NEW ChatGPT Codex Update 🤯

Breaking news. Chat GPT just dropped a huge coding update and this is called Codeex. They've just launched Codeex, a cloud-based software engineering beast that handles multiple coding tasks simultaneously. This software engineer can connect Codeex to your GitHub repositories for automatic bug fixes and improvements. It can use multiple AI coding agents working on different parts of your project simultaneously. and I'll show you how OpenAI's own engineers are using this daily and how you can leverage the same techniques. Chat GPT have just launched a new update launching a research preview of codeex which is basically like a cloud-based software engineering agent that can work on tons of different tasks in parallel. So you can get access to it via chatgbt. com/codeex is available for chatgpt pro right now but they're planning to release it to plus users as well. If you want to see an example of

how it works, let's pull this up and I'll show you how it works step by step. So this is the video from Chat GPT on exactly how it works. So you would just go to codeex in the top left as you can see. And then this is what it looks like right now. Basically what this is designed to do is act as a software engineer that can then go off and fix bugs for you, work on different projects, etc. Right? So you can see here for example codeex agents build features answer codebased questions run tests and draft PRs asynchronously right so you can have multiple different agents working on a software problem or working on different parts of the software as you go along it's pretty cool I really like the UI this has just come to chat cheap it is I said and this basically an autonomous coding agent now

in terms of what it can do so it can for example like draft GitHub PRs in parallel it can navigate your code base really good for finding bugs reviewing code and suggesting improvements. It can run lint and test. By the way, if you're not sure what any of this means or it sounds technical, don't worry. I'll break it down for you in a minute. This is really interesting as well. So, it's fine tuned to work in large code bases. Now, if you're wondering, okay, how does that work or how can you use it, etc. So, you can see on open router, they've just released this literally just came out May the 16th, OpenAI codeex mini. And essentially Codeex Mini latest is a fine-tuned version of 04 mini which is one of the most powerful models specifically for use in codec cli for direct use in the API. We recommend using GPT4. 1. You can actually go into

the chat section over here and use it is available in chat GPT pro. You can see what apps are currently using this as well. And so you can use the API version or direct chat version and it looks exactly like this. So you can connect it directly to GitHub as well, which is pretty cool. And then this is what it's going to look like. So you're going to install an author chat GPT connector. And this is basically linking GitHub directly to chat GPT connector as you can see. And then once you're in there, you can create your starting environment. So you can basically pick a repo from your GitHub account and you can customize the environment later, take the repository, and then edit and improve different parts of your code. So for example, here's some examples of a few tasks to get started. So you can ask it to explain the codebase to a newcomer. For example, like what is the general structure, etc. You can go through your codebase and find issues and propose like one task to fix a typo, bug, etc. And that's what they mean by asynchronous agents because you can have multiple different software engineering agents all working on different tasks at the same time. Typically, it takes I think around 2 minutes to 30 minutes to generate each response. But this is basically what it looks like. I really like the UI as well. It looks really nice. You can see different tasks generating down here. And then you can say pick a part of the codebase that seems important and find a bug. It will download the repo. And then it will set up a separate environment here. Picking apart the codebase and then fixing. So it's reviewing the instructions in agents. mmd to guide my edits and will inspect directories. I'm diving into the SRC for/ agents directory to see what's inside. It's going off working its magic, etc. So, that's pretty much it in

terms of the codebase, how it works, etc. You can see this was just announced as they said we're launching a research preview of codeex and it's rolling out to pro, enterprise, and team users in chat GPT starting today and will be coming out to plus users soon. Now, if you're wondering what it looks like, etc., here's an example. This is the official openai. com article on it. And basically what this is supposed to be is like a cloud-based engineering agent.

Now this is really interesting because if you haven't seen it already, Windsurf, I did a video about it yesterday, but Windsurf have just launched a new software engineering model called SWE. Okay, and SWE seems very similar to this. But the difference here, right, if we go to Windsurf Wave 9, it's interesting here because Windsurf are supposed to be getting acquired by Chat GPT, but in the same week, they both released a model for software engineering. Now, either that's a coincidence or they're going up against each other. I can't figure out, but very interesting times right now. As you can see right here, they said Codeex is powered by Codeex 1, a version of OpenAI 03 optimized for software engineering. It was trained using reinforcement learning and real world coding tasks in a variety of environments to generate code that closely mirrors human style and PR preferences. Here's an example what it looks like, how it works, etc. Really nice clean UI as you can see. And then if you want to see some benchmarks, here you go. So it's comparing codeex one versus 03 high. And the benchmarks are very similar over here. OpenAI internal SWE tasks. Here's the benchmarks. And they've also got an interesting part on building safe and trustworthy agents and they've compared codeex responses versus open 3 responses directly. Now also they've talked about use cases right so one of the biggest use cases here is like technical teams openai has started using codeex as part of their daily toolkit. So is used by open air engineers to offload repetitive tasks, scope tasks like refactoring, renaming, writing tests, fixing bugs, writing components, drafting documentation. So it's quite interesting to see this. You basically got like an agentic software engineer in the cloud and even if you can't code, you can still improve your code and debug it as well. They've also got this guide for

codecs in chatgbt and you can allow yourself to toggle this on, right? So, if you go to chatgbt. com/s settings, go into your settings. Obviously, you need to be on the pro plan and then you can toggle codeex on like you can see right here. And everything is inside a self-contained network isolated container, right? Which means that it's designed for safety and also to avoid any sort of issues. Also, they've said they've also answered the FAQ, which is do you train your models based on my usage of codecs? and they've said for chat GPT team enterprise and edure users we do not train models on your codeex content for users on other plans model training depends on whether you've allowed that to happen or not right so if you want to protect your code or you don't want to share anything or don't want to train chat cheapy then you can change the settings as you can see right there now let's break this down in terms of what it means how it works etc in simple language so it's basically a cloud-based AI agent that can handle many coding tasks at once multitasking for software development, right? It's powered by codeex1 which is the new API from openi just got released as you can see right here. It's basically like GPT4 technology but fine-tuned for coding. Now you might be thinking okay why is this interesting? How does this change things etc. It's basically a major step forward for AI assisted coding. So you and your agency can like automate coding tasks, speed up development etc. And also they've got this completely new codeex interface that lets you interface directly and interact directly with the entire code repository. So you can ask it for to find bugs and recent changes, fix them automatically, etc. Now you might be wondering how long does it take? So each codeex task takes between 1 and 30 minutes to complete task with realtime progress tracking. Provides verifiable log so you can see exactly what it did and then also it can follow guidance from your file. So if you set up a agent MD, this will basically be like a readme for the AI explaining exactly how to work with your repository and it looks way cleaner, right? And also improves the quality of your code as well. Now in terms of real world applications use cases, so basically you've got like on call engineers, you can use it to quickly understand fix bugs from error logs, helps tune alerting systems, acts like a personal software engineering assistant and allows like multiple different software engineers to start multiple tasks at once and also it helps scaffold new projects automatically from natural language commands. So if you want to get access to it, it's available in pro, you can link it to GitHub, etc. So thanks

for watching. If you want to get access to my community that tells you how to make money and scale with AI, feel free to get that inside the air profit boardroom. Comes with coaching support community. You can ask any questions inside there. You can jump on weekly Q& A calls and you get access to all of our best automations like so. If you want to get the notes from today, feel free to get that inside the AI success lab. Link in the comments description. All the notes right here, plus details on my new NA10 scraper agent plus Gen Spark updates as well. This community is full of 5,700 members completely free and all the videos that I showed recently. You can get access to the notes inside the AIS success lab. And then if you want to get more support, if you want to upgrade, etc., then I would recommend checking out the AI profit boardroom. Thanks so much for watching. Appreciate it. See you on the next one. Cheers.