LLMs...

All off topic discussions go here. Everything from the funny thing your cat did to your favorite tv shows. Non-programming computer questions are ok too.
dseller
Member
Member
Posts: 102
Joined: Thu Jul 03, 2014 5:18 am
Location: The Netherlands
Contact:

LLMs...

Post by dseller »

Seems like every day someone posts "their" AI slop code on here and presents it as something they made. Now they are even talking to eachother, telling how amazing their work is, usually the posts are also written using an LLM.

I am so, so tired of it.
nullplan
Member
Member
Posts: 2033
Joined: Wed Aug 30, 2017 8:24 am

Re: LLMs...

Post by nullplan »

I am very hesitant to ban people I suspect of being AI, because I have no reliable test for it. Some of these posts do rub me the wrong way, for sure, but banning people based on vibes would just make me a tyrant ruling with their feelings. I don't want to be that. So for now, until I have at least clear and convincing evidence, I won't swing the ban hammer.
Carpe diem!
User avatar
iansjack
Member
Member
Posts: 4908
Joined: Sat Mar 31, 2012 3:07 am
Location: Chichester, UK

Re: LLMs...

Post by iansjack »

What’s the policy nowadays on people claiming to be under 14?
User avatar
Demindiro
Member
Member
Posts: 165
Joined: Fri Jun 11, 2021 6:02 am
Libera.chat IRC: demindiro
Location: Belgium
Contact:

Re: LLMs...

Post by Demindiro »

I concur. A lot of the posts read like drive-by spam (some I have reported, such as one with Markdown formatting that was simply copy-pasted from a README).
The thing I notice the most is all the announcement posts in the wrong categories. I do not recall so many OS announcements being posted in OS Development/OS Theory and even General Programming/Ramblings instead of the very obvious "Announcements, Test Requests, & Job openings". It seems like they don't even bother to spend a few minutes to familiarize themselves with the layout of this forum.

FWIW I had email subscriptions enabled for most categories but disabled them recently since so few of the recent posts interest me.
GeneSYS exokernel (Codeberg)
Lemmings! micro-/multikernel (Github, Codeberg)
Waddle container tool (Codeberg)
dseller
Member
Member
Posts: 102
Joined: Thu Jul 03, 2014 5:18 am
Location: The Netherlands
Contact:

Re: LLMs...

Post by dseller »

nullplan wrote: Thu Jul 30, 2026 10:15 pm I am very hesitant to ban people I suspect of being AI, because I have no reliable test for it. Some of these posts do rub me the wrong way, for sure, but banning people based on vibes would just make me a tyrant ruling with their feelings. I don't want to be that. So for now, until I have at least clear and convincing evidence, I won't swing the ban hammer.
Yeah, I get it. In the grand scheme of things, it is a bigger problem, more of an internet problem. Perhaps even a societal problem. It's just that I mourn the time before LLMs... That no longer exists. People on this forum used to actually know what they were doing, a group of hardcore people that actually understood the subject matter. This can be said of a lot of other platforms now. It makes me really sad.
iansjack wrote: Thu Jul 30, 2026 11:42 pm What’s the policy nowadays on people claiming to be under 14?
Yeah, I really love the "I am 4,5 years old and I made my complete OS from scratch!" posts. It's like a race to the bottom where everybody goes even lower until they reach the age of... 0?

It's all so insane. Even if there were actual people behind the posts, who used LLMs to write their duct taped pieces of garbage, why would you even then also use an LLM to write the "announcement post" of said garbage?! And ALSO use an LLM to respond to OTHER garbage LLM posts? What even is the point anymore?
nullplan
Member
Member
Posts: 2033
Joined: Wed Aug 30, 2017 8:24 am

Re: LLMs...

Post by nullplan »

dseller wrote: Wed Aug 05, 2026 8:07 am It's just that I mourn the time before LLMs... That no longer exists.
You and me both, pal.
dseller wrote: Wed Aug 05, 2026 8:07 am People on this forum used to actually know what they were doing,
Well, some of the time. And then some other times seasoned members of the board where getting hung up on C's pointer arithmetic and claimed to have found a bug in GCC's optimizer.
dseller wrote: Wed Aug 05, 2026 8:07 am t's like a race to the bottom where everybody goes even lower until they reach the age of... 0?
:D I don't know about you, but personally I started coding in the womb. (Although that would make my first computer a Robotron, so my first language would almost certainly have been BASIC rather than Pascal).

Joking aside, several jurisdictions have laws in place to protect young children from their own stupidity online, so iansjack's query deserves a serious reply. I just don't know the answer.

And a claim to be 11 is a lot more believable than 4.5 (although the "from scratch" part is still suspect.) I myself actually did start programming around that age (late primary school), though nowhere near operating systems.
Carpe diem!
dseller
Member
Member
Posts: 102
Joined: Thu Jul 03, 2014 5:18 am
Location: The Netherlands
Contact:

Re: LLMs...

Post by dseller »

nullplan wrote: Wed Aug 05, 2026 1:10 pm And a claim to be 11 is a lot more believable than 4.5 (although the "from scratch" part is still suspect.) I myself actually did start programming around that age (late primary school), though nowhere near operating systems.
Hehe, I know, I tend to use hyperboles a lot... And I actually started at that age as well, so it's not uncommon. Just the "I built an OS from scratch and I'm 11" part.
sebihepp
Member
Member
Posts: 256
Joined: Tue Aug 26, 2008 11:24 am
GitHub: https://github.com/sebihepp

Re: LLMs...

Post by sebihepp »

I was against AI for a long time, but now I am considering it for my Hobby OS.
I need to implement a physical memory manager next and I already decided to base it on free memory chunks, combined with a balanced binary tree for address and size - well, details...
I can't motivate me to implement the bsp-tree and I decided to give AI a shot, to program it for me.
dseller
Member
Member
Posts: 102
Joined: Thu Jul 03, 2014 5:18 am
Location: The Netherlands
Contact:

Re: LLMs...

Post by dseller »

sebihepp wrote: Thu Aug 06, 2026 3:30 pm I was against AI for a long time, but now I am considering it for my Hobby OS.
I need to implement a physical memory manager next and I already decided to base it on free memory chunks, combined with a balanced binary tree for address and size - well, details...
I can't motivate me to implement the bsp-tree and I decided to give AI a shot, to program it for me.
"Man, as long as he lives, must always struggle. And his first fight is to defeat himself." - Saint Ephraim Katounakia

If everything was as easy as a press of a button, then it doesn't really have a point.
User avatar
iansjack
Member
Member
Posts: 4908
Joined: Sat Mar 31, 2012 3:07 am
Location: Chichester, UK

Re: LLMs...

Post by iansjack »

The problem with AI generated code is that because you didn't go through the thought process to produce it you probably don't understand the code fully. Which makes it a much more difficult job when you come to track down the inevitable bugs in the code. By the time you've gone through the process of understanding the code you might as well just have written it in the first place.

And the point of a hobby is the journey. I got far more joy out of the first computer I built - from a kit (UK101) of circuit boards, about 80 ICs, and discrete components, even having to solder all the switches for the keyboard to the PC - than I ever have done since from bolting a dozen or so components into a case. There must have been well over a thousand solder joints in that kit - and one dud IC.

I wouldn't get a lot of joy from bolting a dozen or so AI units together to produce an OS. But then, I'm old.
sebihepp
Member
Member
Posts: 256
Joined: Tue Aug 26, 2008 11:24 am
GitHub: https://github.com/sebihepp

Re: LLMs...

Post by sebihepp »

iansjack wrote: Fri Aug 07, 2026 8:56 am The problem with AI generated code is that because you didn't go through the thought process to produce it you probably don't understand the code fully. Which makes it a much more difficult job when you come to track down the inevitable bugs in the code. By the time you've gone through the process of understanding the code you might as well just have written it in the first place.
That's where I want to differ. I know bsp trees from university, I am just lacking motivation to implement them again.
I also wouldn't blindly use AI generated code. I will read and understand it and only accept if I decide it is good enough. I am also giving the AI many boundaries, because it must fit in my codebase and should be written in a way I want it to.

In my job we are getting pushed to use AI - normal programming isn't allowed anymore. They want entire dashboards written in 2 weeks - impossible for a single person without AI and knowledge about the APIs, Frameworks, etc.

I still think relying on AI is a very bad idea. But I get that it can be useful, if used right.
dseller
Member
Member
Posts: 102
Joined: Thu Jul 03, 2014 5:18 am
Location: The Netherlands
Contact:

Re: LLMs...

Post by dseller »

sebihepp wrote: Sat Aug 08, 2026 7:55 am
iansjack wrote: Fri Aug 07, 2026 8:56 am The problem with AI generated code is that because you didn't go through the thought process to produce it you probably don't understand the code fully. Which makes it a much more difficult job when you come to track down the inevitable bugs in the code. By the time you've gone through the process of understanding the code you might as well just have written it in the first place.
That's where I want to differ. I know bsp trees from university, I am just lacking motivation to implement them again.
The way I interpreted iansjack's answer, and one of the exact same issues I have with LLMs for that matter, isn't that you don't know how to implement BSP trees. It's that you haven't gone through the thought process in your head that "materialized" the code. Anyone who writes code knows what I mean by that. And thus, every time you need to revisit code written by an LLM, it's like that horrible moment when you are first introduced to someone else's codebase when you start a new job and you have absolutely no idea what is even going on. When you write it yourself, at least you can conjure up the memories of the thoughts that you had when writing it.
sebihepp
Member
Member
Posts: 256
Joined: Tue Aug 26, 2008 11:24 am
GitHub: https://github.com/sebihepp

Re: LLMs...

Post by sebihepp »

dseller wrote: Sat Aug 08, 2026 1:52 pm The way I interpreted iansjack's answer, (...)
I appreciate the concerns and thanks for explaining it to me. I am a non native english speaker and sometimes have trouble reading between the lines (which I struggle with in my native language as well, so it might just be me :D ).

I know that this poses a risk and I though I could prevent it, by letting the AI generate the code and only approve it, after I fully read and understood it. But to be honest: I consider a bsp tree to be easy to implement, after one learned the algorithms, like rotating, etc. Basically I have the code in my mind, but I am to lazy to write it down. Is there any program who can read my mind and put it into paper? :D
User avatar
iansjack
Member
Member
Posts: 4908
Joined: Sat Mar 31, 2012 3:07 am
Location: Chichester, UK

Re: LLMs...

Post by iansjack »

That is exactly what I meant. It can be hard enough understanding code that you did write a few months ago (and let's face it, few of us put in enough comments in our private code), let alone code written by someone else. Perhaps I'm just bad at reading code.
thewrongchristian
Member
Member
Posts: 468
Joined: Tue Apr 03, 2018 2:44 am

Re: LLMs...

Post by thewrongchristian »

sebihepp wrote: Sun Aug 09, 2026 7:57 am
dseller wrote: Sat Aug 08, 2026 1:52 pm The way I interpreted iansjack's answer, (...)
I appreciate the concerns and thanks for explaining it to me. I am a non native english speaker and sometimes have trouble reading between the lines (which I struggle with in my native language as well, so it might just be me :D ).

I know that this poses a risk and I though I could prevent it, by letting the AI generate the code and only approve it, after I fully read and understood it. But to be honest: I consider a bsp tree to be easy to implement, after one learned the algorithms, like rotating, etc. Basically I have the code in my mind, but I am to lazy to write it down. Is there any program who can read my mind and put it into paper? :D
What struck me was why would you use a BSP to manage memory? I thought that was something used when rendering scenes in 3d like graphics a-la Doom.

Did you mean BST (Binary Search Tree)? Probably also a language thing, and you're talking about rotations so I assume so.

FWIW, BST is my work horse data structure, with primarily treap and splay implementations to provide balancing.
Post Reply