>By which I mean Machine Learning, since the term "AI" smacks of not just determinism, but also a distinct lack of VC funding.
Is all we do, in a way or other:
https://en.wikipedia.org/wiki/Greenspun%27s_tenth_rule
> Any sufficiently complicated C or Fortran program contains an ad-hoc, informally-specified, bug-ridden, slow implementation of half of Common Lisp.
Thats why "Interpreter" is one of the common OO patterns on the book.
I done a lot of micro-languages (later I found are called "Domain Specific Languages" by hipsters). I have one that build the Database interface code automatically from the .sql script that build the postgresql database.
I code the database at hand, run "builddb.py" and it spit out all the mapping and part of the ORM layer. I doit with bad hacks, regex and stupid stuff (from a proper compiler construction perspective) but I only spend like half day doing it.
Later, I also build the inserts/updates and other extra stuff that sometimes is useful. Then I also spit out both F# and swift code, with the same terrible hacks...
----
Now, after do this thing like millon times before I start to think that I'm a idiot that could have donde this MUCH better. So, this time as a side project I'm doing the same thing, like a ORM-made-as-language.
And boy, this thing is fun. I not only can run "sql" agains my RDBMS but also with owm in-memory structures, similar to LINQ but, IMHO, better.
Now, the thing is... if this little project get to a point where is good, why not release it as a language?
If you think smaller though, there's a ton of utility to be had in taking the ideas and applying them to domain specific languages. I don't know much about Racket, but it feels like that's its reason to be.