Building or extending Virtual Machines (VMs) to investigate new language features or optimization techniques is challenging in several ways. The overhead for developing a new research VM for an existing practical language is immense, and meaningful evaluation often requires implement- ing much more than just the parts that are interesting for the research question.

In this paper, we propose a different approach for implementing VMs based on wrapping an existing, feature complete VM. Our technique aims for lower implementation overhead by reducing the number of features that have to be implemented to produce a working prototype and thus produces results quicker. While already proving useful for research, our approach also suggests a way to extend legacy virtual machines with new features and optimizations.