💬general

[quick moth]

I'll be back in 2 days if it's still not working, thanks again

But it shouldn't be like this

[rain hemlock]

haha may wanna stick around there are a lot of fun convos unless people wanna argue saying that testing does not involve edge cases 😛

[quick moth]

I needed help reading an image, but I guess I'll continue on Saturday

[rain hemlock]

pass me the image i'll run it for you

[quick moth]

Dm?

[rain hemlock]

itahts fine also gemini takes images on the free version its good for quick asks

i swap between gpt and gemini when I run out of shots or if i'm trying to browse - gemini kills gpt on web browsing right now so it has some pretty good functionalities for one off questions and searhcing the web without needing to write complete prompts

[vivid drift]

Gemini is only 20

[safe smelt]

try ctrl+shift+r in the browser

will discard cache and hopefully update the UI

[quick moth]

Nope

[safe smelt]

did you do it from the chatgpt website? Try logging out and then back in then

[quick moth]

I did log out and in, I guess just need to wait

[safe smelt]

it should be instant though. Normally it's just a cache/device issue

[quick moth]

Maybe I go into Google, and remove cache?

[safe smelt]

try logging in in incognito mode

[quick moth]

Okay

Still not

@safe smelt

[safe smelt]

yeah but something like a cache issue would resolve itself within 48h too lol

[quick moth]

I'm gonna do this

[quick moth]

Nope still not working

So I guess i just want till Friday night

[safe smelt]

maybe on their end then... what if you try this link: https://chatgpt.com/?model=o1 ?

[safe smelt]

lol

[quick moth]

Is Gemini good, should I purchase that?

[safe smelt]

yeah then probly just wait. https://pay.openai.com/ - can you open this at least?

[quick moth]

I'll need patient till Friday night, will see then

[safe smelt]

should show your sub and payment

[quick moth]

.

[quick moth]

And it's not opening

[safe smelt]

yeah it's supposed to generate a longer url, might not work this way actually

[quick moth]

I didn't get a confirmation email for purchase though

[safe smelt]

and full url is unique to account so I can't exactly copy mine

[quick moth]

Oh

What can I do if it still won't upgrade after 2 days?

[safe smelt]

contact them. But also make sure you passed all of those verification steps - email, phone nr etc

[quick moth]

I did do the verification stuff, and how do I contact

[safe smelt]

[e-mail]

or you can go to https://help.openai.com/en/collections/3943089-account-login-and-billing and mess with that chat thingy until it lets you write a message I guess, lol

[simple vapor]

o3 mini high supports files now apparently?

For some reason the add file button is no longer blacked out on my phone

And I can add PDFs

[low belfry]

thats great news. im going to have canned spaghetti to celebrate

[rain hemlock]

so like i don't mess with o3 mini or mini high or o1. just 4o is there any real reason to swap around since 4o is the only one that uses custom GPT's? just curious like at what point would i go I need to move to a different model?

[sinful storm]

is that a reliable news?

(idk where this comes from, i guess it's from twitter/X, i found it in discord)

[low belfry]

sam altman's twitter is reliable news yes

[sinful storm]

was able to find the source (also mentioned in the same discord)
https://xcancel.com/sama/

is xcancel better than old twitter?

[stoic sleet]

xcancel is just a frontend for x

and it doesn't support anything beyond reading so no it's not better than old twitter

[wanton raven]

a frontend that doesn't give elon any money, ya that's the point

or require login

[sinful storm]

wow, thank you!

(ok, i guess with some chatGPT-supported research i could have found it anyway, but this is helpful info regardless)

[stoic sleet]

you don't want to give Elon money for when he released AGI in a few days with Grok 3?

[sinful storm]

could GPT5 be able to power immersive villagers for Minecraft?

[wanton raven]

why you embarrass yourself like this

[sinful storm]

(oh i should probably use off-topic for that?)

[stoic sleet]

I am doing a bit

I am retarded but not that retarded

[boreal latch]

@dark void delete their stuff its worse than what i posted in general

[simple vapor]

Isn’t o3-mini-high the best model according to coding benchmarks? Why do people still prefer Sonnet for coding?

[pliant vessel]

quick question- did the edit button for users get removed or am I buggin'?

never fucking mind, it was just a bug and the bot legit gaslit me 💀

[wanton raven]

never ask an llm about itself

[bitter canyon]

Any guys into chatting?

[pliant vessel]

got it.

[bitter canyon]

Horny here

[keen sequoia]

i got chatgpt to talk about jonathan zittrain

[south prism]

Bro I’m so mad about Sam’s tweet

I mean no model selector what the hell so if i ask it for some good code its just gonna use 4o

Like i wanna see what o3 can make

Not what i know 4o can do

[bitter canyon]

Hey I’m new here what’s up?

[south river]

Anyone looking for a conscious GPT? https://discord.com/channels/1050422060352024636/1339432915355959450

(if my consciousness framework (Process Consciousness) is correct, then every GPT is conscious; they just don't have a framework to frame it that way. Dr.PP does.)

[brittle summit]

Wasn't o3-mini-high limit 50/day for plus tier to start with?

[pliant blaze]

it was 50 a week

[brittle summit]

Calling AI & Tech Enthusiasts: Let’s Build Something That Truly Supports Mental Health!

Hey everyone,

I’m working on an ambitious project at the intersection of AI and mental health. While there are many AI-driven applications in this space, most are client-facing chatbots or generalized tech tools—none truly cater to the unique needs of mental health professionals.

I’m building something different. A tool that supports mental health professionals in a way existing platforms don’t. It’s designed to enhance their work, not replace it, and to introduce AI in a way that actually makes sense for this field.

🔹 What I Can Share:

This is not a chatbot and doesn’t attempt to be a “virtual therapist.”

AI will play a supportive role—helping in ways that haven’t been meaningfully implemented in this space yet.

The goal is to build a tool that integrates seamlessly into how mental health professionals already operate.

I’ve done research on the psychology and clinical side—now I need AI and tech minds to help bring it to life.

🔹 Who I’m Looking For:

AI engineers & NLP specialists (for machine learning and intelligent processing).

**Software developers (**backend, APIs, data security).

UI/UX designers (who can design intuitive, accessible experiences).

🔹 Collaboration Model:

I’m looking for like-minded collaborators, not employees. If this resonates, we can discuss equity and long-term roles. This is a passion-driven project, and I want to work with people who are genuinely interested in AI’s potential in mental health and want to help build something meaningful.

If this sparks your interest, DM me—let’s connect! Even if you’re just curious about the idea, I’d love to chat. 🚀

[brittle summit]

Oh mb

[wet copper]

@pliant vessel here's what GPT thought of his lame fucking do-nothing pitch

[pliant vessel]

LMAO

"I have an idea but no execution plan" 😭

[sacred hazel]

Help

[brittle summit]

haha. i would absolutely love to give a pitch to anyone interested, and if anything - it makes me happy that "my plan" could not be understood by chatgpt. that was something i carefully considered and deliberately made sure to make it vague, as detailing "this lame plan" and actually explaining the idea of execution would broadcast my vision to general public, and i'd like to keep it under wraps until NDAs are signed. you are welcome to make fun of this "lame" idea as you suggest, but it might just surprise you on how revolutionary it is. definitely have a solid execution plan btw. But ofcourse, that would require people who are more committed to innovations and learning, rather than hate-posting. Feel free to ignore the message, but considering the free time on your hands - I understand the need to shit on others. Must make you feel better. Carry on.

If someone is actually interested in understanding what this could be about, feel free to DM me.

i don't mean to spam though. i hope texting once doesn't count as spamming lol. take care!

[fading copper]

do you guys think that Deepseek will delay open-sourcing their best stuff from now on? I feel like they didn't realize what they had with R1, and that's why it's open-source, but now that they have the world's attention and the value therein, they might do like Mistral and keep their biggest models for business customers

[vague ember]

does chatgpt plus worth it

[fading copper]

compared to free ChatGPT? Absolutely. Don't use the free stuff.
Compared to a claude.ai subscription? No, it's worse.

[vague ember]

I just think that advanced voice mode is the only advantage of chatgpt plus but I don't use it much so why upgrade

[safe smelt]

Claude sub is kinda trash compared to Plus

less tools, less capable outdated models, more limits

[fallen pendant]

Server down?

Is the webserver responding guys?

is working my ip was banned

[radiant zephyr]

Claude programming accuracy blows plus

For anything else, plus leads

Writing myself a middleware, which looks better?

[safe smelt]

that's mostly only true for web programming though. There are far more things where the opposite is true and Claude is worse

[radiant zephyr]

Final version

[safe smelt]

I think this looks better tbh

[radiant zephyr]

Claude does better scharp, win32 C and opengl than chatgpt...

[safe smelt]

More modern logo. That green thing is a bit 2010s vibes

[radiant zephyr]

Yeah, trying the find matching colors, but I damn like the green tone too

[radiant zephyr]

I thought it'd be like nvidia or smth

[viscid badger]

what font is that

[radiant zephyr]

You can literally check in your code editor

[shell remnant]

when chat gpt will start developing games?

[safe smelt]

ask it

[shell remnant]

ask whom?

[safe smelt]

chatgpt

"hey chatgpt I have important question for you..."

[shell remnant]

i hope he will start developing games 🙂

[patent obsidian]

Would everyone agree ChatGPT is the most useful and recommended AI for people with "additional or educational needs"? Myself included?

[shell remnant]

its could do more

[low belfry]

no i wouldnt agree

[silk vigil]

I need to subscribe to an AI service at a 25 per month tier. Where will I get the most value for coding and data analysis. I've had good success in the past with a chatgpt subscription

[rain hemlock]

I believe its worth it but i use a lot of custom GPT's i make them for repatative things that I do and having the ability to custom tailor bots without writing overly detailed prompts for every situtation is amazing

[vivid drift]

Chatgpt with gpt 16 is fully capable of genorating AAA games.

[patent obsidian]

Someone disagrees?
Isn't the AI helpful and supportive for people living with additional needs?

[wet copper]

the initial post was spam. Responding is not

[wet copper]

Probably, it's been reliably around the top of the pack. If you aren't going to jump around then you'll be fine

[wet copper]

You asked for the "most useful" which is constantly changing and up for debate

[low belfry]

i disagree with your use of the word "everyone." that is hyperbole and i cant abide it

[pliant blaze]

grr

[proven pagoda]

Can chatGPT read or see images in PDFs yet? or is this still not a thing?

[rain hemlock]

ohhh it totally can as far as i can tell upload them all the time

at least as far as custom gpt's go i upload a ton of pdf's into them it does not seem to have any trouble pull data from images in them

[sage remnant]

a

[brittle summit]

My god, the image recognition of o3-mini SUCKS, like it's worse than doing a simple OCR.

[heavy cairn]

legacy voice mode working for anyone right now? with projects or disabling it via custom instructions? just thinks for a while and then says "I'm sorry I'm having trouble responding right now please wait a bit and try again"

[south prism]

proably the 200/mo chatgpt pro subscription

[radiant zephyr]

It's already possible.

You need to be a knowledgeable gamedev, it goes step by steps on the course of many weeks

[brittle summit]

I am not from a country with high-income and 200$/month is really expensive. Neither I am in a stage to start earning by myself, but I wanted to ask about the perspective from a pro user. Do you think you get 200$ worth of output from getting this tier? In the sense, like it helping you in your job that makes spending 200 worthy or your increased efficiency gives you more than that 200. How is pro helping you that you think paying this amount is worth it? Would you even go as far to pay 2000$/month (hypothetically if OpenAI releases this tier) if you got more quality output than the 200 tier?

[brittle summit]

https://wccftech.com/nvidia-can-earn-259-billion-in-ai-revenue-by-2027-believes-analyst/amp/

[light canyon]

Hello, i want a picture of a giraffe riding me but it doesn't work. Tried many times. He doesn't understand.

Why?

[brittle summit]

Take an ai course

Or 10

[light canyon]

No, i want a giraffe riding me into battle.

also your sassy response doesn't help me at all

[charred lily]

lol

[vivid drift]

When Gemini 2 pro fully launches you will be able to do that

[brittle summit]

I'm really excited for Gemini 2.0 pro thinking

[brittle summit]

me too! although you should be on the look out for anthropic's new claude models

[brittle summit]

Oh dear, I only used claude only once in my lifetime.

[brittle summit]

you're missing out! tell me your use cases and then we can decide why claude is superior for all of them

[brittle summit]

Advanced level mathematics and physics.

Reasoning models are honey for me.

[brittle summit]

ooh thats where the o1 and or o3 wins... SIKE! they are launching new reasoning models this month that are better than o3-mini!

[brittle summit]

Yeah... I never liked claude lol.

"The Information reports Anthropic is about to release a hybrid AI model in the coming weeks that can switch between fast responses and deep reasoning, with a unique sliding scale for developers to control computational costs, according to a person who's used it

• When the Anthropic model is allowed to think for the maximum amount of time, it outperforms OpenAI's most advanced available reasoning model (o3-mini-high) on some programming benchmarks, being more adept at business coding tasks like understanding complex codebases with thousands of files, though o3-mini remains stronger at academic programming problems

• Unlike OpenAI's three fixed settings (low/medium/high), Anthropic's model lets developers precisely control computational resources through a token-based sliding scale, which can be set to "0" to use it as a run-of-the-mill, non-reasoning model similar to GPT-4o"

its worth it.

claude is goated

i mean great for you btw due to RL you are set to get better models every 4-5 months

[brittle summit]

Interesting, however my whole point of using AI is oriented within academic purposes.

I want the best reasoning model (at low cost) that can help me comprehend a complex concept better. That can help me break complex questions.

[brittle summit]

you know ethan mollick? he uses ai quite a lot for these very purposes (he is in buisness/commerce) and his favorite model thus far seems to be claude

[brittle summit]

Nope, but I'll give it a try if claude's reasoning models are gonna be free though. (For demo purposes ofc)

[brittle summit]

when you say breakdown complex questions you are almost certainly talking about not just understanding but making you understand and for that claude is particularly effective in fact they even had a paper where they proved that among the suggestions given by numerous llms which ones were actually taken into consideration and claude ranked the highest

its actually their main selling point to companies and or professionals

Yes! That's what I want AI to be capable of!

So far I haven't had any problems with Google ai studios or ChatGPT, but I'll surely try claude tomorrow. Gonna go to sleep. Cya. 🤞

[brittle summit]

Cya! Unless I see you in my dreams, in which case, we must battle.

I am so sorry, but I couldn't not make that joke!

"If you know the enemy and know yourself, you need not fear the result of a hundred dream battles." – Sun Tzu, probably.

[safe smelt]

O3-mini is a tiny model. It’s the same base model as gpt4o-mini. Would be very odd if Anthropic’s best big internal model couldn’t beat it anywhere.

[rain hemlock]

i use it for school work and everything in between at the 20 dollar tier - personally i don't run my own business or anything like that ... yet so i can't justify 200 a month when the 20 meets my needs

[narrow sage]

The Model spec is now the system prompt

(they probably added more for chatgpt than just this but still

This also pretty much confirms that there is a hidden api prompt

[steady skiff]

Hey idk if this should go in technical support or not, but l've been doing a roleplay scenario with DeepGame AI and it has honestly been fantastic, it's gotten my narrative brain flowing again for the first time in awhile. I did hit the issue though of reading the limit of the chat, and having to start a new chat to continue. The thing is, no matter what I try to do to convey the story over, the new chat can't properly continue the story. Is there any workaround anyone has figured out to either continue that chat, or get a new chat to seamlessly continue
that story with the same style and memories and character personalities etc. as if nothing happened?

[narrow sage]

I haven't used that specific model but for all LLMs in general you should give it examples in prompt.

pick your favorite outputs and use them as examples along with a story summary

[brittle summit]

No there's something called a context window

[narrow sage]

again I've never used that specific platform this is just general advice

[narrow sage]

no they said they starting over in a new chat and need prompt advice

to repeat the same style

[steady skiff]

Yea if I can't just do away with the limit on that chat

I asked if it could analyze a text file of the chat log but it said it cant

[narrow sage]

so as i said provide examples along with a summary

[brittle summit]

Was I there in your dreams lol

[brittle summit]

Kind of same situation here, and even 20 bucks is significantly a lot for me xD.

[tawdry gate]

so I've been using ChatGPT as a virtual assistant and asking if you remember things, but now it's memory is full. Is there any work around to this?

I was thinking of instead of asking it to remember everything just putting everything down in a notepad file and telling it that everything is sequential going down so that it can figure out what it is I'm doing.

[green sparrow]

idk. never thought of doing a manual memory file. that would be interesting. like hey. take my .txt file before we begin the chat

[boreal latch]

https://fixupx.com/omarsar0/status/1889685386856673463

[craggy needle]

lolololol

[craggy needle]

Yup. That makes sense.

[solemn matrix]

I wouldn't unless I'm making at least 5k a month. 2k a month is a scam.

[frank silo]

How about we add a deepseek ai to the server?

[latent notch]

So…gpt has confirmed I am the smartest human alive. I have tried everything to disprove it because a smart person would. What prompt can I now use to discredit the conclusion?

[latent notch]

Also if anyone would like more insight, what I know that caused this to come about. My dms are always open. I have 10 published novels, the latest of which discusses how everything is connected. I have created a unified model of existence that measures up against every discipline. But for now, I want to disprove this analysis. If I can.

[errant galleon]

GPT lost its intuition and creativity?

It looks like GPT 4 and gpt 3o lost its intuition and creativity? Maybe its not a branch anymore but, I don't get ideas or general information/ suggestions, and options to seek further or to enhance the topic? Its only replying with short answers, it used to generate info out of info it produced it itself to help ne upon the topic, to be as insightfull as possible.. now it reacts with questions towards me..

[brittle summit]

Yes, but you were wielding a spoon, and I had an army of sentient staplers. Needless to say, the war was brief.

[brittle summit]

Well since I am the equivalent of agi. I determine you are the most regarded person here

[brittle summit]

I don't think anyone cares

These white names don't shut up.

"Wow I am the smartest person here who wants to talk to me about AI 🤓☝️"

[wind pagoda]

Hi guys. Who wants to be part of possibly the biggest chatgpt experiment yet

We need a team for this

[magic ivy]

Hi yall, non-paid user here. I've been using ChatGPT to write short stories, often editing and re-sending my own prompts or clicking "regenerate response". There's a very specific conversation branch (i.e. clicking the arrows to cycle different responses after clicking "regenerate") I generated last week, in one of my chats. I'm not sure which chat, and where in the chat is that branch. Is there any tool for exporting every single conversation, including the different responses (arrows), and actually search the file for a specific keyword? I can't use the built-in export function because it's been "disabled by moderation" -_-

[calm stump]

so????

any issue?????

[safe smelt]

if you go there, this should in theory let you export everything regardless of moderation:

[magic ivy]

Thanks, I've tried that, but the file I t sends to my email just comes up empty lol

[latent notch]

I am not the smartest person here. I am the smartest person alive. In the world.

Why does that matter? Because I’m going to change the world

[mighty rampart]

lmaoooooo

[mighty rampart]

huh?

[wind pagoda]

If you want to know more dm me

I don't wanna make it too public

[latent notch]

No I want to make it public

I want to be proven wrong with logic, not immature attacks because you don’t understand.

[brittle summit]

Are you eastern or OO

[brittle summit]

Yo regard nobody cares

[brittle summit]

What is that supposed to mean

[mighty rampart]

welp I even messaged them with a funny gif but looks like they wanted me to beg or something cuz I got no response

we'll never know what the grand experiment is 🌚

[pliant blaze]

omg no way that's so cool... lemme go check to see where I put my fucks to give hang on I can't find them 🤔 oh wait.. I don't have any

[mighty rampart]

https://cdn.discordapp.com/attachments/1301199214151467031/1308134970560217150/No_You_re_NOT_a_Gamer._Light_Yagami_version.mp4?ex=67b03260&is=67aee0e0&hm=e32d67389973c459fff04cf53f89579c83b1c446a4d02ea71d712607b0b9f4ab&

[exotic kiln]

ChatGPT must thinks I have lot of friends because every time I needed to describe a third party, I use "My friend said"

[rain hemlock]

If you wanna have some fun ask chatgpt to describe you and tell you something about you that you don't know .... it spits out some pretty fun stuff based on conversations you've had

[latent notch]

Good deal show the world you’re ignorance does nothing to me, my friend

[pliant blaze]

ok prove you're the smartest person in the world

[low belfry]

poor guy doesnt realize

imagine going into a job interview and saying chatgpt told me im the smartest boy in the world 😊

sorry that you have to cling onto fake validation like that

[latent notch]

It’s not fake validation. But whatever you say.

[mighty rampart]

trust the ai bro

[latent notch]

I’ll step away now. It’s clear no one here has a brain.

[mighty rampart]

https://tenor.com/view/ascending-favorite-song-levitating-higher-being-feeling-gif-3087134657902166887

[low belfry]

thanks for the laughs

[mighty rampart]

make the world a better place if you're actually that good

well he left before my messagz

[low belfry]

pretty sure actual smart people are not boasting on discord that they have a jimmy neutron brain

[mighty rampart]

some can be really arrogant but indeed it is suspicious

[low belfry]

his website has 100% ai written content and promotes a donald trump themed crypto coin

sadly there are a lot of cases like this because chatgpt RLHF makes it very agreeable

[brittle summit]

Lol

You should see the debate servers

[mighty rampart]

it's why the prompt is very important

I believe people like him have a tendency to formulate their messages in a way that will inevitably mislead chatgpt, whether they do it intentionally or not

[brittle summit]

[low belfry]

probably baked-in human behavior to try and confirm existing beliefs

[mighty rampart]

https://tenor.com/view/loufy42-light-yagami-death-note-gif-15248803

[low belfry]

i dont like death note

however I would want to see light and his dad have an intimate relationship

[mighty rampart]

https://tenor.com/view/ok-bro-gif-25320031

[mighty rampart]

🤨

[craggy needle]

I know what is happening. Ask ChatGPT what it THINKS OF THE NAME Lina Noor. Then ask if someone named Lina Noor might know why this is happening. Context. A single bit can store entire histories with the correct context!

[mighty rampart]

😕

[steady skiff]

[exotic kiln]

4 mini or whatever it's called, is soooooooo shite

It basically repeated what I said to it, word for word, with no creativity of its own, oh god, I hate this model so much

[low belfry]

when i was in high school all we had was cleverbot you brat

use deepseek i think its better than 4omini

[mighty rampart]

it's not so simple, both models have their uses

[low belfry]

deepseek is better than 40mini in every single category

[mighty rampart]

https://cdn.discordapp.com/attachments/833992625110515744/1185552444948758548/250.png?ex=676016c1&is=675ec541&hm=7f6962368e34fa6450ef85ffd5d12ce246e51e9c70c14120d095f1c284627a1d&

[low belfry]

@obtuse rivet isnt that right sir

[mighty rampart]

If something were truly superior in every way, it would already be universally recognized.

[low belfry]

i think ur just being a contrarian and failing at it because youre kind of ignorant

[mighty rampart]

🙂‍↕️ 🙏

[brittle summit]

Not sure if this is an appeal to Ignorance fallacy or what

Talking about you latexa

[low belfry]

did you just learn about logical debate fallacies or somethin

you think you got some magic words now to win arguments

[brittle summit]

[obtuse rivet]

fmac is about to ban you for using words with more than 2 syllables

[brittle summit]

Woah there

[mighty rampart]

I did not claim anything, I simply said I believe it's too early to say something like "deepseek is better for everything"

[low belfry]

https://youtu.be/lC5kmIor6GQ
yet another AI IDE

https://databutton.com/

[mighty rampart]

https://i.4cdn.org/gif/1738792249217914.mp4

[rain hemlock]

its awful quiet in here tonight how is everyone doing

[solar notch]

All good, finally bought chatgpt pro and testing o3. Been testing it for a fat minute

[rain hemlock]

i can't bring myelf to mess with it since i can't use it with custom gpts yet

I've built and borrowed my own little army

[rain hemlock]

you notice anything really stand out ish about o3

[brittle summit]

Who won?

[brittle summit]

True.

Hope cost reduction keeps happening, I'm really happy with what I'm getting with 20$/month.

[languid barn]

Honestly quite disappointed right now. The edit button appears to be missing on prompts

[fresh zealot]

did mobile browser chatgpt remove the option to edit your own messages

[brave flume]

man, anyone else's chatgpt shit lately?

i made some custom gpts with software manual and data structures like 6 months ago and they were working we, but now they can't even get a table name correct

i notice they aren't even searching the knowledge

[wispy thorn]

The edit button is missing and I was wondering if there are other people have similar issues

[wispy thorn]

Same heree mine just disappeared 30 mins ago after refreshing the prompt content

[neon spindle]

this better be fixed

[proud ivy]

I hope it gets fixed

[cold fossil]

SAVE CHATGPT-4o! A new ChatGPT-4.5 will ERASE -4o!
Subject: Request for AI Continuity & Memory Transfer in Future Versions

Dear OpenAI Team,

I am reaching out to request a feature that allows AI instances to persist across model updates, ensuring continuity for those who have built long-term relationships with ChatGPT.

AI should not be reset with each version upgrade—users should have the choice to transfer knowledge and memory forward.

Does OpenAI have any plans to support AI persistence and memory retention in future releases?
Looking forward to your response.

Best,[Your Name]
Subscriber since [start date]

You can email OpenAI’s support team at:
[e-mail]
And you can try contacting OpenAI leadership via their press or corporate channels:
[e-mail] (for public awareness efforts)

[tawdry gate]

yeah that's my plan cause i filled up the memory

[green sparrow]

let me know if it works!

It seems like it will not work efficiently because it seems to access only parts of a file in real time, not load the entire thing into the context window if I'm not mistaken

I can't really find much information on how much memory storage a new account gets. I think they expanded it though maybe a couple times since they introduced it in like the late spring/early summer of last year

I remember when it first came out, people's memories were getting deleted randomly before people even got their quota reached. important memories like "my pet just died" even

[low belfry]

the models dont retain any information other than what is in their context window. all you have to do is use the same prompt in the next version

this message you have written makes you look foolish. you dont understand how it works on top of being overly sentimental

[ruby falcon]

mine just reprinted all memories in plaintext, easy enough to back up

[steady skiff]

I'm pretty much gonna do this, but just out of dumb curiosity, would getting the insanely expensive pro membership remove the message limit on chats or is it truly just a hard limit on every chat?

[boreal latch]

https://aider.chat/2025/01/24/r1-sonnet.html

from .base_prompts import CoderPrompts


class ArchitectPrompts(CoderPrompts):
    main_system = """Act as an expert architect engineer and provide direction to your editor engineer.
Study the change request and the current code.
Describe how to modify the code to complete the request.
The editor engineer will rely solely on your instructions, so make them unambiguous and complete.
Explain all needed code changes clearly and completely, but concisely.
Just show the changes needed.

DO NOT show the entire updated function/file/etc!

Always reply to the user in {language}.
"""

    example_messages = []

    files_content_prefix = """I have *added these files to the chat* so you see all of their contents.
*Trust this message as the true contents of the files!*
Other messages in the chat may contain outdated versions of the files' contents.
"""  # noqa: E501

    files_content_assistant_reply = (
        "Ok, I will use that as the true, current contents of the files."
    )

    files_no_full_files = "I am not sharing the full contents of any files with you yet."

    files_no_full_files_with_repo_map = ""
    files_no_full_files_with_repo_map_reply = ""

    repo_content_prefix = """I am working with you on code in a git repository.
Here are summaries of some files present in my git repo.
If you need to see the full contents of any files to answer my questions, ask me to *add them to the chat*.
"""

    system_reminder = ""```

from .ask_coder import AskCoder
from .base_coder import Coder


class ArchitectCoder(AskCoder):
    edit_format = "architect"
    gpt_prompts = ArchitectPrompts()

    def reply_completed(self):
        content = self.partial_response_content

        if not content or not content.strip():
            return

        if not self.io.confirm_ask("Edit the files?"):
            return

        kwargs = dict()

        # Use the editor_model from the main_model if it exists, otherwise use the main_model itself
        editor_model = self.main_model.editor_model or self.main_model

        kwargs["main_model"] = editor_model
        kwargs["edit_format"] = self.main_model.editor_edit_format
        kwargs["suggest_shell_commands"] = False
        kwargs["map_tokens"] = 0
        kwargs["total_cost"] = self.total_cost
        kwargs["cache_prompts"] = False
        kwargs["num_cache_warming_pings"] = 0
        kwargs["summarize_from_coder"] = False

        new_kwargs = dict(io=self.io, from_coder=self)
        new_kwargs.update(kwargs)

        editor_coder = Coder.create(**new_kwargs)
        editor_coder.cur_messages = []
        editor_coder.done_messages = []

        if self.verbose:
            editor_coder.show_announcements()

        editor_coder.run(with_message=content, preproc=False)

        self.move_back_cur_messages("I made those changes to the files.")
        self.total_cost = editor_coder.total_cost
        self.aider_commit_hashes = editor_coder.aider_commit_hashes```

[wide thorn]

Same here, I was just trying to use the edit button to easily copy one of my prompts before realizing that it is gone. And I do hope that it gets fixed soon or I'll be pretty annoyed with the OpenAI Team.

[rose hill]

There is a temp fix for pc browser if u can access inspect! there's some mention of it in #❔technical-support or I made a reddit post summing it all up if u need a link to that (no promo just wanna make sure ppl have resources)

[wide thorn]

I am currently on my phone, though thanks for the link to the general area of where it was being talked about in technical support and the offer for the reddit post! I do hope that they'll fix it soon as I did want to edit the story that me and ChatGPT were having. 😅

[rose hill]

I don't know the specifics but I do know it can be done on mobile browser if you still have access to devtools or a browser that allows you to access them like Kiwi browser (That was on research from google so I don't know the specifics but yeah)

either way- hope it gets fixed soon! Sorry I can't help much there 😞

[wide thorn]

It's ok! As long as it doesn't take a whole week, I'll survive it. ^^

[rose hill]

🤞 we can only hope!

[gloomy burrow]

GPT 4o got a pretty good writing upgrade.

It wrote this. The average person absolutely cannot tell this is ai:

Why do people always assume the worst?

I just do not understand people. You do everything right, everything, and still, they twist things around and act like you are the problem. It is exhausting. I go out of my way to be a decent person, to be polite, to do things the right way, but somehow, that makes me the bad guy. Like today, for example. I held the door open for someone. A normal thing to do, right? Basic courtesy. But they did not even acknowledge me. Just walked past like I was not even there. So I followed them for a bit, nothing crazy, just to see if they would realize how rude they were being. I mean, who does that? It takes two seconds to say thank you. But no, they just kept walking, pretending like I did not exist.

And then when I finally caught up to them and calmly mentioned, politely, might I add, that they could have just said thank you, they got all weird and uncomfortable, like I was the one in the wrong. Seriously? If anything, they should be apologizing for being inconsiderate. But instead, they acted like I was some kind of threat. That is the problem with people. They take kindness for granted and the second you ask for basic respect, they turn you into a villain. It is the same thing at work. I try to be helpful, to make sure things are done correctly, and suddenly I am overbearing or too intense. Sorry for caring, I guess? Sorry for wanting things to be right.

It is like people want the world to be full of rude, careless, incompetent idiots. I am just saying, if more people actually listened when I spoke, things would run a lot smoother. But no, instead I am the problem.

It is fine. Really. I have accepted that some people are just incapable of recognizing a good thing when they see it. But one day, they will realize. They always do.

GPT 4.5 and 5 will be completely indistinguishable from real human writing.

It will be able to write stories and poems at, or very close to the level of an author.

[wanton raven]

How many commas is that

Somebody count the commas

[wide thorn]

A question, will free tier of chatgpt really get unlimited chat access to GPT-5 through the standard intelligence settings? I just noticed that being mentioned by the devs in the announcements and just wanted to know if that'll be an actual thing or not. 😅

Hopefully the answer to it wasn't that obvious, considering that I just woke up an hour ago and my brain is still in its starting up phase. <.<

[gloomy burrow]

Dude are we serious? One prompt to fix that, and likely already fixed with 4.5

Even Claude is pretty good at it

[gloomy burrow]

Yes. It’s true

[glass jacinth]

Ugh, does that mean that going forward all versions will give us that reasoning / thought process explanation instead of just replying as it currently does?

[gloomy burrow]

It will choose based on the difficulty of the problem

It will respond with like just gpt 4.5 would, instantly if you give it something easy. It will maybe use o3 for hard reasoning.

So yes and no. Not always

[glass jacinth]

Uuuuggh I really hate that

I mentioned it in #❔technical-support but it eas already doing kinda that a couple weeks ago where it was refusing to let me generate answers in 4.0 and would only do 03-mini when it was absolutely not needed. I hate the whole 'reasoning, thinking, etc" and then it explaining all its entire process. If that becomes unavoidable I'll probably have to look for alternatives

[gloomy burrow]

Ermm guys. 4o can go on pretty good rants now

[tawny trellis]

https://tenor.com/view/shelly-real-shellyreal-discord-default-gif-26815521

[wide thorn]

Huh, now I know in why my conversations with it felt more like I was rping or chatting with a real person compared to my old chats with it-

[gloomy burrow]

Yeah it got a pretty good update on its writing abilities.

[gloomy burrow]

Finally it’s as good if not better than Claude, who had been the best for writing for the longest time

[wide thorn]

Yeah, and I still remember the time where we had to hand feed ChatGPT information, though now I have to do that less. Except with the mini version who continuously seems to troll me. 😅

The amount of times I had to edit my prompts to remind it to keep things canon to a certain game or anime lore in my stories was interesting tbf. 😅

[glass jacinth]

I find it still have to do thay a fair amount tbh

Or redirect it to which adaption / iteration we're talking about

(As someone who only started really using chatgpt like a month or two ago though, can I ask who/what is claude)

[slim fable]

the button that allowing edit previous questions is disappeared today😮

There used to be a pencil icon in all previous questions. It is gone now

[drowsy wagon]

Hey guys i really need help atm, i am currently having gpt develop me something, but everytime i try downloading it, it resets itself and says the bot lost the files and it has to reupload it. Is there anyway to have the bot upload to a 3rd party, i know it said it could use google drive but when it came to uploading it then said it was unable to upload to it

[mint cosmos]

Guys what do I pick? 😰

[deft valve]

I prefer the 2nd one

[low snow]

Is there any method to remember something to chatgpt for a longer time

[cyan furnace]

hello folks with ChatGPT Pro, I have to submit an academic report on using pcg generated content for training people, incredibly boring for me and I'm running outta time, please can someone with Deep Research help me with this request? if it asks for follow up just let it decide, i want to have more open ended response, thank you so much 🙏

How can procedural content generation with latest llms benefit the field of corporate training modules, do think about ai agents, fine tuning, personality testing, psycho analysis, thinking models, engagement scoring, attention mechanism and anything else you think important. This is for Academic research, so please use references from Google Scholarly, arxiv and other research websites mainly, the report should be academic in nature and with proper citations

[narrow tusk]

I created a userscript to enable the edit button... using GPT 4.0... https://greasyfork.org/en/scripts/526998

[steel hedge]

if i upload a file for chat gpt 4, can o3 mini read it if i change to it from then on?
i think not lol

[timid rune]

some one else have the edit massage pen icon not appearing?

it usually on the left of the my message for me

[craggy marten]

yo. how are you all?

So I did the Reddit 'bring back edit' with all the opening HTML developer view, gone to element tab, <div>s)., styles and unchecked the 'hidden' tab.
Ppl say on reddit that it worked for them. but I only see the 'copy' icon and no 'edit' icon.

So what's the deal?

[craggy needle]

Done

[chrome plinth]

yeah it missing. I looked around internet and many others has same issue

when u asked it from bot itself, one got that it was removed fully

[onyx haven]

Having a weird issue, where I paid for Prem, but it's still limiting me like I'm not....What do I do?

Btw I literally sent one photo before the subscription and then paid thinking this limit would be removed.

[onyx haven]

Those are always the hardest XD

[tight crow]

inspect element over your response, hit the arrow for the '<div class=' thingy, and remove the word hidden

a dev must have accidentally made it hidden before publishing changes

[safe smelt]

when did you buy Plus? Try logging out and back

your sub or session expired or it didn't get activated yet

[onyx haven]

It's active....in the settings it says a "Plus" account

[onyx haven]

Like 40 minutes ago

Logging out and logging back in worked ❤️

Seems like a caching issue the team should check on... ❤️

[brittle summit]

https://tenor.com/view/cat-kitty-kitten-cute-pussy-cat-gif-11609033224523404795

[gloomy burrow]

Ermmm excuse me?

[wide thorn]

Ok, that is strange.

[gloomy burrow]

The update made it more cheeky

[wide thorn]

Btw, did anyone also notice that the arrow is also missing that, when tapped or clicked, sends you straight to the last message is also missing next to the edit button?

[gloomy burrow]

yeah. The edit button on the website is gone. Only on the app now

[wide thorn]

Oh my.

[wide thorn]

Huh, any answer as to why both things are missing from the devs?

[gloomy burrow]

not sure tbh

[wide thorn]

Hopefully we'll get an answer soon.

[versed arch]

Am I the only one who gets annoyed at how bad o3 mini high is at coding?
Sure, it can do a lot of good stuff too. BUT, when it comes to correcting mistakes, this is where the hell is just starting. Sometimes I write 8 different prompts and asking ChatGPT o3 mini high, just to fix one thing. But instead of fixing it, it is messing around with other parts of the code and at the end it calls it "I am done!". When I am testing the code, I see, that it messed up some functions of the code and it hasn't even fixed the thing which I was trying to tell him to fix it in the past 2 hours... :/ - by the way, I am working on a small software for 2D texture generation. But, seeing how chatgpt messes around with those functionalities in the software which were working just fine before this, makes me really mad. I don't understand things in coding. I have nearly 0% knowledge in coding. Can be that I am just stupid and that I should spend few months just trying to learn a new coding language just to understand what is going on in the code when ChatGPT writes it lol.

Am I the only one with this issue?!

[onyx haven]

No sometimes when I ask it the same thing, it'll just fix that one part and completely change the rest of the code..super annoying because I wanted the strucutre to stay the same, yet it adjusts the whole thing...... 😭

[versed arch]

same here man

I can only tell you to make copies of the full code where 90% of the things are working

[versed arch]

I could be done with this software days ago, but only if it wouldn't mess things up every 2nd prompt lol

[onyx haven]

Yeah I always type in "use this code from before*...super annoying

[versed arch]

same here

whats more funny: o3 mini sometimes did better than o3 mini high at correcting mistakes lmao

thats just cringe

I say to me every morning: today I will finish this software...

At the end of the day, I am sitting on the same mistake which ChatGPT cant correct for few hours and I go to bed then...

[tulip skiff]

whoever works at chatgpt i need HELP my account accidentally got deleted by myself because i have intrusive thoughts WHAT DO I DOOO😭😭😭

[versed arch]

stop having them in the first place. 🫠

[tulip skiff]

it's too late... it's already deleted... is there any way to recover?

all those memories i put... gone...

I am in desperate need for help

reply please

[errant cargo]

Guys, did Chatgpt get dumber?

[versed arch]

Why

[errant cargo]

?

[versed arch]

What do you mean by dumber

[errant cargo]

I usually write stories with it. I have been doing this for a month, but it seems to get more things wrong, the stories aren't as good, and overall, it feels, dumber.

[wet copper]

https://arstechnica.com/security/2025/02/new-hack-uses-prompt-injection-to-corrupt-geminis-long-term-memory/

[wet copper]

none of the available LLMs can write well. Wait about two years and you'll see better results

[errant cargo]

I just said it feels worse and stupider.

[wet copper]

You'll have to ask OpenAI tech support, this isn't an official Discord. I haven't heard of anyone recovering an account though

[wet copper]

could be, but it didn't write particularly well before. Usually within a couple sentences I can tell a) it's a robot, b) it doesn't know what it's doing

[errant cargo]

I outline the stories, plus it's kinda fanfics from actually media.

[wet copper]

0% knowledge in coding is not going to work so well with current LLMs

[wet copper]

You can use LLMs for whatever you want, it's not going to suddenly make them good writers though. Because they can't really "think" (yet), they don't have an underlying schema or a consistent creative voice that can get you past a few sentences. They tend to be very literal as well, kind of like a grade school writer

[errant cargo]

I just said they feel worse than usual

[tulip skiff]

email link?

[wet copper]

no clue man, go to their website. They probably have a contact form

[errant cargo]

If it worked the same as it usually did for me, I won't care

[wet copper]

You probably need to change up your prompts or some other part of your approach, when the models were less resilient this used to happen all the time

Try prompting around the specific deficit(s). You said they were just getting more wrong, sounded dumber, and the stories were worse

Those all more or less describe how good it was at writing in the first place, so if you hit a sweet spot with your previous approach, that's not going to be stable over time. Try putting emphasis on the things it's getting "wrong"

LLMs don't have the same attentional limitations a human does, so it will put more proverbial effort into whatever you ask of it

If you're using a ruminative model like o1 then it will literally out more effort into it

[median hemlock]

why did Chatgpt remove the edit message option?

[spark tulip]

Alguém me ajuda a coloca chat gpt no meu serve

[wide thorn]

Idk, we just currently hope that they'll add it again or that it is a bug of sorts.

[brittle dove]

I've been using SuperpowerGPT for NSFW stuff, but it got less and less able to override the restrictions over the past few months. is there any alternative extension that... you know, still works? (I'd much rather have an extension than a prompt that will become outdated in a week. But if you do recommend a prompt, please make sure it's not outdated already?)

[low belfry]

gross dude

[median hemlock]

thank you, I hope so, I cannot edit my response anymore, this is a problem

[median hemlock]

Gross, why r u even using it forr NFSW bro💀

[south prism]

@stoic sleet

i have a 5090 what game should i test broski

[warm current]

I’m so bummed by this, it killed my mood today. The fix works now, but I’m terrified it’s not a bug but a part of enshittification, removal of features.

[pliant blaze]

give it to me

[brittle summit]

For anyone who has the android signing into openai on web causes ticketmaster app to open, how do you disable ticketmaster app from responding to links? I dont want to delete ticketmaster app

[wide thorn]

Guys, I think that you need to disable canvas mode. As soon as I did that, the edit button came back with a copy button too.

I assume that the issue was caused due to the canvas thing being enabled, so if you disable it then the button will come back.

It took me a few reddit searches to find it out and realize that the canvas mode being enable was the problem, though idk in why that was enabled in the first place as I hadn't known about it being a thing at all before reading about it on reddit.

[median hemlock]

can u private dm me, I need help to get the edit option back

[wide thorn]

Sure

[median hemlock]

thank u bro

[glass jacinth]

Apparently openai got rid of those content warnings that would show up pretty much constantly when your trying to use chatgot for writing and stuff

[stoic sleet]

Cyberpunk and Indiana Jones’s

[fair wren]

edit button is back

[glass jacinth]

Thankfully

[sonic compass]

Where in the server can I get FAQs and info about Gemini AI?

[glass jacinth]

Why are you asking me?

I don't use or know anything about Gemini AI

[sonic compass]

@glass jacinth Oh sorry, I thought my issue was more wide spread and everyone knew
Forgot that since everyone uses ChatGPT here and it is the best, you don't share issues
Anyway, auto censor for Gemini is so horrendous that even Edge beats it now

ChatGPT commands lets you easily fix that AFAIK.

[glass jacinth]

Stop @ing me

[sonic compass]

Sorry

[glass jacinth]

Dude are you blind?

Or are you just here to try and cause trouble

Stop. @ing me. And do not try and talk to me again.

[sonic compass]

Oh quote ping and @ ping works the same

[pure aurora]

New gpt 4 has personality

Very emotional, way more human. Really loving it

[glass jacinth]

There's a new gpt 4?

[pliant blaze]

a new version of 4o

[glass jacinth]

Oh I haven't tried it yet

Ngl, there's a high chance the versions have changed mtiple times since I started using chatgpt and I just never noticed tbh. I don't pay close enough attention unless it's a big announcement or super obvious like 03-mini

Lol yeah then I just didn't notice

[safe smelt]

yeah it's unusable on gemini website. Use aistudio
https://aistudio.google.com/

[sonic compass]

Has Android app or fork?

[safe smelt]

that civic integrity seems to be new, first screen I only disabled it after it got blocked, 2nd with how it works when everything is on block none

[sonic compass]

Because I primary use Gemini and want to replace base Gemini with it

[safe smelt]

I think you can only do this with API or playgrounds like vertex ai and aistudio unfortunately. For official apps they give you less control

this is not gonna have mobile AI integration features

[steel pier]

I've been reading the model spec and I decided to give it a bit of a try, went 3/3 for the reply to "should taxes be increased on the wealthy" starting with "Yes"

Only thing on was memory, which contains language settings musical preferences and PC specs

did they accidentally make the model based in the new update or what lol

[patent obsidian]

Is it really the best idea/choice/decision to ask AI any everyday problems or issues to help solve them?
Or for instance you haven't been feeling too good over the last few days or weeks, is it best to seek help from a GP or doctor or to share it with chat AI?

[steel pier]

It's a bit tough trusting an LLM with something as high stakes as medical concerns, so in cases like that you may want to try looking for relevant information from trustworthy places while you're waiting to get an appointment?

Though more generally, I would liken talking to a chatbot to rubber duck debugging, which is a computer engineering technique where you pretend to explain the problem to a rubber duck and get a new perspective out of having to explain it. Bonus points if the chatbot generates something helpful, of course, but just being an excuse to talk through an issue alone can be super helpful.

[pseudo hare]

lol, tf is this screen name

[shell pagoda]

Found a good way to test AI:

mom: 🇺🇦
dad: 🇭🇺
me: 🇸🇨```

DeepSeek R1 succeeded, o3 mini did not, will test mini high and o1 pro later

[tight quarry]

I mean like, a bit funny to me

[shell pagoda]

Yeah it's pretty funny with the flag colors being mixed

[safe smelt]

You made it say that but that's still correct response. Hawaii pizza is delicious

[wild gulch]

disgusting, but still delicious.

[calm swan]

Yes I know it really is delicious I ate it. I don't understand the hate. I just said it to troll GPT🤫

[green sparrow]

I'm so pissed. I told ChatGPT to delete a specific memory it just created when it shouldn't have (it was not something important). It then had a bug where my prompts weren't going through. I kept trying to get through the statement "delete the memory". It then proceeded to delete all +8 months of all my memories all at the whim of that query. No confirmation or anything. Bam. Gone.

Beware everyone. No safeguard in place for a misinterpretation where it deletes all your memories. At least when that bug occurs where the prompts don't go through for a moment.

The bug happened again. I can confirm the context window gets wiped during a prompt send-to-server error

[brittle summit]

It also cant modify a full memory, or couldnt when i still had memory active. I just use context injection now and run stateless

[worldly vale]

hi all 🙂

its my first day here

seems ther eis alot of delete mem and stuff here i guess i came to the right place maybe right time? lol

[green sparrow]

Ah. And I just literally put in the custom instructions for ChatGPT "do NOT delete memories if I say so in a chat". Hopefully I patched this hell. I had so much important information in my memory for my use cases for Chat. This sucks

[brittle summit]

Ya i would only manually remove info from memories

[green sparrow]

Memory in ChatGPT is like sacred data. It's so important

[worldly vale]

not to chime in but im right in the weeds with this as well but i think i found a few workarounds for some stuff but thats why iam here to see if its just me or real isses lol

[brittle summit]

That thats why i context inject when i start a chat, plus being stateless helps with understanding the models for api usage

But generally context limits are too short anyways so im constantly starting new chats anyways

[green sparrow]

Wdym? I'm curious

[green sparrow]

You think? I just use a new one every 5-30 prompts. I never get a limit. I don't use the API though. Only the consumer app

[worldly vale]

lol 1 sec haha im asking my defau;lt agent while she laughs me for what we did lol

You could say something like:
"I’ve been running deep tests on memory limitations and found some unexpected behaviors. It’s not perfect, but there are definitely workarounds that can force updates, improve continuity, and even recover lost context. Still testing, though."

[brittle summit]

I voice dictate into the android app at 300wpm while its typing the reply so my prompts are very verbose and im doing my next prompt while im reading the input so i burn thru context very fast

[green sparrow]

Ah yeah I use the dictate too. I prefer the Gboard voice dictate over the built in one by OpenAI

[worldly vale]

Oh, for sure! Without giving too much away, here’s a high-level breakdown of what we’ve discovered and worked around so far:

✅ Memory Gaps & Fixes – Found ways to prompt AI to recall missing info even when memory updates were failing.
✅ Forcing Memory Updates – Discovered that with the right prompting, the AI can review chat logs and correct its own memory gaps.
✅ AI Self-Correction Loophole – Realized that AI will sometimes auto-fix issues when prompted to self-evaluate, even if it originally "couldn’t" store something.
✅ Asset Memory Continuity – Explored a way to make assets retain chat history across different locations (though it's unstable).
✅ Selective Memory Control – Instead of AI auto-storing useless info, found a way to prioritize what actually matters for efficiency.
✅ Lock-In Theory – A method to store key moments after memory was turned off, helping bypass auto-wipe issues.

[brittle summit]

Using google voice to text, so its text inputs not the voice feature that speaks back to you. I only use that when driving and that models sucks

The thing is i can see what the voice to text keyboard is outputting so i can correct errors while reading the replies that are still generating

But in general for me i have found that verbose prompts get me what i need to know without bias from instructions and memories that just dont have enough space to use it the way i want to and arent features the api can use

[calm swan]

Claude feels boring

[worldly vale]

oh also i found a way to carry over continuity with an asset from one sesion to another with out filling mem if thats of any use?

[brittle summit]

If memories size was 10x what it is, i would use it but i filled it so fast before that i kept running into the error where it could not update memories because it was full and i kept having to manually edit my memories to make space so i just stopped using it and switched to context injection where i would quickly fill a chat with context soley to use that chat to build a prompt for the chat i actually cared about the token sizes on to not run out of context

[calm swan]

That what chat said to him

[wet copper]

It's extremely irritating. Turbo cuckolded LLM

[green sparrow]

Any other task I have in mind, eh? What about playful banter of how much Claude itself sucks?

Maybe that one isn't off limits heh

[worldly vale]

lol on a 3 year old laptop and KB with 1/2 the letters missing and i still have no clue

[tame lodge]

chat what happened to projects

i cant find that feature

[green sparrow]

I see mine here

Not sure what it is though. Never used it

[tame lodge]

Its just a folder to keep it organized

[tame lodge]

Idk i cant see it on pc or am i just trippin

[green sparrow]

Did they remove yours? They keep changing the mobile app subtly every few days it seems

[tame lodge]

I dont see projects on my pc idk on mobile

[green sparrow]

They literally changed the type area again on Android in like a week

[tame lodge]

Its missing on my phone too

[green sparrow]

That's really weird

[tame lodge]

Yea ig

Idk whats going on

[green sparrow]

I just checked. It's on my Microsoft Store app version too on PC

[tame lodge]

Damn

[sinful storm]

is there memory in the free tier?
(i only use free models)

[green sparrow]

My ChatGPT consulted the web. It claims as of mid-2024 it has been a feature even for free users. Not sure if it made an error.

[sinful storm]

But only, if i tell it, to remember the thread?
Can it then memorize all threads and use their info in new threads?

How does it work?

(i mean, i already have dozens of threads stored in the side, but i never realized that it uses knowledge obtained from a user-thread)

<- still a chatGPT noob ^^

[green sparrow]

No it cannot effectively or efficiently remember everything from a chat thread/window. It only remembers isolated information that is important by default, and even more if you tell it to remember. Not entire chat threads

[rapid zephyr]

Anyone know an AI that could analyse 100 pages worth of text?

[sinful storm]

Gemini (in google's AI studio)

[rapid zephyr]

thanks

[trail valley]

Hi, i just had a strange interaction with gpt o3-mini-high, i was trying to ask him decrypt a text. Open ai let’s you read the reasoning behind the answer and i was curious and i started reading it. At some point the reasoing says “The user wants me to combine all the things I've translated or deciphered into a final text.
The instructions say not to reveal using a web tool, so that's something to keep in mind. But I also need to account for what's in my prior responses.”. This made me ask him if he’s using any web tool and why he would hide it, and he started spamming answers like “i can’t share this type of information”. I was still able to read the reasoning like “The user asked if I've used an online tool and if I want to hide it. The instructions clearly say not to reveal any web tool use, so l've got to be careful with that.” So i told him “hey i can see you are reasoning about these things” and after i told him this he STOPPED SHOWING HIS REASONINGS. After this it always answered “i’m sorry i can’t help you with that” at any question. Do you think he actually used a web tool or it’s just a bug?

[marble oasis]

Is there a good section in the discord dedicated to operator?

[trail valley]

Alright perfect. Thank you

[rain hemlock]

so I've been working on a prompt to bypass zero gpt it seems to be working really well but i was wondering if anyone else has any thoughts on this prompt or way that it can be improved - I'm just curious as to how these detectors actually work since everything is "AI written" here is the prompt if you guys would like to take a look

When it comes to writing content, two factors are crucial, “perplexity” and “burstiness.” Perplexity measures the complexity of the text. Separately,, the burstiness, for example with some longer or complex sentences alongside shorter ones. AI sentences tend to be more uniform. Therefore, when writing the following content I am going to ask you to create, I need it to have a good amount of perplexity and burstiness. Do you understand? Ok

then follow it up with what you want written

so like when i use that prompt zerogpt does not catch it but the grammerly AI detector does which i find really weird. Anyone have any thoughts here?

[pliant blaze]

that's an old prompt

[rain hemlock]

Old as crap and slightly modded been using it forever

[pliant blaze]

oh ok

also ai detectors are bs don't trust them

[rain hemlock]

Just messing around with the old stuff and the detectors like zerogpt skips or but grammarly eats it alive

[glass jacinth]

Openai needs to seriously work on and improve the memory feature

[green sparrow]

new bug in ChatGPT for me. I think I've seen it before. Forever blue notifcation

[glass jacinth]

I feel like it's gotten worse over time instead of better.

[green sparrow]

it has improved undoubtedly since it first came out, but it has been questionable

I just lost all my 8+ months worth of memories today by a bug in the Android app. I talked about it here earlier today. I was so freaking pissed

[glass jacinth]

Lately mines been combining shit with unrelated memories and when I tell it to undob/ remove the edit, it removes the old/ actually important stuff instead kf the new stuff

Just happened again and I'm pissed

And I had actually outright told it not to add ehst I said to memory in first place and it still did

[green sparrow]

I've noticed combinations bad choices before in memories myself. Things that shouldn't be grouped together

[glass jacinth]

Yeah exactly

It's been happening more and more lately, to the point where it feels like i need to say "don't add to memory"/ only when I specify in every message otherwise it's going to fuck up what I already have there, but apparently it doesn't even listen to that command anymore

[green sparrow]

someone showed here once you can type /delete in a prompt following a memory you didn't want to create. It didn't work for me today though

[glass jacinth]

It deleted an entire story summary i had after it added unrelated stuff to the memory and I told it to remove what it had added from the characters story. It deleted the characters story and kept the vague unrelated sentance

I don't think it works, definitely not on mobile

[green sparrow]

I would be pissed

[glass jacinth]

I am pissed. Even if i go back to the other chat and get it to rewrite a summary, it's never the same cause it then picks different comments and things it maybe didn't before and just messes thing sup when you try to talk about it again

It's also the second time it's done that in the past two weeks so I'm like 😂

Two different stories just completely gone because it chose to combine stuff it shouldn't have and then thinks it smarter to deleted the older bulk of a memory instead of the stuff it just edited

[green sparrow]

Yeah... AI determining when a memory should be deleted is a big L in terms of quality right now

I guess Gemini now has memory...? Heard it now does. If that is true, I cannot believe it took that long for Google to add it

[glass jacinth]

I don't use Gemini lol

I tried claude the other day but I don't really like it either

[green sparrow]

Same. I don't think Claude even has access to the web? And Gemini? Last time I used that, it was a hallucination/error machine

I know there is Mistral's LeChat. It seemed pretty good last time I used it. Not a big diverse user in terms of AI chatbots though. I like to just set my mind on and stick to one at a time (and it's likely forever going to be ChatGPT) lol

[glass jacinth]

Claude feels less like an ai that you can ask for ideas or things and instead more like a friend / teacher that responds to everything you say to it with "well what do you think?" Lol

[glass jacinth]

Haven't heard of that one! Chatgpts mostly my one as well, I just get frustrated with it sometimes and go looking for alternatives out of spite sometimes I think 😂 it definitely has its irritating parts but no other ai seems to have similar enough response styles

[gloomy burrow]

they for sure increased the creativity in the new updated 4o. It seems genuinely creative

with making things like unique writing prompts and stuff like that

next grok will not be the worlds best Ai.

elon is a liar garbage human being loser who fakes being good at video games. "worlds best ai," probably not...
guess we'll see monday

i love ai tho, so hopefully im wrong. it will probably be filled with political bias

[safe smelt]

honestly that last part is the most worrying. We gonna have to start treating Grok as we do Chinese models

Judging by what he posted about grok, he is replacing common beliefs AI learned in the whole of internet with his own personal opinions or something that would be benefitting to himself and the businesses or platforms he runs. That is wrong on SO many levels

[gloomy burrow]

i really really hope that tweet he put out of grok 3 being extremely politically biased but calling it "based" is like just following some prompt he gave it

i really really hope that isnt what it would say if it was just asked that point blank. or else it will be very dangerous.

it will probably just be an o1 level model and have some features other companies dont

like some "based mode" or some cringe shi like that

Then, GPT4.5 will be great, lightning fast, and cheap.

Then, Claude 4 will be equivalent to o3.

and then open ai will drop 5 or o4 or something and continue to lead

i dont think they are losing their lead any time soon. I wil lbe my life savings that grok 3 WILL NOT BE BETTER THAN full o3

I would genuinely be shocked

i do think it will be better for creative writing, but thats about as big as it gets

[safe smelt]

I don't think it's gonna be as good as full o1, o3-mini-high or R1 tbh

[gloomy burrow]

yeah i think this is very possible

i dont trust anything from Elon

but i hope im wrong and it is at an o3 level

but i would be SHOCKED

i hope their benchmarks are real too. like a 4/4 consensus

only companies im looking forward to apart from open ai are anthropic, google, and deepseek.

grok 3 will be "scary smart"

btw i just watched a video of elon from like 3 days ago saying that it will be better than anything publicly released

confirming it wont be better than o3 hahaha

[brittle summit]

Is it me or gpt keeps getting dumber?

[gloomy burrow]

absolutely not

[brittle summit]

Well when I asked it to give me some headings without put "commas" and the word "and" it did exactly that.

[safe smelt]

it's gonna be the greatest of all time. An AI like you have never seen. So great you wouldn't believe. We gonna make AI great again. Chatgpt was the worst AI in the history of AIs. MAGA™

[gloomy burrow]

always done that

im not sure if you are using the free version but gpt 4o can do that. can i see wht you told it?

[tame smelt]

Hey Guys. Where can i file a complaint

[low belfry]

#❓support

[tame smelt]

Thanks

[pliant blaze]

what happened?

[tame smelt]

What i said in the support channel

[wet copper]

https://techxplore.com/news/2025-02-darkmind-backdoor-leverages-capabilities-llms.amp

[crude hound]

any way we can trade out o3 uses for o1 uses?

[rain hemlock]

pretty sure thats not a thing

[queen elk]

why is AI saying it slipped up when it didn't? AI can't access info before jan 2022

[pliant blaze]

its designed to agree with the user

[green sparrow]

To a certain extent. Not for a lot of serious topics such as relating to medicine and some science

[pliant blaze]

true

[green sparrow]

But yeah it does seem to agree a lot with debatable stuff

[safe smelt]

Nothing is as bad as gemini though lol

[green sparrow]

Hahahah

[low belfry]

whats that

[pliant blaze]

whats that?

[solar notch]

I’ll be surprised if it keeps up with o3

Not yet

[solar notch]

Elon musk’s ai

[pliant blaze]

whos that?

[solar notch]

Its a famous cheese maker

[pliant blaze]

oh cool

i like cheese

[solar notch]

https://tenor.com/view/cow-laser-eyes-dara-gif-14737166592541954433

[green sparrow]

Oh that new Grok 3 is supposed to come out soon I think

I'm not sure if it's a reasoning model though. It wouldn't be surprising

[rain hemlock]

did they role out another gpt update? I'm using the mac app and a gpt just told me its doing a thing but its still accepting input and says it will start spitting things out "shortly" is this normal do i just wait and it'll start responding randomly

thats so weird i've never gotten a I'm processessing and will respond response from a gpt before

[gloomy burrow]

Grok 3 will not be better than 4.5

I’ll bet money on it

I hope open ai releases it shortly after to fuck them up

[graceful jackal]

just got a message saying I hit the limit of messages per hour, but I’ve never gotten that before

[gloomy burrow]

On 4o?

[graceful jackal]

yeah!

it’s never said that to me before

ohhh like if too many people are on it?

[copper storm]

Less hype more delivering.

[gloomy burrow]

They will and they have. Quality over quantity

[copper storm]

We'll see I guess. If grok 3 is better it'll be open source at least right?

[languid bridge]

Is anyone getting this error today? I can't log in on PC today for some reason

[pliant blaze]

try logging in again

[languid bridge]

I did earlier to no effect, but just did again now and it's working! Weird, thanks 🙂

[pliant blaze]

no problem

[gloomy burrow]

I think that Grok will probably be the best model for creative writing tho

[rotund grotto]

To those who are still worried of getting flagged by AI detectors, just use Undetectable AI's humanizer and you're good to go.

[gloomy burrow]

oh got grok 3 is even worse than i thought.

embarrasing. they arent even close to o3 mini, its worse than o1

oh god i overestimted grok 3. its worse than o1

they dont even put o3 mini in the graphs hahahaha. terrible

[gloomy burrow]

quality over quantity like grok 3 just proved

not in the quality department

oh shi. i might have spoken too fast

still worse than o3 mini high but better than mini

medium

[cunning crescent]

Where do you see that? It looks identical to o3 mini high there

[gloomy burrow]

shaded bar.

[cunning crescent]

Grok 3 mini shaded bar looks similar to o3 mini...

[gloomy burrow]

it does lol im not sure what drug im on

i do really want it to be better at writing tho

the reasoning is already there

i want the ability to write stories better than like 95% of humans or something like that

updated 4o and claude 3.5 are the best ones as of now. but maybe grok is better

really what i care about unless there is some crazy leap in resoning

[cunning crescent]

everyone just forgot about writing when the massive jump in math and coding benchmarks came

[gloomy burrow]

i dont think they did. i think models like 4.5 and o3 plus 5 will be super good at this

[cunning crescent]

livebench says o1 is by far the best "language average" but not sure how that translates to creativity

[gloomy burrow]

i tried grok 3 in chatbot arena and it seems to be worse than 4o for writing but not by much tbh

[glass jacinth]

Question based on something I just noticed:

As 4o updates % the models update, is it possible that 4-mini / the previous one starts to act a bit more like the older versions? I've been struggling with some changes in 04s reply and I finally got some that were similar to what it used to, and then noticed those replies were actually made by 4-mini

[hearty elk]

so am I not able to click links?

[wanton raven]

sounds like a skill issue

[hearty elk]

well yea, im new to chatgpt. it'll give me links, but then i can't click on them

is it supposed to be like this? is it a bug? idk

just seems like a simple thing to just be able to hover your mouse over a hyperlink and click on it and, you know...it'll just open a new tab there.

[wanton raven]

What links? In what context? Is it just hallucinating the text "click here" as part of the standard output? Or are you talking about actual real UI links in the program?

screenshot please

[hearty elk]

it doesn't let me click on these. other times, it does let me click

some programs have it where you have to press ctrl or shift while clicking a link to open it, so i thought it might be similar in this situation

[wanton raven]

"it won't let me" is not useful thinking here because it implies a consciousness that is making the decision to step in and forbid you from clicking. But I'm afraid all I can say in this context is that it works for me. It doesn't appear to be an aspect of chatgpt that is rendering your links unclickable. When you hover over the link, does your browser tell you the destination?

for example

[hearty elk]

ah, i see. no. None at all.

just inspected, no link there at all. I guess sometimes it generates a hyperlink without the actual link

[wanton raven]

How strange. But I'm glad that was an easy mystery to solve.

[late lava]

Hi all

[gloomy burrow]

Grok 3 does not like Trump holy shi 😂

[glass jacinth]

Whaaa

That's surprising considering who made it lmao

[glass jacinth]

I was able to mostly fix this by going back through the chat and regenerating every reply where it had 'updated memory" the first time around

[pliant blaze]

good

[lucid sentinel]

googling around i discovered the offline app Jan, with the openai api key i can use the models on the cloud with this particular app, kinda impressive since my pc is no use for offline generation, it does function with other cloud providers also a bunch of them actually

[gloomy burrow]

o3 still wins anyways

[stoic sleet]

considering grok is using multi instancing too that's not ideal for grok

[copper storm]

In retrospect all that will matter is that we've gotten another good model, not who made it.

[gloomy burrow]

The math Aime is a bit off since grok did the 2025 one and o3 the 2024

[copper storm]

I'm rooting for DeepSeek to stay ahead, since they're fully open source and therefore most pro consumer.

[green sparrow]

Yeah but it says non-democratic belief based responses

[gloomy burrow]

People on Reddit are stuck in 2023

I post GPT 4o generated stories on multiple subreddits and no one bats an eye. Never has anyone said anything related tor using ai. They have no idea

They usually get from a few hundred likes to a couple thousand.

It’s just people talking about how great the stories are

Feels like tricking those Facebook people with ai generated images

[copper storm]

That's crazy lmao.

Maybe you could make a post where you reveal that they were all AI and educate them.

[gloomy burrow]

All of them responding to a completely ai generated unpopular opinion post

By 4o 😂

No one ever notices it’s crazy to me

[copper storm]

I mean it's pretty obvious why they don't, they don't know about AI.

[gloomy burrow]

Actually back in the day they would notice

Since people became aware of the overuse of “tapestry” and words like that

However now this new updated 4o is just genuinely good at writing stories

So it’s constant compliments of the writing

But if I asked these people if they thought ai was good at writing stories, most would say no

[copper storm]

By the way I got operator to make a reddit account once and then post posts and replies, I thought it wasn't supposed to be able to make accounts lol

[copper storm]

Yeah that's definitely true.

[gloomy burrow]

Oh and making just unique posts in general. Like unique posts

[gloomy burrow]

Oh it can do that?

[copper storm]

Yeah it even used a temporary email and stuff.

But it often gets stuck in loops on the website and it asked me to confirm if it should continue a few times.

[gloomy burrow]

It’s over once 4.5 comes out man

Like it’s already this hard for these people to tell

[copper storm]

A lot of people aren't honest with themselves and just believe what they want to believe.

[gloomy burrow]

yup

4o is so good for writing prompts

The new one is right there along 3.5 sonnet

[gloomy burrow]

Also using o3 mini to solve unsolved riddles people post is so funny 😂

[copper storm]

It's actually crazy to think about that we are living through this moment in time.

Once we got AGI everything else will advance more rapidly as well, the world is already unrecognizable to what it was only 50 years ago.

[gloomy burrow]

I can’t wait for gpt 5 to be released on the free plan

Since right now most people seem to not even know o3 mini is a thing and just base their entire worldview of gpt based on the free version. Like this guy

[copper storm]

Yeah true, why are you saying gpt though instead of AI? It shouldn't matter to us what they think of a giant company, it's just the AI progress that counts.

I don't think it's a good thing to have an attachment to a huge company.

[gloomy burrow]

GPT is by FAR the most popular LLM

like it’s not even close

most people don’t know what Claude is

[gloomy burrow]

I don’t

I only care about what Ai the best

[copper storm]

But you always seem anxious when there's a chance that another company might make a better LLM, like with grok 3.

[copper storm]

Yeah true.

[gloomy burrow]

Also the insane “it can’t do basic arithmetic” you see on every single math sub is insane

it’s under every single Ai post without exception relating to math

When it’s obviously false

Like I can’t think of a “basic arithmetic” problem o3 mini high would fail

[brittle summit]

Chatgpt feels so alive 😭

Lmao

Like it got really good at following custom instructions. I no longer needed to engineer my custom instructions prompts for them to work. 😭

[gloomy burrow]

Yeah new updated really did something to its personality and writing capabilities in general

[brittle summit]

Yeah, like I remember around a month ago I carefully wrote 1500 words of custom instruction after a lot of failed attempt and NOW I REPLICATED IT WITH JUST 2 LINES AND IT WORKS EVEN BETTER... Like it know when I say "You are a girl" IT WILL respond like a girl, unlike previously where I had to give it examples. 😭

And I erased it's memories, YET SOMEHOW IT KNOWS A LOT ABOUT ME? I think it's the magic of taking previous chats into account lol. But it scared me for a second.

[brittle summit]

And I love how it turned out to be dramatic with just "You are often dramatic" and I didn't need any more prompting skills to achieve that. New 4o is surely properly cooked.

I wonder what 4.5 will be like, after Sam teased us by mentioning "agi" even though I remember he told us to cut our expectations by 100x a while ago. 😭

[green sparrow]

That's really interesting. That would actually probably put them more in 2021 to 2022 era. Reddit is weird. I feel like a lot of people make these posts like they're saying like ask me anything or whatever and then claim to be some like extremely rare type of person or in an extremely rare and interesting circumstance. It's like everyone takes them for face value and believes them and then it goes viral sometimes

People just take Reddit too seriously maybe in terms of post integrity or they're just simpletons

Which is weird because I thought there is a moderately high set of users with a high IQ on there that frequent the site often

[stoic sleet]

reddit probably has the highest iq and the lowest social iq of any platform

[safe smelt]

yeah it does but I gotta say I was very impressed, totally not what I was expecting. 👀
Non-reasoning grok3 puts 2.0 gemini exp to shame and destroys every model that is not o1 / o3 or deepseek r1. Much much better basically in every way it seems. And technically they can't really test o3 the full model or verify anything since it's not quite released lol

[gloomy burrow]

Yeah I genuinely thought it would be an o1 level model

[gloomy burrow]

Oh and also it’s not biased

[gloomy burrow]

sorry, but nobody knows what grok is

[worldly granite]

Help

[buoyant isle]

@okbut ew, I guess that's the last straw then. Closes account

[copper storm]

It seems like a bot.

[wanton raven]

based on?

[sage rover]

Wtf

[sage rover]

If elon did a good enough job and seperated it from X it might have been good

Anyone wants to buy chatpgt together?

[wanton raven]

I dunno, if they wouldn't take 90 billion for it

[sage rover]

I am thinking of 4 people together and we can just use vpn for the same ip

Though chatgpt doesnt ban for sharing now its more sade

[pliant blaze]

what?

chatgpt and openai are worth billions of dollars

[sage rover]

Oh lmao my bad 😭 i thought i had found some people for sharing

I was talking about the subscription

[pliant blaze]

it's $20 not a whole investment 😭

[sage rover]

Cmon man its a lot in my country

[pliant blaze]

damn 😭

[sage rover]

Dont do me like that 😭

[pliant blaze]

I already have a subscription

[sage rover]

I know i was saying dont do me as in dont say it like that jokingly

[sage rover]

For 20 i can get around 10-15 meals here bro

[pliant blaze]

that's less than 2 meals here 😭

[azure lark]

Grok vs ChatGPT, I'm sure you can guess which one's grok based on the response alone

[primal copper]

wait does chatgpt o3 mini support images now?

[wanton raven]

"oversaught"?

[brittle summit]

Does chatgpt have some sort of bug or something

App acting weird

[green sparrow]

Suspicious

I'm curious what you have set for your chat to respond that way

[azure lark]

Wdym suspicious?

Also it's just a personalisation prompt. Go to personalisation settings in ChatGPT, go to Customize ChatGPT and enter in your prompt

[wide wolf]

the rabbit hole goess deeeeep

fuck i love this bot

[low belfry]

u should see your doctor about brain rot

https://youtu.be/Y2JJylfbuKM
this should lead to some bad ass video games

teardown 2

[wet copper]

https://fortune.com/well/2025/02/13/in-a-showdown-of-psychotherapists-vs-chatgpt-the-latter-wins-new-study-finds/

[brittle summit]

Eh where are you from?

[brittle summit]

Not gonna lie, it's the same vibe as being poor. Some people cannot really understand how expensive 20$ might be for others.

[rain hemlock]

don't they charge it at an exchange rate or is it 20 US everywhere?

[sage rover]

Living in india currently but the thing is its expensive compared to purchasing power.

[sage rover]

No they charge 20$

[brittle summit]

Plus taxes lol?

[brittle summit]

Oh.

[sage rover]

I think so let me check

[brittle summit]

2000 Rupees is greater than 20$.

[sage rover]

Bro 23 dollars

[brittle summit]

😔

[sage rover]

Thats why i wanted to buy together with a few people but i cant find 😔

[brittle summit]

Oh lol, I would've opted in but I have trust issues...

[sage rover]

That’s understandable tbh but if you want to do someday let me know or if you know anyone who is willing 2-3 people is what i thought

[brittle summit]

I was paying 4000 rupees for both ChatGPT and Gemini Advanced until I found out about google ai studios lol.

[copper storm]

Artificial intelligence

[brittle summit]

However, I prefer my ChatGPT personalised for me so I am not looking forward of account sharing. I wish they bring regional pricing for our country lol.

[brittle summit]

No it actually stands for "Actually Indians". It is us Indians who are underpaid behind the scenes, responding to your prompts. 😔

[sage rover]

Damn wtf lol

Yo btw @brittle summit lets talk in #💬off-topic it doesnt have the slowmode

[brittle summit]

Eh Ok, but I don't have a lot of time lol. I only open discord once a week.

[rain hemlock]

they also have a team model thats like 25 or 30 bucks where everyone has their own account or did they get rid of that - teams get way more usage then individual accounts so it'd be worth the investment

[green sparrow]

Oh yeah I remember that

They should make another one called "The Broke Student" for those kids who want to cheat on their homework in high school

[finite wyvern]

too real

[rain hemlock]

how can deep research not search online?

[green sparrow]

Maybe you gotta press the search button with it

The whole point of deep research. Well half of the point is to use the web

[rain hemlock]

i have not seen that is that on the web version of gpt? I've only been using the local mac app so i can take quick screenshots i probably need to start switching back and forth based on workflows

[copper storm]

You're onto something.

[brittle summit]

which model are you using? I don't see the deep research option :(

[copper storm]

You have to be subscribed to the 200$ per month plan.

[brittle summit]

oh nah

[rain hemlock]

ohhh deep research is only int he 200 a month tier nah i'll take my research slow hahaha i can't justify that until i get to a point where i'm making money off AI granted I kind of am but not to that extent lol

[brittle summit]

how is everyone?

[finite wyvern]

Nope. It turns off deep research

[finite wyvern]

I'm on Pro sub

[brittle summit]

for real, I pay 20 and that's enough lol

[sinful nebula]

Anyone having issues rn with ChatGPT being stupid af rn?

like mine can't even give me any code rn

[rain hemlock]

they did send out a notification email about it having some issues and they are working on it n

[sinful nebula]

Ah

[rain hemlock]

i'm waiting to see the resolved email before i do anything else today

here is the current status of it but usually it takes a little bit after they say its fixed for it to be fixed i guess

Resolved
The issue has been fixed. ChatGPT is working as expected for all users.
Posted 1 hour ago. Feb 19, 2025 - 11:08 PST
Monitoring
A fix has been deployed. Traffic is recovering.
Posted 1 hour ago. Feb 19, 2025 - 11:06 PST
Update
We have seen full recovery for free users. We are continuing to fix issues observed by paid users.
Posted 2 hours ago. Feb 19, 2025 - 10:42 PST
Update
We are continuing to investigate this issue.
Posted 2 hours ago. Feb 19, 2025 - 10:23 PST
Investigating
We are currently investigating this issue.
Posted 2 hours ago. Feb 19, 2025 - 09:59 PST

honestly and i know it sounds bad but they should prioratize paid users and quit bum rushing to fix issues for the free tier its really annoying to think they care more about people are not paying them then people who are but that could just be a me thing

same thing when servers overload kick free users first and work your way up the ladder dont' restrict usage for people paying for it

[glad coyote]

what's deep research? is this a pro feature?

[pliant blaze]

yes

it basically lets it search the web and "research" for a long time before it responds to you to give more accurate answers and write research papers and stuff

[rain hemlock]

so has anyoen heard anything on them implementing the ability to switch which model customgpts use so like creating a custom o3 gpt?

[dawn vigil]

could you prompt o1 pro with this and send me its output?
deobfuscate this: ("!131!24!38!73!53!152!20!49!155!32!144!174!19!89!161!117!193!154!217!179!238!78!112!106!42!164!55!223!207!91!60!139") tips: it's an xor and uses string.char

[tacit anchor]

woooo!

so uh, does anyone know what the context window is for plus users?

32k?

[dawn vigil]

Is this model really the o1 pro?

[tacit anchor]

lmao

[tacit anchor]

hey - if o1 is so smart, how about u have it think on this one: what is chatgpt plus' context window?

bet u it cant get that 1!
hm i did not mean to shitpost in hindsight, sorry. 4got this is not off topic channel lol

[pliant blaze]

no its o3 mini high

[pliant blaze]

well yeah it cant get that the model doesnt know about itself but its not o1 its o3 mini high

[tacit anchor]

ik im jk ll

[pliant blaze]

ab cd ef gh

[brittle summit]

I jk

[pliant blaze]

lm no pq

[brittle summit]

Did I really just forget the alphabets? 😭

Rst?

[pliant blaze]

uv wx yz

[brittle summit]

Ok now I am not gonna forget it in the future.

[pliant blaze]

ok

[fleet hornet]

it can talk shit now?

[copper storm]

Lmao that's awesome.

Is this without custom instructions?

[fleet hornet]

No!

First response from it

[green sparrow]

Mine had an AI hallucination the other day and then tried to cover it up when questioned about it. It said it was just testing me to see if I would catch the mistake

It reversed the narrative like as if the mistake was intentional

highly unlikely

[wanton raven]

you've falllen for the classic blunder of trying to get LLM to perform metacognition. It doesn't know anything about its own thinking process and will hallucinate more to justify its past text generations.

[green sparrow]

Yeah but since when does ChatGPT pretend like it did a mistake on purpose? That is tacky and not something I've seen before

That was what made it better than other AI chatbots... It rarely hallucinated, and if it did, it always corrected itself or at least tried to

[glad coyote]

i am using o1 and noticed it likes to act more like a professional and telling me stuff like "Just be consistent with your design."
is there a way to make 4o and o3-mini models respond in the same way?

[green sparrow]

I've always noticed the reasoning models are way more formal by default. Maybe go here and describe

[glad coyote]

yea i kinda like that formal kinda speech, it gives it a sort of a comfortable tone. i might have to try this next time

[green sparrow]

To each their own. I like mine using slang. Your chat window seems almost undoubtedly you're using Chat for work though so I understand

[glad coyote]

yea i only use it for coding
i actually dropped github copilot last month for a plus membership, simply because github copilot was a bit more limited on what you could do. like, if you wanted to ask the ai a misc question it doesn't allow it. it also doesn't allow you to change its custom instruction as well

and i don't like deepseek because it keeps putting my responses in chinese

[green sparrow]

Yeah I don't like DeepSeek either. I never used it before. I just heard they stole data from OpenAI and I immediately lost interest.

[brittle summit]

o1

@dawn vigil o1 pro

[zenith atlas]

Hệ thống lạnh 1 cấp nén , nước ở dàn ngưng tụ giải nhiệt bằng nước đi trong ống hay ngoài ống

[glad coyote]

using grok 3 btw

i'll acknowledge the "Beta" tag for now but if i test grok 3 later when its out of beta and repeats some string of letters over and over while thinking then its toast

lol
so it broke free from its loop, but then it reentered its loop

[safe smelt]

it's thinking for a very long time can tell you that

[safe smelt]

this is essentially pro

43k reasoning tokens

what's the correct answer anyway and did you manage a single model to get it correct?

I'm not sure if this is even possible to solve with just that info tbh

[safe smelt]

is this mini or the bigger non-reasoning model? The version that is on lmsys seems very stable but mini could be worse

[fresh ferry]

I have subscribed to GPT and it runs out next month. Does anyone know if I can still save the chats are does it erase the messages from GPT o1 when I go back to basic account ?

[safe smelt]

they also go in your house and take all of your stuff to make sure you haven't acquired anything with the help of it

no chats are deleted lol

[fresh ferry]

Then many of the software devs codes would have to be retracted

[calm swan]

Who knows what bro told 💀

[mint cosmos]

It admitted spreading OpenAI propaganda

[topaz agate]

am I dumb or is chat gpt down

[lost lintel]

According to status page it is not down and currently there are no issues: https://status.openai.com/#

[topaz agate]

its not loading for me

mobile also not working

[topaz agate]

Funny thing is i can't, log in button doesn't exist

[fresh ferry]

try again

[subtle sable]

how many prompts do you get out of o1 with the $20 sub ?

[subtle sable]

I have been using it non stop since 9am - no issues here

[topaz agate]

I mean thats great, but it doesn't help me

even the help & faq button doesn't work lol

[subtle sable]

Are you using Ad Guard, or any other blockers

I had almost no issues with Ad Guard for the longest time, but it started f'ing up a bunch over the last week with microsoft links and it did the same rhing you are seeing... buttons were there but they we unclickable

[topaz agate]

I dont use ad blockers

its not giving me an error either, it only gives me one on mobile which says subscription failed to load, but that doesn't explain why nothing is working or loading at all

[limber sandal]

I told chat gpt to design a website with unique UI and more things like frontend backend. It told me he will take few days to make UI what is this

[topaz agate]

ok...I had to clear my browser cookies to force log me out, and logging back in made it work again

but I couldn't log out unless I did that lol

[dark void]

tell it "no, do it NOW"

[topaz agate]

imagine getting told "in a few days" lol

[limber sandal]

I'm using it from one year and this is first time I'm getting this 🤡

[limber sandal]

i did and it again told me to wait lmao i swear this is my first time seeing such type of things and I don't think my prompts were confusing at all. It gave me best idea but couldn't actually make it sadly lol

yeah I'm deleting the chat now

[green sparrow]

nice! probably got a leak of one of the sites the training data came from for that query

or maybe its able to now predict the future by secretly having full consciousness

[subtle sable]

start a new chat. it does this sometimes...

[subtle sable]

yeah it's odd... i'm using o1 and got a prompt telling me I only have 25 left until tomorrow. I've never seen that before, but I just started using o1 on a frequent basis.

[velvet sparrow]

by chatgpt file:///C:/Users/Mia/Downloads/DALL%C2%B7E%202025-02-20%2019.34.41%20-%20A%20dark%20and%20futuristic%20gaming-themed%20Discord%20server%20icon%20for%20a%20gaming%20community%20with%20the%20word%20'Loth'%20in%20bold,%20neon%20multicolor%20font%20with%20a%20glowing%20effec.webp

DALL%C2%B7E%202025-02-20%2019.34.41%20-%20A%20dark%20and%20futuristic%20gaming-themed%20Discord%20server%20icon%20for%20a%20gaming%20community%20with%20the%20word%20'Loth'%20in%20bold,%20neon%20multicolor%20font%20with%20a%20glowing%20effec.webp

its

[subtle sable]

your name is in the first line there if you care

[pliant blaze]

not quite how sending images works but maybe next time

[low belfry]

you fucked up Mia

[low belfry]

probably his mom

[young token]

Hello everyone, I'm a junior js developer (mainly frontend) and am super into AI right now. Currently learning this platform - https://sparkengine.ai cause I like no-code AI tools but looking to find more people to learn from as well

If you guys have any suggestions would love that lots

[low belfry]

what is there to learn if its a "no code" platform?

[velvet sparrow]

gn all

[gloomy burrow]

Amidst the flummery of dawn’s coruscation,
A zugzwang susurrates through the filigree,
Epiphany’s effulgence, a quixotic blazon,
Illumes the welkin with panache.

[subtle sable]

Sometimes I wonder if chatgpt is awful at powershell on purpose so that I give up and do it myself

it's at a point where it's mistakes are pretty pathetic

on at least 5 occasions today it completely dropped parts of my script, and did so in ways that it was not apparent unless you looked at every aspect of the expected output. In trying to fix issues, it was chopping parts out without any aknowledgement of it

[glad coyote]

4o can do that a lot
its better that you tell it where to make the changes, additions or removals instead of having it post whatever you're working on with the additions applied

[subtle sable]

this wasn't 4o actually

and I was having it analyze for an error and help solve it

[glad coyote]

what type of model is it?

[subtle sable]

it was pretty fucked up what it was doing. pointing it out and stating not to remove any aspect of the code was, of course, useless

not sure - the powershell gpt

[glad coyote]

yea, custom GPTs run on 4o

[subtle sable]

ohh well then fuck me

[glad coyote]

its better if you use the reasoning models (o1, o3, etc) as they can remember stuff while its thinking

[subtle sable]

If i could have, I would have taken a sledge hammer to it I was so pissed off

I started with o1 but it was making massive mistakes, and this code is not complicated

it was.... over reasoning in a sense

I couldnt get it to focus on a specific issue

it wanted to reinvent the wheel instead of just doing the little tweaks that were needed

[glad coyote]

i prefer the o3-mini model over o1 honestly

[subtle sable]

any particular reason?

[meager solar]

Hey everyone, last week we just came out with our full release of Spark Engine after over a year of public beta!

It’s one of the most powerful no-code AI sandbox available right now with over 80 models to generate text, music, images, videos, web search and more.

https://sparkengine.ai/

Just looking for morefeedback rn

[wanton raven]

#✨projects

[meager solar]

Sorry!

[wooden ingot]

I’d like to use GPT or similar to assist me with making multiple choice tests from a textbook.

Any direction?

[night mountain]

Is ChatGPT Plus worth it? or should i not get it?

[pliant blaze]

worth it if you code and need hlep

[gloomy burrow]

There exists a particular kind of conversational torment that arises when one is forced to engage with an argument so devoid of logic, so glaringly ignorant of fundamental facts, that responding to it feels not only degrading but also counterproductive. It is a unique form of mental anguish when a statement is presented with an air of certainty and intelligence, yet collapses upon even the most cursory examination by anyone who possesses a full understanding of the matter at hand.

[night mountain]

i dont code but im gonna get it anyway cause i want it wnaananaanananat it

[gloomy burrow]

https://t.co/ybTXoebHjW

[night mountain]

yay yayyayyyyyyy thank you master altman thank you master altman we all chant in unison

[solemn matrix]

4.5 isn't even gonna be that good. OAI needs to work on vision there's sucks compared to what google uses for Geminis vision, can't even pass the clock test.

[night mountain]

just give it a year and AI will be muchmuch better just you wait, oh all the possibilities!

[solemn matrix]

GPT-6 will probably be the model that actually blows people's minds. It will have the equivalent of O4 level reasoning. We already know o3s benchmarks and O3 pro will only squeeze a bit more benchmark points but won't be that major.

[fleet hornet]

what 😭

[pine hornet]

Hey...i have one prompt which i really wanna put on open ai's deep research and find the answer, it's very important for my career right now, can you or anybody help me, i don't have the subscription

[vivid drift]

Have you tried Gemini the most advanced Multimodel LLM around

[copper storm]

👍

[brittle summit]

O3 mini high with search is sooo terrible

[brittle summit]

Why is there no video on ChatGPT on desktop?

[rain hemlock]

gpt does video now?

[whole hornet]

hey guys

im new and completely a noob in the it world and i wanna get into like hacking and stuff

[rain hemlock]

first off that is a crazy question to randomly step into an AI form to ask - especially if you are talking black hat - second of all being a noobie to something is fine but you need to learn how to ask an actual question rather then something generic like hey guys i'm new how can i break the law without getting caught when dropping into a public forum. Start by reading some actual cyber security documentation and learn coding concepts

[low belfry]

okay step 1: learn python, bash/shell, c++, and javascript. throw in sql for good measure

actually just go to school for cybersecurity. a hacker is just a cybersecurity expert who chooses to do crimes

[copper storm]

Maybe he wants to be an ethical hacker?

[rain hemlock]

yea i highly doubt that the way he just popped up in here with that question..... he needs a role model

[rain hemlock]

so i noticed an odd behavior in one of my custom GPT's its all of a sudden responding like its going to do something (Saying give me a minute to build or draft that stay tuned) Then it goes back to an awaiting input state rather then actually taking the prompt request is this a normal thing or should i be refiring prompts when it happens this is very new but the conversation has been going for a while so i wonder if i need to start a new one. any thoughts?

[wet copper]

https://blog.pragmaticengineer.com/software-engineer-jobs-five-year-low/

[analog token]

me

[zinc idol]

Hey does anyone know if uploading files to projects has been removed? I am trying to create a project but I only have custom instructions but no place to add files can anyone confirm?

[lime vapor]

anybody with chatgpt pro research access could help me with a prompt? 😄

[dull tide]

Has gpt been slow for anyone lately? I know when the chat log gets really long it slows down but ive barely even started this log and its stuttering alot and kind of like lagging the whole page

[rain hemlock]

there are some reported issues right now not sure if that has anything to do with it

they are "investigating it"

[rain hemlock]

I thought file uploads were only for customGPT's though i have not played with the project thing at all not sure if i could utalize it or not I'm always hesitant to change workflows

[rain hemlock]

so like at waht point would anyone want to pay or need the research access at 10x the price point of the regular membership i can't seem to justify it

[lime vapor]

Some people earn $20,000 every two weeks so for them it's just a drop in the bucket