Even just reading through an existing implementation can be very enlightening. E.g. JonesForth (where like 90% of the assembly source is comments explaining everything in detail): https://github.com/nornagon/jonesforth/blob/master/jonesfort...

I really liked JonesForth, and I ported it to powerpc about a decade ago (maybe longer, I don't unfortunately have it around anymore to check, nor a machine to run it on). One thing that was very frustrating was finding out that the macro system was much more limited under darwin's assembler—this resulted in a painful translation process and too much M4 usage. If you do go down this route, I highly recommend fully exploring the capabilities of the assembler's macro system before diving in.