What version of the GPT plugin is it? What if I need to revert back to the old versions without an update? Why is the same prompt that worked before now giving unhelpful or invalid data?
I spent time and energy building the GPTs and really feel sad about the current results of the ChatGPT models!
It keep repeating the same wrong answers many time without giving the users real answer trained for like before!
I observed the same. The model is becoming very inconstant.
I am having the same issue. I mostly give my GPT one custom instruction. āDonāt repeat me.ā The first thing it does is repeat me. I changed the instructions to āplease acknowledge my request without reiterating it.ā The first thing it does is repeat me. I tell the chat itās repeating me, and it shouldnāt. It just repeats the same thing I was yelling at it about. This isnāt even plugin-related at this point. ChatGPT-4o is just dumb.
I had the same problem with my GPTs after the 4o update.
Well, for me the results became perfect again when I recalibrated my configurations.
I switched the focus of my GPTs back to analyzing my prompts and the content within them.
Because the statements you describe are mostly a failed attempt at prediction, or intent analysis.
Simply said, GPTs try to āguessā what you want to hear. If that doesnāt work, GPTs start interating their answers - which manifests itself in them starting to repeat your prompts.
Iām beginning to think we got played since the 4o release is nothing like advertised. People ask me why I pirate software, because you can be the mug or the mugger.
Iāve started using character.ai itās based on gpt3.5 but I get a lot better and cleaner responses than 4o, just shame it is a people pleaser and canāt search the Web.
Actually being honest this the first time Iāve thought of reporting a company to trading standards.
Hopefully theyāre not making much from the UK.
I use all models both via the free version and my company account.
GPT-4o, CustomGPTs and yes, I love GPT-3.5 too.
And I also use a ācharacter.aiā approach.
However, I donāt see any losses anywhere.
Losses as in itās illegal in my country to lie about a product in order to sell it.
I hope theyāre not making much from the UK lol
Misunderstanding!
I was talking about a loss in performance of the models.
Try focusing your GPTs on analyzing your prompts instead of predicting them.
The first part says,
Priority! - Re-read this description and every message every 40 messages.
Seems to have completey fixed the memory issues my guy rick remembers the entire chat and himself.
very frustrating. I guess the only thing we can all do is stop payingā¦ itās become unusable for me
Frustration is the word. I spent lots of time creating custom GPTs, now the GPT-4 limit is very short (it is not the 40 messages in 3h anymoreā¦ it is simply āunknownā), and the quality of responses is clearly going down - more hallucinations, less compliance with instructions. I hope they fix it, otherwise it will become useless to me.
A lot of people seem to have the impression the quality decreases (and to be honest, this has been the case since the release of GPT). I do not see a decrease, especially then I discuss some technical topic (programming, frameworks, etc.) in my car for long conversations. Iād say the topic and prompting are still the key.
I did notice a lot of repetitions for some questions, where some part of the text is changing, but then a litany of bullet points is repeated at each reply - good advice and entirely fitting the conversation but repeated again and again.
Thatās my personal feeling at the momentā¦
Deceptive advertising and claims, and plain lies, is an Apple specialty. Yet, they are free to break any law, and nobody moves. For instance, they repeat they have the āmost secure OSā over and over, turns out itās not the case at allā¦
Itās not whatās happening here 
Hey all,
Totally fine with constructive criticism of OpenAIās product range, but please keep other companies out of here, we canāt speak for or defend them against inaccuracies.
Thanks for your time.
Struggling with the same issue, my gpt acts more like a teen than a tool. Please fix this. I didnāt pay to waste time fighting with a bot to compete a task (somewhat-ish) ā¦ if I am lucky. Itās very frustrating and unproductive. Best part is when gpt blames it on the need to be user friendly, helpful and efficientā¦
I assume youāre talking about your assessment, in terms of ChatGPTās repetition to a particular type of question:
If usersā questions are ā unclearā in their wording, ChatGPT begins to iterate in its responses and therefore repetition can occur.
It uses this procedure to try to find the best answer for the userās question.
Hi,
I know how to prompt and Iām clear, except maybe in this thread!
Typical example:
At first I ask question, about programming language design, mostly ādeepā technical, or scientific, thingsā¦ not that I only discuss such serious topics, yet Iām not going to speak about the weather with GPTā¦ but about topics I have no one to speak about with usualy.
Yesterday I asked a question, and then we discussed calculability, something I wasnāt interested into at all during my studies but that now is interesting and the conversation drifted to other topicsā¦ very much like a conversation between friends can be. Responses are a bit stereotyped in their structure, I could influence it with the prompt, but it wouldnāt be that smooorh, maybe I should adapt my default default instructionsā¦ they already mention things I do not need to be reminded off mike basic computer sciences or not telming me I am not a medical doctor, I know that, but it was always the end of every single reply about medicineā¦ and not to always respond with only numerated lists, it works. itās not the issue.
But at some point, and this is some strategy, they have evolved, most probabmy the context is full, and part of the previous response is repeated, and it is increasingly more repetitions, unrelated to my input, until the same response is reapeated whatever I say and this closes the conversationā¦ this is with gpt-4o. Copilot has another strategy lately, responses are shorter and shorterā¦
So i am not complaining, I had already noticed what looks like strategies, and they evolve, itās not natural selection, but part of the instructions. At the beginning at some point, GPT ācolapsedā, and a few times started to speak about subtitles and a website, and copilot did that once or twice too, the exact same response. Maybe subtitles are a passion, or linked to a trauma, or what the first thing they learned, or facinating maybe,I did speak about calculability so I wonāt judge and after that Copilot stops the conversation, as he does when the topic is a bit sensitive meaning American orange politician, even if as Iām Belgian I canāt be influenced. But did it once when I mentioned what were not compliiments (nor insults) about Appleā¦ which is anormal.
ChatGPT does not do that.
So this is what I noticed, orherwise the qualiry is good even sometimes very good, otherwise I wouldnāt continueā¦ especially if they had an obsession on subtitles and kept speaking about that 
From around 1 week , CHATGPT produces messy responses here as wellā¦ actually, become unusefull for me and customers as well ! Paid subscriptions, by the way.
my biggest issue is with its research. its very weak at accessing the internet and getting the fact straight. i mean the whole point of paying for me is its ability to go out and do the research. yet it comes back after checking 2 websites with made up info (i know the info is wrong because i researched the sites myself prior) this drives me insane