27B models run like shit on my hardware, I only get good speeds with 13B models. So I can only assume it was a trick question after all 😭 (Opus 5.5 is the only option that makes sense imo, since it's already enough for pretty much all of my coding tasks.)

And why would you? Especially when it comes from such a suspicious looking cat...

If you read it in good faith, perhaps. But still, I wouldn't take any chances at ending up with an unusably slow LLM because "use permanently" implies I wouldn't be using anything else

Darn, even reading that hurts. Even if she says she meant something else, I can't see this as anything other than a Freudian slip.

This is underspecified. Fitting and being usable are two different things. I don't want to run an LLM at 1t/s, even if it fits my GPU.

Okay I saw your repo, and looks like it isn't using native Windows DLLs at all. So that explains it. I didn't look at what differs your approach from Wine yet, they seem pretty similar at first glance, but it's cool to have alternatives nonetheless!

So this is what heaven looks like...

Wow, things are moving so fast. Are you going to publish Sogen on the Play Store?

kek

了解！

Lol, are the two heroines in the center gremlins or something.

Hi

Happy birthday and happy anki 10k day! 🥳

Turma D: Seitokai Takuindomo

Does that mean Raven is that way because it doesn't use the Win32 comctl32/shell32 DLLs, and therefore isn't using the same rendering code? I guess not, since you're saying it's just emulating system calls... Well, whatever the case, I'm very curious to see what you come up with!

If it runs real Windows binaries natively why is the file dialog and/or button styles not native? I guess the Windows DLLs aren't native?

Next: Can it run GTA V?

This reeks of AI slop, probably it's fake engagement bait.

Recently it's becoming normal for me to wake up to terrible news, wtf is happening to the world...

Back to decomp.me ig