A unified GPU programming system written in C++, aimed at rapid experimentation and prototyping with complex GPU workflows.

Shaders and CUDA/OptiX programs coexist as nodes in a data-flow graph, enabling flexible composition, complex feedback loops, and cross-domain data flow between rendering and compute stages.

The system reduces the structural and boilerplate overhead of GPU workflows that are conceptually straightforward but traditionally tedious to set up—such as temporal volumes that record GPU frames over time, where the logic is simple but buffer management and feedback wiring dominate the implementation.