Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Microsoft denies rewriting Windows 11 in Rust using AI

windowslatest.com

Post reply on HN
Print view View original

21–30 of 113 posts

Re: Microsoft denies rewriting Windows 11 in Rust using AI

#21

Post by Krssst »

charcircuit wrote: post #12
monocasa wrote: post #10

Earlier quoted context omitted.

My read was rewriting one million lines of code per engineer month using ML to do the heavy lifting. Which is absolutely batshit. There's no way that can be reviewed properly, even if it's putting all of the review work on all of the other teams.

With this mindset I feel like you would also think bumping a C++ compiler toolchain version is impossible due to all the different changes to code generation that could happen. This is already done today and has similar issues where technically all the code can be affected, but it's not reviewed via a process of manually reading every line.

There is a C++ standard that everyone writing C++ code follows and newer version are usually compatible with one another regardless of toolchain version. Behavior of the toolchain should not change. Worst case you can use deterministic, reliable tools to automatically detect problematic locations if there really is a behavior change. (compiler warnings/errors for example)

AI code generation is not deterministic and has no guarantee of behavior, thus requires review unless incorrect code is acceptable.

Re: Microsoft denies rewriting Windows 11 in Rust using AI

#22

Post by charcircuit »

Krssst wrote: post #21
charcircuit wrote: post #12

Earlier quoted context omitted.

With this mindset I feel like you would also think bumping a C++ compiler toolchain version is impossible due to all the different changes to code generation that could happen. This is already done today and has similar issues where technically all the code can be affected, but it's not reviewed via a process of manually reading every line.

There is a C++ standard that everyone writing C++ code follows and newer version are usually compatible with one another regardless of toolchain version. Behavior of the toolchain should not change. Worst case you can use deterministic, reliable tools to automatically detect problematic locations if there really is a behavior change. (compiler warnings/errors for example) AI code generation is not deterministic and h…

>AI code generation is not deterministic

You don't have to use AI code generation to be what is generating the code or you could require some kind of proof of equivalence to verify the code that was generated.

Re: Microsoft denies rewriting Windows 11 in Rust using AI

#23

Post by monocasa »

charcircuit wrote: post #18
monocasa wrote: post #16

Earlier quoted context omitted.

There's a nearly incalculable difference between bumping a compiler version and rewriting it in a different language.

A C++ compiler translates C++ to an assembly. This project would translate C++ to another language. It's not that different of a concept.

A compiler isn't using a statistical model of language more complex than anyone could understand with a lifetime of study to do its translation, adheres to a standard for that translation, and if you're important enough (and Microsoft internal teams are for MSVC), you get heads up on what specifically is changing so you know where to look for issues.

This is "lets put our postgres database on blockchain because I think blockchain is cool" level of crap you see in peak bubble.

Re: Microsoft denies rewriting Windows 11 in Rust using AI

#25

Post by dmitrygr »

monocasa wrote: post #10
charcircuit wrote: post #9

Earlier quoted context omitted.

>is pretty wild How is it wild? On social media I kept seeing things like people falsely expecting the end goal would require manually reading through a million lines of code. It seemed more like people making up reasons to be mad or trying to dunk on the author.

My read was rewriting one million lines of code per engineer month using ML to do the heavy lifting. Which is absolutely batshit. There's no way that can be reviewed properly, even if it's putting all of the review work on all of the other teams.

What makes you think any existing recent code added to Windows has been reviewed by anyone? This is the company that broke the start menu and the login screen in two consecutive updates.

Re: Microsoft denies rewriting Windows 11 in Rust using AI

#26

Post by Arainach »

charcircuit wrote: post #9
sva_ wrote: post #5

The original LinkedIn post is pretty wild. I wonder if he did a fat line of coke before writing that, or if it actually were any concrete plans that have been worked out.

>is pretty wild How is it wild? On social media I kept seeing things like people falsely expecting the end goal would require manually reading through a million lines of code. It seemed more like people making up reasons to be mad or trying to dunk on the author.

If you're "producing" a million lines of code (that's 50K lines per working day) and not reading them, that's even worse.

Re: Microsoft denies rewriting Windows 11 in Rust using AI

#27

Post by gorgoiler »

monocasa wrote: post #10
charcircuit wrote: post #9

Earlier quoted context omitted.

>is pretty wild How is it wild? On social media I kept seeing things like people falsely expecting the end goal would require manually reading through a million lines of code. It seemed more like people making up reasons to be mad or trying to dunk on the author.

My read was rewriting one million lines of code per engineer month using ML to do the heavy lifting. Which is absolutely batshit. There's no way that can be reviewed properly, even if it's putting all of the review work on all of the other teams.

You’re right about the impossibility of reviewing for style, clarity, and coherence. For correctness though, Windows is famous for being insistent on backwards compatibility over timespans measured in decades and that must surely be automated to the hilt.

As a third-party developer in the late 2000s I remember my boss giving me a CDROM binder (binders?) of every single OS release that Microsoft had ever put out. I assume he’d been given it my his developer-relations rep at Microsoft. My team and I used it to ensure our code worked on every MSDOS/Win* platform we cared to target.

I expect that, internally, the Windows team have crazy amounts of resources to implement the most comprehensive regression testing suite ever created. To that extent, at least, you’d be able to tell if the Rust version did what the old code did even if you didn’t read the code itself.

Re: Microsoft denies rewriting Windows 11 in Rust using AI

#28

Post by checker659 »

Sytten wrote: post #8

I am sure why people are so terminally online to care about this. Like Rust for Linux is also here to stay and we can expect that in 10-20y a large portion of the kernel will be in Rust. Obviously the post was boasting, it was meant as a recruitment ad. They actually probably got a few decent candidates from it.

What makes it inevitable?

Re: Microsoft denies rewriting Windows 11 in Rust using AI

#29

Post by stackghost »

dmitrygr wrote: post #25
monocasa wrote: post #10

Earlier quoted context omitted.

My read was rewriting one million lines of code per engineer month using ML to do the heavy lifting. Which is absolutely batshit. There's no way that can be reviewed properly, even if it's putting all of the review work on all of the other teams.

What makes you think any existing recent code added to Windows has been reviewed by anyone? This is the company that broke the start menu and the login screen in two consecutive updates.

Also the company whose start menu ads made the interface so laggy their "solution" was to just preload the bloat.

Re: Microsoft denies rewriting Windows 11 in Rust using AI

#30

Post by overgard »

charcircuit wrote: post #9
sva_ wrote: post #5

The original LinkedIn post is pretty wild. I wonder if he did a fat line of coke before writing that, or if it actually were any concrete plans that have been worked out.

>is pretty wild How is it wild? On social media I kept seeing things like people falsely expecting the end goal would require manually reading through a million lines of code. It seemed more like people making up reasons to be mad or trying to dunk on the author.

LLMs generate lots of security issues and bugs. Just being "Rust" doesn't automatically fix that. Generating that amount of code means no human review. How could this not end in obvious disaster?
Post reply on HN