We present a preview of HVM2, an efficient and massively parallel GPU and CPU evaluator for a system of extended Interaction Combinators. When compiling non-sequential programs from a high-level programming language to HVM2 and executing them on the C and CUDA runtimes, we achieved a large parallel speedup as a function of cores available. We give an overview of HVM2’s theoretical foundations, implementation, early benchmarks, current limitations, and future work.