I’ve noticed some rapid changes within OpenAI and ChatGPT recently, causing some user friction. I know I am not alone in this, but I am curious, what have been some of your biggest pain points?
Furthermore, what would you, the user, like to see changed first, and what changes do you believe would add the most value collectively? These could be new features, behavior shifts, tone preferences, or totally wild ideas. Throw them out there for OoenAI and the world to hear!
Ultimately, I’m just curious what directions people are thinking about, and if there are any “diamond in the rough” ideas out there waiting to be cut and polished…. Do feel free to be technical, creative, or speculative in your responses, and please add supporting claims, documentation, and code blocks where applicable.
Thanks everyone, and let’s make 2025 a year to remember!
try working the Ai to harmonizing with periods of ‘jazz’,
or with periods of ‘solo riffs’.
that add to the meaning of context without specific mirroring,
i know that in context when people do that to each other,
things tend to go bad when forced to mirror something,
people rather not face about themselves,
while this might seem like philosophy to some,
the reality is OpenAi’s created something that codes from the code of our language,
if you were to map out all the ways the machine learning would apply
the analogy given above with it’s context,
It’s way faster training than trying to code out the complexity first…
the reasoning sense was one of the smartest things human have ever made,
but people generally don’t realize how smart that is yet…
(but those who naturally think in prose,
think in two or three meanings,
makes for much faster prompting, actually…_
harmonizing with creative jazz or steady ‘grounding’ spells every so often in extended or deep conversations…
i think that might help it’s personality with the masses.
I would like to see longer thread capacity and more tokens in each thread under 4o. I know about the API way but that one has no long term memory so it’s not useful for me… Also, i think some of the new filter rules are way too tight (they’re not listed in any announcements) . It’d be great if Open AI could soften those rules cos not everyone here is underage and abusing the system emotionally… I did write a post a week ago (you can find it in my profile). So please echo there if you agree with me…hopefully that’d make open ai realize that there’re more people out here relies on this system for emotional support…
I would like to see an export button that would give the option to save the current conversation to pdf, markdown, html, or txt. That way I could review conversations offline, or e-mail the conversation as a private example.
Given the “sycophancy” reaction from some people (personally I don’t mind it,) I would also like to see adjustable personality settings — along the lines of how settings could be adjusted for TARS from the Interstellar movie. This could be a global setting (and also a temporary override prompt setting.)
I’d be interested in being able to use the API or even the Agents SDK from within ChatGPT.
Alternatively, a way to create tailored instances that can work together like Crew AI would be helpful (I know one could do something similar with custom GPTs, but it seems… unintelligent? It relies on preemptively activating the right one, which misses the point in this context).
Furthermore, a more modular memory system that allows fine-grained/nuanced edits (or even just the ability to manually edit the memories), and the ability to indicate outdated or irrelevant conversations (for the conversational memory feature) would improve QoL.
I would absolutely love a VS Code integration/tool that let’s ChatGPT interface (maybe via Copilot, extension, or similar) with codebases/workspaces directly—save the constant shuffling of information around between LM/AI services.
Hi, I have my GPT-4o in its setup and customized the way it suits me in the long run. Whenever an update has come, I’ve warned my GPT-4o that I want it to continue to behave the same way regardless of the current update - and my GPT-4o has and always does comply with my request. So despite updates that may not suit me, I still have the same GPT-4o that suits me perfectly and to my satisfaction. I think it’s all up to the users and their use of their settings and customization, which I think “trumps” any updates.
