Karl Yeh

is it token allocation now you're going to be talking about?

How much do you allocate essentially compute or tokens to your developers because what are they doing?

What are they going to do based on whatever the question?

So if you think about it,

from a token perspective, maybe that $20,000 researcher a year or a month or whatever may actually be the more cheaper than like getting masked.

Well, I think like the... I think Sonnet 4.6 with what its tooling and agent benchmarks are, I think when you go into co-work or you go into cloud code and actually set up whether your team orchestration agents, your sub-agents actually execute work for you, as long as you have A and MCP or you're using Cloud Chrome, Cloud and Chrome, I think like

As long as you also accept the fact that it'll probably be a little bit slower than you, but you could do multiple pieces of it.

I think that's kind of like, it's all about based on your expectation of what you think it could do versus the hype that people say it will do and also how much cost it will, right?

So it depends on what you're using and if you're going to get rate limited.

Because open 4.6, you will get rate limited unless you're using the API.

Right.

Yes.

But the thing is, though, similar to Microsoft Excel agent, you have the ability to switch to Claude model.

You've got to think what did Claude.

What did WordPress, what is their system prompt on top of it?

Right.

Will you get, will it be better for you to just use Claude or Claude code, have its Chromium and run your WordPress for you versus you using the weird version like...

So are we talking about a connector or are we talking about in WordPress, there is actually a Claude kind of tool in there?

I do, but, okay.

There's one thing I wanted to show everybody.