I also really didn’t expect it to get worse. I thought it could only get better, but I was wrong too… I figured that it got worse regarding programming or coding stuff which was a big part for me besides text generation and SEO tasks etc. It got worse in every shape and form I used to use it.
1 Like
Yes, I am one of the first ChatGPT users, and since January 2024, the service has been declining. Now, even the audio chat doesn’t work properly because if you tap to stop and try to speak again, you end up speaking alone. You need to power cycle the chat to make it work again. I’m tired of it, so I canceled my Plus subscription a week ago.
I agree. It’s so FRUSTRATING when ChatGPT can’t follow simple orders. Like, I’m saying it can’t use certain words, and it does, I say “no stylizing text”, it uses bold and cursive, I explain to it what it did wrong, it continues to repeat the mistake. It’s an indescribable level of frustration.
1 Like
ChatGPT is an absolute nightmare, the level is so low it is very frustrating. It can’t follow simple instructions, can’t be consistant on the output start to become dumb after 3 to 5 requests. It has become an absolute trash. And the playground/ API is no different.
2 Likes
Are you guys using 4o by accident?
Nobody here is talking about which version.
4o is dramatically worse than 4, which is still available. And they will quietly downgrade your chat. Remember to check, when it strangely degrades in quality.
1 Like
Yeah. Totally. And then I realised it has started switching me to 4o-mini automatically recently and I hadn’t noticed. It’s worth checking what model it is automatically starting you with before you embark on something more complex. I gave up in frustration yesterday and subscribed to Claude to get the job done, and after the question from @kazvorpal I took a look and - well, that’s frustrating - defaulted on 4o-mini. You might have thought that if you are having to redo stuff time and time again, it might suggest you try the next model up - that in itself would probably stop them bleeding customers, because when people are focused on getting something done, they don’t necessarily look to see if the default has changed suddenly.
Having said that, I have also been using 4o and have noticed a considerable degradation over time. And 4 cannot cope with long scripts at all, it seems - which is why I actually found 4o to be a step forwards (for a while) - being able to keep lots of source code ‘files’ in context and coordinate them well. There was a sweet spot back in around May time, and then it all gradually fell apart - no idea why - I suspect they’ve dumbed it down so they bleed less money, but that doesn’t keep you customers. Seems like the next model is struggling to be performant enough to be usable though.
Looks like I’ll be flipping between Claude and ChatGPT for some time. Which isn’t a biggie - often generating several things on-the-go so just means more flexibility, if less money in my bank at the end of the month.
I do feel silly though. ChatGPT-mini on a complex coding task isn’t going to cut it!
There are indeed something going on but…
My fix(doesn’t always work) is usually to disable or clear memories. Sometimes ChatGPT is using something it ‘Learned’ from you to answer something you asked.
a confirm, and glad I am not the only one seeing this. chatGPT is soooo bad from month to month, that it gets on my nerves. it seams we helped them train it for their own use, and now they deliberately plugged it to such a dumb version…
it does not follow so many instructions, it lies, it repeats itself, it ignores, it seams this is deliberate at times, targeted even.
i do not trust this company anymore unfortunately. And I rooted for them soo soo much…
another example where power corrupts the integrity of man…
this is sad and very concerning…
emailed them many times. they gave me the typical corporate reply but nothing else…they enjoy the money they get from us but forgot the respect…sad
100% agree, I tell ChatGPT to remember instructions and it will be lucky if you get the same format twice. The old version even with its limitations seemed better and more consistent.
ChatGPT has gotten worse and worse, forgetting so many things. I am just about to cancel my subscription because it is too frustrating to have it make so many mistakes. I am not sure what openai did recently, perhaps they are throttling resources to save money, but the product is definitely becoming less and less useful.
Coding is no longer a joy with any of the models. You end up wasting whole days trying to undo the damage it caused to working code. The o1 model is basically the same as the others. This new “thinking” part is probably just a way to save resources. Frustrating, useless and harmful.
1 Like
its amazing how stupid chatgpt has become, and im talking about the pro plan. i think its becoming a platform for writing emails and letters and less towards complex task.
1 Like
Trying to speed up my workflow and asked Chat GPT to create a list of 50 tiktok accounts for italian restaurants in Berlin.
Did not accomplish the task and after several prompts i got only 5 restaurants, suggesting that it is difficult to work with tiktok and that i should expand the research myself.
Gave another couple of prompts asking to just search the web and give me a list of 50 websites instead.
Chat GPT gives me 5 sites and suggested that i should expand the research myself.
I finally asked if it is possible to accomplish the task just checking google maps, but obviously task was not accomplished and dropped it.
Now, what’s this?
I even paid the subscription for 1 month!!
Wtf !!
It has definitely gotten worse. In many cases, it forgets the previous chats. If they don’t fix it it’s gonna push many people to find an alternative
2 Likes
Code writing features got much worse. it always forgets parts of the code and writes in complete code even if I ask for the full response.
I regret that I bought Plus here instead of claude
1 Like
My dears,
I believe that every effort is being made to improve and improve the tool. We cannot forget the exponential growth of users, in addition to other factors, such as the emergence of other AIs competing in the market. Despite the notable loss in intelligence, I believe everything will be resolved with time.
To add to your rant… Today my gpt that worked somewhat okayish stressed me out so much I deleted it… It seems like they are making 4o even worst so users can upgrade to the bigger packages to get better results. I have used the free version before upgrading and now the paid $20 package functions like the free version… Dumbbbbb
i stopped paying and hardly use it now. its rubbish and i waste time i dont save time
I am so sick of it. I used to have a coding assistant with chatgpt3.5. Now i have a idiot work experience ai trying to make up rubbish