This is very impressive to me. No matter what your thoughts are on LLMs as code generators, getting 1 million lines of code that compiles and passes tests in a week would have been unimaginable to me just a year ago.
I personally think LLMs are going to write a lot of code in the near future and if we like it or not this is going to become more and more common. I took the extreme path of fully embracing this on my side project to learn what can go wrong and how this stuff is going to impact software development in general.
My side project (https://tsz.dev) is also about 1 million lines of Rust. Since my token budget is a lot more limited than theirs, and also I am not rewriting something line-by-line into a new architecture, things have been A LOT slower than Jarred.
It's a strange time to be a software developer. There is a codebase now that I know by heart (paid work) and refuse to accept large pull requests on it as the team lead. Yet I'm doing this crazy experiment where I let agents write and evolve tsz almost entirely automatically. I am not sure the former role can sustain for too long, where you can read and understand all of the code.
Code reviews are also getting much harder to do. Almost every PR I review is AI generated and reviewed! So I have to really look for the 10k ft view of things and fully understand the system the PR is modifying. It is really exhausting because the quantity of PRs has 10x'ed since LLMs started writing acceptable code.
I hope I'm wrong about LLM coding, but from what I'm seeing the profession has changed a lot. Nobody is fighting Tabs vs. Spaces fights anymore... the passion about every line of code is mostly gone around me...
Yeah the next generation is understanding the codebase vs understanding how to prompt the info out of the codebase.
I too took on the rust port of a working software, learned some LLM roadbumps, and it looks very much like the future is LLM coded and human "managed"
On one hand I kinda feel validated for having jumped ship on Zig 3+ Years ago[1] and moving everything to Rust[2], with the language simply being too unstable and unsafe in my eyes, despite my love for comptime and people arguing that Bun and Tigerbeetle were proof that it wasn't the languages fault.
But I also feel bad for the Zig project to loose one of their flagship projects, because while I find the project ultimately anachronistic, I know what it's like to pour your sweat, heart and soul into something,
and having it replaced within a week is a sobering experience even from afar.
A couple years ago this would have been unthinkable because of how slow legacy codebases and rewrites are.
I wonder if Tigerbeetle will also have problems arguing for their solution now that the other project they can point to for customer assurance is gone. And I wonder if they will follow suit eventually simply due to marketing pressure (after having been bitten by the Zig compiler I was surprised that they were putting their super duper high reliability database on top of it at all, but with another big player using it there was at least some peace of mind for their enterprise customers).
Does anybody know what model they are using? Is that Mythos?
This is impressive. I honestly don't know what to think about programming careers, need to reflect on my role in this area. The only thing that comes to mind the that quote from Cypher (Matrix):
"Buckle your seat belt, Dorothy, because Kansas is going bye-bye."
My hot take is Anthropic is going to have a really rough year as they’re going to start to feel the weight of all their vibe coding and products.
I have a hard time believing they have something figured out that the rest of software engineers don’t. I’ve been using LLMs to help code since before claude code and heavily since. I have a hard enough time managing the “slop” from 4-5 agents. I can’t imagine this is going to go well having this done in like a week.
Anecdotally, I am seeing tons of visual bugs in software and apps these days.
Issues of rendering, too many menu options, mobile sizes not being accounted for per se, non-functioning buttons, half-load issues, etc. Consumer facing, major platforms, commercial services.
I can’t ascribe all of it to LLMs but there’s been an objective change in my impotent cursing from “what is wrong with them” to “did anyone even test this?” over the last six months to a year. If the backends start acting like these frontends we might get to see a bunch of cognitive/technical debt blow up at scale.
Given how terrible their in-house support bot is, I don't think they've figured out jack shit. If you truly had a half-decent LLM, you'd think the first place you'd put it to demonstrate the capabilities is handling support.
Regardless of your optimism about LLMs, it seems crazy to expect an entire runtime to be stable after this kind of refactoring pass. I would be skeptical if humans did it too. Rust isn’t magical; it can still compile with bugs despite what some people claim.
This is very impressive to me. No matter what your thoughts are on LLMs as code generators, getting 1 million lines of code that compiles and passes tests in a week would have been unimaginable to me just a year ago.
I personally think LLMs are going to write a lot of code in the near future and if we like it or not this is going to become more and more common. I took the extreme path of fully embracing this on my side project to learn what can go wrong and how this stuff is going to impact software development in general.
My side project (https://tsz.dev) is also about 1 million lines of Rust. Since my token budget is a lot more limited than theirs, and also I am not rewriting something line-by-line into a new architecture, things have been A LOT slower than Jarred.
It's a strange time to be a software developer. There is a codebase now that I know by heart (paid work) and refuse to accept large pull requests on it as the team lead. Yet I'm doing this crazy experiment where I let agents write and evolve tsz almost entirely automatically. I am not sure the former role can sustain for too long, where you can read and understand all of the code.
Code reviews are also getting much harder to do. Almost every PR I review is AI generated and reviewed! So I have to really look for the 10k ft view of things and fully understand the system the PR is modifying. It is really exhausting because the quantity of PRs has 10x'ed since LLMs started writing acceptable code.
I hope I'm wrong about LLM coding, but from what I'm seeing the profession has changed a lot. Nobody is fighting Tabs vs. Spaces fights anymore... the passion about every line of code is mostly gone around me...
Yeah the next generation is understanding the codebase vs understanding how to prompt the info out of the codebase. I too took on the rust port of a working software, learned some LLM roadbumps, and it looks very much like the future is LLM coded and human "managed"
On one hand I kinda feel validated for having jumped ship on Zig 3+ Years ago[1] and moving everything to Rust[2], with the language simply being too unstable and unsafe in my eyes, despite my love for comptime and people arguing that Bun and Tigerbeetle were proof that it wasn't the languages fault.
But I also feel bad for the Zig project to loose one of their flagship projects, because while I find the project ultimately anachronistic, I know what it's like to pour your sweat, heart and soul into something, and having it replaced within a week is a sobering experience even from afar.
A couple years ago this would have been unthinkable because of how slow legacy codebases and rewrites are.
I wonder if Tigerbeetle will also have problems arguing for their solution now that the other project they can point to for customer assurance is gone. And I wonder if they will follow suit eventually simply due to marketing pressure (after having been bitten by the Zig compiler I was surprised that they were putting their super duper high reliability database on top of it at all, but with another big player using it there was at least some peace of mind for their enterprise customers).
1: https://github.com/triblespace/tribles-zig
2: https://github.com/triblespace/triblespace-rs
> Security risks: None. This is dead-code removal
Thanks Claude! XD
https://archive.org/details/flash_allyourbase
The PR was closed because it did it too soon.
RIP bun, it was good to know you.
Dupe: https://news.ycombinator.com/item?id=48132488 ("Rewrite Bun in Rust has been merged" - 170 points - 5 hours ago - 159 comments)
Does anybody know what model they are using? Is that Mythos?
This is impressive. I honestly don't know what to think about programming careers, need to reflect on my role in this area. The only thing that comes to mind the that quote from Cypher (Matrix):
"Buckle your seat belt, Dorothy, because Kansas is going bye-bye."
My hot take is Anthropic is going to have a really rough year as they’re going to start to feel the weight of all their vibe coding and products.
I have a hard time believing they have something figured out that the rest of software engineers don’t. I’ve been using LLMs to help code since before claude code and heavily since. I have a hard enough time managing the “slop” from 4-5 agents. I can’t imagine this is going to go well having this done in like a week.
If you think that it will, I implore you to attempt the same in your large existing codebase and report back the mess. Theirs is littered with TODOs like this https://github.com/oven-sh/bun/pull/30412/changes#diff-6c2b5...
Anecdotally, I am seeing tons of visual bugs in software and apps these days.
Issues of rendering, too many menu options, mobile sizes not being accounted for per se, non-functioning buttons, half-load issues, etc. Consumer facing, major platforms, commercial services.
I can’t ascribe all of it to LLMs but there’s been an objective change in my impotent cursing from “what is wrong with them” to “did anyone even test this?” over the last six months to a year. If the backends start acting like these frontends we might get to see a bunch of cognitive/technical debt blow up at scale.
> My hot take is Anthropic is going to have a really rough year as they’re going to start to feel the weight of all their vibe coding and products.
sed -e 's/Anthropic/everyone/g' and I agree.
"move things and break fast" all around.
Given how terrible their in-house support bot is, I don't think they've figured out jack shit. If you truly had a half-decent LLM, you'd think the first place you'd put it to demonstrate the capabilities is handling support.
The linked PR has been flagged as AI slop, but there seems to be an equivalent PR that's still open: https://github.com/oven-sh/bun/pull/30683
> This PR has been marked as AI slop and the description has been updated to avoid confusion or misleading reviewers
They are already hiring for (experienced) Rust and C++ developers to clean up the slop that used Claude to create.
But then again, this is all marketing content for Anthropic for large scale production-grade rewrites such as this.
It might be the time to flag Bun as AI slop
Regardless of your optimism about LLMs, it seems crazy to expect an entire runtime to be stable after this kind of refactoring pass. I would be skeptical if humans did it too. Rust isn’t magical; it can still compile with bugs despite what some people claim.
its always been slop though, even before the AI