Destroying trust in the Linux kernel, for one.

I don't maintain a pastebin or something. You can look at why he defederated lemmy.dbzer0.com if you want.

Nonono, you must understand, every patch Linus made before using AI was flawless and required no review at all.

"pivot to ai" is the first i've heard of him. if you have examples of this behavior in other contexts, i'd love to have a look at it.

I don't know about the specific drama you're talking about, but I had the impression somebody did a presentation on some event presenting a "vision of an AI-friendly KDE" or something, that at first glance looked like it was being presented by KDE.

Lol let there be no mistake about who exactly these bad actors are

If you were around when Debian adopted Systemd, you will recognize this mass hysteria among those who do not code or maintain.

I like how we keep trusting humans to not take shortcuts.

You know humans, the most legendarily lazy, shortcut taking, lying ass fuckin species.

Come on.

The kernel isn't slop yet. I'll let real life experience tell me what's more likely.

What I said doesn't indicate that at all.

AI actually basically called my code slop (and it was), and made changes that improved performance, and fixed issues with incomplete frames. The solution it came up with was far better, and far cleaner than mine. I never would have come up with it..

I mean, these are your words.

What I said doesn't indicate that at all. Not sure how many websocket servers you've written, but there is a reason that there are entire test suites to test them

Even worse, there are all kinds of combinations of half finished frames, masking, etc. The data size isn't predictable and you can also have fragmented frames too.

We also need to connect to all kinds of clients which do things slightly different

And, I had limited time to write it in as well.

If you're not getting any bugs, you're either not doing anything new, or you're designing really basic apps that offload the hard work to someone else.

Big shocker, but the performance fixing came down to bit handling of the header.. and yes, that made a huge difference overall. The code wasn't wrong...

We're working with embedded systems limited by RAM and CPU that operates on a single thread for everything, with lots of clients .We don't have the luxury of throwing in a better CPU or just pretending it's an issue. Even simply enabling logging drastically changes the CPU usage

So you are just a terrible developer then. Not sure what you are trying to show here with your anecdote other than your inadequacies to do your job properly?

The entire concept of hating ai is fundamentally retarded.

Hating LLMs because of how it is made and the resources it takes isn't retarded. Companies are literally stealing work of millions of people with no consequences at all and the environmental cost is added on top of that so that Uncle Fuckface can generate an image of himself as a vampire or next to Abraham Lincoln.

I work with a client that has a full engineering team and they outsourced their project to us because they didn't have the bandwidth and pretty much every response we receive is a AI generated email or document for a design they provided themselves that they don't understand. When challenged about their design, I get met with deer in the headlights stare because they don't know what they gave us.

There are many reason to legitimately hate LLMs

I think they call it "engagement" these days, but it still feels like enragement to me.

That is David's modus operanti for years now and he and hid crew routinely bash Foss projects and communities based on exaggerations and hearsay. These are people who will deliberately choose to disbelieve what people who know more tell them, just so that they can remain upset. I'm Speaking from multiple run-ins with them.

but that quote isn’t from anyone really involved in the patch at all.

Reminds me of the recent panic about KDE becoming a AI Desktop Environment. Blog posts were written, news articles were made and posted to Lemmy, people were outraged.

... Turns out, it was just some random dude who made a "hey someone should make a AI version of KDE"-post on a forum.

Nothing brings people together like hate.

Your account is basically just ragebait about stuff you know nothing about, isn't it?

Your accusation is the definition of slop. Sounds convincing, asserts to have sources, turns out there's nothing of substance.

If something hasn't been noticed for 10+ years, is it really a "bug"?

Yes. If the intended functionality fails under conditions not often seen, it's still a bug. Bugs can be lower priority but that doesn't make them not bugs.

I wasn't going to name names.. lol.

I think part of the problem though, is it's like vaccination, where people have jumped online with no actual knowledge and isolate themselves in online communities where they ban anyone who isn't on their side.

There was a comment here which spoke about the recent "Racing Condition", which is further evidence they aren't a dev (devs have been writing "race conditions" since the kernel started)

But, in practice, Linus has no choice here realistically I feel. It's a huge competitive disadvantage and a security issue NOT to be using it.

It does suck.. And I look forward to AI companies HOPEFULLY getting taxed into oblivion

i'm worried about the direction linux is going, and i am particularly worried about linus himself.

that said, the pivot to ai guy gets his bread buttered by being anti-ai. a number of people on the fediverse maintain their followings in part by constant, vociferous ai bashing.

i also hate ai! i think it's dangerous and one of the worst things humanity has yet created!

but i can read the mailing list, i can look at the patch, and this is so overblown it's nuts. nothing about what went on with this patch matches the descriptions people are giving it. someone even pulled a "i don't understand the rcu that well" quote in order to make their point, but that quote isn't from anyone really involved in the patch at all. it's just cherry picking a loosely-related message in order to have something to scream at.

and, i hate to say it, but this is pretty common with dave and his crew of hangers-on.

it frustrates me to no end that the people who are most visible in the anti-ai camp are people like this, because it undermines everything when it's so trivial to point out how they're blowing things up way out of proportion in order to maintain their rage machine

It's not called a Racing condition, its called a race condition (so, I already know you're probably not a dev anyway, and are just repeating info)

The kernel has always had race conditions.. A good example is: https://www.networkworld.com/article/4200802/linux-xfs-has-a-decade-old-race-condition-allowing-full-root-access.html (this was decades old code)

In fact, race conditions were a big factor for dropping SysVInit.. They're even easier to do on a multi-threaded system, and become more common when optimising and multiple things are happening at once

And, the simple fact is, the development process is DESIGNED to catch them (and it did), but they are so common, that the news never reports them. If anything, AI is actually making them LESS common

Nope. I assisted system test back in '04 when I was doing build/rel for Unix and ent Linux, so it's no big deal: if you have a test plan I'll run it too.