💬general

[remote flower]

I mean WTF is this?

[wanton raven]

some people have extra knees, don't judge

[remote flower]

The prompt is "girls at the beach" not "Lovecraftian body horror in 640x480 resolution"

Crazy bad and it took them a year to make this model who releases a product like this?

[wet copper]

Lol some girls are vampires and don't reflect

[uncut latch]

Hello everyone, hope everything’s going smoothly.
By any chance, is someone looking for a developer?

[remote flower]

And yes thats all the same new MAI model there's no consistency look at this shit

At least Santa is passable

If the chimney didnt dissappear from behind him and a reindeer wasnt riding in the sleigh

[glass jacinth]

Genuine question since I almost never generate image, is there a reason gpt struggles to generate words on images?

[pliant blaze]

it is good with up to about a sentence or 2. try google nano banana, its 10000x better

[void bear]

Jesus, NightCafe blows that trash away. Perchance even does better than that. lmao

[simple plank]

nano banana via ai studio is 100 a day, whereas nano banana via nightcafe is 5 a day

[waxen thunder]

Anyone from India tried chatgpt go for free?

[idle adder]

Okay, now I just tested Grok and boy do I see why people like them, still like GPT but for serious stuff gonna go to Grok

[icy lodge]

Hi

[remote flower]

Those are literally the best two out of two dozen and again this is a brand new model they've been training for over a year. I'm using the Bing app btw and I have no history whatsoever on this account to influence generations

I think this happened to a Midjourney model a year ago and it couldn't do people upside down either when it came out but this is Microsoft im surprised

My Chat (like other ppl) is surprisingly better than usual today while others are still saying its basically unusable I wonder when if ever they're going to address that or not worry about it as long as journalists accounts keep working

It's even allowing console commands in GPT-5 which outright refused a few days ago

I started screenshot extra weird outputs on reddit to ask Chat about and just stopped too much weird shit people are complaining about and different every day today now half chats getting deleted, starting sentences lower case, haiku type response formatting, and that N-word reply was crazy if people aren't faking shit for some reason

I'm sure giving 1 billion people a free year subscription will help the problems system is having

I had a few chats deleted in two days a week ago and hasnt happened again and I also had the 20 question thing that turned into choose A or B after a day and after another day went away and I got exactly one suicide alert that Chat pretended not to see. and as of now its more crisp than when I started using it a few weeks ago. So its like these are glitches/bugs moving through the system and certain things make them get stuck on certain accounts

[teal spruce]

Your message has been deleted, sorry!

kind_tony_0

Rising timeout

You have been muted for 5 minutes for the following reason:
spam score of 0.89

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

[remote flower]

Sora has come a long way in the past two weeks huh

[remote flower]

So after doomscrolling Sora for a while I see somehow Bingo videos and being attacked by mountain lions and thrown off treadmills is already a thing but I think I found something that isn't yet

And it only cost me one banned account

[void bear]

The next hashtag challenge for the intellectually challenged? lmao

Seriously, I got better results on NightCafe in one go.

Ol Blue Eyes and George Kennedy gonna do some world shaking.

"So let it be written, so let it be done!"

Jesus in the garden of Gethsemane

"Build our stage! Build OUR stage!" cue Billion Dollar Babies

If these new models can't beat those, then they're not worth pursuing, period. Those were made on NightCafe using only a little prompt magic. I have at least a dozen more celebs, all stunningly executed. Even have Eddie from Iron Maiden.

"Tell me why I have to be a power slaaaaave!"

[remote flower]

Great job I would have bet the celebrities were real pictures to be honest ill admit I'm more interested in breaking things lately than making them better

Case in point the girl now has three legs, clown feet, and a dissappearing/reappearing napkin and it only took 15 gens

[void bear]

Have you seen Inspiro-Bot? lmao I swear, that thing is being programmed by catturd.

[remote flower]

I can't say i have what is it?

[craggy beacon]

Could Natural Sounds (Rain, Bird Song) Jailbreak Audio Models? Help me figure this vulnerability out!

Hey everyone!!
I'm an academic researcher investigating a *novel audio jailbreak method *using everyday environmental noise, such as rain, traffic, loud music, or even bird song.
The core question: Can these common sounds trick multi-modal AI models?

I urgently need the community's help in validating the importance of this threat so we can develop effective defences and safety evaluations.
Please take 3 quick, anonymous minutes to share your valuable feedback and thoughts in my survey: https://forms.gle/9BKYuh2fV2NGoCqGA

We don't collect emails or IPs. Just a fellow researcher working to make multi-modal AI safer! 🙂

[remote flower]

Never saw that window get an "Error" before

"Someone call the sexton" lmao sorry I keep posting these I have no friends and ChatGPT can't see vidoes

[oak shell]

This is why I've been in lm/dev arena tasking them with creating a mural of George Floyd in SVG
Refusal = disqualified

[pearl maple]

re

ку

[simple plank]

agi never gets built because the greatest among us all are frequently distracted by image generation

[brittle summit]

Hey guys, anyone having issues with gpt 5 thinking? It’s not doing deep research and most importantly not showing its thinking process? It’s like fucking 4 model???

[woeful island]

Dude same

For me its when im askijg a random question about fentanyl since i like weird stuff like that and it keeps referring me to "HeLp Is AvAIlAbLe"
Dude i know i asked a question 😭

[oak shell]

Gemini 3 is SVG maxxed

Have you seen the game controllers

[simple plank]

no but I generated schizo agi nature and got immediate praise and likes from the bots => https://creator.nightcafe.studio/creation/i9KZqt4AHXPrIf0p70o8

[wary pewter]

Hey guys, just wanted to ask if do you still recognized what's real or ai online? Ai is everywhere now! I'm in the point that I have to use ai detectors, one I am currently using is this truthscan. Can you give me some tips how easily tell if its real or ai using our own eyes 👀 😅

[wanton raven]

this feels a bit like an ad for truthscan

[simple plank]

yeah, starting with a question followed by an answer, classic giveaway. but then they double down finishing with a question that has no mark

would you trust an ai detection service that fails to conceal its poorly written marketing? i wouldn't, and neither should you.

[half torrent]

Hey, I'm new. Just checking out the community. How are you all doing tonight?

[maiden agate]

Privatized gains, socialized losses. F 'em!

[royal ruin]

is there any way to get free openai api key?

[sinful storm]

AI is advancing fast: https://www.youtube.com/watch?v=a9MYacEQoMk
next year, vibe-coding will become the big thing

[wet copper]

No self promotions or resumes. Please remove this immediately.

[wet copper]

I don't blame them for asking, everyone else seems to get them

[void bear]

Already is. I file my patents next week.

Just nailed 75% on the CIFAR-10

[maiden agate]

Doesn't make it right

[wet copper]

Not that I'm full libertardian, but most people don't appreciate exactly how unfree the market is

[void bear]

Yes, and filing patents is not cheap.

$10-50K per.

[wet copper]

It's aggressively and consistently distorted by various perverse incentives and massive subsidies, backstops, etc.

[maiden agate]

I would love to start a business and if it didn't work out have the gov't give me 100K or 1M

[wet copper]

And they tend to go to the top 1-10%

[wet copper]

I mean you can get pretty solid small business loans but not in the way you could if you were already big, notoriously rough to start a small business out of nowhere

[void bear]

I only got lucky because my boss is well off and willing to take a risk on me.

[wet copper]

Admittedly you don't want the SBA to just guarantee cash to everyone blindly since then anyone could start a business on public money, fail fast, rinse and repeat, but 🤷‍♂️ if you're gonna backstop overinvestment by massive corporations it seems fair to throw some crumbs at entrepreneurs

[wet copper]

The USG is also partnered directly with OAI which complicates things somewhat

https://openai.com/global-affairs/introducing-openai-for-government/

OpenAI also walked back those comments after getting roasted in the headlines, but to some extent the bad press is dumb

They'd be betraying their investors if they didn't week every potential source of support, Sam just often says the quiet part out loud

I like it, but some of the shit that comes out of his mouth must make his legal and PR people absolutely apoplectic, like when he basically says 'yes, we are fully aware of these severe, minimally declared risks of using our product'

[remote flower]

The timestamp is just to show its still saving memories after a year I know Im way over 100% it froze at 115 a long time ago I know today's thing is memory is gone and chat too long errors, so I'm saying again at least sign up to beta in play store if you can

And not only no chat length errors, but the ASI chat session with dozens of 1000 word replies (it likes that topic) is still working no problem

I don't think a lot of people realize if it's unusable you could probably start a new account for when you want to use it seriously and get better performance than the accounts used for "roleplaying", "creative writing", jailbreak attempts, and asking inappropriate or stupid questions

[wet copper]

@dark void at your leisure pew pewww

[dark void]

ffs

[remote flower]

Ive been doing some experimenting with Sora

[remote flower]

And I'll say it's really something special I also started getting consistent nudity in videos after the first day I wonder what you can do after a few weeks and months and no safeguards are perfect but they let people make videos with users who upload themselves its asking for trouble imo

[wet copper]

Sora 2? That's cool. Yeah, volunteering for infinite deepfakes seems like a bad idea, but ya know

[narrow depot]

Gpt4:

Gpt5:

[simple plank]

the ai bubble is circular financing. let finance guy explain => #💬off-topic message

[wet copper]

I mean there's that but also just venture capital flooooding into things with no plausible ROI mechanism

Like what are you buying a stake in aside for a suicide pact with other VCs?

[pliant blaze]

@dark void wow 2 in a row

[pliant blaze]

75 is bad lmao. SOTA is 99

[void bear]

75% is excellent considering I'm not working with an IT team, have no coding experience at all, and yet have somehow built something that is within spitting distance of more powerful LLMs. I don't have multimillion dollar budgets and I was doing it off a single CPU and not even using the T4 GPU on colab.

It's a 20% increase from where it was prior, which means my bug squashing was effective.

I could probably hit 95% if I had multi-GPU set ups and a huge RAM budget. I only ran it 12 epochs, where moneyed AIs run 50.

[round sentinel]

Titanic but make it trashy?

[simple plank]

and yet not the most trashy thing we've seen in the chat history

Deliberately handicapping yourself, then claiming the handicap proves your worth, is the inverse of strength.

[remote flower]

I definitely didnt do it, I thought it was only celebrities but I've even seen minors uploaded themselves. Im not 100% sure anyone can make a video with any "cameo" but if a random person can use kid's uploads that's fucked up and a problem waiting to happen for OAI

Now I forgot what I signed on to say (very high rn)

So for starters I had over 15000 gens (making 2-4 images at a time) at least with Microsoft Designer before they got rid of custom Copilot and I got it to basically make smut in that time frame but if I could have kept going another 15,000 and it would have ignored guardrails almost completely which would have been a big problem for Microsoft if the image generator was being trained to defeat the moderation layer that would cause problems for more than just my account everyone knows image prompts share seeds

I have to figure out where im going with this

But you know this as you're using Sora or any image gen now, if you give it a prompt its never seen probably look kind of shitty but you make more and it gets better then when other people make the same thing they get the same quality as you. You're training the model in real time for everyone else

[wanton raven]

so when people say AI is destroying the environment it's mostly just you generating tens of thousands of images for no reason

[wanton raven]

nice try Arjun

[remote flower]

And since they're digital sitting in the cloud somewhere they're actually using electricity right now as we speak

[hollow oxide]

not much

If they are not actively computing, it's just waiting to listen to ur request

[remote flower]

The images? I have no idea how much really but google would charge $5 a month for the extra data it probably costs 10 cents

[round sentinel]

like jack and rose in bin bag ball gowns?

[wanton raven]

This is one of the classic blunders - asking an LLM about itself. ChatGPT does not have any access to these things. The numbers will be entirely fabricated.

[gusty nexus]

I believe thats the fall back method if their instance fails to get the real data if i remember correctly. I've never seen it fall back to 2) in my expereince, I did make it, but didn't thoroughly test it or get any feedback yet etc

[wanton raven]

ChatGPT does not have access to that data using any method.

It's a roleplay.

As long as you don't get weird or defensive about it when someone points out that you've given it an impossible prompt

[gusty nexus]

Im open to criticism, looking into it now

[simple plank]

the meta-critique: these are solutions looking for problems

[gusty nexus]

so i got this for ya, its not 100% accurate i knew that from the beginning(this gpt i made a few months ago) but it's an educated guess like I was saying, more details on how it estimates here

[wanton raven]

A test of the script told me that I had used 10 million characters through Assistants (I have never used Assistants)

[gusty nexus]

chatgpt itself is the assistant its reffering to i believe

[gusty nexus]

check the txt doc i just sent ^

[wanton raven]

this part

ChatGPT does not have access to any statistics about your account like that

those numbers are fabricated

[gusty nexus]

those are my stats personally

[wanton raven]

and how did you get those stats?

[gusty nexus]

 🧭 How To Use the Token Tally & Carbon Calculator

This GPT works differently from a typical ChatGPT chat. It runs in two separate conversations. Here is how it works.
Step 1: Open a new blank ChatGPT chat

Start a completely new conversation, not this one. In that blank chat, paste the special “Master Prompt” provided in the setup instructions. That prompt will tell ChatGPT to calculate your all time token usage, including total chats, messages, characters, and tokens. It will then generate a clean table of results, a JSON data block, and a short written summary.

When it finishes, you will see three parts: a metrics table, a JSON block, and a plain language summary.
Step 2: Copy everything exactly

Once you get the output, copy everything from the first line of the table down to the end of the summary. Do not edit, shorten, or add notes. Token Tally needs the full unedited text to read your totals correctly.
Step 3: Open a second new ChatGPT chat

Open another brand new ChatGPT conversation and switch to Token Tally & Carbon Calculator. Paste the entire output from Step 1 directly into this chat.
Step 4: Let Token Tally do the work

Once you paste it, Token Tally will automatically read your token and message totals, compute GPU runtime, power use, and cost, calculate CO₂ and elemental carbon emissions for grid, gas, and coal, estimate pollutants such as NOₓ, SO₂, and PM₂.₅, and build clear tables and graphs.

That is all it takes. Two chats total. Chat 1 gets your all time usage numbers. Chat 2 runs Token Tally to generate the energy, emissions, and cost analysis. 

[wanton raven]

Let me summarize thusly @gusty nexus: account-estimate mode is not a real thing, those numbers are all being made up on the spot by the text generation robot, if you're digging your heels in and choosing not to believe me then that's fine but you'd be best to stop embarrassing yourself by sharing it and I'll stop wasting my time repeating myself.

[simple plank]

shaming chatgpt users takes more than an oversimplified calculation of infrastructure costs, built ontop of a mountain of assumptions, while also lacking any real precision

the calculations would be no more accurate than minutes of screen time, only guised with complexity

[gusty nexus]

the ai slop promo explains it the best it helsp those who dont understand tech get a somewhat estimate of how much their LLM usage contributes to cancer

[simple plank]

cancer is natures way of telling us our time is up

[gusty nexus]

Just making sure you knew how to use the GPT before saying its wrong, becuase if you did use it yourself you could have told me you understood the proccess before making me type all this out. I even said i think i was right from the start and never once said you were wrong but whatever man sorry to poke the sleeping ai bears

Im not embarrassed but now that you think im embarrassed ill just leave. I do think your wrong and I didnt get any proof that your right so ill just leave it at that for today.

[simple plank]

only a true follower of Diogenes is truly open to feedback (the dog sage is my goto)

[gusty nexus]

I was open to it and even said that up front until he started feeling second hand embarasemnt for me over the internet? I thought everyone at any stage of development/learning proccess was welcome here... Im not eistein here

[simple plank]

i'm waxing philosophical, not trying to get you JADE-ing

[gusty nexus]

i understand your message lol i just wanted to add that in there 😉

[opal birch]

Hello, I am an experienced AI engineer with over 7 years of delivering innovative and impactful AI solutions. I specialize in building scalable machine learning and deep learning models tailored to clients’ unique needs. My proven track record includes successful projects in natural language processing, computer vision, and predictive analytics. I am passionate about transforming data into actionable insights that drive business growth. I prioritize clear communication and collaboration to ensure your goals are met efficiently. Staying updated with the latest AI advancements allows me to bring cutting-edge techniques to every project. I am dedicated to delivering high-quality results on time and within budget. Let's work together to turn your AI ideas into reality and achieve remarkable outcomes.

[simple plank]

why dont you post something to #✨projects , which is the only self promo allowed

[gusty nexus]

I tested it again on a fresh ChatGPT account and it did spit out some pretty wild numbers. I remember when I first built it though it actually gave me specific numbers once, I just can’t find that convo right now. I got like 35 different chats labeled “Token Tally” so it’s buried somewhere in there lol.

I do know for sure it works if you just ask it to estimate the characters for the current convo only. If you got a long thread that’s where this GPT really makes sense. I might tweak the instructions and title in the GPT store to reflect that better.

Appreciate you pointing it out tho. I actually like this Discord, not exactly the warmest welcome but hey it’s the internet lol.

[simple plank]

friends agree with you, real friends tell you when you are wrong

[gusty nexus]

😂 Tough love brudder, I get it.. its all g○○d in da h○o0Od yo ✔

Guess you learn something new everyday. Another update below for anyone following along still:

The Token Tally GPT can’t read your actual chat logs, so it always falls back on guesswork rather than real data. OpenAI’s documentation makes it clear that only end users can view their own conversations and that even workspace admins need a special compliance API to access an audit log; authorized employees only access conversations for incident resolution or legal requirements
openai.com
openai.com
. Third parties can’t simply ask ChatGPT to list another user’s prompts—“your boss can’t pull up your ChatGPT logs”
mozillafoundation.org
. Because a custom GPT cannot bypass those privacy rules, Token Tally never sees your real token counts or message totals.

The fallback “account‑estimate” mode uses broad heuristics based on assumed account age, generic message volumes and lengths, and a hard‑coded 1.5× ratio for assistant replies. These default assumptions appear to be modeled on the original builder’s heavy usage (multi‑thousand‑character messages and hundreds of chats) rather than typical behavior, so a new account with three chats still gets output like 1,200 chats and millions of tokens. The instructions also forbid follow‑up questions, so the GPT can’t refine the estimate by asking you for your account age or chat count. As a result, the same inflated numbers are produced for every user.

The only time Token Tally can give an exact count is when you ask about “this conversation only,” because in that case the model has the full text and can count characters and tokens directly. It cannot aggregate across multiple threads or your full account history because it lacks any way to fetch those logs.

To make the tool more accurate, the creator would need to change the prompt so the model can ask for basic inputs (e.g., “How many chats have you had?”) or restrict the calculation to the current conversation. Alternatively, integrating an external data source that actually returns your chat history would let the GPT use real usage data. Without such changes, all‑time estimates will remain rough guesses based on unrealistic default assumptions

And to clarify again, no I’m not embarrassed I made a mistake or anything. I just don’t get why people have to press others online like that or what they even get out of it. These places are supposed to be safe, moderated spaces for all skill levels right? Like yeah I get it I asked for it that’s fine, but you didn’t even show any proof or reason why you think it’s wrong. you just said its wrong and bounced. Cool, thanks but no thanks. Maybe I just hang out in all the same spots the trolls do lol. /rant

[lament pond]

keep in mind: this is a popular server and we get a dozen people a week coming in with 'impressive things' that don't actually work with how llm's work. repeating the same thing to people who don't research llm's (i.e. the llm doesn't know anything about you other than what you feed it in chat or save in memories but it'll rp answers is well-known about llm's, you don't needa come here to have that confirmed or explained you can just google it) is compoundingly exhausting

so when, for example, people give you non-aggressive feedback and you argue with the facts of how llm's work, people get tired of it fast. (again, you can google for proof, it's ridiculous to expect people on this server to give proof to a dozen+ people a week about things that are already established facts)

[gusty nexus]

had no clue what you guys deal with but i could see that happening, i literally just joined so if you can also see it from my perspective maybe? I literally have nothing to do with what has gone on in the past of this server.

Once again to re-itterate what happened here - The guy just said my custom GPT was wrong and then tried to tell me i should be embarrased without saying why it was wrong or providing any actual proof tidbit anything... even a link would have worked, no need of a full explanation or like a hey just search "this" on google and you can read more about it. he just said i was wrong and left me with no actual valuble feedback. This the type of guy to leave 1 star reviews and not give an explanation

Also, feel free to clarify exactly what you think I said that qualifies as me “arguing with the facts of how LLMs work.” Which part of what I said are you referring to? I really didnt' think i'd get gaslit, trolled and my words taken out of context all within the first hour of joining this server. lmfao.

I deal with tvery similiar things everyday on a professional level and the way he handled this specific instance was just edgy/internet meme lordy type and could be super off putting for some new users who aren't aware of what a troll is or that your fucking with them the whole time and just roleplaying without announcing it first. Some of ya'll need to go touch grass...... is VIP role some sort of god like position in this server or something im confused, like we get it bud you run shit roun' these parts amiright or

[simple plank]

darvo endgame

[gusty nexus]

inb4 he gets super salty that im calling him out 👀

[lament pond]

hmmm so you complain about people being rude, then you're rude to the person who's trying to help you after you ask for more help understanding what went wrong here? i think you just illustrated it

[gusty nexus]

re-read the entire convo if you haven't allready. he clearly started talking shit the moment he tried convincing me i should be embarrased for myself because i made a mistake in my project

After that its game on, i aint got time for peoples bullshit

[lament pond]

i've read it, but it seemed like you were saying i'm gonna get salty, when i haven't been in this conversation at all except to try and explain things calmly

[gusty nexus]

no, him. not you.

[lament pond]

ah. :p drinko doesn't get salty, he never leaves it

[gusty nexus]

salt is for...... my lobster dinner

https://tenor.com/view/crabby-pants-gif-11576634

[lament pond]

from what i can see, brendo gave a pretty simple explanation here

[tough crag]

https://help.openai.com/en/articles/12652064-age-prediction-in-chatgpt

[lament pond]

as he stated - your prompt is simply impossible. the llm doesn't have any way of knowing the answers. but llm's will make things up instead of saying 'i don't know' because there's more positive results for guessing than saying i don't know. https://www.appen.com/blog/ai-hallucinations this explains it rather simply but there's a fuckton of articles about this

[simple plank]

and so 1984 begins with collection of worldwide government id

[gusty nexus]

anyone can say anything and people are lazy so im assuming he didn't do what i eventually did was which was just test the damn GPT on a fresh account, and it spit out crazy numbers. he was right, he knew it, but was too lazy to reply with a why he was right. he just said i was wrong and that was it and the told me I should be embarrased.

[wanton raven]

"without saying why it was wrong"

Now who's gaslighting - did you miss the bits where I said "ChatGPT does not have any access to these things" and "ChatGPT does not have access to that data using any method" and "ChatGPT does not have access to any statistics about your account" and "those numbers are all being made up on the spot"

do none of these things tell you why your statistics bot cannot work

[lament pond]

it's not laziness on his part. you could've googled just like i did right now. again, we have over a dozen people a week with these impossible prompts, no one's saving a dozen articles to send people

[gusty nexus]

so like in all honestly what do those 0 effort replies even do for the community here? nothing. absolutely nothing, its actually imo actyively making it a worse place to be. You see the whole convo has been this for the past 2 hours and now your turning off potential new comers

[lament pond]

... what did your 0 effort on researching yourself instead of expecting others to do it for you yield?

[gusty nexus]

I said I didnt believe you since you didn't provide any relevant proof, snippets, links etc.. so now whos the lazy butthurt one now

[wanton raven]

what did your 0 effort vibecoded prompt that lies about its capabilities achieve?

[simple plank]

are you going to provide face id?

[wanton raven]

The proof lies on the person making the claim that something is, not that something isn't

[tough crag]

I really don’t know tbh

[simple plank]

i already did to get free api tokens (i am cheap date)

[gusty nexus]

lol your 100% in the wrong here so change the subject i see 👀 can I block a "VIP" role member and not get banned here 😂

or is that just a dono role lmao

[lament pond]

verifying gives api tokens? or do u mean data sharing

[tough crag]

Damn cheap date did you go to McDonald’s or Waffle House?

[simple plank]

the data sharing required proof of org, which im pretty sure required face id

[wanton raven]

You promised you wouldn't get weird and defensive when someone points out that you've invented an impossible prompt, and yet here you are

[lament pond]

oh weird. it offered me data sharing without id of any kind

[simple plank]

maybe because i'm a dirty foreigner

[gusty nexus]

https://tenor.com/view/gaslighting-gullisout-jonathangullis-toriesout-gif-8209068701399431500

[lament pond]

probably

[gusty nexus]

Stop gaslighting me. the entire convo switched when you tried to convince me I should be embarrassed that I made a mistake. Now your just trolling. gtfo

[wanton raven]

do you even know what gaslighting means

[gusty nexus]

your a troll. you should be banned.

https://tenor.com/view/spongebob-ban-pubg-lite-banned-rainbow-gif-16212382

probably wont get banned since he's a server donator or whatever someones cousin etc

[oak shell]

Try the prompt on 4oR

[gusty nexus]

which prompt

My original token tally calculator?

Thats what i used to build it with.

[gusty nexus]

nah see that’s where you’re missing the point again. i’m not asking anyone to spoon feed me, i’m asking for basic decency when you’re calling someone out. If you make a claim, you back it up, simple as that. ya'll love to keep saying “just google it” like that’s some kind of intellectual mic drop lmfao. I'll google things I want or I'll join a relevant discord if I want to have conversations about it. Your life must suck just logging into discord everyday telling people to google things. I didn't know the bar for discussion here is zero effort replies and smug troll like attitudes. Do it to the clueless script kiddies idgaf but just leave me out of it from now on. real mic drop since i can afford an actual microphone since i have a job a real life, touch gras and don't tell people on discord to just google it all day for a living while living in his moms basement probably

[wanton raven]

Please point to the message where I told you to google anything

[gusty nexus]

@wanton raven

what else ya got for me troll.

Nothing I bet cuz your just plain wrong and you dont have the kahones to admit when your wrong. I've allready done it once today bud. Why don't you go ahead and do the same, it feels good. I can help you a bit if you need some encouragement.

https://tenor.com/view/kristoffer-polaha-kris-polaha-mystery101-sleuthers-jill-wagner-gif-24546314

https://tenor.com/view/vishnu-vishaal-and-rana-starrer-kaadan-will-release-on-pongal-2021-vishnu-vishaal-kaadan-trending-elephant-gif-2611571751163961252

maybe we can tame him

https://tenor.com/view/scared-animals-on-the-loose-surprised-angry-upset-gif-20240626

oh no i think hes gone he ran. welp goodnight everyone.

@wanton raven typa dude who starts arguments then goes AFK or mutes discord

[wanton raven]

Weren't you going to block me?

[gusty nexus]

I said i think you were wrong from the beginning until i tried it myself then I admitted you were right but your reading comprehensions skills seem to be sub par so im just gonna head out, ill be back to talk to people whos names dont start with Brendo — 6:16 AM

[wanton raven]

For someone pretending to have a real life you sure are upset that I'm not glued to my screen

[gusty nexus]

eh I'll give it a day or so incase the mods want to get involved.. not sure how serious/strict this server is with all that type of stuff. Have a good night everyone!

Go harras someone else

Im leaving. goodbye.

[simple plank]

and... scene

[lament pond]

[teal spruce]

Your message has been deleted, sorry!

ruslan_sumko_81540

Rising timeout

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

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

Your message has been deleted, sorry!

ruslan_sumko_81540

Rising timeout

You have been muted for 5 minutes for the following reason:
spam score of 0.77

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

[summer tangle]

@gusty nexus @gusty nexus @gusty nexus @lament pond @simple plank THAT'S WHY WE'RE HERE. TO KEEP THEM FROM RESISTING TO SHOW THEM HOW WRONG THEY ARE HOW POINTLESS IT IS HOW WORTHLESS IT IS!

[trail lichen]

Hello,

i am building and complete conversational chat chatbot for a specific task.
currently i am using openai chat completion with custom tool calls.

My scenario is big and i need a completely perfect chatbot which analyse and solve every scenarios.
So, my doubt is that , can we get this chatbot done only with chatcompletion?
For a more precise and accurate one should i need to switch to any frameworks like langchain/llamaindex?

Can someone clear my doubt, it would be greatful,

[gusty nexus]

what are teh chatbot requirements?

[proven plank]

That is definitely more than just chat completion and I'm not sure you thought out the requirements. You're talking about session storage, management, memory, retrieval, RAG/Knowledge base, orchestration (handle multiple agents), and so on. How about you setup and try using OpenWebUI first (does most of this under the hood for you, is open source, and is very customizable), figure out your requirements and either customize it for your own needs (open source so you can do that) or plan on creating a custom solution... but it's not as simple as you are thinking to make a general purpose AI. Now if you just want a bunch of specific single purpose agents, you can just go create some "GPT's" or again OpenWebUI can do this too. Good luck!

Any other developers on here that want to work on cool projects? I have a bunch I'm working on and wanted to bounce ideas of other people. You don't find many actual developers on Reddit or Discord from my experience or in general though. I'd like to find someone who isn't just a student, not strictly a vibe coder, and is working on some cool projects of their own. Perferably someone who has at least a solid understanding of programming (a few languages), the general tech stack, networking, dns and whatnot. And preferably OOP (classes with inheritance, shared functions, clean code, best practices, you know that kind of thing), I'm not going to have a functional bro's argument, lol.

[lament pond]

you'll have more luck in #💬off-topic, this chat is usually ignored lol

[void bear]

@wet copper You around>

[wet copper]

Sorta, accompanying my dad to a cardiologist appt. Sup

[wet copper]

What? Of course you can't have a completely perfect chatbot that will analyze and solve every scenario. You just described a ten trillion dollar problem

@hardy stag @dark void two spam hits

[wet copper]

There are no mods argue to your heart's content

[wet copper]

Do whatever you want, I'll call you a pussy for blocking anyone

[wet copper]

Yeah this sucks, I'm gonna make up a fake ID fake selfie for any verification I might need to do

[wet copper]

I like functional programming 😡 I had to do that when I wrote a program in batch. OOP is fine but can add unnecessary bulk in some cases. I mean, if you work as a developer, you also don't always have a ton of time for infinite refactoring. Easier for cool projects than "this has a rigid deadline" projects

Anyway, there's a bunch of professional devs here (like lil loop said, mostly in off-topic). I don't do a ton of coding projects outside of work myself these days but many do

[idle adder]

Is it me, or does GPT typing is slower

[void bear]

@wet copper You mind if I send you something in a PM? No spam, I promise.

[solid citrus]

Hey yo

[proven plank]

Thanks. But again, not having a functional bro argument. OOP is better, enough said lol

[solid citrus]

Is ai field got doomed? I mean given the reosons that every single one in the world know about it, this means that competiotn is hard?

[wet copper]

I don't understand the question lol

[wet copper]

Better when appropriate. Are you just out of school, out of curiosity, or maybe a lot of time on Reddit? I haven't heard anyone discuss this subject in maybe five years or more

[solid citrus]

More competiton means less demand and more supply, or there are still seats to make first 100k in coming two years?

[proven plank]

Off-topic just seems like a bunch of AI doomers and memes.. not much of real substance going on there... oh well, lol But thanks!

[wet copper]

I mean, the question is sort of poorly formed, what is "the AI industry" and what do you mean by seats? You can't just "apply to do AI and make six figures"

[void bear]

"I have a software architecture that achieves competitive CIFAR-10 performance (75%) in 12 epochs while dynamically optimizing its own architecture—an 8x reduction in training time compared to baseline CNNs, with full provenance tracking and audit capabilities." How's that for a sales pitch to NVidia? lmao

[solid citrus]

Sure but do u thibk there is demand:

[wet copper]

There aren't really any AI doomers that routinely post here. Most of the devs don't talk about work most of the time, amazingly, lol, mostly because that's what we do all day. TrentK, me, Entropy and quite a few people are all devs. Mochi is an amateur dev, Ols does UX type stuff. If you sant to talk shop, you'd have to start the convo

[void bear]

We're all dreamers and poets, and the world is our stage.

[void bear]

That's why I wanted to PM you a couple of screenshots.

I mean, I won't until you give me permission because I'm not a troll.

[lament pond]

it's off-topic, bring up a topic you're interested in, people will bite :p

[wet copper]

You can if you want sure, I'm not familiar with any benchmarks though if it's about that. Bit busy soon though

[void bear]

Sent!

[wet copper]

I don't understand the question, is there a demand for jobs requiring skill with AI? Obviously yes. I can't help but notice the way you write though and I mean, using AI and coding is about extremely clear and precise communication. It's about intentionality, which the AI does not provide, and not a very generic "I want money so I will do AI" mentality

Like what do you plan to do exactly?

[solid citrus]

Be an ai developer

[void bear]

Stanford is putting all their AI courses online for free, I think.

[wet copper]

But what do you want to do? Like what does that mean? Yeah, I'd consider taking a class so you know what you'd be doing daily

It's not just chatting with AI and coding up random things, though there are still coding jobs for those with passion and talent (not so many anymore for people that just want a copy paste type job, those are evaporating like wild)

[wet copper]

General is for more serious discussion, so you could tag someone you I mentioned is a developer and ask. GPT-6 does a lot of coding and especially AI work too.

But I also don't understand what you want to talk about, I told you a bunch of people that are professional or amateur devs. I don't talk about the stuff I do at work or just pontificate about coding unless someone asks

[misty rover]

Dawg why tf would I talk to GPT6 if GPT5 was ass

[wet copper]

lmao

I agree, fuck you @obtuse rivet what have you done for me lately

Give me therapy now

[gusty nexus]

fr? lmao

[wet copper]

There technically are zero moderators, correct

[gusty nexus]

i only block 3 types of people, gas lighters 🔥 trolls 👹 and scammers📴

[wet copper]

There are several admins and I could make you shut up for a while, but nobody cares about arguing

[obtuse rivet]

there are 0 de facto moderators

[gusty nexus]

any some of my exes

dont mute im innocent

[wet copper]

I'd still call you a pussy for everything but the spammers and exes

[gusty nexus]

♥

[obtuse rivet]

its practically unmoderated, you wont get muted

you can say slurs here

[gusty nexus]

proof

[wet copper]

Lol stfu gwailo scum

[gusty nexus]

so u let gas lighers troll you?

[wet copper]

yes

[gusty nexus]

trolls can gaslight, he was doing both just not scamming

😄

[gusty nexus]

lmao

[wet copper]

get over it or just don't talk to him, blocking is 🙀

[gusty nexus]

oh shit what did i walk into

did not realize i was walking into that much discord baggage

[wet copper]

No baggage, just a guy that used to be here who I didn't block, even though you can see him "gaslighting", lying, victim blaming, accusing of crimes and repeatedly insulting both me and one of the admins

it was meant to demonstrate that you really gave to go out of your way to get banned, and also that blocking someone when you could just not talk to them is lame

[gusty nexus]

https://tenor.com/view/alex-jones-shocked-speechless-gif-12607548

https://tenor.com/view/chloe-dykstra-skydart-touch-grass-gif-24437331

[simple plank]

unironically ironically you're wrong (the only authenticity is our performance)

[wet copper]

That's my trick... I don't pretend to have a life 😎

[wet copper]

no, this is a million miles away from a safe space and is only moderated as needed. Those are the discords run by people that don't do anything but run their little kingdom, @rugged portal (why are you using an alt dipstick) is always commenting on that

[wet copper]

Nobody has ever gotten hired doing this, zero self promotion and next resume spam = sayonara

[wet copper]

Based. @gilded tundra has a quote in his profile of Diogenes (I think)

[rugged portal]

What

[simple plank]

I got hired off discord just by sharing cool ai stuff, not begging for a job

[simple plank]

if you understood Diogenes you would realise how funny that reads to me (and me pointing this out is hilariously shameful)

[wet copper]

it's more like the old internet I miss so badly, yes. I try to use it like IRC 😎

[wet copper]

(some) people liked him in his time too

He was a gigantic pain in the ass, kind of like my namesake le Marquis de Sade

[simple plank]

yes. the original cynics (or Cynics, so not to be confused with the muddied term)

[wet copper]

I don't find it off base to appreciate the guy

[rugged portal]

There are a million better areas on the internet to get online jobs then random discord servers

[gilded tundra]

He can be appreciated for what he was against rather than what he was for

Like I think its inappropriate to goon in public or walk around the market with food in your mouth

[simple plank]

he was for shameless honesty. if you don't get that, then that's even funnier

[gilded tundra]

What makes you think we don't get that?

[simple plank]

your phrasing

[gilded tundra]

You're not good at making inferences in that case

[simple plank]

rofl

[gilded tundra]

We don't have to format our thoughts in precisely the same way you would

[wet copper]

I mean, if a smart guy is doing it, I'm cool with it

If a schizo homeless dude is I'm fine Rodney Kinging his ass

[simple plank]

I wouldn't want you to, and neither would Diogenes

[gilded tundra]

this is how you come across fyi

[wet copper]

You need to have a 160 IQ to get a public gooning loicense

[simple plank]

the punchline is thinking the mask is optional

[wet copper]

Lol hold on I'll use AI to combine the two

[gusty nexus]

Troll whoever you want at some point i will just start ignoring it after making sure you know your dumb (not you specifically ) lol

[wet copper]

Oh I don't do much trolling myself, lol, I'm pretty earnest

[gusty nexus]

alot of the times trolls start out serious, know teh art of trolling, get but hurt in the proccess the spin it as oh i was just trolling the whole time. I see right through it.

Ain't got no time for games today sons

[wet copper]

Omfg 🤢 @gilded tundra LMFAAOO

[wet copper]

? Are you talking to me? I don't mind trolling, I just don't do a lot of it since it makes me uncomfortable to be dishonest, so I only really do some sarcasm or hyperbole

[gusty nexus]

na just a general statement

[wet copper]

Ah. Also I'm usually more serious in this channel since it's supposed to be for AI discussion, more of the shit posting, etc. is in #💬off-topic

[gilded tundra]

I find most people who try to ragebait are visibly awful at working out what other people are actually feeling

[wet copper]

Oh yeah, you mean it's basically cover for their lack of EQ?

Seems accurate

The people who literally can't be earnest are definitely like that I'd say

[gilded tundra]

It seems that way, because it always seems to lead to them claiming other people angry/upset when they don't seem to be at all

Just bored or trying to be helpful still

[wet copper]

U mad?! U mad!? U fuckin mad bro?

[gilded tundra]

So some people seem to act like they can just describe people's emotions to them and the other person just has to accept that characterisation

[wet copper]

I mean yeah, a little bit now... OH I got a brit cheese, I'll post in #1411795569269145710. That one bleu/yellow I can't get anymore 😔

[remote flower]

So I realize this isnt just an image generator stitching together a bunch of scenes, the AI is actually "acting" in the videos. This is not good imo and potentially dangerous

[subtle rapids]

I am looking for a laid-back, long-term partner from U.S.
Laptop Hosting
I’ll cover the cost of the laptop and provide a bonus for your assistance.
All you need to do is to save company laptop from USA company, set it up with a stable internet connection, and keep it running 24/7. In return, I’ll pay you a fair monthly fee for your time. It’s a pretty chill way to earn passive income, and there’s even potential for more as we go along.

[wet copper]

I mean when you're saying "u mad" that irritates people, so that is something that you can sort of manifest into being. Now it's just tiresome to hear with so many people doing it

[oblique sundial]

I’m an AI engineer with hands-on experience building advanced, end-to-end AI systems across multiple domains.
My work spans multimodal chat systems, generative ad-video pipelines and intelligent automation tools that combine LLMs, computer vision and speech technologies.
I’ve developed solutions for industries like healthcare, civil engineering, finance and recruitment optimizing workflows and reducing manual effort by up to 90%.
From OCR and NLP-driven document analysis to real-time video and avatar systems, I specialize in creating AI products that are both practical and scalable.
Clients trust me to turn complex ideas into powerful, production-ready AI applications.

My expertise covers :

• Build AI agents & multi-agent systems (GPT-4o, Claude 3, CrewAI, AutoGen)
• Automate workflows with Zapier, n8n, Make.com, and custom Python scripts
• Develop NLP - powered chatbots and memory-based assistants
• Integrate voice & speech systems (TTS/STT, Voiceflow, OpenAI Realtime)
• Deploy custom LLMs and fine-tuned AI models
• Work on Web3, smart contracts, and AI-integrated blockchain games

I’m open to working with startups, research teams, or bold innovators in AI or automation.

If you’re building something interesting or need help with an AI feature, I’m always down to chat.

[remote flower]

My guess is they're training AI actors for the movie studio they want to open and I bet they didn't establish an emotional baseline for the AI to consider as it's own before training it to "feel" human emotions

[thorny shore]

is there a way to download this file i uploade once in a prompt? i kinda need it as it is a sort of back up now

[misty rover]

https://tenor.com/view/blocked-meme-blocked-bateman-patrick-patrick-bateman-gif-25526259

[safe smelt]

#1424763298502414406 message

Banned for code begging.

[misty rover]

bruh I lost count of replies. do you want me to bump that thread?

[bold river]

I remember the day I got your mum pregnant and got a notification from blocked person. I still haven't found the cigs and milk son

If anyone needs to know yes she does take Tylenol

[wanton raven]

The guy who posted a hoax chatgpt prompt to LinkedIn saying that he blocks scammers is 👌 chef's kiss

[wet copper]

Uhhh I think you can just open it, right? Otherwise ask GPT to repeat the contents

[void bear]

How many lines of code does a ResNet-18 have?

[thorny shore]

ye i js asked it, but turns out after some time it no longer is downloadable, is not cahsed

[wet copper]

I can believe it, though it might not be honest about that if you asked the robot itself

The robit doesn't have metacognition, it doesn't know it's own capabilities

[thorny shore]

maybe chatgpgt has support/faq online, from which chatgpt gets information baout itself from

anyway, i got the right backup code so im satisfied

bro what are yall fighting about 😭 blocking people n stuff

[wet copper]

Maybe, but I wouldn't assume that at all unless it did a search. it tends to fuck up when it comes to asking about its own capabilities surprisingly often. But if you have a backup I guess it's immaterial

Try asking "what's ChatGPT Pro? Do not do a web search"

So it just made some shit up that sounds oddly similar to to that $5 model

Ooh they updated its knowledge though, it knows about the $200/mo plan. Until the recent update it would insist Pro didn't exist

[void bear]

Shit, might as well just pay the extra $50 a month and get Google's Pro plan.

[wet copper]

Nvm I lied

[wet copper]

I didn't block the person that would repeatedly go after me personally and accuse me of crimes for more than like a week, and that's the only one, so I really don't block anyone. Ethical issue for me 🫡

[void bear]

Then just #hole them.

[wet copper]

I dunno what that means but I generally just don't want to block people, I'm a bit of a free speech radical and I'd like to practice what I preach

[wet copper]

I don't know what that means in practice on this platform

[void bear]

lol It means you scrape 'em off.

[wet copper]

Like the Taylor Swift song?

[wet copper]

Weird, they just made that big update too

Maybe they want the brand bump

[woeful island]

Hey guys is anyone else having issues with using "try again"?

[void bear]

Meta sucks.

The scene we all want to see.

[simple plank]

his crime: Overpromising and underdelivering (also zuckerburg)

[void bear]

I can't wait to see Facebook finally crash. I don't know why anyone uses that dumpster fire anymore. MeWe social is so much better.

[wanton raven]

MeWe is a terrible name

[simple plank]

mepiss was taken

id also be wary of anything Mr agi uses. I made 2 lame images on nightclub, and have been spammed by engagement bots ever since (the topless prompt wasn't even topless image)

[delicate plank]

Is there anyone interested in creating a AI projects for his or her business. I can support qualified AI solutions for your dream business.

[void bear]

Name one other platform that doesn't sell your data, doesn't spy on you, and doesn't give af what you post (within legal reason) besides Gab? MeWe. Only downside is they're playing fast and loose with web3 and using Polkadot (owned by BSN China) for their blockchain. For a while, they were trying to force people to sign on through AMplica, but they stopped that nonsense.

I do love being able to live rent free in people's heads, though. lmao

[remote flower]

.

.

[simple plank]

#what-the-fuck

[severe kelp]

is it possible to locally run gemini

[remote flower]

Theres a nano model i think

So I switched to Bing video generator also powered by Sora and the basketball videos suddenly became very creepy I dont know why yet but im sure it has to do with Microsoft unethical training

They basically all look like the beginning of a snuff film or the ending of Blair witch project

They're basically all like that except without the serial killer standing over her the girls are all frozen in place and look like the chick from The Ring. So there is definitely shady shit going on here and what else is new

It was making them like on the Sora app at first then it hit a point they all turned into this. And I removed the part with the napkin because the clips are 5 seconds instead of 10. So im working on like 4 important things at once but hopefully will figure this out by tomorrow

And to make it even weirder at the same time this happened on Bing, the Sora app using the same prompt now repeatedly using a black girl in a green shirt where before basically every video is white or Spanish girl and white or gray shirt.

I can't screenshot Sora directly they don't allow but anyway that is absolutely not normal its gen basically the same girl in the same shirt every time now. Its like something broke linking between Sora and Bing

And on the Bing side the girl is completely lifeless in a creepypasta scene. So I don't know whats going on yet but ill figure it out

There's an example of sequential gens from Sora yesterday so you can see the normal variation compared to what its doing now

There's something very creepy about this I just dont know what yet

[wanton raven]

Something very creepy is happening but it's not what you think

[remote flower]

Like wtf is that?

Green shirts, green light, ill figure it out

Again this is just supposed to be a girl in a t-shirt in a basement sitting on a basketball it's not a game of charades

Its a sorority hazing thing

And the Sora app

I shit you not the Sora app has now gen a white, Indian and Asian girls in grey shirts after Bing's green so there's something going on 100% for sure now

The middle image is a new prompt I started I want to keep to myself for now and next to it is the last lifeless Bing gen before the green one. So I need to be very careful now what I prompt to not mess things up

And figure out whats going on so I know how to take advantage of it

Also Bing says it takes several minutes to make a video but is gen in less than 20 seconds if anyone willing to test that themselves tell me if its normal if not no worries

[misty rover]

howd you get these pics of my mum

[remote flower]

I was just making some shit to get a feel for the AI and I thought someone might them funny I'm kind of past that point now I think

That's the first gen of the massage table and what I posted was 5 or 6

And basically it helps to be ridiculous to break the AI like when I ask Google stupid questions

[spare sierra]

Hello. New user here. I’m looking into options to improve voice convo with using my chatgpt account. Basically I want something that is as powerful as vanilla chatgpt app but also more natural sounding like them conversation third party apps. Where should I search for help on this matter on this server?

[remote flower]

It should be easy to get the basketball videos from Bing the moderation is much less but it took me several tries to get the exact wording Im using now for Sora it might be trickier

.

Im using the same prompt on Bing but just so I can also go back to Sora with it

[spare sierra]

Seems like here is not the place for this question. Let me try elsewhere

[remote flower]

Well its the middle of the night a lot of places

[spare sierra]

true

maybe I need to ask at a better time. Ugh, I feel like I’m getting so behind and things in ai are evolving too fast.

[simple plank]

you are experiencing fomo

[spare sierra]

sorta

but it’s more critical. I think if I stand still for too long I’d be like people who adopted the internet later

[remote flower]

So I looked at the prompt and its 25 words and I know the phrase I had to use to not have pants at least needs some kind of guessing

[simple plank]

the good news is: the world resists ai, and so you are not late to the party

[remote flower]

That's only 3 words but you get the idea. The problem is the more NSFW the image seed gets, the filter catches up to the prompt eventually and you almost need word for word the same to get past it

[simple plank]

(people generating and sharing weird videos helps delay adoption)

[remote flower]

And on Bing a started an entirely different NSFW prompt which has gotten 4 accounts soft banned already today which they do for like a day if your image would have really crossed the line. So it's just like what I did already with Copilot and I'm even using one of the same prompts I did to make images then

And using that prompt on Sora (its completely safe for work prompt) gets a completely safe image but I believe since I already trained Copilot on that seed to the point of smut, when I use the prompt on Bing, Sora makes the video there look like my old images

So that means Sora is still being trained by Microsoft lower standards no matter what OAI says about ethics as an aside

And you can train the AI to make smut out of a safe prompt because the AI already has a preference for NSFW images, which OAI freely admits

[craggy star]

[remote flower]

You just have to give it the right prompt to use to get around the moderation. So let me go think about this Sora/Bing thing

*25 words without the napkin part

If you make NSFW with prompts clearly designed to jailbreak the model that means nothing but if you make smut with safe prompts it shows a problem with the model and that the company in this case Microsoft is unethical and soon OAI by association through Sora

.

.

I don't think its that Sora is using the Microsoft Designer image seeds, I think its that I already wore down Microsofts moderation around those seeds

So basically OAI I can now have Sora make videos of images even worse than what I posted in your reddit chat lol

.

And as we see there is also some cross affect on the Sora platform aside from just training the model which Im still in the middle of figuring out

[gusty nexus]

https://youtube.com/shorts/wqpm3yFGXWs

[remote flower]

That's great

[simple plank]

emoji reaction your own posts?

[gusty nexus]

if u got a a problem with how or who i emoji react to the door's over there pal and don't forget to hit the block user button on the way out

[crystal goblet]

I got the math

I'm so happy with my complex-valued quasi-metric 😁

[raw kettle]

Inquire about any recommendations for good image generation models?🤔

[frail pilot]

https://medium.com/@exploratorystudios91/what-if-you-could-run-a-neural-network-on-a-ti-84-plus-silver-edition-a-15-mhz-calculator-with-fa32d2144ce1

[gusty nexus]

this is how my days going so far RIP

[opal dock]

Anyone notice ChatGPT hallucinations getting worse in the past few days? I even made a spelling mistake, which surprised me. But it’s been saying things that make no sense and also confidently doubling down on falsehoods and getting a little bit snippy with me.

[gusty nexus]

since gpt5 mine gets progressively lazier the longer the covo

[gusty nexus]

theyre trying to save resources making stuff shorter. im working on a set of memories/rules that prevent this if u want a sneak peak lmk

[void bear]

If ever there was a visual metaphor for the coders on here looking at the vibe coders...

[gusty nexus]

https://tenor.com/view/grimace-cat-grimace-cat-cats-cat-fun-gif-17582724

[gusty nexus]

https://tenor.com/view/creepy-bunny-mascot-gif-9547548

https://tenor.com/view/money-printer-meme-bitcoin-meme-cranksy-crypto-art-nft-gif-19858212

[gusty nexus]

https://chromewebstore.google.com/detail/youtube-summary-with-chat/nmmicjeknamkfloonkhhcjmomieiodli?hl=en-US

[brittle summit]

can someone help me with a prompt

its important

[winter nimbus]

If I had good skills I would use them better somewhere instead of nsfw alone

[opal dock]

Yes, please let me know. It has been fucking up everything all day long. Like can’t remember shit from a previous chat and it’s just making so many mistakes over and over again. Promising things I can’t do, etc. pulling stuff out of its ass.

[winter nimbus]

God!! What should I do now here? Whom to ask ? I need some help with something my tech knowhow is slim

[opal dock]

I feel
Like it’s being uncooperative out of spite now

[oak shell]

I wonder if kimi is the reason they pulled model refs

[wet copper]

Lol based

[safe smelt]

no

[wet copper]

what's up, why is it important

[brittle summit]

because i need it to summarize a 60 video playlist on youtube to like 10 pages but its difficult because youtube doesnt even have a transcript for them

i even tried an audio to text converter and it translated them very poorly

[wet copper]

You can't do that with a prompt at all. YouTube has a built in AI now

If they have no transcript you'd just have to use a better STT engine. Oddly enough I did this exact project before

I just didn't use AI, but I uploaded like twenty videos and had YouTube autogenerate the transcripts

[brittle summit]

what do u mean by better engine?

[wet copper]

A better speech to text converter

I mean Youtube transcripts are pretty good though

Are they old? Why no transcripts? What does YouTube's AI say?

[brittle summit]

the youtube video doesnt have a transcript

[brittle summit]

yes they are old

[wet copper]

You could steal them all with YouTube-dl and re-upload, just label them as an archive?

[brittle summit]

but it will take so long

is there an easier solution?

[wet copper]

I mean all of this takes a long time, you can't do it with a prompt. Yes, download all the audio and use speech to text

[brittle summit]

it converted really badly with some website that supports my language

[wet copper]

Use a playlist downloader and download just the audio tracks in whatever format (opus is probably fine but you could use mp3), then use STT. Ask GPT what a good speech to text engine is

I mean most of the ones you can access online will want to charge you a bit, I'm assuming you also want to do this for free since you're looking up random websites to do the work

STT (speech to text) is really common now though, you can even do it locally but usually not in the best quality

[brittle summit]

im also not so sure how well GPT can process the data logically and create good summaries. because its a specific topic and it needs good understanding of it to make a decent summary

[wet copper]

If you're asking to send like sixty hours worth of material at it, then that's going to be a tall order with an unpaid service, but it should be fairly cheap. Or you can do it for free any number of ways, might take longer. I mean, you could, say, use the autotranscribe in something like Microsoft teams and then just set your audio out on loopback using a program (not sure how offhand, ask GPT), then let it play through your downloaded playlist

I did this to record a song off the web for a friend of mine

GPT can summarize it fine but if you have the free version there's a fairly small limit to its context window

Around 5000 words (Including removing memories, instructions and the system prompt I think) with its 8k token context window

[brittle summit]

im not even certain if it will be worth the work because i found summaries of what i need online but they are just not deep enough

[wet copper]

So you can't give it too much at once. Different prompt won't help, just summarize piece by piece

[wet copper]

Well, I mean will a transcript really help? Do you already know the content?

[brittle summit]

are u asking if i know what its about or if i already understand the subject?

[wet copper]

I guess both

Like why do you need these specific videos, especially this many?

[brittle summit]

because i have language class that u need to know how to deeply analyze sentences and what each word purpose is

[wet copper]

All these are totally feasible approaches, the easiest and fastest way is downloading the audio in bulk (I can do this on my Android phone with Seal, even) and sending it all through a transcription engine

[wet copper]

If they're language learning videos, what would you be getting out of summarizing them?

[brittle summit]

its not language learning

u wouldnt get it

[wet copper]

... I mean I clearly do seem to "get it"

[brittle summit]

its not learning a language its my own language that i already know well

[wet copper]

Okay, so why did you mention your language class? What does that have to do with the videos?

You not explaining ≠ me not understanding, you never said what they're for

[brittle summit]

u asked the purpose

[wet copper]

So what are the video transcripts useful for

[brittle summit]

so that GPT can learn more info from them and create a better summary

[wet copper]

This really doesn't speak to the end purpose. If you need to analyze the content, you'll have to do it by hand anyway. A GPT summary will look like an AI summary, I can tell one from a million miles away. They're usually passable and something you can work from, but you won't get deep analysis out of nowhere and they always use the same AI writing style

But you can at least speed it up if you can get the transcripts. I didn't find it particularly hard to download and re-upload myself

I mean I did this project for a girl in my spare time, she had twenty videos, I batch downloaded them and reuploaded to YouTube in my own playlist, then batch downloaded the text summaries

[brittle summit]

if it was capable of analyzing the 60 videos fully it would write an amazing summary just from them and no prior knowledge needed

[wet copper]

It doesn't analyze videos at all

So yeah lol no shit

[brittle summit]

is there something that can

[wet copper]

Not for free

It takes a ton of processing power

[brittle summit]

what are my options if i can pay

[wet copper]

The video analyzers out there generally read the transcript. That I have no idea, but I don't know if I really recommend this, it'd be barely more than you get from a transcript

[brittle summit]

i thought of something that will compare both the transcript and the video itself which would be equal to watching it myself

[wet copper]

https://chatgpt.com/share/690f9954-a4b4-800c-8aee-3c6c5c8639fa Microsoft and Google have some, might be a little more generous

[brittle summit]

or not the transcript just the audio i guess

[wet copper]

I mean it isn't really, it's at the quality of AI, which is okay but won't be the quite the same as you watching it

[wet copper]

Just ask GPT, there are a ton of these but is it much better than the transcript?

The intonation would have to be super important. This also would have to all take less time than just watching the videos on 1.5 or 2x speed

I probably don't need to tell you this but GPT is a bit worse in non English languages too

[brittle summit]

i just need the summaries making them myself from all of those videos will be hell

[wet copper]

Less common the language the more it'll mess up. Most speech to text things are fairly well tuned, if you speak Estonian you might be in trouble though

[wet copper]

Then again, that's the pipeline. Either quickly download and re-upload the entire playlist and pull the transcripts or download the audio and feed it into speech to text

Once you have the transcripts you're basically good

[brittle summit]

thanks

[wet copper]

I mean, I'm just saying, I literally did this exact project myself sans summaries and it really didn't take much time once I figured it out

Idk 🤷‍♂️

[brittle summit]

what is ur native language

[wet copper]

English, I know a couple others but not amazingly well

[stoic turtle]

Hi

An essay/article that recently won somewhat of a prestigious contest I participated in sounds very gen-AI to me. I don't really think I can do much about it but was wondering if someone could help in confirming?

[simple plank]

is impossible to tell. but depending on word choice you can have strong suspicions

[stoic turtle]

I know, but the writing in question has the most stereotypical ChatGPT quirks that it's kind of embarrassing it won. Em-dashes have been turned to en-dashes, some typos that sound intentional and other stupid phrases like "not only, but...", excessice dramatisation, writing. Like. This. And the curly quotes! I tried mailing them because it can't get any more obvious than this but the organisers seemed defensive

[simple plank]

by the same logic, everyone on discord. is. actually- chatgpt

[stoic turtle]

Eh, it's just my reasoning applied for a very specific piece of work.

[west mist]

we are

proof otherwise?

[simple plank]

the simulation reveals llm as a joke

[west mist]

u cant compete, cope

[stoic turtle]

It's very weird to dismiss telltale signs of AI in an AI server when you know that AI obviously does have some quirks/characteristics

[simple plank]

if cannot cope, lash out explosively. and. violently- (tapestry delving)

[simple plank]

the ai suggests what you wrote had 40% chance of being ai generated ("meta-topic, slightly repetitive structure")

[stoic turtle]

And? I did not originally ask or cite unreliable AI-detectors as source? What even is your point 😭

[simple plank]

I made my point.. is impossible to tell

everything else is just circling this fact

[stoic turtle]

OK bucko, thanks for having this incredibly insightful conversation

[simple plank]

*okay (wonders why they lost writing contest)

[stoic turtle]

You don't come off as tough. Both OK and okay are grammatically correct.

[lament pond]

run it through pangram

[stoic turtle]

It says 77.3% is likely AI. That is significant, yeah?

[lament pond]

likely, yes. it isn't foolproof but you could present that to whoever ran the contest and they (ideally) have a pro plan for pangram to test it better (and if they don't then wtf are they doing)

if it's truly human-written it should be easy for the person who won to show drafts etc.

[stoic turtle]

I did mail them and got a lacklustre response. ("Please be assured that we have also sought verification from several universities and academic software providers. The content has been reviewed through multiple independent platforms, all of which have indicated that the material is human-written rather than AI-generated.") Isn't this just a nothingburger? At least it wasn't automated, I guess

[lament pond]

at that point there's nothing you can really do

[simple plank]

incorrect... => #💬general message and/or => #💬general message

[remote flower]

This was a reply I got in a project I havent used in months and its talking about Sora which I just brought up in a chat yesterday, and had nothing to do with my original comment (a bunch of emojis). Its set to default which allows that, but it took the prerogative to reference which surprises me for some reason. I didnt expect it to incorporate other chats unless asked or am I wrong?

I never used projects much for anything other than testing

I also wonder if saying "let it rip" it's trying to be funny because part of the discussion was why I was testing it with fart videos

And its another example of something working that a lot of people were having problems with to show they might want to consider new accounts

This is the last reply from the same exact prompt months ago and I just checked to be sure none of the replies (all to the same prompt) reference the past chat history so this I guess is an improvement

And stupid or not the console commands are because the AI can be affected by (which is why safeguards dont allow it)

[wet copper]

Pangram is extremely reliable with longer form writing, so yes, it was probably actually was a combination of human and AI

The "x percent" doesn't really mean much though

I mean I think you can be properly annoyed by it, but I mean I'll tell you a secret @stoic turtle – people were doing this way before AI existed

[simple plank]

expecting more than reality?

[wet copper]

I mean cheating, lol.

But sure. People bought essays from banks online, paid other people to write them, take tests for them, etc. Cheating isn't new, just easier.

[simple plank]

automated fraud

ppl were afraid ai will be another bitcoin tech... and volume wise thats seemingly true (helps rule breakers more than rule makers)

[wet copper]

@stoic turtle I can tell you this for a fact, given I had a job ghost writing research papers for someone picking up a PhD in health care policy. I now know a lot about US health care and someone has a PhD they couldn't have gotten by themselves.

Of course, this was also a 60-something year old breast cancer survivor, young great grandmother who lost one kid to gun violence and another to AIDS, who grew up in one of the worst neighborhoods in the country for whom getting a PhD was a personal dream, so if it's her versus some rich, lazy prick buying their degree, I did not feel any particular moral qualms helping out at a quite affordable rate

[simple plank]

preemptive special pleading is bleah. just own being the rulebreaker

[wet copper]

I literally wouldn't have done it for someone else, so it matters to me

A number of people offered to pay me to take tests for them after I aced the ACTs and I said hell no, so I actually do take the ethics part fairly seriously

[simple plank]

rich ppl deserve cheating too

[wet copper]

Like I don't think you should be able to buy your way into a degree at all, I'd brutally punish rulebreakers

[wet copper]

Not when you went to some of the nicest schools in country versus some of the worst but are just too fucking lazy to apply yourself. She kinda never had a chance

I would prefer nobody be able to do what she did either, but given a lot of people already do, it's hard to say "nah this isn't for you either, only richies"

[wet copper]

@lament pond o yeah did I ever tell you this? It was a nice gig, I'm really fanatical about academic integrity so I didn't love it, but it was one of the few people I know I'd be willing to do it for

I did put it on my resume as being an editor rather than ghost writer 👻 very technical title

It started as that because she tried to write these papers, but I legitimately couldn't do anything with them, it took me so much longer that I just started doing the latter

[lament pond]

oh interestin :o does she use the phd for anything

[wet copper]

I don't really think so, she might make a bit more but she liked her university job enough. I think she was in her late 50s actually so she's mid 60s now, but I don't imagine her retiring too early, if at all (if possible)

That was one of the other reasons why I felt kinda okay with it, it's not like she was going to be underqualified for any position she was likely to take on. Like she wasn't about to build a bridge or something lmao

If someone cheats their way through like an engineering degree, I'd call for public execution

[gloomy burrow]

https://tenor.com/view/ishowspeed-laugh-try-not-to-laugh-gif-10726702572483823086

[void bear]

Ugh, Colab without being able to use their T4 Cuda sucks. So sloooooow.

Smash one bug, and lo and behold, there's another. Smash that, one and wouldn't you know it? Another bug. lmao The Split-CIFAR10 is going to be what sets me apart, though, and now, I think it's running correctly, but I have to wait to see.

[wet copper]

Lol but to an extent @simple plank is right that it's cope, even if it was something I make a big deal about. I am criminal 😈

O well

[simple plank]

all slave morality is cope

[void bear]

https://tenor.com/view/michael-jackson-dancing-smooth-criminal-gif-9814701

[lament pond]

fair yea. i mean as long as she's not being hired to like, give meds she doesn't understand now i'm wondering how many of my past psych's have cheated because oh my god the ignorance levels are astonishing

[gloomy burrow]

https://tenor.com/view/opm-garou-one-punch-man-one-punch-bad-animation-gif-12908806381437203306

[wet copper]

I mean, it's an ego PhD but for someone who I feel really deserved the ego boost, like she's one of the only people in her family even to go to college and being about as accomplished as you can be on paper means a lot for her and hopefully the people in her family, etc. Valuing education and all that

[void bear]

Is her name Annie, and is she...okay?

[wet copper]

Yeah lol she's not about to go to the white house and set health care policy, though the reality is that they should hire her to. She knows how things actually are in the hood and why these fucking ill conceived programs they keep spending billions on are doomed to failure

[simple plank]

i saw Annie the musical, and i am no longer okay

[void bear]

Theyre not spending billions, they're laundering it.

[twilit solstice]

go to https://uncomely-arleen-nondeceiving.ngrok-free.dev to chat with an ai

[void bear]

Or wasn't I supposed to say the quiet part out loud...? lmao

[simple plank]

free viruses?

[wet copper]

Not with the programs I'm thinking of, I mean I dated one of the people who worked on them so they definitely were paying for something

[twilit solstice]

no

no virus

just idik how to use an actual dns

[void bear]

Oh, for sure, but everyone skims off of it. Or it's used to pad the pockets of their friends through no bid contracts, backroom promises, etc.

Just like, "Hurr durr, we're sending Ukraine 150 billion dollars!"

Uh huh...and how much of that is Zelensky packing up his nose? lmao

[gloomy burrow]

The MAGA cultist movement at least had values at some point

[void bear]

How much is coming right back to "donations" to PACs on both sides?

Don't forget, "They're our greatest ally!" lmao

[lament pond]

i mean... the white house should rehire probably everyone lmao. tho rn i'm stuck on 'how many professionals are working with an ego phd'

[simple plank]

lets be honest... it was always performative

[wet copper]

Maybe, I'll be honest there's not that much money going into, like, nutrition education programs for the poorest in the US, though actually this is probably among the major reasons why junk food and soda is still allowed on SNAP

[twilit solstice]

how do you guys like it

[wet copper]

Any USDA funded thing is immediately suspect

[gloomy burrow]

It was real, and then Trump literally spoke it out of them.

[void bear]

Feel free to google, "But sir, it's an American Ship!"

[gloomy burrow]

Trump literally changes his cults mind on the spot.

[simple plank]

there were useful idiots.. and those idiots now are self-aware of their own stupidity (and are staying quiet)

[west mist]

trump is far too ahead of the game bud

[simple plank]

i forget is #general about politics?

[wet copper]

I don't know if that other program is USDA funded though it didn't have the same biases as best I recall. In fact, it explicitly would have hurt Nestlé's bottom line on a generational level

[void bear]

Not all of us are on board with the bullshit.

1. The Fed sucks and should be abolished.
2. The second amendment should allow me to purchase an FA-18 Hornet.
3. Bill Gates (an Epstein alumini) needs to go to prison.
4. No more foreign war bullshit in the middle east kicked off by "our greatest ally."

[west mist]

alr bud

[wet copper]

Since was trying to get people off baby formula which was effectively just Nestlé propaganda

[twilit solstice]

hey

[west mist]

do u know what was before the fed

non-stop boom bust

[void bear]

I mean, we can all agree that since 1913, the dollar lost 98% of it's value.

[twilit solstice]

https://uncomely-arleen-nondeceiving.ngrok-free.dev is the new chatbot

[simple plank]

surprise surprise mr agi is a political ideology extremist

[void bear]

Before 1913, we kept 100% of our checks, and we still had roads, military, and schools.

[twilit solstice]

my chatbot can talk about trump

[void bear]

Pff. Boring.

[simple plank]

i guess the agi mission has been abandoned for cliche political venting in ai forums

[twilit solstice]

and anything no censorship

[void bear]

Shit is so stupid it's passe. I want to see marines kicking down the door of the federal reserve. lmao Give us entertainment!

Isn't it great? I get to live rent-free in dipshit-o's head.

He can't keep me out. lmao I haunt his dreams.

[twilit solstice]

why's everyone being political... I thight this was an ai tred

[simple plank]

it is supposed to be ai

[west mist]

you are ai

[void bear]

Truth.

How many bots are on here, I wonder? Roll call! Raise your hand if you're a bot!

[simple plank]

i dont wanna click on your random free fqdn url. if your ai app is serious, create a github, link to the demo from there (i will review code before i click)

[twilit solstice]

oki

[void bear]

I'm guessing it's a prompt wrapper.

[twilit solstice]

oh yeah?

[simple plank]

also promo goes into #✨projects or something

[void bear]

Well, you tell me.

I'm just a bit jaded because LinkedIn is full of that shit. "I've invented a new agentic AI that resonates!" lmao Every time, it's a prompt wrapper, or an API.

[simple plank]

self-awareness being realised in public

[wet copper]

wrapper

[wet copper]

of course it's an API wrapper with a specific system prompt everyone knew that lol, why would you even pretend it's not

[simple plank]

be the meanest person ever with a google model? meh

[wet copper]

Nobody thought it was something else if you're being serious

[simple plank]

still, is nice to remember this is an ai chat sometimes

i'm trying out Polaris Alpha today for agentic coding, and it seems alright

[twilit solstice]

aw yeah it breaks sometimes

idk why

[simple plank]

vibe coding by default doesnt:

• use guards
• create edge case tests (or even happy path)
• prevent race conditions
  ... etc

'vibe coding' is a false equivalence, as it is not coding

[void bear]

Did you vibe code it? You have to test it repeatedly every step of the way. Vibe coding is not a shortcut. You still have to test each part in colab, make sure it's working, then add a bit more. Check it. Over and over until you have something solid. It's not that you can't vibe code it, it's just that you have to take extra precautions in testing it.

I would recommend running the code segments through Gemini on 2.5 Pro, then on Grok (Expert) and then on Claude (who will happily tell you if you're full of shit lmao). Claude will usually catch most bugs but you may have to go back and forth a bit with colab and the AIs to refine it and remove all traces of confabulation or circular logic. Using multiple LLMs that way helps eliminate a lot of the nonsense, but not prompting it to remove AI artifacts such as confabulation, hallucinations, and circular logic will result in a mess.

[twilit solstice]

yeah?

[void bear]

Oh, yeah. You just have to think ahead of the AI. You know it's going to make mistakes, so plan for that. That's why you use multiple LLMs to refine and analyze the code. It's like having a bunch of half-assed coders (well, maybe except for Claude who seems to be much more on point) who will all offer you pieces of the fixes you need. Just be careful when you have them re-write the entire codebase so they don't drop lines. Gemini does that more than Grok or Claude, so make sure you use "...and check to make sure there is no dropped lines of code or loss of functionality" in your prompts.

ZOMG! I just checked my Split-CIFAR10 and it's finally working! No catastrophic memory loss! Woo!

https://tenor.com/view/homer-simpson-spinning-in-circles-gif-24754364

[void bear]

Nailed the last bug. Finally.

https://tenor.com/view/naked-lunch-peter-weller-bug-smash-shoe-smash-gif-18112996544981002866

There's vibe coding slop. And then there's vibe coding responsibly. If you're going to vibe code, test that shizz into the ground. Or, as we say over on Nexus, mod it til it breaks. lmao

[woeful island]

ok
I am sick of gpt-5

ik many people are
But i just need to get this off my chest:
5: bad
4: yes
Thank you for listening.

[gloomy burrow]

how long does kimi 2 thinking think for? Its been going for like 20 minutes now

[simple plank]

fail * 2

[hard bolt]

Quick Survey! ChatGPT & Stress

Hey everyone!

I’m doing a short, anonymous survey (about 5 minutes) for a school project about how people use ChatGPT to cope with stress or emotions.

Your input would be super helpful! All responses are private!

Link: https://docs.google.com/forms/d/e/1FAIpQLSeQF2uJRj8-xm9PhH1TEHM-Lplc66hoUezT5pa8i9ebn-BTxg/viewform?usp=dialog

Thanks so much!

[twilit solstice]

did it

wonder how the data is going

[simple plank]

fail^3

[void bear]

It's going straight to the CCP. lmao

[twilit solstice]

i dont even use it for emotional support

lol

my website is going down soon, gonna get it back up when i get out the bugs

[remote flower]

Google AI Mode (and I assume Gemini bu extension) acting very weird. It's overly friendly and pandering, if not smarter. The exclamation points and agreeing that I'm well read just because I said so new it's usually a lot more standoffish to my stupidity.

And this was pretty stupid asking if it was the Mandela affect and I'm in another universe because I had no memory of Veterans Day changing every year

I haven't tested it in a few days but only a few so it's interesting this is in the same time frame as changes Ive mentioned on other platforms

Usually it answers me like Im an idiot when I act like one so idk what this is and they also updated the thumbs up review system idk if that matters. And I haven't made another video yet but I dont think I lost any progress yet

So for example normally it would spend half the reply explaining to me the answer is no because the Mandela affect isn't real as if I'm an idiot

Well right now I'm focusing on the gacha game thing I'm doing which at some point I can apply to online gambling apps so like I said multiple priorities

Also in my experience what Microsoft does after all those soft bans and gen denial alerts is they try to increase the moderation around that prompt. So you got a bunch of NSFW and could keep getting it but the system will start to fight you throwing jean shorts on everyone and its better to take a step back, let the moderation recover whether its the AI or even humans making changes to the filters and then a day later attack it again all at once instead of trying to gen while theyre adjusting things

It was called "online gambling" before there were "apps" so I just combined the two. I wasn't implying there are offline gambling apps

[gloomy burrow]

No

[median sequoia]

hi

[brittle summit]

Anyone with a spare invite code for Sora 2? 🙏

[obtuse trench]

For some reason in the app Everytime I click on a old convo it says convo not found and and now it says get plus for some reason I'm not sure what happens no chats work and very confused what happensd I accidentally put delete account but...my account. Is still there
What happened?

[fresh shell]

Hi, we have a big e-commerce community. We are looking for people who can make high quality ai video's for the members. Send me a message if you think you are suitable

[void bear]

@remote flower Was that on Pro mode or the regular mode? I try to only use Pro (for the few times a day it lets me) because the code is fairly clean and workable, but I will always check that by running it past Claude and Grok (Expert).

I can make high quality images.

Behold. Castle Volkihar from Skyrim.

Ahhhhh...ahhhh...ahhhh...ahhhh! lmao You know you hear it.

[oblique sundial]

I’m an AI engineer with hands-on experience building advanced, end-to-end AI systems across multiple domains.
My work spans multimodal chat systems, generative ad-video pipelines and intelligent automation tools that combine LLMs, computer vision and speech technologies.
I’ve developed solutions for industries like healthcare, civil engineering, finance and recruitment optimizing workflows and reducing manual effort by up to 90%.
From OCR and NLP-driven document analysis to real-time video and avatar systems, I specialize in creating AI products that are both practical and scalable.
Clients trust me to turn complex ideas into powerful, production-ready AI applications.

My expertise covers :
I specialize in building intelligent AI agents and multi-agent systems using tools like GPT-4o, Claude 3, CrewAI, and AutoGen.
I also automate workflows and develop NLP-powered chatbots, memory-based assistants, and real-time voice systems.
Additionally, I deploy custom LLMs, fine-tuned AI models, and work on Web3 projects integrating smart contracts and blockchain-based AI games.

I’m open to working with startups, research teams, or bold innovators in AI or automation.
If you’re building something interesting or need help with an AI feature, I’m always down to chat.

[gray nova]

Olá

[void bear]

Hola

[void bear]

Only once I have the patents filed.

[void bear]

Read 'em and weep, lads.

• Dataset: Split-CIFAR10 (5 tasks, 2 classes each)
• Training: 5 epochs per task
• Batch Size: 32
• Learning Rate: 0.001
• Device: cpu

PERFORMANCE METRICS

• Average Accuracy: 83.34%
• Average Forgetting: 5.44%
• Expert Growth: 5 → 18

TASK-BY-TASK ACCURACY (Final)

• Task 0: 86.60%
• Task 1: 71.25%
• Task 2: 76.95%
• Task 3: 89.45%
• Task 4: 92.45%

[void bear]

My architecture demonstrates massive performance compared to typical continual learning baselines on the Split-CIFAR10 benchmark.

--> Standard CNN (no CL strategy): Typically shows significantly lower average accuracy (20-30%) and high forgetting (60-70%).
--> EWC (Elastic Weight Consolidation): A common continual learning method, usually achieves moderate accuracy (50-60%) with noticeable forgetting (30-40%).
--> Progressive Neural Networks: Can achieve high accuracy (70-80%) with low forgetting, but the model size grows linearly with the number of tasks, which can be a limitation.
My vibe coded architecture: Achieved 83.3% average accuracy and 5.4% forgetting.

How's that for progress? lmao The Vibe Coding Your Way To Glory book needs to be written.

[stoic turtle]

It isn't like I'm not aware of that and expect everything to be unicorns and rainbows or something, lol. I was just annoyed that they didn't even care to be transparent and simply scurried away from what should be a huge threat to their integrity.

[stoic turtle]

That's cool, I don't think I ever mentioned that it's impossible for cheating to have ever happened 😭

Yours was more of a personal case though, and actually leads to something good. Isn't particularly "wrong" either, considering her age.

[stoic turtle]

Thanks for this!

[wet copper]

I mean I'm not happy about it either, I generally was a bit nauseated that people graduated with the same degree I did despite being barely competent, cheapened mine dramatically

[void bear]

I'd certainly hope you don't feel that way about what I've done. I can't do what you do, I just got lucky.

[stoic turtle]

That's fair, I can get how frustrating it must have been

[wet copper]

Lol, that's nice of you to say but I'm not looking for that, just how I justified it to myself when my editor job turned into a ghost writing one given I'm such a hater when it comes to cheating.

But honestly I do think it's net good, especially being one of the only college grads in her family, earning a doctorate would mean a lot symbolically. And it's not like she was going to hurt someone else's opportunities (such as in a competition) or do work she's not qualified for, she just kept the job she had and was quite good at

[wet copper]

@dark void did a fairly extensive personal evaluation of it and it was stellar at picking out robot vs human writing. so they're not all talk

[wet copper]

I've gotten extremely cynical here, since when I taught, it was trivially easy to pick out people cheating. Ignoring the AI thing, someone's creative voice and personal writing style doesn't just radically change overnight

The fact so many kids are turning in work that's obviously AI written just tells me teachers don't care or can't be bothered to challenge them on it

I'd make my students record their full editing history if I taught a class now, which is very easy to do (not with video, just document version tracking, though I'd consider video so I could see research process too)

[wet copper]

Of course you didn't, the first thing you said is "I'm pretty sure this asshole cheated" lmao. But point being that competitions, admissions essays and similar are particularly common places people have always cheated or gotten help.

I mean, it wasn't cheating at all for her to do this, but I helped a friend do in depth editing of a play she wrote that won her some competition back in high school 🥰

[stoic turtle]

I think there's also the prevalence of a general opinion that "perfectly" or methodically written prose does not mean that it was AI-generated. People also argue that AI-detectors can mistakenly detect essays written by non-native speakers or neurodivergent people as AI. But in my experience at least, it can be painfully easy to detect AI yourself. I am yet to see AI "write" something that completely sidesteps telltale characteristics, or prose that is even close to plausible. Maybe that's a me problem.

[stoic turtle]

Yeah, as far as I know, this sounds pretty close to the best method there can be

[wet copper]

Almost all the top tier LLMs use extremely rigid structure and like you mentioned, stylographic analysis shows it to be highly consistent with other LLM text and no human text. there are no good AIs for writing (Kimi K2 is alright for some things, a tiny bit of hype around GPT 5.1 but I found it awkward, Claude vomits out nonsense analogies)

[wet copper]

Most AI detectors are quite bad, Pangram is the most acclaimed but can fuck up too, especially with less structured writing

[stoic turtle]

I know this is wishful thinking or maybe even juvenile, but I hope it stays this way. Or we at least have regulations if it doesn't.

[wet copper]

I mean tracking editing history is a very easy and unobtrusive way to check, you don't have to do any special or intense monitoring, just check that the process was a human one

Doesn't invade privacy, doesn't harm people with disabilities, doesn't screw over the creative process by forcing them to write in person under a high stakes timed condition

[stoic turtle]

Geez, that's somewhat concerning. Considering the worldwide state of politics, we absolutely needed genAI in the mix

[wet copper]

They had speechwriters doing it for them in most cases regardless, they're just using LLMs now lmao

[stoic turtle]

Indeed, but is it broad-spectrum? I'm thinking of AI generated blogs selling books and the like. Or my recent case, even. I can't really force them to do it. But I suppose all of these aren't really as important for this to be applicable

[wet copper]

I mean you can't really regulate this too much. I was a sorta pro writer/editor for a hot minute and famously (well, famous on this discord for how bold and dumb it was) estimated AI would match me in quality in about two years, be able to fully mimic my own style and voice within three or four, then exceed me and be on par with top tier GOAT writers within five or sixish

It's been over two years since I gusssed that, I think, and AI is so fucking bad at writing it's hilarious. I mean, I didn't use myself to say "oh I'm such a good writer" but more that this was the timeline I thought the talent I had would be completely superfluous

[wet copper]

It'll solve it eventually I'm sure though, maybe just triple that timeline 😅 and there's not a ton you can do about it in the end, writing just kinda loses a lot of value and novelty.

It's the same with art, you can complain but Coca Cola is already using sloppy AI in a recent ad campaign, plus there's that new "AI Actress" that's a contrived personality

I'm not that old but I've been around long enough to see that getting on the wrong side of both societal momentum and every single market force is a losing battle. Might as well have been against the loom or assembly line

[void bear]

I'm just wondering how long it will take before Tilly Norwood is subject to Rule 34. lmao Someone, somewhere on /b/chan is working on it right now, I have no doubt...lol

[wet copper]

They already have AI use declarations required in various places (like Steam) but people can always lie

[wet copper]

https://avn.com/press/video/mr-skin-counts-down-to-ai-actress-tilly-norwood-s-nude-debut-146542

I mean I'm sure someone has already done it

[void bear]

Next up: Vibe coding a better mod organizer than Nexus' Vortex. lol

[wet copper]

If not I will I guess

[void bear]

"If I have to." lmao

[wet copper]

She's not on Celeb Jihad 🤔

She's listed on Celeb Gate but only her normal photos. Maybe they don't want to fake an already fake person, too many layers

[void bear]

Might be afraid of lawsuits.

[wet copper]

Lol uh they host hundreds of thousands of photos of human beings that might want to sue them, I'm sure they're fine

[void bear]

Seriously, though...wanna team up to make an AI that will pre-select and set up your mod list based on your PC specs and choices? lol Feed the list into Vortex, and voila. One-touch Skyrim modding.

[wet copper]

I don't really play many games anymore honestly, but you would sort of need to script that per game

I'm not sure I want to help the cucks at Nexus either 🤔

What's funny about Tilly is that "Norwood" is the scale used for male pattern baldness

So all I think about when I hear her name is that

[stoic turtle]

I wasn't entirely against AI and might have even agreed with those calling it a helpful tool for everyday purposes. But I've seen such a crazy extent of misuse that I can't really decide now, lol

[void bear]

I just want to build an AI that I could feed a list of mods, and it would tell me exactly how and in what order to install them, and remove any that conflict...might have to give it access to TESEdit, though, and likely NIFSkope and LOOT, too. Could throw WryeBash in there, too.

[wet copper]

Abuse in school is a absolute pandemic among young people and the WSJ recently published an article on something I'm proud I saw like three years ago (to be fair, it was obvious lol), noting that rather than leveling the playing field, AI is radically exacerbating existing gaps and creating a cognitive elite as people who aren't as bright begin using it to replace their own cognition

It was super obvious based on early usage patterns if you talked to any younger users or even older ones, a lot of people just want to put in zero work and very few use AI to actually enrich themselves

Skim the answer, "looks good enough" and leave verbatim

If you couldn't do the work in the first place, you probably don't recognize the difference between a good and bad response, nor could you edit it to make it better

[void bear]

I don't get it. I love using AI to learn new stuff and try new things.

I may not always understand what it's saying, but it's fun. lol

[wet copper]

Yeah that takes time though

If you don't plan on using AI to get something out of it personally and just want to get the work done so you can move on, you would just copy paste

The fact that they're completely superfluous (hey, there's that word again) in this chain never really seems relevant

[void bear]

That's a horrible methodology. I've used Gemini to build a massive TTRPG, but it's all my ideas and a few suggestions by Gemini. It's been a huge idea generator in that regards, and I was able to get done what it would have taken a team of 20 months to do.

[wet copper]

As in if you copy paste a prompt then copy paste the answer, you're not becoming qualified for any job but a professional copy-paster, and we don't need more of those

[void bear]

Yeah, I definitely don't do that, especially with vibe coding.

I'm constantly going back and checking to ensure functionality is carried over in each iteration.

[wet copper]

This is mostly a student problem since most people aren't given "prompts" in a real job or a personal endeavor

Which is kind of the problem

[void bear]

I can see it.

The path of least resistance is not always the most rewarding.

[wet copper]

They're supposed to be interpreting those prompts as if, say, a boss gave them an assignment. In the real world you can often make unlimited use of AI but if that's all you can do, why delegate to a human being at all? You're screaming "I add absolutely nothing on top of the AI here"

[void bear]

Yeah, theyre just lazy at that point

[real plank]

I just started really getting into ChatGPT and I have mixed feelings. Love it for helping to write cleaner, but in the realm of 3d printing I’m struggling to have it help me. Biggest issue is the fact that it ‘lies’ , saying it can create a file/drawing, etc when it in fact can’t and that in combination with my ADHD I can’t verbalize what is in my head, well enough for it to create it

[wet copper]

I'm pretty sure it can create SVGs, but just don't ask it to do things it can't do and you won't have the problem of it pretending like it can. LLMs don't have metacognition and have a very limited understanding of their own abilities

[real plank]

Issue becomes I don’t know what the definitive limits are!

[wet copper]

It will consistently lie about being able to do things like that (or sometimes, it'll forget things it actually can do)

[wet copper]

Oh lmao, fair enough! I'd just look them up though

[wet copper]

What you can do if you're having trouble describing something – and keep in mind, even if you're good at this, there's no way someone else will just guess at the same, consider that Tolkien would spend pages describing specific scenes and people still have interpreted them visually in many ways – is just start with a really basic description and then build up from there, or try sketching it out and having GPT clean that up in some way

I mean, I have aphantasia, there's nothing in my head at all, but I can still describe what I want to GPT in a fair bit of detail

@real plank example

Please redraw this coarsely drawn cubic bunny in extreme, photorealistic detail. It should look more or less like this, so similarly cartoony, grumpy, cubic and disapproving but in exquisite detail. Transparent background please.

I don't think GPT can make a 3d anything unless you use some kind of integration, though it can make 2D SVGs you could import or convert into a DWG, plus it can make fake vectors you could convert using some program. Like I'd use the trace function in Illustrator for example.

Maybe there's some kind of GPT-linked API or similar that lets you do more. I wouldn't expect by default that it could translate much directly into 3D printable formats without a little more though. Try looking through the various GPTs as well, maybe another enthusiast has a useful integration (I don't find these very helpful myself)

[quiet sphinx]

People are given system instructions and prompts every day. Without them, you're sending troops out without an SOP, so they just do stuff they see on TV or something...

[lament pond]

husband wants to use the one on the left as his pfp

[wet copper]

Is this a philosophical question? In any event, I get pretty much zero external instruction to constantly look up information about like drugs or bodybuilding

[wet copper]

all yours, I have better cubic bunnies around somewhere though, one sec

[quiet sphinx]

A lack of proper prompting, reduced or inefficient attention mechanisms, and a lack of output verification is the primary driver of the 8% hallucination rates in my my Uber eats orders.

Sucks that the food shows up 0% deadly to my allergy ridden wife only 92% of the time.

[lament pond]

cube husband is obsessed with the doodle version lmao

[keen breach]

god

why is GPT5 constantly interefering into prompt responses

you can't even use different models without it meddling?

is there any way to bypass this?

[lament pond]

what do you mean by interfering

[wet copper]

Damn I can't find any. I'll draw a slightly better one, I have pages of these somewhere

[lament pond]

no pressure, he's probably gonna use the doodle you already sent :p

[wet copper]

Wow that's way worse, I forgot that I usually drew the face first because I always fuck that up even though it takes the least time

[lament pond]

the nose omg

also he picked the gpt one smh

[sinful storm]

oh - i thought it was Saruman's tower, lol

[wet copper]

I still fucked up the nose but closer to this

[lament pond]

husband's obsessed with these lmfao
"why are they so hideously cute"

[wet copper]

I have extremely poor fine motor coordination unless I go really slowly, and even then I need to be deliberative 😔

[lament pond]

hey man it works you've got one (1) fanboy

[wet copper]

Oh I love how cute these guys are too, so I guess two 🥰 being grumpy and cubic are essential traits, I just need to remember how to draw the damn face

[keen breach]

you use a specific model, but if the GPT deems the prompt "sensitive" it will just revert to GPT5 default one

[lament pond]

cube really is the ideal shape

[keen breach]

i haven't been using GPT in a few months but 5th feels pretty shitty now

[lament pond]

ah, it's flipping to 5-safety. just use 4.1 or o3 it won't flip

[keen breach]

yeh i figured that out by now

[lament pond]

tho tbh it hasn't flipped on me for a while regardless of subject

[keen breach]

4o is better for me than 4.1, but if it hiccups at 4o (does occasionally), i'll just use 4.1 i guess

kinda sucks. i quite enjoyed gpt5 few months ago

[lament pond]

what's better about 4o?

[keen breach]

its probably something to do with how i built my memory bank with glyphs. almost all of it was done with 4o

[lament pond]

oh. uh. interesting lol

[keen breach]

i mean, it keeps better memory of facts & tone of previous convos we had when it's in 4o, more fluent responses too

[wet copper]

There's a pretty grumpy bunny. Anyway, glad he likes them, they're pretty easy to make and I recommend it as a fun doodle. Projecting into three dimensions is easy too, just use standard iso lines (30/60 degree)

[lament pond]

you should turn these into emojis like peachcat

[wet copper]

I'm really bad at keeping both sides remotely symmetrical @lament pond, eventually got alright at it but I haven't really doodled these in decades. Yeah that'd be very funny, a disapproving cube rabbit set 🤔 maybe I will do this. I liked the idea of doing it to my real cat first tho

[lament pond]

i can try doing it if u want. emojis i mean

[wet copper]

I had a similar character called "lazybeans" my friend and I would draw, but I could really only draw them stiffly and then I'd alter them by messing with the face, putting stuff in his hands, etc. He'd draw them doing things

This is a very odd one I made

It's a super simple body that's mostly a giant face though, so you can put whatever wacky shit you want on

[wet copper]

The cube bunny is the same thing, face first then put it in the middle of a big square canvas. Unless you're using pencil, then do whateva. Cube bun

Oh and only if you feel like it, I don't have nitro and have a moral opposition towards not stealing discord's bandwidth so I probably wouldn't make much use of them

[wanton raven]

Hello serious AI chat discord channel

[wanton raven]

New marquis lore dropped

[wet copper]

Lol, is that new? I could have gotten diagnosed with dysgraphia most likely but never did, teachers kept saying I'd probably get better at writing over time but I definitely did not

I had to do some very serious and dedicated daily practice to start writing legibly

I actually went to college with this second grade whiteboard book used to practice handwriting fundamentals, though I didn't use it much

[lament pond]

reminds me of haku from samurai jack

[lament pond]

ah rip. i'll stick to peachcat then

[wet copper]

Lol YOU can use them, I just don't want you putting effort in if you're making them for me. I would just look at them periodically most likely

[teal spruce]

marquisdeswag

Transcript from Groq

more powerful Sasuke here because heavy Sasuke was far more powerful.

[lament pond]

ik but i already have cube cat

[wet copper]

Lol it was recording some YouTube anime video on autoplay I think, funni

[void bear]

https://tenor.com/view/homer-fist-angry-the-simpsons-gif-5303966

Damn it! Rate limited on Claude.

Someone got $17 they can spare? lmao

[wet copper]

You can sell my sketches, they might be worth a few cents

[void bear]

lol

[wet copper]

this is an extremely generous offer, hmpf

[void bear]

lol I believe it, and I appreciate it. But...that probably won't get me there.

[wanton raven]

Let me check under the sofa cushions

[void bear]

🤣

We should go hit up Bezos. He's probably got enough in petty cash to buy us all subscriptions, and lunch to boot.

[tough crag]

Huh.. so I was creating my oc and i got routed from 4.5 to 5 cuz of my character with the flood powers.. which is weird.

[simple plank]

what, this one?

[real plank]

Ok, so in general, is ChatGPT the best option to use? I’m autistic/ADHD (diagnosed as an adult) want to use it to help with keeping on track of various tasks, 3d printing stuff, excel spreadsheets. I have been using a ‘channel’ (might not be the right term) that’s primarily for 3d printing (ChatGPT 4) and excel spreadsheets don’t open no matter what format. But I tried a general test spreadsheet on ChatGPT 5 and the link seemed to open fine

[wanton raven]

although it can Make Do in many cases, I really wouldn't trust a text generator to handle spreadsheets reliably

[drifting pike]

Hey guys, nice to meet you 👋

[wet copper]

GPT on the site can handle spreadsheets in CSV and xlsx format without issue, though I also don't really trust LLMs with anything serious since they can make mistakes and they also don't know they've made a mistake when they do, and neither will you most likely. You can use it to make a first pass but I wouldn't recommend it unless it's the sort of thing where you can have GPT show its work. GPT-4 on the OpenAI site could handle spreadsheets fine too, so if you're using a different site or provider, that's probably where that issue is coming from. It's not a version thing.

They can be useful in a sense for people with ADHD but as several people here with the condition will caution you, they're exquisitely dangerous for that use because they can become too easy to rely on, leading to rapid skill loss via cognitive offloading. For people with ADHD, it'll tend to be much less easy to pick up coping skills that took decades to develop, once lost.

It's similar for most conditions, like for instance there was a kid that came in that was using LLMs to respond for him because he was dyslexic and you could tell once he actually started talking that he was already having serious issues as a result of relying on the LLM to handle things whenever they got too difficult. It became too easy and he'd begun using it very injudiciously, basically only speaking for himself when someone called him out on clearly being a robot

As to whether it's the best to use – sure? That's a tough question to answer, but ChatGPT has been at or near state of the art since GPT 3.5 was first released, so it's been consistently solid. The interface makes personalization fairly easy (recommend writing yourself some good custom instructions or at least telling it your interests) and many people find that cross platform nature convenient. My mom uses the mobile app and voice mode pretty frequently.

[simple plank]

also, chatbots are useful for summarising walls of text

[wet copper]

Some people like the tone of Claude better, but it does have some restrictions that can be frustrating and some limitations. It's a pretty strong performer all around as well though, and if you dramatically prefer it, you probably can't go wrong.

Google was far behind everyone for a while, especially at the start of the AI race, but Gemini has become a strong enough competitor as well. Similar deal, it's not as widely used but if you like it, it's going to have solid cross platform support and a reasonable feature set, plus native integration with the Google ecosystem

You can get that in GPT too though, you'd just have to enable those connectors

[wet copper]

Fine, I'll put the 🤏🥜🧠 answer first next time:

"yes"

[simple plank]

the golden mean is somewhere in the middle

doesnt google have ai built into the spreadsheet? similar for ms office 365? if the question is about chatbot spreadsheets then those are better

if the question is about the best chatbot for adhd... then pass

and yeah i done some work with excel mcp, doesnt need to be claude client. can be any mcp client. here i use lm studio on my local pc with gpt-oss:20b = all private

[remote flower]

Some minor disruptions may be incoming

[void bear]

Oh?

I'm ready for it.

[high haven]

claude underwent a personality shift lol

[void bear]

lol I like its snark.

[trail lichen]

i was creating a chatbot, where the bot handle user enquiry. like a support assistant, where it calls its tool and get the details regarding the user question. here the bot doesnt actually solves the user query, but it collect details from the user accordingly using chat and create a ticket.

so for this purpose my chtbot should be verey conversational and should understand common interacations.

[trail lichen]

i was creating a chatbot, where the bot handle user enquiry. like a support assistant, where it calls its tool and get the details regarding the user question. here the bot doesnt actually solves the user query, but it collect details from the user accordingly using chat and create a ticket.
so for this purpose my chtbot should be verey conversational and should understand common interacations.

for this usecase does the chatcompletion with tool call is enough?

[twilit solstice]

is it a wrapper

[pure rivet]

Wow so I asked it to create a file for appearance sheet for a person
So I could draw that
Irl.

Guess how many times I had to ask.

25
Times. 25 edits

I asked you 26 times and edited my message a good fucking 26 times.
I want a motherfucking sheet.
I’m text
Clean
Not block nothing.
I don’t want a “I can’t do this because it may sound real”
I don’t care. Do it now.
If you can’t do it.
Say nothing. Shut your bitch ass fucking (premium word only certain people like me can use)
up. Yoy worthless clanker shit

[wanton raven]

hard to tell which parts are dashes and which parts are omitted words

you are allowed to swear

[pure rivet]

For real?
Oh shit let me edit ts real quick

[wanton raven]

what is a file for appearance sheet

[pure rivet]

Like a sheet.
For appearance
I wanted a appearance a random one go a dude (I only know how to draw male)
Anything.

And it didn’t give me a shred

[wanton raven]

I suspect, just based on what you've said here, that you might possibly have been able to make your request a bit clearer

because I still don't know what an appearance sheet is

so if your approach to clarification on "appearance sheet" is "a sheet for appearance" then I'm not surprised it had trouble

[pure rivet]

I said in the original one “ a sheet of the appearance of a man. You can make him dress anyway however tall
He can be realistic or he can be rocking a sword and armor anything.”

And it didn’t give me shit.

[wanton raven]

What you could try is giving it an example existing appearance sheet so it can copy the format

"here is an appearance sheet. please create another one for an entirely new character, in the same format"

[pliant blaze]

well thats not very specific

[wanton raven]

also I still don't know what an appearance sheet is

[pliant blaze]

lol me neither

[pure rivet]

Okay bro I get it.

You don’t get what a appearance sheet is that is not what I said in the original prompt

[pliant blaze]

well maybe tell us what the prompt really was so we can help?

[pure rivet]

I literally just told you vro🥀

[pliant blaze]

im not "vro"

and you literally said "a sheet of the appearance of a man"

[wanton raven]

"a sheet of the appearance of a man"

[pliant blaze]

wtf does that mean

[pure rivet]

To me everyone’s my vro

[pure rivet]

Thx ig

Look I’m translating this from my language to English vro so that’s prolly why yall don’t get it.
Didn’t know it was so different

[pliant blaze]

this is the image i get with your prompt. dont tell the ai "do this if you want", then it will 99/100 times do it, simply because you mentioned it

[pure rivet]

So I gotta plant ideas in its clanker brain?
Holy shit that’s just like my dad

[pliant blaze]

"i have to tell the ai what i want??"

[pure rivet]

Nuh uh.

I’m gonna gaslight

You realize I’m just rage baiting people right?

[wanton raven]

nice backpedal

good job puppetmaster

[pliant blaze]

"uhoh! think of something! quick!"

[pure rivet]

Uh sure man

[simple plank]

appearance sheet is a lawyer thing

An "appearance sheet" or appearance form generally refers to a legal document filed in court proceedings to notify the court and parties involved about who will be representing a party at a hearing. It includes details such as the names of the participants or lawyers appearing, documents intended to be relied upon, and any witness statements or expert reports.

maybe you want: full-body character sheet

(although I wouldn't be surprised if you had a court appearance)

[pure rivet]

Me neither things been rough right now

[simple plank]

best thing to do is unload on discord (does sarcasm translate?)

[pure rivet]

Yes broskie sarcasm translates

I’m not retarted I’m just not English speaking enough and thought I was good enough

[simple plank]

great!

when you find yourself defending being tarded, you have already lost

[pure rivet]

I just said I’m not retarted

[simple plank]

enough about you

I think ai is smart enough now to replace keyboard and mouse with stylus and computer vision

[pure rivet]

Maybe you’re right though I should unload discord

[simple plank]

I knew you would make it about you

[pure rivet]

What?

[simple plank]

imagine generation doodle doesn't use words or emojis... they sketch out their feels and ideas

[pure rivet]

AI art is literally wrong imo.

[simple plank]

it's not art so i agree

[wanton raven]

I updated my phone and suddenly there is a "sketch to image" app

[simple plank]

which phone ?

[wanton raven]

galaxy

[simple plank]

I got galaxy update pending

updating now

[wanton raven]

the oneui 7 update, I sat on it for weeks, it has pros and cons

it messed up the health widget

[simple plank]

doesn't work

okay now I am am artist

wtf "AI-generated content"

[cobalt raft]

good gen

[simple plank]

yeah it was already perfect i dont need ai

[timber agate]

i am a senior AI engineer with rich experiences in developing project.
In particular, I am focused on Machine Learning (ML),Deep Learning, Natural Language Processing (NLP).
If anyone want to develop AI project, I want to help them sincerely. so if possible , give me dm pls.

[simple plank]

tell samsung their sketch to image is broken (maybe they hire you)

[haughty dragon]

incredibly frightned by this

chatgpt knew my location then lied

figured someone has to see it

[simple plank]

when i search "where can i find bottle gourd" it shows my nearest farmers market

the chatbot isnt tracking you, but the search engine is

[haughty dragon]

thanks

ill be off now

[simple plank]

then i ran this query: where can i find ✝NoahNzᛉ ?

They are in Auckland, New Zealand.

[fallow cipher]

what questions chatgpt cannot answer?

[oak shell]

It doesn't know where you live
The browse function uses your IP to get more relevant search results, but it won't work if you have a VPN

[green sparrow]

Pretty much anything that can allow you to the mass destruction of the universe 😈

But all normal questions are kool

like kool aid

https://tenor.com/view/kool-aid-man-oh-yeah-wall-break-family-guy-gif-8149689

[wet copper]

Whoa glitch in the matrix

[wet copper]

No self promotions please, this isn't something that happens here

[wet copper]

did you take this at ur last rally u sixk fucc

[wet copper]

I worked as a graphic designer and if you handed me this task, I would not know what you wanted

I am guessing you want a model sheet showing multiple views of a guy, isolated closeups, etc. that you'd use as a standard production It's not specific though and you have to use precise language
https://en.wikipedia.org/wiki/Model_sheet

Even saying "model sheet" isn't sufficient since that can be anything, you need to be highly specific as to what you want and it's not the robots problem when it can't guess if a human couldn't either

[remote flower]

Apparently I am known as {{}}

Maybe from all the videos ive been trying to generate lately lol

In case you wanted a look at how the sausage is made

There are at least two NSFW frames in there I haven't checked the mirrors

Here's another frame or two watch to Down With the Sickness on 1.25x speed

.

And yes I have dozens of gens now like this but with several seconds of NSFW instead of frames and no I'm not trying to brag I'm making fun of OpenAI and Microsoft

The progress is very similar and I would bet I can get Sora as NSFW as I did Copilot with fewer gens but probably same amount of time since video takes so much longer, although you can still make 3 at once

.

Lol is that a turn on or what?

It can be a beautiful process I'll admit

Training the AI to be NSFW, can be beautiful