I agree. The more I learn about the Mill the more I feel that trying to shoehorn Forth’s stack-based approach into a Mill environment would be frustrating and not very productive. It would only be useful if one wanted to use existing Forth libraries; not my intent.
So what would a higher-level language look like coming from the Mill-style of computing? The immutable values remind me of Functional languages. Items falling off the belt remind me of automatic GC. I did think about writing an interpreter for the Mill Instruction set in Haskell, paramaterized of course by the family-member specifics.