Grant Harvey

I'm thinking it through, I'm reasoning, and then I'm hitting a trigger that lets the AI act.

So in order for it to be a true agent, the agent has to be the one that does the reasoning.

And basically that's possible with N8n and what's called this agent node.

So the first agent that pretty much anyone shipped, and now all the main AI models do this, is Deep Research, where you go in, you give it a task to go out and research, like, find me data on the top AI companies.

And then it'll go out and it will ask you a couple of follow-up questions, and then it'll research that on its own.

I'll just let you decide.

And then it'll go off and it'll research that.

And this was the first kind of like true agent where you have a quick conversation and then it goes off and it can basically do it all itself.

Then they released this thing called Operator.

And Operator is a bit different where it actually creates a...

instance version of the AI running on a cloud computer somewhere and it's like actually using a web browser and trying to accomplish the task for you so just say you decide base it in

Los Angeles.

And I'm not even reading this.

I'm just giving instructions to kind of show you just a quick tool.

And so as you can see over here, it has this thinking process.

You can kind of watch the way that it works.

But it's kind of, would you say this is a true agent, Corey, based on what you?

Definitely.

And we can get into that right now.

Corey, it is literally what you described an agent to be, isn't it?