Hacker News new | ask | show | jobs
by phil21 122 days ago
It’s really just easier integrations with stuff like iMessage. I assume easier for email and calendars too since that’s a total wreck trying to come up with anything sane for Linux VM + gsuite. At least has been from my limited experience so far.

Other than that I can’t really come up with an explanation of why a Mac mini would be “better” than say an intel nuc or virtual machine.
1 comments
steve1977 122 days ago
Unified memory on Apple Silicon. On PC architecture, you have to shuffle around stuff between the normal RAM and the GPU RAM.

Mac mini just happens to be the cheapest offering to get this.

link
cromka 122 days ago
But the only cheap option is 16GB basic tier Mac Mini. That's not a lot of shared memory. Proces increase bery quickly for expanded memory models.
link
WA 122 days ago
Why though? The context window is 1 millions token max so far. That is what, a few MB of text? Sounds like I should be able to run claw on a raspberry pi.
link
tjchear 121 days ago
If you’re using it with a local model then you need a lot of GPU memory to load up the model. Unified memory is great here since you can basically use almost all the RAM to load the model.
link
steve1977 122 days ago
I meant cheap in the context of other Apple offerings. I think Mac Studios are a bit more expensive in comparable configurations and with laptops you also pay for the display.
link
yberreby 121 days ago
Sure, but aren't most people running the *Claw projects using cloud inference?
link
phil21 121 days ago
Local LLM is so utterly slow even with multiple $3,000+ modern GPUs operating in the giant context windows openclaw generally works with that I doubt anyone using it is doing so.

Local LLM from my basic messing around is a toy. I really wanted to make it work and was willing to invest 5 figures into it if my basic testing showed promise - but it’s utterly useless for the things I want to eventually bring to “prod” with such a setup. Largely live devops/sysadmin style tasking. I don’t want to mess around hyper-optimizing the LLM efficiency itself.

I’m still learning so perhaps I’m totally off base - happy to be corrected - but even if I was able to get a 50x performance increase at 50% of the LLM capabilities it would be a non-starter due to speed of iteration loops.

With opelclaw burning 20-50M/tokens a day with codex just during “playing around in my lab” stage I can’t see any local LLM short of multiple H200s or something being useful, even as I get more efficient with managing my context.

link