just in case you didn’t know that “it’s only for code reviews, I don’t see the problem” is always bad faith horseshit: https://www.phoronix.com/news/Linus-Torvalds-Claude-Code
-
@davidgerard @datarama @zzt i suspect they added that "no insults" thing because they are using prompts as training data and that its bleeding through into models,
@Bredroll @davidgerard@circumstances.run @zzt I think it's because they're mentally unwell and think Claude has feelings.
-
just in case you didn’t know that “it’s only for code reviews, I don’t see the problem” is always bad faith horseshit: https://www.phoronix.com/news/Linus-Torvalds-Claude-Code
> Linux creator Linus Torvalds has been known to use AI to help with kernel debugging and outside of the kernel tree has also done some vibe coding with Google's Antigravity. He has previously said he views AI as a useful tool. It turns out he's even now using Claude Code for working on new Linux kernel features.
> "The code, the test programs, the review behind the list above and the first draft of this text were done with an LLM coding assistant (Claude Code) under my direction. The measurements were run on my machine."
so just a full vibe, done with the LLM that can’t make a web browser, C compiler, rsync features that aren’t riddled with bugs, or JavaScript runtime
this is the shit we were warning you about
was there a reason why you didn’t listen?
@zzt I find it deeply ironic there is an anti-crawler / do busy work landing page sitting in front of kernel dot org where this sludge patch is sitting
-
@zzt One more reason software projects shouldn’t be run on the “benevolent” dictator for life model.
@cjgibilisco @zzt my thoughts this past week: maybe the BDFL wasn't a clever idea after all
-
@zzt It depends, I suppose, on what "under my direction" means. Given who we're talking about, that might very well mean reading the code and responding with angry, expletive-laden rants when he isn't satisfied.
@zzt This, in fact, is the reason why I think that if there's anyone in the entire world who has a chance of actually pulling something like this off, it's probably Linus.
Not because he's a great superprogrammer or anything, but simply due to experience: Reading code and complaining about it if it's not to his liking *has basically been his entire job for 25+ years*. He's also used to dealing with sycophants - though not, as far as I'm aware, with the kind of "malicious correctness" that LLMs tend to do.
(Of course, this also means that if he can't pull it off, then it's probably safe to say that nobody can.)
-
just in case you didn’t know that “it’s only for code reviews, I don’t see the problem” is always bad faith horseshit: https://www.phoronix.com/news/Linus-Torvalds-Claude-Code
> Linux creator Linus Torvalds has been known to use AI to help with kernel debugging and outside of the kernel tree has also done some vibe coding with Google's Antigravity. He has previously said he views AI as a useful tool. It turns out he's even now using Claude Code for working on new Linux kernel features.
> "The code, the test programs, the review behind the list above and the first draft of this text were done with an LLM coding assistant (Claude Code) under my direction. The measurements were run on my machine."
so just a full vibe, done with the LLM that can’t make a web browser, C compiler, rsync features that aren’t riddled with bugs, or JavaScript runtime
this is the shit we were warning you about
was there a reason why you didn’t listen?
@zzt I take some solace in knowing that, while my C chops are not nearly as good as Linus's, I will soon surpass him in skill as he gradually loses any semblance of technical ability by outsourcing it to the bobot and I don't.
-
@cjgibilisco @zzt my thoughts this past week: maybe the BDFL wasn't a clever idea after all
@tipoqueno @cjgibilisco @zzt "Benevolent" has always been wildly open to interpretation.
-
@zzt "Under my direction". Sure.
-
@tipoqueno @cjgibilisco @zzt "Benevolent" has always been wildly open to interpretation.
@randomgeek @tipoqueno @cjgibilisco @zzt "benevolent" for those not on the "malevolent" end of the deal...
-
surely this time. surely the horrid thing that can’t produce stable software for the xteink x4, to the point where people are abandoning their devices out of frustration, will be fantastic at making features for a much larger, much much more complicated kernel that has to work across a gigantic variety of hardware
surely this time
god, remember all the little fuckers who used torvalds’ “not a social warrior project” proclamation as entryism to push for LLMs in other projects “for reviews only” because that’s what the kernel was doing
all of those little fuckers just got the outcome they were waiting for
-
@zzt the biggest problem is that LLMs will create unmaintainable code and maintainers that cant really maintain the code, the projects will see a sharp decline and stagnation
@Bredroll @zzt that is the problem. LLM / vibe coding does not require knowledge of how a system works. It will create high level code that may work, but the "coder" has no idea why. When bugs come up then there is no one with the knowledge to fix it. Can one of these coders troubleshoot a memory or stack issue?
Second to your point, I have yet to see these engines create low level code that is required for high speed, close to hardware applications (routers, infrastructure, kernel, drivers, etc.).
Reminds me of a PHP script that is 200 lines and a Perl script that is 5 lines that does the same thing. In the end the Perl script is faster.
We may get there but by that time will there be any programmers left?
-
J jwcph@helvede.net shared this topic
