Soft Bodies, Hard Physics: Evolving Digital Organisms in CUDA

What it does ALIEN is a 2D artificial-life simulator built entirely on CUDA. It models soft bodies, fluids, and heat as particle networks, then layers on genomes, cell constructors, and neural networks so the particles self-organize into digital organisms that replicate and adapt. You can interact with the simulation in real time, design organisms by hand, or let mutation and selection do the work.

The interesting bit The project treats evolution as both a research tool and a toy. It won the ALIFE 2024 Virtual Creatures Competition, yet the README is equally honest about the fun of smashing thousands of machines with your mouse cursor. That duality—serious open-ended evolution research wrapped in a polished, real-time “god game” shell—is rare in academic simulators.

Key highlights

• Entire simulation runs on the GPU via CUDA, handling millions of particles in real time with OpenGL interoperability for rendering.
• Organisms are built from particle networks controlled by neural networks, with blueprints stored in genomes and inherited by offspring.
• Includes a built-in simulation browser for uploading, downloading, and upvoting worlds, plus extensive editors for genomes, particles, and geometry.
• Ships with a headless CLI for batch runs, performance testing, and automated parameter sweeps outputting CSV statistics.
• Aims to study pre-biotic evolution and ecosystem dynamics, but doubles as a generative art engine.

Caveats

• Requires an NVIDIA GPU with compute capability 6.0 or higher, and the same card handles both CUDA compute and rendering, which can cause crashes if multi-GPU setups are misconfigured.
• Documentation for the previous major version is outdated; current guidance lives mostly inside the program’s tooltips and help windows.
• Known build failures with recent toolchains: GCC 12+, Visual Studio 17.10, and CUDA 12.5 are currently problematic.

Verdict Grab it if you want to experiment with artificial life or need a visually rich physics sandbox that runs entirely on the GPU. Skip it if you lack a recent NVIDIA card or need a mature, well-documented API to script against; the focus is on interactive simulation, not library integration.