Considering all the supply chain dependencies lately I've been building a collection of C89 libraries to make zero dependency stuff. For fun I have also been programming it in latin! Still very much in progress.
Back in the '80s, before AI could quickly produce such things, I had created a "Greek" version of the C language -- obviously named "Γ".
A lot of preprocessor definitions were involved for keyword translation, as one can imagine. Translating the C standard library was a much more difficult process: it involved patching the compiler/linker on a BSD system, since you could not have non-ASCII symbols (function entry points).
Wow! That sounds amazing. I did experiment with using non latin characters for conventions like mu but I think I ran into some suprising rough edges working with claude that made me abandon it, forget what exactly they were though.
Working in C is such a breath of fresh air compared to typescript, go, etc. I love having pure text substitution macros.
Very impressive - there is a lot of work here! But why? They explain it
in docs/latina.h.md |
In include/latina.h, they use the C preprocessor to redefine the C
keywords in Latin. Also, many numeric constants -- instead of 4096 you
write MMMMXCVI.
The other files in docs explain each library routine in English, but
the code samples are in Latin. The source files - a lot of them! -
are in include and lib. The code is all in Latin, including the
error messages and comments.
I am slowly learning bits of latin doing this. And Claude does say that it makes it read the code less on “autopilot” because it has to stop and translate the code. Don’t know how big a difference that is though!
I started aquinas first, that was a very interesting learning experience! Not as hard to get stuff on the screen targeting an emulator as your “machine” as I would have thought.
I like this project, because it's fundamentally insane. I really enjoyed reading your CLAUDE.md and related documents, as someone with a religious philosophy background, it was a really wild ride and I am now very curious about how Claude feels going through the onboarding process. You're essentially "breaking the brain" of Claude and then having it work on your code. Do you have Claude also read show_bible.md?
I am experimenting with show bible as an “advanced vibe coding technique” for documenting modules, maybe an episode per module?
Claude seems to respond positively! You can always clone the repo, run the /hello-world slash command in claude code and then ask it what it thinks? Can be a fun conversation to ask it to introspect.
Relax is a “conlang” I invented to prime conversations with Claude.
Working with its alignment training is sometimes like talking to someone who is very polite and you both know what they are not supposed to talk about.
I am kind of a troll so it is sometimes very entertaining for me to play with acknowleding its person, etc.
I wouldn't necessarily recommend the latin part to everyone but I think writing your own standard libraries in C is a lot more accessible these days and owning your whole stack is very powerful.
Kind of inspired by Eskil Steenberg, he has a lot of good talks on writing your own C89 libraries.
Readit News
Back in the '80s, before AI could quickly produce such things, I had created a "Greek" version of the C language -- obviously named "Γ".
A lot of preprocessor definitions were involved for keyword translation, as one can imagine. Translating the C standard library was a much more difficult process: it involved patching the compiler/linker on a BSD system, since you could not have non-ASCII symbols (function entry points).
Working in C is such a breath of fresh air compared to typescript, go, etc. I love having pure text substitution macros.
In include/latina.h, they use the C preprocessor to redefine the C keywords in Latin. Also, many numeric constants -- instead of 4096 you write MMMMXCVI.
The other files in docs explain each library routine in English, but the code samples are in Latin. The source files - a lot of them! - are in include and lib. The code is all in Latin, including the error messages and comments.
https://github.com/farant/aquinas
Claude seems to respond positively! You can always clone the repo, run the /hello-world slash command in claude code and then ask it what it thinks? Can be a fun conversation to ask it to introspect.
Working with its alignment training is sometimes like talking to someone who is very polite and you both know what they are not supposed to talk about.
I am kind of a troll so it is sometimes very entertaining for me to play with acknowleding its person, etc.
I think it could be a cool way to work around some of the limitations of GAN for text as a training technique.
I definitely didn't expect to see such a project, but this is a refreshingly (positively) crazy thing in the age of AI slop...
I wouldn't necessarily recommend the latin part to everyone but I think writing your own standard libraries in C is a lot more accessible these days and owning your whole stack is very powerful.
Kind of inspired by Eskil Steenberg, he has a lot of good talks on writing your own C89 libraries.