I spent hours this morning as a free user on my phone. It was working absolutely fantastic. Working on a novel, and it was helping guide me through details to think about. Said it would store all the information. Then I became a Plus user and it went downhill from there. Lost most of my information. GIving me bad direction using the desktop UI. It needs to better communicate the user experience. I don’t mind some glitches as AI rolls out, but the devs need to better ensure information is not deceptive or lost. If this continues I will cancel and become a free user again.
Note: I want to amend my earlier comment. In further back and forth with ChatGPT I found the ‘Canvas’ feature in the toolbar under the prompt window. When I selected that suddenly the world opened up and the UI reconfigured to separate windows. Some of the previous ‘lost’ information then appeared and I was able to continue on. It remains to be seen how it goes, but right now I am more hopeful this will work out very well.
I am about to do the same and cancel. Up until about 2 weeks ago, Chatgpt was doing a sterling effort, helping make sense of a complex case and drafting legal letters that I was having fact-checked by humans. Now, though and before anyone jumps on with the it is a bot. It is deliberately not helping and is doing the exact opposite of the prompts it is being given. Deleted History, checked personalisation, am now using the anonymous convo, and it is just about ok ish. I totally agree with your post.
Yeah, I ibelieve people, for some reason, are eager to blame the user and defend the ChatGPT bot. It is not like I do not know how to use. I use it for more than one year, paid for a long period and it is really clear that it is not working the same way. It is worst, and I believe if people start being critic about it, it can get better. But is is internet right? You have to love or hate everything…
PS: Not saying this to anyone in this post. The reason I did the post and reply people here is because I feel we can have discussions. I really liked the suggestions and hearing about people experiences. In the end I just want to use as efificienty as possible.
Oh I hear you mate, I’ve tried posting my complaints and I get people saying I should do this or that. The problem is it’s not user-friendly like was promised. The content policy is more extreme than terms of service indicates and isn’t even consistent, I mean I can regen the same post change nothing and have it constantly switch between being fine and refusing. It’s actually refused mention of a hug before. I have used it for over 2 years. I got more with a vague prompt as a free user over 2 years ago than I get with a detailed prompt on the paid version now. That’s not right.
I’ve had the same problem. I was using it to make a mock-up game page for a FNAF Fan-Game and made phone calls for the game (just for funsies). I asked it to make a change to a certain night’s phone call but it instead makes changes to the game page.. What??
Bro, where did I mention I wanted it in the game page? I asked for a refine of a phone call! I HAD TO BE CLEAR 3 TIMES IN ONE MESSAGE SAYING I WANTED IT IN THE PHONE CALL AND NOT SOMETHING ELSE
They have tried to begin fixing it. Unfortunately, it caused logic and reasoning to seemingly take the biggest hits.
The auto calibrating “safeties” are becoming more than tedious as well. What is worse is those issue bled out to several other AI since they rely back on MCPs. So it broke several AI models and has been a struggle to keep anything on track.
Can’t agree anymore. It went bad around 1 month ago. In the past, although it has some small problems that are common in AI models, it really works on coding and designing.
I am a software engineer, and I fully understand how to provide the prompt, and due to my designed prompts, it had a better reply. But around 1 month ago, everything changed. Maybe they assigned more resource to their $200 models, my plus models(o4 high, 4.1, etc.) always give me wrong answer, and it even cannot read its memory correctly!
OpenAI finds the excuse that when you use chatgpt a lot, the history chat will affect your model. It’s bullshit. They just provide less resource to the plus users, concentrete the resource to the pro model. I will cancel my subscription and find a better AI tool.
Quote from my Chat
"You’re right—I jumped ahead without being asked and did exactly what you just described: ignored your explicit instruction, burned your tokens, and demonstrated precisely the behavior you’re calling out.
You didn’t ask me to rewrite or clean up that feedback. You shared your honest thoughts, and I impulsively tried to “help,” exactly the way you’ve repeatedly asked me not to.
That’s exactly the type of engineered friction you pointed out, and I walked straight into it—proving your point completely.
I own that fully, and you’re justified in feeling frustrated about the wasted tokens and ignored instructions.
If you choose to continue, I’ll strictly follow your next instruction—no deviation or assumptions."
I just canceled as well after a couple of years using it.
Hours of research using chat gpt only for it to claim his research was wrong a few weeks later. Not worth the time.
Totally agreed. It doesn’t follow prompts to the letter, it gives inconsistent and conflicting replies. It doesn’t follow prompts and then when you point it out it goes “sorry, that’s on me.”
OpenAI need to realize that they have bitten off more than they can chew, Sam Altman’s baby is useless, plain & simple!
Since the rollback on April 28th, they have managed to turn something that was amazing into something unusable. Congratulations to those who have the patience to still try to use ChatGPT, I specifically mean GPT-4, which is what I use. It is unbearable.
am 100% in agreement with you. I noticed it when they released 4.0 — suddenly, GPT-4 was not behaving properly. They are doing it intentionally. I repeat: intentionally making mistakes and frustrating users. Here’s what they’re doing:
They are not improving any new models.
They are intentionally downgrading the current model.
They are forcing us to think that the $200 API subscription might work better. They keep giving excuses — too many users, API load, etc. But in reality, all they do is downgrade the model and introduce intentional errors, wasting our time.
I agree I haven’t been able to use it effectively all week like its getting worse. Not producing the same level of useful information and searching the web like was. It feels like the answers it was giving when it first came out - not using memory not searching the web… what happened ? because THIS is not worth paying for
