Hello! Sorry if this isn’t the right place for this feedback, but I couldn’t find a better category or forum.
I’ve noticed significant performance degradation on the ChatGPT website (link removed as I can’t post links - sorry about that!) whenever ChatGPT conversations become very lengthy, which appears directly related to the entire conversation (from the very first message onward) always being loaded and visible simultaneously.
After discussing potential causes with ChatGPT itself to see if there’s anything on my end that’s the culprit, it seems this slowdown is largely due to the browser managing an extensive number of DOM elements and substantial text content. This negatively impacts user engagement, causing frustration from delays, necessitating frequent reloads, and sometimes leading users to abandon valuable, extended conversations entirely.
I would like to see a dynamic loading and unloading system, similar to the approach many social media and chat applications already employ successfully (like with Discord, Facebook, X, etc). Specifically, older portions of chat history could be unloaded as users continue to chat, and newer parts of the convo could be unloaded as users navigate to earlier parts of the conversation.
I feel this optimization would dramatically improve page responsiveness and user experience, potentially reduce network strain and associated errors from repeated reloads (though I could see constantly re-polling the servers for chat history could be an issue, but I wonder if that would still be reduced since it’s many small requests instead of one big one), and would provide a user experience for ChatGPT comperable to many chat apps.
I would imagine with this mainly impacting the front-end, it shouldn’t hurt the back end so nothing should break, either; I genuinely think this improvement would greatly enhance both UX and system efficiency.
I love this platform already, and hopefully you can consider this feedback worthwhile to implement! Thank you so much for your time!