OpenAI Unveils Codex Update: Multitasking Agents and Image Generation for Developers Now Live
The company OpenAI has introduced an update for the Codex tool, which expands its functionality and demonstrates an approach to creating a future desktop application. Earlier, it was reported about the development of a solution that should combine ChatGPT, Codex, and the Atlas browser, but it has not been presented yet.
What is known
According to the head of Codex, Thibault Southeau, the current release is aimed at developers. It features multifunctional agents that can perform tasks in different environments and interact with applications on the computer. The user can specify a specific application or allow the system to select it independently for performing the task.
Particular attention is paid to performance — OpenAI claims its optimizations allow agents to work in parallel with the user without overloading the system. In addition, the company added 111 new plugins that combine integrations with applications, development tools, and server connections to expand the context.
Codex also received an integrated browser with a commenting function, allowing changes to projects directly through prompts. Additionally, image generation based on the gpt-image-1.5 model is integrated — it can create layouts, interface designs, and graphic elements.
Among the innovations are memory functions: the system remembers previous actions and suggests proactive steps, such as reminders about comments in documents. The updated Codex has already started being implemented for desktop ChatGPT users, initially on macOS, with further launch in Europe and the UK.
Source: OpenAI
