Moonlake

The simulation and world modeling infrastructure for robotics

[Moonlake @moonlake]

Introducing Moonlake's 3D Agent.

Our agent acts like a technical artist that can build and reconstruct articulated assets and large-scale editable scenes with hundreds of objects from a single image and can improve its generations continuously.

Learn more in the thread below.

[Sharon Lee @sharonal_lee]

Been seeing a lot of Astra for real2sim, but what you need for robotics: physics and affordance.

Our SOTA recipe specifically solves for:
• Affordance: model object details that convey functionality.
• Physics: calibrate simulation dynamics against real-world observations.

Our agent has the action space of a “full-stack simulation engineer” and jointly considers geometry, materials, colliders, articulation, and physics. It automatically calibrates physics against real trajectories and matches task-specific dynamics.

Simulation environments is useful only if it can provide meaningful policy level signals. Methods, comparisons, and results below.

[Sun, Fan-Yun @sunfanyun]

Cohosting an event at IROS with @PathRobotics, @AGIBOTofficial, and @FaradayFuture on 9/29!

If you’re looking to put more 🤖 to work, we’d love to meet! Limited space, sign up here: https://t.co/OIeCQdE9Xa

[Moonlake @moonlake]

Our team shared the mission of building towards embodied general intelligence at AGI Summit.

@pmarca: software will eating the world.
@moonlake: verifiable world model will eat the physical world.

[AGI Summit @agisummitai]

🌐 Just wrapped this morning at AGI Summit—happening live now in San Francisco!

“Beyond Language: World Models, Spatial Reasoning, and Embodied Intelligence” featured Fan-Yun Sun, Christopher Manning, and Christian Laforte, moderated by Mercedes Bent.

Stay tuned for more live from the summit. 🔴

@sunfanyun @chrmanning @chrlaf @mercebent
#AGISummit #WorldModels #SpatialReasoning #EmbodiedIntelligence #PhysicalAI #SanFrancisco

[AUTONOMOUS @autonomousevent]

The sessions end at 5. The conversations continue long after.

@moonlake is hosting the official AUTONOMOUS after hours party in SF directly after the event.

Spots are limited: https://t.co/1IVCu5aFFN

When one room closes, another opens. Exclsuive access just for #AUTONOMOUS2026 attendees.

[Moonlake @moonlake]

"We are producing controllable and manipulatable models of the world which we can then use to assess how good our render is against physical reality, work out the deviations, revise the code and make better and better representations"

[Sun, Fan-Yun @sunfanyun]

@chrmanning is sharing our approach to embodied general intelligence at AI Engineers!

Room 2006

[Moonlake @moonlake]

@sunfanyun @nvidiaomniverse 🫡

[Sun, Fan-Yun @sunfanyun]

Excited to share that we are working with Nvidia to bring @moonlake's 3D Agent capabilities into NVIDIA's simulation ecosystem (@nvidiaomniverse) to help teams build assets, digital twins, and scenarios at scale.

Here are 8 pairs of input images and the generated simulation-ready scenes.

Read more here: https://t.co/XMRz23Lb8m

[Moonlake @moonlake]

We are happy to announce that Moonlake AI has been named on the 2026 @CBinsights AI 100 list.

One of the 4 on infrastructure and world models. Read more here: https://t.co/gXpT7IqgDM

[Moonlake @moonlake]

@vokusz https://t.co/sBLh1ab2K0

Can try this!

[Moonlake @moonlake]

3D reasoning and modeling capabilities represents a 100B+/year category of work in simulations, games, and film/VFX because the verification process depends on having spatial, physical, and geometrical priors of the world.

Read more here: https://t.co/ViI4AkFmjZ

[Moonlake @moonlake]

Introducing Moonlake's 3D Agent.

Our agent acts like a technical artist that can build and reconstruct articulated assets and large-scale editable scenes with hundreds of objects from a single image and can improve its generations continuously.

Learn more in the thread below.

[Moonlake @moonlake]

Our agent can integrate natively with your existing pipelines; connect it to your digital asset management system and launch hundreds of jobs in a click. Don't spend hours on checking object states and naming or asset cleanup. Save time for what matters.

If you'd like to use our agent, reach out to us at [email protected]

[Moonlake @moonlake]

A deep dive into our latest work on controllable world models at Moonlake.

This Wednesday 5pm at Stanford.

[Stanford AI Club @stanfordaiclub]

Pumped to be hosting @sunfanyun from @moonlake this Wednesday at 5pm for our next Speaker Series!

This event will be available to Stanford students only, but will be recorded & posted to our Youtube channel afterwards.

[Christopher Manning @chrmanning]

It was great to talk with @swyx and @sunfanyun on @latentspacepod about world models.

🔥 Below I give a few more long-form thoughts on what is needed for successful world models. 🧵👇

https://t.co/QEEm8RKiWc