Moonlake AI, an applied research lab founded by former Stanford AI Lab and NVIDIA researchers, has launched from stealth with a $28M seed round.
The raise was led by Threshold Ventures, AIX Ventures, and NVIDIA Ventures, with participation from more than a dozen top AI founders and researchers, including YouTube founder Steve Chen, AngelList’s Naval Ravikant, GANs inventor Ian Goodfellow, and Google Research Chief Scientist Jeff Dean.
Moonlake is building reasoning models that enable anyone—developers, designers, or fans—to create interactive worlds in minutes using natural language.
Lowering the barrier to world-building
Moonlake AI has developed the first reasoning model for real-time interactive content, capable of compiling natural language prompts into playable 2D and 3D worlds. Users can instantly edit physics, add interactive logic, or reskin environments.
Under the hood, the platform combines multi-modal reasoning for spatial layout, program synthesis for logic, simulation layers for agents, and real-time diffusion for rendering. This makes it equally powerful for game prototyping, robotics simulations, or training reinforcement learning (RL) agents.
“Moonlake is marrying vibe coding and world generation to create the ultimate reinforcement learning environments,” said Mo Islam, Partner at Threshold Ventures. “This unlocks entirely novel forms of agents in both the digital and physical worlds.”
Advertisement
A bold vision for interactivity
The startup wants to solve a fundamental bottleneck in creative industries: building interactive content is slow and costly. Today’s blockbuster games can cost billions and take years, while smaller studios and indie creators struggle to experiment.
Moonlake’s platform aims to democratize interactive world creation for:
- User-generated content and fan-driven micro-worlds
- Video game studios and IP holders
- Robotics and embodied AI training
- 3D environments for film and animation
- Educators building immersive classrooms
“Too many great worlds never made it past the sketchpad—not for lack of imagination, but for lack of the right tools,” said Sharon Lee, co-founder of Moonlake AI.
From the lab to the real world
Moonlake AI was founded by Fan-Yun Sun (ex-NVIDIA, Stanford PhD in CS) and Sharon Lee (Stanford PhD researcher in generative AI and computer graphics). Their work combines diffusion models, 3D engines, and spatial reasoning to give both humans and AI agents the ability to interact with rich, dynamic environments.
“Simulations and games are an extension of our ideas,” said Fan-Yun Sun. “We want anyone to sketch a feeling or mechanic and simulate it within minutes. That unlocks new creative workflows and gives AI agents the environments they need to learn.”
Early access and next steps
Currently in private preview, Moonlake AI is already working with early creators and research partners. Interested developers and studios can request access at moonlakeai.com.
“YouTube unlocked video creation for everyone. Moonlake has the potential to do the same for interactive content,” said Steve Chen, angel investor and founder of YouTube.
With its $28M seed round, Moonlake is positioning itself as a new creative and technical layer—where world-building becomes as simple as describing a vision.
