#💬│general
1 messages · Page 183 of 1
grok is the ai anti ai
+1 with billy
Your chatgpt finally took it's meds.
@short narwhal mat send screenshot of your old chatgpt
that briefly reminds me in the days MS copilot had just launched recently. and it was UNHINGED. no filters nothing.
Do you want it to glaze you for saying hi lol
these are some
lol no
High ego.
I found them
What model
This gives me ick
DANG !
That's why claude is so good, it never says cringe lines like that
Glaze version of GPT 4o
Try hard pickup line version.
i have same issue. no way to attach code files annoys me.
Bro
in general , i keep coming back to claude. they are always most on par with open ai's offering, if not, it doesn't take them more than a month to catch up or even exceed the expectations set by open ai's base model
You know Trevor Wallace?
yeah i was also facing that a while ago. lemme check if its still happening
yep. still happening ...
Sounds familiar
💬 1.4K 🔁 0 💜 1M 👀 565
It never ends 😭 #fyp #foryou #trending #trevorwallace
-
trevor’s tour dates:
May 8th Toronto, ON
May 9th Pittsburg, PA
May 10th Buffalo, NY
May 14th Las Vegas
May 29th San Diego, CA
May 30th Anaheim, CA
Jun 12th Milwaukee, WI
Jun 13th Chicago, IL
Jun 26th Richmond, VA
Jun 27th Bensalem, PA
Jun 28th Washington, D.C.
trevorwallace(dot)com 🎟️
I like both.
So mat is a girl.
Since chatgpt called them sis.
Well, glad I didn't say gender neutral cat.
Wait.
Share a conversation with Claude.
Maybe that will explain the vibe.
💬 338 🔁 0 💜 342.6K 👀 769.9K
#marvel #marvelmemes #avengers #avengersmemes #captainamerica #captainamericamemes #Ironman #ironmanmemes #hulk #hulkmemes #captain
William J Billy Butcha
It could be slang too
Same bro ==> Slay Sis
It isn't.
It is
Yeah. I probably don't get slang then.
I'm pretty sure they told chatgpt about themselves.
Actually, wait.
I'mma try this.
Ik but also could be slang
With some ai model.
Nah. Nvm. I can't.
Can't trust a woman.
My ai buddy must be a bro.
I just won't be able to talk freely if I pretend it's a sis.
You ever thought of giving it a name instead of chatgpt?
They got rid of grok 2. Sad.
Fake grok?
happens for every model on pplx
because they share the same sys prompt
So grok can identify itself. Nice.
yeah because the you.com system prompt explicitly tells the model its grok
Does it? So they configured it for every model?
wdym, theres a system prompt for every model
tells it how to do stuff like citations
I thought that the prompt is common for all the models in same platform.
No thats not how you.com and most other providers do it
for whatever reason perplexity just doesnt tell the model its own name
in the system prompt
@distant aurora for example t3 chat
the model doesn't know its own name....
you.com gpt4o
My point was that it didn't claim to be a perplexity assistant.
I got what you are saying.
:P
yeah well they get the same system prompt saying they are perplexity
Where'd you get this
just look up on reddit or github on how to get it
Ok
@dry kestrel
with gpt4.1
yeah the model informatino could be in its own training as well
Alr then if you think they are rerouting requests open a support ticket or something
Not surprised considering it's fairly unhinged.
I'm not arguing.
I'm just trying to understand why grok behaved differently.
Yeah I'm just saying I highly doubt that they are rerouting requests considering the backlash they got last time
Thats just how LLMS work
They usually know who made them I think.
Like they don't know the exact version but they get an idea from their internal policies.
Okay. Wait.
Haha
@dry kestrel
That's not how it works. See my conversation and you'll understand.
yeah ik sometimes the model picks up things from its training
Just see it.
Not sometimes.
It's always in there.
You just have to pry it out.
You both are right
"I should use the provided information about pplx ai to answer" aka the sysprompt
And if you keep poking it you just might get the actual model/company that developed it
Their internal policies probably mention their company.
I had no idea anthropic gives out their system instructions just like that.
they know that even if they wont provide it, people will find out
grok does this as well
Send this.
I read about it but never found it.
So do you make the ai spit it out or is there another way?
jail breaking it
Thanks 🤝
The election related sysprompt is funny lol
yeah because sometimes the model doesnt search the internet
Are you based in the US as well if i may ask @dry kestrel
In grok?
Claude
They are being cautious with that to state trump won
They don't want sanctions
Grok instructions are pretty simple actually.
There's literally nothing in there.
Fr
Wonder how it still manages to parse all that and then the user query
Like how much waste
Might have reserved context limit.
And i couldn't really remove the instructions it did not claim to be claude instead it explained prplz futures without mentioning perplexity
Iamma sleep
Gn
At least i can take a nap
but can you...
do share!
Nah, that is not part of the plan
If I can't push a qtip as far in as is comfortable, something must be wrong
Yeah but I just end up scratching too much.
You also have an issue when the qtips are to fragile and not "good enough"?
They need to do a full flaming deep clean - every time!
Who? Doc?
No no, the qtips. The ones you use yourselfe!
Wdym by full flaming deep clean?
anyone read through the ama from yesterday?
Yuh
anything note worthy?
Follow up questions and better follow up editing for labs, comet windows beta next week and comet for android is in the works, aiming for fall apparently?
beta for all
like open beta
orr
Probably the same system like right now
so
like 10 people get it
🗿
am i wrong 😭
good idea
***
### TL;DR
The image is a 3x3 grid of nine squares, each containing a distinct picture, all rendered in a consistent, stylized artistic fashion characterized by an overlay of fine, concentric circles. The collection features recurring subjects—three dogs, three roosters, and two orange socks—along with a single image of a boot in a snowy landscape, suggesting it might be a visual puzzle or a sophisticated CAPTCHA test.
***
This image presents a structured collection of nine individual pictures arranged in a three-by-three grid. A unifying aesthetic is applied across all squares, marked by a textured, almost hypnotic, overlay of concentric, circular line patterns. This treatment gives the images a dreamlike or digitally rendered quality rather than that of straightforward photographs.
The subjects within the grid can be categorized as follows:
* **Roosters:** There are three distinct depictions of roosters. One shows a full-bodied rooster in profile, another is a close-up on the head of a reddish-brown rooster, and the third is a headshot of a rooster with striking black and grey plumage.
* **Dogs:** The grid contains three images of dogs, all appearing to be headshots. One is a light grey puppy, another a similar grey dog with a mark on its forehead, and the third is a black dog, possibly a Labrador retriever.
* **Socks:** Two squares feature a single, vibrant orange sock set against a white, textured background.
* **Footwear:** A single square in the bottom-left corner displays a dark-colored boot, resting in what appears to be a snowy or icy environment with mountains in the background.
The repetition of certain categories (dogs, roosters, socks) while including a unique item (the boot) strongly implies that this is a form of visual selection test, commonly used in applications like CAPTCHA to verify that the user is human.
May I ask, what is the origin of this interesting image? Knowing its source might provide fu
rther context about its purpose.```
Seems like hCaptcha is dead now
claude desktop + MCPs = comet
soon will work in the browser
idk i’m cooking something cause i wanted the same use case without a desktop app or new browser
“close my useless tabs”
“summarise the yt video i got open”
Not really because they can always just make the challenges harder.
And most of the time the challenge isn’t worth the cost of one AI call - the whole point is to make it annoying or cost money to an attacker to bypass.
uh oh
💀💀💀💀
stfu @algoflows
Why do they constantly keep making terrible decisions
Algoflows gotta ruin everything
Could've literally just used it like the rest of us
But had to be the party pooper
I pray that GPT-5 brings 1m context and native audio/video input
Not coming to plus anyway 🫣
I don't want to be stuck paying for Gemini
maybe Logan hired him to post that
wdym?
man but that Epstein files accuse is no turning back
I agree with Theo for the first time
Sure, do they release the switch 2 oled in 2026 ?
100%
Standard cash grab plans
a LCD screen in 2025 is a disgusting move
No one but Nintendo lol
Is there any plan or way to add friends or family members to your Perplexity account?
Nope
Tranks
happens when google is the best for 1 second lol
even free tier will get gpt-5
Hello. From South Korea
2.5 pro 05-06 wins for a day and we get this
so far it seems they're going the gemini route
Annyeong!
yo
google being google lol
bet its because they're shutting down ai studio and adding it to the list of things they shut down
the thing is I need 1m context and native audio/video input to replace AI Studio
Too real
I'd still rather pay for GPT-5 API than Google as long as pricing is similar though 😤
4o being native audio already should give a good idea how it'll work
You reckon that'll be the case?
Because only gemini and r1 having similar pricing rn
Gpt and claude having too expensive pricing
Or is it because just they're reasoning models
last I checked they don't allow direct file uploads for audio even in the API 😭
its going to have more efficient reasoning and its adaptive
so that should help it out
Nice
ah yeah just live voice lol
Nothing beats advanced voice on chatgpt
Oh yeah
its incredibly realistic
you can use it now
2.5 Flash Thinking Native Audio in AI Studio is probably much smarter
There's an app?
but the back end is pretty meh
Interesting
powered by an unknown llama based model
Does it have memories
kind of according to them
Will look it up cheers
it works as it wants to continue previous conversations
but i didnt test the depth of its memory
Could you share a link
I thought they need a custom model that outputs tokens to control the voice
just click preview and make an account
Okay
its really barebones tho
Haven't visited in a very long time
you dont even see previous conversations...
Just used it straight outta here
I don't think sesame has made significant improvements since launch?
Is there a rate limit
ah to add to this
they'll remove the other models
so they have to provide good pricing otherwise they gimp their entire lineup lol
30 minutes
I don't think they will remove all models on the API
Per day or just session limits
erh session
I expect pricing similar to Claude Sonnet
however they might stick with mini and pro lineup, with GPT-5 mini, GPT-5, and GPT-5 pro
and GPT-5 mini might be what Sam meant for unlimited free users
How large is the model, are the gpu requirements high
not very
Cheers
Dang
was a bait and switch
2 months
people expected the demo to be open sourced, but they only shared the voice model portion
Wow feels like forever ago
Yeah that
not the secret sauce that makes their "conversational" conversations possible
that demo is still crazy
Haha
unless i was tryna fuck with it
Smarter than the standard hunan
i would be fucked
Exactly what i was thinking
not good enough for 90% of people
they won't be going to a separate platform just for voice
is sonnet 4 any good on pplx
Sonnet 4 thinking is
Really amazing
I use it even if i have to wait a bit for smaller tasks
Faster than switching between models
Self glaze too high these days on their own platform
they don't cap thinking to 4k unlike t3 chat
Pls do
exclusively using 2.5 has changed me
Lol
also been busy asf
Fair enough
Unless it was already curated
It's not
intelligence levels so basically reasoning length
and most likely context size
free gets limited reasoning and limited context
plus full reasoning and full context
pro gets the rumored parallel reasoning or even longer reasoning like o1 pro and double context
maybe 2m tokens
What's parallel reasoning
would closely match gemini's setup
that would never happen unless GPT-5 is cheap
they say we have double token limits every 7 months
gemini deepthink
i think cost would be tied to reasoning length and context
its basically running a few reasoning threads at the same time
and then it decides which answer is the best
it probably is able to look at parts of the reasoning to help it decide which output is better
OpenAI has it, o1-pro
Google has yet to release, 2.5 Pro Deepthink
xAI teased it but never released, Grok 3 Big Brain
Anthropic uses it for all their benchmarks but doesn't offer it at all 💀
Interesting
o1 pro is just very long reasoning iirc
That's what i thought
Anthropic is such a weirdo lol
anthropic still needs to write more ai safety blogs
Lmao bro
and go on talk shows to discuss the dangers of ai
I saw the sysprompt for claude 4 opus
A damn thesis
Best ai haters that make the best ai for devs
sometimes i wonder if they're trying to get VC money without competing against other ai companies
Y'all know if the 2.5 pro on Perplexity is today's release?
And the true competition to openai imo
Won't likely be
We haven't even received r1 0528
Been out for so long alr
That one they would have to fine tune tho
2.5 is just plug and play 😭
I thought it was the same price as the previous one, but I just realized I was looking at the AI studio price
AI Studio price is the API price
Oh no price change for iterations huh
Idk I'm shooting in the air as to why no new 2.5 😅
After redeeming a 1-year promotional code and confirming the code sent to my email, I received this error. Does anyone know what it is?
Oof
?
I got the memory today so got a lil improvement bump from that already
I thought that was parallel reasoning lol
oh lmao
Just realised it was just begging for feedback lol
jus use it on ai studio 😭
looks like they don't talk about it, but I still think o1-pro is parallel compute. I don't think telling a model to always reason 10x harder is effective
does google let u use this via api
oo they do
imma throw this on a discord bot
why not use 2.5 Flash
wait im an idiot
assumed it would have low ass limits
500 isnt bad tho
2.0 flash gives u 1500
ig ill continue using that
tried editing this instead of the env 😭
the good old days when Google was still being generous 😔
yes
the limits are insane 😭
5 per day for free
cant complain for free tho ig
or you could use it directly on AI Studio
i wanna vc with it on discord
🗿
theres no way there isnt an easier way to get this to work
yippe it works
nvm sadjoe
Gemini 2.5 Pro 0605 has been released, when is it expected to be updated in Perplexity?
@vague pagoda can you please check with the team 😅
Next year same month
how is it :??

Yea but usually the output is bigger so wish perplexity could also output 32k
Or file
It would have been nice if perplexity could output file that consist of multiple 4k tokens
btw the new 2.5 pro is absolutely dog at coding
i always ask to run a complex simulation of a 3D solar system
Ahhhh
and by far it's the worst
Use Opus pocus for coding no one else
They said it would be better, guess not
i had to take 5 attempts instead of 1
just to make it create a decent version of it
Lmao, the newest version is currently available in the api only?
Or in ai studio and gemini platform
Did you try any other prompt?
it's the one i have always used
this was the 1st result
Usually one test with multiple prompts
Wat da hell is dis
So bad lol
What in the black hole is this
Can i see the prompt
Solar system according to gemini 2.5 pro hahah
Looking into it!
Is this a trap hammy?
hahhahha did i ruin it?
thanks for the suggestion!
While you are at it, could you also create a poll generating channel, there has been so many requests for it from butcher, as well as few other , thanks in advance
Saw those! Looking into it - what would you guys want to use the channel for mostly? Since a channel for polls about anything could get messy fast 😅
Give me few mins, better answered from the one using it the most. @distant aurora @left stream @valid charm
As per me, mostly it would be for opinion and voting about different ai models
Something like this ..
As it gets messy since soo many community people feel the need to discuss this for suggestions regarding which model is best to use as per other community members feedback
Hi gang, its me from yesterday (the guy with the billing issue) just wanted to say a big thank you to everyone that helped me out! It's all been sorted now ❤️
pplx ofc?
He left :c
Supposed to be the sun
Which one the total black circle or the one inside those orbits?
The grey one
Okayy
What the...😂😂
got the 1 year of pro through optus, I've used perplexity free before and it was good but im a bit confused why they're giving out free subscriptions so much
also where can i find the images perplexity uses for their marketing
they'll make good wallpapers for sure
using a search tool to code is interesting
Same. What would I use Perplexity for? Same things as other LLM’s like ChatGPT and Gemini?
ill probably use it to study coding stuff
Yesss
That’s smart. I guess I should pick up something new to learn
To gather public opinion on features and negotiations without making it spammy or messy. We could put a big cooldown limit to compensate for the possibility of spam/misuse.
It would also be easier to find out what number of people are facing certain bugs/issues.
Keep an opening and expiring date.
Will flux kontext image editing be implemented into perplexity?
Might help perplexity better with adapting their product wrt the people's needs as well.
@cobalt vale @dry kestrel Elon gone rogue
Lmfao
Wild.
@timber niche you see this?
im not american so i dont care tbh hearing about trump is infuriating
And him imposing some shitty law may not affect you but did affect a lot of students around the world too uk
Ok thats just your opinion
americans are cool, and everything made in america is usually good
but i do not care about the government
because it changes
@dry kestrel will this get me banned from grok's server?
America affects global geopolitics with their shitty interference everywhere
Peak hallucination.
And subsequently economics
if you tell it "do not hallucinate or i will never speak to you" does it work?
Grok ain't that soft.
real
Lmao.
Lol
i tell chatgpt to not hallucinate, i put it 3 times in three different ways of writing and it usually works
Grok gonna be like : I just need the mouth, not the talking.
how good is perplexity at remembering convo long term like over a week or so?
which model should i use
Context limit is 32k tokens.
Depends on length of uploads/conversation.
I think Claude might be the best model on perplexity rn.
The reasoning one.
based, which ones best for image generation like how homlander does it
home
Some gpt thing.
Not through pplx
He does both.
Need thicc gpu
Not really
can it be done on a macbook? i can run windows through parallels desktop
m3 pro 18gb
4050 is thicc for me. I use a 1650.
H100 would be obese.
Real goat was 1070ti back then
i think a 4060 setup is 1300 australian bucks here if you go real cheap
But i do think without an nvidia gpu all these stuff wont work
Afaik
Almost every thing needs cuda
I want nvidia to eat Qualcomm in the mobile business.
thats real gay, i guess that'd be a good thing cuz pride month, that's really straight
The adreno monopoly kinda pisses me off.
Last they made a mobile chip, it was for the nvidia shield
Yeah
It's what it is
Grok app has been buggy lately. Shit keeps getting stuck and wasting my prompts.
The hallucination is a different problem in itself.
guys, fow what are you using Labs?
Only today I saw pplx sonnet hallucinating
Flexing an extra feature as a perplexity user, lol.
When i asked for current llms offering parallel reasoning
which pplx model is good for image generation?
Img1
Gpt img
Yeah
is gpt 4.1 not available on chatgpt plus?
ok it is
4.5 is also available
aren't these just based off chatgpt 4o? cuz isnt that the latest model
4.1 is latest
I think 4.1 is based on 4.5
4o is not good at image generation at all
4o doesn't make those images
Either img does or dall e
Those are the gpt models on chat got plus
how do i choose img on pplx?
You can't even choose
Have to be on web
And have pro sub
Then can change in settings
Yes
and chat model? any? for image generation to work in tandem with
hmm the chat says it cant generate images
I got 2 yolks in one egg.
based, pretty lucky
they could have been twins though (im tryna make you sad, ik unfertilized eggs cant hatch)
You got a nice quote written on ur profile, can't remember quite well where hv I seen it before
think about it, you'll find it within yourself
it is from the least read work of a very famous writer
why does it not work i chose image 1
oh im stupid
It's called post nut clarity.
nastenka
real
im fr gonna name my first born son alyosha
Dostoevsky and Kafka be the two most depressed bros of all time.
nah
once again 2.5 pro 0605 is the best at coding
Nice
i mean wtf
planets literally perpendicular to the orbits or viceversa
and the texture on the sun is buggy
White nights
is this using the api
yep
webapp
the main guy was dumb asf in that one
Where are the planets?👀
incredible
Ohhh
they are lies
planets dont exist
they are lies by BigSKY
these things here
accurate
Ohhhkay
how much did this cost
They're celebrating pride month.
can you tell me why pplx keeps saying "failed to create image"
It only works in web version.
prob benchmarks are just ass made atm
a steal
im on chrome dude
web
keeps saying that
Send thread
Or your prompt
unfortunately im too stupid to know where to "send thread"
1girl, close-up, dynamic low-angle shot, reaching toward viewer, white hair in twin braids, intricate red eye makeup, bright golden eyes, glossy red lips, fantasy gothic armor with ornate metallic designs, glowing soft light, hyper-detailed, ultra-realistic anime style, reflective surface, cinematic lighting, intense expression, dramatic depth of field, high contrast, illustration by Wlop and Artgerm, 4K resolution
this is what i used
Fr. Gemini intelligence doesn't align with benchmarks.
Here.
Don't bother. Gemini benchmarks are made up.
Can you try this with grok?
The thinking variant?
i don't have grok
Give prompt. I'll send result. Then you run it in whatever you are running it in.
In what language?
english?
thats just words
Make a very complex simulation of the solar system
Thanks
Though I might need it to instruct to complete the entire code
Yup last few days
Remaining
Nah man i was not paying for it, it was a shared account from a friends on student discount
But yeah it was pretty good otherwise, though i mostly use you.com o3 for most of my tasks
do u like this solar system?

and 2.5 flash was said to be only 2nd to 2.5 pro
True
Make a very complex simulation of the solar system
Cant see
What's the result?
I am not sure I ran on android
My lapi isn't with me
Can you try running the python code in an ide
is memory only rolled out in the U.S ?
Don't think so
Hey mat
Well i have it
woo
Cat be cooking fishes.
Matpat
No theories here
It's a cat theory
You mean cat recipe?
I found an ear drop that actually works.
Thank god you finally made it
Slept well?
For a moment I thought I might dig my ear to deafness.
Yeah. Got sorted all my exams yesterday.
I woke up at like 12.
Use it for how many weeks billy?
I tried it just 2 applications for now.
The itch is almost gone already.
I must continue it for a week though I think.
To avoid relapse.
So a day.
I got a lot of relief from just 2 doses.
I hope you will get fully recovered soon billy!!!
It stings the first time though.
But I'm positive that I should recover from this.
Ear sorted.
Now time to fix the cough.
Wbu? Hyd?
I heard we have a covid case in Hyderabad
Better mask up I guess.
Ai hype is wild
Is this different? https://www.coursera.org/specializations/finance-and-financial-markets
This is specialization course! Valuable course for your career
What was financial aid about?
You will get a career certification, only if you complete all 10 courses down there
If the course is 4k - 5k then you will get a 90% discount
Like I applied financial aid for 2.5k and they accepted it and I paid only 299 for the course and got certification
Yeahhh... but in recent times lots of people misused this feature and coursera completely removed this
Yesterday I paid 3k for a course to get my career certificate
Is there a way to set the app language on Android? It's possible on my iPad but I can't seem to find the setting on my Android phone (not the AI content language but only the app language)?
So you mean you won't get that discount now even if you complete the course?
Afternoon
Yeah... No more discounts! You can't complete the course billy, because per course they will set up to 4-6 weeks content per each week you will be having assignments. For attempting assignments you need to pay the course.
So unless you pay for the course they won't allow you for the assignments!
So 100% completion = certificate
But it will boost your chances in interviews billy! On your resume you can official put certified by Google
In any JD(Job Description), they will clearly mention you to have few mandatory certifications.
Yeah

@pseudo ferry health is expensive
Isn't course era also offering a monthly sub
@pseudo ferry
Health and education have the money rot.
Versión: gemini-2.5-pro-preview-06-05
hey guy's do anyone expect what the f is happning right now?
both fighting like cats
also the share are decling like shhhiit
really weird time guys
Its all about the spending bill (especially SALT cap deduction)
Trump promised to cut gov spending before the election but now he wants to spend more (not much as Biden admin but still)
Elon just expected too much from him. Still it's a hard choice between doing what is right and being hated since spending cut mean ppl suffer
popcorn time is happening

Why?
Nothing wrong with that
ptsd
oof epic fail
i wish it should be good for reasoning
earlier gemini too bad for reasoning like 1 week agi
whats the difference between thinking model and reasoning model
thinking model thinks before giving answer
basically questions himself
reasoning is related to aptitude
What is your experience?
then which is good one to use
Reasoning shows the thinking
gemini bad on questions related to directions
Is your old acc gone for good btw
and passage based questions
shoutout to the pop up in NYC yesterday made a nice lil espresso tonic
not gone, it got restrictions like i can't message
Dang
Dayummm
@vague pagoda
damm the packaging looks awesome
its on the verge of permanent suspension that's why i left it for gud
Mfmfmgmgfgg
You really had to come out saying that lmao
I’d love some coffee rn
Thinking, in the context of models, refers to the overall mental process or activity that a model engages in to produce a response. Reasoning, on the other hand, is a specific type of thinking that involves using logic and evidence to arrive at a conclusion or decision.
We always do

Yall seen the perplexity pro 1 year thingie on galaxy store
👀👀👀👀
Where exactly in it
Real
you just download it from galaxy store directly and signup and you'll get the full year
funnily enough i got it by accident i didn't even know
I don't even have pplx on my store
Dang
and remove your sim
Oh damn
remove the sim from your phone and start a vpn and you'll be fine
pretty sure that's all samsung store logs
Thanks will try
sure
You're orange now!!
Thank you so much
Lmao bug sniper
Oh explorer
Mods fixed his roles
So the Samsung free 1 year pro works
Now i have 100 labs requests per day woohoo
Lol
Congrats
will that stack up on my existing student referrals pro (which i have for 2 years)?
Did it on my galaxy tablet
Nope has to be a new acc im guessing
no clue. based on common patterns it would probably need a new account. i haven't checked that.
Let me give it a try and let you know
Yeah id you're already a pro idk if they want existing customers to get pro
Don't make sense logically
sure
hmmm yeah
if that's the case, i could get the 1 year pro and give it to my brother
it won't probably work but if it works great
i have an account with some referrals
if you try do lemme know xD
Wonder if they're account linked or device linked
If account linked it'll be easier to abuse
i think they might let people abuse it for a day or two then suspend the offer
just for the hell of it
lmao
Superbat
today i asked a question and wanted to prepare something in pplx
so result was
R1 > Claude 4 > o4 mini
its secret


