In my personal view, this issue is purely related to the web interface interaction and has nothing to do with the underlying model.
This implies that the server and model functions are working correctly, but a low-level bug has emerged in the user’s web interface interaction. This web interaction problem is preventing the conversation from continuing and blocking the saving of new dialogue segments within the existing conversation.
I believe there are two possibilities now:
The issue is so basic that it’s embarrassing to admit.
The problem is so complex that even employees at a star company couldn’t reproduce or fix it within half a month.
I’m facing this issue too, I’m now using the [alt + enter] method plus using branch chats as the save point then deleting the new branched chat as a method to keep the broken chat going since it has valuable info. Ive also moved all the previous chats in the google drive folder so the new branch chats are the only things populating it and the only things to delete in it. Feels so backwards but GOOGLE doesn’t want to fix their premier AI experience website.
it also seems that the system is either failing to count tokens properly or being unable to. because one of my chats used to be at a stable 800k tokens but ever since the bug affected it, whenever i scroll up and down and the process “finishes” i get varying counts between 600-780k tokens used. I was wondering if they scrubbed spare tokens at the backend and cut out a lot of information i’d sent over to it, but no, when I was able to chat with the bot it still remembered some fair few details sent to it over the course of the chat.
so i’m not sure why the token calculation is so unstable
Thank you for your patience while we looked into this issue. We’ve rolled out a fix and you should now be able to continue submitting prompts even in longer sessions.