Marco Arment

that would be another applicant.

Not that Apple's going to make an $8,000 computer to play games on or whatever, but that's an easy one.

But no, the AI thing is just such a no-brainer now because we are in such an early part of that curve where the models that everyone wants to use literally can't run on anything.

Even on a desktop, they can't run.

They only run on $60,000 NVIDIA rack mount things and probably multiple ones of those given how many tokens they use.

And

That's a place where we haven't been for so long, where there is a thing we want to do in computing that cannot fit on your desk.

And AI is currently that.

It cannot fit on your desk.

I mean, I guess it could if you spent a few hundred grand and put a rack there, but...

Like, it doesn't fit in a normal desktop computer of any reasonable size and dimension.

And it's going to be a while before we catch up with that.

And why do we ever need to do that?

Why wouldn't we just always run it on the server?

Because local has advantages, obviously.

Like, you know, setting aside all the privacy and other stuff like that, being able to do stuff locally is advantageous for multiple reasons, which is why we never fully converted to thin clients where everything is in the data center.

So, yeah.

And yet it was still a useful product for DJs, and you can't even say that about the Mac Pro.

Oh, 2019 was amazing.

What are you talking about?