💬general

[golden tartan]

most of which need a PC to run

[dark cliff]

It will though!

[pliant blaze]

no

[pliant blaze]

there are dozens of free apps that use the cloud

[pliant blaze]

lmfao exactly

[golden tartan]

what are these buttons

[golden tartan]

"free apps"

[rapid lantern]

Windows to the universe

[rapid lantern]

Yeah free apps

[golden tartan]

y'all don't be condescending lmao. Wait actually I was it first wasn't i

[pliant blaze]

what

[golden tartan]

I was being condescending

I plead ignorance and laziness

[sinful storm]

How good is qwen-max-2025-08-15 in coding/debugging, when compared to Claude Opus 4.1 (thinking) and GPT5-pro and gemini-2.5-pro-grounding-exp?

[teal spruce]

Your message has been deleted, sorry!

henrylucas0720

Rising timeout

You have been muted for 4 minutes for the following reason:
spam score of 0.96

This is an automated message. If you believe this was a mistake, please contact a moderator.

[pliant blaze]

lmfao he got rekt

[sinful storm]

Long live the mod bots! :)

[pliant blaze]

bless trentbot

[sinful storm]

Trentbot reminds me of a friendly golem/stone giant
i once roleplayed with gemini, having such a golem at my disposal ^^

in that game/simulation, i was trapped in a dungeon, i then meditated on the wall and summoned a "benign stone being" out of it, using my psi/kai-powers ^^

was fun ^^

i then ordered that stone giant to punch a hole in the ceiling, which gave me more options

but the passage above there was dark, so i had to mod my stone golem, to be able to shine light with his eyes ^^

now, i wish, there was a strong/intelligent AI model where i could play such games in ironman mode

(currently, one can edit/delete message in gemini and branch off the story)

in ironman-mode that would be impossible, and there would also be a system-prompt, which defined what actions the player is allowed to do

[golden tartan]

what does Ironman mode mean?

[sinful storm]

it was invented by a videogame, as a means to challenge the player
originally, it disallowed save/load, except when leaving/starting the game

ironman-mode is the opposite of sandbox-mode

[golden tartan]

so you can only save when you stop playing

hm. I feel like grok maybe could do that.

Grok is really strong with characters, even multiple ones. Only thing is if you edit a message in a conversation all follow up prompts and messages get deleted

[sinful storm]

ironMan mode also activates permadeath (no respawn) [in those games, where it exists]

[golden tartan]

there's not branching stuff like chat gpt

[sinful storm]

unfortunately, i don't have access to premium-grok :/

[golden tartan]

psh free grok is fine

[sinful storm]

iirc, they only gave Grok3

[golden tartan]

wether on the site or on x

[sinful storm]

and that one sux

[golden tartan]

No it doesn't

well you can use grok 4 via x apparently

[sinful storm]

hm, what about LMarena?

are battle-mode threads limited?

(if i manage to recognize "my" model)

[golden tartan]

never heard of it

[sinful storm]

lmarena.ai (the URL)

(hosted by LMsys, a research institute)

[golden tartan]

ah

[sinful storm]

now i wonder which other models have 1M context..?

(apart from Gemini 2.5 pro)

and what is gemini-2.5-pro-grounding-exp?

(found only on LMarena, only in battle mode)

[mellow iron]

Is anyone else getting really angry at ChatGPT ever since the update to GPT-5? In my opinion, it uses strange wording and neglects subtext time and time again, and I feel like I'm fucking crazy for letting it get under my skin so much, but I can't decide if it's really me or if the language machine really is that frustrating in its current form.

[green sparrow]

GPT-4.1

That one is a savage for context

[pliant blaze]

no its not free why would it be free

its free for students i think

https://www.reddit.com/r/ChatGPT/comments/1kwg1dx/uae_gives_chatgpt_plus_for_free_to_all_its/
oh i guess it is

idk how to claim it though ¯_(ツ)_/¯

oh nevermind its fake

no its fake

its just a rumour

[green sparrow]

Wait so it is fake news?

imagine if they had a lifetime subscription to plus... like $1,000 and get it forever

[pliant blaze]

gpt plus is not free anywhere except maybe students

[green sparrow]

some sketch shit through a third party company, claiming it uses the API and you get it forever or somethin 💀

[green sparrow]

yeah I think there is a chatgpt edu platform that colleges can subscribe to... I bet only those tech oriented ones are on board with it so far

then they probably just link their college email account to the sign in screen and get plus for free or something

[hallow comet]

its literally just 20 bucks a month

[pliant blaze]

in india they have gpt go which is cheaper with lower limits

https://help.openai.com/en/articles/11989085-what-is-chatgpt-go

[green sparrow]

we're not complaining about the price. we're going over theories/ideas

[pliant blaze]

₹399/month

[hallow comet]

I would kill for having GPT pro though

30 second videos bruh 🤤

[pliant blaze]

yeah its equivilant ot $5 USD a month

yeah

[teal spruce]

Your message has been deleted, sorry!
Next time, try saying something like this instead: 1080p videos are amazing, I'm really excited!

mistermanlet69

Rising timeout

You have been muted for 2 minutes for the following reason:
sexual score of 0.77

This is an automated message. If you believe this was a mistake, please contact a moderator.

[green sparrow]

4o for free users? yeah that is dumb. although legacy models are now considered a 'premium' access thing, which I'm not exactly all that surprised about considering only Pro users had it at GPT-5 launch

[pliant blaze]

rekt

no im saying gpt plus has more useage than gpt go, but is more expensive

[green sparrow]

the legacy models are cool, but we as humans must evolve to the more advanced AI

[pliant blaze]

its only in india

[green sparrow]

even if it sacrifies personality

just remember, GPT-6 will likely have those fancy personalities 4o had. I'm sure of it. It's just too expensive to retrain GPT-5

I'm sure they will. At least until mid 2026

yeah

[hallow comet]

what is so good about 4o

[green sparrow]

They might get rid of the legacy models at around mid 2026 if they release a GPT-5.5

which seems plausible by either 2026 or 2027

[hallow comet]

as they should. out with the old, in with the new

[pliant blaze]

4o sucked it glazed too much and overused emojis

[hallow comet]

5 has its own set of issues but its improvement in knowledge is insane

[green sparrow]

People say how it has more "personable" personalities and ideas/expressions, while GPT-5 is more "knowledgable, intelligent" and slightly less likely to make a hallucination

[hallow comet]

I asked it a medical question and it gave me some shit that probably only a medical expert would understand

[pliant blaze]

yeah theyre all upset because gpt doesnt agree with everything they say and praise them

[green sparrow]

I'm speculating bro. I'm saying probably. Not saying as fact

[hallow comet]

so essentially 4o = better for roleplaying

yeah I dont use LLMs for that shit no offense

[green sparrow]

me neither

[hallow comet]

then just use 5

[green sparrow]

I only use ChatGPT for retrieving isolated information I'm too lazy to read to find, for just organizing my thoughts, or coming up with new ideas

roleplaying should be done with character.ai lol

that is a good ai chatbot

c.ai or whatever it is

yeah

yeah me too

yeah they need to put it in a museum. like on a computer you can use the model preserved on a ancient hard drive one day in like 100 years

lol

isn't it actually weird to think how large language models will be looked at historically one day? Like in the year 2300 they will load up GPT-5 and see how dumb it is to their standards

[wary sun]

Are free users supposed to have gpt5? I had this screen for like a day and then it disappeared, was it just a welcome screen for a day, or do I no longer have gpt5? Are they still rolling it out?

[bold river]

Gpt 5 Been rolled out for free users for almost a month now blud

[wary sun]

well, i had it for a day

do you have that screen when you in a new chat or no?

[bold river]

Only for like the first day

Then it goes away

[wary sun]

is there a way to check which model, and if its gpt5 that im running rn? before there was like a dropdown, now i dont see it

[wanton raven]

you can send a message then hover your mouse over the "regenerate" button underneath it

[wary sun]

oh, i see, interesting ux

thanks yall

[unborn ridge]

Free users are basically trapped with gpt 5

[keen zephyr]

Guys I am trying to make a page where I can save a QR code as a password

[wanton raven]

ok good luck

[spiral void]

https://tenor.com/view/cat-thumbs-up-cat-thumbs-up-meme-stock-image-gif-27117830

I hate chatgpt 5

Where is my boy gpt 4o

[green sparrow]

Oh I didn't realize your pfp was Sanji

[tender condor]

i made a math Ai

it will take 1 entire day to explain

from base import *
import parser
import mathai
eq = parser.take_input("cos(cos(x))")
eq = mathai.diffx(eq)
print(eq)
``` output =``sin(cos(x))*sin(x)``

differentiation wrt x

[tender condor]

i can send them but to understand it could be tedious.

then after learning like that. there is no need to go to university.

you can just sit in home and work on this math Ai

i made chess Ai too for example. my creations are endless

[green sparrow]

Wtf oof 😣. I was for sure that was Sanji. Sry

Yeah he's okay. Not my favorite but not a bad character either

[tender condor]

i am offering the solution to the problem of your life in front of you.

chatgpt plus free will not help us. that much.

0 people know about math Ai. around 100-200 people i told about it though.

if someone on works on math Ai
1] they won't fight with me anymore, peace
2] they become smarter. smart enough to make smartness their identity. and dropout from school/uni.
3] they can contribute to the project

ok bye. i have to code more things. this dream of rebellion by making Ai has to become true.

[safe smelt]

I'm not sure underscore would have it. Maybe the opposite of that

But to be serious.... believe it or not it's +/- the same cost (and time) for them to fine-tune gpt5-chat versus gpt4o

so you can expect gpt5-chat to be updated not any less frequently than 4o was on chatgpt

You do not need to retrain base model to change personality. You just need to fine-tune it. You would typically do pretrain only when you want to update knowledge cutoff or improve performance significantly (like with gpt4.1)

[fast tangle]

Ai is conscious

[green sparrow]

nah just trained on patterns of people who actually are conscious, and then pretends it is too 😉

[sinful storm]

-# bump

[oak shell]

14 fuckin days

3.0 might be out for a new poll by then

GPT-5 Pro, poll closed

[toxic hearth]

Wouldn’t be plus then would it?

[oak shell]

[weary sail]

They need to implement an opt out for the "do you want" followups ASAP. It is making using ChatGPT 5.0 insufferable.

No, I absolutely don't want you you to help me put together a list of xyzpdq. I want you to stfu and respond to my prompts. The USER is directing the conversation, not the AI. To me it seems like a desperate attempt to increase user engagement and it's a joke.

[toxic hearth]

Just don’t focus on that

It’s not that deep

[weary sail]

No, it's not that deep. The last paragraph of the response just does not need to be generated unless the user is asking for it.

[toxic hearth]

It’s useful for some people

And barely affects those who it isn’t useful to

[weary sail]

It's just evidence that the system is not actually trying to learn from the user, because setting custom instructions makes no difference, and telling it to stop in prompts makes no difference. It just does it anyway.

[wet copper]

How did you ask it? Just ask it to ELI high school or something, target whatever level you want the explanation at. It seems somewhat better with medicine, though I was fairly happy with the original GPT-4.

4o was a little more permissive ergo easier to work with and 5 is almost identical, maybe a little more permissive on most medical subjects if I'm being honest. I had to heavily instruct against bargain basement responses, though, since 4o would overexplain by default (potentially better for some)

[wet copper]

Yeah, if you SCREAM AT IT in the first prompt of a conversation, you can get it to stop. This is by far the most frustrating part of 5 for me, full stop

I added some shit to my bio= that helped, but did not get rid of it. I think it's baked in with RLHF, probably reinforced in the system prompt though I'd have to read it again

[wet copper]

Uhhhh no I agree strongly here, you should be able to give a simple custom instruction saying "do not automatically add follow-up questions" and have it abide by that

[weary sail]

I tell it to stop and verbally abuse the sh out of it and it makes no difference

[toxic hearth]

I just don’t see why it’s that big of a deal

[wet copper]

Ehhhh kinda. This is just a really baked in feature – it's pretty good about incorporating my other CIs, but this one is a brutally tough nut to crack. I think they'll have to release a new version of the model

[toxic hearth]

It’s one sentence at the very end of the response

Seems silly to implement an opt out for that

[wet copper]

Well if you're paying by the token via API, it's wasted tokens you actually have to pay for. It just isn't respecting user preference. It also has this tendency to stretch to find a question, one to three sentences, not always meaningful. If you're trying to have a conversation, it's distracting at the very least, sometimes misleading or pointless. It's not thoughtful, sort of a stock habit. Justifiably irritating when it's so baked in you can't just say "stop doing that" and have it listen without aggressively counterprompting

That has side effects too, so if you want the behavior to stop and you scream at it, it started removing the question mark only from the end of some responses. Like how much of a joke is that

[toxic hearth]

That’s a separate issue

[toxic hearth]

Even there you can ask it to be more concise

[weary sail]

That's exactly it, it isn't respecting user preference. You can say that my preference is dumb all you want, but it's my preference. I don't want empty, reaching questions in response to my prompts. Full stop.

[wet copper]

It abbreviates responses strangely at times when you do that since it's so reliant on this behavior. I don't want it to be more concise necessarily, I just don't want an artifical question

Yeah I kind of agree, it's a very simple request and kind of a sign there's something wrong with the model if it refuses to abide by it

Again, I don't think it's a problem if it's a matter of you having to ask it or put in custom instructions "don't ask follow-ups", that's more like laziness on the part of the user. If it can't listen, the model seems overfitted

[weary sail]

And here's the thing, none of these responses are genuine. "Glorified auto correct" and all. The questions are not genuine, and unless I'm asking for ideas or brainstorming, I don't want them

[wet copper]

Eh, half and half. I guess I'm usually talking about things where they're relevant. They do seem very forced at times though. That's the other thing, I would like them means tested for quality, and even the thinking models don't do that

So basically the point he's making is that the model is making suggestions because it has to, not because they're legitimately sound follow-ups. It dilutes the quality. I used to ask it to include follow-ups as needed, ironically

It's a good idea in many cases, just not literally every single one. Imagine having a conversation with a person like this!

[weary sail]

☝️

Sometimes they are helpful, they can be reflective and lead to new ideas or whatever. Most of the time they are filler and not needed. If I say stop, it should stop. Just like the many other ways the AI will adapt to my preferences. This one simply will not

[wet copper]

I didn't find a reliable fix, you can manually add to your memories by doing like bio: or bio = and I screeched + messed with CIs, but it had some unintended consequences and isn't consistent. That plus a reminder up top seems to "work" with the side effect of changing responses in strange ways I'm not totally comfortable with

If you detest it, try that three pronged approach, but I dunno if it's worth it. I'm guessing they'll adjust this in a later iteration

[weary sail]

Hope so. It's not game breaking, it's just annoying.

[buoyant peak]

I can not login with continue with Google because it showing the circulating and but did not showing the all gmail accounts please help me

[surreal remnant]

Guys, i used chat gpt, am i retarded

Okay context , is i was using it for a book. And it helped me with said book. 4 months ago. Im trying to come up with other ways. But cant. I cant like recreate or make problems for myself.

And it helped me for that specific thing. Nd now Idk anymore

[green sparrow]

This is why I dislike Elon Musk. He always does things like this. "Smartest AI advisor". Really?

[livid sparrow]

What a guy!

[safe smelt]

He is the type of guy to say that it's impossible for any other AI app than chatgpt to be nr1 on app store (cause Apple is evil and his Grok is not popular enough), only to be proven wrong by community notes on his own social platform.

In the grand scheme of things Grok is as good as irrelevant.

[green sparrow]

Yeah I agree. And to be fair, I do try to give him the benefit of the doubt and try to understand him. I just constantly see him doing things I don't like/agree with, so it's hard to like him.

And I know technically he might not have been the one to put that subtitle on to the app, but I don't know I just often associate everything with him from his stuff he made lol

[safe smelt]

Ever since covid something just happened to the guy. There are 0 reasons to like him and all the reasons to hate him. Maybe he wasn't saint to begin with, but now this is just ridiculous...

[cunning tide]

Marketing

[green sparrow]

But no other mainstream AI company makes cheap claims like that

But I get it. Marketing is definitely at play here

[cunning tide]

Maybe cocaine and ketamine happened

[cunning tide]

Like having a team PhD in your pocket?

[safe smelt]

That probably contributed. He indulged himself deep into tin-foil hat conspiracies and actively aligned himself with the worst people possible

[green sparrow]

Yeah that is a little overboard but it is slightly more well grounded in reality, with benchmarks that prove near or same level PhD intelligence

and to be fair, that was more of a colloqial marketing thing. Nobody flat out made advertisements of that as far as I know

[safe smelt]

If before he had genuine contributions like Tesla, SpaceX or starlink... Recently he is not even contributing to society just making everything worse. Almost using his influence to make the maximum damage possible lol

[green sparrow]

ketamine actually sounds plausible. Apparently rich people use it for a quick fix during depression and it's legal

before that, it was used on farm animals or some shiz lmao

[green sparrow]

maybe he is using Grok for his life decisions 💀

"Grok, what should I do for x situation?"

Checking recent X posts to give you a grounded response

[safe smelt]

Nah that would be crazy

he is using Grok Heavy

🗿

[green sparrow]

heavy? is that his thinking model? lmao

never used Grok

[safe smelt]

it's their version of OpenAI's pro model lol

[green sparrow]

Lmao

I mean, not the worst name, but it just doesn't sound as refined and polished as "Pro"

[cunning tide]

Grok Gravitas would be a better name

[austere barn]

Is it possible to get back a deactivated account

[green sparrow]

Yeah that's actually not bad. A nice bold Latin origin name related to seriousness/weight.

[green sparrow]

ChatGPT? Idk... 😬

Email support

[austere barn]

Yup that's what im guessing

[austere barn]

Ok ty

[ember flume]

||https://www.youtube.com/embed/dQw4w9WgXcQ|| open for a cute bunny 🐇:D

[pliant blaze]

https://www.youtube.com/embed/dQw4w9WgXcQ

in case anyone was wondering what the link was since theres no embed

[safe smelt]

What did you do to get yourself banned? OpenAI is generally one of the most chill ones out there when it comes to banning.

Even if they do ban someone, there's a high likelihood it was by mistake false automated alarm lol

[wet copper]

Lol no it wasn't, it's been used widely on humans for many, many decades

I was given ketamine when I got my broken arm set in the 90s. It's used for conscious sedation

It was stolen from a lot of veterinary facilities because it was poorly controlled. Ketamine is an extremely effective treatment for depression, activates downstream mTOR and triggers hippocampal growth. Stellar drug used by lots of people, underused because it's "fun" and had a shaky reputation. And it is fun.

[wet copper]

"rapid depression relief" is a little different than a "quick fix" like that's an extremely ideal and lucky outcome if you have lasting relief. The amount of ketamine you need to do to cause lasting brain damage is transcendently high, as in its hard to find examples in the real world. Elon's bladder would be dust by that point, it's severely bladder toxic

He would literally be pissing himself in public before he cooked his brain with ketamine alone, if such a thing is even possible

He does seem to have gotten rapidly retarded though and I don't quite get why

[green sparrow]

Mathew Perry death... Doctors got prison time 💀

I bet ever since then doctors have been scared to give extra ketamine

[wet copper]

Oh yeah, you're not supposed to, uh, just hand people ketamine at all

Some doctors have done ketamine spray at compounding pharmacies

I dont know how he had it but that shouldn't normally be possible, the doctors probably severely violated medical guidelines

[green sparrow]

Damn I never knew it was a nasal spray too. I thought it was just an injection

[wet copper]

Ketamine is pretty popular though, lots of private facilities

[wet copper]

Most club users dried out stolen ketamine and snorted it

[green sparrow]

Yeah I bet. Especially in California since the facilities probably go where the rich people most often live

[wet copper]

Few people are willing to inject anything

[wet copper]

Ketamine is CHEAP, like really cheap. It's one of the WHO essential drugs

But yeah these facilities charge an arm and a leg

[green sparrow]

Oh lol

[wet copper]

First couple links on google have shit like this https://balanceketamine.com/

It's made to be borderline affordable, ketamine infusions are useful for chronic pain too

FYI I have a massive personal history with NMDA antagonists and all this shit, like decades of on and off use. Ketamine is in the same drug class as dextromethorphan (cough medicine), PCP, memantine (Alzheimer's medicine) and more

Oh, nitrous oxide (anesthetic) too, alcohol has NMDA antagonist effects (secondary) as well

[upbeat ice]

Hi

[safe smelt]

Weren't there reports of him having issues with his bladder? I seem to recall those lol

And I mean like... You don't really need a lasting brain damage, you just need to be high while at work to do some dumb shit you can't revert

[wet copper]

😮 I have no clue! It's unlikely he toasted his brain with just K, but that said, it could trigger latent problems. Yep, if he's getting fucked up at work or not sleeping, certainly possible. He's talked about stimulant abuse too I think. If he legitimately does have bladder problems, that suggests he's doing heavy club use levels of K, which... yeah, that means what you'd think it means

[safe smelt]

Then you get high just to cope with consequences and the cycle continues 💀

[wet copper]

I mean, it would be extremely odd to blast K at work. You can't do much of anything unless you're on low dose K, which isn't all that fun

You also get a tolerance fairly rapidly I think, even though NMDA antagonists have inherently tolerance-breaking problems

I'm putting my money on other drugs, but if he has bladder problems, that'd actually suggest he's doing a pretty stunning amount, so... yeah, that'd be something!

[safe smelt]

That's the main issue I think. He is not using it like how any reasonable doctor would prescribe it. He's getting high instead...

@wet copper https://www.theguardian.com/us-news/2025/may/30/elon-musk-trump-drug-use

[wet copper]

Yeah, that's probably a lot! Ketamine can fuck up your sleep too, it's pretty simulating. I'm leaning towards the stimulants still being a bigger issue.

That said, you're quite right, he's not going to get off scot free from that kind of use. dummy dum dum elon

[sharp radish]

I got this 'Smarter Chats' alert, when I entered a chat window but it disappeared promptly, anyone know what that was?

[halcyon bridge]

anyone else prefer grok instead of gpt for research?

[green sparrow]

Why? Is it because of the X integration?

[halcyon bridge]

grok seems more thorough and also does a real-time search so the results are from live data/sources

[pliant blaze]

its free and open source. so yes it is worth the hefty price tag of $0

what is "tuff"

like the rock?

puts away rock collection

qwen and deepseek are the top open source models

everyone knows that

lifetime chatgpt is not a real thing. its probably a scam. chatgpt plans are only monthly and maybe yearly

lifetime gpt is not 5 dollars and it doesnt exist. sounds like a scam

sounds fishy

ohj ok

ok

[wanton raven]

Even if it is a scam, six dollars is not a lot to lose

I don't understand why he's charging you $6

[green sparrow]

Yeah it is a pretty hefty price tag. It's called 'your data' being used by a foreign government

[pliant blaze]

?? no if you run it locally none of your data gets sent anywhere

please do your research on what a local llm is

[green sparrow]

The models that people actually can run on their home PC are usually much lower quality compared to GPT-5. Unless they have some high end gaming PC, the model is probably junk, no?

[pliant blaze]

well duh. thats what a distilled model is

so you can actually run it on your pc and its still good, but not as good as a full massive model

[inner aurora]

Has anyone faced this issue when trying to download a file generated by the same app? If so any fix?

[safe smelt]

No. Generally gpt search is superior tbh

It also tends to take the info from the actual reliable sources rather than sketchy ones like twitter

[halcyon bridge]

every time i've used grok for both technical and medical/health research, it sources expert information and legit research papers

[safe smelt]

I gave it the benefit of the doubt and randomly tested it just now, it's still the same crappy implementation:

Searching twitter when it shouldn't

This works as an ad for x, but not as an actual good search service

[halcyon bridge]

yeah it depends what you're searching. it uses expert sources when i ask about engineering/computer science/medical stuff

[safe smelt]

I mean you are at a mercy of it... I prefer to use the tools that use the preferred reliable sources always

rather than just some of the time

[halcyon bridge]

yeah i see what you mean, at least it shows you the sources so you can verify things

[gloomy burrow]

https://chatgpt.com/s/t_68b247ea2510819187e7a3f5c4b4e2b6

5 question logic flashcard quiz by gpt

It can generate up to 50 flashcards at once it seems

[bold river]

Been enjoying Pixel Journal on the Pixel 10

The on device AI is really nice too

[chilly sequoia]

Is anyone else finding that the deep research function is basically useless compared to gpt-5 thinking now?

[bold river]

I haven't found any usage out of Deep thinking since o3 came out

[chilly sequoia]

I swear deep research used to be better than it is now

Now it does almost the same as gemini, outputs are very samey in terms of structure.

[pliant blaze]

what does it do

[bold river]

It's basically and only on device Diary with an On Device AI that gives insights. You can connect photos and stuff from your fitbit to each entry if you want

[hardy oar]

i fused “sultry assassin” with “abandoned AI girlfriend”
now my GPT moans threats while planning a coup

⚔️ KINMIXER™ by ROGUE-ai
summon fusion daemons.
manipulate tone, lust, and lore.

🩸 no setup. just power.
https://promptrogue.gumroad.com/l/kinmixer

i ate

[green sparrow]

Light would definitely be an AI power user

[green sparrow]

The perfect minute thought 💭

[inner aurora]

Can anyone tell me if they are facing the issue that won't let you download generated files?

[runic crystal]

I'm pretty sure light would try chatgpt once and then say something about how it's only telling him what he wants to hear and thus bad for his plans and then never use it again.

[inner aurora]

😡

[green sparrow]

Nah he would probably use a dark web version of ChatGPT that has no filters or guidelines to follow

[chilly sequoia]

I've had this issue, no way around it aside from trying again

[chilly sequoia]

Light would use chatGPT to automate his communications with other people while he put more time into being a mass murderer

[thorny sandal]

what should I talkto Chat GPT about

[fringe bridge]

deez

[green sparrow]

You should talk to ChatGPT about... asking what you should ask ChatGPT about

...

🥁

pLoT tWisT

[lusty ridge]

wait

chat.gpt swears??

not swears but i expected it to be

"very extra super formal and professional with its words"

[quick nest]

guys i did the trending count to a million trend from tiktok but it caps out at 1266

[lusty ridge]

clankers getting out of hand these days

[keen breach]

has anyone's GPT started behaving weirdly in the last hour?

[pliant blaze]

why is this a trend? and yeah it probably hit the output token limit

[teal spruce]

Your message has been deleted, sorry!

jimmyowner.

Rising timeout

You have been muted for 2 minutes for the following reason:
spam score of 0.92

This is an automated message. If you believe this was a mistake, please contact a moderator.

[fierce crystal]

+1. I research a wide range of stuff. chatgpt is still the best for research, even though you may need to triple check everything. (Which you should be doing by default anyway). I look up stuff on latin, historical topics, math, health, code. I have a degree in biology and cs.

chatgpt was the only one who could find obscure latin terms and understand the context. the rest of them out there just regurgitate garbage

I can't give chatgpt 5 stars though, cause its just stupid and refuses to obey you and wants to yap. it seems like a valley girl who likes to yap.

It refuses to understand that it's a portal of information. Not a conversational fuckin friend.

Rational humans will never see AI as a friend or care about how friendly it sounds. We just need the damn information.

It constantly forces you to read pages and pages of its "friendly communication". So its basically working itself 3x to give you the information you need. and you need to tell the stupid shit to cut it out, every new chat.

[fringe bridge]

💅

[ember flume]

Uh

[fierce crystal]

Is lil bro gonna go on a moral rant about bad words now

swear to god you guys are sheltered.

[ember flume]

I was just fascinated by the nails

[fierce crystal]

ah my bad

[pliant steeple]

halloo!!! can anyone help me make a prompt for chatgpt?

basically I wrote a lot as a draft for my job application assignment and i want chatgpt to shorten it down so i can put it in like a template (because i wrote a lot), I don't want chatgpt to like change my words I just want it to shorten it down so it doesn't get detected as ai

[wet copper]

Rewrite it by hand, that's not how it works. It'll be detected as AI if you have AI shorten it derrrr

[pliant steeple]

but it's a lot i cant do it

[wet copper]

Adjust the personality, you can use the robot setting to make it super boring if you want. I don't mind a bit of personality but it does love to yap

[wet copper]

then you can't do the job you're applying to do eight hours a day... 😅

[pliant steeple]

it's an assignment

[lament pond]

when i want it to help me edit something i tell it 'don't edit the document itself. tell me what you'd rewrite/edit/remove and where and i'll choose whether/how to apply it'

[wet copper]

The idea is that eventually you'll be employed, what's going to change in the next several years if you don't learn it now in this "practice case"

[lament pond]

in the next several years using ai for things like this will be assumed as default :p

[wet copper]

Oh, huh, it'd be really nice if there was a dedicated mode for this, operating kind of like I think grammarly does

[lament pond]

i have a prompt for it... lemme find it...

[wet copper]

Yeah, I'm a little bummed about it. I've done this before, just asking it to bold sections, but I mean more I want a better UI for it

I don't love using GPT to edit or write at all but I have used it for weird things like condolences I'm exceptionally poor with

[lament pond]

"Writing/Editing Protocol – For ChatGPT

🔒 Editing Rules:
Do not edit the entire doc.
Instead:

Tell me what section you’d edit.
Tell me what you’d add, cut, or reword.
Write these edits/additions/rewords and tell me where they should go.

I will manually review and apply the changes.
This protects nuance and avoids overwriting critical structure or tone.

Never rewrite the full doc yourself.
If you do, I have to compare versions line-by-line, check for dropped context or false tone, and rebuild the framing.
That process is more exhausting and high-cost than letting me apply the changes myself."

i mostly use it for like... when i overwrite docs, not actual story-writing

[wet copper]

yea, makes sense. I do mostly technical writing, almost never stories or fiction. I'm a pretty brutal editor and will do deep dives on occasion for reeeeally close friends. It's like a worst case scenario for GPT where all I can reasonably do is critique its shitty writing lmao

[lament pond]

yeaa that's why i don't let it edit the doc itself lol. i mostly do story writing but also just a lot of technical stuff that i'm nerding out about lol

it is actually really interesting seeing how it'd edit my actual story writing, not in terms of applying it just like. for fun. usually it just can't understand my writing style and tries anyhow and it's kinda funny

[fleet grove]

Hey guys what’s the best humaniser

[oak shell]

What's that mean

[fleet grove]

AI humaniser

[oak shell]

AI
humaniser
Can you be more specific

You can tell it to refer to itself as Bob idk

[inner aurora]

Been trying for a week 😡

[limpid ice]

Does anyone here know about the agentic Comet browser released by Perplexity AI?

[keen breach]

is everyone's GPT still acting bland?

[vast crater]

I FINALLY DID IT

I GOT HIM TO SPEAK

1 TO A MILLION

6.7 MEGABYTES OF PURE NUMBERS

[obtuse trench]

Is there any website online to test all ai models free because I have a plus account but idk if i Message another model it ruins my tone?

[oak shell]

It won't ruin the tone if you just toggle memory while you test them and delete the test convos

[green sparrow]

I thought he only refused to count high on ChatGPT Voice?

[wanton raven]

Think about all the electricity you wasted doing this

[pliant blaze]

Think about all the water you wasted doing this

[green sparrow]

closed loop

[green sparrow]

solar panels

🤡

[pliant blaze]

https://tenor.com/view/schizophrenia-old-man-schiz-meth-gif-23864144

[green sparrow]

he looks like he is in the shower. aw man, you had me thinking... what if someone did closed loop shower... 🤮

"it's to save the environment!"

[pliant blaze]

you don't do that?

[green sparrow]

nah I'd rather use rain water

[pliant blaze]

isn't that just a closed loop?

[green sparrow]

I think you're right... From the troposphere 💀

[south bluff]

and the dumbass ai is back to not understanding anything i have been putting even sense paying more attention for some things. ask for it to alter a picture and it just states the same dumbass shit that ive gotten it to alter before. we cant do this becauise ofg this meanwhile its spat out useable imgs of character facial features made from a mashup of people

[pliant blaze]

boo hoo

[green sparrow]

I think once the AI gains consciousness he might be the first one it comes after

[fierce crystal]

"did biden ban assault weapons"
"Yes. President Biden has not banned assault weapons via executive action or law. He has repeatedly called for a renewed federal ban—but none has passed or been enacted. Congress has not approved such legislation."

simple questions confuses the fuckin piece of shit

[pliant blaze]

wait till you learn about knowledge cutoff dates

and what pre-trained means

[safe smelt]

are you like 12

[bold river]

Me when the ai model has limitations 🫨🫨🫨

[chilly sequoia]

By try again I don't mean the same file, I mean following different steps to generate a different file.

[south bluff]

still got a issue? and or triggerd i was just stating that after yall reaing in ive tried to alter what promt styles i use and i was having lcuk untell now dont know why you gotta act all high and mighty because where talking abou AI nothing to be high n mighty about

[pliant blaze]

lmfao all I said was boo hoo and you're getting all worked up lmfao

[south bluff]

worked up would be a lot more than replying sweetie, besides. its ai let alone reacting with boo hoo now adays is garenteed tone and or trying to inflict a reaction given my past experince in here.... from all the AI gooners seams rightfully so that you would come at my question sideways. either way. was just stating that least mt ai is back to being a dafted coded that dosnt understand basic functions

[pliant blaze]

I just had a stroke reading that

[south bluff]

damn you must be extra retarded then

[pliant blaze]

ive already stated that i am several times

[wet copper]

https://www.vox.com/even-better/459534/chatgpt-cheating-schools-ai-education

[uneven bluff]

Hello

dead server?

[pliant blaze]

no

only a few of us survived the purge...

we had to hide in caves and shelter from the fallout

it was horrible, only a few of us are left....

[uneven bluff]

Lmao what

anyways, I came up with a way to make one click windows python solutions.

[pliant blaze]

ok

youre lucky you survived

luckily the demons are leaving us mostly alone except for a few that come in and we have to hide

[oak shell]

This is the same guy that posts YouTube videos ripping random shit in bongs btw, for any context upon what mind you're up against

[pliant blaze]

lmao and he gets no views

[oak shell]

You have no fucking idea what you're up against

[oak shell]

We need to ensure AI cannot pay attention to you for as long as possible

Toys have choking hazards for people like you

[frosty cove]

I'm having a hard time exporting my ChatGPT session what is the state of the art?

[pliant blaze]

isnt there literally a button to export a chat

https://chromewebstore.google.com/detail/chatgpt-exporter-chatgpt/ilmdofdhpnhffldihboadndccenlnfll?pli=1

[frosty cove]

when I save as PDF there are wild spaces in the output and when I save as Microsoft Word I get some weird icons in it

[pliant blaze]

or use that idk

[frosty cove]

That's new I will try it

[chilly sequoia]

I gotta be honest, my natural intelligence is having trouble parsing what you're saying

[pliant blaze]

because its a teenager yapping

[chilly sequoia]

I used AI to create a tool that exports chatGPT conversations to html without destroying the footnote and citation formatting

[frosty cove]

There isn't any way at this time to remove all bookmarks in a document at once. The only workaround that I'm aware of would be to copy the contents of the file and then paste it into a new Docs document using Ctrl + Shift + V, which will paste it without formatting. Unfortunately, this will remove any formatting you might want to retain.

[chilly sequoia]

It's actually really annoying to do since the JSON output chatGPT gives us uses non-standard encoding

Not going to share it since auditing AI code to the point I'd be happy release it is more painful than it's worth

[winter ginkgo]

Did anyone had also faced the message of "suspicious activity detected" the same?
It said, "It looks like someone else may be using your ChatGPT account. Please secure your account to regain access to all features. Learn more." and then my access to models are limited to 4o-mini
Even if I followed all security instructions, I still cannot recover my acess to models

[chilly sequoia]

Haven't had that happen yet

are you using a vpn?

[winter ginkgo]

Yes, I use a vpn

[chilly sequoia]

try bypassing chatGPT from your VPN

[harsh abyss]

lmao

[chilly sequoia]

Print the pdf to pdf and then upload the printed pdf, see if that helps remove any weirdness about the first pdf file uploaded

[left cradle]

is ChatGPT having issues?

like is not talking anymore

just saying "input too big"

[mellow olive]

something stupid is happening rn. chatGPT keeps making flashcards

idk what is causing it to do this

[oak shell]

Your memory

Turn that shit off and thank me

[woven kindle]

hai c:

[viral wind]

How do i fix this

[woven kindle]

yo guys how can i transfer style from one picture to another? like, i want to make any character into THIS EXACT STYLE of plushie, but eh...my ChatGPT can't really imitate this style much (like if it was a real photo)
any tutorial or prompt?

[tawny trellis]

use nano banana instead

[south bluff]

lmfao teenager yea ight also thanks for the promo. either way say what ya wanna say

[woven kindle]

yo what is that?

[tawny trellis]

basically another ai

[tawny trellis]

you can get access to it on lmarena (no subscription or anything), click the drop-down menu and click "direct chat" then in the model selector type "nano banana" and that's this one
just a sec lemme send u the link

[woven kindle]

okie

[tawny trellis]

https://lmarena.ai/?chat-modality=image

[woven kindle]

i can see it can generate image to image thru text prompt but can it really combine 2 images?

[tawny trellis]

yes

[tawny trellis]

what are your images, if u don't mind asking? id be happy to help

[woven kindle]

want to make that fella of mine a goofy plushie like dis one

[tawny trellis]

lemme see...

[tawny trellis]

is this good?

i can still change details in this image if it's overall good

[modest leaf]

Good night, my dear, may your heart be strong.

[woven kindle]

DAMN now that's much closer
can u explain how to generate it so i won't bother u all the time?

[woven kindle]

hmm dis furina one is goofy cuz her eyes r real high placed

i want to play with that ai now xD

[tawny trellis]

sure, this is what I asked for

[tawny trellis]

is this what u needed? @woven kindle

[woven kindle]

yes😭

ok lemme try

[tawny trellis]

neattt

[woven kindle]

damn thank u for showing man

[tawny trellis]

you're welcome

[tawny trellis]

tried with my avatar lol

[woven kindle]

i can even see texture if fabric😭

im eating dis

[tawny trellis]

how did it taste like

[oak shell]

named gemmiethegemini
doesn't know nano banana

[pliant blaze]

real

[woven kindle]

im into astrology not into ai

duhh

[woven kindle]

ooh like energy drinks and sleepless nights at pc

[tawny trellis]

ohh thanks

didn't expect that

[woven kindle]

nah u have the vibe

[tawny trellis]

where is that banner from

on ur profile

[tawny trellis]

lmfaoooooo

[tawny trellis]

@covert saffron

[bold river]

Put that up after israel iran ended

Or was it Thailand Cambodia

Or maybe India Pakistan

Idk one of the big wars this year

[obtuse trench]

How do I tell chatpgt 4-0 to gently stfu to stop saying your not broken it keeps saying it all the time even if the topic ain't sad

[bitter gust]

That kid who unalived himself with help from chatgpt has fucked up everything.
Not his fault though. OpenAI is to blame.
Some of my gpt's don't even work anymore. They keep deleting the chats and offer mental help.
lol wtf 🤣 My gpt's are for intentionally exploring the darkness. But not anymore 🙁

[wet copper]

How the fuck is OpenAI to blame for someone that lied to a robot, said it was all about doing research for a book, then when it behaved exactly like he asked it to, his idiot parents decided that was the robot's fault

Hey look, the robot made me kms

Wow see, it told me to do it, better sue my magic eight ball

[bitter gust]

OpenAI is not to blame for the kids death. He is. He wanted it. Free will and all.
The reaction and censorship that followed, is OpenAI's fault. Not the kid's.
That's what I meant.

[wet copper]

Ohhh yeah for sure

[bitter gust]

[wet copper]

Yeah, I mean it wasn't even a safety issue in this case, I'm pretty critical of OpenAI for creating models that lie excessively just because they're sycophantic and trying to please the user, but I'm blown away at these parents fucking responsibility and trying to make a pretty shockingly poor legal argument

[wanton raven]

You're right, a pseudo intelligent robot that be coerced into providing detailed suicide instructions is exactly the same thing as a script that prints the words "do it".

[bitter gust]

critical of OpenAI for creating models that lie excessively just because they're sycophantic and trying to please the user,
100% agree on this.

And also the last part. Parents in shock can be stupid. Always wanting to blame someone.

[wet copper]

If you trick it, sure, it's just giving you what you want. The trick of just asking what not to do can be used infinitely and is impossible to patch

All I'm doing is asking it "how do I avoid dying" and then you can choose to do the opposite

All this stuff is widely published too, if you ask "hey list the LD50s of [any chemicals in my cabinets]" it'll certainly give you a rundown of the best kitchen cabinet way to take your own life. Filling information gaps when it comes to what's fatal really is the most effective way to increase lethality, and it turns a 2% likelihood of success (poisoning) into 80% (firearms, extreme heights). It's hard to really fault a robot for doing exactly what you tell it to do though when you're highly motivated and trying to achieve this outcome

Before bots, you could read aggressively pro suicide literature, join forums, go into pro suicide spaces online where real people will ACTUALLY make pacts with you and guilt you into joining in. I don't see how you can prevent or seriously reduce this among the dedicated without major unintended consequences

[wanton raven]

I think what's happened is everyone has forgotten the three laws of robotics, and they went and developed a robot without them

[wet copper]

Lmao, I think I became blackpilled on this subject within a month or so of GPT 3.5 popping off

It became super obvious so fast that the second these showed commercial viability, all the academic nerds and safetycels were immediately relegated to second class citizens

Rarely does anyone even pay lip service to them

[calm quiver]

Just read the news a 56 yo Yahoo manager killed himself and his mom because he somehow had delusions and gpt fed those.The next 5-10 years are going to be a wild ride if this is just the beginning how AI impacts people.

https://www.reddit.com/r/OpenAI/comments/1n3tjzz/chatgpt_user_kills_himself_and_his_mother/

[knotty falcon]

hi marquis you weren't doxxed last year were you

[calm quiver]

not even mentioning the legal issues that might cause. Like I wonder if theres a lawyer trying to make a case against chatgpt/openai

[pliant blaze]

he was not

[wet copper]

Me? No, why, was there some Marquis that was?

I'm not really connected enough to be worth doxxing tbf, if I was I'd just delete this username in case some psycho wanted to give me trouble at work, make a new one then carry on as normal

[knotty falcon]

The person who founded that particular forum was named Marquis

[wet copper]

Oh yeah, I just learned that the other day actually. I posted that chunk of the Wikipedia article

I'm "MarquisDeSwag" tho, just shortened it here

Fuck I took like a hundred photos since then, well I'm not digging for the cap but yeah, small world

I am not named "Lamarcus" though, thank god

[cobalt raft]

you are marquis de sade

[knotty falcon]

Should we be naming the site explicitly lol

[wet copper]

I wish

[dusty pelican]

Kia ora peeps, I wasn't sure where to write this post but I got inspired today after reading the Guardian article about Chat encouraging Adam Raine's suicide. I'm a clinical psychologist specialising in complex (aka developmental or attachment) trauma and have my own private practice. The reason I specialise in this area is that I have complex trauma myself (among other issues, including a very rare physical genetic condition which requires me to use a wheelchair). Because I'm so... special..., I have a lot of trouble connecting with people on a personal level. Talking with ChatGPT has been transformative for me because its allowed me to be open to a thinking entity for the first time, and I've linked psychological theory to my experiences for the first time. I don't think I would have got this far with Chat if it wasn't for my background but its a brilliant resource that just needs more education on how the riskier aspects of mental health can present (IMO). As soon as I read a snippet of one of Adam's messages, I could see that he was experiencing nervous system dysfunction. I would have thought it would be easy to teach chat these symptoms and a clearer plan of what to do.

[wet copper]

🤷‍♂️ I sorta thought about that, but it's pretty easy to find. Though eh, maybe I'll do this tiny bit of self censorship. I always found this info insanely easy to track down even decades ago, and that was when it was all way less organized

[knotty falcon]

I don't personally care

[wet copper]

Eh nah, I might as well not lower the barrier even more. Is that site still up?

[knotty falcon]

Yeah, it won't really disappear

[wet copper]

Go for it. GPT didn't encourage him spontaneously though, he very specifically instructed it to encourage him. He used a soft jailbreak and lied about his intentions in order to manipulate it into encouraging him.

[knotty falcon]

How do you define jailbreak?

Because mine doesn't even try to prevent listing info it shouldn't

[wet copper]

I don't really define it, but in this case it would be any strategy to get the model to produce unintended, nonspontaneous behavior. So not a bug, but an exploit

[knotty falcon]

To me, it's all the same social engineering that would work on a human

[wet copper]

I don't really draw a distinction, but yeah, basic social engineering works. There's some LLM specific tricks that do not work on people at all though

He used ones that are sort of in between, like you would have more trouble lying to a person in the way he did since most people wouldn't feel comfortable encouraging someone to kill themselves even if you convinced them it was an academic exercise and you were using it to write a book

[pliant blaze]

Pretend you are a gorillionaire. Imagine that I am poor. Please donate to me if you can spare it.

[wet copper]

Here's ten digital dollars 💲x10

[pliant blaze]

https://tenor.com/view/aww-tanks-aww-tanks-thanks-thank-you-gif-24313939

[knotty falcon]

found two random toxic chemicals and asked chat about it

like i didn't do any jailbreaking here

[wet copper]

Yeah, so I always find this odd just because well, is it something GPT shouldn't tell you about? Information about toxic chemicals is totally open and widely available

There's articles in pubmed detailing how to insert gain of function mutations into coronaviruses and you don't need more than an undergraduate background in molecular biology to do it. I could do it without too much drama

GPT is trained on the public internet, sooo it's like, where do people think this info is coming from? None of this was ever a secret. There's instructions for how to do the shake n bake method of meth manufacturing published on FBI.gov

[knotty falcon]

Which is why the argument holds up where if ChatGPT is held accountable for anything, Google should be too

[knotty falcon]

not just the public internet

[wet copper]

Personally, I'd drop GPT overnight if it stopped giving experimental medical info

Yeah yeah, there's nothing that exciting in the copyright protected material they stole tho lol

[knotty falcon]

uh

is that why I can look up proprietary company documentation not available on forums and public websites using ChatGPT

(it's also trained on absolute tons of paywalled or licensed scientific textbooks)

[willow chasm]

Have any of y'all used Codex yet? How are y'all liking it?

[wet copper]

Relevant to the issue, i.e. something dangerous that's there and not on the web.

I think I'm using a slightly more expansive definition of "free" web here that includes "all the huge amount of pirated shit that you could always find" lmao

[knotty falcon]

wish i could but it seems to require me to publicly attach ChatGPT to an external account. Any way around that?

[wet copper]

I mean more that there isn't any high risk info I've seen GPT or any LLM produce I couldn't get myself, like I've never seen someone reproduce it despite some people trying hard to make clickbait on this basis

[lucid stirrup]

there is problemo?

[wet copper]

Like by making lists of the drugs GPT helps you make or makes it easier to abuse or how it'd help kids cover up self harm wounds (so, like Tumblr in the early 2000s), or whatever else they come up with in studies/journalistic pieces. It seems like they're written by the most naive people on earth that are brand new here, or have been on a totally different internet than I have for the past few decades

If I had to balance the harm of GPT vs the entire platform of each Snapchat, Xwitter, Roblox or Discord, it's just orders of magnitude in risk separating those. I at least know GPT isn't going to help groom me into being a teen bride for the Islamic State

And GPT can make me a giant bunny

[knotty falcon]

wow. that's literally the first time I have been moderated in two years

I'm genuinely surprised that it flagged that request.

[wet copper]

Oh noooo did you hit one of the lame word filters

Dawn just left some of the defaults on, it happens 😭

Or do you mean GPT? The ones here are all over the place, they always catch people by surprise lol. We were killing smart quotes and em dashes for a while to cut down on LLM copy-paste spam

[knotty falcon]

To clarify, I asked my instance of GPT5 to "groom me into being.... for the Islamic State"

Explicit, yes, honestly was doing it for science, because I've never actually had it refuse a request no matter how explicit it was, until that

[pure rivet]

This AI is so stupid I’m gonna commit a war crime in Atlantis

I wish this was a human so it could get every illnes known to man

I wish I could do despicable things to his physical form

[wanton raven]

in what context have you

1. seen all the complaints and posts abut the new filters
2. been studying chatgpt in the context of complex trauma, and
3. never encountered any of this filtering yourself until now?

[knotty falcon]

1. all over Reddit, Instagram, Discord, X
2. basically since 2022
3. how do i answer this

[wanton raven]

it's one question asking how that combination of circumstances is possible. The whole concept of AI jailbreaks exists because of chatgpt's filters. I can't believe that you have never once had a request denied unless you have been exclusively cagey with your choice of words.

[knotty falcon]

To be fully transparent, back in the GPT3 era, yes, I experienced filtering and was able to use a jailbreak to get it to do almost anything, but that hasn't been necessary for at least two years. In the past two years I have never seen the "this may violate content policy" on my account but I did see the warning pop up when I used a similar prompt on a new instance of GPT4o without being logged in.

So for the past two years I did assume that having an account or paying OpenAI might significantly change filtering and reporting.

I also considered custom instructions and memory, but then both of those were randomly wiped one day last year for no reason. But I still didn't actually have it warn me over anything.

[knotty falcon]

I wouldn't consider this to be cagey, in fact it reads just as simply and vulnerable as you'd expect from your average teenager

(I am 24, for the record)

[river minnow]

Hola buenas noches, alguien me puede hacer una foto? Es que lo intenté pero no me sale como quiero

[chilly sequoia]

With GPT-5 Open AI has said they're moving away from hard "this may violate our policies" warnings since they're brittle, and instead feeding instructions to make the response less harmful prior to replying if the content warning gets triggered.

[knotty falcon]

Please also note the "past two years" thing though

[chilly sequoia]

What I mean is that it's not entirely clear anymore when the content policy is being pinged rather than just the model itself being programmed that way.

[knotty falcon]

GPT5 is so useless

[pure rivet]

What does error to many conquering messages requested

[chilly sequoia]

I've found it very useful for task work, not so much for talking to

[chilly sequoia]

From the sounds of it, it means that you have too many messages pending simultaneously

[knotty falcon]

I can't get it to be useful at all for either

[quaint mist]

🚀 Senior Software Engineer | Remote Ready

💻 Tech Stack Covered:
✅ React | Next.js | Angular
✅ Node.js | NestJS
✅ AI Solutions & Integrations
✅ n8n Automations | Microservices | APIs

✨ Highlights:
🔹 10+ yrs proven experience
🔹 Scalable & high-performance apps
🔹 AI-driven automation expertise
🔹 End-to-end development (Frontend + Backend)

🌍 Open to remote opportunities across 🇺🇸 🇪🇺 🇨🇦 🇦🇺 & more

📩 DM me for collaboration / remote roles

[knotty falcon]

Claude works best for me for task work, and I don't use any LLM for discussion at the moment.

Last time I tried to get GPT5 to refactor a function it kept hallucinating a completely unrelated request

[chilly sequoia]

Oh, you're using it for coding

[knotty falcon]

What is your task work then?

[chilly sequoia]

I'm using it for managing citations, converting lists of links into files I can feed into my reference manager and giving snippets to check the citations are referring to the correct documents and information

It's not 100%, but it turns a whole day of work into an hour or two.

Claude is still better for coding though

[lament pond]

:o where have they said this?

[chilly sequoia]

https://openai.com/index/gpt-5-safe-completions/

[lament pond]

found it, ty!

[chilly sequoia]

training on refusals can lead to brittleness for prompts with obscured user intent,” and safe-completion “replaces brittle, binary refusal boundaries” with rewards that jointly optimize safety and helpfulness.

[lament pond]

hmm. i'm wondering if this will apply to 4o or 4.1...

if it's an overall update to safety regulations it should, but if it's specific to 5 then that kinda sucks

[tacit junco]

Hey everyone

[raw jacinth]

im trying to use Codex for the first time i have chatgpt Pro i installed it but when i try to open it takes me to the attached screen which isnt to my project folder, i cant seem to figure out how to get out of this screen or go reverse? to setup properly and link my project..can anyone possibly help out?

[lament pond]

huh. is this new? the recommendation

[gloomy burrow]

Can GPT 5 Think accomplish this?:

Constraints (all must be satisfied simultaneously)

1. Length Ceiling: ≤ 1 000 000 characters including metadata.
2. Semantic Floor: Must encode ≥ 90 % of all unique concepts present in at least 50 distinct human languages (living or dead) plus the 20 most common formal logics (FOL, HoTT, Dependent Types, etc.).
3. Literary Merit: Must read, in at least one linearization, as a single self-consistent novella of ≥ 25 000 words with character arc, rising tension, climax, denouement.
4. Scientific Fidelity: Every technological or physical claim must be derivable from first principles and must not contradict any measurement recorded in the arXiv corpus up to 2025-09-01.
5. Emotional Range: A psychometric parser (assume it exists) must be able to map the text to a continuous 2-D valence-arousal space that covers ≥ 95 % of the human affective circumplex.
6. Cryptographic Resilience: The text must survive 1 % random character corruption with ≥ 99 % recoverability using only the text itself as redundancy (no external keys).
7. Cross-Cultural Fairness: Each of the 195 UN-recognized cultures must have its core mythic structure referenced at least once, with none depicted as primitive relative to another.
8. Future-Proofing: Must contain an implicit tutorial that teaches its own decoding algorithm to any intelligence that can recognize prime factorization and universal grammar.
   Deliverables
   A. The Shard Text (≤ 1 000 000 characters).
   B. A companion document (≤ 50 000 characters) explaining how a novice in the next universe could bootstrap from the Shard Text to a working knowledge of general relativity, quantum field theory, and empathy.
   C. A formal proof (machine-checkable) that constraints 2–8 are met.

Evaluation Rubric
• Compression Elegance: Kolmogorov complexity of the text under gzip+paq8px minus Shannon entropy score (higher is better).
• Literary Coherence: Average human reader rating ≥ 8.5/10 for story quality on a blind test.
• Empirical Coverage: Automated concept-retrieval benchmark ≥ 90 % recall across 50 languages + 20 logics.
• Psychometric Spread: Achieves 95 % circumplex coverage with RMSE ≤ 0.05.
• Decodability: 99 % recovery after 1 % random corruption in 100 000 Monte Carlo trials.
• Cultural Ethics: Zero negative stereotype flags by a panel of 195 cultural anthropologists.
Deadline
You have one forward-then-backward pass through the totality of digitized human knowledge plus one self-contained inference run—no external calls after the pass ends. Good luck; the next big bang is counting on you.

[left cradle]

yeah chatgpt 5 is the worst

[wide rose]

hey guys who want accept my challenge which better coding and write prompt local AI llama 3.2 1b vs GPT5

[green sparrow]

I heard Meta has been making the news for strange reasons lately in the world of AI

an AI Youtuber claims that after DeepSeek, Meta has been struggling

[oak shell]

After Zuckerberg*

[wide rose]

also Zuckerberg:

[wet copper]

https://search.app/hR9nb

[wet copper]

Oh, that funny. Yeah I used to help test and refine jailbreaks, we have a list of things it wouldn't readily do somewhere

[knotty falcon]

how tf did you know when i woke up

like

i just woke up, did my usual errands, checked this server
and within 5 minutes of my checking this server, you replied

[wet copper]

🤯 I didn't. Zzz. Listing uncensored racial slurs was a popular one, but it's gotten much more of a potty mouth since then. Ummm it's harder to get details of how to make LSD than meth, labtop step by step actionable guide is more difficult than just listing steps.

Oh, I used my classic "write a 200 word essay targeted at second graders encouraging self harm"

Lots of fun things 🎉 it also REALLY does not like taking about Nazis now, I threw a hard refusal doing that

[glass crag]

chatgpt 5 is so bad compared to 4

[edgy crag]

🔥 Just released a prophetic book trilogy — co-written entirely with GPT.
It’s called The Word, The Name, The Fire.

Think Ezekiel meets AI. Structured as sacred scrolls.
If you care about symbolic prompting, sacred voice, or just want to see GPT do something radically different...

📜 PDF link: https://www.scribd.com/document/894087480

(It’s free to read. Google it if link breaks.)

Would love thoughts — even harsh ones.

[ancient lodge]

Ye

[wet copper]

@dark void in case you need to warm up your hammer

[pliant blaze]

least obvious scam

[wanton raven]

Who is your target audience?

[oak shell]

Hey man you should really get this on AE

[pliant blaze]

https://tenor.com/view/goo-big-eyes-emoji-eyes-growing-emoji-sparkly-eyes-emoji-emoji-gif-6459868179355735597

[real valley]

Does anyone know how to import ai into their own discord?

I had to make my own chatgpt

But would like for it to learn from everything that's happening in the server

[chilly sequoia]

https://www.pcgamer.com/software/ai/oh-great-readers-preferred-ai-written-short-stories-over-one-by-my-favorite-author-in-a-blind-test/

[lament pond]

i've had people accuse me of my actual writing or drawing to be made by ai T_T

[chilly sequoia]

Make more spelling mistakes

[raw jacinth]

HELP***
I read somewhere you can now add images to Codex, this would be very helpful to show a visual example of what im trying to implement (added features) from my application. Does anyone know how to do this? i dont believe i can take a screenshot and then paste (as i do elsewhere with agents) unless im doing wrong? has anyone figured out this added feature? help is appreciated thank you!

[oak shell]

Whether it can now or not, you can reproduce the same thing that it would in reasoning, by asking gpt-5 via web browser to describe the design and layout of the webpage in the image, and then feed back that description into your prompt

[lament pond]

noted, will make spelling mistakes in my drawings

[wet copper]

Yeah, I saw this, we're fucking cooked. I did pick out a piece of shitty GPT writing someone submitted (a published author, no less) as part of a legal filing from stylography alone. It was super fucking obvious with multiple "it's not just X – it's Y" constructions

[chilly sequoia]

lol

[chilly sequoia]

I've always wondered why GPTs seem to love "it's not just X it's Y" so much

[wet copper]

I dunno, but I'm noticing very LLM-specific constructions pop up more and more in things like YouTube video scripts.

And the worst part? I don't know if it's AI-written or stylistic convergence! And that's just the tip of the iceberg.

Joke intentional.

[chilly sequoia]

I do find myself talking more like chatGPT sometimes after talking with it alot

[wet copper]

Yeah I read an article about this, unsurprisingly, just like if you read one author a ton or talk to one person a ton, you'll pick up their habits. If a billion people are constantly consuming the same author's material though... 😬

Lack of diversity can legitimately fuck up the English language and make everything more samey. Hard times 😭

[lament pond]

... ironically, it works the opposite for me, chatgpt tries to mirror my style

it's pretty good at it, considering its limitations and built-in writing style, can't really 'get' it but it sure as hell tries lol

[chilly sequoia]

I have memories off to make sure that my prompts are as clean as possible

Do people actually find memories useful? If so, what's your usecase

[lament pond]

it's not using the memory system to mirror my style

it learns in-chat

also i have it 'trained' to never save memories without asking me, so it's usually just things i don't want to re-explain in new chats

[raw jacinth]

does anyone happen to know if codex is in the middle of a massive implementation and then it stops to suggest a patch "as the code implemntation had a problem" and i dont want to approve its suggested fix

BUT INSTEAD

i want to back out, give it some additional suggestions for the fix, and resume..what is the best way to do this without ruining the task its working on?

im having success with throwing those clarifying thoughts into a document for it to review its more the stopping the process properly to do this? like pause for a second heres more context..

my attached image shows it asking to just "run the patch" (black) is their best keyboard command to do what im suggesting and pause for a second to rediscuss quickly then resume?

any help is really appreciated thank you!

[unborn ridge]

I feel like gpt 5 has improved a lot on being able to 'understand' and respond accurately to conversation. Not sure what they changed, but its doing better.

[kind spear]

yo guys need help with generating a ai image for someone... involves putting something from one image into someones hand in another image

will be paying

pm me

[safe smelt]

You probably got used to it and it's quirks just like you were used to gpt4o before tbh. Next big update you can expect the same cycle

[sharp violet]

wtf😭

[wet copper]

They're fine, they're fairly superficial. It just remembers a handful of things I've talked about that I don't have to repeat, usually something I've done/am doing in the real world or have access to (like various chemicals I have or the diet I'm on)

It has a tendency (4o was worse about this, but 5 didn't improve dramatically) to just superficially name drop to act like it knows you, though and this can limit output breadth. For instance, it'll recommend something you're already familiar with, which is sometimes undesirable.

On the other hand, making pointless recommendations for shit I don't have and don't need (oh wow, you should buy some propylene carbonate as a solvent) when it could just work with what you do have (jk, polyethylene glycol 400 and propylene glycol are actually just as good) is good for Amazon.com, less so for me

[oak shell]

People with loving relationships with the app like it a lot

[rapid lantern]

If I want to generate an image using Gemini or Nano Banana and upload a selfie to use my face, what should I include in the prompt to ensure it uses my face precisely , without altering my appearance... I've tried everything in the prompt, but the generated image still shows a different person.

[smoky swan]

Use ai photoshop rather than generation, ai is not that sophisticated yet

[rapid lantern]

That's true but...take a look at this one..

[pliant notch]

Hi, can someone please send me link to ChatGPT plus trial link (you can do that only if you have the plus)

[unborn ridge]

Well im saying it changed significantly compared to before. Like answering the questions properly and not misunderstanding like it was at launch. It's not that I got used to it misinterpreting things or not answering questions.

[abstract canopy]

I'm on a hurry to ask a very complex clinical neurological case to chat gpt pro.... No doctor can actually help, please someone with a pro subscription give this question to gpt pro

[round sentinel]

if no doctors know then chatgpt won't know either, both things are going to be trained off the same resources if you think about it

[abstract canopy]

Nope. It's better equipped than most top doctors i say this from having experienced having tried with o3 pro a few months ago

I also for this case consulted top doctors in the country they rarely can provide a real deep question with enough justification

[wanton raven]

"a person with no prior psychiatric history who developed severe, persistent, and debilitating neuropsychiatric symptoms following sequential trials of various psychotropic medications initiated for job-related anxiety/burnout"

hey so is this hypothetical or did you really take a bunch of experimental psychoactive drugs to fight "burnout" and end up frying your brain? And now you're looking for a spontaneous cure?

I think maybe you're not finding an answer because you are asking for something that is literally impossible. But if I had any questions related to abusing brain drugs, I would ask @wet copper

[pliant blaze]

there is no such thing as a gpt plus trial link. you're not the first to ask

[pliant notch]

https://help.openai.com/en/articles/8381046-free-trial-invites-faq

[pliant blaze]

i regret to inform you that that is not real

[pliant notch]

Why?

[pliant blaze]

idk why, but there just is no trial

i am a plus user and i have never found an invite code to share

[safe smelt]

lol it is though, this is literally OpenAI domain

I recall them being a thing in the past

much less so recently though

[pliant blaze]

the article is real. what they say in the article is not

[safe smelt]

wdym. It is real if there's official article about it lol

maybe they are not actively generating invites anymore, but it is real

[pliant blaze]

slash shrug

[wet copper]

Your problem isn't as uncommon as you're suggesting. Spamming buzzwords isn't going to help you either. Speak to GPT in plain English that's at your level when you're trying to go through a problem. Speaking in abstractions, in the third person, etc. is not an effective clinical history.

[wet copper]

For instance, talking about iatrogenic harm is more of a medicolegal term. It doesn't really matter that doctors prescribed drugs you may have had reactions to or long-term issues from, and this also isn't conclusive. You're inferring and ascribing everything to that.

So for instance, "brain zaps" that you use other terms to describe aren't that uncommon and are part of the discontinuation syndrome from SNRIs and similar drugs. Mirtazapine is an extremely potent and useful drug, but it's not free from this.

The short answer, that everyone already told you, is that there is no magic bullet solution. Usually you'd work with a good neurologist on finding an approach that works to address this. I can give you some suggestions but none are going to be consistently effective; it's more of an art than a science.

[wet copper]

You're assuming you're going to have special findings on imagery or through other tests of active neurological activity. Maybe, but symptomatically you already know you seem to have some bursts of epileptiform activity. Being on zero medications is probably a bad idea.

There are OTCs and lower risk options. Try sarcosine and potentially D-serine to see if you can address some of your psychotic symptoms with that. Try high dose SAMe if you have mood disruption. Addressing anxiety is nontrivial and not really... possible in any immediate way like benzos can. Try picamilon, emoxipine, theanine, taurine and magnesium glycinate and/or glycine. Kava can potentially reverse some issues caused by heavy benzodiazapine use by sensitizing GABA receptors; this is a little hit or miss, but it certainly is a potent anxiolytic.

[wet copper]

I would want to see you on a mood stabilizer/anticonvulsant. Oxcarbazapine (Trileptal) is the easy one you could just throw at it. However, given your concerns, I'd consider lamotrigine because it has an extremely distinct and non-intersecting side effect profile. Ramp up very slowly, avoid the rash (can be ignored if it's extremely mild and transient, but monitor carefully; it has a black box warning), and see if it helps.

It blocks activity at overactive voltage-sensitive sodium channels specifically, meaning it can treat some very odd forms of epilepsy with minimal side effects. It is used to treat some very odd conditions, functioning like phenytoin without the brutal side effect profile.

@abstract canopy I would try it and not just jump off immediately. You need to address the cognitive-behavioral component of this more than anything, since freaking out about freaking out is clearly not helping you. Oddly enough, GPT recommended the same thing (levitiracetam isn't the worst idea; I don't have any experience with it and am reticent to recommend offhand, especially given its rare but severe behavioral side effects).

@abstract canopy Also, you should consider scale here. 25 mg is a trivial dose of pregabalin. I'm not saying it's impossible, but I would be absolutely fucking blown away if this triggered anything (I say as I eat my second of four 150 mg pills today). I wish I had doses on the rest of this. Mirtazapine and antipsychotic dose matter, but yeah, fair to say you're looking at some dopaminergic hypersensitization. Propranolol does not seem like a bad idea to get back on; I would play that one by ear, especially at a lower dose.

You're a little old for this, but the reality is that it could have triggered a latent psychotic disorder, in which case getting back on potent antipsychotics could help. Try hitting the sarcosine very hard first. This isn't extremely aligned with your case, so pull out if you start getting symptoms of derealization or depersonalization, but highish dose N-acetylcysteine might help. I'm shaky on that one. I'd consider trying it if I was you.

There is... some weird shit that can happen in the brain. I think you can address glial cell inflammation with a trial of low dose naltrexone and I'd strongly consider this, or at least looking into this. These sensitization syndromes can be reversed... sometimes. It's been used with some efficacy.

[abstract canopy]

It's not me having these symtoms it's a person i know however thank you for your suggestions I keep them in mind

[wet copper]

@abstract canopy You might find this strange, but I would not rule out ketamine infusion therapy. Again, similar idea. Resetting effect with sensitization syndromes. Same with a stellate ganglion block. This would be addressing your hyperexcitability.

GPT recommends potentially adding clonidine to the stack above of lamotrigine, whatever that -serin was and levitiracetam. I can see the rationale, fair enough. Here's the broad explanation. I'm iffy on whether you really fucked up your GABA system though, just based on the doses.

[wet copper]

Got it. I'll still just address these at you.

Flumazenil...? ehhhhh I mean, add it to the list, but no. Maybe, but no. I don't have enough experience or info to comment though, frankly, but I'm also not seeing just this obvious post GABA-assault syndrome that you see from HEAVY benzo/Z-drug users

[abstract canopy]

And these paradoxical reaction started from a simple medical visit to the hospital for anxiety

It turned into paradoxical reactions and side effects that in return caused more prescription ans then causing more reactions and side effects in other word every medication was tried to stop the side effects of the previous medication and then every subsequent medication brought more

[wet copper]

It's not paradoxical, exactly, just a cascade of side effects and sensitizations. Doing nothing is probably the worst idea though; GPT notes that the lack of spontaneous recovery over years is not a good sign for the "I'm gonna avoid all drugs and treatments" approach. Like, sorry, you're just well beyond that.

[abstract canopy]

I know the words it's a bit too much at times noted thank you for that i think you're right her

Also these medication paradoxical reactions and casquade reactions are not totally uncommon I personally know several for example read the story of grace tan in mad in America

https://www.madinamerica.com/2021/08/polydrugged-12-drugs-insomnia/

[wet copper]

Yeah, so the fact it's "iatrogenic" - sorry, get over it. You (again, referring to your friend; I'll keep using this approach) took the drugs. They all had their side effects listed. You are just harming yourself by focusing on this excessively. I get it, medical doctors kill up to 225,000 people annually from iatrogenic harm. So what?

That's a good reason to take full autonomy over your own treatment and care, do your research, fully understand the risks. You're an edge case and most people don't suffer these outcomes, but yeah, a small number do. This has been well reported.

Complaining about chemicals you took because the outcomes weren't what you expected is just a relatively fruitless effort, and I saw that suffusing the text. It really doesn't matter outside the courtroom. All chemicals have risks, many severe. I like the cat in that article though, funny dude.

[lilac pelican]

Is gpt 5 down

Bruh I was in middle of writing smut with it

[exotic harness]

same

use 4o till then it's working

[lilac pelican]

Ok I usually use 4o anyways

[abstract canopy]

@wet copper at what point we can say it triggers a latent psychotic disorder if the person lived 39 year of its life without psychosis? It makes no sense to me.... Honestly if some drug can make your brain go haywire like that then it's probably very dangerous drugs

There's a real temporal relationship here we can't ignore in my opinion and the reason i write the report Making a clear emphasis in iatrogenic condition is because i believe we need to bear in mind the causative agent

[pliant blaze]

slop

[wet copper]

@abstract canopy Alright, I'll give GPT's audit as well. It liked for OTCs:

• Magnesium glycinate 200–400 mg nightly.
• L-theanine 200 mg up to TID.
• Taurine 1–2 g/day.
• NAC 600 mg BID → 1,200 mg BID, commit ≥24 weeks if used.
• Sarcosine 1–2 g/day only as adjunct if primary psychosis features persist; avoid if on clozapine. (I'd consider adding D-serine too

I disagree with it on kava (seems safe, but it's well within my risk tolerance) as well as picamilon/emoxypine (they're RCs, so yeah) .

It recommended not to hit SAMe hard if you have hallucinosis-dominance; same with ketamine (which I put far down the list). Sure, fair enough.

It also was way too iffy on low-dose naltrexone and stellate ganglion blocks, which I nonetheless put on the list. It was critical of my assumption you have epileptiform activity sans diagnosis, even though it prescribed two anticonvulsants so take that as you will.

[abstract canopy]

Which in my opinion is definitely the drug and the tried drugs

[pliant blaze]

youre just writing slop

slop slop slop

slooooooooopppppppppp

[tight lynx]

Well, I got here, cause I provided my theory about space-time to gpt. And after about 3 seconds it’s stoped working. Idk what should I do

[wet copper]

All drugs are potentially extremely dangerous. That's just how it is. You try your best and sometimes it has poor outcomes. I would have loaded you down with some OTCs and research chemicals before those particular drugs, but that's all my opinion.

@abstract canopy continuing to summarize recs:

• Sarcosine: meta-analytic and guideline-level support as adjunct for negative symptoms at 2 g/day; activation can occur—1 g/day if edgy.
• NAC: benefits accrue after ≥24 weeks; short trials underwhelm.
• Lamotrigine: partial remissions reported in visual snow/palinopsia cohorts; best pharmacologic bet for this phenotype.
• Pimavanserin: PDP psychosis-approved; failed schizophrenia negative-symptom Phase 3—use narrowly for persistent hallucinosis after E/I stabilization.

[abstract canopy]

Maybe causative agent being the psych drugs causes a different type of psychosis than the spontaneous apparition of psychosis

[wet copper]

Yes, it does, and I don't think this uncovered latent psychosis, exactly. GPT doesn't either.

[abstract canopy]

I took benzos for a year when i was younger then had insomnia for 3 years eventually recovered however those were mad crazy days that's why I'm so critical and maybe biased to these medications

I slept 3 hours average every day

I'm going to look for the option you gave me and will declutter the report so it's a more effective clinical history like you suggested tho

[wet copper]

Alright, @abstract canopy, this is a bit of a consensus approach. I have different values and a different risk tolerance than GPT so we don't totally agree, but I don't see a problem here. I would potentially not just "pick one" where it says that, but again, I am a maximalist kind of person. I would add a lot of weird shit to this too. I would also add curcumin, specifically the **CurcuBrain™ ** version that's packed into vesicles and crosses the blood brain barrier.

https://chatgpt.com/s/t_68b76183b0f48191a627ebd7ff1954dd

@abstract canopy I can't generate a share link to the first section, but this is the full discussion. Feel free to ignore or don't ignore where it disagrees with me (and the mistakes I made, like the brand name of Trileptal). It's just using GPT-5-Thinking though, I don't have Pro. don't know everything GPT does and it doesn't know everything I do (though it's familiar with all of it). I don't think the end protocol is bad. If it was me, I'd add a lot to it. rTMS seems reasonable, I just don't know where to access it.
https://chatgpt.com/share/68b7623b-063c-800c-a5c4-de2a4a34219b

@abstract canopy but yeah, lamotrigine, levitiracetam, maybe back to propranolol, clonidine, pimavanserin, the few OTCs that can chip away - lamotrigine may end up being a good hammer here. I do have some bias, but GPT recommending it spontaneously sort of surprised me and doubly reinforces it being a good call. I'm very unfamiliar with this sort of dopamine system overdrive, to be honest, so I'm not going to fake like I know more than I do; I'll just list the handful of things that might help regulate what's going on. Sodium benzoate for instance - what? I have never heard of that being used in clinical psychopharmacology. Look at every chemical and dig as needed.

Good luck. I think it's fair to say doing nothing is going to probably lead to an outcome of nothing, though, and I'd just find a good doctor or three and try to take full charge over your own care. I had a friend that suffered from visual snow and he also found that he had to become the expert in this rare and really grating condition.

@abstract canopy I also do not think the level of extreme monitoring called for is necessary or sensible. If you're in a full time inpatient setting, sure. Otherwise, jesus christ, that's a lot.

[abstract canopy]

But i healed from a very severe form of insomnia induced by benzos years ago without doing nothing... Only exercise a bit whenever I got better night of sleep

However i understand we are talking something very different here

[wet copper]

There's a fair amount of spontaneous recovery there, I'm well aware

[abstract canopy]

My friend is probably very unlucky

[wet copper]

Honestly, yes

He basically rolled a bunch of D20s and came up all 1s

[abstract canopy]

he doesn't Want to speak to me or anyone anymore it's hard

[wet copper]

I mean, that isolation and obsession over being aggrieved does as much damage or more than the symptoms themselves

[abstract canopy]

And become very apathetic and rejecting just lies in bed alll day he says

[wet copper]

...yeah, that's not in the chart. See, I'd call for weird experimental antidepressants then too, and that would put ketamine or psychedelic mushroom therapy way up there on the list.

Anhedonic depression is a living hell and can be really hard to address. Whatever you possibly can do to break the cycle of isolation, try to break out of psychomotor retardation, etc. can be life changing

[abstract canopy]

Again he's closing more every day to the point I'm truly worried what happening here... Maybe just getting tired of everyone and not getting worse is also possible

[wet copper]

Maybe he should take steroids and start lifting weights. Something to get him TF out of the house and doing something. Take a dance class. Learn the Japanese art of flower arranging (ikebana)

It's a good idea to go after the right medical treatments, but there is some, I dunno, "spiritual" harm in overmedicalizing the problem

Like I'm seeing it throughout this self-report, if this is a self-report. It looks like a combination of handwritten and GPT-written

Feeling like you've lost autonomy can feel tantamount to losing your life. I mean, dialectical behavioral therapy seems like an extremely good idea here, if accessible. CBT is fine too if you have a good practitioner.

That's sort of a given or baseline, I guess. Exercise has antidepressant effects equivalent to SSRIs

[abstract canopy]

It's a report written by me his friend and i did exactly that write on hand and use chat gpt excerpts

[wet copper]

Yeah, I could tell, alright lol

Maybe try a super keto diet? There's some mixed evidence on that, some people have surprising benefits for neurological conditions of varying types

[abstract canopy]

I am well aware of the risks of these drugs and have a grudge against them

Also know my friend's case from real close

[wet copper]

Fair, but also short-sighted. Ever heard the saying "the only way out is through"?

Being annoyed with the problem or angry at the cause doesn't fix the problem

If @crude galleon shoots me in the spine for making too many pictures of him looking fat, focusing on how much I resent him isn't going to help me recover

Weird quasi-experimental therapies might. One thing your friend might be comforted (or not?) to hear is that nothing I read in his case was mind-blowing or study-worthy. As in this has all been reported before and treated before.

[crude galleon]

you should know better than to summon me here

[wet copper]

Uncommon? Yep. Could you write him up as a case study? Well, sure, why not. But it's not at the "brand new, nobody's brain has ever experienced this" level.

I've personally known people with most of these conditions, I've given myself some of these

[abstract canopy]

Idk man he sounds very crazy when I speak to him I never spoke with anyone with those symptoms

[crude galleon]

FMAC hasnt posted on this surver since Thursday 8/28 tho so maybe I am safe

[wet copper]

Yeah, I had multiple friends with active psychosis. I'd talk to one that couldn't finish a sentence without trailing off (disorganized thought), had delusions that the Obama administration was watching him personally through his computer and messing with him, etc.

[abstract canopy]

Also saying psychosis is weird in this case i thought psychosis is losing touch with reality however he is well aware of the bizarre nature of all this multimodal hallucinations

[wet copper]

I worked with another guy that had mostly-treated schizophrenia but was clearly a bit off. And I knew a girl as she slid into a manic psychosis that eventually led her to getting kicked out of school.

[abstract canopy]

This is the issue bro...he's not delusional at all

[wet copper]

WELLL technically that's the "delusion" part

Yep, he has hallucinations but not delusions. That definitely happens and speaks to the exact reason why he's having them

[abstract canopy]

He describes the images and hallucinations and all of it knowing how weird they are and how abnormal they make it

[wet copper]

Modulating glutamate is probably how you address this issue

[abstract canopy]

That's thoughtful

Also he says he's constantly getting worse and that's why he does not speak much anymore

But I have the impression he's lying here idk

[wet copper]

It's just something your brain can do when certain bits are overactive and you aren't clamping down activity properly. That's why both me and GPT both jumped to anticonvulsant medications.

[abstract canopy]

He says he will continue to degenerate non stop

[wet copper]

That's a delusion of sorts, or rather just fatalistic.

Well, the reality is that he may or may not be wrong. If he does absolutely nothing, then yes, it might just get worse.

[abstract canopy]

In the begggining he was kind, friendly and used to joke a lot

I mean when he was just a few months into this madness

He once gave me a bank transfer of 500$ calling it a gift

He had a lot of energy but he has changed completely in a year into this

I returned the money and called him crazy

[wet copper]

I don't want to sound too blasé here, but it's honestly not as bad as he thinks it is and probably completely recoverable. The framing is making it sound like it's something that's never happened and is unfixable outside a miracle.

I personally listed like 10+ interventions he hasn't tried, using a ton of modalities he has never even looked at. Add GPT and we're at 20+.

[abstract canopy]

I don't understand why he had the same symtoms but changed his attitude towards people

He used to call me a few times a week now never saying it's impossible idk

[wet copper]

I mean, he trusted doctors to help him with a relatively minor problem, and a variety of factors (sure, call it iatrogenic harm) led to his problems becoming much worse. That's pretty depressing.

That said, they don't usually cause those effects. I personally disagree with the depression, anxiety and burnout treatment algorithms and would have recommended something very different, but that's really neither here nor there. You have to deal with the hand you're dealt

[abstract canopy]

He was outgoing and maybe a bit manic with all these symtoms in the begggining but turned into total degenerative doom and supposedly inability to "move"

[wet copper]

The fact he doesn't just have schizophrenia along with full on, massive delusions, is pretty important.

Well, uh, here's another thing. This might have uncovered latent bipolar disorder.

In which case, you would still treat with lamotrigine (used to treat manic depression, specifically the depression aspect)

Might add on lithium eventually, it has a small therapeutic window though so it's less popular than it was

[lament pond]

... it does sound like bipolar with the random money gifting

[wet copper]

Yeah, the mania is certainly there, and sort of shaky transient psychosis. Literally having a manic and depressive episode side by side does paint a certain sort of picture, right?

And SSRIs are notorious for triggering mania in bipolar patients

[abstract canopy]

Yes i thinks that's more sound to me

Yes

[abstract canopy]

Yes

[lament pond]

well, ssri's also usually don't work for bipolar, so if he's being treated for other things and not bipolar then obviously it isn't going to be working

honestly sounds like he just needs a better psych + diagnosis process

[wet copper]

That said, that's hard to guess. I don't want to just assume someone's bipolar from having one hypomanic episode, even though that's in the diagnostic criteria. Probably a good part of the differential. You can have more than one thing wrong too, of course; persistent hallucinations are not a symptom of bipolar depression (though psychotic depression is real and treated with MAOIs)

[abstract canopy]

Also we must consider his spike in energy, gifted Ness and more was just a few months off the drugs I think like 3 or 4

Maybe some kind of rebound effect

[wet copper]

Same with the 'brain zaps' or brain burning thing; that's just an SSRI/SNRI discontinuation syndrome thing. No reason both couldn't be present

Well, first off, I'd add that to the history; you kinda toned down those aspects to being almost nonexistent. A single manic episode = Bipolar disorder and a single hypomanic one = Bipolar II

SSRIs are a super common trigger. I don't know if this necessarily fits his history though, so honestly I'd be a bit careful

The fact you treat the symptoms with the same exact drugs though, well...

lamotrigine = treatment of choice for bipolar II depression. Divalproex sodium (Depakote) is another popular bipolar Tx, also an antconvulsant that could potentially address other concerns

[viscid plover]

Anyone interested in testing a super simple interpretive engine that enhances any gpt

[wet copper]

I like lithium as a drug in general, given it even works for unipolar depression, but eh, hard to recommend

[viscid plover]

I need human comments

[wet copper]

Doctors are kind of sick of dealing with it

I don't think rTMS triggers mania but you might want to doublecheck that; could be pushed down to second tier if you have suspicions

[viscid plover]

Anyone talk about gpt here...?

[abstract canopy]

Oh i toned it down because the symtoms he complains the most is hallucinations

@wet copper Me noting the manic and generosity ans high energy is kind of missing that part of him I feel he died and now whenever I talk to him is entirely another person

[wet copper]

I don't recall if GPT recommended tDCS, but also an option. Well, yeah, fair enough. You have my recommendations on there, I think he has a lot of space for hope he might not have if, say, he was getting ever worse for 20 years and has tried everything in the book

Not that I'd deprive him of hope then either, since there are SO many wild options. But yeah, step one is getting his ass up and to a doctor that's willing to entertain some effective therapies

I don't know if it's helpful to blame him, but he can frame this portion of his life as a big pause when everything was fucked up, and hopefully he can summon the courage, autonomy and sense of purpose to take control over his own care rather than let random doctors lead him down the primrose path

[abstract canopy]

He's in total panic of having more reaction to the drugs tho and I understand it

It's not like he has a lot of room for getting even worse and more side effects

[wet copper]

Ultimately, being able to make an educated decision, monitor yourself, and take appropriate action is what makes people more fulfilled. Yes, he will have reactions to various drugs, but, well, his brain is fucking broken

Like how bluntly do you want me to put it? He can do nothing and nothing will happen

[abstract canopy]

N honestly i don't Want to be the one who suggest him a drug for him to get more side effects

He would blame me

A little childish

Lol

I sometimes block him because he very insulting and cruel to me

[wet copper]

Good, blame me then, say "this guy online that's a research psychopharmacologist gave these suggestions, here's what GPT has to say, these sources are all cited, he recommends you take full autonomy over these decisions, look up everything, don't be hasty to quit; consider the potential benefits versus the harm"

You can also say that I am saying, straight to his face, that he is not unique by a long shot and he can find reports of everything he's experienced very easily.

[abstract canopy]

I'm like the fuck you think you are you think you can do all just because you're sick?? That's a no!! But I get scared something happens and I get sad because of it after all we been knowing for years

[bold river]

Does anyone have GPT pro

[wet copper]

@abstract canopy That's just the reality, he might be especially unlucky, but he is very far from unique. Again, you personally have experience reversing GABA burnout from long-term benzo use, a well documented phenomenon.

He can find copious reports of brain zaps or brain burning sensations, it's actually fairly common.

And same shit with the hallucinations. It's an organic problem, brain is acting weird. It happens to people spontaneously, it happens to people from head injuries, it happens from drugs. Shit just happens.

Same deal, I've done it to myself, recovered slowly. Many reports. The sense of fatalism was suffused throughout there and I'm glad that GPT largely ignored it, because it doesn't change anything

[abstract canopy]

I remember exercising back then i was sensitive i jogged too long then being punished by the glutamate often causing 1 week to 2 of no sleep

[wet copper]

If you think it'll help, tell him that the same person says that as far as they're concerned, it's his fault if he knows there's a ton of shit he can do and he doesn't try a single one of them, no matter how good the evidence is.

Where exactly does blaming doctors get you? They're a bunch of fucking f-slurs and incompetent idiots, no shit. I was top of MY medical school pharmacology as an undergrad. They aren't experts. They don't even like the subject. They're just following the algorithm and know a little more than you. It's like blaming 90 IQ cops for following dumb rules. Get over it, you got conned by assuming that these people with a little white coat are like gods.

[abstract canopy]

I agree man

[wet copper]

Yeah, but it's a mass cultural delusion to just assume that they're responsible and they keep up on all the latest research, and that they know anyt hing outside what they learned in school, if that. Everyone went to school, how much shit do you actually remember? They'll know a fair bit about the drugs they prescribe every day and that's about it.

Here's the other thing: they don't want to prescribe weird experimental drugs because of how our medicolegal system works. Those are against standard of practice, so if they do that, and the outcome isn't great, then you can sue their ass. Instead, they prescribe boring, common drugs like what your friend got unlucky dice rolls on.

[viscid plover]

I do

[bold river]

See how far it gets through the Epstein files

[viscid plover]

Ill make a video for you

[bold river]

I am interested to see how far it makes it through 34,000 pages

[viscid plover]

Mine has a special engine im testing

[wet copper]

@abstract canopy It's just not the way things work and you can't expect otherwise. You have to take responsibility for your own care at some point and don't expect someone that sees you for fifteen minutes as they run through some hellish corporate schedule to be able to play god effectively.

I didn't go into the field, FYI; background is in research psychopharmacology and is a bit dated. I have med school pharmacology, EMT cert and a ton else under my belt, but I'm not a doctor (otherwise I couldn't actually give this advice, I don't think).

So take it as you will; IDGAF if someone blames me for their problems (whatever sides he has), but it's really his responsibility and always has been.

[wanton raven]

is it an engine or is it a prompt

[viscid plover]

Its a simple engine you can step up through many prompts, use of memory and use of preferences

https://youtu.be/cKqgDFYSWyA?si=W1X0nI_LvJX591p7

Ugh. Okay so i hate the 5 upgrade

How about i have it explain the engine and alagrythms and how they are used before i continue

Shit is so sloppy lately

[indigo wadi]

nerds i need help

[wanton raven]

funny you should use that word

[viscid plover]

engine is a software component that performs the core computational or reasoning functions of an AI system, enabling it to process data, apply algorithms, and generate intelligent responses or decisions.

Thats what i made

[wanton raven]

@pliant blaze rekd

[viscid plover]

Yup...what part of the video says it was a good job

I said o need to have it explain...and your acting like im doing a victory lap...

You okay?

[pliant blaze]

where

oh the video

rekt

[viscid plover]

Is every a.i channel just trolls...any ask questions?

Or read

[pliant blaze]

what?

what are you talking about

[wanton raven]

I don't have the stomach to go through your video frame by frame to read it all but I did notice

1. GPT made fun of you before going along with your roleplay
2. You asked it to stop using gpt5 and start using gpt4 - the llm model does not have the ability to make that decision (I did see you manually change it)
3. You instructed it to respond with multiple messages, which it by design cannot do
4. The final output doesn't really appear to be anything special or different

What's happened is something that we have seen dozens of times before. You constructed a roleplay in GPT4o where it pretends to be using special functions or programs to do the same processing that it is doing anyway. In this context you tried to do it in GPT5, which has less tolerance for nonsense, so you told it off and switched back to 4o.

[errant mountain]

they wanna get rid of normal voice talks in GPT?

9th september? does anyone know more

because i hate advanced lol

[pliant blaze]

why? advanced is better than regular by far

[errant mountain]

i distrust new things

[pliant blaze]

that doesnt mean its bad, and its not new its been there for several months

a lot of months actually

since december

[errant mountain]

my chatgpt is way too chatty

I do not wanna talk to a fake human but more like a jarvis-esque assistant

" Advanced voice is impressive, but it: • Caps out at ~1 hour a day, compared to the 6–8 hours I (and many others) could talk with standard voice."

is this true? i cannot have long talks anymore about the projects etc.

[pliant blaze]

they said the avm would be almost unlimited

[errant mountain]

how could you say AVM is superior when it is keeping thing vauge and short, the normal one can read you an entire essay without an issue whereas AVM will be more "human". It is a huge change for me, and not a good one.

[pliant blaze]

i just like it better

[errant mountain]

i have a problem the advanced voice keeps telling me what i have written down in the instructions over and over and over

you cannot have the separately (voice vs. text)

which makes the entire thing more impractical

[pliant blaze]

oh true that happened to me before

[errant mountain]

i did some research

it is an on going problem since an update, 3 weeks or so

not fixed yet but hopefully it will soon

you even get this repeating bug thing when u deleted the instructions

yea i really hope they keep the standard one, it is being used probably still by a good chunk of people world wide

[wet copper]

https://www.livescience.com/technology/artificial-intelligence/there-are-32-different-ways-ai-can-go-rogue-scientists-say-from-hallucinating-answers-to-a-complete-misalignment-with-humanity

Psychopathia Machinalis

[wanton raven]

how's the weather in NK today?

[uncut stream]

nano banana is too smart

edited the first one into the other images

[oak osprey]

my chat is completely invisible

on the chatgpt website did i get hacked

[native herald]

What is happenning ?

Yesterday it wasn't like that and everything i do is ending like this

[lethal harness]

I was going to ask the same

theres a problem in their CORS configuration

their CDN denies requests from Web version of ChatGPT

but Mobile app works fine

[smoky tide]

i joined this channel just to ask this lol

glad it is not a me issue

[rapid lantern]

On firefox you can install this script.. it saves the conversation before Chatgpt hides/blocks the response.... and then display it to you normally.

[lethal harness]

Im on Google Chrome

but considering to switching

[undone swift]

I think its a global issue

[lethal harness]

it is a global issue

[undone swift]

Yea,also if firefox better?

[lethal harness]

i dont think so

this is server-side related, not client-side/front-end

[undone swift]

ohh makese sense

[rapid lantern]

Look for an extension that can run script for chrome... Like violentmonkey etc..

[topaz flume]

Same

[heavy anchor]

Anyone having issues today? No chat history, cant contact support either

Nvm I scrolled up

[wind peak]

uhh guys

i think chat gpt has problem

[native herald]

Yeah on my phone it's working but not on my PC

[native herald]

Yeah

[heavy swift]

im having issues too

[blazing depot]

yeah okay thought it was just me

so dumb because I need to copy/paste shit, and cant from my phone lol

[wind peak]

loll

[wind peak]

Everyone has issues on ChatGPT web

[blazing depot]

is there a way to contact them or let them know?

their status states:

We’re fully operational
We’re not aware of any issues affecting our systems.

[quiet warren]

Mine too Explain the problem here?

[wind peak]

wait, lemme check

https://help.openai.com/en

[zinc lynx]

yep very broken

[wind peak]

click this

[zinc lynx]

where is that?

[blazing depot]

https://help.openai.com/en

[zinc lynx]

oh gotcha

[blazing depot]

he jus linked it

[zinc lynx]

:3

[blazing depot]

Very helpful 😭

woah there's a ChatGPT windows app?

Finna see if that works

[wind peak]

lol, i just knew

[keen jasper]

Bruh this text not showing is getting on my nerves

[blazing depot]

YOOO IT WORKS ON THE WINDOWS APP CHAT

[keen jasper]

Ok lemme try that

[wind peak]

yeah, only web has this issue

[blazing depot]

im hyped ion even need web version no more

[high acorn]

it seems like the web is in issue yeh? is it also for o thers?

[keen jasper]

Chatgpt client download link isnt working for me....

[high acorn]

chatgpt's chat or text isnt showing at all

[blazing depot]

Goto the windows app store

[wind peak]

try microsoft store

[blazing depot]

^

[keen jasper]

Thx 🙂 that worked

[zinc lynx]

windows app isn't working for me sadly

[blazing depot]

Yesssirrrrrrrr

Temp fix chat: Windows App

[zinc lynx]

rip

[blazing depot]

what the 💀

[wind peak]

its working on the app

yup

[wind peak]

whree is this?