I've always been a bit sad that more Forth-inspired DSLs never became a thing. Assembler Forths and C Forths often feel like DSLs on top of their host languages/environments, so doing the same thing as a network or build-system DSL would be interesting. I do recommend folks play around with more mature, big Forths like gForth to see what building an application up in a Forth like system actually feels like.
Commercial systems are also now available for free for trial or non-commerical use, that offer native code compilers and IDEs and Graphic interface tools.
How about Forth as a compilation target for other languages. But you can dig in and define words for common patterns and the higher level code will be compiled down to that custom vocabulary instead.
Forth is just not very easily optimized - it has a lot of indirect jumps that do not play well with today's instruction pipelines.
There are a few optimizing forth compilers, but I never have seen compelling evidence that they would be faster than an optimized c compiler.
Forth also has seen comparatively little work lately - only a small number of people use it (for better or worse). Micropython might be eating Forth's lunch on many embedded systems that have a lot more memory/ resources now.
