jchw

32 ms·

1.
by jchw 5d ago
That is pretty clever. In a particularly C++ heavy program I was able to find decent splits by starting around where the vtables were stored and searching from there, which was helpful since they also often pointed into the relevant part of
2.
by jchw 5d ago
How in the world are you finding reasonable TU boundaries, without debug info? I've always been puzzled by this in particular.
3.
by jchw 5d ago
Oh yeah - absolutely. This isn't even just a PowerPC thing - it happens more generally, because usually if you're forwarding both arguments and return values for a function that has the same calling conventions, you don't rea
4.
by jchw 5d ago
My eternal advice to anyone doing heavily LLM assisted projects is to do them a bit less LLM assisted. You should still use LLMs, in my opinion, they're wonderful tools. But I know reading this page that this text is mostly LLM-generat
5.
by jchw 5d ago
Yeah, this is probably true. I've done non-matching decompilations of modern software up to a few hundred kilobytes worth of code - it is challenging but doable. I have no idea how hard it would be to get to matching with LTCG no matte
6.
by jchw 6d ago
Most people are working on builds where debug information was either inadvertently or sometimes intentionally included (e.g. for beta releases sometimes debug builds with debug info would ship for sake of making things easier.) These builds
7.
by jchw 6d ago
Making a function fully matching by virtue of hacks like this is mostly not harmful to the ability to understand the code, but is a useful tool in ensuring that the code as a whole really is matching and identical to the original. Otherwise
8.
by jchw 8d ago
I feel like it has happened, but I certainly can't remember a specific time. It feels in a similar vein to the NSA releasing Ghidra to the public as open source software after the Snowden leaks. I mean, on the contrary, imagine if th
9.
by jchw 8d ago
The answer to your question is very simple.
10.
by jchw 9d ago
The scammy part was definitely making all of the claims as if they were already true. A lot of people sincerely believed it. My recollection was that it took months before some of this was fixed to explain that it wasn't available yet,
11.
by jchw 9d ago
Well, to be fair, you're correct. I am asking for a higher threshold of evidence. It's a stronger claim. I feel a stronger claim deserves stronger evidence.
12.
by jchw 9d ago
You're simplifying the exchange a little too much. I said: > Anthropic doesn't appear to use Claude for blog posts My claim is literally the lack of evidence, which, yes, can't prove anything. This claim can be contested
13.
by jchw 9d ago
I think you can't trust Pangram in a high stakes situation, but it is absolutely better than random noise at detecting AI-generated text. Which isn't surprising. If the distribution of probabilities can yield blatant Claudisms, it
14.
by jchw 9d ago
To be honest with you, I don't think I would be able to identify with high certainty that the bottom text is AI generated, so it definitely goes a long way to obscure the AI-generated nature of it, but I also think it still feels unnat
15.
by jchw 9d ago
Prove it.
16.
by jchw 9d ago
Sure, but even Anthropic doesn't appear to use Claude for blog posts. (I don't think anyone should. The prose stinks.)
17.
by jchw 11d ago
> you were complaining how nobody writes decent documentation They still don't. > Now that the README actually describes what the project is all about you don't read it because "it's AI slop". Going from no do
18.
by jchw 12d ago
I've gone from being blown away by what LLMs could do with web design (I mean, it's impressive, especially what it does purely without vision...) to wishing it couldn't. Everything has become "same"y in a way that i
19.
by jchw 12d ago
> But, why would we not create a fully AI-operated factory/chemical plant/fab as soon as it is economically advantageous? Or a missile silo, as soon as it seems tactically necessary? Well, for one thing, because it would be dan
20.
by jchw 12d ago
Edit: Actually, nevermind. You guys are not capable of having a reasonable discussion about this.
21.
by jchw 13d ago
To me, it is a feature to have a password required at boot to unlock the drive. Relying on TPM for keymatter is convenient but comes with caveats I don't like. I don't personally trust that the boot chain and OS on a modern system
22.
by jchw 13d ago
Some websites like PC Gamer have some shitty middleware (that seems to be very careful to hide its identity, only appear once per session, and not immediately register on page load, probably to try to delay browsers banning the ever loving
23.
by jchw 14d ago
Though that is not the type of abandonment that IKEA is worried about in this case. In fact it isn't really clear what form of trademark abandonment IKEA would have to be concerned about for this action to be justified. IANAL but IKE
24.
by jchw 14d ago
Yeah, but that doesn't sound like AI at all. Not even close.
25.
by jchw 14d ago
Oh that's simple. Both Tesla and John Deere are bad guys. And yes. I am happy to expand that list to literally every other manufacturer you have in mind. They are all bad guys too. If manufacturers were altruistic good guys, they would
26.
by jchw 14d ago
Trademark is about preventing someone's brand for being confused for another. Unless someone was confusing this for being an official IKEA product I'm not sure why it would be in the ballpark of trademark. There is also trade dres
27.
by jchw 15d ago
There really isn't any moral argument against distillation, which is itself pretty goddamn benign. It's pretty simple. Someone pays for Claude access. Claude outputs tokens that are not copyrighted. Then you train on those tokens,
28.
by jchw 19d ago
I've actually been thinking a lot about this. I do absolutely 100% lament that we are possibly going to see the loss of the art of programming. Not entirely, sure; some people will do it by hand the way some people do anything by hand.
29.
by jchw 22d ago
I hate to invoke Poe's law but, I've now flip-flopped like six times over whether this could be serious. I think it is serious. In which case, I gotta say, it really seems like you didn't spend much time thinking about this.
30.
by jchw 25d ago
It absolutely isn't. It is only ludicrous if you hold Apple to some weird standard of ethics that they absolutely do not deserve. If you think this practice isn't common, that's purely a matter of ignorance on your part.

More ›