ChatGPT Desktop Absorbs Codex, Confuses Users

The unification makes strategic sense—one desktop surface for chat, delegated work, and coding—but the information architecture makes each mode feel like a separate product with overlapping boundaries. For developers, the biggest risk is losing confidence about which context, history, permissions, and usage pool a task belongs to.

• –OpenAI defines Chat as quick help, Work as multi-step deliverables, and Codex as repository-centered coding; those distinctions remain easy to blur.
• –Separate histories and local-versus-cloud execution make portability and discoverability harder.
• –Shared plugins and agentic usage pools increase reuse but make permissions and resource consumption less obvious.
• –The rollout feels more like a product architecture migration than a polished desktop redesign.
• –For AI developers, this signals Codex becoming a platform layer inside ChatGPT rather than merely a coding client.