John Siracusa

Yeah, I think I had this in the notes ages ago, but this is essentially Google's answer to private cloud compute.

Now, the tricky bits are always like...

How Apple did a bunch of stuff where they bent over backwards to say, and we'll give you this binary image of the thing.

And this is the way that you can prove to yourself mathematically that when you connect to our servers, what you're actually connecting to is the binary thing that we gave you.

So we know we're not just giving you one thing and using a different thing on our servers.

And lots of people looked at that and they said, it's great and this is very helpful.

But in the end, you do kind of sort of have to trust Apple to some degree.

Same thing with the Google private compute.

I'm sure they're doing all the same things.

But in the end, you have to trust Google.

You trust that Google has correctly and honestly implemented the thing that they said they implemented, which I'm sure they have because, again.

you know, they want to do this for Apple.

They want to do it for themselves.

Being able to privately process stuff that is too expensive to do on devices.

A useful service to offer, and this is their offering that they announced in November.

So, again, no announcement from Apple about, I mean, there's some rumors.

I don't think we're going to cover them in the show.

Maybe we mentioned last time, but like the bifurcation of the Apple's rollout of their LLM tech where like 26.4 will have some stuff, but then like the

The chatbot thing will come later.

And the rumors about Apple using Google's TPUs and their data centers was about the later chatbots and not the 26.4 thing.