A really great place to reach for this is Elixir/Phoenix now that the set theoretic type system is in place with LiveView. I have been doing some pretty large/complex codebases gradually working towards fully integrating working this way for the past 5 years and it has been great.
The general challenge becomes integrating the fractional front-end code written in your backend-language that compiles to JS with the rest of the JS ecosystem.
JaneStreet have a love of writing their own stuff from scratch so it doesn't apply to them but it might to you.
Hence most people end up with frontend-as-backend rather than backend-as-frontend.
> Is there some limitation that WASM has that JS doesn't?
You need a JS trampoline to call your WASM and make browser primitives available to it, and IIRC calls into browser code incur some extra overhead, but those are pretty manageable.
Additionally: for high level languages, source code is _much_ smaller than compiled binaries. If your initial needs are simple, your users are likely downloading more than 10x as much code.
WASM does not (yet) have access to the DOM or web APIs, meaning every direct interaction with the browser must first go through JS interop anyways, with a resulting performance penalty. There are proposals, and maybe we'll see it happen by 2035, but as of now WASM is best suited for heavier workloads where the edge in application performance outweighs the cost of JS interop with the browser rather than being a universal solution.
FWIW if you listen to Signals and Threads (the Jane Street podcast) they talk about this.
They have dedicated UX designers who do relatively little code but focus entirely on making the UX as optimal for the traders as possible and one of them was on an episode.
They talked about how most of the normal UX rules for SW go out the window because quants and traders have such different needs than "normal software".
Is there any professional software that wouldn't benefit from high information density? Most software design is optimized for the non-users who have to approve it.
"non-users" are also big targets for software. If you run a software business, chances are you want people who've never used your software to start using your software. In that case, discoverability trumps high information density.
this - most of the internal software at my job that's actually important is very dense, with keyboard navigation and shortcuts. The ones that somebody makes to sell themselves to upper management while angling for a promotion look very pretty, and never get used.
And yet, we make such beautiful and vacant software ourselves for our clients, because again, its b2b, so you aren't selling to the people who actually have to use the software. Im glad at least we arent forced that often to use specific internal tools, the better ones just get adopted organically
This matters. Beyond performance (a universal good), the definition of "good" UX is strictly dependent on the user. Fintech platforms for traders who open a multi-window thick client application(s) and spend hours in it every day have little in common with consumer mobile apps designed to hold your attention for maybe a minute. It's mind-bogglingly lazy to choose something like MUI (Material UI) as if it were some kind of reasonable foundation irrespective of context.
Curious how this compares to Melange which is used by Ocaml shops as well to double up on Ocaml for both front and backend (ahrefs being the major user and sponsor). Does this mean giving up a lot of the JS ecosystem (React, graphql, etc)?
They're a trading firm that hires really clever people to use math to trade. Their strategies are, presumably, very complicated and ever changing, and secrecy about them is the point - you get bigger margins for longer if people don't learn about your 'one weird trick' (that and also that there is presumably another side to your trades losing out on upside, and they might not be best pleased to learn about it.
This doesn't seem quite as clear-cut as you're making out.
The Securities and Exchange Board of India (SEBI) have accused Jane Street (JS) of market manipulation, JS claims it was legitimate arbitrage, and the case is still ongoing [1].
As I understand it, the ban only applied to Indian securities (not Asian markets as a whole), and was lifted back in July 2025 after JS paid ~$560 million (the alleged ill-gotten gains) into an escrow fund. They don't appear to have resumed trading, though [2].
The above might be too far-fetched as it only applied to India? They did pay a fine to the regulator and resumed trading (might not be the same strategy though)
To OP’s original question, they made $20B last year and have a niche in several financial products
Can someone who understands web UI programming tell me if this would be good for my local agent to use to produce HTML based reports and outputs for me? Or for TUI outputs?
The main thing that Bonsai gets you is a nice programming model for making your UIs interactive. If you're mostly producing static reports (with maybe some basic interactive visualization widgets or something), it will not get you much over generating some variant on HTML + JS.
It makes more sense for applications where you have complex interactive logic in the UI, with a lot of different components that need to control and display some shared state.
There was this submission two months ago: https://news.ycombinator.com/item?id=48365904 . Basically they have a TUI frontend for the same thing. And the blog post says: "Bonsai_term would feel familiar to anyone who’d ever done web development here, and it had the huge advantage of being especially amenable to AI assistance. It was actually somewhat of a mystery to us how good the models were at writing Bonsai_term code, given how relatively obscure it is [...]"
There's a full terminal implementation of Bonsai as well. I actually use it in my personal workflow these days to manage my contacts database! https://anil.recoil.org/notes/aoah-2025-9
Custom infrastructure; I use my site as a playground for various things over the years. It ran as the first MirageOS unikernel for many years, and recently it's been running on a zero-allocation OxCaml webserver https://anil.recoil.org/notes/oxcaml-httpz to help me get to grips with the new extensions.
Jsoo supports self-recursive and mutually recursive tail calls, but for general tail calls it would need to trampoline every single call, since the only major browser with proper tail calls is Safari.
It looks like a nice little library; but oh boy must this be so limiting for the product teams that are forced to use. Everything looks like it's straight out of the 1990s.
This is great. It focuses on utility and information density over design. It looks like someone took a terminal UI and transplanted it to the web, Bloomberg terminal style.
I'm pretty sure you can build tools with this that are fast and pleasant to use.
Janestreet was banned from asian markets for market manipulation. They took in billions and fisted the average-joe investor. As a high profile ocaml user that casts a big shadow on ocaml too, as many involved in janestreet are also part of ocaml core team in some way.
I wont buy ignorance, as devs clearly know whats going on, so staying silent is pretty much being just as guilty.
If you read up on it that was basically India protecting its corrupt cronies. Janestreet showed severe problems with some Indian companies and instead of fixing that and creating a healthier and more transparent market they decided to go after the messenger.
At Jane Street we're super excited by Functional programming and by CAML in particular, so when we need low latency software, we use OCAML, when we need hardware, we write out own langauge - HardCAML, and when we need a Web UI, we build a Web UI framework in CAML. Because we fucking love CAML.
But AI do benefit a lot from a strict compiler, and having a simple language improve a lot on type-safety, so I don't think it is to throw, even today.
JaneStreet has enough free cash to not worry about those issues. If one of their key persons dies in service, they can go out and contract the worlds leading expert in that domain, and their annualised rate isn’t even a rounding error on their bottom line.
It might be a very performant UI framework in OCaml, but all the UI elements look extremely unpolished to me, like what a really smart high-school kid would build the GUI for his side project on Windows in 90s.
Thanks, I can use JS as a functional programming language.
> And because Bonsai is written in OCaml, it becomes possible to use the same language and types on both the backend and frontend.
Finally! I was waiting for this to become possible!
There's also Fable for F# but I believe this commenter is being sarcastic. Javascript is a common backend language.
A really great place to reach for this is Elixir/Phoenix now that the set theoretic type system is in place with LiveView. I have been doing some pretty large/complex codebases gradually working towards fully integrating working this way for the past 5 years and it has been great.
Similar attempts include Scalajs.
The general challenge becomes integrating the fractional front-end code written in your backend-language that compiles to JS with the rest of the JS ecosystem.
JaneStreet have a love of writing their own stuff from scratch so it doesn't apply to them but it might to you.
Hence most people end up with frontend-as-backend rather than backend-as-frontend.
Such attempts are quite common. Some I remember:
https://ocsigen.org/ in Ocaml too
https://websharper.com/ for fsharp and csharp. Really good when I used in in fsharp
Maybe https://melange.re/v7.0.1/ too? (Not sure)
I took this as a joke in reference to the popularity of JavaScript on the backend for the last two decades.
that's because most people's intro to CS was how to build a webpage.
Couldn't WASM solve that problem once and for all? Is there some limitation that WASM has that JS doesn't?
Disclaimer: I am very inexperienced at front-end development.
> Is there some limitation that WASM has that JS doesn't?
You need a JS trampoline to call your WASM and make browser primitives available to it, and IIRC calls into browser code incur some extra overhead, but those are pretty manageable.
Additionally: for high level languages, source code is _much_ smaller than compiled binaries. If your initial needs are simple, your users are likely downloading more than 10x as much code.
WASM does not (yet) have access to the DOM or web APIs, meaning every direct interaction with the browser must first go through JS interop anyways, with a resulting performance penalty. There are proposals, and maybe we'll see it happen by 2035, but as of now WASM is best suited for heavier workloads where the edge in application performance outweighs the cost of JS interop with the browser rather than being a universal solution.
Yup, and Clojure/ClojureScript!
> Finally! I was waiting for this to become possible!
Sarcasm?
Do you want to have a talk as to why JaneStreet wasn't built on JavaScript?
I am sure it's very performant, but to me it's extremely ugly; Surely someone can fix margins and still have it be performant.
I’m no UI expert - what’s wrong with the margins?
inconsistent and maybe too tight, but I would think those are issues with the sample UIs and not this library
FWIW if you listen to Signals and Threads (the Jane Street podcast) they talk about this.
They have dedicated UX designers who do relatively little code but focus entirely on making the UX as optimal for the traders as possible and one of them was on an episode.
They talked about how most of the normal UX rules for SW go out the window because quants and traders have such different needs than "normal software".
Is there any professional software that wouldn't benefit from high information density? Most software design is optimized for the non-users who have to approve it.
"non-users" are also big targets for software. If you run a software business, chances are you want people who've never used your software to start using your software. In that case, discoverability trumps high information density.
this - most of the internal software at my job that's actually important is very dense, with keyboard navigation and shortcuts. The ones that somebody makes to sell themselves to upper management while angling for a promotion look very pretty, and never get used.
And yet, we make such beautiful and vacant software ourselves for our clients, because again, its b2b, so you aren't selling to the people who actually have to use the software. Im glad at least we arent forced that often to use specific internal tools, the better ones just get adopted organically
This matters. Beyond performance (a universal good), the definition of "good" UX is strictly dependent on the user. Fintech platforms for traders who open a multi-window thick client application(s) and spend hours in it every day have little in common with consumer mobile apps designed to hold your attention for maybe a minute. It's mind-bogglingly lazy to choose something like MUI (Material UI) as if it were some kind of reasonable foundation irrespective of context.
From experience I can say that traders prefer extraordinarily little whitespace in their UIs.
Probably because they skew towards higher information density
I'd kill for tight UIs now. Everything insists on having 15000-mile wide margins, and at most two items visible on an 8K screen.
We are at a point when TUIs from 1990s running in 12" screens with at best a 480x240 resolution display more info than almost anything we have now.
Curious how this compares to Melange which is used by Ocaml shops as well to double up on Ocaml for both front and backend (ahrefs being the major user and sponsor). Does this mean giving up a lot of the JS ecosystem (React, graphql, etc)?
Interesting! What are Bonsai's dependencies? I use OCaml but - so far - none of Janestreets libraries etc.
Basically most of Jane Street's published code[0] via its dependency on Core.
[0]: https://github.com/janestreet/bonsai/blob/f31661450eb133fe89...
What does JaneStreet do besides sponsoring nerdy Youtube channels and writing UI libraries?
They're a trading firm that hires really clever people to use math to trade. Their strategies are, presumably, very complicated and ever changing, and secrecy about them is the point - you get bigger margins for longer if people don't learn about your 'one weird trick' (that and also that there is presumably another side to your trades losing out on upside, and they might not be best pleased to learn about it.
> really clever people
Sam Bankman-Fried and his League of Legend rankings would like a word
Print money?
Governments print money, high frequency traders extract money. They would say that they provide liquidity to the market as a service.
Market manipulation. They got kicked out and banned from the asian markets for fraud.
EDIT.
Janestreet PR team is downvoting. Like always, radio silence and just downvotes.
I had expected more from janestreet.
This doesn't seem quite as clear-cut as you're making out.
The Securities and Exchange Board of India (SEBI) have accused Jane Street (JS) of market manipulation, JS claims it was legitimate arbitrage, and the case is still ongoing [1].
As I understand it, the ban only applied to Indian securities (not Asian markets as a whole), and was lifted back in July 2025 after JS paid ~$560 million (the alleged ill-gotten gains) into an escrow fund. They don't appear to have resumed trading, though [2].
[1]: https://www.reuters.com/legal/government/us-trading-firm-jan...
[2]: https://www.reuters.com/sustainability/boards-policy-regulat...
Trading the underlying to offset your derivative position isn't fraud.
The above might be too far-fetched as it only applied to India? They did pay a fine to the regulator and resumed trading (might not be the same strategy though)
To OP’s original question, they made $20B last year and have a niche in several financial products
Yes, the fraud was done in india. Why does that matter?
Can someone who understands web UI programming tell me if this would be good for my local agent to use to produce HTML based reports and outputs for me? Or for TUI outputs?
The main thing that Bonsai gets you is a nice programming model for making your UIs interactive. If you're mostly producing static reports (with maybe some basic interactive visualization widgets or something), it will not get you much over generating some variant on HTML + JS.
It makes more sense for applications where you have complex interactive logic in the UI, with a lot of different components that need to control and display some shared state.
There was this submission two months ago: https://news.ycombinator.com/item?id=48365904 . Basically they have a TUI frontend for the same thing. And the blog post says: "Bonsai_term would feel familiar to anyone who’d ever done web development here, and it had the huge advantage of being especially amenable to AI assistance. It was actually somewhat of a mystery to us how good the models were at writing Bonsai_term code, given how relatively obscure it is [...]"
Probably not, because your agent won’t have much info about this tool in its training set.
I really want to like OCaml. One day I'll get around to learning it.
Not sure why, but reminds me of Fog Creek's Wasabi.
Looks like it's Web-only, no mention of the native UI support (terminal UI excluded).
There's a full terminal implementation of Bonsai as well. I actually use it in my personal workflow these days to manage my contacts database! https://anil.recoil.org/notes/aoah-2025-9
Seconding the sibling comment - what are you using to run your personal site? It is very nice.
Unrelated but what do you use to manage your personal website? I love how everything is interconnected.
Custom infrastructure; I use my site as a playground for various things over the years. It ran as the first MirageOS unikernel for many years, and recently it's been running on a zero-allocation OxCaml webserver https://anil.recoil.org/notes/oxcaml-httpz to help me get to grips with the new extensions.
The “The thinking in bonsai” link 404s.
This says it is based on Elm. So it has the same clean immutable state structure?
Oh it needs a userland trampoline!
> JSOO does not have tail call optimization
Jsoo supports self-recursive and mutually recursive tail calls, but for general tail calls it would need to trampoline every single call, since the only major browser with proper tail calls is Safari.
Why does this library get posted here, what feels like every month? I remember it seeing at least twice before.
Been here 5 years, never saw it
HN is obsessed with anything done in niche programming languages, is my guess.
Also, Jane Street.
It looks like a nice little library; but oh boy must this be so limiting for the product teams that are forced to use. Everything looks like it's straight out of the 1990s.
I wish there were more OCaml shops out there
This would of been cool back in 2014.
Ocsigen and Eliom were started 2005 and are pretty similar, although I can't find any well-known projects using Eliom :/.
Presumably this has learned some lessons from it, so it's probably better.
This is great. It focuses on utility and information density over design. It looks like someone took a terminal UI and transplanted it to the web, Bloomberg terminal style.
I'm pretty sure you can build tools with this that are fast and pleasant to use.
Related:
Janestreet was banned from asian markets for market manipulation. They took in billions and fisted the average-joe investor. As a high profile ocaml user that casts a big shadow on ocaml too, as many involved in janestreet are also part of ocaml core team in some way.
I wont buy ignorance, as devs clearly know whats going on, so staying silent is pretty much being just as guilty.
If you read up on it that was basically India protecting its corrupt cronies. Janestreet showed severe problems with some Indian companies and instead of fixing that and creating a healthier and more transparent market they decided to go after the messenger.
The "Why Bonsai?" I found really funny.
Let me re-write that section for you:
Why Bonsai?
At Jane Street we're super excited by Functional programming and by CAML in particular, so when we need low latency software, we use OCAML, when we need hardware, we write out own langauge - HardCAML, and when we need a Web UI, we build a Web UI framework in CAML. Because we fucking love CAML.
And there's nothing wrong with that. Many languages attempt to do the same thing. Why bother with 15 different languages if you can leverage one?
See also LiveView and Hologram for Elixir.
Agreed, looks like it's reinventing the wheels. Frameworks does not really matter today, AI agents write the code anyway.
But AI do benefit a lot from a strict compiler, and having a simple language improve a lot on type-safety, so I don't think it is to throw, even today.
But what about compilation speed? Hot reload? Maintainability? What is the bus factor for these kind of technologies?
Yea, its cool to have strict type safety, but what's the point of it doesn't have all the benefits from other Frameworks that took years to polish
JaneStreet has enough free cash to not worry about those issues. If one of their key persons dies in service, they can go out and contract the worlds leading expert in that domain, and their annualised rate isn’t even a rounding error on their bottom line.
AI certainly writes Internet comments so that the flock does not have a single minute to get independent thoughts.
Better frameworks will still help AIs avoid silly mistakes. A smarter framework means you can be productive with a dumber/cheaper AI.
All your money and retirement funds are safe; I’m too dumb to work at JaneStreet; every time I see OCAML I feel less than and confused.
It might be a very performant UI framework in OCaml, but all the UI elements look extremely unpolished to me, like what a really smart high-school kid would build the GUI for his side project on Windows in 90s.
Thanks, I can use JS as a functional programming language.