Codex adds Images 2.0, Windows sandbox

// 26d agoPRODUCT UPDATE

Codex adds Images 2.0, Windows sandbox

OpenAI updates Codex with native Images 2.0 support and a secure Windows sandbox for "agentic" design workflows. The release enables autonomous agents to generate UI mockups and execute shell commands safely within isolated environments.

// ANALYSIS

The evolution of Codex into an agentic orchestrator signals a move beyond simple code completion toward autonomous project execution.

–Native Images 2.0 support allows Codex to generate context-aware visual assets directly from code structures and CSS definitions.
–The new Windows sandbox uses Restricted Tokens and Synthetic SIDs to grant agents secure, isolated shell access for file operations.
–Integration of "Skills" lets developers modularize agentic tasks, connecting the IDE directly to Figma, Linear, and CI/CD pipelines.
–Multi-agent orchestration with isolated Git worktrees solves the coordination problem for complex, multi-file refactoring tasks.

// TAGS

openai-codexai-codingagentimage-gendevtoolmultimodalinfrastructure

DISCOVERED

26d ago

2026-05-14

PUBLISHED

26d ago

2026-05-14

RELEVANCE

9/ 10

AUTHOR

OpenAI

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE26m ago

v0 converts prompt secrets to environment variables

Vercel has introduced a new security feature to its AI-powered UI generation tool, v0, which automatically detects sensitive information like API keys or secrets in user prompts. Instead of hardcoding them into the generated application code, v0 extracts them and securely converts them into environment variables, preventing accidental credentials leakage and streamlining development.

UPDATE27m ago

StackBlitz drops free domains for Bolt Pro

StackBlitz's browser-based AI development agent, Bolt.new, is offering a free custom domain to all users on its Pro tier. The perk allows builders to connect custom domains directly from the browser to turn AI-generated prototypes into branded web applications, with claims open through June 12th.

UPDATE28m ago

Pika MCP adds video language swap

Pika has released a new "Language Swap" skill through its Model Context Protocol (MCP) server, allowing users to translate spoken dialogue in videos while preserving the speaker's original voice. This capability streamlines localization and multi-modal content creation by bringing advanced video and voice editing tools directly into conversational AI assistants.