Most AI Interaction Will Go Through Your DA

The stages of technological evolution (click for full size)

A long time ago, I wrote about how things tend to start off as ideas, then become websites, then applications, and eventually move into the operating system. Like an app maturity ladder. I don’t quite agree with that exact order anymore, but I thought it was a useful exercise.

Ever since writing The Real Internet of Things in 2016, I’ve thought that the final stage is not actually the operating system but rather your Digital Assistant.

And this is the way I still see things, and it’s how I’m interpreting this whole move to everything being an MCP server and everything being an AI browser.

The Final Stage of Development

I like to think about what the final stage of development for a thing is. If you think about wanting to do coding, or really anything at the computer—making art, writing a book, designing a comic book, building an application—I think something fairly close to the final stage is having a bunch of screens around you in a kind of Minority Report situation.

The iconic gesture-based interface from Minority Report (click for full size)

So basically, you’re talking to your screen and it’s doing different things.

Who Are You Talking To?

So the question is, who are you actually talking to when you’re talking?

One way to think of that is you have a specific AI app that you’re talking to, like ChatGPT or Claude.

Another way to think about it is you’re talking to a browser or you’re talking to your operating system.

I think your operating system is the closest version, but not quite correct.

The problem with the operating system is that it’s just an OS. It’s just a piece of software, and you might want to give it a personality or whatever, but ultimately it’s macOS, Windows, or Linux. So by its very nature, it’s kind of impersonal.

I think the much more natural and therefore inevitable location for this is you are talking to your digital assistant, who in my case is named Kai.

The Inevitable AI Future

So Kai knows everything about me, as I explored in my post about AI’s Predictable Path.

The seven components of AI’s predictable future (click for full size)

Kai already knows everything about me: how I like to code, what apps I’ve already built, my website, how I like to communicate, what I mean when I pause or give certain comments. Ultimately, Kai knows everything about me in a way that my operating system does not.

Now you could argue that if Kai is sitting on top of my operating system, then my operating system would have access to everything that Kai does. I would say that’s a separate technical question. But in general, I would say the answer to that is no.

Even my operating system should be treated as somewhat of a third-party.

Everywhere You Are

Ultimately, Kai is going to be everywhere that I am. I could do coding or building or whatever when I’m talking to my phone, when I’m out on a walk and I’m talking to my AirPods, or when I’m sitting at home with all my monitors and all my tech around me.

My operating system is not a guarantee. I want to be able to switch operating systems or phone carriers or tech stacks or whatever, but Kai will always be with me.

The Battle Heats Up

This is why we’re seeing the battles heat up so much with OpenAI, Anthropic, and Gemini. I don’t know how much they’ve figured this out versus if they’re just going in the same direction because it’s just natural.

But OpenAI’s recent release of Operator and Computer Use is taking us even further in this direction.

OpenAI’s Operator performing tasks autonomously in a browser (click for full size)

We’ve already seen computer use by multiple vendors, and now we have this agent thing which is basically computer use that’s even more powerful.

The other piece that they have that’s a massive component of this is obviously the memory.

The Missing Piece

The one piece that they haven’t added yet, which I’m sure is soon to come, is actually naming your assistant. At that point, you will have basically a personality around all this knowledge and all the capability.

And we will be at the exact place that I talked about in The Real Internet of Things.

What Do You Think?

Keep an eye out for it. If you have any ideas about how I could be wrong about this and how the final destination will be the operating system, the browser, the mobile OS, or some other place that you can think of, let me know.

I’m curious what your thoughts are.