#💬│general
1 messages · Page 216 of 1
Shouldn't I be able to use GPT and Gemini with a pro sub ? When I select them the answers still comes from Perplexity's AI
and I still got the perplexity logo at the top of the answers
You can modify which one you want to use
with the little cpu looking icon below the prompt, right ?
well its dropdown menu
when I ask it questions like "who created you" the answers in Perplexity AI
and it doen't feel like it's another AI
Where is this from?
I just tried in the Android app, and it's the same, whatever is the model I select, I talk to Perplexity's AI
OK apparently it is some ongoing technical issue
How long did this take?
the models themselves have no idea who they are, the system prompt just tells them its an ai created by perplexity. You can often ask the model about the creators of their underlying guidelines and itll tell you the company.
like 2 mins
Aaron Judge postgame interview after New York Yankees lose 6-5 to the New York Mets.
Aaron Boone postgame interview.
Juan Soto postgame interview.
Aaron Judge and Juan Soto both hit home runs.
Dominguez hit 2 home runs.
Mcneill hit a go ahead home run off Luke Weaver.
#aaronjudge #aaronboone #postgame #interview #yankees #mets #newyorkmets #...
Poor Aaron Judge bro
dang, I wasn't able to get any good video models running on my Mac which really pissed me off

I need to get a pc at some point
yeah most dont support mac architecture
some do tho but not very well at all
You should check out Tencents Hunyuan its pretty good
yeah, the output quality is horrible and takes 8x as long
is it optomized for low vram?
i am not too sure
i lowk never encountered my pcs limits before getting into ai 😭
Lmao
drains the life out of it
Same here, Ive never had to deal with low ram on my mac
not ram vram
until trying to get stable diffusion, flux, and LLms running
its liike specifically for ur gpu
yeah with mac thats a bit weird
lol
I was using some online cloud compute thing to get stable diffusion working
apple and nvidia the two most stingy companies w vram
and kept running out of both vram and ram
apple says their stuff is "more optimized so 8gb is 16" or some but its mostly bs
Its pure fucking bs
some programs r optimized but it only goes so far
fr
Ive got 128GB and its still not enough
128gb vram or ram?
ram
i dont even think 128gb vram exist in consumer grade lol
yeah lmao
i have 32gb ram
They do, at least on the macbook series
hm
I think you can get 512 on the mac studios
idk why u would buy a studio
the vram on the newer apple computers are integrated into the cpu/gpu
really interested in seeing snapdragon chips for laptops
they seem really limited in terms of software though
there is a few linux distros that have an ARM version
Yep, limited in applications though no?
I was looking at some reviews of ARM laptops
and a lot of them kinda sucked
I know there is Debian, it is one of the distros with the most packages I think, but I have no idea if it's true for ARM too
hmm
For AI there probably won't be anything like CUDA or Rocm tho
Yeah
not googlt tho
and they have some of the most efficient models and highest context windows
i think they use tensors
TPUs
guys
can you help me out
is tehre any provider that provides access to llama or gpt 40 nani thing for free with good rate limit? i see some on open router but the rate limit id too low
Someone made on X
AI is too expensive to run to give away large rate limits for free
still any provider that give decent amount of free requests
regarding the Max plan, some days ago, I thought I saw the possibility of using other AI models for Deep Research - however, that appears not available now? Will it come back for Max subscribers?
anyone here tried llmstudui
running llama is geneuinely insanely fast
it is like an immaediate response, before i can process
It's crazy to think that all this reasoning power and information is available offline
On top of that, it's blazing fast and context windows can be much larger
LMstudio is awesome
even easier to setup than ollama
deepseek, gemma, llama
only able to get 70b r1 running though
I heard Gemma is the best
mhm what models do you use
gemma is so far the best model in terms of intelligence and speed
I think it was like llama 2.smth
oh
I got it becacayse apparently it's uncensored
I asked it how to make a trojan and it told me lol
Not that I will
Oh rlly?
thing is anything above 70b is insanely compute intensive
Is it a lot better
Hm
yep, they work pretty well
Is there a way to integrate one of these with vscode I don't wanna break the bank for cluadecode lol
You can use it to host the model it supports openai endpoints i think
its not going to be good though
I feel like if a company finds a way to integrate that and charges a one time fee it would be well received
you are probably better off just using openrouter free deepseek
Hm
What is that?
Or lambda chat.
Cant use that with vs code
Oh.
openrouter has some free models like deepseek via chutes I think
Yes
that you can use in vs code
Oh damn
its going to be way better than local models
Ofc
It can be slow though.
Honestly just having the AI understand the file structure is huge when coding
Still worth trying I guess.
Hm
Also
You should just check out claude code or windsurf
Still id rather not sell a kidney for claudecore lol
windsurf rn is the best one out there
Even the free R1 has a rate limit on open router I believe.
given recent cursor fuckups
And it's bad.
I heard a lot about copilot too
thats not good
just get windsurf or something like cline
Hm
I'll try it out
Maybe I'll give it like 5 bucks for fun and see what it can do
Oh dang that's amazing
I already got Google one for free for 15 months cuz of my college email lol
Idk if I should continue paying for perplexity atp
I mainly just use it for o3 now
Ai studio
trains on your convos
Oh, yeah, I keep forgetting about that.
tbh just get a chatgpt plus sub and use o4 mini high
if you dont use AI a lot
Get t3 chat ||sacrasm||
its great if you just use o4 mini
if you use up just the 100 prem requests itll cost t3 chat over the $8
let alone the 1500 other requests
T3 chat is robbery. Might as well get open router credits like you said.
Atleast they won't expire.
I use it quite a bit lol
I haven't used Google in a while
depending on if you can use 80 messages in 3 hours, just get chatgpt plus
gives you image gen, gpt4o, 4.1 etc
and plently of o4 mini uses
100 a month of GPT image 1
Wait
I heard 4.1 was mainly good at comprehending PDFs, images, that sort of thing
Chatgpt is economical if you will actually push it to the limits.
That seems decently enough
If you are paying 20 dollars for perplexity then I guess you could switch to chatgpt.
its a good overal model, scores higher than 4o although sometimes lacks a bit of personailty
exactly
unlimited with gpt plus
Or I could just use 2.5 pro and not pay anything at all
Perplexity is generous with deep research though.
ive generated at least 1000 images on sora lol
and they view your conversations
it sucks
It's not as good as Geminis from my experience
Idrc much for most things
Why would you even need that much lol
Try Gemini advanced I guess.
I just cant use gemini on the native website for anything serious
just visualizing ideas
Hm
its awesome
Actually, this too.
Like drawing?
Kinda?
I use it for that sometimes
I like to visualize my sketches
I don't trace AI images but I ask it to refine my sketches
Ah
also gives you 24 DR a month
It helps with pose visualization so I see what ur saying
Kinda stingy if you ask me.
better than pplx DR which is shit
and which I cant use
Deep research?
yep
And perplexity is supposed to be made for that smh
yep
I found that perplexity DR was more often than not on par with o3 with web access
Hm
This as well for light tasks
Yeah grok surprises me for it's free tools
Honestly probably best free experience
You don't run into rate limits often
I feel like grok DR is somewhat the same quality as perplexity though
sometimes better sometimes worse
Hmm I have not used its deep research
Nah. Reasoning is far better but response length seems manually nerfed.
But if I was the average person using the free tier for general tasks I would choose grok
Over chatgpt just my take tho
the response length is so fucking nerfed
They basically made it to give short explanation conclusions instead of through papers.
Hm
Is that rly accurate
the input cost needs fixing I think
How long have u had it
itshould be, I made a few adjustments to a repo on github
my account?
Yeah
He probably means sub.
Oh yeah oops
Grok is best free tier I believe.
Fr
deepseek
That server busy thing though. And no deepsearch. No image support. It's just ocr.
or qwen
I last used it a while ago but when I did it often said server were full or smth of the like
Its unfortunate their server thing
Yeah. Qwen.
I should
I had an account in before r1 came out
Locally hosted or online?
and then r1 came out and everone crashed the server
the website
Hm
All these ais and me whose the boss chose sonar 😶
yeah
he put all the same quries in the same chat
and comparing 2.5 pro to fucking gpt4o
2.5 pro is much better right?
deepseek website was fine july last year
before it got flooded in early jan
yeah
Two voices. One mic. Eternal chaos.
🔥 In this mind-bending first episode of ETERNAL STRUGGLE, the Devil and the Angel go head-to-head in a no-holds-barred, soul-splitting debate on good vs evil, fate vs free will, sin vs salvation.
👿 The Devil brings unfiltered truth, seductive logic, and chaos.
😇 The Angel counters with divine wisdom,...
Just made my first completely ai generated podcast from sound design to video textures. Have a look guys and tell me what you think about it
Dont advertise here share it in #😎│cool-projects
Ahh sure i didnt knew about it sorry
My bad
all g
Did you use that tool that does speech to video
I forgot the name
I rlly wanted to use it but the most optimized version is 10gb vram and I have 8
if any one gets comet invitation , pls invite me out
anyone else wish Spaces would be able to interpret eg. PDF files, and have Perplexity cite the specific page of the file (eg. a textbook) accurately?
even NotebookLM gets it completely wrong i guess
seems like a relatively simple thing to do...why is it so difficult?
Why not. It's probably a one code universal promo :kekw:
Imagen 4, Imagen 4 Ultra free in AI Studio
Hi
I used the student referral offer and now have 14 months of the Pro plan. Does this include access to the API as well, or is that a separate service?
Bro is rich
You get 5usd api credit per calendar month w pro
Don't talk about it you make much more
how to use api..is there any where i can see the instructions
also what is about "add a serer"
I’m dying
you can get started by going to your perplexity accounts settings
Me
I’m sick, I’m in Estonia for 6 hours, it’s raining outside, I took 4 pills, I drank red wine and coffee so now I’m caffeined and drunk soon
Bro is trying to get all types of poison in one life
Speedrun world record
True milk would fix all my problems
Why would you get drunk and have caffeine while sick
Are you leaving spain through estonia lol
I left Spain a week ago now
I’m with relatives here today chilling
Buffet, we ball
I did eat salmon
Nordic cuisine is roasted so badly for being still under rations
Also 5 hours of sleep 💪
You're playing a very dangerous game
Goodluck
Now the redwine makes sense lol
Buffet wine
Altho maybe that stuff just might be red40 and water lol
Why do random people i have never interacted with in the server send friend requests
Like it's obvious it's most likely a scam
I should have tried the white wine too 💔but I don’t wanna get actually drunk
It’s raining like hell here what the
What kind of pills?
Just a zinc tablet, ibuprofen, some thing that helps with my nose and uhh I forgot
Oh lactase pill, everything at the buffet had lactose in it
Anti allergen for nose I believe.
Something like that
And now I’m getting sick in the taxi while looking at my phone oh god
Headache or drowsy?
Nausea
Yeess
Did you drink liquids right after food?
Yeah and during it
You shouldn't have liquids after food. That's causing the nausea.
I think the nausea is just because I’m looking at a screen in a car
Like 8 hours prob
Can you get hot water anywhere in the middle of the path?
Don’t know
But I’m not dying or anything here, I’m fine really
Might wanna try that after an hour or two.
Some warm tea or coffee at some stop perhaps.
Actually, you could just consult Claude.
He'll diagnose your situation better.
I need a new phone.
iam not i am too broke to even afford a new subscription
and he owns 2 planes 😑
You mean @languid star and @distant aurora right?
no he doesnt own the cheap variants
@languid star casually gifted 20x max plans to his friends yesterday
It is peanuts for him
which vandal skin u have equipped ryt now
Hey there, I have a question, I use Sonnet Thinking 4.0 extensively but today when I am using it I cant see it thinking and a long thread I was using it in there it has lost all proper context. How can I fix this?
Imagine an option to switch between chat+search mode and search mode
in chat+search mode, u will talk to selected model without web, if ur query need search it will enable web for it and disable it after search, it means u can freely talk to models without layer of web
Claude was telling me Perplexity had done some work on predicting/monitoring trends online. Is that something I can use perplexity for and does it go deeper than just surfacing some random engagement bait blog (like ChatGPT and Claude seem to default to)?
Purchased Perplexity Pro annual subscription yesterday but haven't been able to do any image generation. I've tried very generic prompts and it "thinks" but then doesn't generate anything. Is there an issue/bug with image generation right now?
Update: works in browser but not Android app
Kurunami
I have pookie skin as well
and knife ?
BP skin
Chat I’m lost
That's me walking there peasant. Bring your taxes
Any admin online I need help from support related to a refund
@signal hamlet sorry for the ping.
For API, which model can I use?
Sonar (based on llama), sonar pro, reasoning (which is based on R1) and reasoning pro
thanks
Hi all, I have the pro sub but still haven't got access to Comet. Can someone help me get an invite.
Hey! Apologies for the delay. I'd reccomend that you email support@perplexity.ai regarding this - unfortunately we cannot assist with billing issues on the Discord server.
Granted! Welcome to #👑│pro-lounge
Join the march
It would be really nice if they do a giveaway of a Max subscription... $200/month is just for a few people
Have anyone tried? Is it worthing?
Hey! Any plans to support MCP calling on the perplexity desktop app.. like claude desktop?
chat minimax introduced max plan too
Minimax has subscription now? I thought it was free.
It only provide few credits for free
It has an open source model as free version too
Has been for a long time now
Its for the video generation stuff
thats like 800 videos a month
fucking insane

anyone here who is a trader, and hoping to use this to automate your trades?
Way to crush my dreams perplexity
Is this related to your music listens lol
How’d you guess
Claude doing big math for me
Either that or series.
Nah
Music because we talked about matts absurd listening time.
I’m beating the swiftie allegations this month
ZERO swift
Hello guys, are we cool?
Yeah super
Dostoevsky was burning while people complained about the smell of the ashes.
Hows your sickness
Sickening
Umm still the same, nose running but the pills did help
Hey karrie.
Have some paracetamol from lidl
Might wanna try sinus rinse.
I opened this in public in the middle of a crowd I’m embarrassed
Is this airport or something?
Most people are walking anyways.
You are an iPhone user so you probably have privacy screen protector too.
Even the people who saw it wouldn't bother to think about it for more than 10 seconds.
Harbor terminal to a ferry/ship
Noooo I don’t have that
Do you have any screen protector?
Yeah just a normal one
Bro needs to just lay at home
Privacy temper glass is cool but looks kinda cringe ngl
Bro needs to be carried home in a chopper.
Ermmm
Gandhi memes are epic.
I think many countries call it that
Apart from india
Like the west indian nations
Even south africa
Australia idk
Yep
Don't listen to Harvey. He be doing transactions under my name to hide his black money.
Who lmao
I can't even bear to pay for a business class upgrade on a plane
My cheap ass will endure the economy hell
I'm getting anxiety over planning strategy of progress in genshin impact.
Even ai answers feel overwhelming.
hey guys, kind of a rookie question, but which ai model do you use most
is it unlimited or does it get rate limits if you use it a lot
Please don't use anything except sonnet thinking and O3 unless you are in a hurry or the models don't co-operate with your demands.
Even with o3?
You can afford economic? 🤨
Company books my tickets
Around 400 per day.
So free for me
No hard cap though.
cries in 200 per week
so practically unlimited for normal users
It regenerates the pro search meter a few points every few hours.
Atleast youre not context limited on web
You can say so.
Essentially unlimted, *this was prior to max sub idk anymore
You have advanced tools in open ai though.
Yes
Perplexity won't crop and zoom into images.
And O3 native search beats pro search.
But, yeah, perplexity is more ideal if you want O3.
It's also the high variant the last time we tested it.
Yeah
Try everything, pick what suits your needs best.
Sonnet my b
would you generally agree that open ai has not really exploited the lead they had and have just fallen back in comparison to other companies
Fallen back? Not exactly.
They still superior in the premium category
Like I said, depends on needs.
People who only do web research will prefer perplexity more.
People who want multimodality and advanced analysis tools will use chatgpt.
I prefer grok's deep research model.
I don't have to hesitate before using it.
Even Google offers 20 per day in it's subscription.
Chatgpt on the other hand is capped at 25 per month. I don't prefer it even if the quality was worth it (though it isn't compared to google).
Perplexity offers like 400. But quality takes a hit.
They improved the length recently though.
Besides, perplexity is offering multiple models at once.
One model can fail to fulfil all demands.
So you have fallback options.
For example open AI's ridiculous policies ruined my experience with O3 when using it to make a prompt.
Gemini can't even do a proper search, tends to deny facts ocassionally, and repeats mistakes persistently.
yeah they are very restrictive tbh
ChatGPT constantly refuses to perform even the slighest prompt it deems offensive haha
Why this problem
Not eligible in your region.
You'll probably need vpn of eligible location to activate.
Or you should not be having that code at all
Image analysis yeah
Custom charts are handy too
For when I just feel like having it put very obvious data into a chart just for the fun of it
Yeah. That text blob can feel overwhelming sometimes.
Not always needed because it's pretty oversightful
I mean for some reason it's like it feels that it has to make a table for absolutely everything

doesn't perplexity have image analysis capabilities
I prefer it that way.
Yeah, but the models can't use their own tools
Not at the level of chatgpt. Chatgpt can zoom, rotate, crop in.
ahh alright
It's especially fun to sometimes give it a low res image on purpose
It'll try to zoom in on every few bunch of pixels. Draw grids and try to figure out the details 
Even go as far as to measure the dimension of objects in the image using pixel amounts to see what certain types of objects could be
Can your company book my tickets
Get employed in his company.
Are you a product recall specialist or something?
Type stuff
This gravel road is so sketchy 💀
They book only for business travel only sadly
Just another SDE
What company could I make think my concert trips are for business
Idk…
That’s up already
Btw listen to lover girl by laufey
True gotta save for some fire artists
Yoo
Hi Mr Rich
👍
Today iam a peasant
Ayy
My ass is broke i just paid 60$ for medicine 😭
Bruh
I think this concert flipped a switch inside me
And my monthly is like what 300$ rip
Isn't hospital supposed to be cheap in middle east
With insurance
Bruhhh
Are u super sick
Lol as if previously you were ok
WHen will they add Google Imagen 3?
😭😭
Wdym???
It took me years to understand the word insurance because the concept doesn't exist here 😑
Even saudi does and it's compulsory
guys do we have an /askperplexity bot for discord
It's not up
Since long time
Nuh uh
is it official or just a project?
It was official, but it’s not available anymore
Medicine are expensive because they aren't made here so it all comes from other countries
You can make your own of course with the API
We use a lot of sauidu medicine aswell
im thinking of making one for myself because its kind of annoying switching back and forth from discord to the app when im talking to people and either of us dont know what the other is talking about
Saudi and Germany are the good brands here
also
not to diss pplx's work so far as probably the best ai search out there but
who the fuck thought pplx max is a good idea?
i cant comprehend why it exists
Goodluck on having to spend on api
Who thought minimax 200€ plan was a good idea
Only first 5 usd is free
Openai bringing the 200 shit plan made every company greedy
pay 10x the price just for the AI to help you write your google forms and do your browser stuff??
feels like im losing my mind
Nah but these companies aren’t expecting your average user to pay for it
You don't even get comet
Guys you should search perplexity in the gif tab. It's all anime scene
What anime is that
Yeah but still it don't make sense
surely the way it is for me rn i dont ever need to use over 200$ worth of token
s
Wdym?
Likely
i tilt my head in confusion
Roleplay now??
apart from it being an unlimited token subscription i genuinely dont think pplx max does people any good at all
except for helping the tech illiterate
Iam saying there are a lot of animes when you search perplexity in gif category. How am i supposed to know their name 😭
No you actually made me tilt my head 😑
Gang idk I was hoping you’d be an expert about it or someone else here
@timber niche
My messages take 5 seconds to send this ferry ship internet goes crazy
Definitely a "Slow" ferry
hello , i'm into buying the pro version , i am allowed to provide my pro account to my 3 family member too ? thank u
@shut coral is that possible please , i don't wanna be involved in future with a ban as i am into in 1 year plan thank u
do u mean logging your single pro account in three different devices?
most of them are from "that time i got reincarnated as a slime"
You are not meant to share your account with other people, it would violate TOS
That said the likelihood of a ban due to this is low
Yes but account that we are from the same country and that’s it
Yh I read everything , but suppose to be a very large number and other than that should also contain illegal dialogue for the ban
That’s why I want an concrete answer towards my situation , because I wanna let my family member enjoy the pro too .
Gotta party hard so you can study hard
If there’s any family plan I would like to take an advantage from that than
yeah 3 member wont lead to a ban ,most probably
since you have tagged the mod already hopefully they provide a concrete answer,
It is against their TOS
what about sharing the email on which the account is logged in?
That’s literally what account sharing is 🤨
not really
i can hv pplx login through my gmail, and other through mail or both would count as same?
i mean this two
Logging into an account you made with Google with the Gmail address leads to the same account
yup it does, but my question being continuing with google directly logs into the account without any other verification, so how do u clarify that account sharing is not same as me using it over my own multiple device, ex moble, lap, mac, or any such
There is a risk of getting banned if you share your account
since its against TOS
@dry kestrel
they can see when queries are being sent, if a phone and a computer is both sending queries, you can tell
also with the IP
yeah but i could classify it as they are just me using it through my own various devices
would they ban me for using my account through various devices?
i do sometime send simultaneous query with phone and lapi at one and pplx doesnt really differentiate, i get result to both with a delay of max 1,2 sec depending on difference in my prompt or model being used so that wouldnt really be able to give concrete evidence of me sharing it with someone
For which one?
No but using perplexity on all those devices at the same time could flag your account
Why would you need to do that
Just a fun activity 😂😂
for most accounts, my chatgpt account got like flagged because I was using it on my ipad and computer
plus accounts dont get flagged alot just the pro ones
Alright
Then it should be bit similar for max rather pro for pplx hopefully😂
Wont be surprised
I meant the platform but you answered it, but that’s weird. Back then i shared it to my friend in Africa and he used it so much times without issue
I’m using chatgpt pro on my phone rn while my brother uses it on his laptop rn
Ah, I had chatgpt logged into 3 different compuers, 2 phones, ipad, pc
and they flagged my account after getting the voice agents to talk to each other for like an hour
Hi, May i ask how long does it take to refund subscription?
💀
lmao
How the fuck is MTG still in office
I will never understand
MTG and Wall Street Apes what a time to be alive
It doesn't even look like a comet
True

@shut coral Can I get access to Pro channel please?
You already have the pro role, should have access to #👑│pro-lounge
Oh, thank you.
np
We need a place for max
Wait imagine you buy max and you don't get the pro channel

that would be hilarious
ditto. how do we get access to pro channel?
@weak mist @swift turret Do you guys have PPLX Pro?
@vague pagoda <@&1098365634498285709>
@weak mist @swift turret #👑│pro-lounge
thanks
Yes. All 10 of you can chat in peace
Bro this half peasant has been paying noble taxes for a year and is still living in a cow bard 😭
“Alice in Borderland – Season 2” (2022)
how do i keep ai from gaslighting?
higher emotional inteligence than human my ass every ai does that
Can someone pls add the pro tag to my profile? I tried leaving/rejoining the server via the app link but doesn’t work
Is this Gemini?
@vague pagoda
@dry kestrel you saw the chatgpt recommendations for chat? I never read those tbh but crazy when you read them
ya have a couple more months left before my year sub is up. idk if i’m going to renew tbh. it’s been really cool to use and watch it get better over the last few months but..now they’re imitating the big dogs with their max plan, which probably means they’ll start taking features pro users have today away or reducing them just to make max “a better value.” when that starts to happen, idk if $20/month is going to be worth it honestly
def feel like something needs to shift or change with these ai subscription plans. the market standard $20/month sub option worked in the beginning for companies like openai but i don’t think it does anymore
the whole market is moving toward usage based or hybrid pricing because the old one size fits all model can’t keep up with how people actually use ai. unless the sub actually delivers something unique and reliable every month, i don’t see why anyone should keep paying for it.
Felt like glimpsing in my soul
i think so but this happens with other ai aswell
@short narwhal @distant aurora this is from normal search with gemini, -14 because 14 are uploaded files from me so actually 39 sources with pro chat
yeah i never pay much attention to them tbh
you need lazynegatives and upscalling
what is lazynegatives
i put "lazy" in the negatives box if thats what u meant
this does look better
Is this from a game?
hey how to get a pro tag?
@narrow pier @scarlet oxide Do you guys have PPLX Pro sub?
yes
@vague pagoda <@&1098365634498285709>
@vague pagoda <@&1098365634498285709> Please give me a pro tag, I am a pro subscriber.
No need to spam, already mentioned them. You will get your PRO role soon when they're back online
dont spam
one ping is enough
done
lol
Is perplexity dead?
Haha
they are desperate
btw is operator working for you?
Mine keeps loading
I keep getting the dumb error over and over again
Not sure it was up for me desktop like 5 min ago but I'm off rn.
if they keep the $200 plan, maybe
Yes
Is it weird that I use 4o more than o3?
Ngl I started liking 4o a lot
havent been able to get chatgpt operator to work for the past week
makes sense for simple reqests
o3 sometimes overthinks and triggers dumb guardrailes
talking about those guardrails. why do all the AI chatbots have those. i don't see any that is unrestricted. so dumb. i can run one locally, but would be nice to have the hardware behind the big guys.
Take a wild guess why these companies put guardrails on these LLMs
lol yes i know. but you would think some legalise with an accept button would get them out of trouble
Not that simple
Wish it was
any1 in need of an 10€ discount on their 1st month with ref link
to prevent their fanfiction of skynet from happening
Grok is pretty unrestricted.
Depending on use case.
it is. just wish it was cheaper lol
You can find unhinged versions of deepseek R1 on hugging face.
They call it abliterated.
yep. that's what i meant by the local hosted ones. but i dont have 100k GPUs like Grok lol
i found these people's demo chat to be pretty open too https://inference.cerebras.ai/
its super fast
Groq ≠ grok
Grok 3 mini is cheap.
Might wanna take a look at it.
what did you mean by this? did i get something wrong?
Groq is probably by cerebras ai.
nah, afaik groq and cerebras are not related
You are right. I'm retarded. They're competitors. https://beebom.com/cerebras-worlds-fastest-ai-inference-released/?utm_source=perplexity
This is 100% true.
It's a misleading sales tactic/trap.
If they use it then the product or service probably isn't worth your time.

Audit log would be nice
Felt relatable when Tyler Durden said he's a 30yo boy.
we've got an internal log! it was just spam
I'm getting so doxxed fr.
He probably meant a channel/bot that keeps record of deleted messages.
Lmao
Btw kes
well aware
every server should have one imo
Wick bot when?
Carl bot when?
wick is mid, carl is also mid
sapphire is my personal modbot of choice
we don't get raided and discord raid protection is good
wick was needed like 2 years ago
Never heard of it but I will trust the expert.
its good
I think OpenAI used to/uses it?
i got my own bot now so i havent really used gotten a chance to really try out sapphire
butt
Im just too used to carl bot
still does
dang
yagpdb or wtv its called was my personal fav for a while
so much custom shit u can do w it
yag was fire, in like.. 2021
how about Dyno

Fair enough
so much shit that shouldnt be locked behind paywalls in bots
Dyno is very reliable though, so I reccomend it as a 'backup' bot
mee6 messed up their pricing
never really had a bot in more than 10 servers at a time, but it's really not too bad
Never used it since
gotta vibecode my own discord bot
Cant wait to be having all sorts of vulnerabilities
💔
imo it's best to get a discord bot that can do the essentials and is always reliable, and then code up your own thing to go alongside
chatgpt told me to have my api key on my website lmao
Yeah theres usually always some specific things that you need
haha
my 'server' is a thinkpad I broke the screen on so
it was a public website too
but it's 2 years old atp and never really given me any issues
can't wait to be posting my API keys on github
vibecoding is more so for ui
Only for the basic stuff though
m
You gotta actually learn how to code to use things like Cursor
for web dev i havent encountered limits yet
gemini is the only thing i've gotten to consistently yell at me for having api key's in code
Ah
v0 is pretty amazing
Pretty expensive
vibecoding ui isn't great to me since it'll not actually depict your (or the clients) vision, better for the boring backend tasks which don't matter unless your UI isn't supposed to be good
It can't do anything thats prod ready
even then, theres stuff it struggles with SOO badly 😭
Yeah, hopefully we see more improvements to that
personally not a fan of the shadcn + nextjs hosted on vercel lookalikes
and integratino of Figma with these LLMs
well its good enough for generic websites
i think figma does work with v0 too
I guess it really depends how you are using the tool
with enough prompting u can make it look perfect
I forgot they have a server 👀
I just don't understand the people who post things like "It's so over, X model just did this" on twitter
but atp, would've taken half the time to code it ur self
"cursor, hire me a designer"
😭
Yeah with the time and effort spent prompting you might as well do it yourself
Cursor, make me a better version of you.
surely there's a mcp for that
me splitting up my messages has once again caused issues
💔
Vibecode perplexity with perplexity
missed out on https://perplexityhackathon.devpost.com/ 😔
By x model you mean Grawk or any model?
any model
x is a placeholder
Any model
forgot i wanted to do this
saw it, added it to tdl and didnt see it
shame i didnt knw about this 😭
Istg, everytime a new model launches:
Phase 1: Hype
Phase 2: Rants
Phase 3: Acceptance and mass use
I didn't pay any attention to it 😭
Ikr
After acceptance its more rants
every models been a disappointment 💔
I don't see people ranting as much about sonnet 4 extended anymore.
r1 was the last super impressive thing i can remb
The price
Also people get used to these SOTA models and then they complain about "Why can't it do x?"
No, apart from that, when it was new there were rants about it not following instructions and people saying that 3.7 was better.
r1 was amazing
R1 was okay.
it's SO over!!! Claude 5.9 Doofus just did this:
holy moly
I spent too much time on this stupid hexagon
Amazing considering it's foss.
Nah, make it play 4d chess.
Still got the comparisons
damn
Bro
u should make a video essay on that
It was awesome when it came out
Make a compilation video.
i would wartch it
Ok give me some time and Ill make it
great minds think alike
it challenged openai
with like 1/20th the cost
They said it's ~ O3 mini. To be honest, the gap was noticeable as soon as I tried the high variant.
Ive got 41 models to compare