We can just put all "builtins" into a file, then ship together with SLC and let it use them.

I think this is cleaner than adding syntax that makes the parser and interpreter just unnecessarily more complex...

So there does not need to be the following 4 variations:

: regularword
:: compileword
:builtin regularword
::builtin compileword

^^ if I somehoe let the interpreter know what the builtins are and define them all together in a differen parser state...

So the parser would in that mode prefix everything with the prefix - and in the symbol table of course enter the real C builtins
with the prefix too. Also in the calls / code of these, we would prefix everything to look up well in the symbol table so the code
for builtins can be totally unprefixed and all.

This really helps to simplify the parser-statemachine in my opinion - also the interpreter state machine should be unaffected!