New Shen Ports

[Reuben]

Hi all.

Just found out about this group when I made a PR to add a recent port to the github.io page. I recently discovered / got really interested in Shen. I've been exploring ways to leverage its unique capabilities to improve AI-driven coding systems. You can see some of my experimentation w/ it here. I wrote a blog post about this, had some good feedback on HN as well.

As part of my experimentation, I decided to work on several ports to expand the ecosystem: 

• Fork of existing `shen-go` port to include a bytecode VM: github.com/pyrex41/shen-go
• New `shen-rust` port: github.com/pyrex41/shen-rust. It's not quite as fast as shen-cl cold, but got pretty close.
• New `shen-lua` port using luajit: github.com/pyrex41/shen-lua. Slower than shen-rust and shen-cl on my machine, but faster than shen-go. This lets you potentially do really interesting things, like use Shen to author nginx runtime policies in OpenResty
  

All 3 ports use the newest 41.1 Shen kernel. All of this 100% AI driven, and documented clearly as such. The existing porting guide + the existing test suite provides excellent deterministic context to drive AI coding loops, and the small size of the KLambda kernel made this particularly tractable.

I used Simon Willison's `showboat` tool to create executable demos. I haven't seen a ton of people using this tool, but I've found it really useful as way for LLMs to demonstrate what is built in a way that's easy to digest for both humans and LLMs.

I'd appreciate any feedback, particularly on ways that I can extend tests in useful ways. 100% pass on the Shen test suite != guaranteed bug-free.

I'm really excited to dive deeper into the Shen rabbit-hole. I've long gravitated towards typesafe, functional languages (I learned Elm so i wouldn't have to learn JS to make websites, and I've gravitated a lot towards Elixir for other projects), and have only doubled down on this w/ LLMs. I also love lisp -- I taught myself to program reading Paul Graham' and Peter Norvig's books on Common Lisp. Shen combines all the things I liked about those, and so much more...

[Juan Cadena]

I guess we’re all coming to shen because we’ve realized that LLMs aren’t all that great on their own and we’re hoping to bridge the gap. 

I tried a small experiment 

https://github.com/juangrukat/logicbox

And it works about as well as it can. 

Since you’re making ports, try making one that has good ui support, and is within a lisp family. I was looking into racket. 

Why? Because when you’re translating from shen output to json or some other non-native structures you might have translation issues which will corrupt your analysis and you’ll try to figure out why. 

--
You received this message because you are subscribed to the Google Groups "Shen" group.
To unsubscribe from this group and stop receiving emails from it, send an email to qilang+un...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/qilang/216231a2-78aa-4a42-be21-77ffd9d8d64an%40googlegroups.com.

[Bruno Deferrari]

First of all, it is very cool that you did this and the way you approached it.

You probably found them already, but a few months ago I added some files for agents to the Shen sources repository on GitHub. If you are not using them already, give them a try because they add useful context.

For the shen-cl updates I saw you worked on, you may want to check the Shen/Scheme compiler code and use that as input too. The shen-cl port on GitHub is based on an older version of the Shen/Scheme port, so most updates since then probably still apply.

--

You received this message because you are subscribed to the Google Groups "Shen" group.
To unsubscribe from this group and stop receiving emails from it, send an email to qilang+un...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/qilang/216231a2-78aa-4a42-be21-77ffd9d8d64an%40googlegroups.com.

--

BD