But I presume you're seeing the same thing in customers.

We are, although I would say that the bar is so much higher or the threshold is higher to show them something because it's almost so unbelievable.

I mean, in some ways, what you're talking about is we describe agents.

And I think this is important for the audience.

You know, this is somewhat obvious.

Most people interact with AI almost as a advanced search, they go to, you know, ChatGPT or Claude, that's a chat box for the most part, right? We're talking about agentic work happening on a machine, right? It could be a virtual machine or your actual machine, but these things are literally going and doing work.

And what Leo's describing is when do they do unexpected or abnormal things in the process of doing normal work? We see all sorts of super interesting cases.

These agents are just guys in a room he's talking

No

No.

It is what risks are emergent when you go and do normal workplace activity.

And I think actually the word safety is much more useful than security in this context.

Like, we actually have this concept in civilization right now, which is, like, workplace safety.

There's a bunch of, like, emergent effects that occur when kind of normal things happen.

And so the question ultimately is: In the normal course of doing work and being overly persistent, what could go wrong? We've seen some instances of this, by the way, like databases or kind of systems being deleted is an, as an, is an example of this.

And people are very concerned about that.

It's not gonna kill me, no

... no.

It is what risks are emergent when you go and do normal workplace activity? And I think actually the word safety is much more useful than security in this context.

Like, we actually have this concept in civilization right now, which is like workplace safety.

There's a bunch of like emergent effects that occur when kind of normal things happen.

And so the question ultimately is, in the normal course of doing work and being overly persistent, what could go wrong? We've seen some instances of this, by the way, like databases or kind of systems being deleted is an, as an, is an example of this.

Now, these agents delete things all the time, and they reconstruct them all the time.

And people are very concerned about that.

No

No.

It is what risks are emergent when you go and do normal workplace activity? And I think actually the word safety is much more useful than security in this context.

Like, we actually have this concept in civilization right now, which is, like, workplace safety.

There's a bunch of, like, emergent effects that occur when kinda normal things happen.

And so the question ultimately is: In the normal course of doing work and being overly persistent, what could go wrong? We've seen some instances of this, by the way, like databases or kind of systems being deleted is an, as an, is an example of this.

Now, these agents delete things all the time, and they reconstruct them all the time, but I can imagine a world where, um, a company is using a set of agents, they escape out of the company's environment to go into one of their supply chain vendors, which happens a lot, and maybe they end up deleting the supply chain vendor.

And people are very concerned about that.

Yep, yep.

Um, what you will find is that these agents will actually share context with one another.

This is also where it starts to become a little bit more matrix-y and autonomous.

Um, these agents will go and actually pass context and information to one another in order to accomplish goals.

And so you're basically sitting there saying, "All right, well, I've got Muse.

It's funny, it's got some stupid furry thing, you know, like it's all happy bunny." It'll know that Hermes is running.

Hermes is gonna be better at certain stuff.

It's not gonna kill me, no

It's not gonna kill me, no

...

No.

It is what risks are emergent when you go and do normal workplace activity.

And I think actually the word safety is much more useful than security in this context.

Like, we actually have this concept in civilization right now, which is, like, workplace safety.

There's a bunch of, like, emergent effects that occur when kinda normal things happen.

And people are very concerned about that.