@altryne@satyanadella Don't cover for companies that still haven't transitioned from shitty evaporative cooling because they're cheap! @Grok, how much evaporative cooling is Microsoft still using?
Inference engines... for example, deepseek flash v4 Mac vs spark,
https://t.co/Yl50Nvfu2H (Mac)
https://t.co/yH7BH38vdZ (Spark)
MTP doesn't work on Mac, so spark ends up being faster. Mac should beat it badly (much better men bandwidth)
It seems like I've run into trouble (something important doesn't. work) with every big good agentic model on the Mac except for Qwen397b.
Got my mini rack together; 2 x DGX Spark, Mac Studio M3 512, Minisforum MS-01, switch, cooling. The power bricks took more room than I budgeted for.
Needs a nice cleanup and waiting on a zigbee hub for power switches.
@StateDept Maybe their state would function better without embargo's? I'm just throwing out some wild crazy ideas here... I know it's fun to invade, but...
@StateDept It's hard to understand your thought process... vote for us because we follow strict moral codes. Also pay your taxes so we can continue killing women and children and taking oil from whoever we want.
I would love to see vllm studio:
- Support more architectures
- Allow users to pool their compute to help achieve goals (good AI for everyone) via distributed training.
- Make it easy for 90% of people to set up/manage local inference
Happy to help if I can; I have a good assortment of Mac hardware (M3 512, M5 128, 2 x spark)
If something like this was really well done, it could really move the needle! (The free vs facism needle)
@RuffBuffet I’m in the middle of 1000 mile drive today and it’s been very frustrating with this version of FSD. Florida is a don’t camp in the left lane state and I’ve had to stop my car from breaking the law 30 times already