OpenAI's ChatGPT Can Now Help From Your Phone

OpenAI has made it possible to use Codex—the coding tool built into ChatGPT—from your mobile phone. If you connect your phone to your computer, you can now see what's happening with long projects and give the system approval to make changes, all from the phone in your hand. The feature is available as a preview on both iPhone and Android across the world. Support for Windows computers is coming later.

How It Actually Works

When you use Codex on your phone, the system does not run the heavy calculations on your phone itself. Instead, your phone stays connected to your computer or laptop, where the real work happens. Your phone is more like a remote control: it shows you what your computer is doing and lets you make decisions about it.

Think of it like checking on a slow oven from another room. Your phone displays what your computer has been working on—the conversation history, what changes the system wants to make, and special tools you have set up. None of that context is lost when you switch devices. The real computation stays on your computer, where it runs much faster and more reliably.

Why This Matters for People Who Code

Software developers often write code that takes a long time to run. While the computer is working, the developer might step away, grab coffee, or work on something else. But sometimes the system needs a decision from a human: "Should I make this change? What should I do about this error?"

Until now, if you were away from your desk, you had to go back to your computer to answer. The mobile preview changes that. You can approve changes, check on progress, and give guidance from anywhere—all without losing the context of what you were working on.

This is especially useful for teams spread across different time zones or for people who work from home and move around their space. It reduces the friction of checking in on your work throughout the day.

What OpenAI Says the System Can Do Now

OpenAI has documented that Codex can now handle longer and more complex sequences of steps. Instead of needing constant supervision, the system can work through multiple steps, recover if something goes wrong, and continue toward a goal. This is important because the more a system can do without interruption, the more useful it becomes to people managing it from a distance.

Research groups studying these kinds of systems have found that their reliability improves over time. Tasks that would succeed half the time now succeed most of the time. That steady improvement suggests the underlying technology is genuinely getting better, not just being tweaked at the edges.

Under the Hood: Things to Know

For the system to work, your phone needs an internet connection to your computer. This introduces a delay—what technologists call latency—when you interact with your system from the phone. If you need immediate, split-second feedback, that delay matters. But for checking status or approving a change, it is usually fine.

The system has to handle what happens when you lose your connection—maybe you move between WiFi networks or your phone goes to sleep. OpenAI's approach likely stores enough information locally to pick up where you left off without losing any work.

Security is worth flagging here. Your phone is now a gateway to your computer and its projects. That means your phone's safety matters more than before. If someone steals your phone or tricks you into giving them access, they could access your work. OpenAI has built authentication and checks into the system, but these security boundaries become more complex when more devices can connect.

What Comes Next

The preview status tells us OpenAI is still learning how people use this feature and what works well on a phone's smaller screen. Mobile interfaces for technical tools need time to get right—people interact with phones differently than they interact with laptops. History suggests multiple rounds of refinement before this becomes a standard feature.

Looking at how remote development tools have evolved over the past two decades, we have seen this story before. When cloud-based coding environments first arrived, they started with basic functions—editing code, compiling it. The real value came later, when all the complex tools—debuggers, deployment systems, testing frameworks—became accessible from lightweight devices. This Codex release follows that same path.

The broader direction here is clear: AI coding assistants are moving from tools you use at a desk into tools available everywhere you are. That shift changes what becomes possible—not just in how often you can check in on work, but in how work itself gets organized. It may sound like a small change, but it is the kind of small change that has historically reshaped entire industries.