What I like about Forth is that it can be expressed at the lowest level of computation, and that it can be used to bridge that to the highest level of computation. For example, Forth only requires about 12 opcodes to run, which can be implemented in a few dozen chips. But now that you have that, since it's Turing-complete, you can now pull across a lisp or C compiler, and build a working operating system from there. Granted, that would be a lot of work, but it's relatively straightforward work, and that's always impressed me.