Claude Code uses Bun written in Rust now
simonwillison.net
simonwillison.net
Humans and agents share one thing: they are both non-deterministic. He talks about the issue of tracking memory lifecycles manually in zig so it can be explicitly freed. As expected, this leads to a long list of bugs where people missed things.
Rust does this automatically. It removes an entire class of errors from his backlog. From an engineering management perspective, this looks like a pretty good trade.
The bonus here is that compiler errors are exactly the kind of deterministic guardrail you need to put around coding agents. Claude works really well if you give it a way to test for correctness and "make it compile" is a pretty good target.
There's a general version of this: the artifact you expose plus the test you run on it. Deterministic tests turn stochastic output into a hard guarantee. Wrote it up here if useful: https://michael.roth.rocks/blog/verification-surface/
You can just do that, and then Zig is really no less robust than Rust. But if you want to do "managed language" style allocation patterns (like what llms generally prefer), it doesn't make sense to use it.
I cannot take this seriously as tutorials on robust Zig Allocation Pools will store a deinit method for each item within the pool, so when the pool deinits, all internal objects can be deinit'd.
That is just RAII & dtors from first principles, except with extra overhead of manually storing fat pointers yourself (and the bugs that come with this). Instead of using a language with builtin guarantees & optimizations around handling this so your object pools don't need to carry around a bunch of function pointers. C++ has aggressive de-virtualization passes so at runtime a lot of the 'complex object hierarchies' can be flattened to purely static function calls.
Here I would just like to mention that if you have to rely on "de-virtualization" passes, you're in a miserable situation architecturally. If you have code where the overhead of virtual function calls might be too much to pay, don't do virtual functions then. End of story.
To deconstruct a pool of objects, I don't see what should ever be wrong with a function pointer. The overhead of loading the function pointer will get divided by the number of objects being deconstructed. Care to explain what's the issue here?
1. You're writing code you don't have to
2. That adds runtime overhead
3. That when you screw up has non-trivial security & resource management side effects
This is objectively indefeasible in nearly any vaguely professional context.
2. And the code being compiled is abstract & generic, it won't be instantiated for every type and bloat the executable or instruction cache.
3. Security concerns: With C++ virtual methods every object carries a mutable pointer too (to a vtable containing function pointers). What resource management side effects please?
You claimed there were problems many times for sure, I don't think you came up with any evidence of those problems.
People use them because by default values have scope based lifetimes and destructors unify the handling of stack based values which require no explicit cleanup with data structures that heap allocate, which require no explicit cleanup once the destructor is implemented.
This is the opposite of C where even though the resource cleanup needs to happen on scope end, it is not automatic, and so becomes a source of bugs. This is exaggerated when there are multiple returns, gotos, macros and other sources of branching especially for error handling.
This automation also removes boilerplate so there is less management that needs to be done when using data structures.
What was your evidence or explanation?
I have given a variety of arguments why I think the C++ RAII specifically has downsides. You can accept them or ignore them, but don't act like I didn't give arguments.
> People use them because by default values have scope based lifetimes and destructors unify the handling of stack based values which require no explicit cleanup with data structures that heap allocate,
Once again you're explaining beginner C++ do me. Are you still doubting if I understand this argument? My response to this was and is, if you have primarily stack scoped lifetimes, you are writing beginner programs. This is scripting and plumbing. It's not interesting to me. Systems programming is not so much about stack scoped lifetimes.
RAII systematicing object lifetimes and cleanup, thus enabling exceptions and implicit or uncontrolled control flow, may sound nice on paper, but turns out we shouldn't want them in the first place. What RAII leaves for me is a system that will automatically call nested object's destructors when I delete something. It's not something I want in general, I think it has more downsides than upsides for the code I'm writing. I've actually tried to use RAII many times and I've concluded it doesn't work for me.
> This is exaggerated when there are multiple returns, gotos, macros and other sources of branching especially for error handling.
If I were you, my response would be: The burden of proof is on you, show the evidence. But my answer is, yes that is right, but everything is a tradeoff, and the issues caused by manual cleanup are also somewhat exaggerated by C++ people, and the issues caused by blindly buying into the corset of C++ types are not well understood by C++ zealots (I've done my best at explaining them).
Issues caused by C approach are also a matter of design and approach to programming in general. It also depends on the actual problem you're solving. The linux kernel for example doesn't exactly have the easiest or most beautiful code to follow, but surely it's one of the most technical codebases out there, one that solves difficult problems (performant resource multiplexing for programs that it doesn't even know). On the other hand, the debugger code base I've referenced doesn't have any gotos for example, nor even early returns (or almost none, not sure).
Find me a single example of programs that run as fast and are as productively maintained as these two codebases in their respective areas? I believe there aren't any.
This should be plenty evidence to a reasonable mind, but that isn't you.
I don't think you did, I think you just said it has downsides over and over.
Once again you're explaining beginner C++ do me.
Don't ask for basic knowledge then get upset when you get it.
My response to this was and is, if you have primarily stack scoped lifetimes, you are writing beginner programs.
This is what the vast majority of values have. Some escape one scope and get cleaned up in another one. This also includes values inside data structures.
It's not something I want in general, I think it has more downsides than upsides for the code I'm writing. I've actually tried to use RAII many times and I've concluded it doesn't work for me.
Again, this isn't an explanation, it's just you saying "I don't think it's good, I don't like it", but you aren't explaining why. This isn't evidence, it's just you restating "this is bad". Why is it bad? "I told you it's bad!!".
thus enabling exceptions and implicit or uncontrolled control flow
Exceptions aren't uncontrolled flow and you can also not use them. Something else being enabled doesn't mean the unrelated feature is bad. This also implies that every other language that isn't C is terrible because they have some sort of automatic cleanup. Now java and python are unstructured by your own definition?
If I were you, my response would be: The burden of proof is on you, show the evidence.
You don't have an explanation of why the burden of proof is on me. Everyone uses these techniques in system software now. You are the odd person out, that's why the burden of proof is on you.
I also gave you a very good explanation and you said "you're explaining basic C++ to me". I am, you asked for it and you didn't poke any hole in why it's wrong.
the issues caused by manual cleanup are also somewhat exaggerated by C++ people
It means memory leaks and crashes which everyone has been fighting for 50 years.
This should be plenty evidence to a reasonable mind, but that isn't you.
Someone running a marathon without shoes doesn't mean shoes are bad. Just because an old program is written in C, that has no bearing on if other tools are good or not. This is not a logical conclusion. Unix was written in C too, does that mean destructors are bad? No one said it's impossible to write a program in C now that C++ exists.
Why can't you give a single basic explanation instead of making the same claim over and over?
Show evidence of your claim.
> Don't ask for basic knowledge then get upset when you get it.
Where did I ask for basic knowledge? You are hallucinating. I've repeatedly and explicitly told you to shut up explaining beginner things. You are trying to refute my criticisms of RAII with your silly explanations of how RAII works. That's not how an argument or "evidence" works.
> Exceptions aren't uncontrolled flow and you can also not use them.
I would rather call it _implicit_ flow but they also lead to losing control over control flow because of that. Yes, I can just not use exceptions, and also avoid gotos and use them in a principle way, and do cleanup in a principled way. And then I can also do without RAII and without all the baggage I'd have to buy into in order to use it.
> You don't have an explanation of why the burden of proof is on me.
Please p* off man. You're being ridiculous.
> It means memory leaks and crashes which everyone has been fighting for 50 years.
Like you have been on your Linux, Windows, or Mac machine? Maybe you are running a linux machine with an uptime of weeks, months, or even years?
> Someone running a marathon without shoes doesn't mean shoes are bad.
You claiming to run a marathon with a 25kg backpack doesn't mean you don't lie, and it also doesn't prove running without it isn't the better approach.
> Why can't you give a single basic explanation instead of making the same claim over and over?
The only person not giving explanation beyond basic boring introductory C++ RAII school, that's YOU my friend. I've been giving a lot of explanations what I think is problematic, somehow it seems you're mentally unable to acknowledge them.
Show evidence of your claim.
That would be proving a negative, which is another logical fallacy.
Please p off man. You're being ridiculous.*
If that were true you could explain why.
I would rather call it _implicit_ flow but they also lead to losing control over control flow because of that
But it's just running something you need to run anyway and it's optional. Destructors don't need branching, they can just run a single function like free.
And then I can also do without RAII and without all the baggage I'd have to buy into in order to use it.
Again the "baggage" is the part you can't seem to explain.
Like you have been on your Linux, Windows, or Mac machine? Maybe you are running a linux machine with an uptime of weeks, months, or even years?
Are you trying to say people never write memory leaks or forget to clean up resources in C? Something being done with huge amounts of efforts doesn't mean a better way to do things is bad.
You claiming to run a marathon with a 25kg backpack doesn't mean you don't lie, and it also doesn't prove running without it isn't the better approach.
This doesn't mean anything. Now you're trying to say (without evidence) that someone is lying? Destructors don't add any size to a program.
Is this your only argument after dozens of messages? Running a function with no branching destroys control flow and running a tiny function that runs what you need to call anyway is bloat?
I've been giving a lot of explanations what I think is problematic,
No, you just said it's problematic over and over, that's not an explanation.
"Please give me examples of great systems software written in plain C or C style procedural C++"
"Please give me examples of great systems software written in modern C++, as opposed to procedural C style C++ or plain C".
Why does it seem like the majority of core infrastructure, like Linux kernel (also vast parts of Windows/Mac OS), PostgresQL, Sqlite3, Git, OpenSSH, GCC, but also e.g. MS Office/Excel, are written in plain procedural code and not in modern C++? For the latter, I get stuff like "ScyllaDB NoSQL database" or "Envoy Proxy" or whatever, what is that even and why should I care or how do I know they're actually technically solid products and not just hype?
My claim: The arguments that I've given (but you keep insisting I didn't give them) play a huge rule in why that is so. You can use C++ RAII for plumbing and high-level stuff. But if you try to actually implement technically interesting things, it won't help you at all, it's just getting in the way.
Why don't you do it so you can make your own argument.
Why does it seem like the majority of core infrastructure,
Why are projects started decades not using modern C++ that came decades after they were started?
This doesn't have anything to do with destructors.
The arguments that I've given (but you keep insisting I didn't give them)
You didn't give arguments, you made claims based on other people doing something different decades before modern C++ existed. Someone managing to dig a hole with their hands doesn't make a shovel a bad tool. Animals dig holes all the time, people are better at it.
But if you try to actually implement technically interesting things, it won't help you at all, it's just getting in the way.
This again is your claim which you haven't been able to explain in any way.
If you just don't write bugs, then yes all languages are equally robust, including assembly.
Zig, like C, is simply not a robust language. I don't know why this feels like something contentious? It's clearly not intended to be robust?
I don’t think it makes sense to say Zig is or isn’t intended to be robust in general. Like, we don’t say Rust isn’t robust since it doesn’t add dependent types and general purpose static verification that can do more general proofs. It’s focused on eliminating one class of memory bugs in particular, exactly the class of bugs that are the biggest challenge for software like Bun, and other software with complex lifetimes (it originated from Mozilla and Rust is perfect for browsers)
Zig is intended to be robust for software like TigerBeetle, or the Zig compiler itself, where memory lifetimes are simple.
I’d say the focus on built in tests, fuzzing, debug memory allocators and safe mode shows that Zig is absolutely intended to be robust, within the scope of what the language aims to be. Far more than C itself or most of its popular compilers ever did.
Give it a few years! I've noticed an explosion in interest in formal verification recently, especially since nowadays the bar to entry is so low: just ask your LLM agent to give it a go.
Despite TigerBeetle being one of the highest-profile remaining Zig projects, I actually don't think they're representative of the average Zig project at all.
Not a lot of people write programs this way.
"Kelley describes why he created Zig, when other options including C, C++, Rust, and Go already exist. He said he set out to develop a digital audio workstation. He tried Go, but found interoperability with C libraries difficult, and said the garbage collector caused audio delays. He tried C++, or coding C-style using a C++ compiler, but found that small mistakes led to memory corruption bugs that took weeks to fix. He tried Rust but "really struggled to write code that would satisfy Rust's rules," and spent a month trying to make font rendering work."
https://www.theregister.com/software/2026/05/28/zig-creator-...
I've even seen it on some simulation software's core that was written in the 80s originally; at the time memory was much more constrained so allocating upfront meant you could check upfront whether the simulation could actually run or not vs crashing out part way through.
See here for a post on their deterministic simulation harness: https://tigerbeetle.com/blog/2023-07-06-simulation-testing-f...
And I think embedded software is a field where Zig will be at its best. The only thing it is missing is maturity. When project lifetimes are measured in decades and changing a single byte can cost millions, no one in his right mind will pick a language that is still in development. Things will become interesting when it reaches 1.0.
Allocating in large chunks is often not very performant which is why people came up with tools like the borrow checker, you often want to allocate and deallocate dynamically on a need-basis but that's exactly where bugs occur.
Of course you could argue that on average, most programmers are not going to have the right practices and skill, so on average you should prefer Rust. But that's unrelated to the argument I was making, and in any case not a very interesting point in my opinion.
The problem is mostly people graduating from school thinking that somehow there is only stack and heap, and malloc/free is how you do heap. That view completely ignores that the essence of programming systems is mostly to understand the machine, and then doing conceptual and architectural work on a solution (and also on a problem). The act of writing actual code is then mostly just translating those concepts into the digital world verbatim.
"Extraordinary claims require extraordinary evidence" -- Carl Sagan
You could write a JS engine with Zig-like idioms (arena allocation, static initialization), but that would require re-writing the whole JS engine from the ground-up (though I would definitely be interested in it if someone actually tries to do it!)
The C++ interop of Swift is perfectly fine, to such a degree that FoundationDB is now using it effectively alongside its C++ origins.
Arena allocators & static initializers are not novel. You'll find them in high performance C++ projects as well, such as LLVM or JavaScriptCore. But arena allocators have the quite significant limitation that they only help when everything being allocated in them have approximately the same lifetime. So they don't help when you need to allocate memory to provide the native implementation of a JavaScript object, for example (eg, FFI).
Could you actually? That seems like a bad fit for a JS engine to me. Predictable memory requirements are great when you can have them, perhaps you can avoid complexity then, but for a JS engine?
That's just it, using Zig required more rigorous engineering than the Bun team were capable of.
Who are these "people" you speak of? It's possible to write software in low level languages that don't have these problems. Not a "non-zero" it might be possible, it can be done thoughtfully, and the popular notion it can't be done is backed only by incomplete anecdotes.
Should everything be written in low-level languages? No, that would be absurd. Is it a simple fact of life that not every person/team/organisation is capable of meeting certain standards of rigour? Yes. That's not to say anyone in the Bun team could not become sufficiently competent in the future. For whatever reason, current experience, incentives, and personal motivations did not make for a conducive environment to make Bun watertight in Zig.
Zig does help you. Array slices, explicit nullability of pointers, defer errdefer, explicit allocators, built-in leak detection, bounds checks, overflow detection, the list goes on. If you need to play around on that side of the fence, Zig gives you a lot to make sure you don't mess it up. If we were talking about C I'd give you your flowers, but we're not. The most common issues and vulnerabilities that crop in C from manually managing memory are strongly mitigated by a quarter of that list.
The good news here is that we have more than just anecdotes to support this, we have empirical evidence.
Grouped allocations works especially well in parsers & ASTs where the lifetime is very bounded. Since the Rust rewrite, we still use arenas for Bun’s parsers and the bundler but not a ton elsewhere.
Not if you use segmented arrays
"At the time of writing, about 4% of Bun's Rust code sits inside an unsafe block (~13,000 unsafe keywords across ~27,000 lines / ~780,000 lines), and 78% of those blocks are a single line — a pointer that came from C++, or one call into a C library. I expect this number to go down over time as we refactor from a faithful Zig port (which had no greppable unsafe keyword) to idiomatic Rust, but we are going to continue using C & C++ libraries like JavaScriptCore so it will always have more unsafe than pure Rust projects."
It's unsafe to call a C library that gives you a raw pointer, that can be a single line. It's unsafe to use that pointer, that could be a second single line. Carrying that pointer around, the data structures it's in, that's all safe, and doesn't implicate lifetime checking at all, so Rust will let you do silly things with the actual lifetime.
A better metric would be absolute unsafe keywords (because each carries a need to review), or "code in a module that uses unsafe anywhere" vs total code, because module boundaries isolate unsafe.
In short, if you use an unsafe block, then potentially any part of your code is unsafe.
The way I think about it is that the unsafe keyword is a promise that you’re going to maintain the safety invariants yourself, manually. The program is memory correct if it correctly maintains a certain set of invariants. Unsafe punts responsibility over those invariants to the programmer. If you use unsafe and mess up, the program may be in an invalid state. Yes - it might crash, anywhere. But the bug is still almost always in an unsafe block.
It’s often possible to make fully safe wrappers around unsafe code which maintains all those invariants manually. (Either statically or dynamically.). A lot of the rust standard library does this. For example, Vec, Box and slice all use unsafe code internally to create safe APIs.
That's entirely dependent on how you write your Rust code. If you're derefing an invalid pointer then the bug is usually in how you calculated that pointer value, but the only part that actually requires 'unsafe' is the deref, not the bugged pointer calculation.
Now in properly written Rust code you should be marking all of that code as 'unsafe' in that case and documenting what invariants need to be maintained, but that's entirely on you to do. The only part the compiler actually enforces is that you mark the specific spots where you make use of the operations that 'unsafe' allows.
There was a world where everyone just used the unsafe keyword everywhere, and Rust code crashed & had memory issues all the time.
The big thing Rust did wasn't invent borrow checking or memory safety; it's the ease of use, the defaults, and the social aspect of "if unsafe is used, a memory bug is in the unsafe part".
Yes, I wish there was a way to mark code as unsafe without also allowing unsafe operations. Its quite common I have some "safe" code that generates values which are eventually used in an unsafe block. If the "safe" code is wrong, my unsafe code will fail in memory-unsafe ways. But there's currently no way to annotate this sort of "safe" code. Rust provides the unsafe keyword - but that keyword is generally reserved for code which needs to actually make unsafe operations (like derefing pointers or calling other unsafe functions).
The fact that 'unsafe' also enables all unsafe operations in the function body just makes it even messier, since you definitely _don't_ want that unless the entire body is really nothing but unsafe operations. Thus the "mark a safe function as unsafe" has a clear downside since it allows all the unsafe operations you didn't want to use.
This changed in the 2024 edition to now fire a warning lint, asking you to wrap your unsafe operations in an unsafe block, even inside an unsafe function.
This is a common misconception, in a literal interpretation. It doesn't invalidate your point, but since people get the wrong impression: unsafe doesn't turn off any of the checks the Rust compiler does.
It allows you to perform 5 additional operations, and that's it. Using those operations wrong is what breaks the safety promises of the language. As an example, you could dereference a raw pointer and tell the Rust compiler it lives forever, when it's really a pointer to an object that's about to be freed.
A garbage collected language does this automatically. Rust still requires thinking about and tracking memory lifecycles, but the borrow checker will complain and keep you from doing it wrong. That's why LLMs like Rust. It gives immediate feedback on what to fix. By-default constant reference parameters helps prevent major performance problems.
Ownership is automatic. You don't have to explicitly drop things when they go out of scope. (although the responsibility for that is split between the compiler and the library code).
Lifetimes are not automatic, but also have no effect on memory allocation. They are purely a static analysis path and you can make a functioning rust compiler that completely ignores lifetimes.
That's a semantic distinction which does not matter in the point OP is making. And contrasting rust static approach to general GC.
Once raw pointer is turned into a T, &T or &mut T, the borrow checker is on.
The only thing unsafe does is let you have an unbounded lifetime. As I said, it doesn't check those:
fn get_str<'a>(s: *const String) -> &'a str {
unsafe { &*s }
}
> if you generously sprinkle pointer dereferences in unsafe code, you effectively disable the protection provided by the borrow checker
You don't disable anything. You wrote a "trust me compiler" block, and compiler trusted you.
Rust won't ever protect from all possible problems, just the ones the compiler handles.
As for the actual unloaded question, "What's the point of unsafe in Rust?" it is to contain and make it easier to identify sources of UB.
Maybe in the future the unsafe code will go down to 1%, bringing that to two orders of magnitude.
Of course, only time will tell if that is true or not, but from experience I’d be willing to bet it is.
Obviously you should try to avoid writing unsafe Rust to begin with.
With unsafe, you’re telling the compiler “I’ve taken extra care to make sure that what I’m doing is safe and doesn’t break your rules” and the compiler can go ahead and assume that you don’t, in fact, break the rules, and therefore can verify everything else as if the rules never got broken.
In less safe languages, the entire program is “trust me, it’s safe”, while in rust only the parts flagged as unsafe are.
The point is that you should only use unsafe when 1. It’s absolutely necessary for functionality or performance and 2. You have verified and are very certain that the code is correct.
That’s a very useful property to have.
No, you're wrong: You can create any lifetime. Proof:
fn oof<'desired>(x: &u32) -> &'desired u32 {
let ptr = x as *const u32;
unsafe { &*ptr }
}
> You don't disable anything.
I said "effectively disable". For example:
fn trust_me_bro<'a>(x: mut u32) -> &'a mut u32 { unsafe { &mut x } }
fn main() {
let mut x = 1_u32;
let reference_a = &mut x;
let reference_b = trust_me_bro(reference_a);
*reference_b = 2; -- Whoops
println!("reference_a: {reference_a} reference_b: {reference_b}");
}
You just listed examples of unbounded lifetimes.
> I said "effectively disable". For example:
Not sure what you meant by this example since it doesn't compile. It seems the borrow checker caught your mischief. So much for effectively disabling stuff :P
You haven't effectively disabled anything; you just (tried to) wrote unsound code that washes one mutable ref as another. This stuff is allowed provided shared refs are never accessed at the same time (for example, panicking upon reading reference_b).
What you probably meant is https://play.rust-lang.org/?version=stable&mode=debug&editio...
But you know what? If you're dabbling in unsafe, you have this big button called Tools in the playground. Choose Miri, then run your code; it will display large Undefined behavior. It even highlights the `trust_me_bro` function.
Hell, run this with UBSAN, ASAN, and other C tools. They will probably catch any such behavior.
You're just splitting hairs and trying to weasel around the fact that yes, you really can create any lifetime you want using unsafe.
> Not sure what you meant by this example since it doesn't compile.
That's because the crappy HN formatting ate some of the characters. Here's the original version:
https://play.rust-lang.org/?version=stable&mode=debug&editio...
> What you probably meant is […]
If you know what I meant, then what's up with your snarky comment about "So much for effectively disabling stuff"? In your playground link, you did effectively disable the borrow checker in safe parts of the code. Next thing you're moving the goalpost, now it's not about external, static analysis tools: "run this with UBSAN, ASAN, and other C tools". I don't think you're arguing in good faith here. Goodbye.
No. I said borrow checker allows unbound lifetime. I didn't disagree with you. I noted you didn't read the argument.
> If you know what I meant, then what's up with your snarky comment about "So much for effectively disabling stuff"?
Because your original code looked like trying to cast mut u32 to pointer.
> you did effectively disable the borrow checker in safe parts
So this code (https://play.rust-lang.org/?version=stable&mode=debug&editio...) runs now?
Effectively means achieving the goal in satisfying manner.
Writing unsound code is less of effectively disabling borrow checker and more of a hack.
Think about it, if using Java unsafe I gain access to underlying HashMap array and do weird stuff to the HashMap invariants am I effectively turning HashMap into Array or am I doing a hack job?
Edit: By that logic since there is so much unsafe in the code base isn't borrow checker effectively already disabled? You can argue semantics but no, borrow checker isn't de facto or de jure invalidated by unsafe blocks.
It's invalidated by unsound unsafe and that's on code writer to fix.
That’s the same thing.
E.g. in C you can write code and say "don't call it outside the situation that this function was written for" and then blame [0] future users of the API.
In Rust you need to actually make sure that your code works, i.e. your safe interface is not unsafe.
[0] The blame game is not a technical solution to a technical problem...
Garbage collection is a method to make programming easier, not to make the resulting program better.
Even with the huge amount of "unsafe" rust currently in bun? https://news.ycombinator.com/item?id=48967630
Of course they’ll iterate and remove the unsafe bits which were necessary for the transition
[0] https://andrewkelley.me/post/my-thoughts-bun-rust-rewrite.ht...
News at 11.
The fact that you don't think I realize what I did is also rather entertaining. Thanks for mansplaining, though! I was lost until you saved me.
“Safe rust isn’t expressive enough!” -> then use unsafe rust.
“Unsafe rust lets you do anything! Even crazy things!” -> then use safe rust. Or just don’t write crazy code?
Does bun actually do anything insane like that in its unsafe blocks? Or are you just fear mongering?
> Does bun actually do anything insane like that in its unsafe blocks?
Who knows? At 10k unreviewed uses of unsafe, I'd guess there are quite a few incorrect ones. LLMs don't produce perfect code (neither do humans), so there's a high probability that at least some of those create UB.
How would anyone even know, it's vibe coded.
All of this is just a (succesful) marketing stunt by Anthropic.
But only as far as "make it compile" is a good predictor of runtime behavior. In C++ for instance, I can "make it compile" and it still crashes at runtime or does other undesirable things.
Yes, and again the "you're holding it wrong" people or "you are not a good enough developer" people will try to do juggling with a chainsaw and lose a couple of fingers in the process
Claude Code's endorsement (and real-world testing) speaks louder than internet discussions that are at this point 30 years old (and probably more)
Personally my take on the entire affair is quite negative, whatever Jarred or Simonw says about it.
I think Bun owned by Anthropic and the entire rewrite with AI is not the real point (even if it's quite interesting, though).
My take is that Jarred, and Bun,didn't demonstrate a serious, adult approach, from "this is my branch, you are overreacting" message to just proceeding with a 1mil+ PR merged in less than month.
The communication is the issue, and it was handled very badly, in a way that impacted trust and divisions.
Was it so difficult to adopt the approach that the TS team adopted for 7.0?
Seems like this is delivering business value.
Is there a huge need to run that typechecking on browsers?
It's not just typechecking, the typescript library is also the reference parser for TypeScript and reference emit. Emitting JS from TypeScript is non-trivial and non-local.
It's not just in browsers, you might want to run the typescript library on the edge or in some restricted environment where JS/WASM is OK but native code is not.
You may want to use the typescript library as a dependency in your own package. TypeScript becoming non-pure JS means that your library also becomes non-pure JS.
Secondly, WASM isn't the right target for the browser anyway, JS and the DOM have to be what you consider "native", for the most rigorous UI projects at least (which building an IDE is). If you want to build a cross-platform UI product that doesn't require installation and has Emacs-like levels of extensibility, JS is the end of the line in terms of language selection. There are no other candidates.
calibre (with python+Qt) begs to differ. And I believe VLC runs on every current platforms. There's also Code::Blocks and Blender. Cross-platform UI is not rocket-science.
A full rewrite in different language that fails to be idiomatic is a step backward operationally, even if you stand to gain on issues the new language just eats for free
Evidence: see the amount of nonsense in the postgres rewritten in Rust story - https://news.ycombinator.com/item?id=48841676 where there are two contradictory claims made:
"postgres is so stable I will never trust a rewrite."
"covering 100% of postgres regression suite doesn't guarantee you have replicated every behavior."
Folks who are okay with LLMs think the regression test suite is the spec and is the guarantee of stability. How else can it be? If you are depending on some behavior not covered in the regression tests, how do you know the next minor release won't break you?
Folks who are against seem to imagine a platonic ideal of PG which conveniently is the original PG implementation by tautological definition. So no rewrite can ever meet their bar.
Decades of intrinsic knowledge. Which a rewrite lacks.
imho the issue is the language used. AI rewrites are cheap and cheap exercises require a great deal of scrutiny and proof of correctness. Simply using regressions test lacks the intrinsic knowledge of a decades old codebase stuck in developers minds.
Claiming a rewrite is better because it passes all tests is a flex, a new version requires a boatload of evidence for people to accept it as an improvement and not just "passes tests, written in rust via LLM so it must be better". Run it in production for a year in a sufficiently large system and you might be somewhere.
You mean the ones encoded in the regression suite? I think your argument is valid in many medium-to-faang firms where the application is going to have encoded business logic that isn't explicitly tested for. If anything, projects like PG are the exact opposite: no business context and regression suites that test every possible scenario due to the accumulation of bug fixes and context over many years.
>> Run it in production for a year in a sufficiently large system and you might be somewhere.
How do you know every release of PG doesn't break in the setting of "a year in a sufficiently large system"?
Spend time enough with a codebase and you’ll know stuff about its behavior that is not encoded in a test suite. Especially when you need to adjust an integration tests due to the modification of an invariant in a dependency.
> How do you know every release of PG doesn't break in the setting of "a year in a sufficiently large system"?
Because the postgres team is professional and will take care of publishing a changelog for what has been modified since the last version.
Regardless of the technical choices, whether Rust is better or worse, whatever -- pgrust popped into existence thanks to one person driving an LLM through 7000 commits in ~2 weeks. It produced something that passes the regression tests. Even as an LLM-sceptic, I think that's amazing.
From that point, though, it appears to have been completely abandoned. There hasn't been a commit in a month, other than a brief tweak and a note that an as-yet unpublished version that's even betterer is in the works. IDK. I don't think we've acclimated to the shock of the change LLMs create, but if the outcome is a forest of exciting new projects that have a bus factor of 1 and little to no collaboration, I think that's a disservice to this profession.
I think this deserves a real response.
First, an analogy. I drive along a cliff with no guardrail. How do I not drive off the cliff? By knowing how to drive. Sometimes people mess up and drive off the cliff
In practice people are operating Postgres as a machine, more than an abstract spec. Minor releases exist, but the changes are made by people operating the machine and who have a fear of making the machine break.
There are also performance characteristics that are part of an informal spec. While you can definitely write regression tests on performance, theres loads of value in stability of internals because people who have problems look into the machine and discuss it.
The internals might change, but there’s a lot of friction. So… you can have a lot of informal knowledge about.
If everyone working on a software stack is bathing in this informal knowledge, then the decision making is based on that. Things like what is meant to be in a minor or major release is understood. And people make those judgement calls.
After all, even if you have regression tests if you’re making changes you’ll need to write new tests! How do you know your new tests are right? That the new behavior is right?
PG is the entire machine. Fortunately we have version numbers, migration strategies, etc. But “here’s a new binary that passes the test suite and… maybe changes everything not covered by tests maybe doesn’t”. Why bother suffering when there are totally reasonable gradual rewrite strategies?
And of course… every bug in the world… got through despite a test suite! What software out there doesn’t have bugs?
And if the behavioral difference in the rewrite does affect people… I guess that’s something right? “Oh this isn’t a bug because it wasn’t covered in regression tests” is not that tenable.
They are not informal AFAIK - https://github.com/PGPerfFarm. A lot of these heavy duty OS projects have explicit Perf testing/nighlites (see Lucene's at https://benchmarks.mikemccandless.com/)
If you are saying the PG regression suite doesn't cover these perf tests - that is fair. I consider perf to be part of the "regression framework" generally.
>> How do you know your new tests are right? That the new behavior is right?
This is a more generic question in the LLM world. Reliable verifiers is what drives LLM loops. If you don't have these, you have no idea if you are actually making progress or you just generated code that returns 42 for every question. You needs something to actually ground the LLM output against. For existing features, the reliable verification is the existing regression/perf suites. For new features, the regression/perf suites should expand to fit. There are ways to go at this that range from ad-hoc (line/branch coverage) to fully formalized (verification-aware languages like Dafny).
They are in extremely short supply right now.
It’s a mix of serious-wanna be, quasi religious, quasi-technical people running the show right now.
I am so happy to I got to read about the 90s-2000s tech culture, and experienced the post-2008 startup culture.
Just about 10 years ago people were really serious, but chill-looking. Somehow that got flipped.
Sorry, but I'm calling bullshit. I know there's no absolute smoking gun here, but it's all a bit too cute to be unintentional. They are very very smart people. They knew what they were doing.
Wow, so you can't even find circumferential evidence of circumferential evidence? I guess you can believe whatever you want, evidence be damned
Terrible project governance.
TS 7 is a good example of what a responsible team that cares about their users should have done.
It’s clear seeing your post here a couple days later what’s going on. It’s deliberate. I can only speculate on their reasons why but it seems to be an intentional form of disrespect on the engineering culture. “Your norms, your work, your principles” mean nothing. We can spend money and invalidate your life’s work. We don’t need your consent or support. That’s my personal guess at a surface level.
And so, the FOSS project "Bun" silently dies in darkness and is now something completely else. I'm glad I still had "Investigate if Bun is worth it" on my TODO list and hadn't yet gotten to it.
What is the governance structure for Bun by the way? Couldn't find any documents/explanations about how it's supposed to work. I'm guessing it's essentially just "Anthropic decides what gets done and accepted" today?
Though most will be forgiven to not reading it since well it's all AI anyways. I don't know how I feel about all this yet, maybe someday.. ooof
Where? Is see no CHANGELOG in the root. I do see LATEST, last modified 2 months ago, and it says 1.3.14
> Upgraded the bundled Bun runtime to 1.4
(Though that doesn't clarify that this is the Rust rewrite)
But this is HN, a subsection of which places a high importance on what language something is written in, moreso than what it does (feels like).
In the current world of cyber weaknesses everywhere, infrastructure must be solid and rust is a better choice than zig for something like bun.
In other words, bun is not a demo of zig programming that happens to do useful things. Bun does useful things, and must do them securely. That is its number one priority, language is secondary, inasmuch as it helps doing useful things securely.
Rust protects against a rather small group of memory error, if and only if the unsafe sections are actually correct. And nobody knows if they are correct or not, plus there can be millions of logical errors that can be exploited.
See how many CVEs uutils has… quite a lot for a supposedly secure rewrite of GNU coreutils.
But I guess my opinion is null and void since I "slop-code"
Anyone who actually believes that I can't have ANY sound opinion on Zig after 31 projects building with it should get their head examined for brain tumors. I mean, it's not like I have to deal with all the positive and negative effects of things written in it. It's not like I've had to solve problems in it that emerged when the LLM didn't do things correctly (and yes, I can tell when this is so, and not just when things are buggy). And it's not like I don't know (by heart, not by LLM) at least 15 other languages that I can draw on (C, for example, for which I not only got an A in a class at Cornell once, but was in fact the person with the highest score in a class of ~150 people... A perfect score, in fact) to form valid opinions in combination with how I see Zig working.
So please take your baseless skepticism and baseless assumptions and shove them up your ass, dude. I have work to do, and I'm far more capable of it now than I was, thanks to the LLM boost.
And no, I don't think you can have any valuable opinions on Zig since you don't write it. In your analogy you're not even driving the car, but it doesn't surprise me that you can't figure that out.
the ones learning how to run the looms safely and productively are going to continue to succeed and scale upwards and outwards while you get paid pennies to use knitting needles on bespoke hand-coded solutions. And at this point, I don't give a damn anymore because you lot deserve your upcoming rot 1000%, especially with a stubborn pissant plebe fucking answer like that. Holy shit. You're literally going to stubborn yourself into poverty just to make a point. Enjoy.
> Rob Pike, an unsafe guy, may insist that a small language and garbage collection are enough. But the typesafe visionary Grzegorz Wielbodłąński understands the deeper truth: every invalid state permitted by a compiler is a tiny act of civilizational sabotage.
https://x.com/typememetics/status/2078572052680790237
There is no sense pretending anymore that languages with as many gotchas as C and Zig are "simple". Quite simply, they are not. The complexities lie in what's left unspoken, unaddressed, and of course undefined; failure to comprehend these completely can be catastrophic. Rust is so thoroughly and obviously the correct thing that using anything less correct should be taken as a sign of engineering incompetence.
This is where I’ve arrived too. After nearly 20 years of programming I feel done with leaving foot guns lying around. Languages that are full of them, but “fine” as long as you take extreme care are not interesting to me. I’ve started calling this “simple as long as you avoid all the foot guns” idea the sword juggling fallacy[0].
It’s too bad about Zig because it has many nice ideas, in particular allocators, that I hope Rust or some future language will adopt.
0: https://hugotunius.se/2026/06/15/the-sword-juggling-fallacy....
there is a potential future where an llm with unbounded patience would have an easy time literally walking through the zig code looking for one of the footguns might miss a footgun in rust because it's abstracted away. the ones in the stdlib should give you pause. what is hiding in some third party library's code?
AT&T was the one doing the Cyclone research.
And then we have a bunch of folks assuming they know better about C that its own authors?
Barthes aside, C's changed a lot in the years since they left.
Still they are the authors, and withdrawing from WG14, going their own direction with Alef, Limbo, and finally Go, kind of proves the point they considered C done for its original purpose.
I don't think it proves anything other than the wanted to work on something else.
Something was designed in a certain way, accomplished its goal of writing portable UNIX, and then design by committee took over.
The authors, with their experience moved on, the committee and the audience doubled down on the flaws.
> One probable outcome of an unsafe-language-dominant world is full Andrew communism, which is precisely what Andrew “Smelley” Kelley proposes: abolish ICE and replace the market economy with a globally distributed network of community-maintained allocators.
> Rather than a commercial product, software becomes a “public good,” ultimately provided by the Zig Software Foundation as manually managed digital infrastructure.
Calling him "smelley"? Really? Pulling in the political views of the founder of a language and claiming that they inform his language design as part of a grand conspiracy against Rust/capitalism? Where are they getting all of this? Free, open-source compilers and related tooling (both of which Rust and Zig have) are already public goods managed by global communities of mostly volunteers
> This future strikes me as a dystopian hellscape, but I have never met an unsafe-language advocate who could explain how this ecosystem is supposed to sustain itself without unpaid maintainers, ideological volunteers, subsidized infrastructure, and endless donations to compiler projects that still cannot protect users from basic memory errors.
How is Rust any different than Zig in this respect? Both are projects largely run by nonprofits relying on communities to promote, maintain, and improve them.
Frankly, that entire "Institute for Type Safe Memetic Research" seems like either an elaborate satire or an account run by genuinely insane people who think type and memory safety are paramount societal issues.
So. Rust evangelists ? :)
But yes, it is wild that people can honestly quote such drivel as evidence of the language and its philosophy being superior.
Each time I try to evaluate Rust, I do so on its own merits. Going online I am rapidly convinced that I don’t want to be in the same room as a large chunk of its userbase. Statements start to sound a bit like that one time I was dragged into a pentecostal meeting.
BTW did you hear about this:
https://babylonbee.com/news/democrats-say-they-are-perfectly...
OR
Turnabout is fair play.
> Frankly, that entire "Institute for Type Safe Memetic Research" seems like either an elaborate satire or an account run by genuinely insane people who think type and memory safety are paramount societal issues.
You've been had. Your entire rant means the bait worked.
It's a shitposting, hyperbolic bait account. As funny as calling Andrew Kelley Smelley, it's not meant to be taken as gospel.
Or they will not… how could you possibly know.
Everything is in the open: PR #30412
If you can't find an update at that place in the codebase saying 1.4.0 you can ask where it is. It seems after doing that and not receiving an answer would be the appropriate time to start making the claims that people are throwing around.
This seems like a candidate
https://github.com/oven-sh/bun/commit/b18bf6d1d0a92238f240bf...
Ok, go ask those people what they think then, I feel no outrage over a project switching language or merging whatever. But when once-semi-transparent projects become more opaque after an acquisition, I'm starting to readjust anything to the project I had in mind.
> The most recent release of Bun on GitHub is currently v1.3.14 from May 12th, so that v1.4.0 version number in Claude supports them shipping a preview of a not-yet-released Bun version.
> And so, the FOSS project "Bun" silently dies
There’s no implication that it’s because of the switch to Rust.
I was responding to your GP comment. For weeks, it was uproar over the blanket AI move to Rust and now it’s this reason. That’s why I said the goalposts are being moved.
What?
Now, /you/ are trying to move the goalposts into an entirely different argument.
Very likely it’s just anti Anthropic altogether.
Like I said, a meme.
Anthropic hasn’t done anything wrong but if they move a single inch then people will complain. Perhaps it is just a preview version?
I’m sorry but this is a nonsensical debate.
> Now, /you/ are trying to move the goalposts into an entirely different argument.
You said something and I responded to it. That’s invalid too now? There can only ever be one single objection in a conversation? I’m not permitted to react to the things said? Like I said, nonsense.
I'm just one person though, and have no problem with a move to language X, using tool Y, who cares?
For "goalpost are being moved" to be valid here, you have to be able to point to someone saying one thing at one point, then same person putting a different target later. You're folding a huge group of people into one person, when that's not really accurate to do.
It's true that we don't know exactly which, and whether or not there are any other changes added on, but that is the case for every single release of Claude Code that has used bun.
from strings:
bun-v1.4.0
f6d0fcd24abd48061873c2f1a6fb2a67eee487b8
Upgraded.
Welcome to Bun's latest canary build!
(In contrast, earlier builds that I have locally have commit IDs that can be found in the public repo). I don't think it's particularly damning for them to vendor a canary release, mind you.
Stay your pitchforks.
At this point, charging is the least of your issues.
I think it’s valid to object to. If 1.4 is good enough to be used by Claude, why not us?
…but they have? Claude users are public, no? If they were dogfooding internally I’d agree with you.
You just can't win.
You could! You could not release a “pile of vibe coded slop” onto people’s machines. There don’t even appear to be any particular advantages to Claude upgrading early. If there are no one has been made aware of it.
I believe you've just made your opponent's point for them.
But more to your point: Do you believe in "guilty until proven innocent"? Do you believe that they're doing bad things with the bun code, and you can't be convinced otherwise in your judgment of them until they release the code?
> Do you believe in "guilty until proven innocent"? Do you believe that they're doing bad things with the bun code, and you can't be convinced otherwise in your judgment of them until they release the code?
Hell yes, and anyone who trusts one of the AI labs for anything important is insane.
Whether it's their initial mass piracy and then lying about it, dealmaking with the US government, or continually shifting opt-out lines on what they are and aren't training on, they haven't earned the trust to skip the "but verify" step.
Their economic incentives are too transparently aligned with ends at any costs.
That might even be the reason why they switched to rust.
Don't worry too much.
Theres thousands of PRs from claude agents or whatever.
Yeah really feels like a worthwhile place to contribute lol.
FOSS as in FOSS - Free and Open Source Software. Has nothing to do with if it's community driven or not, I'm strictly talking about FOSS.
I feel weird having to defend reality; reality being that it was merged nearly 2 months ago and tons of people have had their pitchforks out without a shred of actual evidence that this made bun worse in any measurable way. But they still insist it was a mistake.
I’ve never met Jared or the bun team but I don’t understand all the personal attacks, I just feel the need to correct the facts. Literally who cares what language a JS toolset is written in?
It’s not like they ported JavaScriptCore (the actual JS runtime) to rust even. All that stuff is largely untouched.
Maybe you're being coy by asking a rhetorical question you already know the answer to but I'll answer as if you asked sincerely...
There are 2 different groups interacting with software products:
(1) end-users : this is where the "Who cares what language it's written in?!?" is usually applicable. E.g. The finance guys using MS Excel don't care whether it's written in assembly, BASIC, or C Language.
(2) code contributors and/or programming language enthusiasts who see other projects as "validation" of the whatever language they've invested in: these people definitely care.
For all the decades that computer languages have been debated, Group (2) will always discuss projects language choices. E.g. reddit.com switching from Lisp to Python, the Linux kernel fiercely debating future Rust contributions , the Typescript compiler switching from Javascript to Go, Bun switching from Zig to Rust, etc.
People try to lecture others in Group 2 about "don't make a programming language your identity" ... but people are human and they can't look at all the above language choices as totally detached observers. They like to talk about it!
If one is a Zig coder that contributed to the previous Bun Zig codebase, we can't expect them to be neutral observers.
There are not nearly enough disenfranchised Bun-on-Zig contributors to make a dent in this conversation. There are lots of Group 3s in every similar convo, for any combination of technology, project, and acquirer you can name.
And for most part I don't think it was communicated well apart from the last blog post.
Personally I chose not to disclose my findings in public to avoid this kind of outcome. Having a good laugh with others is not as important as keeping the result of this experiment undisturbed.
This also eliminates the chance of another round of marketing of LLM ability based on actual human contributions and I suggest everyone who is able to analyze the code to do the same - so that they cannot misattribute our ability to LLMs.
You just AI generated 1 millions lines of code claiming it's for safety. Who exactly is to make any kind of security guarantees about this?
With regards to AI-generated code, this means the copyrightability (at all, not with regard to who owns it) turns on whether or not a human was substantially involved in its creation.
Existing decisions require evidence of pretty heavy and continued human guidance to qualify.+
To wit, autonomous agent created code (prompt -> machine churns for hours -> output) is explicitly not eligible for copyright.
Functionally, this is a double-edged sword.
On the one hand, it means anything coded with autonomous agents by Meta, Google, et al. can be legally reused if it leaks (because no one could hold copyright on it).
On the other hand, it leaves copy-left open source licenses in a weird place. If you convert open source (even MIT/BSD-style) code into something else with an autonomous coding agent... the result has no copyright (nor can ever in the US).
In this instance, Bun was MIT-licensed, no? Then it was shoved mostly-autonomously through an LLM for the port.
Now Bun-Rust is technically still MIT licensed, but if push came to shove it seems like US law's current position is that Bun-Rust would now have no copyright license (because the manner in which it was developed renders it ineligible for copyright).
That's on the copyright side.
On the infringing usage side (i.e. whether you were entitled to shove a copyrighted work into a coding agent to produce something)... that's still TBD.
^ https://media.cadc.uscourts.gov/opinions/docs/2025/03/23-523...
I think it's the "infringing usage" question that is more interesting. If the LLM trained on GPL-derived code, what does that mean for the end result?
Either it means nothing at all, or it means that a substantial portion of the code produced at some of the world's most valuable companies over the last two and a half years is GPL.
Issues on GitHub was mass closed claiming zig is no longer relevant without fixing the real issues.
In fact a lot of "feature requests" like npm compatibility related or missing specs etc ALL also got closed for being "zig" because reproduction and suggested solution were of course zig.
Again doesn't mean it got "auto fixed".
Anthropic, apparently, not the least because they blanket-closed all issues submitted before the rewrite. If the language were irrelevant, it shouldn’t matter for the existing unfixed issues.
the language is relevant for political not technical reasons. you could say, blanket closing the Issues, getting rid of this thing called Jarred from your stack, etc., is really an attack on the GitHub and the GitHub Agitator Lifestyles, which is why so many commenters on Hacker News actually care.
You'll need a citation for that. While I'm still not sure whether Claude Code's issue tracker is worth much (Anthropic doesn't seem to engage with individual issues a while lot), it still has 5k+ open issues ranging back to March 2025. Bun also still has all of its Zig-timeline issues, all 5k+ of them. So what's this about?
> What is the governance structure for Bun by the way? Couldn't find any documents/explanations about how it's supposed to work. I'm guessing it's essentially just "Anthropic decides what gets done and accepted" today?
And this is a valid question. You are not "defending reality" by refusing to listen to it.
In summary and based solely on my understanding:
- Jared misled people about the intentions of the migration. It's not the worst thing in the world, but it's certainly worthy of criticism.
- Jared has commented before about locking out human contributors from open source projects. Whether he was making a larger point is irrelevant as his comment stands on its own.
- Other Bun contributors, past and future, outside of those employed by Anthropic, did not and likely will not have equal access to the model Jared used for the rewrite.
Jared, working in the public Bun repository, used tooling not available to his community to experiment with a signficant migration. He dismissed all concerns and told people it's just a bit of fun, and that it shouldn't be taken seriously. Most of the controversy would have been avoided were the experiment done in private.
None of this is a big scandal but questions about the project are entirely justified.
Similar to the Claude Code 60s timeout incident [1] it could just be "Jarade decides what gets done and accepted"
It seems like when some folks see “open source” they think the project maintainers owe them something.
An open source MIT license of the source code is very different from being the original copyright holder.
This is even more interesting given that prewar Bun was not a well respected code base. This matters if the code was bad, were the tests bad / not comprehensive as well? If so, the translation to rust whose sole / primary target was test-passing will have a fair amount of undocumented bugs in it.
> poster child for Zig programming language actually being the prime example of How Not To Write Zig Code
https://andrewkelley.me/post/my-thoughts-bun-rust-rewrite.ht...
I don't consider the following comment to be good, honest, and open communication:
> There’s a very high chance all this code gets thrown out completely.
... and a few days later he decided that the code wouldn't get thrown out. That doesn't make the comment itself dishonest, he didn't have a crystal ball.
You can absolutely argue that the rewrite was not communicated clearly once they committed to it, but I don't think the comment you linked to there stands as evidence of poor communication at the time it was written.
You don't create PRs like this when you're working "completely in the open"
test ci #30412
No description provided.
Rewrite Bun in Rust #30412DRAFT: DON'T MERGE CI TEST
and then they forget to close em when they go another route and they sit around for 2 months until they auto close or they fix it up into something reviewable. It's been a thing at literally every place I've worked are you really saying you don't see that?
The rust code is not yet in a stable release, and CC only switched to it in the latest version, so there's not been 60 days of widespread battle testing yet, so i don't see how we could possibly say it will have no negative difference. It hasn't, because nobody is using it yet. It still might.
>Literally who cares what language a JS toolset is written in?
I don't care what language it's in, but i do care how it is written and maintained. I'm not going to trust my projects to a runtime that was written by an LLM in two weeks with practically no human review
I'm concerned that the complete rewrite in an entirely different language is not a sound technical decision and instead is a ploy to shed copyright claims from past contributors.
Now, based on comments from this thread, the formerly FLOSS project is somehow granting special access to a corporation that apparently is invested in going way out of their way to push implementations that consume the complete rewrite before the world has access to it.
I for one won't be touching Bun. This doesn't pass the smell test. It feels like a bait-and-switch in progress.
Even if this was a terrible technical decision, it is a good, maybe even unavoidable, business decision. And I don’t think moving to Rust was a bad technical decision at all, given Zig is unstable and constantly changing, and that it’s not memory safe, which is a big problem for something like a JS runtime!
The reason it’s a necessary business decision is that Bun is owned by an AI company and Zig has a policy to not accept AI assisted contributions, making it impossible for Anthropic to contribute to a language that is still evolving rapidly, not to mention the bad marketing of using a language that basically is hostile to your product.
No,not really. There is only one point: does Bun acknowledge that past contributors still hold copyright over the Rust rewrite?
And this contribute to oss voluntarily but hold copyright and use it as some sort of leverage is bullshit, in my opinion. Don't contribute to someone else's project if your concern is copyright.
However, people also report that the canary version doesn’t identify itself as version 1.4.
Doesn’t sound like responsible software deployment
They are basically testing the rust port on user machines.
Who come that since the rise of AI all principles of software development are thrown over board?
We have run it in production for a month: https://www.prisma.io/blog/bun-rust-rewrite-prisma-compute
Now it is a Deno clone, both without anything enticing over node.js.
I don't care about meh Zig being rewritten to bad Rust if it does the same thing, but taking what is presented as" look at this funny experiment I did" and then taking that into production with barely any announcement is what kills off the interest to me.
Bun has been a great ad for whatever LLM they were using but the cool factor is gone now, and that's really what set it apart from basic NodeJS in my mind.
It's cool that the Zig project exists, but I'm unlikely to ever use Zig so that’s sort of orthogonal to my interests.
The normal process would be "hey, this experiment is going well, we've made this plan, come help us shake out the new codebase for the switchover in X time." None of that happened. There wasn't even an announcement, just a silent commit that trashed hard work from hundreds of community members.
All those PRs can be fed thru Claude and converted to rust and then manual polish can be added.
Bun has bugs? yes. Do they affect vast majority of users? Not really. Been using bun for about 3 years; never had an issue with it.
"Because the discussion was interesting" is why I care.
The worst is "hey they are running proprietary fork and not open source anymore" when the code is available in plain sight. Canary build is available with one command. The stupid article made a huge stink about it without this key fact. Now they added the update but damage is done.
If you would like to join us in this thread now that this misunderstanding has been corrected, please, share your thoughts on the Bun community, what the "Bun community" even is, and how it has been impacted by a sudden, unannounced rewrite.
Bun "user community", is not affected by this at all (yet) except the ai doomers/zig fanatics.
When v1.4.0 is fully released, and if Bun team can't get it to full feature (and no new serious bugs) parity with last zig version within a few versions (let's say v1.4.2), then there will be real effects.
I will be testing v140 after release. As a bun user community member, I expect no new type of crash, slightly faster perf/less memory usage and move on with my life.
A rewrite like this is unprecedented but doesn't mean impossible. I am cautiously optimistic.
This is, as you say, unprecedented. Not impossible, cautious optimism is a fine response. But to bring back a point I made earlier in the new old light: Then let's be cautious. Let's do some announcements ahead of time. Let's get some people in and get some physical hands on the codebase. Let's not go from "Stop freaking out, it's just an experiment!" to a silent merge to main in a week. That causes reduced trust in the project from those even just a hair less optimistic than you, and that's clearly unhealthy for the project no matter how you define "the community."
Jarred earned the trust by providing value. People who say trust is reduced without actually seeing the result, never had trust to begin with.
It's not like they merged and dumped it on to the world immediately. It's available as canary. They are doing their own dog fooding. Prisma is using the canary successfully.
As long as oven team can deliver a zig version equivalent stable bun, it is a success in my book. After that they can focus on resolving as many unsafe as possible and then the real benefits will materialize.
Did Bun v1.4.0 release weeks ago and I missed it? I would call a formal release the point that it goes "into production", and I think dogfooding it via Claude Code is yet another form of testing Bun-in-Rust before the latter goes to casual users.
“bun the JavaScript runtime ” is not dead though.
Any review would get to the simple conclusion that this should not be released before all the obvious bads are sorted out.
In Zig, every single memory operation is unsafe.
And Bun must interface with C code that has no safe interface, necessitating a ton of boundary-level unsafe behaviors.
There's too much, sure, but can we at least be honest and reasonable?
> must interface with C code that has no safe interface
Yeah so the sane first step is to create encapsulated, safe interface for them, especially in a project like this. Deno for instance have ~0.2x as many unsafes.
And mind you if you haven't read the code, the vast majority of unsafe blocks in bun are for raw pointer access to local (Rust) objects because their ownership was a mess both before and after the rewrite. Also funnily enough a lot of the access patterns are wrong (in the Rust sense), leading to hundreds of new undefined behaviors.
> be honest and reasonable
Well, well. Talking about dishonest and unreasonable behavior, why is bun releasing a new version before solving any of those glaring issues? I'd remind you the current new version is not an improvement compared to the previous one, both in terms of correctness and maintainability.
Except, you know, multiple real companies saying that it is and using it in production. And the fact it closed all known memory leaks. And that nobody has a really pointed to a single actual issue the new version introduced after two months of endless, ceaseless bitching.
I'm also fascinated by all these people upset at Jarred for harming the readability of a codebase they've literally never cared about before it become drama. Bun has almost exclusively been maintained by Oven employees since it's inception.
> Deno for instance have ~0.2x as many unsafes.
A project written ground-up in Rust idiomatically, with a smaller surface area, still has an unsafe footprint within an order of magnitude compared to this automated rewrite from an unsafe language with different practices and known bugs? That's not exactly the slam you think it is.
I personally know 3 categories of newly introduced issues spanning the code base with 200+ occurrences, that leads to undefined behavior and memory issues. Not doing any contributions to the code is my conscious decision, which includes not pointing to the specific problems in public.
It's nice to hear that people are doing the same.
Another point of comparison is density of unsafe: the number of unsafe blocks per line of code and/or file. By this metric, Deno has a bit over half the unsafe (because the bun rewrite is significantly more lines of code).
I don't understand why folk are having such a hard time understanding why you do large projects in multiple steps? 80/20 rule? Perfect is the enemy of good?
Was nobody here for moving billions of lines of JavaScript to Typescript? It starts with declarations, then turn on type checking gradually inside the codebase: piece by piece until done.
Known unsound code should not be merged, let alone released to production. If you have good enough tests and run them under MIRI or ASAN, maybe you can get away with it for a time, but most Rust experts would not sign-off on such a project.
If somebody put a gun to my head and told me to make the best of such a codebase, I would try to figure out how to turn off the LLVM optimizations that assume the Rust references don't alias. With these optimizations this codebase is scarier than most C or C++ code.
Can't help laughing at this point.
I do regret that I've also read the code though, otherwise I'd probably live a little happier with ignorance right now. Sadly as they don't know they don't know about the issues, there's no indication of those in the two public materials they have released.
I don't see why either of those would be true. The project isn't dead just because it's in a different language now, and for liveness purposes it doesn't matter how it got to that different language.
A fast rewrite of Bun in Rust has effectively alienated most of the people in the Bun community, so in that sense the “open source project” has died. It’s no longer a community project, it’s just become a personal project again.
I think even this is largely beside the point. The acquisition of any project by Big Capital (whether tech or VC) is usually a killing blow.
It's not about making the dev experience better than node, that use-case is now secondary.
I'm talking about the 1.4.0 release:
> Claude supports them shipping a preview of a not-yet-released Bun version
Maybe Bun/Anthropic fixed this after Simon's initial release of the blog post, but seemingly when we both looked, it wasn't public.
Version was changed in Bun in May 18
https://github.com/oven-sh/bun/commit/b18bf6d1d0a92238f240bf...
So it has been canary since then, most likely, if docs are accurate that latest main is canary.
Nevertheless, the principles of open source never demanded more than providing the source upon request.
~$ bun upgrade [2.42s] Upgraded.
Welcome to Bun's latest canary build!
Report any bugs:
https://github.com/oven-sh/bun/issues
https://github.com/oven-sh/bun/compare/f161e0311...98f664962The fact that Anthropic felt the need to buy a runtime so they could make their TUI better speaks more to the quality of engineering than anything else IMO.
If rewrites are so easy, why not rewrite CC in a native language? Would’ve been a hell of a lot cheaper.
I'm sure there was some logical reason for shoehorning web technology into this stack given that we have a good 40 years or so of experience with interactive terminal programs that use curses/ncurses, alongside emacs, vim, almost the entirety of MS-DOS, and so on.
This does not even speak to it pegging my CPU and keeping the fans running on full blast even when the laptop is supposed to be idle overnight.
You could test this by running "yes > /dev/null" in Ghostty. In Activity Monitor on my Macbook it shows `yes` using 100% and Ghostty using 0.5% of cpu.
Never had to dig through the xterm bug tracker to figure something out. I don't even know what version I'm running, or even what the versioning scheme looks like. I haven't changed the configuration since the first time I set it up. It just works.
With js you get 1 thread at 100% utilization. Power usage and heat scale non-linearly with cpu utilization and 100% utilization on a single threaded js app means you will have ui lag. Other languages like golang would split work across 8 threads and have 8 threads at 20% utilization and this would result in less power usage. Claude Code would have been better off performance wise being an electron app because it would be offloading rendering to the browser and gpu.
Also, the architecture of Open Code is actually a lot better here than Claude Code. ClaudeCode does everything, including rendering in a single thread and it's all in js. OpenCode has a zig based tui renderer it offloads that work onto.
But I will also say these coding agent tuis do get unfairly maligned because they launch subprocesses and those subprocesses tend to be expensive. If you are using Rust analyzer with claude code, it's rust analyzer that's causing the majority of your problems.
Bad coding is just bad coding.
It's also all cpu bound work. The majority of the stuff on your screen is being rendered by the gpu.
You are really, really not. Not at LLM inference speeds at least. Frontier models output maybe 200 bytes per second on the fast end. I guarantee you that idle BSD box is still processing more than 200 bytes of whatever at any point in time. Streaming 200 bytes of LLM output through syntax highlighting, markdown rendering, and diff formatting, will still leave the box basically idle if done correctly. Claude Code has not done it correctly.
> they need a high powered game engine rendering loop
Don't believe the bs on twitter. Claude code source was leaked, there is no high powered game engine rendering loop.
> It needs to stream text to stdout
These apps are doing real work taking text that is streaming in and doing syntax highlighting, calculating diffs, and rendering markdown. If the work was split across threads they would not use anywhere close to as much power and there would not be a slow ui.
If you think this is actually nothing, prove it. Put the code up on github that shows taking a constant random stream of markdown, code, and diffs and displays it in the terminal can be done cheaply on a single thread.
The reason the computer gets hot and uses more power is because the cpu is getting close to 100% utilization.
Put up the code and prove it. We already know exactly how fast golang is in comparison to nodejs on basic numeric computations, it's roughly 40 to 50% faster.
> These apps are doing real work taking text that is streaming in and doing syntax highlighting, calculating diffs, and rendering markdown.
The first argument is that these are expensive operations. They are not. And the second argument is the assumption that these are desirable things for an agent system to be doing. That's a personal preference but, personally, I don't want them and would appreciate a way to disable all of that to reduce CPU use.
A lot of people go “js is bad” reflexively and they are right for the wrong reasons. Its not because js (specifically v8) is slow. I have many websites and web based apps open and none fuck up my laptop as hard as claude code.
The UI itself requires laughably simple tech. The idea that a markdown renderer that can produce HTML at 50MB/s would use much CPU to process incoming hundreds of bytes per second is insane. CPU should be idle even if doing that constantly. Nevermind that between turns it has nothing to do whatsoever and should use zero CPU at all.
Casey Muratori addressed this years ago with the windows terminal drama, https://github.com/cmuratori/refterm
I picked the P90 because I remember running Linux ~1.0 and/or NT on it, with real multitasking and services etc. Those soaked up many of the cycles.
Another story, around Y2k/Win2k was asked to write a C program to record all the files/sizes on the C: in to file. Then something would be installed, my program run again, and the results diffed. Despite thousands of files the program was done before the enter key came back up. Boss told me that he thought the program was cheating.
That’s just how fast modern computers are these days, I said. When running C code anyway. That perf is still available today, compare astral’s uv to slow competitors. Just takes some investment, though probably less when boosted by VC bionics.
— Henry Petroski
I don't doubt this, but I would love details and citations here.
Then something is waking it up and it's not asleep at all. Asleep the CPU shouldn't be running.
I have been trying to keep away in the last couple weeks and it was all win for me. I still come down here sometimes when I am stressed with real work since it is a strong addiction to see “how terrible the plebs are doing”.
My question is moreso “why was it ever JS in the first place”
& the cost of a native rewrite would be cheaper than acquiring Bun no matter how you shake it
I hear some version of this anecdote constantly
yet despite using all these prompting tools since 2019 I have never once heard a lab or professional say “x model can do anything”
Can you provide me whatever specific source you heard for this?
Dario Amodei said in January that he expected software engineers to be replaced [0], so I suppose the quote should be extended to “can do anything… a software engineer can do”
And I think we can all agree that SWE is on the more complex end of all white collar jobs. So, my thinking is that if SWE can hypothetically be replaced, so can many other jobs. Hence, the claim the models can “do anything” intending to capture the majority of white collar work.
This, of course, is if you take their claim at face value - which I certainly do not.
[0] https://www.entrepreneur.com/business-news/ai-ceo-says-softw...
So then its misleading in the extreme to say “I’ve been told AI can do anything”
It amplifies hype when people make claims like this by including hyperbolic statements as though they are reflective of the state of the discussion.
That said, with respect what I clarified, I’m curious your thoughts.
If I missed your point it was unintentional. Instead of being rude, try gentle clarification.
But now I’ve lost interest. Have a good day!
Yeah, they "merely" say that they're about to break the singularity in a couple of years, that it's too capable to be safe to release, that it's going to take 50% (or all, depening on the day and audience) of programmers out of work, that it's smarter than Nobel prize winners, that their programmers don't write anything themselves anymore, and things to that effect.
Again… provide me a quote from someone doing this every day that says: “they're about to break the singularity in a couple of years”
The only people saying this stuff are random Internet commenters, breathless reporters and reactionaries.
The only person on record for a statement like this was Geoff Hinton about radiologists and he has since walked back his statement
They chose it early on, it works, and it makes obscene amounts of revenue. End of story. That doesn't mean it was the "greatest" choice, or has a perfect technical architecture.
Rewrites are never easy, even the bun rewrite. But a non-UI developer tool with a rigid API surface contract (and associated tests) will always be easier to trust after a rewrite than a partially tested UI tool with ambiguous functionality.
And the question of “if AI is so amazing, shouldn’t it enable easy development of a native TUI even if JS is easier at first?”
Don’t get me wrong I find great value in coding agents daily. Just finding the hype cycle tiring.
the era in which the tech industry built software to last was over long before LLMs, especially VC-backed startups.
I'm pretty sure "built to last" was only ever more true in specific contexts, and wasn't particularly more true than today.
Edit, the sonnet question shouldn't be taken as proof, it knows I'm a rust dev.
Nobody but a JS dev would choose react-on-terminal shenanigans over it
I could see CC getting a ratatui port eventually actually
It's like "why did you go all in on buying scamcoin 3.0 as your investment strategy?" -- "I 5xed my money! End of story! It was fine!"
Or those choices just don't matter, it's fundamentally just "tabs vs spaces".
This isn't a minor nitpick, it's a pretty major UX issue.
Not saying it's like a massive business downside because I'm just one of a few users, maybe this affects their bottom line a little bit, but probably not by much.
Regardless, switching to pi has been a nice breath of fresh air. It just renders well and smoothly and handles terminal resizes well, which is especially important when used in a terminal window in PHPStorm.
Truthfully, the terminal in Jetbrains IDEs just isn't that great. It's gotten a lot better with their rewrites (yes, plural) of the last years, especially on Windows, but it can still be pretty dodgy.
I am 46 y/o and still excited to code and solve problems with code and I still have trouble to admit to myself sometimes that people make dozens of millions with PHP and the tech hardly matters when you can throw bodies at the problem (and when that throwing of bodies at the problem actually solves it, of course).
Or many billions. eg. Facebook was originally written in PHP.
As a 52 y/o who cares deeply about software development as an art I don't have trouble admitting that the tech hardly matters relative to the economics, that was obvious to me very early in my career.
I have just never really cared because there has always been a path where you can have a decent career being paid to work on and with good technology in spite of the fact that the technology wasn't important to the economics.
I do fear that the rapid adoption of LLMs will probably cause the path I took to narrow considerably. There will likely be less companies willing to pay people to work on elegant technology when they are all competing against endless code generation machines (and this is assuming software companies can exist at all in this future). But at my age that's more of a concern for younger people than me.
But IMO the endless churning of code will slow down, if it has not already. Quantity does not equal quality as we know, it even erodes it. There are already consulting companies that are "vibe-code fixers".
I am also not sure it's about "elegant" code. What we might call elegant has multiple measurable axii f.ex. well-isolated effectful code layer so you can easily static-code-check it, domains and enforcement of boundaries between them, bespoke linting to enforce business processes and standing code discipline, and many others.
Tabs vs spaces makes no tradeoffs. Tech choices do
They couldn't even fix flickering in their TUI and their new renderer is complete garbage.
They succeeded despite their tech.
However, that's not the same as the technology choices not mattering at all. As others have mentioned the product is buggy, and it appears Anthropic are struggling to continue to release new features within the existing architecture.
That's also not to say that Anthropic made bad choices - there's always going to be a tension between building quickly to capture the market now vs moving slower and building something more maintainable in the long-term.
This is just blatantly ignoring all of the glaring issues Claude Code has had over its lifespan.
Most people would choose to swim a lap through a pool of pee to get a billion dollars at the end. The pool of pee itself wasn't the once in a lifetime opportunity (for most).
Sorry, I'm bad at analogies.
It is rarely the case that technology choice is the make-or-break when it comes to whether a product is successful and achieves widespread adoption. Some choices are less ideal than others, but at the end of the day if you manage to make something that people want, the rest won't matter much.
A ringing endorsement!
Your counter argument would be valid for a 2000 or a 2020 business decision about some tech stack.
But the whole point of their product is that it supposedly nullifies such "business" concerns around the use of technology, by making it cheap and fast to build whatever you like automatically.
That they wont, or worse, couldn't, speaks against that.
There’s two arguments in competition:
1. LLMs make it cheap (in the time sense) and easy to build
2. Rewrites and/or writing something in a native app or program is harder and more time consuming
I think I am willing to take it as an axiom that a native version of CC would be superior from a user perspective. Performance, etc.
I just don’t see how one can say that building things reliably good is easy now when the company providing these tools can’t even do it well.
I don't know how you can't see it, to me it's blindingly obvious: risk aversion.
Let's say the risk for a problem is 1%; hell, let's put it at 0.1% even. For a company at this scale even that amount of risk is too much.
I trust Opus/Fable to drive prod database migrations and backfills. I don't trust it with our financial ledger. I trust it with part of the infra. I don't trust it with backups. Etc.
You and others are arguing against a premise that nobody defended, namely "Claude can rewrite everything, for free and with zero mistakes". A bit of a straw man, don't you think?
And that's not even touching the fact that writing a GUI app is difficult for LLMs due to difficulties in it getting feedback and a "feel" whether it delivered what was asked (though I know people are working on it).
> I think I am willing to take it as an axiom that a native version of CC would be superior from a user perspective. Performance, etc.
How does that follow, and from where? I never once noticed any visual jank/lag in the TUI; not in iTerm2, not in Kitty, not in Alacritty, not in WezTerm and not in Ghostty. And even if we exclude those two, me and many other devs are quite fine with a TUI and don't miss a GUI program for everyday coding.
Not saying that our preference is superior -- but it'd be strange to blatantly claim: "for dev purposes, GUI > TUI/CLI".
0 - https://finance.yahoo.com/news/anthropic-ceo-predicts-ai-mod...
Now whether they can do it -- that's a separate, and a much more interesting, discussion.
Why would it be harder and/or more time consuming rather than the opposite? It's like having full specs and a full test-suite to match against the result you want.
What do you mean? Claude cli is wildly successful. Why do they need to re-write?
Cheaper, but not free (if you don't buy into the marketing promises too much). The bigger the project, the bigger the cost, even with a discount.
At the same time, the early versions weren't very good and you can be sure that any rewrite will also need to be similarly iterated upon until it is also good enough and polished.
If you do that and don't spend enough effort on making it be something polished --> your competitors have a better product and you lose.
If you pause feature development to give enough effort to the initiative, you don't get to add new features quickly enough --> your competitors have a better product and you lose.
Maybe their priorities lay elsewhere, like how I've noticed that the desktop app version of Claude Code has gotten both faster (no 2-4 second lag when switching conversations), more stable and usable over time, to where I enjoy using the models because of it, not in spite of it (though not that they haven't had bumps along the way, like that one cache invalidation issue, or how people didn't like the auto accept timeout thing). I don't doubt that you can get pretty far with gradual patches and improvements, instead of only big rewrites.
Honestly it's really cool for me to see Kimi having their own CLI too, same with OpenCode, Pi, Hermes (well more of an agent than just a coding harness but you get the idea) - there's so many competing solutions out there, each good or bad in unique ways.
Just wish we'd see similarly many GUI solutions, for now OpenCode GUI seems like the one I've settled on (cross platform and supports most models), though it's not exactly ideal either (feels a bit barebones, especially in regards to sub-tasks and progress/plan tracking, even ZCode seems a bit better in that regard, it was actually surprisingly good after they pushed out some updates).
This is a spectrum, it's not just 0% vs 100%. Even Fable frakked up a few things really badly in my professional work (though to its credit after a very detailed 2h chat it self-corrected and fixed all the blunders).
I would also challenge "fast" -- Fable (and I assume Mythos) are wicked fast and efficient and even they can't compensate for f.ex. slow recompilations or test suite reruns or security scans, linters etc.
Reminder that the Bun's Zig-to-Rust rewrite took 11 days with dozens of agents working 24/7 and the author put the cost they'd pay (if they had to pay) at about $168k.
Billions of dollars of annual revenue go through Claude Code, and the people who work on it must be a lot of millions in headcount.
The time matters a fair bit, it's probably time someone can't spend breaking things in the name of new features, but if rewriting the stack had even a tiny impact on retention or driving higher usage, it would pay back that $168k.
You might get more maintainable code for $168k by hiring humans, but if its something bespoke it isn't going to get you far (or far quickly).
Vibing is straight out gambling right now, the thing is, I don't know who the house is any more.
In my work I have Fable/Opus add bespoke linters and tie them to verify/precommit tasks (including in CI) so nothing of the reworks and 15+ initiatives we have in flight ever regresses.
Some discipline will remove that worry of yours. It's not nice to immediately assume that the rewrite's orchestrator (Jarred) did not use discipline. He even outlined a few false starts.
For an entity like Anthropic that's not even the cost of a single developer for a year. It's closer to what they pay a chef on premises.
That the showstopper for a better Claude agent is that they'd need to pay $168k or event $1M or even $10M in costs, can't be used as an excuse.
The argument I responded to argued that cost might be the prohibiting factor. I'm saying that for them, the costs mentioned are negligible.
What I argued was that you seemed to have been arguing an extreme point i.e. LLMs can one-shot stuff in any PL and that it does not matter in what PL they'd rewrite Bun.
If I misunderstood -- my apologies.
What I said is, those AI companies push LLMs as the replacement of programmers, that can do everything code-wise, say their programmers "don't write code anymore" etc.
So how come they can't have the LLM write something better than "react for cli" for their flagship agent product?
Claude Code is likely not well-specified and it's a product of rapid iteration and now nobody really knows how the whole thing works. LLMs don't do well when requirements are muddy.
Does anybody really believe that it only took 11 days, one engineer and $168k in tokens?
Jarred used many agents and they worked 24/7. It tracks for me.
So yes, I do believe it. What makes you skeptical?
The major open harnesses are in the ~800k lines of code size ( https://simonwillison.net/2026/Jul/15/grok-build/ ) and CC is the oldest and quite-likely largest judging by features.
I can understand people being more sold on Rust compared to Zig -- there are legitimate differences on the memory safety front and I can see how people will prefer one or the other.
Plus I believe Jarred (Bun's main dev) went out of his way to explain why he chose Rust over Zig in his blog post.
Of course the marketing angle might indeed be the explanation; not arguing against that, I simply followed with interest and Jarred seemed very reasonable to me.
they rewrite themselves, Bun team is under Antrophic umbrella
This can be true, while "nullifies such "business" concerns" can be false. Both are not dependent on each other.
Automation doesn't "nullify business concerns", it just changes them.
In 2 years? A well-engineered TUI could be the difference between the Blackberry and the iPhone for all we know.
Eventually? I am sure they would agree. Currently it's you (and lots of people like you) who are doing the supposing, not Anthropic.
Any kind of scrolling back, copying text, using their menu system - basically anything that isn’t typing characters has had/still has unaddressed bugs.
OpenAI shipped a competitive model and I’m over in Codex now. I have yet to hit a bug.
If you’re holding the SOTA crown, people will put up with your buggy mess. As soon as that crown slips your pile of trash becomes a huge liability.
Oh, that's buggy too. I just tried Claude Code on win10 powershell, and the first typed character goes in the wrong spot and can't be backspaced.
It is by far the the least reliable program on my machine, and every time I have to interact with it I feel like walking in eggshells.
So weird that the same exact people telling you that programming careers are now obsolete are the same group who haven't been able to fix screen flickering bugs for like a year...
But the "coding is mostly solved" narrative kinda doesn't match right? If good, correct, high-performance software is like, free now? Wouldn't you want it to be slick as hell, really reliable, all that? Even a little breakage costs a lot of money at that scale and pricing, it would be better than a wash if you put the magic code thing on the case.
"Claude. Do all employee work. Make no mistake. Notify in slack when revenue is double."
I'd argue these tech companies got popular because they have good models, nothing else.
Even OpenAI took two years to fix basic chat scrolling.
The engineer is suggesting that it could be done cheaper and maybe with better outcome. Ironically, this is a classic business case.
If you have unlimited access to the magical development tool, then why would you not?
But they haven't.
Programs which need more stability and performance than exploration are better done in statically typed languages like Rust. Therefore Bun is a platform for JavaScript programming, but itself is written in Zig or Rust.
You can make anything work when you have enough money to buy and radically change the entire runtime you’re relying on.
One must suspect that if they did not have insane amounts of money to burn, they could have tried other approaches to fixing the problems. Maybe engineering, perhaps.
The point is that if you were starting a new TUI LLM harness today, you would basically use CC's architectural decisions as a guide for what not to do.
Much less fragmented compared to "write the app 3 times in 3 different languages".
Could it be that they aren't being entirely honest?
Unless you use an ancient teletype, you don't have to redraw everything. That would make any interactive applications way too slow/flickery.
You can move the cursor arbitrarily in the terminal and start overwriting characters from there. So you need to track state to know what is "dirty" and needs refreshing. Occasionally you issue a full redraw to catch missed artifacts left behind or when the terminal is resized (SIGWINCH).
This is especially useful when you're trying to evaluate behaviors while changing state surgically.
It's hell to maintain for not much gain (for a use case like this at least). As much as it's become a meme, JS and web tech in general has become extremely portable and stable.
I also don't think Anthropic bought bun to make their TUI better. They could have forked it, they bought bun because it incidentally was excellent for the way agents prefer to work and they wanted to capture that audience.
Their new flagship product (earning them billions of dollars in revenue per month) was dependent on a platform maintained by a tiny startup.
Buying Bun was a very rational way to reduce that risk, epically since it also got them some top tier engineering talent.
Thinking about that further, I wonder if that was part of the rationale for switching from Zig to Rust that they haven't talked about?
Zig is a much riskier bet for your multi-billion dollar cash cow than Rust is.
But saying that out loud would be rude - they took steps to NOT openly criticize Zig, even after Zig's founder did not show them the same courtesy.
If this does not get Anthropic 'The Presidential Medal of Strategic Restraint' I don't know what will.
"I'm using technology I know that will get us there" is not the same as treating every problem as a nail. It's making a practical choice that probably also had time constraints and other factors we don't know.
How does this support your argument? The tech didn't even exist in the 80s or 90s.
Saying "They wouldn't have made the same choices 40 years" is meaningless in this context. There's a lot of things they wouldn't have done back then that we do today. And this is a field that has been changing by the week recently, which makes even less appropriate.
This idiom now apparently means "I disagree with your tech choices" or in your case, "Someone 40 years ago would disagree with your tech choices".
I did not say they wouldn’t used react in the 90s (obviously). Anybody familiar with terminal uis (a technology which did exist) and perhaps any other ui system would not have chosen this design.
And no it’s not just me as evidenced by the many confused comments any time this is brought up.
Hammering in a screw is using technology that the hammer-wielder knows will get there. You’ve just rephrased the statement.
What you seem to be saying is that in some cases, the hammer may be the appropriate tool. No-one disputes that.
My point is that the idiom has now come to mean "I disagree with your tech choices".
This is close to the same "why Spotify is a chromium embed?" question. Because it works, and users are ok with it.
Is this because Codex is written in Rust and not JS? I dunno. I think it's more just "lighter" in general, or it certainly feels that way. It's probably possible to make something with a similar feel in JS, just perhaps not with the big honking mess they've created.
Claude Code seems to be an insanely complex program will all manner of cutesy features and telemetry features. The net result is approximately the same as Codex, but it’s pretty common in software engineering to find a simple thing and a complex thing that do more or less the same thing.
I suspect user facing/fast moving code (UX layer) will move to dynamic systems with fast iteration times. Infra layer will move towards safe systems level environments like Rust. I'm not sure where Java/C# lands in all of this - it's kind of the middle ground between these two worlds but the tradeoffs change drastically with LLMs - my gut feeling says that TS/Python is good enough for UX work and Rust is better for systems work so it gets less popular going forward.
[1] https://avi.press/posts/2026-07-10-after-7-years-in-producti...
Also GHC has an interpreter. It powers the REPL. No need to compile everything if you’re just experimenting. I do find that in the Haskell world the REPL is criminally underused compared to Python.
And the rest of your comment doesn’t make sense; this shop is rewriting their infra (not UI) from Haskell to Python.
To me it comes back to the premise I hear a lot from AI hype men (of which I do not subscribe but it’s worth clarifying that I find great value in using it for code) that the code doesn’t matter anymore. I just don’t know how to square that claim with the idea that we’re limited still by JS/TS being easier to use or approachable despite inferior TUI toolsets.
No, you are not crazy. They do crazy things with unlimited budget, and still their chat app flickers when using.
They should just port pi-tui from pi coding agent, since they have no clue.
It's like asking "why does everything run on Windows?" Because everything runs on Windows.
Proves LLMs don’t help with taste.
Yeah, good question. OpenAI decided to rewrite Codex in Rust about a year ago[0].
In fact, since rewrites are that easy, what do we need Bun for? Why doesn't everybody just port all of their Javascript code into Rust?
I've read and heard in videos that Javascript is a pretty good language for AI to write code in. Apparently this is because there is so much training data out there. Also Javascript avoids problems of multi threading and memory management that can mess up the AI in other "more performant" languages.
So maybe Javascript is not the worst choice for writing software fast with AI
The latest frontier models are competent at Rust and Swift and all manner of other less widely used languages.
The more important factor is how good they language's compiler is at kicking out actionable error messages, since one-shot code generation isn't as important once you have a coding agent loop.
Like Ada or Ocaml or Haskell or something.
Or even require like MC / DC testing or MISRA C verification.
Or even some of the languages that apply Hoare checks, like Ada SPARK or FRAMA-C or VALE or whatever.
I don't have the budget to test how that would work but it seems interesting.
Heck, a buddy and I once chatted about the likelihood of k8s running as the control plane in a prototype autonomous vehicle.
top/btop/htop on the mac are always fun to run and see what's up.
But then again, it's crazy how people are disingenuous. You're in a thread about Bun being rewritten from a non-memory-safe to a memory-safe language and everyone's shitting on it because it's a useless rewrite. How does this mesh with what you say about caring for stability?
Using AI you can actually start caring about things such as memory, footprint, stability and speed because it's crazy cheap to start optimizing for this, when before you couldn't afford to make the tradeoff.
It was unpleasantly surprised when I learned the hard way that LLMs are not much better at translating than writing from scratch.
The more you look into how they work, the more you see that it doesn't really give them a huge advantage, if the classes are big enough. You can tell them to break each function up and translate them one-by-one, but the errors compound, you can't test most of it until you have a lot done, and in the end, it really isn't much faster (and sometimes it seems to be a lot slower) then just telling them to start over from scratch.
The downside is... If you have a system that already works, you don't want to start from scratch and test everything all over again...
Because React is the only UI framework that takes the problem seriously. Everything else is stuck in the dark ages.
How HN gets this so badly backwards I'll never understand. Everyone on this site talks a big game about "there shouldn't be so many competing tech stacks, why can't everyone work together on one framework that does things right", and then as soon as that framework actually appears this site hates it more than anything.
Couldn't they have rewritten Claude Code in Rust directly? No more need for a JS runtime, better performance, etc... If their agents can do Zig to Rust, why not JS to Rust?
Claude Code isn't the only user of Bun (it's probably not even the only user of it internally at Anthropic), and this way they get to keep the Javascript runtime (and other tooling) that their coding models may one day, if not already, prefer to use when given the opportunity. For those kind of apps, you'll probably also eventually find that--what do you know!--Anthropic also has their own cloud offering (instead of Bun being the one to build one) that specializes in running and managing those applications for you. Even if all they get is a community of developers that choose to use Bun over other options, that gives them power and seats at tables they wouldn't have if they just rewrote Claude Code in Rust.
That'd be my best guess, anyhow.
This thread is filled with non-sequiturs.
I think a JS runtime is fine because the ecosystem of tools is very large and plugins are easy.
The answer is because you still need at least one human to develop and test any tool, integration, feature.
Unless AI overlords lied and we do need humans
For their work they extensively test using humans with deep expertise
But while marketing they say just vibe code bruh
And here they can't even rewrite CC in rust? Why not! Use 1000000 agents!! Write it in platform specific binary! SAAS is dead, that's what he claimed. So what's stopping him from writing the harness in Rust or C than ReactJS?
When someone says all Software engg could be automated within 1yr they don't mean 'oh use our agents and keep human in loop'
They mean 'AI superior than humans fire all humans and all SWEs will be jobless'
I am not going to reply to you going forwards as I can't speak with someone who lives in a different universe
They did not claim there is no human involved.
https://youtu.be/02YLwsCKUww?is=UsJrcLRgdSUd38Jt
Here's a transcript: https://gist.github.com/simonw/0050b4ce40439b97598d4ffdb273d...
Not all of the memory usage is the fault of Bun/JS. Some of it is likely memory leaks (holding on to data it shouldn't). However, some of the blame does go to using JS. In Rust you can tightly pack your data if you want to, but in JS you can't have nested structs for example. Every object referencing another object is two separate chunks of memory, each with a header, and a pointer from one object to the other. GCs also use more memory. The overhead adds up.
Building a CLI program in JS is a bad choice, and nobody should be defending this decision. Especially since Claude Code is very much able to write Rust code, and they've shown it can port code. Just port Claude Code to Rust directly.
If it run as good as before or even better, then that's kinda impressive.
I'm a developer so I really don't like it when AI might took my job, but if everyone on this planet could create a software for themselves exactly as how they wanted with just a few simple demands, that will change the world for the better.
Think of it as a democratization of technology. You don't want Microsoft stealing your data? Just ask a AI to write an OS for you. You don't want Google to listen to what you're saying? Just ask a AI to design a phone for you. If one day the AI ended up doing that, it will be the ultimate technology self-sufficient. In front of that, your job security is insignificant.
It is also why keeping the tech open source is that much important. Otherwise, it's still the same old shit again and again, and, you lose your job too.
Then you must know that the hardest part of software development is getting clear requirements...
> everyone on this planet could create a software for themselves exactly as how they wanted
Even with perfect AGI this will never happen because people will never be able to express clear requirements
The secret is individual humans are already largely irrelevant to our society. That's how we got here. The trend will only continue.
Incredibly naive take when these models are closed behind (for now subsidized) paywalls.
This "democratization" argument is so nauseating, seems like the Bun port to Rust as a marketing piece (the primary motivator) sold it well!
LLM is here to stay, keep denying the fact won't work. What does is to make the tech so abundant, it flushes the bad actors, such as for-profit AIs, completely out. It's either this, or get crushed by for-profit AIs, pick one.
The same tool, in the hands of someone that doesn't know how to handle it? I'll let you finish that
Basically, a good developer can handle the tool better by asking the proper questions, setting good guardrails and tweaking the output where needed
> Maybe I’m taking crazy pills, but I’m still stuck on “why the hell does a TUI need to run in terminal React by way of JavaScript”
> The fact that Anthropic felt the need to buy a runtime so they could make their TUI better speaks more to the quality of engineering than anything else IMO.
> If rewrites are so easy, why not rewrite CC in a native language? Would’ve been a hell of a lot cheaper.
It turns out, there's some similar sentiments in the last year ish:
https://news.ycombinator.com/item?id=47043325
https://news.ycombinator.com/item?id=47105720
https://news.ycombinator.com/item?id=48001113
https://news.ycombinator.com/item?id=46134570
https://news.ycombinator.com/item?id=46124596
I'm not a fan of any of the technologies, companies, or people related to this. I just couldn't shake the feeling that I've seen similar comments.
September and all.
I just thought it was interesting. I'm a fan of TUIs, both positive and negative about JavaScript (the modern language is quite nice, given you avoid the historical warts, in my opinion - the ecosystem is unfortunate), and I think the best description for how I feel about LLMs is that I'm bearish. Also Simon is a fantastically prolific and intelligent person though, even though LLMs are not my cup of tea.
https://news.ycombinator.com/item?id=48960912 is a recent comment by someone I also respect related to how LLMs recreate human-like content (Widgets, really), making actual human-created content feel fake.
Sometimes it sure does feel like ELIZA is the main news source these days.
Tell you what I sometimes think about though: The fact it has clickable links and complex formatting rules for markdown, the most interactive and highest-quality clickable interface I've ever seen in a terminal, and somehow manages to work. That actually blows my mind.
THEN, I'm reminded that it's written in React, and I think "Huh, guess that does make it a ton easier than using ncurses or something."
And done.
And your "it's just ANSI escape codes" is the weirdest take on any kind of software.
Not sure why you’re trying to bring up experience as a justification for using React to render a tui and all the downsides that come with that.
Take it from someone who, in those 24 years, has written UIs in jquery, react and ncurses (and yeah I've tried out Ink and it's great).
"all the downsides that come with that" yeah well, while you're stuck implementing your app in the Best Possible Library, millions of people are happy using claude code and it's crazy useful.
If they could migrate an entire runtime to Rust why can’t they migrate CC? It’s not as if React is the only way to do declarative UI.
Go has such frameworks, same with C# (which you can AOT compile) and so on. You're not stuck between javascript and C.
[0]: https://ratatui.rs/
If you're crashing every 5 minutes, either you have a cursed build, or you need to do a memtest.
This is a product by a company worth supposedly $1trillion and they can't even make one of their starship products not shit the bad within 5 minutes? These are the same engineers telling us for the last 3 years that coding is solved?
Like it's pathetic, and it's pathetic you're falling for it too. Have some standards, good grief.
I have a literal Star Trek conception on my literal Star Trek device. Impossible to even think about how inconceivable this stuff was years ago.
And you’re complaining because they use react.
The epitome of entitlement, first world problems and privilege all together.
As it is now, your post is meaningless to me.
So, from the start, people will be skeptical of a "we rewrote xyz in Rust" because it rarely works out. But there's a lot more to this, in fact it's a little bit loaded with "tech bro" ideas:
- Rewriting in Rust
- Rewriting using AI
- Making terminal apps in JS
- Zig (a lot of people's dear language) is involved somehow
This attracts every single skeptical developer. The only thing missing is maybe blockchain, if that was still cool. It just hits all the "what? why?" spots.
I was already writing a comment, when I realized that what I wanted to write had already been said by the top comment. I suppose that's why comments read so similarly.
Is all you need to know to consider how much of an abomination it is
2026-05-14 23427db 13907
2026-05-14 19d8ade 13861
2026-05-15 4d443e5 13840
2026-05-17 172afa5 13803
2026-05-17 80a06a8 13849
2026-05-18 fba43af 14026
2026-05-19 303cd28 14052
2026-05-20 21db682 14243
2026-05-22 a06a00a 14239
2026-05-23 49c97de 14090
2026-05-28 472a06a 14076
2026-06-01 a0d1472 14071
2026-06-04 8553428 14032
2026-06-09 717542f 14053
2026-06-10 1c90e5a 14043
2026-06-11 6e91d24 14031
2026-06-16 bd8edc7 14086
2026-06-17 6ef5977 14104
2026-06-20 315ed50 14106
2026-06-22 c6be834 14120
2026-06-23 ea7e44f 14108
2026-06-23 03042ab 14128
2026-06-29 86d32c8 14046
2026-07-01 6640fcf 14077
2026-07-04 51074e3 14099
2026-07-06 9d0e93d 14186
2026-07-08 ab6eb2d 13953
2026-07-09 86caf6e 13936
2026-07-10 91675d0 13930
2026-07-13 73b6c14 13951
2026-07-16 4bbe075 13978
2026-07-16 57e30a5 13995
ChatGPT written script for counting here: https://gist.github.com/simonw/b1015bcadcedd1a781cedb7af9cbb...
Now it's smaller, faster and has fewer bugs. Also its every potential memory issue is neatly annotated by an unsafe block so you can go and refactor them out one by one with confidence.
All this seems like a pretty huge improvement to me. Why is this an abomination in your eyes?
True.
> has fewer bugs
Nope, this is demonstrably false because Rust has its own invariants around its types and the codebase is violating a lot of them.
> every potential memory issue is neatly annotated by an unsafe block
"Potential memory issue" can originate in unsafe blocks and safe code that are able to alternate the input condition of these unsafe blocks. Guess what? That still counts towards 100% in this code base, hence the abomination remark.
> refactor them out one by one
Not as easy as it sounds. They are like threads in a yarn ball, if there are one or two ends visible it's easier to sort them out. The actual situation is more like we have tens of thousands of ends (all the raw pointer code that comes with every shared object), to fix them it's basically a requirement to unwind the whole thing (rewrite all the callees and reorder the data flow as needed, redesigning all the APIs during that).
It's too early to declare it as anything remotely close to a win, optimistically saying.
That being said I've found a case that doesn't do much harm even if it's fixed, and is technically not my contribution. You may take a look at that:
The second that you start to introduce nonsensical axioms (or bugs in unsafe sections) all your proven correct theorems will be worthless.
It’ll most likely still do 99% of the things people need it to do, there’ll be a an issue created for that broken 1%, and eventually it’ll be iterated upon and fixed.
Unless you’re working on a security boundary of course, there you should treat it like maths.
bad axiom -> worthless theorem.
bad unsafe section -> verified safe section is actually not safe.
The proof that there are no (for rust's definition of "no") memory bugs becomes worthless, not the entire software itself.
Maybe? I'm not sure.
> The proof that there are no (for rust's definition of "no") memory bugs becomes worthless
I still don't think so, the proof that there are no memory bugs is definitely not worthless, since while the whole proof itself is binary (it is either fully correct or not), its effects act more like coverage, and the more of it that happens to be partially correct, the better.
I think cve-rs is the perfect example of this. Its very existence proves that the borrow checker is already unsound, and with that, the very foundations of Rust's safety guarantees are currently flawed.
TypeScript is another good example, its type system has been repeatedly proven to be unsound, so it's also flawed.
But so what if it's flawed? Many smaller properties of these proofs still happen to hold, and they might prevent 90%+ of CVEs even as flawed as they are.
I'm not sure I'm following what you're saying here. In the rewrite blog post it has been stated that the Rust port landed in a place where it passes 100% of the existing test suite and does so without some of the known memory leaks of the Zig version that they were aware of and didn't know how to address.
Sure, there's no guarantee that the port didn't introduce new bugs and most likely it did, but the 1.4 version has not been released yet, so I don't think a direct comparison of a long-time "stable" version and the just-landed full rewrite is meaningfully helpful.
> "Potential memory issue" can originate in unsafe blocks and safe code that are able to alternate the input condition of these unsafe blocks. Guess what? That still counts towards 100% in this code base, hence the abomination remark.
But that just means that once you get rid of the unsafe block, you're pretty much guaranteed by the Rust compiler to be memory-safe. This means that all your potential memory landmines are annotated in the code and once you disarm them all, you're memory-safe.
I don't understand how what you're saying contradicts any of my statements.
> Not as easy as it sounds. They are like threads in a yarn ball, if there are one or two ends visible it's easier to sort them out.
Still easier than the Zig version, where all the same things you've mentioned apply, but also there's no visible indication of progress and issues are unmarked and could be pretty much anywhere.
> It's too early to declare it as anything remotely close to a win, optimistically saying.
Based on the rewrite blog post it's pretty clear that the Rust version is outperforming the Zig version by pretty much every measurable metric at this point.
I think the backlash is mostly focused on the ideological opposition to fully AI-authored 1MLOC+ PR approach they took.
They can't mention what they don't know. And that's the exact reason I'm against this: doing a port without being able to tell what is correct is asking for a disaster, obviously.
I've checked their posts and blogs for a few times, and I'm quite sure they don't know any of these to write about them.
> the 1.4 version has not been released yet
Let's hope the repeated delays (it was scheduled for last week) are not for quality issues, otherwise I'll only be half correct. It appears that all the MIRI errors are not stopping them though. (They have a script to run it but most probably it just fails and they don't care. Personally I'm not executing any of bun's code to avoid summoning nasal demons, so I actually don't know for sure)
> get rid of the unsafe block
Would be possible if they are not disregarding everything else going for numbers. E.g. they are actively increasing the unsafe surface by rewriting some code in SIMD C++ (https://github.com/oven-sh/bun/pull/33032). As if no Rust solutions already exist.
> Still easier than the Zig version
Well, this is true but only realizable if they know how to deal with the new problems emerging during the process. My current conclusion is that they don't.
> ideological
For me it's methodological as explained above.
who claimed that? Are you suggesting that the rewrite did not introduce any new bugs? The correct answer is, by the way, that no one knows since it's millions of lines of code no one has properly read.
> smaller, faster
I thought this has already been debunked. You could just write better zig and make it smaller, faster (and have fewer bugs!)
And like, this is just the beginning of the port. They did a mechanical port basically line by line, next step is to make it idiomatic rust.
I thought by now people would’ve learned to stop betting against this rewrite.
As with all transpile ports, the true test will be how well it can be extended and maintained over time. Historically working with the output of a transpile is not pleasant and requires heavy rework to get it to the point where you can be comfortable enough to extend it.
The existing community is now either going to need to learn Rust or new Rust developers are going to have to join the project and they may not invested enough to refactor the transpiled output when they could work on something else like Boa.
This is I think a preview of the future of software development. LLMs are having the same effect as compilers, frameworks, etc - they are making new kinds of software development feasible which were not previously so.
I'm not quite as optimistic. "Unpleasantness" might sound like subjective disgust, but it often relates to objective criteria like complexity, that increases the surface area open to bugs in a real way. Similarly, if reworking has a non-0 chance of bugs, its gonna have some form of risk, and a major one will of course be higher.
LLMs may be capable of a lot, and they miggt change what projects people can take on, but they definitely don't write bug free code.
The bun rewrite/transpile did use some neat ideas like lots of testing and per-file translations, so hopefully they have more in their toolbox than "LLMs hopefully won't be affected by complexity or write bugs"
You can tell what will happen when they release it before sorting out all the new bugs introduced by the not-exactly-line-by-line port.
Given the success of the port so far and the fact that CC is now running on rust Bun, that seems highly unlikely to me.
I'd also like to inform you that
- the current success metrics solely consist of their advertisement, my eyes looking at the code strongly suggest otherwise
- the Bun team lacks knowledge to actually make it more idiomatic: none of the Bun team has written any Rust and relying on model knowledge is already proven insufficient.
The release decision itself is presumably driven by that tbh. It's only LGTM from there when nobody knows how to review unsafe Rust.
Do you know how silly this sounds? Good engineers can work in any language. At the top companies (of which Anthropic is certainly one) they don’t have “C++ engineers” and “rust engineers” they just have engineers.
They were up and running and super-productive with Django within a week.
Given that they are investing their time into actually learning the language before diving into this mess. No, unsafe Rust isn't supposed to be something you can handle without prior knowledge.
Jarred was even asking in public how to use traits without runtime overhead during the rewrite. And the blog post phrased it like a normal problem to encounter while doing a million lines port... That's two months after the initial rewrite so you can guess if they learned.
It would have made far more sense, for reliability, efficiency, cost, etc., every metric really, to use or write a source to source translator that preserves as much structure from the original code as possible. Typically if you do a rewrite there are lessons learned from the existing code base that you want to take into account when doing the rewrite; using a bunch of agents to do the porting file-by-file buys you none of that. In either case the code will be an unidiomatic translation, just with LLMs you get an added source of indeterminism and a huge bill at the end of the month.
Anthropic seems to be all in on 100% AI code.
When I started, I discovered to my shock that it was by far the worst TUI I had ever touched. Rendering glitches, keyboard input screwups, and just all-around jankiness.
Despite that awful baseline, I have to agree that it's gotten noticably worse in the past week or so.
It's started mangling my terminal sessions somehow, so that not all characters I input are visible in the UI.
Backgrounding it, doing a 'reset', then re-foregrounding seems to fix it.
Of course when I asked it to diagnose the problem, it assured me that Claude has no such big and it must be something else I'm running.
All I'm running is tmux and Claude, though, and the behavior surfaced while I was running Claude.
So, yeah - if they just recently updated Claude to use the Rust rewrite, this increased crappiness might actually be due to that. It does seem to at least roughly correlate.
Yes if we had a Zig->Rust transpiler maybe it would produce similar output. But I can’t find one. We’d have to use an agent to build one first :)
cpython has ~7,000 open issues, but ~70,000 closed ones, a 1:10 ratio. Bun's ratio is above 1:2.5, while for Claude Code it's slightly over 1:6.
Much more importantly, though, Bun's oldest issue is from Sep 2021, while Claude Code's is Feb 2025. cpython's oldest issue dates from June 2000 (presumably migrated from an older tracker). Claude Code has very nearly the same total number of issues (open + closed) as cpython at around 77,000, but Claude Code has done it in a year an a half whereas cpython has taken 26 years to get there.
It's a good comparison to pick, as it can't be explained by popularity. For all the real-world use Claude Code sees, it's nowhere near what cpython sees, even over the year and a half in question. Bun, meanwhile, is hitting these numbers while not even being the mainstream choice for what it does (node still holds that crown afaik).
The rewrite may work out on a technical level or it may not. Let's wait and see. The Bun project did not swear a blood oath to use Zig forever, and it's ultimately their choice to switch to a different programming language.
Beyond that, people seem to be incredibly emotionally involved with this, for reasons that entirely escape me.
A dirty secret of AI data centers are that they’re only getting 40-60% efficiency out of their GPU clusters and because the moneygun go brrrrr they just buy more.
You wonder why they’re so afraid of the Chinese competition… they can’t afford to be as wasteful
Faster startup
Typescript support out of the box
Better stdlib than node
Stdlib includes yaml, sqlite etc so you need to pull in fewer deps, so you can avoid the left-pad/is-even node_modules explosion problem to a greater extent
AI is pretty good at rewrites and ports. Quite good infact.
Project-wise, nothing has changed.
Bun was always great because of the fantastic dx - it was just really easy to use , with stuff like out of the box typescript (unreal that it took so long for node, and it's behind a super long flag, wtf...). And it didnt have the weirdness of deno, it maintained backwards compatability with node api, and it just worked.
But it was never stable. You'd have to be fool to believe that a single project could stably do everything bun covers. It's always been an insane project. It was built on top of zig, a langauge that hasn't reached 1.0, and is constantly changing, and throw in how he was rewriting his own custom zig stuff. Like c'mon, let's apply some common sense here.
For me, little has changed. I am still going to use bun as a nice dev tool, and use node for production.
I am curious how this will work going forward from FOSS perspective. Will humans be allowed to modify the generated code? Only the .md prompts for the agents? Or what?
Can you select text with shift + arrow keys now in the command line client? :)
And the crowd is cheering.
Claude Code is closed source, doesn't let me extend it beyond hooks, and these I write in bash anyways.
That all said, why should I bother about this change? Feels like a nothing burger to me, as an end user. This should matter more for those that are internal to Claude Code and Bun developing it.
you are welcome
When this happens, a link shows up to report the issue. It’s not clickable (likely due to the segfault), and perhaps more important: it’s encoded, so you can’t see what you would be sending in your report.
Hope it gets better.
If nothing prints "Segmentation fault", this just sounds like a hang
From what I could tell it looked like it was in JS code that had been JIT compiled. I haven't attempted to troubleshoot further beyond setting all my future Claude Code instances to pipe their stderr to disk so that I don't lose the stack trace next time.
And I can believe you because I’ve seen JavaScript-heavy Safari tabs crash.
It becomes completely unresponsive to any input except scrolling. I can't select options or even cancel out.
(I am in no way implying this is related to Bun)
I actually used to have issues with run away memory leaks causing my computer to hang which often required a reboot to recover from. I haven't experienced that in a couple of weeks now. Too early to say just yet but I think its definitely possible that the rust port fixed a bunch of memory leaks that have tangibly improved the experience atleast for me.
In Chrome or Firefox, a segfault is usually automatic CVE and in most cases a bounty of thousands of dollars…
If projects like Bun can be substantially rewritten and shipped to millions of users, it suggests we're entering a very different phase of software development.
Today's AI-generated rewrites may not produce code that humans would consider high quality or maintainable. But I'm beginning to wonder whether that will even matter in a few months. If AI is the primary consumer, maintainer, and refactorer of code, human readability becomes far less important than correctness, performance, and the ability to iterate.
This feels like a shift where software may no longer exist as a long-lived artifact in its current form. Instead of writing and maintaining applications for years, we may generate, adapt, and discard them continuously for each use case.
We'll develop faster and better tech. We'll find resources to feed that. Use that to build better. Access better resources. We'll mine asteroids. We'll harness much more from the sun. The factory must grow.
Earth is in a cooler period and it will get very warm. These crappy climate change activists will not get anything done but to hold us back. We need more power, more tech, and more abundance of things. We will need to engineer our way out of real climate change.
At that point, even Bun itself doesn't matter. All intermediate tools don't matter if LLMs can reliably write something large.
The problem is that LLM is not quite there yet. The rewrite was only possible because they mostly stick to 1-to-1 translation resulting in non-idiomatic Rust code. So, what from there? I don't think they can really build up a sane codebase from that state. They only shot themselves with a bigger gun.
That’s patently false, just read Jarred’s own blog post describing how that was the first stage only, they went through many more to get the amount of non idiomatic, unsafe Rust code to an acceptable level.
> If projects like Bun can be substantially rewritten and shipped to millions of users, it suggests we're entering a very different phase of software development.
exactly. this wasn't a technical project. this was a marketing stunt that worked on you.
Keep in mind that you are not Anthropic, and you'll likely NOT consider rewrite at all if you don't have unlimited budget
Software refactoring and tech debt is absolutely where LLMs will shine as is evidenced by the Bun rewrite.
(I was actually half way through writing about something related, when I thought it would be interesting to see if I could prove that I had the Rust version of Bun on my laptop already.)
Please don't post insinuations about astroturfing, shilling, brigading, foreign agents, and the like. It degrades discussion and is usually mistaken. If you're worried about abuse, email hn@ycombinator.com and we'll look at the data.
If anyone else did that on HN, they would be accused of slop with their domain blacklisted.
Knowing Simon - he’s very much both :)
(Unless photos of actual pelicans count as "AI" content now.)
I just shipped this filter UI so you can see my other non-AI-tagged content in one place: https://simonwillison.net/search/?exclude.tag=ai
The "ai" tagged posts (2,100+) significantly outnumber "sighting" posts by orders of magnitude. Even searching HN, this post is the very first mention of it.
Perhaps that is why everyone here (except you) overlooks that or why we don't see much more of those sightings posts here either.
I would be very surprised if one of my sightings posts made it to Hacker News. Maybe if I manage to snap a photo of an actual pelican perched on a bicycle some time.
You can see my list of previous weekly blog sponsors here: https://simonwillison.net/dashboard/sponsor-history/
None of them get any influence on my content. That would hurt my credibility, and my credibility is the reason my site is worth sponsoring in the first place.
Honestly, if OpenAI or Anthropic offered to sponsor I'd probably turn them down. The optics of taking sponsorship from companies that I frequently write about are not great.
But look at your own disclosure page [0]: OpenAI paid you for your time at a GPT-5 preview, and you regularly get early access, embargoed previews, and free API credits from OpenAI, Anthropic, Gemini, and Mistral. So "I've made no money from Anthropic" is narrower than it sounds, access and preferential treatment are compensation too, even if cash didn't directly change hands.
Same goes for the sponsor list. Microsoft is OpenAI's investor and infrastructure (or was anyways, I've got no clue what the status of that is these days), Amazon and Google are both major Anthropic investors, and most of that list runs on Azure, AWS, or GCP, or sells tools built on these labs' models. Taking a Microsoft sponsorship isn't really separate from an OpenAI one, it's just a single hop removed. The entire software ecosystem and all the players in it are entangled in this bubble.
I'm not accusing you of consciously hyping any of them. Your posts are high quality even when I disagree with the contents of them. You often list downsides too, prompt injection, security, slop issues but the throughline still repeats: this is inevitable, get on board. Product-market fit found, November 2025 as the inflection point. That kinda framing shows up again and again throughout all your posts.
It's refreshing that you're upfront about the previews and credits and all that jazz, but surely you understand why some people stay skeptical even with that disclosed as it confirms the access exists.
Whether you realize it or not, I think they are indeed paying you in some sense, which is part of why I'm skeptical when your posts consistently land at the top of HN. Not because the writing isn't good, but because I think the same incentive shaping the coverage probably shapes some of that reach too.
[0] https://simonwillison.net/about/#disclosures:~:text=2024%2E-...
Nobody is unbiased, no matter how hard they might try. I like to get my information from sources where I understand their motivations, so my approach is to try to make my motivations as clear as possible.
(That's why I spend an unreasonable amount of energy arguing with people on Hacker News who call me a shill!)
I have a long running side project I migrated to bun, and I'm starting to regret it. I don't want to build on top of this much churn.
While everyone here in this forum kept arguing (and fighting and yelling at each other) whether tis moral/right/secure/cheap for ppl to rewrite and ship a major software package with LLMs/agents, one of the main drivers behind AI actually did it, without consulting your opinion, and most same everyone actually slurped this decision without paying a a notice.
Boring is good, but this boring is super massive major thing that happened, and precisely because security is still intact.
Just that?
I was expecting more from that rewrite. Maybe Rust is not so worth it after all.
You wouldn't notice either of those if you were a user, unless you happened to hit one of those bugs.
The benefit of a rust rewrite is memory safety improvements, but currently they've just rewritten zig to unsafe rust so they don't have that either yet
The latest Claude installer (2.1.215) possibly does not have it anymore https://github.com/anthropics/claude-code/issues/4953#issuec..., but more tries on various machines are needed to confirm that.
My anxiety before merging a 2K lines PR is greatly reduced after 3 frontier models (Fable, 5.6 and kimi K3) finds no issues in it.
Just 6 months ago (Opus 4.6) this was not true, a big PR would have countless number of issues.
Aside from the human drama, the message to all of us is - these things are ready for whatever your imagination can throw at them.
I think that's exactly what Anthropic wanted to communicate.
…if you have the compute and capital to run that whole agentic system