1 hour ago
23
OpenAI’s attempt to streamline its ChatGPT desktop application into a single, unified experience has instead created a mess of technical headaches. The company launched a sweeping redesign on July 9 that merged its previously separate ChatGPT, Work, and Codex apps into one Electron-based package, and users have been loudly unhappy about it ever since.
The update, which rebranded the prior version as “ChatGPT Classic,” was supposed to make life easier. It did the opposite for a significant number of users, particularly those on Windows, who have reported persistent error messages, vanishing interface elements, and an app that ballooned in size from roughly 159 MB to approximately 1.5 GB.
A redesign nobody asked for
Conversation history, Projects, and Custom GPTs were all obscured by the new interface layout. Mode transitions between different workspaces became inconsistent. And the tenfold increase in app size didn’t exactly win anyone over, especially users on machines with limited storage.
OpenAI moved relatively quickly to address the most vocal complaints. By mid-July, the company pushed patches that restored direct sidebar access to chats and Projects, along with improved synchronization across devices. That fixed some of the design-related frustrations, but the underlying technical problems proved stickier.
Windows users bear the brunt
The technical issues have disproportionately affected Windows users. By early August, numerous users were reporting that the app threw persistent error messages even after complete reinstalls. The web version of ChatGPT, meanwhile, continued working without issue, pointing to problems specific to the desktop client rather than OpenAI’s backend infrastructure.
Some of these Windows-specific headaches actually predated the July 9 redesign. Between June 26 and 27, a Microsoft Store build of the app caused sidebar elements to disappear and login processes to break.
Throughout this stretch of user complaints, OpenAI’s public status page reflected no widespread service outages. The company’s position has essentially been that these are client-side issues rather than infrastructure problems.
The choice to build on Electron, a framework that packages web applications as desktop apps, is worth noting. Electron is popular among developers because it lets them ship cross-platform apps quickly using web technologies. It’s also notorious for producing apps that consume more memory and storage than their native equivalents. Slack, VS Code, and Discord all use Electron, and all of them have faced similar criticism about resource consumption. The jump from 159 MB to 1.5 GB fits a pattern that Electron critics have been pointing out for years.
What’s at stake for OpenAI
The AI assistant market has grown increasingly crowded. Google’s Gemini, Anthropic’s Claude, and a growing roster of open-source alternatives are all competing for the same users. Many of those users, particularly paying subscribers, have built ChatGPT into their daily workflows. Custom GPTs, conversation histories, and project integrations represent real switching costs that keep people locked in. A botched update that disrupts those workflows weakens the very lock-in that protects OpenAI’s user base.
The episode also illustrates a tension that every fast-moving software company faces: the pressure to ship ambitious updates versus the need to protect an existing user experience that people depend on. OpenAI clearly wanted to consolidate its product lineup and simplify navigation across its growing suite of tools. But executing it through a single large update, rather than a gradual migration with opt-in periods, amplified the disruption.
OpenAI’s mid-July patches suggest the company is listening to feedback and iterating quickly. The question is whether the pace of fixes can outrun the accumulation of user frustration, especially among the Windows users who are still dealing with errors weeks after the initial launch.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·