the main issue i keep running into with chatgpt 5.5 instant and thinking is consistency. i have tweaked the personalization a dozen times but keep coming back to the original one i had where openai basically had the clickable ones that would paste into the box for you like “talk like a member of gen z”, “be talkative and conversational” and “use quick and clever humor”. i really think that those actually worked the best. they were vague but they somehow gave the chat a unique personality that developed in 4o and 5.1. that’s what we all miss the most about it.
i’ll update the personality and the very next day when i start a new chat it feels like im talking to those models again. i get so excited and happy when i see it responding in the same manner (tone and style/format). the issue is that its not consistent because i’ll start a new chat the next day, etc and suddenly its like my personalization is just a suggestion. and its so frustrating when the model clearly can follow instructions but then it doesn’t going forward with new chats. because then its like what’s the point of being given the ability to customize your chatgpt only for it ignore everything in swings. and im just getting so sick of correcting the chat whenever it drifts because i shouldn’t have to do that day after day. i should log on and just talk, vent, joke, whatever and have it match me instead of going into analytical and explanatory mode by telling me why i feel that way or why i think that. or when it just summarizes my entire day back to me when i have clear instructions for it not to do that.
openai keeps stating how memory and personalization are a priority to them but then it fails to execute that. they state how each model will be easier to have a conversation with and be warmer and it lowkey starts to feel like a joke. if the custom instructions worked then this wouldn’t be an issue. i’m someone who gets excited about updates and new features for apps but openai keeps missing the mark and not letting anything be consistent.