▲ 213 points
back
88 comments
As a more-or-less neutral observation (I still kind-of like the project): What I find curious - perhaps a sign of these times: After just 2-3 months in development (judging from the repo commits, obviously I have no insight into what came before), the developer already communicates Jolt as a feature-complete, polished (perhaps even mature) Clojure dialect with multiple Jolt-adapted libraries from JVM-Clojure, FFI, Fibers, already with a professional looking and marketing-focussed website, detailled documentation, nREPL support, etc. Things I wouldn’t expect so soon from a language developed (mostly?) by a single person in such an early release stage.
I don’t know how to get a feeling about the actual state of projects like this anymore, so I always remain a bit suspicious. In Clojure, there is the (Rich Hickey inspired) design philosophy of “hammock driven development”, where we don’t rush things, let them simmer while lying in the hammock, sleeping on them, taking a walk, etc. Makes me wonder if projects like this are too rushed these days without thinking things through, discussing design decisions and getting ideas from other people, letting them mature over time. This isn’t specifically about Jolt (I don’t know much about it or its creator), it just inspired the thought, especially when comparing it to projects like Jank which has been years in development, with lots of feedback from the community and Clojure core maintainers as input.
The reason it's possible to move fast here is due to the fact that I'm porting an existing language to a new runtime. All the hard work and hammock time has already been spent by the core Clojure team and the community. I'm not aiming to reinvent the wheel here. Furthermore, having worked with Clojure for nearly two decades now, I have a very good idea of what the language does, and what I need to implement on the host platform.
Once I had the basics bootstrapped, it was really a matter of getting tests passing, and then getting more and more libraries to run on it which allowed me to exercise their own tests. And that's what really made the whole thing possible. It's sort of the ideal scenario where you have very well specified inputs and outputs, and a ton of existing tests to ensure you are implementing the behaviors correctly.
The only other aspect here is performance, and that's addressed by having an extensive benchmark harness which compares with JVM timings. This allows me to see where performance gaps are, and focus on addressing them.
I would be careful in extending my experience of being able to port Clojure to Chez Scheme runtime to working with LLMs in general because in most cases you don't have such well defined parameters when you start.
This is the germane point here. I'm not into LLM coding myself, but was talking about just this with my CS PhD supervisor yesterday, and porting language implementations is basically a perfect use case for them. He's been using them for porting examples from one language to another and was saying it was perfect - he already knows what he wants to see, understands both languages, and needs a syntax aware typist, essentially.
This. Good stable foundations allow for quick software development. Even if you use AI for coding, working with Clojure as an experienced developer produces extremely good results quickly because of the great foundations.
I love the expression "the hammock time has already been spent": it is actually very accurate and I'm very grateful to the Clojure team and community for spending that time!
Great work. It just simply worked for me.
> I don’t know much about it or its creator
The guy literally has published book(s) on Clojure - not some random, unknown, AI-wielding teenager from somewhere in Nebraska. I'm pretty sure they know what they're doing.
But of course, nothing is wrong with being suspicious about any new projects these days. Overall, I agree with your sentiment here.
Not slop, though. The main developer is Dmitri Sotnikov (yogthos), who's been around in the Clojure scene for a long time and knows what he's doing.
However, you're right to point out that the development is moving really fast.
I'm building a Clojure compiler for machine learning (https://github.com/sheaf-lang/sheaf), and after 9 months, it's still very far from being production-ready.
From what I’ve seen so far, I would also not dismiss it as slop, even though I’m generally against vibe-coding (but more for ethical and environmental reasons, not because the code is inherently bad/slop/whatever). Many people seem to have a hard time acknowledging that LLMs actually can produce good stuff, but it highly depends on the experience of the developer who guides them.
If a dev is competent and has enough experience actually writing code and making good design decisions, I see no reason to judge their work on the basis of what tools they used to get the job done. However, it may still put me off, since I find it hard to value LLM-written code the same way I value human-written code, even though it may be equally good or even better in quality.
The website is clearly slop. I KNOW it has been generated because it has the bog-standard unique claude LLM card-mistake that humans who hand-code just DONT make.
> Not slop, though.
What would you call 2k commits in 2 months by a single committer if not slop?
You're right to be suspicious: it's slop.
This is such a disingenuous, childish, immature and ignorant take - I can't even remain calm without getting angry.
Have you ever, kids, tried building anything useful? Anything practical, anything that actually works, anything beneficent? With or without AI?
Well, if you have in fact ever tried, perhaps you'd know that building reasonably complex software remains a difficult, challenging, unappreciated work. With or without the help of any tooling. AI-assistants included.
Maybe try building something good first. Earn your right to throw the first stone, come engaged, come prepared, ask questions.
What you're doing is a fucking slop. Human slop of ignorance. Worst kind.
Honestly LLM’s really foster hammock driven development.
It's really unfortunate that Hacker News has become an echo chamber of AI trash, but I suppose that's to be expected from people who have a vested interest in OpenAI and Anthropic, to the detriment of literally the entire world. I'm quite confident that YC is taking deliberate action to portray LLMs as a good thing much like Fox News.
This appears to be vibecoded but not disclosed as such. 2k commits from a single author starting from June 1st with really long commit messages and source code comments. The GitHub org has 19 additional projects that are all recently created.
edit: Yup, found a very long blog post that says it's made with LLMs near the end https://yogthos.net/posts/2026-07-02-jolt.html
Vibecoding means you don't pay much attention to the code, not the case here.
I think the author has been very open about the use of LLMs. And while I am very wary of LLM slop, this is one use case where it makes complete sense to utilize them to the fullest. A Clojure implementation already exists for LLMs to reference, and tons of test cases to verify behavior. It doesn't get better than that for LLM usage.
Besides this and jank, there is also Janet; playing in the league of Lisp to standalone binaries. It’s also cool!
I spent a couple of days digging in to Janet and what surprised me was how even though at a surface level it resembles Clojure, under the hood its actually quite radically different.
It has a lot of very interesting ideas going on, though ultimately i didnt end up using it (i need my persistent datastructures and seq col abstractions)
It feels very minimal and scheme-y and very different from anything ive used before. Definitely worth exploring
How much does this overlap / complements something like jank ? [1]
How much "production ready" / "battle tested" would this be ?
(Not to start a flame war, I'm genuinely curious about the differences.)
I am the creator of jank. I am also not interested in a flame war.
The key differentiator for jank is the seamless C++ interop, which is a problem not many people are willing to tackle, due to its scope and complexity. jank is a novel approach to this, providing incredible JIT and AOT support of arbitrary C++ libraries alongside your Clojure code, including a Cargo-inspired native build system for building your native deps along with your program (or finding them in the installed system).
For some people, there is another key differentiator, which is the amount of AI-generated code involved. My understanding is that Jolt and other newer dialects like Glojure have a lot of spunk and are exploring new ideas, given the velocity that AI-driven coding can provide. I don't see jank ever being categorized in that way.
Thanks, and thanks for the effort on jank !
Since you're here ;) : I read on the alpha doc that protocols and other part of the clojure object model are not done yet - is your goal to include them eventually, or is there a part of the lore of clojure that makes them not indispensable ?
I have a slightly different goal from Jank because my goal is to provide a runtime that supports existing Clojure libraries. I've spent the time to map out Java standard library APIs that popular libraries use and to create shims on top of Chez that allows them to run seamlessly. You can see a list of libraries that are fully tested and officially supported here. https://jolt-lang.github.io/docs/libraries.html
On top of that, I expose the API for creating shims in user space, so people can easily add their own for the libraries they want to use that might not be covered in the core. And I leverage this functionality myself to create libraries for JDBC layer or crypto that rely on doing FFI to shared system libraries that aren't part of the core runtime.
In terms of how production ready Jolt is, it's still fairly new obviously, so there will inevitably be bugs. However, it already passes full https://github.com/jank-lang/clojure-test-suite from Jank, and has its own conformance corpus https://github.com/jolt-lang/jolt/blob/main/test/chez/corpus... which is a superset of that. On top of that, I'm now able to run original test suites for the libraries I support and compare them with JVM outputs to ensure there aren't unintended divergences.
The other side of it is the benchmark harness which I use to ensure performance stays reasonably close to the JVM, in most cases it's within 1.x, and the worst case is around 6x right now. https://github.com/jolt-lang/jolt/tree/main/bench
So, Jolt basically aims to be a drop in JVM replacement for running existing Clojure code.
But isn't it kind of a one way street? We can extract clojure code from the JVM but then we cannot use its ecosystem anymore, which kinda feels like a los.
Since we can use (n)REPL in all ecosystems (jvm, browser, node etc...), could we use it or a variation to have interop?
Jank is Clojure hosted on C++/LLVM, and Jolt is Clojure hosted on Chez Scheme.
Jank compiles to C++, C/C++ interrop is direct, you just import and use any C/C++ library directly as if you were coding in C/C++. There's no FFI, it feels the same as how you use Java from ClojureJVM.
Jolt compiles to Scheme, but doesn't yet have Scheme interrop, though it is planned. Scheme doesn't have a large ecosystem of libraries. So Jolt is almost more of a "pure Clojure" runtime. It mostly relies on using other Clojure libraries, and provides a subset of Java for common I/O and threading ops that you'd use interrop for in ClojureJVM to get most commonly used Clojure libs working (ring, reitit, integrant, malli, hiccup, etc.)
Jolt has a C FFI, but it's not very convenient, you have to declare each C function yourself, manually manage type conversions, memory, etc.
They can both take your Clojure code and make a compiled native single binary out of it. Jank also plans to allow Jank programs to be compiled into static or shared libraries such as .so files, so they can be embedded in or depended on from C++ and other native applications.
Jolt actually does have both Scheme interop and FFI
https://jolt-lang.net/docs/host-interop.html
and it is possible to make .so libraries with Jolt as well right now
What a wonderful excuse to dive back into Clojure! This looks fantastic. I wonder what neat stuff might be made in Clojure thanks to a C FFI that wasn’t doable on top of the JVM.
Note that nowadays with Panama, there are more FFI opportunities in JVM implementations as well.
Although it is rather verbose, and it is easier to have a C header and run jextract on it, instead of the manual boilerplate.
No idea how well AI tooling would manage.
Coffi is a great library that makes Panama FFI/FFM painless without interop. https://github.com/IGJoshua/coffi
For example look at this arc from sqlite4clj https://github.com/andersmurphy/sqlite4clj/blob/master/src/s... it's very elegant.
(Also can plug my own libvips wrapper using coffi https://github.com/outskirtslabs/vips)
Using FFI/FFM "vanilla" with java interop is also viable, and in my experience the SOTA models do just fine with it (with or without jextract).
I've been diving back into Clojure just recently and this will be something I'll be trying to use immediatley.
Author here, glad to answer any questions about Jolt.
Awesome work, have you looked into using Racket's fork of Chez? It supports some additional extensions and architectures from base Chez that would seem to beneficial as a compiler target, at least that was the case a couple years ago.
Congrats, this is awesome! I have followed your writing and have your book, so was excited to see this was you.
I have one question, can one drop into Scheme the way one can call Java in clojure, or is one totally firewalled from the underlying Scheme implementation?
Great to see Gambit getting love too. Gambit and Chez are amazing achievements.
It's neat!
How does this fit into the ecosystem compared to something like Babashka?
I tried running a cross-platform (JVM, CLR, JS) Clojure project I'm working on, but it failed trying to load the JVM's `System/in`. Does Jolt handle reader conditionals? Is the intention for it to always follow the `:clj` branch or are you planning a `:cljolt` or something?
Chez Scheme is highly underrated! Jolt is amazing work.
I also have been using it for Jerboa, a Gerbil Scheme compatible language.
https://jerboa.sh
8,820,174 lines of 100% vibe coded code.
I just saw this - works great to build small self contained apps. (I just tested on macOS, also supports Linux).
Clojure hosted on Chez Scheme (same as Racket). So cool. I tried building a standalone executable for the Chess game in my Clojure book using "jolt build -m chess-game.cli -o chess_jolt" and it simply worked great.
I tried again with a datomic demo and it needed the source code to the open source version of datomic, so I need to do some work on that.
it's good to have even more (and more slim!) ways to run Clojure
(not just JVM and JavaScript runtimes)
Nice to see further adoption from Chez Scheme.
Have to give it a try.
I understand some don’t like having a JVM, but for running in the browser, what does Jolt on Gambit offer over plain ClojureScript?
Oh I would definitely recommend ClojureScript or Squint for running in the browser. Using Gambit to compile to Js was really just an illustration of how Jolt has a portable layer that can be compiled to a completely different Scheme backend. I don't really intend to compete with the existing frontend solutions because you really want to keep your dialect as light as possible to minimize the size and friction with the existing Js ecosystem.
Does it have tail-call optimization? I scanned the docs but don't see it mentioned.
Presumably basing Jolt on Chez removes the limitation of the JVM? It could be an interesting selling point.
Yup, Jolt has TCO inherited from Scheme
How many years have it been there? My memory reminds me of Jolt but I very hardly remember what was it, all I remember is that it was a LISP.
You're probably thinking of Ian Piumarta's COLA project (Combined Object Lambda Architecture), which had parts named after cola soft-drinks (Coke was a lisp-like language, Pepsi was a Smalltalk-like language, and Jolt was an early implementation of Coke):
See also lg which is a clojure implemented in go:
https://github.com/nooga/let-go
Fun fact. It also runs on plan9:
Jolt Physics: https://github.com/jrouwe/JoltPhysics
What does this have to do with Jolt, the Clojure compiler other than in name? Is there a port to Clojure or other connection?