One AI Model Now Simulates Sand, Cloth, Fluids and Folding Proteins

Photo: Hippopx (public domain). Illustrative image.

What happened. Researchers at MIT, the University of British Columbia and partner institutions have built a single AI model that simulates six kinds of physical behaviour — sliding sand, deformable objects, fluids, cloth, and folding proteins among them — where each previously needed its own dedicated solver. The system, called WorldParticle, is described in a paper published in ACM Transactions on Graphics and announced by MIT News on 9 November 2026. It will be presented at the ACM SIGGRAPH Conference and Exhibition on Computer Graphics and Interactive Techniques in Asia.

How it works. WorldParticle represents every material it simulates, from a dune of sand to a protein molecule, as thousands or millions of tiny three-dimensional particles, using what is known as a Lagrangian representation. A neural network takes the particles’ positions and physical properties as inputs and predicts where they will be in the next frame, one second at a time. It is, as co-corresponding author Peter Yichen Chen puts it, “a similar architecture to ChatGPT. But instead of predicting one word at a time to write a piece of text, we are predicting how these particles will move, one step at a time.”

The crucial difference from traditional simulation is that nothing is programmed with a governing equation. Normally, scientists must convert the equation that defines how a material behaves into a formula a numerical solver can handle — repeating those steps for each material and process. WorldParticle’s neural network instead learns to solve governing equations in general, eliminating the need to design a new model and solver for each one. Before each step, the system separates out external forces — like the impulse when an ocean wave slams into a seawall — then integrates them back in after predicting the particles’ initial states, correcting the prediction for how those forces affect the object.

Why one model matters. “If you study deformable objects and fluids, you would need two simulators — one for each physical regime. But if you unify them into one model, maybe you could couple the physics of these two processes and gain insights into new problems?” said co-lead author Minghao Guo, an MIT postdoc. The researchers drew inspiration from world models: AI systems that learn the rules governing the physical world so they can predict how objects change over time. Their vision is a foundation model for simulation. “We want to have a GPT moment for simulation,” said Chen, a former MIT postdoc who is now an assistant professor at UBC. “Every time people want to build a simulator for a particular physical phenomenon, they need to code it up from scratch. Our vision is to have a foundation model that works out of the box for any scenario you want to simulate. This is a first step in that direction.”

What it demonstrated. The team showed WorldParticle modelling six different physical phenomena with higher accuracy than numerical simulation methods. It captured how a granular object collapses into a pile while some baseline methods left sand particles floating in the air, and it modelled how protein structures change over time about 100 times faster than numerical simulators. Handling millions of particles is computationally demanding, so the researchers used a merging technique to combine and compress particles while still capturing fine-grained behaviour.

Who built it. Chen and Guo are joined by co-lead authors Caoliwen Wang and Siyuan Chen, UBC graduate students; Wojciech Matusik, an MIT professor of electrical engineering and computer science who leads the Computational Design and Fabrication Group within MIT’s Computer Science and Artificial Intelligence Laboratory; and collaborators at Georgia Tech, Inria, Meta, the University of Pennsylvania, the University of Utah, the University of California at Los Angeles and the University of Hong Kong.

What comes next. The researchers want to adapt WorldParticle to more physical systems — soft robot manipulation, RNA evolution, and astrophysical phenomena like the merging of galaxies among them — and build toward a full, general-purpose simulation pipeline for scientific discovery. “We see the unification of computation and science giving us new insights and tools to discover physical phenomena or explain some unexplainable physical phenomena,” Guo said. “With a unified model, we could unlock many discoveries that we can’t even imagine.”

Sources

• MIT News — “A single AI model simulates the behavior of many substances, from sliding sand to folding proteins” (9 November 2026): https://news.mit.edu/2026/single-ai-model-simulates-many-substances-behavior-1109
• Wang et al. — “WorldParticle: Unified world simulation of Lagrangian particle dynamics via transformer”, ACM Transactions on Graphics (arXiv:2605.15305): https://arxiv.org/pdf/2605.15305