SOM, a lightweight Smalltalk dialect, has become a staple in virtual‑machine research and teaching. Originating at the University of Århus in 2001, it evolved into variants like SOMNS and SOM++ that support actor concurrency and generational garbage collection. Over the past two decades, universities and research labs have adopted it for experiments and coursework today.

In 2010, Michael Haupt used SOM to teach a graduate course at Technische Universität Darmstadt, while earlier semesters at the Hasso Plattner Institute and Lancaster University showcased its pedagogical value. The language’s modularity enabled extensions such as TruffleMATE, a variant that incorporates the Mate Metaobject protocol for fully reflective execution in research sessions across multiple institutions.

Research teams have leveraged SOM variants for performance studies. Stefan Marr’s 2015 work compared tracing and partial‑evaluation meta‑compilation, while 2018 papers explored deterministic record & replay for actor languages and concurrency‑agnostic debugging protocols. The 2017 Grace implementation, dubbed Moth, demonstrated that a SOMNS foundation can host a full‑featured language with minimal overhead in real‑world environments.

Because SOM’s source is open and easily modifiable, it remains a testbed for VM adaptation research. Recent work on virtual‑machine product lines and type harvesting illustrates its flexibility. As a result, developers and educators continue to cite SOM as a baseline for building lightweight, experiment‑ready runtimes that blend language design with efficient execution in academic settings.