I'm still waiting for someone to bring forth some reproducible stats and state something along the lines of "I did a comparison of Bun 1.3 vs 1.4, and found these user-affecting issues in the latter which are not in the former".
For the internet at large no, but the comment doesn't make sense. It's not as if volunteers don't post all kinds of checks, and reports, and benchmarks, and deep dives anyway.
I would assume Sapolsky understands how terms are defined by their context, and that someone saying “free will” in a forum discussion about software doesn’t literally imply the strictest, academic definition of the term.
I'm curious as to what other definitions of the term exist, other than the academic one? I had never though of using the term in any other context. But maybe that's because I've been concerning myself with this specific topic for too long.
This reads as another weirdly personal attack against Jarred for slighting a technology stack the author prefers, citing the even weirder personal attack against Jarred written by Andrew Kelley as the sole source to justify their "biggest worry", code quality.
People _really_ don't want there to be success stories for llms and they will go through all lengths to discredit and disparage folks successfully exploring the space.
The most obvious benefit Rust can give them is being safe from memory leaks. This is hardly controversial: it has been a Zig design choice that memory leaks are the developer/user's responsibility. They have already demonstrated to have fixed several of those [0].
Marketing-wise, it has been an obvious success. Dunno if that matters.
Yes? I’ve made that mistake, it’s not particularly hard to do. Stash an Arc in something that lives too long, or forget to clear its inner data structure when you should, and you have a leak.
That’s not like … a design flaw or something, just a common class of error Rust can’t save you from.
Seems pretty simple to me: if the software works, the port has been successful.
If this is unsatisfactory, another way to think about the question is to ask "what would make you conclude it has failed?" That requires a sharp choice.
Both of these are true. I recall Anthropic's lofty claims about their C compiler. It just did not work, and they claimed all of this and that. It was clear they were driven to believe that agents alone could write a C compiler, and refused to believe they couldn't. Not that I think that AI couldn't write a C compiler. It definitely can... with a human at the helm and sufficient time.
But, for people who don't like AI, they also refuse to believe that AI could be good at any particular workflow. Which is just obviously not true, there are workflows that AI is really good at. Language re-writes are one of them. They're almost mechanical in nature and trivial to check via tests. So, the perfect use-case for an LLM.
I'm not saying the Bun re-write is any good, but the underlying concept is sound. If ever there was a tool to do a language re-write, it would be a high quality LLM.
you really went digging with that 3 months old hn comment. What's wrong with that comment btw? Are you saying he was lying and they did in fact commit to the rewrite at that point?
I didn't dig. It was prominent at the time this occurred, because like I said, it wasn't communicated at all until they were already in the middle of it.
And yes, I think he was not genuine there and that they had indeed fully committed to the rewrite prior to that point and comment.
If you're not aware of any of this, how are you arguing it was communicated clearly?
Indeed. This article didn't age well... I'm sure that the creator of the original piece will do a thoughtful self-analysis if they might have gotten something wrong, and make sure they didn't have any sort of unspoken bias in their thinking and writing!
I'm really skeptical about this rewrite
I chose to adopt bun about a year ago, and I've honestly lost count of the number of issues i've submitted and the amount of time i've wasted dealing with them. If I had to make the choice again, I would probably stick with node and pnpm for serious projects
We're constantly being sold on promises of -XX% ram usage, but in practice, it's often builds randomly failing for no apparent reason, memory leaks bringing applications down, and instability that ends up costing far more in engineering time than the resources saved
That's why i'm worried that bun 1.4 could end up being a step backwards, once again putting into question the small amount of stability we've managed to achieve so far
But I still strongly believe the project makes a lot of sense, and I genuinely think these issues can be solved by 2028, if the project keeps its focus on stability instead of another SDK for SQL, web sockets, image compression, or whatever comes next
> But I still strongly believe the project makes a lot of sense, and I genuinely think these issues can be solved by 2028, if the project keeps its focus on stability instead of another SDK for SQL, web sockets, image compression, or whatever comes next
If you check the maintainer X, he's asking Claude for what features to implement next. Not only stabilizing isn't the priority, even the features implemented are random.
What does innocent even mean in this context? I don't care if it's innocent. It is. Naivy too. But millions of people use the system and were expecting stability, not it to be a random AI experiment.
> If you check the maintainer X, he's asking Claude for what features to implement next. Not only stabilizing isn't the priority, even the features implemented are random.
To me that sort of stunt sounds like blatant attempts to distract the audience away from relevant topics.
> The last three months have not looked good for Bun.
Apart from pro-AI or anti-AI posturing how have the last few months not looked good for Bun, exactly? I use it daily and I've seen basically zero regressions. I get it, you don't like AI or you like Zig over Rust, or whatever. I just haven't seen any serious argument that Bun has somehow become worse software.
> The project has over 5k open pull requests, which is the largest number of pull requests I’ve seen.
Terrible argument, and not really an argument at all.
> The biggest worry is, of course, the code itself.
I agree, so look at the code and point out what's wrong with it.
Insofar as Andrew Kelley is concerned, it's obvious he has an axe to grind and is salty about Bun embarassing Zig (which he freely admits). Not sure why you'd invoke an unreliable narrator as some sort of final nail in the coffin.
All that sounds great. The promised of shipping for months that never came is a concern, comparatively speaking to history of the project and what is happening. But if we have user feedback suggest otherwise may be it is good.
For most, they are still waiting for official release so they could make their own judgment, the problem is that it is still not released. I guess we should just wait a little longer.
>> The project has over 5k open pull requests, which is the largest number of pull requests I’ve seen.
> Terrible argument, and not really an argument at all.
Technically, maybe.
But when you consider for a moment that in this pile of slop some PRs were crafted by actual people who trusted Bun and Jarred... and whose trust has been flushed down the toilet by the Rust rewrite. Well, maybe that's the price of progress. Or maybe that's a grave mistake.
The bun rewrite is looked to as proof that this sort of full-throttle
vibecoding is the future of software development, at least for large porting projects like this one, and for many people on this site that is a high stakes question.
To prove that, a proper release is needed, and the Bun community needs to adopt it. There's a huge partisan eagerness to declare the Claude Code release or anecdotal experiences on the canary as victory, but the viability of 1.4 hasn't been demonstrated until it has been released and adopted by the bulk of the community.
There isn't actually any hurry on that. Nobody needs Bun 1.4 tomorrow. It's Jarred that keeps saying it'll be released tomorrow, while the code churns at an astonishing rate (see below). People with a skeptical outlook will inevitably suspect that he doesn't have confidence in the release and is struggling to acquire it.
Github insights for the last week:
> Excluding merges, 57 authors have pushed 349 commits to main and 6892 commits to all branches.
> On main, 2026 files have changed and there have been 189,502 additions and 133,836 deletions
Sure but that’s how you beta test a runtime, by using it in a codebase you control. Claude is publicly released and used widely, that’s still only a single codebase
So they are being more cautious and spending more time on polish after a never been done before historical massive rewrite...? Stop the press! They are not shipping as fast they used to! Omg!
> So they are being more cautious and spending more time on polish after a never been done before historical massive rewrite...?
That's not credible at all. I mean, after a major release there are tons of low-priority low-hanging fruit issues that can be quickly sorted out. That's the nature of a release process.
However, Bun stopped releasing anything. Even after lauding it's AI push,which can easily chomp through small tickets.
It's been a couple of months since the major push. I repeat: the release cadence of a mature, stable codebase as the Zig one was at 2-3 weeks. A messy rust rewrite packed with unsafe code, which is prime ground for small bugfixes and correctness fixes, led the project to grind to a halt.
Not surprising, but Jarred has been teasing "tomorrow I swear" for months and Bun keeps missing his own deadlines.
There's clearly something wrong in their confidence about Rust Bun: If it isn't ready they should have said "whenever it's done", and if it is they'd have met some deadline by now. I suspect the speed was supposed to be part of the stunt, yet they seem to have encountered the 90:90 rule.
The rewrite was a lot faster. Actually shaking out the bugs from the rewrite (or just getting enough confidence in the new version from use by early adopters) is probably not going to be drastically faster.
That was 11,000 lines of dead code removed from pre-rust Bun. In a million lines of vibed Rust Bun I'd be astonished if there was a mere 11,000 unused LoC. Noticing unused code isn't a strength of LLMs in my experience.
I mean, how do you check that your frontend code (possibly not even managed by your team) is calling and needs all of your backend endpoints? I'm sure it can be done, and it probably should, but saying it should always be 0% is not very pragmatic.
Dead code likely means chunks of code not used by other chunks of code, which should be easy to determine statically in most modern languages. Whether an exposed endpoint is being hit (used) or not as an API is more of a test-code-coverage thing, different from the concept of "dead code".
True, but Bun actually has a similar problem. It's creating bindings to a JS runtime. If the JS glue code then no longer uses it, the code is still marked as alive because it's registered with the runtime.
In a previous life I managed to deliver a system with (basic-block) coverage measurement compiled in (gcov) but instead of the mess of files gcov generates, the whole coverage structure was streamed to a remote server (it compressed very well). Once that was in place it was an amazing telemetry tool.
Later on used Intel Processor Trace in a similar fashion for even finer (mc-dc) coverage.
Coverage tools are very useful, if a bit hard to use...
First of all, I'd need to see a diff on that. It's insane to make that kind of claim without backing it up. Second of all, dead code (as in truly dead code: like an if statement that is guaranteed to never be true), is optimized away by the compiler. So even if we had all this dead code, it is not an issue that would ever impact the final product. Third of all, the Bun runtime consists of over a million lines of code, so that figure (again, even if true) is insignificant.
> It's insane to make that kind of claim without backing it up.
A bizarre accusation. It's a fork of Bun that says they removed the dead code. All you have to do to see the diffs is look at its history. Most of the commits are code removal.
It's a fork of Zig bun, though, not Rust, so hardly relevant to the AI argument.
This reads like one of thouse articles in mainstream media where the author has a obvious agenda, and they think they are really clever when they write an attack piece that they think looks like objective , when it really really transparently isent.
For me the real test will be maintainability going forwards. Not a JS user.
I suppose it’s one thing to rewrite code and make tests pass. But a lot of my own CPU cycles when coding go to making the code understandable, modular, editable etc. IME Claude is not great on these.
Maybe it doesn’t matter? Maybe the spaghetti makes sense to Claude, and when you say “hey add this feature”, no problem?
But maybe it will turn into a worse ball of spaghetti, with no nicely curated tests, hacks on hacks on hacks and never ending loops of whackamole of “just change this one line and rerun tests”.
They were pleased to be faster than Go in the "Hello World" benchmark, but unfortunately, they lag behind Node.js and Deno in complex real-world workloads.
Bun has been nothing but amazing wins for me, and I really like what it enabled me to do. Two examples:
- I have a couple of projects that handle images. I was including Sharp, but since these are side projects in a small VM, often a redeploy that recompiled Sharp just crashed the full VM (out of memory). Bun includes Bun.Image[1] natively, which is built around Sharp's API, so swapping Sharp out was very easy, and now deploys are a breeze (and swapping it in would be just as easy).
- Bun's JSX support is a godsend. I've replaced almost all my side projects that were backend rendered from Pug, Handlebars, and other various templates I used to have to just pure JSX. Heck, this made it trivial to make a side projects where the Favicon was a dynamic SVG [2]
That's a fair criticism... they move slower but mostly because they were the first. They would have to build a new Server and Request/Response primitives from scratch, or deprecate what they already have, or somehow shoehorn the existing types into the new ones.
So I don't think they're opposed to standardising, but it's trickier when you have some past cruft built up. Lets not forget that the fetch() standard was built for browsers, and is a bit of a stretch to make it work on servers at all. You have to diverge from the fetch spec to even make it make sense in a server environment (with things like Cookies handling off the top of my head).
Same thing can be said about Streams in Node.js as well. They had a working streams implementation before the Web Streams API was in browsers (iirc). So implementing the standard now means removing what's already there, or breaking a lot of existing code.
Which leads to confusing things like the fact that there are two streams APIs in Node now:
Fwiw I think it's good they're not removing the legacy code, but they should probably do a big cleanup major release and lean into the modern standards and remove the old stuff. But there's obviously push back in doing so... and probably stuff you can do in the old modules that you can't do in the new and vice versa.
Yes, I've dealt with streams a lot and I've seen the pain. I'd say it's the 2nd biggest "compat breaking" of Node/JS after the require vs esm, but while the require vs esm is mostly gone (we all use esm nowadays, except for aham the default npm) the stream types I expect will last for a long while since that's not so "high" in the productivity list for most. Also, that's actually one of the main pains I have with Node NOT supporting Fetchable, the fact that the Body is not a Readable.
When we used to code by hand, Bun used to have an amazing list of articles on how to convert from almost any kind of streamable to almost any other type. I used it a lot, and that was when I started liking Bun so much.
It consolidates a lot of really annoying nits and quality of life aspects like bundling, module resolution, testing, env handling, etc. Node is just the runtime and still would require libs like esbuild, jest/vitest, dotenv, and so on. It's genuinely nice to have those out of the box, and generally in Bun the performance is quite significantly better in places that matter.
You need a "separate" tool called npm to install packages. It's dead slow so they added corepack. Another extra tool. Then you use it to install yarn or pnpm. Another extra tool. It might still be slower than bun...
In the past you need to transpile typescript with other tools. It can only run js out of the box. The later fix was a wasm build, which means slower than it could be. Another 1/2 job.
Many more examples in many areas but anything Node tries to do itself feels 1/2 baked.
Bun and Deno are faster than Node because it is a 1/2 job at its core. It has v8. It has the tools. It could be done but no. It did nothing for years until Deno and Bun came on the scene to nudge it.
From the perspective of someone trying to package nodejs, building it is a complete dumpster fire. GYP (generate your projects) is a now defunct and poorly thought out build tool that node heavily relies on with it's native C/C++ modules. Cross compiling is completely broken and I am honestly surprised it builds reliability for the native target.
C++ addons are great, that is what makes JavaScript on the server bearable, when performance matters and we're stuck with it, instead of using Go, Rust, Java, .NET.
I think this is a great chance for us to evaluate how capable AI is and at what cost does it deliver such capabilities.
I would love to know how much token they spent on the rewrite starting from deciding to rewrite till they actually ship. I assume since the beginning of this whole rewrite saga they have been spending tokens on it every day so the number that Jarred mentioned in his blog post has increased a lot since then.
Also assuming the future releases will be done through AI as well I wonder how much it will cost for each of them.
Jarred wrote a lengthy blog about the rewrite a little while ago and IIRC the token burn numbers were in there (or at least the burn up to that point in time).
Lots of people are apparently happily using the canary version in prod. Anthropic has been using it for ages now, as has Prisma.
Why not release 1.4 (or 2.0 or whatever) then? It won't get picked up by lots of tooling and won't get allowed in many "no pre-release in prod" shops until then. It had plenty of issues before (many thousands of open issues) so a few thousand more is hardly a big deal...
No, the article shows that Jarred is claiming to be working on releasing it. And maybe he is, but he's also not provided reasons for the repeated delays, so for all we know he's goofing off drinking pina colladas while Claude is running the show and Tweeting on his behalf.
I think the rewrite will end up being more expensive (for the same quality) because AI was used.
At the same time this specific rewrite is being used to drive the narrative that reading the code is counterproductive.
If they had delivered their promises, they might have convincingly delivered that argument.
...but the cost keeps going up and up and up and up! If it was only the token costs it might still be economically viable, but I'm imagining that when you add in the pro-rated salaries of the human engineers this would start to look like a financial blunder even if it released today.
It's true that I don't want a success story told about a failure, because then replicating the "success" would likely produce lots more bad outcomes
This remains one of the strangest and frankly most annoying beefs I’ve ever read about all decade, and I say this as somebody with no particular affinity for bun, rust, or zig.
I'm quietly hoping it fails. This will once and for all silence all those paid-shills that are calling out that A.I. will replace all software developers.
It may also lead to the A.I. musical chairs game stopping and a multi-trillion dollar collapse of the stock-market and, most likely, the financial system. Hopefully we'll be able to buy DRAM and other stuff that uses it cheaply again.
I mean there's so much cruft on the internet about people talking how revolutionary these code generators are and that they fixed 5 years worth of technical debt in 2 months etc. etc. Considering the multi-trillion dollars involved I don't believe we can trust anything we read anymore. Many of these proponents are nothing but paid shills.
> Our conclusion is deliberately modest: under the conditions Prisma Compute cares about, the Rust rewrite behaved better than the stable release we had been testing. That was enough to change what we shipped.
Yet another clickbait making outrageous assertions. Fwiw I don’t think anyone’s reputation is on the line and so far, you’ve offered only anecdotal evidence at best. Things like this should be backed by stats and community surveys and a whole lot more. I didn’t see anything about number of downloads or any of the usual metrics used to judge software these days
Your whole scientific endeavor in this seemed to thrive on the rock solid foundation of tweets.
And all those referencing of the Kelly post makes this even more of a clickbait. All you had was to put all the ingredients of a post that will draw the crowd but offer very little on the subject matter it purported to discuss
