynamic, unpredictable environments (source: Google DeepMind Twitter, Nov 13, 2025).