Models & Releases

Codex for (almost) everything

The new Codex release transforms from a code assistant into a multi-modal AI agent platform.

Deep Dive

The Codex application, a popular AI-powered development tool, has launched a significant update that fundamentally expands its role in the developer's toolkit. Previously focused primarily on code generation and completion, the new version introduces computer use capabilities, allowing the AI to act as an agent that can control applications and perform tasks on the user's computer. This is paired with integrated in-app browsing, enabling the AI to search for information, read documentation, and interact with web services without leaving the Codex environment. The addition of image generation via integrated AI models and a persistent memory system allows for more contextual and creative assistance across a wider range of projects.

This evolution positions Codex as a central hub for AI-assisted work. The new plugin architecture lets developers extend functionality and connect to other tools and APIs, while the combination of agentic behavior, web access, and memory aims to create a seamless, continuous workflow. Instead of switching between a code editor, a browser, a design tool, and a terminal, developers can potentially orchestrate complex tasks through natural language commands to Codex. The update signals a shift from tools that merely suggest code to intelligent systems that can actively participate in and execute multi-step processes, significantly accelerating development cycles and reducing context-switching overhead.

Key Points
  • Adds 'computer use' agent capabilities, allowing the AI to control apps and perform tasks on the user's machine.
  • Integrates in-app browsing and image generation, moving beyond pure code to handle research and asset creation.
  • Introduces a memory system and plugin support to create a persistent, extensible AI assistant for full-stack workflows.

Why It Matters

It transforms developers from coders into orchestrators of AI agents, potentially automating large parts of the software development lifecycle.