AI in Everything, Everywhere

After I published AI is the New Netflix, Surj Patel, a former colleague, intellectual sparring partner, and long-time reader, asked a sharp question. Given my argument that AI will drive the next wave of upstream traffic, are we headed toward more of a superscaled peer-to-peer model, away from the hub-and-spoke, client-server architecture that underlies the web? And what technologies or business models does that require?

AI, as we have come to know it, will seep into everything the way cloud and mobile did. It has to, because we are incapable of processing the machine-scale information, data, and processes of the post-digital world at a human scale.

The client-server model was built for human-scale consumption. A person makes a request. A server answers. The session ends. That worked for so long. But now it doesn’t. Everything is always on, and needs to be in touch with other devices, software, storage and servers all the time. Whether it is security updates in our home routers, or resident intelligence in the security camera, everything will happen in an invisible fashion. Every simple thing is so big and complex. It’s really strange to think we are still looking for the big bang app, when what’s actually happening is a death by a thousand cuts.

Each small upstream flow (your doorbell, your connected car, your AI assistant syncing its context, your workflow checking in with a model) is individually negligible. Together, as I wrote in the upload nation piece and then again in Say Hello to the Internet of AI, they are changing the network from the edges.

So yes, the direction is toward something more distributed. But I would resist framing it as peer-to-peer in the classic sense. Agentic is just how the software underlying our world will work. Forget the buzzwords. Candidly, I don’t know what they mean when they say “Agentic.” (This has always been the way of the tech industry, whose greatest invention is marketing.)

We did that during the cloud era, and look where we are now. The cloud didn’t replace client-server so much as it abstracted it. The pipes got bigger, the servers moved, but the fundamental architecture remained. AI will likely do the same: the shape of the network will evolve, but the evolution will be incremental, not revolutionary.

The business model question is harder. Surj compared it to power networks and microgrids, a distributed generation and consumption model replacing the old hub-and-spoke grid. If agents are running persistently on every device, if every device is both a consumer and a producer of data, then the billing model, the capacity model, and the ownership model all have to change. The telecoms didn’t see the last transition coming. The distribution players may not see this one either.

Technology and internet are a continuum. The next architecture is already being built. We just aren’t calling it what it is yet.

---

Related: We Are Now An Upload (Broadband) Nation · Say Hello to the Internet of AI · AI is the New Netflix