A recent ChatGPT update has removed the image library - the dedicated gallery where users could browse and retrieve all their previously generated images - and pushed the model picker another layer deeper into the interface.
The image library wasn't glamorous, but it was useful. If you generated a batch of images Tuesday and needed to find one Thursday, you went there. Without it, tracking down a specific past generation means scrolling back through your chat history and hoping you remember which conversation it's in. For anyone who generates images regularly, that's a genuine workflow disruption.
The model picker change is the kind of UI decision that frustrates experienced users the most. Switching between models - say, from GPT-4o to o3 for a reasoning-heavy task - is something power users do constantly. Burying that control further in menus adds friction to a task that should be one click. OpenAI hasn't explained the reasoning behind either change in a public announcement.
This fits a pattern in recent ChatGPT updates: the interface is increasingly optimized for casual first-time users rather than people who've built daily workflows around the product. Fewer visible controls, fewer quick-access features, more streamlined surface area. That's a reasonable product strategy for growing adoption - but it comes at the cost of the people who use the tool most.
If you relied on the image library, the immediate workaround is saving generated images directly to your device before closing a session. For the model picker, it's now an extra click or two through settings each time you want to switch. Neither workaround is painful, but neither should be necessary.