OpenAI, Codex Superapp에서 웹 검색 기능 테스트

OpenAI is laying the groundwork for a major Codex update that would push the platform well beyond its origins as a coding agent. Hidden references in the current Codex client reveal a new onboarding flow that will ask users to choose between a basic setup and a professional, developer-focused configuration — a clear signal that OpenAI intends to serve two distinct audiences from a single surface. This aligns with the company's publicly stated plan, led by applications chief Fidji Simo, to add agentic capabilities to Codex that go beyond programming before eventually merging ChatGPT and the Atlas browser into one desktop super app. Among the unreleased additions is a dedicated pull request management section in the navigation menu, a frontend UI preview panel that lets users see what they are building in real time, and the ability to leave inline comments on those previews to direct exactly which changes should be made. Combined with the previously spotted Scratchpad, a to-do list interface that can launch multiple Codex tasks in parallel, these tools would turn Codex into a full-cycle development environment where planning, building, reviewing, and shipping all happen in one place. The timing matters. OpenAI has been under pressure after Anthropic gained strong momentum with Claude Code and Cowork, and Simo told employees the company could not afford to be distracted by side projects. Broadening Codex to handle non-developer tasks, normal chat is already possible through the app, positions it as the foundation for OpenAI's super app strategy, gradually absorbing ChatGPT functionality so that everything lives in one place. The release is expected this week, making it one to watch for anyone tracking how the AI platform war is shifting from standalone tools toward unified workspaces.