The ChatGPT Phone Is Not What You Expect
It’s been widely reported that OpenAI is working on a consumer electronics product. Most of the recent commentary has been about a portable smart speaker, sort of like an Amazon Echo but with a battery. It makes sense, if you think about what the service called ChatGPT is today. It’s an AI model that you can talk to. I don’t know how many people like talking to ChatGPT by voice rather than typing. But certainly a lot of people like to do that.
Talking to an ambient computer program is very sci-fi and is useful in some cases. I’m sure there are people who will even use this mode for complex chatbot projects like making a web site or app. But that gets awkward if you want to see your work: you really need a screen to see the web site you are making. And probably tapping your finger on some kind of input is going to be way better than talking out loud in many situations.
Look, this web site in particular has a long and proud tradition of making incredibly wrong speculation. So I’m not going to stop now.
The smart phone is the most important consumer technology product ever created, and probably will remain so until biological implants of some kind are safe and widely adopted. So probably forever!
A smart phone is only called a phone by the happenstance of how it had to go to market. It’s really a tiny computer, really the first and best truly personal computer. It has a screen, touch input, microphones, speakers and cameras. The only thing it doesn’t have is a smell sensor! With that input and output, you can do just about anything a person would need with software. You just need to write the software!
Apple and Google control the smart phone ecosystem with their operating systems. They decide how all other software is allowed to interact with the user and the world. They decide that software from third parties lives in things called “apps,” and what the rules are for waking up to bother the user. Controlling a platform is a great business! It gives both companies both a strong source of revenue and a moat that can be used to thwart competition that could usurp their control.
So companies like OpenAI and Anthropic are on the leading edge of an incredibly useful technology. The investment they have taken suggests that they want to be the platforms for the next phase of technology. But if Apple and Google control the computer in everyone’s pockets, how can they ever become a platform? To be a platform, they have to be the gatekeeper for everything, and the incumbents aren’t going to give that up.
It’s clear today that Apple and Google preference their own AI models over those of any third party. In the case of Apple, their new model that is coming this fall actually is mostly based on Google’s Gemini AI model.
So OpenAI is going to make a smart phone. Of course they are. They aren’t trying to hire every industrial designer and hardware engineer from Apple so they can make a smart speaker. Come on. My most charitable reading of the OpenAI donut-shaped portable chat speaker is that they need to create the institutional structure to ship a consumer hardware device. So it might as well be something they can slap together quickly.
But making a smart phone is stupid, isn’t it? Amazon tried it, it was a spectacular flop. Facebook tried it, but it never even launched. Neither of those efforts had anything compelling to offer. Why would anyone want a smart phone that was, uh, super good at shopping, I guess, and basically the same as an iPhone or Android phone in every other way?
Won’t OpenAI have the cold start problem? There’s so much you’d have to build to make a phone that did what people need their iPhone for. The essential functionality of messaging and communications, a web browser, sure, that’s a lot of work but you could build it. Especially if your software team is all using AI to build it. But then you have to get all the other apps in the world, more or less, to build for your new platform. And then what do you have? You just have a device that is identical to an iPhone or Android in almost every way. You’re not going to win in the market with that.
So here’s where I get into wild speculation. OpenAI (and Anthropic too) are not trying to make a super intelligent chatbot. OpenAI kind of stumbled into a massively successful product by making a chatbot, of course! But that’s not the goal. The point is to take software to a place it has never been, to get to a level that is as far beyond apps-in-a-grid as apps-in-a-grid is beyond green text on a terminal.
OpenAI should build a phone that is made for the AI of 5 years from now, not the chatbots of today. I think that’s what they are doing.
Remember way back in 2025, you could use AI models to write software. It was pretty crappy. You definitely could, even as a person without any experience as a programmer, create software. If you, an experienced programmer, or if you had a friend who was one look at the code produced by the AI, they would be horrified. “Ha ha! That’s complete garbage!” they (or you) would laugh. “No robot will replace me, with my 10/15/45 years of software development experience! Ha!”
Just a year later, many programmers are using Claude Code or whatever to do their jobs all day long. Old timey programmers like me are going over all the apps or web sites they had thought about making, and using AI to make them real. “Golly, it sure is great to have this tireless army of junior programmers working for me. It’s a good thing I have decades of experience building good software systems. No robot will replace my skills as a software architect! Ha!”
Keep pulling on that string. How quickly will AI models be able to assemble exactly the software each person needs, tailored to exactly how they want to interact with it? It sure seems like it will come sooner than we expect.
Imagine the screen in your pocket just presents a UI for exactly what you want to do without having to find an app icon in a grid. You can talk to it or type to it. Or maybe it just shows you what it thinks you want, and is right most of the time?
Is it writing new software for you all the time, just in time? Probably not. There’s probably a lot of common stuff that everybody gets. And there’s probably a suite of “you” apps that get developed and refined, but they aren’t recreated every time you pick up the phone.
That’s the future OpenAI is expecting as they build a phone. Certainly, many businesses will have to make their stuff available to the ChatGPT phone: Uber, Netflix, Spotify, DoorDash, travel agents. There’s still value in doing things in the world for people. It’s not as clear if there is value in developing an entire user experience. For things like Instagram, it’s a bit murkier. An ad-supported entertainment product like that kind of needs to optimize engagement and attention to the maximum degree possible. You kind of can’t just hand that over to five million unique self-coded apps that consume the Instgram feed.
But maybe that doesn’t matter. Maybe businesses that rely on fine control of every pixel are not going to survive this next era, just as magazines and newspapers are being crushed today.
It’s a huge mountain to climb. But from here, you can kind of see the shape of it. Do Apple and Google realize where it’s going? I expect they do. Can they do what needs to be done and transform the smart phone before OpenAI gets there?