Ali Eslami

And that's something that we're familiar with.

Earlier, we heard about agentic commerce and agentic kind of chat experiences.

I think one thing that perhaps is less familiar for a wider audience is these...

People still use the word agentic for them, but it's a slightly different concept where an AI is actually in control of a full computer and it acts on your behalf with your credentials, possibly for hours or days on a task that you set it.

Exactly.

So open-claw really kicked this off.

Anthropic now has cloud code.

OpenAI has OpenAI Codex.

Google has anti-gravity.

So these systems tend to start from a perspective of, hey, this is an AI that writes code for you.

But it's pretty clear that this is going to

go on to do much more than just coding and will soon do all kinds of information work for you, like analyzing your finances, sending emails, organizing your calendar, creating documents, making spreadsheets, writing code, shopping, et cetera, et cetera.

And obviously there's people who are involved in building that technology, but

there's also an element of rewiring yourself to figure out how to use that technology in a way that makes you effective at whatever your job is.

Now, we as AI researchers, we sit in this mind-bending space of using the technology that I just described to build the technology that I described.

So we use those systems to build those systems.

And that gets a bit kind of meta.

But that's the thing that kind of where...

Yeah, that characteristic has been around for a long time, though.

So I would say for the past 10, 15 years, perhaps more, we're kind of used to building things that we don't fully understand.