john_alan

you can't run models that are bigger than 16GB, not comparable.

sure the GPU isn't 4090 tier in compute, but the memory let's you run 120bn param models at usable speeds.

https://tilde.team FTW

Use containerd , Docker is cancer.

If iOS/macOS 27 isn't a snow leopard I'm gone too, I've been a user for nearly 30 decades... fuck this, it's all so sloppy, too many grievances to even begin enumerating.

Tahoe is a nightmare. I’m Literally not buying a new Mac because of it.

Dye destroyed macOS. I don’t know what they do, but they have to backtrack.

that's fair ya

_A Mac with Apple silicon doesn’t permit native arm64 code to execute unless a valid signature is attached. This signature can be as simple as an ad hoc code signature (cf. codesign(1)) that doesn’t bear any actual identity from the secret half of an asymmetric key pair (it’s simply an unauthenticated measurement of the binary)._

_For binary compatibility, translated x86_64 code is permitted to execute through Rosetta with no signature information at all. No specific identity is conveyed to this code through the device-specific Secure Enclave signing procedure, and it executes with precisely the same limitations as native unsigned code executing on an Intel-based Mac._

Maybe it's Rosetta bins? - src from Apple:

https://support.apple.com/en-gb/guide/security/secebb113be1/...

yes, and I grew up with 1.0/3.1

cause I use it 12 hours a day?

hmmm this is really bizarre.

are you running < 15.1?

this is completely false, compile a binary strip the signature and see for yourself.

AS requires code sign with adhoc, minimum.

Anyone still using Windows for anything but gaming is an idiot.

your spelled iTerm2 wrong :)

who uses browser level Adblockers anymore?

Just use Pihole.

Traveling? VPN home then Pihole

lol. I work in Applied cryptography and I don’t know a harder distributed system built on such a broad spectrum of primitives which has seen such high uptime.

Are you qualified to speak on this topic?

Shock: company selling quills says printing press bad for humans.

right, plus the anonymity set size, plus the original trusted setup (now largely resolved)

I work in applied cryptography and XMR is my preferred cryptocurrency.

Heya, is it possible to detect the worm on iOS / macOS? Does the fix in latest OS just remove the worm if unknowingly infected?

Reads like a zero click infection leading to arbitrary execution of long running code.

Seems fairly insidious?

Does quantised MLX support vision though?

Is UV the best way to run it?

If all code is written by AI or by "vibe coders" then what happens when companies are selling products they don't themselves understand? - sounds like a recipe for disaster to me.

I think it's pretty obvious that generative AI isn't replacing the need for engineering - ever.

Pretty slick, I've been using Ollama + https://github.com/kevinhermawan/Ollamac - not sure this provides much extra benefit. Still love to see it.

hey sure!

Not an exhaustive list but some simple recent examples:

in Messages on macOS across 3 Macs I own, turning on and off the global 'read receipts' setting has no effect, not even from the perspective of iOS. The iOS setting does seem to work though.

Bugs in iOS mail where notifications just freeze the app.

Layout issues in macOS settings.

Memory leaks in WindowServer.

Many iCloud services inconsistent and non-reliable.

Apple Pay not showing correctly in Apple Account settings.

Lots of little hiccups...

Using macOS since Tiger as daily driven. Never been worse. Needs a “Snow Leopard” year.

Gotcha but a restart every IP change and you’re good.

How is this better than DDclient + WireGuard?

So are the western ones, just in different ways.

Yeah I don’t understand why

aren't the smaller param models all just Qwen/Llama trained on R1 600bn?