sroerick

It may do the opposite! My sense of LLMs is that simplicity beats training.

If you have 1/5th the context to consume, that's going to generate better code than the comparable typescript, even if the language is an esoteric one.

I mean - how specifically are you using knowledge graphs?

> Describing it as interpretation is often wrong, and leads people to the extremely common (even here on HN) mistaken belief that Lisps are strictly interpreted and never compiled.

Well, I literally said in my post that basically all Lisp was compiled. And I was replying to a guy saying that I thought all lisp is compiled. So hopefully I'm not contributing to that misconception.

Hey, I'm probably punching out of my weight class here - I'm basically a Lisp beginner. But I will say that I don't think the manual for Lisp 1.5 gets as much airtime as the original John McCarthy eval statement, which does describe an interpreter. I do think there's a lot of conceptual overlap between an interpreter and a REPL. And while I've not read SICP my understanding is that they teach a Lisp interpreter. And also I think the author of this article is describing an interpreted language.

So I don't think I'm over my skis here. But I appreciate your clarifying point! I've not read the Lisp 1.5 manual but I probably should!

I asked a similar question above, but can you talk more about the ergonomics of keeping the source and the image in sync? Is this generally a manual process in CL?

Awesome. Got any more info? Were any of the Lisp Machines image based?

Can you explain how you sync back to code, in practice?

In other words - if you modify the run time, are you then able to mirror those changes in your codebase and then reboot?

I am not a Lisp developer by trade, but I accidentally made an interpreted Lisp as a habitat for LLMs. My research has mirrored exactly what you are saying.

I think conceptually, Lisp was an interpreted runtime. However, the combination of chasing efficiency with compilation and perhaps "everything is a file" consuming mindshare, interpreted Lisp seems a largely academic pursuit.

It's really great for running agents, though.

I've been playing around with a fully interpreted Lisp as a habitat for LLMs to code in, and honestly it's really great

Can you elaborate on this?

Hi, thank you both for this. I built my own, worse, harness in my own, worse, lisp and I will also be joining the chat

Cool! I used Larbs.xyz for a while. I love a good dotfiles repo.

I guess I think there's a lot more in Omarchy than a dotfiles repo, I was very impressed with the demo video. Even just creating the menu and updating system is non-trivial. Mirroring the entire distro through Cloudflare seems significant also. I've not seen those in another dotfiles project.

The demo video is comprehensive and impressive

My understanding is that it only goes to Omarchy, and only with explicit permission from the user, so maybe "automatically" is a misnomer - but the diagnosis process feels very snappy

I think it's implemented and demoed in the Quattro release

Omarchy Quattro is pretty dang good. It's a lot more than any shell script I've written, certainly. If you've written a more complex one I'd love to see it.

Having agents in the default environment, having everything configurable through text files, and having default agents automatically file bug reports on crashes - these are tangible quality of life improvements that you can't get on other systems.

Maybe Windows and MacOS should write some shell scripts.

I have a pretty similar approach to this. I built a sexp DSL for spec and I have a "compiler" which does static analysis checks and linting.

I built a Lisp interpreter and hosted it on an OCaml http core, and it does exactly this for me.

I've been using it for the last three months for basically all my computing. I always wanted a Smalltalk type environment, and this finally scratches this itch.

Creating an endpoint is just defining a lisp function.

There's still a billion things to clean up but it works really well.

My LLMs say that if you followed this same pattern in pure clojure or racket it would be comfortably <10k lines of code.

That said, I do enjoy implementing my own lisp and I think the OCaml core gives it some stability and security.

I'd love to know more about your experience on this, generally. What have you been doing in Lean? How have you approached this?

I hadn't heard of BDD before, but the idea of "executable spec" is certainly very interesting

That's good to know. The open weight models are still pretty awful getting stuck in this loop

Thanks, appreciate this also, vibe coding in my own lisp and this has been frustrating

I'm interested in this but why those four separate languages

Edit: okay I read the code, it's actually four separate implementations

How do you teach LLMs to close parens properly??

Yeah, I felt this way too, and it's probably beyond the capacity of the federal government to organize in this way - but it is theoretically possible that the DOE could build a copyright respecting model (which it seems like they are doing here) and then a like minded DoJ could come a calling for the Labs. It may never happen, but something to ponder.

It would be extremely interesting to me if the usgov produces a model which honors copyright and is also useful. This would give them extreme leverage over the labs, who may be violating copyright in significant and obvious ways.

They are both the differentiator.

AI previously provided speed but not quality. As soon as quality reached an acceptable threshold, the speed became the reigning factor.

In my opinion the quality is still much lower, but speed means the cost is significantly lower also.

Solution is simple, run slower!

Even if you're sour on American exceptionalism, this seems like a valid and good product to me, both from a privacy standpoint and from an ethical labor standpoint?

Am I the only one who remembers the suicide nets at Foxconn?

My approach has been an interpreted Lisp running on a web server backend and it basically does everything described here