The biggest problems:
- actually contributing to the interpreter was most of the challenge. I'd rewrite it so that essentially, we work on the specification of a language, not the implementation
- the challenge selections for scoring. ideally, to get a general good golfign lang, you need much better variety of challenges, and potentially even to keep them "hidden" so there's no obvious special-casing
- I tried to make it as "open" as possible, to allow people lots of different ways to advance (hence the list 5 different memory models). it would've been better in the long run to decide on a …