I did the same thing, shouting.h, defined the uppercase version of many types and keywords.
Had a good uncontrolled laugh during a team presentation with a colleague. It was a bit disrespectful for the poor presenter who had nothing to do with this…
Reminds me of my "BASIC" #define statements I made in college to make C look slightly more BASIC-like, I think everyone does this the moment they realize they can do this in C.
Stephen Bourne was a fan of Algol 68, and the original Bourne shell was written in "Bournegol", which is C with enough #define malarkey to make it more Algol-like:
This is a good example of where it’s important to be more up front about the role of AI in the making of a thing.
Making a language that compiles through LLVM is no small task and takes a lot of expertise. Most of the time people do it because they have a point of view and are highly technical.
Making a joke language via AI is an entirely different exercise. Not without value but not the same, especially when evaluating what it means about the author.
I made a “fully” functioning programming language that is kind of like Rebol on luajit using Claude code - and I haven’t really done much serious programming since college 20 years ago. It’s fun though.
> This is a good example of where it’s important to be more up front about the role of AI in the making of a thing.
Maybe, but we don’t always discuss our tool stack before showing our work because a lot of times it’s assumed or not interesting. AI is a tool, it’s a lever that you push on to multiply your effort. The product of your effort speaks for itself, as it always has.
Nicely done Geoffrey, I think we are coming full circle, instead of porting projects to different languages (for whatever) like going from zig -> rust -> zig, now we can also add additional hops like zig -> rust -> curse -> ts -> ocaml -> english -> zig.
I'm writing the C backend by hand and using AI for the rest, so how did this author manage to finish an entire language in just 34 hours? I've been steadily catching and fixing what the AI writes, so it's amazing to me that they ended up with a complete language. It makes me wonder if the way I'm building a compiler is just wrong.
If you tell it to write a spec -> then write the tests -> then implement, the LLM should be able to pretty much one-shot a compiler frontend. LLMs really benefit from the kind of task that has a built-in validation loop.
I'm working on something similar, but unlike the author, my progress has been pretty slow. It's tough. I do write about a fifth of the code myself, but I keep getting stuck on the rest.
GPT/Fable/what have you is not building this language, as its laid out, and "working" in a true sense that fast.
Either OP is completely full of it, the language didn't actually work (and likely still doesn't), or the language is far less sophisticated than it seems from the examples - it's and the examples are minimal, so it's kind of hard to tell what it actually does...
Excellent. At last, I can I confess a far worse crime.
Late 2020, pre-AI, which I'm not sure if that makes it better or worse...
-- Obviously this one also runs DOOM ;)I did this in 1989, as if C had been written by the folks who brought you COBOL:
Needs more DIVISIONs. Though I can imagine this being what would happen if the designers of PowerShell were responsible for C.
How right you are! I use PowerShell when I have to.
Seems like PS was designed under the motto, "The choice: why make it simple when we can make it complicated?"
I did the same thing, shouting.h, defined the uppercase version of many types and keywords.
Had a good uncontrolled laugh during a team presentation with a colleague. It was a bit disrespectful for the poor presenter who had nothing to do with this…
Can't find shouting.h anymore unfortunately.
Reminds me of my "BASIC" #define statements I made in college to make C look slightly more BASIC-like, I think everyone does this the moment they realize they can do this in C.
Stephen Bourne was a fan of Algol 68, and the original Bourne shell was written in "Bournegol", which is C with enough #define malarkey to make it more Algol-like:
https://www.tuhs.org/cgi-bin/utree.pl?file=V7/usr/src/cmd/sh...
this gave me such a good chuckle :p
uwu, this is amazing ;3
Thanks! I hate it!
In awe at whatever inspired this though
This is a good example of where it’s important to be more up front about the role of AI in the making of a thing.
Making a language that compiles through LLVM is no small task and takes a lot of expertise. Most of the time people do it because they have a point of view and are highly technical.
Making a joke language via AI is an entirely different exercise. Not without value but not the same, especially when evaluating what it means about the author.
I made a “fully” functioning programming language that is kind of like Rebol on luajit using Claude code - and I haven’t really done much serious programming since college 20 years ago. It’s fun though.
> This is a good example of where it’s important to be more up front about the role of AI in the making of a thing.
Maybe, but we don’t always discuss our tool stack before showing our work because a lot of times it’s assumed or not interesting. AI is a tool, it’s a lever that you push on to multiply your effort. The product of your effort speaks for itself, as it always has.
It's cool that AI lets you cheaply experiment with language design, but I wish people would stop using it for the writeups, too.
Buried near the end is a mention of per-frame arena allocation, which is an interesting idea for a game engine (although not a novel one).
Nicely done Geoffrey, I think we are coming full circle, instead of porting projects to different languages (for whatever) like going from zig -> rust -> zig, now we can also add additional hops like zig -> rust -> curse -> ts -> ocaml -> english -> zig.
Thank you for keeping the token furnace burning!
What if I don't `evict`? How different is it from forgetting to `free`?
I'm writing the C backend by hand and using AI for the rest, so how did this author manage to finish an entire language in just 34 hours? I've been steadily catching and fixing what the AI writes, so it's amazing to me that they ended up with a complete language. It makes me wonder if the way I'm building a compiler is just wrong.
If you tell it to write a spec -> then write the tests -> then implement, the LLM should be able to pretty much one-shot a compiler frontend. LLMs really benefit from the kind of task that has a built-in validation loop.
I'm working on something similar, but unlike the author, my progress has been pretty slow. It's tough. I do write about a fifth of the code myself, but I keep getting stuck on the rest.
Well... The fact that it took 10x longer for Doom to build than the language tells a lot.
Doom is not harder to re-implement than a language.
The language does not "work" by any sense of the word - which is why it took Doom so long to implement.
You could get GPT to "self host" a "language" in 5 hrs. That's not impressive.
The language actually working and being non-trivial would be in 20x the time.
Sure, GPT can build yet-another-Lisp in 2 minutes. You could copy Lispy yourself in that time: https://www.norvig.com/lispy.html
GPT/Fable/what have you is not building this language, as its laid out, and "working" in a true sense that fast.
Either OP is completely full of it, the language didn't actually work (and likely still doesn't), or the language is far less sophisticated than it seems from the examples - it's and the examples are minimal, so it's kind of hard to tell what it actually does...