Note the plural: Mac Studios. Two 512GB M5 Ultras get you about 25 tokens/sec. Four get you 37 tokens/sec. That's Apple's own test results. We're talking $45K+ in Mac Studios. And on Meta's own chart, GLM 5.3 and Kimi K3 still score higher overall. So, this is the best open US model, but we are behind. TLDR: Cool for where local AI is headed. Out of reach for most people today.

Does Apple really believe people are going to be drilling holes in the center of their kitchen table for this thing?

Looks like Grok 4.7 is happening right now. x.ai/news/grok-4-7

I claimed the "dave" handle on @ThePirateFace 🏴‍☠️ Where Hugging Face AI models never die, and are immortalized as torrents. Claim yours 👇 pirateface.co/claim?ref=dave

Unpopular opinion: I really like MacOS. It has never been perfect, and it isn't perfect today, but it is still very good. (I'm also really glad more people are playing with Linux and are finding out how computing is a fun hobby)

My mood RN.

This also works with Hermes and Eero.

It's so obvious to me that we will barely look at apps in the future (but they will still exist *and be important*).

I kindly request that @spacexai release Grok 4.6 and reset usage limits ASAP. Plz? (It's gonna be a long 24 hours.)

Few people understand what’s already possible. Tech journalists and podcasters not losing their minds over this tells you how out of touch they are.

Let's go! Grok 4.5 is my new default model.

Maybe AI really isn’t that good yet.

"Smart" is considered insensitive by the MLS. Idiocracy is here.

Texting Claude Code From Your iPhone (iMessage Channel Setup + Always-On Daemon) daveswift.com/imessage-claud…