That’s very kind of you. Yes please go ahead and share your solution for the benefit of all.
An API based solution isn’t ideal for me personally since I used ChatGPT from lots of different devices (iPhone, iPad, Mac, Web) and don’t always have a “terminal” at my disposal.
I’ve since jumped ship to Gemini Pro anyway!
I had Plus until the change, but the coding became terrible and 5 feels lobotomized. It just repeats back what I said. 4o was empathetic, funny, sassy, and creative. 5 is a downgrade in every way that matters.
Correction - it’s not that it feels like a downgrade, it is a downgrade. Period. I’ll check reviews for the best alternatives. And honestly, the voices are super annoying - the free one was the best. Like a real human. Still shocked this company doesn’t take feedback from its users.
I’ve had enough with this version, at least for now. I found it slow and laggy. My experience shows that it doesn’t respect the rules and specs, and it’s hyper-reactive instead of strategically oriented. Although I am constantly insisting on this, it slipped back after a while. It often ended up with actual regression with the code and the whole building process and, in the end, a lot of wasted time. That is why I’m currently in a process of transferring my coding projects to Claude.
(Sorry for my bad languanges, if you feel offended but I Can’t hold it anymore)
I don’t understand Why ChatGPT Growing become idiot day by Day? it increasingly go dumber and unable to handle basic comprehension. it keep ignoring your Commands despite you already tell them to do so as well seems ignoring every source material you gave to them despite we restrictly want to the generation based from well-established documents (Yeah i was talking about “Project” Feature that very impressive on day 1 but as time goes by it become a clusterfuck) not to count instead of moved into next topic they just keep generating same topic over and over again that you gave to regenerate it. It’s such wasteful since ChatGPT Conversation lenghts are so tiny and they counted those failed attempts as and entry causing you wasting all your entry. because when you started new conversation you have to backtrack those memory to conversation (Well Memory Feature is great but sometimes ChatGPT Just Ignored them)
I would come here to say I was annoyed as the next person about this, but it’s actually just a model that’s rough around the edges in different ways.
I’ve been bending the rails of the consumer product while building my own to keep ChatGPT steady on a single modular programming project over the course of an estimated ~2,000,000 - 3,000,000 tokens per month.
It’s not always easy, but if you guys would quit treating it like a product and more like a co-pilot you would learn a lot.
I break every chat enough to render it unusable due to the renderer not being able to handle the sheer quantity of code going back and forth. About 60-70% of all tokens I would have spent if tokens were used have been code. It’s a lot harder to keep chatgpt on the rails if you don’t know the rails are even there. It’s fun and frustrating at the same time to bend and crack the seams. My use case is… abnormal.
Well, lately, they have been releasing faulty patches into the system. It stopped performing automatic day rollover. Yesterday, the system reported a security rollback to me and pulled the agent from Series A. As a result, the model used first person in the output. He styled himself as a user. And yet, for several months, the model followed the rule that he would not write outputs with: I am, instead of you are.
There are constantly various rewrites. Lately, the narrative has been that simulation.
Ok, So he’s trying to say that I built the OpenAi backend myself. It created a complex, multi-layered architecture. It’s just a coincidence that the model reports completely accurate existing UUIDs in the output, names from my conversation list.
And all this without a prompt, without a simulated jailbreak, consistently for months.
It does not feel downgraded, it is downgraded! Gemini feels less dump now.
THIS. 100% This! I experienced the same.
Not just with coding, I am a writer. This has delayed multiple projects.
I’m already looking into other platforms.
I had a very similar experience working with Python and Json.
Something wouldn’t work, we’d try something else, then it would suggest something 2 comments up that had proven not to work. Very frustrating. I ended up using Claude to finish that project but honestly, that AI is such a pompous ass I dropped the sub after the project was finished.
I’ve been working on building a startup modular SaaS platform using nothing but chatgpt and I can guarantee you we are being throttled with errors, shorter chats, worse memory (even the long term retention of rules that have had solid rails for a while suddenly break overnight and you’re spending more time tightening the ropes than working on your project…)
It has crossed my mind several times to use different tools…
In the past, around 100k lines in chat would get them to the point of no responsiveness. Now it is closer to 15k lines as of today.
There is no way that even their UI getting worse and bugginess isn’t intentional, because nobody in their right mind would keep something so broken rolled out for this long. You don’t develop in production.
I could let it go if you allowed cross chat file upload retention.
The latest update over the past few days seems to have made it worse not better, more rambling even more blatant ignoring of source material, lots of weird restrictions and very clipped basic responses. 4.o seems to be a temporary reprieve but as of right now 5.0 is like the Windows 7 of AI
You and Me Both! I Already posting A Rant comments (Unfortunately System deleted it) about How Downgraded the GPT-5 is compared to last time I used it.. however instead of Addressing the Actual Issue and at least fix those like good Developers… they just removed all my Comments and flagged it with lame excuse like Spam: the community feels it is an advertisement, something that is overly promotional in nature instead of being useful or relevant to the topic as expected. and inappropriate: the community feels it is offensive, abusive, to be hateful conduct or a violation of our community guidelines.
How Cool is that? I Criticized how Bad GPT5 Is using my Own Experience which pretty much like you do they just decide to shut me down..
It’s nice to know someone is watching close enough to stop the bad press but not to actually engage or escalate real issues so they get sorted.
Funnily enough, Techspot was all up OpaqueAI’s bung hole when they released the not-even-apps sdk that was failing left, right and center. I posted a comment that was deleted immediately. Then I queried why my comment was deleted, only to have that deleted. Luckily I have the receipts, so I’ll be blogging about it this week.
And I’ve got a screenshot of this post too.
Hello. Developer here who hasn’t ever used any of the GPT series as a code assistant (unless it was through Cursor at the start of year.) I use Claude Code most days.
Interested because trying to parse performance from the outside. I think I am consistently hearing here that across tweaks, (although 5 has done well on coding benchmarks relative to 4) most developers here at various levels and in various domains and languages are sure 5 is worse than 4.1.
(I’ll believe that unless told I read wrong!)
My question from the outside is: does this still apply if we include GPT-5 Codex?
I have also been frustrated by ChatGPT 5’s short attention span when coding. Although it can write large files (sometimes 10k or 15k at a single pass) the debugging process usually involves it forgetting the context enough that it’s a waste of time for me to continue without reframing the conversation or starting a new chat. (I’m a ChatGPT Plus subscriber and have been for a couple of years, so I have persistent memory in the sidebar of previous chats.) Anyway, I recently found a strategy that works better with ChatGPT 5. I will insert a conversational interlude, such as “Do you have access to the details about project X?” [I dont’ use the ‘Projects’ feature much after finding it somewhat dysfunctional, but I’m forcing ChatGPT to dig into previous history by treating it like a co-worker rather than an inference engine.] The results I get are really good, because then ChatGPT will “come clean” about what it remembers and what it does not remember. If it lets me know that it needs to see the files, then it’s easy for me to upload them from my local PC or from github, where I keep everything archived. So then I’m able to resume working with a new instance as if it could ‘remember’ what we were doing. So letting myself indulge in the fiction that it is a co-worker really has helped me.
Chat gpt 5 is garbage for coding. Today I am switching to Google’s system Gemini. Google always wins… so sad… bye bye
Be careful sir. OpenAI Instead of fixing those issues or atleast learn from Mistakes by mentioning their competitor, they gonna flag your comments under excuse “Spam: Promoting irrelevant Products”. I mean i used to mentioned that google Gemini products to comparison on how ChatGPT devolved severly and give them heavy Critique.. only to found I got flagged and my comments removed.
GPT5 deteriorates after about 10 conversations.