CHATGPT does a pretty good thing to create and classify code in the text prompt. Now Openai has added a new coding agent to Chatgpt and is not just a programmer who needs to be excited.

CHATGPT’s CODEX takes AI programming to the next step.

Openai begins to preview Codex, Codex, “Cloud -based Software Engineering Agent.” This feature runs with Codex-1, a version of the Openai O3 model optimized for coding and software engineering. The CODEX-1 is also trained to close the output with the “Human Coding Preferences and Standards”.

If you are a Chatgpt Pro, Enterprise or TEAM user, you can find the feature in the Chatgpt side bar. After opening, you can enter the prompt and input to allocate the coding work. password Use the button or ask about the codebase ask button. You can find information about Codex’s work list and progress under the prompt bar.

This new agent can perform multiple tasks, including adding new features, correcting bugs, and answering questions in the existing codebase. Each work runs in a separate isolated environment and pre -loads it to a codebase or repository. Codex can also read and edit files.

Openai presentation It claims that the agent will take between 1 and 30 minutes to complete the work assigned according to the complexity of the work. As usual, while using a browser and computer, you can monitor progress in real time in real time or run multiple tasks at the same time.

The CHATGPT helps you create a full project that can create, download and test your code, but it doesn’t work well for software repository and codebase. Codex’s ability in a typical software engineering infrastructure means that it is much more useful than Vanilla Chatgpt for companies and individuals that maintain multiple projects in repository.

Codex generates a cleaner code compared to CHATGPT, which can be integrated into human review and workflow or codebase. It also runs the test until you pass all the test cases and conditions given. When the work is completed, Codex commits changes to the environment and provides “verified evidence of behavior through” terminal log and test output “.

Why is Codex a big problem?

Codex is a big problem for all industries. You can write Excel macros, automate reports, place files, and do everything you need expertise in some programming or scripting languages.

Of course, CHATGPT can create code and scripts for you, but in my experience, I can’t be reliable. You need a relatively good knowledge about the general ideas of programming languages ​​and code debugging you work. However, Codex automatically checks the code and runs the test to check if it works in the desired way.

This can be improved CHATGPT’s new GPT-4.1 modelBut it’s not a perfect solution. knowing Which CHATGPT model should be used Because it can greatly affect the output, the customized model for coding is better than the general purpose model.

Of course, in the case of a programmer, Codex is integrated with the Github repository to help you manage repetitive tasks and test cases. This allows you to develop and deliver the app faster, without falling into some of the software development processes that maintenance, testing and other tasks are generally part of the software development process.