This year can be said to be the year when the application of intelligent agents explodes. From conversational assistants that can only chat, to digital employees who can independently run code, make PPT, and even edit videos in the background, the development speed of AI is indeed ridiculously fast. In this context where everyone is developing intelligent agents, OpenAI once again dropped a blockbuster.
At the Intelligence at Work live event last night, OpenAI officially announced a major decision:
They plan to integrate Codex's core capabilities directly into the ChatGPT application in the next few weeks, and the existing Codex will still exist as an independent programming tool.
As soon as this news came out, many foreign netizens were discussing: Why did OpenAI put this killer product into ChatGPT?
In fact, the reason is very simple.
In the enterprise market, most companies only know ChatGPT, but are unfamiliar with Codex, which is highly developer-oriented and requires a special entrance. Many companies don't even know in what scenarios they should use it.
Seamlessly integrating Codex's Agent execution capabilities into ChatGPT allows users to directly call these capabilities in a familiar chat box. Not only that, OpenAI has also launched six major industry plug-ins, including sales, data analysis, and creative production, directly turning complex technical tools into daily helpers for small and medium-sized enterprises.
I have to say that Codex’s previous performance on the Mac and Windows versions was extremely amazing.
Just last month, OpenAI has implemented the function of remotely controlling the computer Codex through the ChatGPT mobile terminal. The realization of this cross-terminal linkage means that users can have their home computer automatically run code, change files, and even build a lightweight web page by sending a command on their mobile phone.
This further integration means that this cross-platform super Agent experience for computer control will be completely and seamlessly spread to everyone.
So, will other big model manufacturers follow suit?
The answer is yes.
At present, Microsoft has deeply implanted various Copilot agents into Windows systems. In fact, domestic companies such as DingTalk and Feishu are also accelerating the integration of the underlying execution capabilities of large models into a unified user portal.
After all, users’ energy is limited, and no one wants to install dozens of AI tools for different scenarios on their mobile phones.
Combining the powerful underlying capabilities of large models with the Agent of specific scenarios with one click, leaving the complexity to the system and the simplicity to the user, this is the general trend of AI applications in the future.
Ordinary people do not need to understand what code or development tools are in the future. You only need to express your ideas in plain English in ChatGPT, and the rest of the dirty work is left to the underlying Codex to perform it.
This storm of popularization of Agent applications is indeed coming faster than we imagined.