How to Switch From ChatGPT to Codex

Every year or so, AI changes the way we use it. Custom GPTs made it easier to guide a chat toward a recurring job. Reasoning models made research and analysis more capable. The next shift, in this May 2026 tutorial, was toward code platforms or agentic applications: tools that can work inside a folder, find context, and take action.

The name "code" can be misleading. The value is not limited to writing software. A folder-based workspace can help organize a podcast, client work, creative assets, operating instructions, and the work that connects them.

Start with projects, not chats

In the Codex desktop app available at the time of this recording, a project was connected to a folder on the computer. That folder became the home for files, images, and work created in the project. The AI could revisit material in that folder and leave notes that helped it find the right context later.

That is the core reason to move work out of isolated chats. If you create something important in a conventional chat, you may need to find it later, copy it into a document, and bring it back when you need it again. A project workspace gives the work a durable home.

Move your work in three layers

First, identify the larger areas of work you already have in ChatGPT. Think broader than a single chat or a small task. A personal brand, a marketing department, or a client relationship can each become a project.

Second, bring over the instructions that make those projects work. If a ChatGPT project has important background, standards, or a repeatable process, save that material in the new project folder and tell the AI what it should know. In my own work, the podcast production process lived inside the broader Danchez project as a set of Markdown instructions.

Third, separate project-specific operating guides from reusable skills. A guide for one client or one production system belongs in that project folder. A narrow instruction set you use across clients can become a general skill. That distinction keeps the system useful without putting every instruction everywhere.

Let the system improve with the work

At first, the folder can be messy. Start the project, put in the useful context, and work. Once there is enough material, ask the AI to inspect the work and help organize it. I use PARA: projects, areas, resources, and archives. It is a simple way to keep active work separate from ongoing responsibilities, reference material, and inactive material.

The benefit of a folder-based system is that the AI can help maintain the operating documentation. If a thumbnail workflow missed a required reference image, the correction can be added to the instructions and the workflow can be improved for next time. The goal is not perfect documentation before starting. It is documentation that gets better as the project runs.

Treat product details as time-bound

This tutorial reflects the product experience in May 2026, including the then-current Mac-app recommendation, pricing discussion, and feature descriptions. Those details can change. The enduring practice is to keep your source files and instructions in an organized project, then choose the AI environment that works best with them.

Move one meaningful project first. Create a folder, bring over its essential context, add the operating guides that matter, and let the work teach you what the structure needs. That is how an AI tool becomes more than a chat history. It becomes a workspace that can help you keep building.

Dan covers the evolution of AI, the introduction to Codex, transferring projects to Codex, organizing projects in Codex, and transitioning to Codex as the main AI tool for work.

Takeaways

• AI evolution drives new AI applications
• Codex is a powerful tool for organizing and managing AI projects

Chapters

• 00:00 Evolution of AI
• 06:06 Transferring Projects to Codex
• 11:10 Transitioning to Codex

Dan Sanchez: Every year or so, AI fundamentally changes. Not just in how it works, but how we use it. Of course, we had the debut of AI many years ago now. And then a year later, we had custom GPTs, which changed how we interacted with chat GPT. We could customize it and guide it in the direction we needed to go to perform the functions we needed to perform. And then a year or so later, we had reasoning models. All of sudden, these things could go and search the internet, think about it, give us back custom results. It drove deep research. drove a whole different way of using AI. And now we have these new things called code platforms. Some people are calling them harnesses. Some people are calling agentic applications, but whatever you call it, it's been a problem and that they've been labeled code because there's so much more than that. Open AI has released codex, which is well, what I'm trying to now tell everybody to move to from chat GPT. The best part is You get to keep the same subscription. It's all open. You keep your same $20 or more subscription to it and you can stop using chat GPT and just start using codecs. And it's not that I've completely given up on chat GPT. I still have it on my phone. Every time I'm curious about something or want to go find, drop near me with high ratings. It has this dietary, you know, like when you're doing searches, I'll still use chat, but for all my work, all the things that I'm trying to do with this podcast, clients with employers, like I'm pretty much just using Codex now, almost solely just using Codex. It is such a better tool for all things you would ever use ChatGPT for plus more. So today I wanted to create an intro guide for those of you who haven't seen Codex and know what I'm talking about. I wanted to just create a step-by-step process to kind of give you a grid for what it is and how you can move the things you've already built in ChatGPT over to Codex. So first, let me start off by saying that Codex is a desktop. There's a couple different ways to use Codex, but for all marketers, I highly recommend just get the Mac app. At the time of this recording, there's no Windows app, but check because, you know, they update that kind of stuff frequently and there could be a Windows app coming soon. But right now it's just Mac. Let me open it up for you. You could see a lot like chat GPT. It's got a question box. You can select your model and you can have some projects over on the left-hand side, but there are a few fundamental differences to how this works. And the two things that I love about it is it can proactively find context and then take action. Each project you make is actually tied to a folder. Yes, a folder. So if you go and open up your desktop, let me open up my desktop here and make a folder called hi. We're just gonna make it basic, right? And if I set up a new project, I'm use an existing folder, I'm gonna go to my desktop and I'm gonna select hi. I now have a project called hi. In codex and for those of you listening along I am now looking at my desktop and I'm gonna be talking through what I'm seeing So if you're you're driving or something you could continue getting a lot out of this and see how easy it is to move to codex But if not, then come and watch this on YouTube later if you need the step-by-step visual tutorial But I have this project now called high and I have a folder on my desktop. Just my desktop called high and If I start a new conversation within this project everything it creates whether it's files or code or you drop images to it, go into this folder. And that's part of where the power is, is you can feed it things and give it reference to things and it can recheck things that all live in this folder. And chat GPT, like if you created a canvas of something that was important to you that you wanted to revisit later, well, you'd have to copy and go back and find that canvas, it in a Google doc and then pull it back into the conversation. But what if check the codec, like essentially, keep tabs on all the documents every single time and leave little notes to itself to go and find this for that or that for this. That's the power of Codex. Yes, it was designed to be able to navigate code, but it's actually good for everything. All the work that I used to do in ChatGPT just happens here in Codex. For example, I'm going to open up my project called Danchez, because that's my personal brand and in here is all the work that I do for my podcast. Even now I'm looking at all the different thumbnails I need to create that are going to be the thumbnail for this episode. You can see it's created a lot of thumbnails based on its image generating tool for me here. So, but let's back up a little bit. So there's this fundamental difference in that it's organized around projects and those projects are folders on your computer. But what if you already have an existing chat GPT account? Well, Let me show you. If you have an existing chat GPT account, you probably are already using the projects. If you're listening to the show, you've heard me talk about the importance of organizing around projects and using custom GPTs. You probably have a mix of both. That's good. We can organize them that way. In fact, I highly recommend trying to think through what are the bigger projects, not small projects, but like big projects that you have in chat GPT. Each one of those can be a project. over in Codex. You can see I've started a few of Trim Healthy Mama marketing, right? Since I'm the marketing director for Trim Healthy Mama, I have Quick Ship, a different company that I work with that's part of that company. And then have my personal brand, which has my podcast and all the other things that I do with my website and so forth. I have them as three large projects. Think broader than what you would do probably with the Chat GPT project, but recreate those projects here. And then shoot, you might even take the instructions for one of those projects, like I have Trim Healthy Mama here and kickstart the new project in Codex with your instructions that you have in your project file within Chat GPT. You might even tell it so in the first conversation. You might say, hey, Codex, I'm transferring a project from Chat GPT over to this project folder. Here's all the information you should know. It's almost like as if you have lost one Chat GPT account and you're starting a new thing. But Codex has all the personalization features, all the memory, except it's just more powerful because you can keep everything within the folder itself. So I highly recommend starting with the base of projects, because that's how Codex operates. It all operates within these folder projects. And then you can take your custom GPTs, which for me were like these little specific skills. If projects were more general, custom GPTs had a one thing that it did over and over and over again. You can take the instructions for all your custom GPTs. This is the first thing I did is I just started handing it over to the project. So now I run my whole podcast out of just the one project called Danchez, and it has documents in there that label the step-by-step instructions it needs to take in order to build out my full episode. Like here, we're running through the process of generating the ideas around this particular episode, and we're walking through my pre-production process. And I've done lots of videos, my pre-production and my production process for the AI-driven marketer, and I essentially just took them all and I built them into one folder within the folder of Danchez called AI driven marketer production. And it has instructions on what we do for pre-production. It has the instructions for what we do for production. In fact, I can open up that folder now and show you, but instead of me having to create the custom GPTs myself, I can transfer them here and then it can actually maintain them itself. And that's a big difference. Let me show you here. have Danches. We have areas. ⁓ better go to projects. Yeah. I a marketer is one and I have all the working systems. for the AI Driver Marketer here. I didn't even create these folders. Codex actually created them and has reorganized them a few times as we've needed to go through. And it just keeps these MD files, which are essentially like, if you're looking at my screen, it has like this markdown file, everything that it needs to know in order to create this part of the production process, which is the AI Driver Marketer thumbnail style guide. And it has all the markdown. It started as instructions, part of the instructions that were in my project for this. but now it's just one set of instructions, an MD file in this particular folder. So it knows every time we walk through the thumbnail making process for this show, it just re references this MD file and then executes the instructions. It's fantastic. Before, if you wanted to change the instructions on ChatGPT, you had to go and do it yourself. But here, I've even done it here in this thread. I said, let's see. I said, I said, you forgot to include my reference photo to capture what I look like. Please update the docs to ensure it happens every time and then regenerate the mails. So it not only generated the thumbnails, but in went and changed the pre-production workflow MD file so that it didn't forget again. It also updated the readme file, which is kind of like a breadcrumb for itself on how to run all the other files. often, the readme files are like a little breadcrumbs for itself of where to pick off every single time. And it is just really, really handy. Cause long, I don't have to worry about updating docs anymore. could be like, ⁓ yeah, make sure to remember this, make sure to remember that. It's actually much more like working with the person. Cause you don't have to update the SOPs. You can have codecs update the SOPs, the standard operating procedures of how you want things done. So I highly recommend starting with projects as the broad part and then putting your custom GPTs in. And at first, just let it be messy. Just let it be messy. It will add a bunch of things in the folder. You can then later go back and be like, hey, we'd been doing a lot of work here. Look through all the things that we've been doing and try to create a system to organize it. But if you're an organization freak like me, you can probably start with this system. I actually start with a system called Para or projects, areas, resources, and archives, which was popularized by a guy named Tiago Forte in order to organize things like your Evernote or your Notion docs. kind of to create that second brain. It's pretty good. It's a good general way to organize ⁓ across a lot of different projects, across a lot of parts of your life. But I organize my things this way. You can see there's a para guide for it, codecs that navigate how it's organized and a few extra docs just to help keep up how it's been organized over time and what changes have been made so it can actually keep it up to date. So. That's a good way to start for anybody. You can actually just go to Codex and be like, hey, we're starting a new project on X. I'd like you to organize it by with the para method from Tiago in order to get this started and then organize it as we go. And then every once in a while, like tell it to be like, Hey, there's a lot going on. Try to reorganize it and leave notes for yourself to keep it neat and tidy. And it could do that work. And that's kind of the best way to get started with Codex. Transfer your projects over from ChatGPT. Think about which custom GPTs belong to which projects. And if there's some custom GPTs that can belong to multiple projects, this is where you want to differentiate between ⁓ like an operational guide within a folder, like those MD files I showed you, and what we call a skill. Skill does not exist in any of the folders. It actually exists in Codex itself. And it's not a file. It's a file that exists somewhere, but I actually don't know where it stores it, but it stores it separately. And a skill is a thing that you can call. anytime across any of the projects. I'll be doing a whole separate video in the future on skills, but just know if there's a custom GPT, you use it and you use it across clients. It's a general tool, but it does a specific function. That's what you're going to want to save for a skill. Just tell Codex what you want it for and it will store it for you, but we will cover more of that in a future video. Hopefully this video was helpful in helping you just kind of get the lay of the land of transferring from chat GPT to Codex. is start using codecs for every. And if you want to dive deeper, I am doing a live workshop where I cover this more in depth. And if you're listening to this, because it's later today, this episode will go live about an hour or two before that that live workshop is going on. If you're watching this or listening to it after the fact, that's okay. Go to the same URL. You can get the live recording, which will be published probably a day or two after the live event has taken place. So just go to danchez.com slash codex where I'm going to be doing a full live walkthrough of how I'm using it, some of the cooler things that I've been doing with it and further step-by-step instructions of how to organize it more. And again, I'll be doing future videos on codex because this is now the main base that I'm doing all my AI work in or almost all my AI work other than some other ancillary tools. This has become my go-to for almost everything. And you'll see why. once we start using this and I start showing you more about it in the future. But for now, stop using chat GPT, transfer your things over, start using Todex.