3. Also, the speed hit is not as huge as it sounds, as regular reallocation of the token buffer was already necessary to facilitate the language design. It's not really meant to be a fast language anyway despite the low level paradigms. Though that might change when I eventually tackle compilation

2/ Okabe is structured such that it is practically impossible to depth out, however the interpreter's VM was still accruing call depth. Now, while slightly slower due to the regular reallocation of the interpreter's token buffer, it is now infeasible for the VM to segfault from recursion.

1/ The idea was to take the code from anonymous recursive calls, tokenize it, and inject the tokens into the interpretation stream instead of directly interpreting them. This flattens the call structure, limiting recursive depth to the actual nested depth in-lang.