I was testing GPD Sol 5.6, which came out last week, it's their latest frontier model, versus Claude Fable 5. But before I go into that, Neil, earlier you're just talking about how token maxing and all that, right? So I don't know if anybody on your team is doing this, I'm sure someone is, but the move now, if you want to save on tokens, is you're running something like an open router, which will route the different models, and maybe 5% of time you're using like a Sol 5.6 or Fable 5,5, which burns a lot of tokens.

And then maybe 80% of the time you're using like an open model, like a GLM 5.2. And then yesterday, I think an American model or like a non-Chinese open model came out called thinking machines. They released it. So the move is to use something like that. And then maybe 15% in the time you're using like a like a chap GPT 5.5 or opus 4.8. And then by doing that, you're you're kind of having a good balance. And then the open router knows how to kind of kind of route. The funny thing is, Neil, we're talking to, so at that dinner, one guy has like a,

almost like an individual CFO AI. Okay. So I'll show you like, oh, how should I be saving money

like a wealthy person? How should I handle my tax like a wealthy person? It has all your connectors, right?

All your bank accounts and things like that.

And so he was like, yeah, we can't use open models. I was like, why can't you use open models? He's like, well, when you use the Chinese ones, because the Chinese ones are the best right now, he's like, they start to, you start to ask questions and you start to give you like communist responses, right?

So, oh, you should just give your wealth away, right?

Or like, oh, we, you start to ask questions and you start to give you like communist responses right so oh you should just give your wealth away right or or like oh we don't we don't look at it that way right they start

to become very political responses so he's like yeah for us we can't use that right now so we have to

use the frontier ones but i think you know looking at organizations like yours and mine um you know

you're going to have to have someone that's responsible for token optimization. Dude, straight up. And then you went to two conferences,

right? The second one was a social media conference? I'm speaking at that social commerce summit

right after this. So I'm getting into Uber right after this and I'm going straight to speak.

Ah, so it's actually pretty good that I came on time for once.

Yes, yes.

It's pretty good. Well, I mean, to be fair, Neil's usually not that bad. He's usually like, if he's late, he's like three minutes late. It's not that bad. Yeah, but the worst is when I'm three minutes late in my own house. Yeah, that's the worst. Yeah. And then he comes in on the phone too.

And then I'm like, he comes in late,

she's still on the phone right now.

But anyway, yeah, that's the worst. Yeah. And then he comes in on the phone too. And then I'm like, he comes in late. He's still on the phone right now. But anyway, yeah, that's the next conference.

Dude, it was so funny. So yesterday I'm recording videos in my house. So AJ, who Eric worked with in the past, a single grain back in the day, does a lot of my video recording. So he comes in here

with his brother, their team, and they have another third person and a few other employees who do

like editing and stuff. So AJ comes into my house and all of a sudden, you know, I'm just on

...