MIT's WorldParticle uses GPT-style next-step prediction to simulate sand, water and protein folding
thisdudelikesAI · x · 2026-11-11
MIT researchers unveiled WorldParticle, an AI simulation system that works like ChatGPT's next-word prediction — but for physics. It breaks objects into thousands to millions of particles and iteratively predicts where each particle moves next, frame by frame.
- Replaces the old paradigm where every substance (sand, water, etc.) needed its own hand-written expert simulator; one model now covers sandpile collapse, liquid hitting a flexible wall, molecular motion, and protein folding
- Protein folding simulation runs 100x faster than standard simulators
- Paper headed to SIGGRAPH Asia
- Peter Yichen Chen (formerly MIT, now faculty at UBC) calls the goal "a GPT moment for simulation"
The team says the system could aid development of soft robots and medicines.
Related event: MIT's WorldParticle Simulates Sand, Fluids and Proteins with One AI Model(3 posts)→
More from Research
- David Bau's Cornell Tech talk on interpretability and introspection now on YouTube — yoavartzi · 2026-11-11
- Claude Opus 5.5 runs an architecture search and finds AttnRes-Canon, a new top Transformer design — SonglinYang4 · 2026-11-11
- 5 research habits worth knowing: read papers backwards, reproduce first, wire coding agents via MCP — cneuralnetwork · 2026-11-11
- Running Qwen2.5-Coder-3B on Android: Jupyter + Pydantic Evals on-device eval pipeline — alex_b3b · 2026-11-11
- cloneofsimo releases a video explaining determinantal point processes — cloneofsimo · 2026-11-11
- Mapping Qwen3-8B's internal touch body map into an anatomical plate — repligate · 2026-11-11