The DSL approach to Lisp programming was mainstream in the 80's. It comes natural to any system that has enough frictionless syntactic plasticity and expressiveness.
These days it still exists in Racket (and possibly others), see language oriented programming ideas.
I would like to add though, that "language" does not have to mean macros and DSL. It can also simply mean what kind of things one considers primitives for the problem to solve. Those primitives can be built in already, they can be user defined functions (or words) or macros or whatever else makes an identifier or thing one can reference later when solving the actual problem.
Agreed that the semantic of language-oriented is sometimes hard to disentangle from the aesthetic of "parsing and compiling". In a majority of cases you can go part of the way in that direction by overlaying a FSM, stack machine or a trivial string parser onto whatever feels cluttered and obnoxious to keep track of.
If you want to do the parsed exterior DSL thing today, the magic wand that probably does it the best is Raku and its souped-up grammar system. I actually am exploring that now and I think it has a good toolkit for prototyping anything resembling a textual interface.
