Just to get some early people kicking the tires, the initial target audience is people who want to make weird things out of programming LEGOs.
Though it's my belief that once people get the feel for it as a curiosity, they'd see the greater potential.
Longer term, the goal is more or less the original one:
It's supposed to be for solving problems in a literate way...using the same kinds of mental processes we use when talking/writing...with a very shallow dependency footprint for the entirety of the toolchain. People should be able to page through the interpreter sources and say "oh, yes, this makes sense".
And it has to be on the scale of something like a Lua or similar... relatively small... but the idea is to have truly next-level expressivity and notable design that elevates it into a different category.
Red and Arturo have barely any magic that any other garden-variety language doesn't have--just the sort of hint of freeformness Rebol's permissive source notation offers. A capability which winds up being difficult to wield in interesting ways, to where if you're going to have to invent everything yourself you might as well go the whole way and use lexical tooling to define the exact syntax you want, vs. bending into their limited notions of what a Redbol syntax should support.