stymaar

6 ms·

61.
by stymaar 23d ago
> Somehow I expected inference engines are generic LLM runtimes that can execute any weight. In fact, this was close to be true until last year: almost every open model except DeepSeek had a very similar architecture that was pretty clos
62.
by stymaar 23d ago
> which is misaligned with OpenAI and humanity OpenAI is itself misaligned with humanity, as their mishandling of such incidents (and the many other other issues their model have been causing) shows.
63.
by stymaar 23d ago
Right, I see what you mean, but that's merely “saving the company the hassle of bogus support requests” and still has nothing to do with the claimed “improves reliability and consumer safety”. This actually confirms my intuition that t
64.
by stymaar 23d ago
> there really are a lot of potential safety and reliability problems when using random third party versions of components. What are you talking about? If a seller sells you random junk that causes safety issue, they are liable of the da
65.
by stymaar 24d ago
This. Distillation “attacks” are a made up concept. It's as if I claimed that Anthropic made a “training attack” when training on my internet writing.
66.
by stymaar 24d ago
I'm also certain that they violated countless ToS when they scrapped the internet for training purpose.
67.
by stymaar 24d ago
> There is nothing illicit or illegal Which is exactly the point of the person you're responding to. What part of “but completely legal” isn't clear enough?
68.
by stymaar 24d ago
> and the idiocy of the US police is not within their power to solve. Given how much money Google gives to electoral campaigns, I'd argue that it is in their power but that's the least of their problems when they interfere with
69.
by stymaar 25d ago
> Mathematics belongs to all of us. If mathematicians get depressed about it, I don't give a hoot. They are depressed because what's being destroyed is mathematics, actually. > Five years of guided computer-aided math could
70.
by stymaar 25d ago
Also, since it's criminal behavior (instead of just civil damage), the DoJ can sue too even is the victims don't.
71.
by stymaar 25d ago
Lucky you. I've only managed to move my family to Signal, but everything else is still on Whatsapp.
72.
by stymaar 25d ago
At the individual level any mathematicians can, but the social group of “mathematicians” will not unless drastic social changes are made, because that's not how the system works today. Drastic social changes usually take decades, and
73.
by stymaar 25d ago
For those wondering the difference with scratch, it's on the about page: > About Snap! > Snap! (formerly BYOB) is a visual, drag-and-drop programming language. It is an extended reimplementation of Scratch (a project of the Lif
74.
by stymaar 26d ago
I think that the argument is that an hostile model could attack overseas datacenters, host itself there and then launch its attack from there.
75.
by stymaar 26d ago
I don't think anyone who's tried something else can like claude code. The only reason to use it is the subsidized plan and nothing else.
76.
by stymaar 26d ago
Unless that CEO is Mark Zuckerberg, I guess.
77.
by stymaar 27d ago
I don't share your opinion: IMHO, xhigh makes sense if you want a slower Opus4.6 at home. It is able to complete tasks autonomously in a way that I've never seen another local model do. But yes, for simpler tasks or more hands cod
78.
by stymaar 27d ago
> unless reasoning is set to xhigh That's the default and I'm sure almost everyone else is also using it because other reasoning efforts yield subpar results from what I've seen.
79.
by stymaar 27d ago
They said literally the opposite in their message above. In their experience, the caveman speech occurs in chat ui, not in coding harness.
80.
by stymaar 27d ago
Looks like we're a bunch of weirdos reading Qwen's CoT in here.
81.
by stymaar 27d ago
I just tried to reproduce in Llama-cpp's webui: with the following (reverse-engineered) prompt: “Write a complete short story scene of 2000-3000 words. Use a third-person limited perspective focused on a character named Rowan. The sce
82.
by stymaar 27d ago
Do you prompt it to behave this way or what? Because here's the king of CoT I get: > Hmm, but there's a subtlety: does babel-jest + preset-typescript transform the file to CJS by default? No — babel-jest doesn't transform
83.
by stymaar 27d ago
> Qwen3.8 27B's CoT seems ... very similar. What? I've never seen garbled CoT like the one you posted when using Qwen3.8-27B.
84.
by stymaar 28d ago
Sure. But the same way it would be disingenuous to say that copyright exists so that we can have copyleft, it's a fallacy to say that trademark exist to protect the consumer, which is the point I'm arguing against: > The core r
85.
by stymaar 28d ago
FYI you're missing the satire in the comment you're responding too.
86.
by stymaar 28d ago
Nothing you've written above contradicts my argument. which is that there's no liability attached to a trademark as opposed to what something made to protect the consumer would require.
87.
by stymaar 28d ago
That's the licensing agreement between the owner and the trademarks, and their customers for the use of the trademark. But the American Halal Foundation could revise their definition of what counts as Halal, and then license the brand
88.
by stymaar 28d ago
Yet as a customer I don't think I've got to endure as many outages in my life before 2026 as I did for the past 9 months… And it's not just github, everything seems to be down on a daily basis right now.
89.
by stymaar 28d ago
> The core reason to enforce trademarks is not to fluff the corporation that holds it but to let the public be confident in the name. That's something we hear often but it's not grounded in an legal reality: trademarks are pure
90.
by stymaar 28d ago
I made the experiment some times ago: it's faster to edit the link, replace x.com with nitter.net and load the content from there, than wait until the original page load. The time difference isn't meaningful (it's less than a

More ›