OpenAI's New ChatGPT UI Criticized as Cluttered

koltregaskes · x · 2026-07-11

The author argues that the new desktop and web experience for OpenAI's ChatGPT/Codex is "more powerful but messier," offering several improvement suggestions:

Unclear Modes & Tiers: OpenAI uses Chat, Work, and Codex for chatting, long tasks/deliverables, and software development, but they act more like different views. The author suggests simplifying to a "Simple / Advanced / Developer" tier system for instant clarity.
Inconsistent Terminology: Web and desktop use different terms for "branching to a new conversation/continuing a task." The desktop app also simultaneously features Chat and "Open side task," causing confusion.
Scattered Features: The new desktop app accesses some chat history, but Projects, Scheduled Tasks, Images, GPTs, Library, and Pins remain scattered without a unified hub.
Unify Extensions: The author recommends merging Agents, Skills, and Plugins into a single "Extensions" menu and improving filtering and sorting in the plugin directory.
Overly Complex Model Selector: Simple mode uses unintuitive names like Sol, Tera, and Luna, which confuses average users. Advanced mode combines multiple models, effort levels, and speed settings, making it overwhelmingly complex.
Missing Voice Mode: Standard web ChatGPT supports full voice, but the new Work and desktop apps do not. The author argues voice should be universally available across Chat, Work, and Codex.
Performance Issues: The article concludes by noting that the new desktop app still suffers from slow load times.

Original post →

OpenAI's New ChatGPT UI Criticized as Cluttered

More from Apps