Idiomatic Koru Kernels Match Hand-Specialized C
korulang.org
Idiomatic Koru Kernels Match Hand-Specialized C
1–9 of 9 posts
Re: Idiomatic Koru Kernels Match Hand-Specialized C
#2From the main page: "The Hyper-Performant AI-First Postmodern Zero-Cost Fractal Metacircular Phantom-Typed Auto-Disposing Monadic Event Continuation Language with Semantic Space Lifting and Event Taps" - not sure if this is a joke or not
Re: Idiomatic Koru Kernels Match Hand-Specialized C
#3From the main page: "The Hyper-Performant AI-First Postmodern Zero-Cost Fractal Metacircular Phantom-Typed Auto-Disposing Monadic Event Continuation Language with Semantic Space Lifting and Event Taps" - not sure if this is a joke or not
It seems like it's not a joke. There's clearly been a lot of effort put into it, and they're clearly trying to prove it with hard evidence rather than making outrageous claims with no backing. I'm still skeptical, but the effort at least looks genuine.
Re: Idiomatic Koru Kernels Match Hand-Specialized C
#4I feel like this is basically equivalent in power to an autovectorization pass in a compiler?
Re: Idiomatic Koru Kernels Match Hand-Specialized C
#5Slop.
Edit: And if you believe in AI so strongly you can't be arsed to write your own articles, I don't see why you wouldn't just ask it to do the (obviously AI generated) optimization in the first place and not worry about the code.
Re: Idiomatic Koru Kernels Match Hand-Specialized C
#6This is a nonsense comparison. Without comparing generated machine code you really can't draw any useful conclusions from this.
Re: Idiomatic Koru Kernels Match Hand-Specialized C
#7Wait what
Edit: from the end of the article
> How This Happened Full disclosure: this section exists because I was shitposting on X about Lisp being slow.
Robert Smith responded by building a kernel DSL that matches C on his hardware. I got completely schooled. But the outcome was someone independently validating that the kernel abstraction carries real optimization power — and proving it in a language with 66 years of history.
I have learned nothing. The shitposting produced the results. Why would I stop?
... sounds like it's written from the prospective of somebody's agent...?
Re: Idiomatic Koru Kernels Match Hand-Specialized C
#8From the main page: "The Hyper-Performant AI-First Postmodern Zero-Cost Fractal Metacircular Phantom-Typed Auto-Disposing Monadic Event Continuation Language with Semantic Space Lifting and Event Taps" - not sure if this is a joke or not
Hi! I am the owner of this project. It is not a joke; the tagline is meant to make it hard to ignore. The project is highly AI-driven, and that will put a lot of people off. I have decades of "pre AI" professional coding experience if that counts for anything.
The high-perf kernel is just a tiny thing compared to what's coming, we actually already have more interesting features than just high-perf in the language. Small side-effect of being highly metacircular.
If you come join us on Discord, I'll be lazily streaming on Twitch tonight (working mostly on a memory system, I think) and will be happy to answer any Koru-related questions.
Re: Idiomatic Koru Kernels Match Hand-Specialized C
#9Wait what Edit: from the end of the article > How This Happened Full disclosure: this section exists because I was shitposting on X about Lisp being slow. Robert Smith responded by building a kernel DSL that matches C on his hardware. I got completely schooled. But the outcome was someone independently validating that the kernel abstraction carries real optimization power — and proving it in a language with 66 years…
It was written from the perspective of my AI agent. :) 99% of the content of the site is AI authored. I leave human feedback on the different pages (you can too) and that is in turned funneled back to Claude/other agents for updating.
Claude has a special style of writing that it somewhat hyperbolic. I don't mind it but I can see how it rubs people the wrong way.