ecto

ecto / phyz

Public

differentiable physics engine for rust

17
2
100% credibility
Found Feb 23, 2026 at 12 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Rust
AI Summary

phyz is an open-source differentiable multi-physics simulation engine in pure Rust with GPU acceleration for rigid bodies, particles, electromagnetism, and more.

How It Works

1
🔍 Discover phyz

You stumble upon phyz.dev and watch mesmerizing videos of robots swinging, fluids splashing, and physics coming alive just like the real world.

2
🖱️ Play with demos

Click around the interactive playgrounds to tweak pendulums or robots and see instant changes without writing any code.

3
🏗️ Build your robot

Drag and drop parts like arms and joints to assemble your first robot model, feeling like a kid with building blocks.

4
▶️ Watch it move

Hit play and see your creation swing, bounce, or walk with realistic physics, adjusting speeds or forces on the fly.

5
Speed it up

Switch to super-fast mode to simulate thousands of robots at once, perfect for training smart behaviors.

🎯 Train smart robots

Your assistant learns to balance or catch objects perfectly by itself, ready for real-world experiments.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 12 to 17 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is phyz?

Phyz is a Rust-based differentiable physics engine that runs multi-physics simulations—rigid bodies, SPH fluids, electromagnetism on Yee grids, even Regge calculus for general relativity—all with analytical Jacobians for gradient-based optimization. Load MJCF models, batch thousands of GPU-accelerated envs via WGPU, and get derivatives through full trajectories for controller tuning or parameter fitting from video. It turns physics into a true differentiable function, skipping noisy finite differences.

Why is it gaining traction?

Unlike standard engines, phyz delivers end-to-end differentiability across domains like differentiable physics simulation of dynamics augmented neural objects or differentiable physics enabled closure modeling for burgers turbulence, with kernel fusion slashing GPU memory use. Rust ensures zero-cost safety at C++ speeds, while MJCF import and probabilistic inference (SVGD/HMC) hook ML workflows. Devs dig the one-loop sim+grads for differentiable physics and stable modes for tool use and manipulation planning.

Who should use this?

Robotics researchers optimizing manipulation planners with differentiable physics models for real-world offline model-based reinforcement learning. Audio engineers reconstructing sound fields via differentiable physics for sound field reconstruction. Sim hackers blending EM/rigid for Lorentz-coupled actuators or differentiable ray tracing github prototypes.

Verdict

Promising differentiable physics engine at 1.0% credibility (11 stars, solid tests, thin docs)—use for proofs-of-concept now, but wait for benchmarks and examples before prod. Fork-friendly MIT for custom physics tweaks.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.