Can AI Agents Cooperate? Health Evidence and the Huawei-Qualcomm Deal

But for now, we're going to talk about this emerging AI agent economy.

It turns out that Walmart and Meta have kind of put together an initiative that starts some conversation about the standards that might be required to drive all that.

There are other consortiums out there that are kicking around the same idea.

But Drew, are we on the cusp of something here? And how do you think this might all play out? Because it's one thing for me to tell an AI agent to go do something.

It's another thing for me to give it access to my money.

Yeah.

And sadly, the first thing people want to do is have it fix their budget.

Right.

That may be the last thing you wanted to do.

But yeah, eugenic economy.

I've been joking for a while now that like the future world is going to be based purely on tokens.

Right.

Like a token is a fundamental unit of work, either compute or with a robot like physical labor.

Right.

But it's a total mess.

No one knows what a token is.

It's completely malleable.

There's no good way to sort of compare them.

So it's a perfect currency if you're building an agentic economy.

But that said, the desire to have standards is obviously super important.

Unfortunately, those standards discussions are really in their infancy.

So what they're talking about for the protocol is, use BOAuth, that's kind of it, and more controls to come later.

Effectively, what they're trying to do, I believe, is creating sort of standard ways of describing the permissions that an agent should use and leveraging things like OpenAuth to help that.

But it's not groundbreaking.

And there's, like you said, lots of other initiatives in this front.