deaux

50 ms·

901.
by deaux 9mo ago
I called one out here recently with very obvious evidence - clear LLM comments on entirely different posts 35 seconds apart with plenty of hallmarks - but soon got a reply "I'm not a bot, how unfair!". Duh, most of them are
902.
by deaux 9mo ago
Same, I'm still not sure where it shines though. In each of the three big domains I named, the respective top performing closed model still seems to have the edge. That keeps me from reaching for it more often. Fantastic all-rounder f
903.
by deaux 9mo ago
That's the setup you want for serious work yes, so probably $60kish all-in(?). Which is a big chunk of money for an individual, but potentially quite reasonable for a company. Being able to get effectively _frontier-level local perform
904.
by deaux 9mo ago
It's incredibly bad on this article. It stands out more because it's so wrong and the content itself could actually be interesting. Normally anything with this level of slop wouldn't even be worth reading if it wasn't
905.
by deaux 9mo ago
I see this by far the most on Github out of all places.
906.
by deaux 9mo ago
> Takaichi is much more popular than Truss was That's about as clear as 1+1=2; I wouldn't be surprised if Truss was the least popular PM/President of the last century in the OECD at "start of tenure D+5". > J
907.
by deaux 9mo ago
> And btw "it's just TARP fraud" is an idiotic thing to say: the fraud was in misrepresenting the banks' positions prior to the crash during the TARP audit. People want CEOs to go to jail for knowingly playing a role
908.
by deaux 9mo ago
You're blaming the public for not caring about 200 irrelevant bankers going to jail for mostly other reasons (e.g. TARP fraud), when there's no reason for them to. There might as well have 0 bankers gone to jail. Getting pedantic
909.
by deaux 9mo ago
As someone like you, this isn't the case, because it _entirely_ depends on environment/culture. Not just for you and me, though it's more extreme for us.
910.
by deaux 9mo ago
Similarly, it's probably not a coincidence that the best F1 driver happens to be the one who started driving karts the earliest (4 years old) and spent the most time doing it out of even all those elite ones.
911.
by deaux 9mo ago
They're 16% owned by Google and Amazon, so they're already a minimum of 16% "bad guys".
912.
by deaux 9mo ago
Hey there, I quit a job over similar concerns, knowing it would lead to a >70% decrease in comp. Without a significant nest egg or wealth, whether personal or through family. Now let me say the same: But those tools buy Teslas and $8 don
913.
by deaux 9mo ago
I disagree, based on having used it extensively over the last week. I find it to be at least as strong as Sonnet 4.5 and 5.2-Codex on the majority of tasks, often better. Note that even among the big 3, each of them has a domain where they&
914.
by deaux 9mo ago
> Codex is surprisingly just as good This completely depends on the domain, as always. Each of the big 3 have their strengths and weaknesses.
915.
by deaux 9mo ago
Some people spend $50k on a new car, others spend it on running Kimi K2.5 at good speeds locally. No one's running Sonnet/Gemini/GPT-5 locally though.
916.
by deaux 9mo ago
And that's at unusable speeds - it takes about triple that amount to run it decently fast at int4. Now as the other replies say, you should very likely run a quantized version anyway.
917.
by deaux 9mo ago
Until last week, you would've been right. Kimi K2.5 is absolutely competitive for coding. Unless you include it in "frontier", but that has usually been used to refer to "Big 3".
918.
by deaux 9mo ago
It's interesting because the latest Cellebrite data sheets showed them to support all iPhones including e.g. unbooted, but apparently not lockdown mode? It also showed they hadn't cracked GrapheneOS.
919.
by deaux 9mo ago
It looks like you do pull shenanigans like these [0]. The person you're replying to even mentioned "ChatGPT 5.2", but you're specifically talking only about the API, while making it sound like it applies across the board
920.
by deaux 9mo ago
It directly disproves this wild claim > You still need $500k in GPUs and a boatload of electricity to serve like 3 concurrent sessions at a decent tok/ps. as being patent bullshit, after which the burden is squarely on you to back u
921.
by deaux 9mo ago
Just clicked on your first link. The appendix is there? Page 18 of the PDF.
922.
by deaux 9mo ago
The mistake you're making here is assuming that > The hysteria over "assault weapons" basically outlawed guns that _looked_ scary, while not meaningfully making anyone safer. This wasn't the goal by the congresspeop
923.
by deaux 9mo ago
If the modern right's obsession with "ivory tower academics" was real instead of a stick with which to beat ideas they don't like, the field they'd focus on would be economics, not gender studies. Most of it is astr
924.
by deaux 9mo ago
Try Kimi K2 (not the new 2.5), it's known for its default voice being decidedly casual and different from most models.
925.
by deaux 9mo ago
Or 0.028% of the net worth of Musk.
926.
by deaux 9mo ago
It's very unlikely that a majority of Anki users only uses the iOS app. 1. Anki isn't your everyday application with your everyday audience. 2. The number of people willing to splash $25 on an iOS flashcard app without first havi
927.
by deaux 9mo ago
http://infolab.stanford.edu/pub/papers/google.pdf stanford.edu, and the appendix is there. In fact on the link you gave the appendix is cut short - looks like an OCR/copying issue but then at a glance it does
928.
by deaux 9mo ago
Yup, though outside the workplace is usually the bigger one - might depend on the country. It's just not going to happen without tobacco though.
929.
by deaux 9mo ago
Can't similar be said for capsaicin?
930.
by deaux 9mo ago
> It isn't a social lubricant like alcohol It is, and I'm not a smoker. Ironically mainly because of the indoor smoking bans.

More ›