#š¬āgeneral
1 messages Ā· Page 372 of 1
In term of security and u have more sources to choose when using pro search
premium data
Nah AI is not of 2022 rn
@twin grail cat can you overthink for me
Yeah it does fr
Is that a bug or something
Not a bug I think
It has message limit
..
See netflix from a long time analysed users from AI and recommended them the things
So we can say the billionaires had what we have today like 10 years ago
So what the billionaires have now?
How advanced
....
Also sometimes it says here is your file but nothing comes
Contact the services
Email them
How do I set up a profile so that Perplexity is aware of my information, including my job, etc., so I don't have to notify Perplexity every time I chat?
Nah bro figure it out yourself
Life shall not be too easy
Alr
for my queries in perplexity gpt5 thinking + web search is giving me more reliable answers than deep research of perplexity, this is the only thing which is holding me from long time to shift to perplexity completely
Thatās undeniable
Pplx giving u fast deep research
But the data is 60-70% compared to chatgpt or claude
Itās not as detailed
But the general answer is accurate
Create a space then provide that info to it then any chats you have in that space will have that info just say you want any chats/questions asked in that space to review that info in case its relevant to the info its providing
lol its appearing officially now @twin grail bro
+1
This dude is going to destroy my eye sight
Hate you bro
One more
š§
Avoid excessive messages, all-caps shouting, mass pings, or repeating similar posts. Use channels appropriately.
in perplexity how to change sources in same chat conversation?
Guys rate it out of 10
I don't know if I'd compare that with a 'breakup'... rejecting an application isn't a breakup if there is no relationship even started, and plus even if there was a relationship with an employer is not equatable with a love interest.
Perplexity comet ios and android app when š
True
You should complete the chess game first
Most of those rejection replies are probably just automated and never even passed by human eyes. That's the state of HR today... smh
I propose that people try visiting the headquarters and talking to the executives directly by just walking into their office with a resume to show your dedication. š
<gets escorted out by security>
Not possible
Are false positives always so crippling and bad? I literally refunded my purchase immediately, can't believe it...
Back on GPT... š
Is it possible with spaces by editing the space documents?
It depends on the model, but I think the API version of GPT-5 that PPLX uses is prone to false positives quite a bit. It's not comparable to the normal GPT-5 version.
(The versioning of GPT-5 is extremely confusing, even though they tried to simplify it, it seems)
I never even got to chose a model... It just started arguing with me over a funny message I wanted to send a friend as a test and that was that...
I guess better luck next time, when a GPT mess up might make me give it a second shot in like a decade or whatever...
Ah, switching the model is here - it just defaults to Auto:
Sure, but the default does not work... That's all I need to know...
I've NEVER experienced an AI instead of trying to help starting to full on argue with me... š
Hillarious if it weren't so incompetent
sorry it's full
Gorg ā¤ļø
Heyo š
Did you buy it?
For real?
I mean it's enterprise
so...
The normal pro cost twice
325 for enterprise i think it's suitable
is that for the year
for freeeeee?
yes
normal pro ?
that will last me to december tbf and im happy with that, just whilst income isnt steady
i appreciate the work you guys put in
you want more?
oh u used all of the refer >
?
I have a few as well tbf
how?
does my friend need to buy it or something
like for student
Nope if u are verified as a student
U can refer ur friend who is also a legit student
i thought it was monthly lol
how do i verify as a student ?
u both get a month free
School email id, student id card
It's the same, i plan on buying it for a year
so
Doesn't make any difference
ah alright
i have mine till next year
should be fine for now
github, notion we got few more
I am the git to the hub
Hey all, any way to get someone about a billing issue? They said discord is the fastest way to get a response
email them
I did but it said discord was faster
yeah it is , contact a mod , dm him
like alex romanov who is online rn
Can't DM him, his DMs are not open @signal hamlet I have a billing issue, can you DM me? I am unable to DM you.
Hi, please contact support@perplexity.ai for further assist
I did already
You will need to request human support about the problem related to billing
ok got it
the billing team should take around 3-14 business day to reply to you
If you are in enterprise or higher plan
u should get a faster reply
I am
around a few days
But donāt snitch on me because Iām in their student program rn
also convert it to standard units
you know how long business days are
Will Perplexity respond on š again? What happened to it?
3-14 business days is wild. I'm on a pro plan
ye mb working days
but it should take around
a week
is average
arigato
There is
i paid for a year of enterprise pro a month or so ago but saw that max came out today, if i want to upgrade to enterprise max does that mean that i have to pay for another full year rn? what about my pro sub payment?
I think it will be refunded to ur account
You are not allowed to share you referrals here
Post that in #šāsharing and delete it here
I have one question. I have the pro version of perplexity and after 2 days of image generating, now it says i have reached the limit. Is this true?
To add the ability to tailor the platform for use with any institution's data, I think Perplexity had to compromise the quality of their core product. I'm noticing a lot more both-sideism and equivocation on data that is clear. It acknowledges this when the user references the actual data. This is not the previous performance, even as recent as early August. Whatever has changed has made the platform less useful and reliable.
is a banana cutter microwave safe
Anything metal isn't microwave safe.
Especially if it sticks to a magnet.
Tf is a banana cutter?
aye
@young hollow damn. Qwen cooked.
80b model. A3B. 250k context.
235b was like 128k context.
Hey! please don't advertise here
yup

67
@tame current hi
hi
Ur from
India
Same
erm~
In india where?
I noticed that gemini 2.5 pro in perplexity is way dumber and worse in performance and benchmarks than in google ai studio why is that
why is gemini a part of trend nowadays
it was useless before right
Its pretty smart and basically free
But in my benchmarks gemini in perplexity scored nearly 3-10
In googles ai studio it was 9/10
Just with 80B Parameters Daym
whats special with the max account?
You watch your credit card
out quicker
unlimited use
priority support
Having Max mode for comet
- instant access to comet
and 2 premium ai model
claude 4.1 opus thinking and o3-pro
This couldnāt be typed in one message?
I respect the hell out of that
are the automation different?
Thanks for doing what you do for the economy
burning discord resources
Longer reasoning
makes sense
Giving Microsoft one meta data at a time
still making the most out of it
Unpopular opinion for the discord today dove Cameron slaps while making ābreakfastā
Try this specialization if you're interested.
Is it Free?
I think it's since it Coursera
Oh it literally says it's Free
Tysm Man Another Resource for My Study
It's from Vanderbilt University | Coursera
what is this?
It's a specialization course from Vanderbilt university on Coursera platform ducky!
Yeah Exactly
It's a combination of 3 man, so you need to complete all the 3 courses to receive your specialization certification from vanderbilt university
Oh the Certificate is Free too?
I thought they were used to be Paid at least
I guess at present you need to pay for it! Before 3 months, I applied for financial aid and got 75% off on all 3 courses
I would personally recommend you to try it because I completed this specialization few months back, learned a ton, and personally built 3 SaaS websites.
I see gotcha
As long the courses are free is enough for me
The certificates are not necessarily yet
It's up to you! You can still learn it on YouTube for Free. Good Luck!
Surely there are Countless Resources for The True Learners
Tysm Again and Good Luck to You as Well
perplexity being perplexity
gym is $50/month?? Mines like $28 max
Im aight
is the artstyle perplexity uses on their ig posts ai or human made?
AI, might be touched by hand at the end though š¤·
They have their own design team, its def not just AI
Ask chatebt
@dry kestrel first time I see you using def
Howdy
Congratulations on learning lingo unc
yeah seems like a bit of both
i love it tho
They have their own style
Fake news
b l i n d
totes
?
?
matt.exe crashed
What
Sounds like matt has a rat problem
It peaked around 2012-2014
shh
Bro you still have that kimi tag
Drop it. Arvin is never gonna adopt it š
Imagine not having the PPLX tag couldn't be me
3 months for training ahh
Got bored of it
No I like Kimi
How come
I mean i like kimi but i didn't expect you to like kimi
Deepeek + Nvidia is my favorite
But sadly it doesn't exist
Btw gpt 5 got 50% cheaper for a week
meow
Dunno if you are aware or not
Quack
Only on openrouter
furry!
Is that an ad by berd????
wdym
š GPT-5 is 50% off on OpenRouter - for a week
- No opt-in needed š
- Discount limits to 20 RPM, standard price after that
- From Sept 17th, 10 AM PST to September 24th, 10 AM PST
Give @OpenAI GPT-5 a try https://t.co/9OhAL1iwSY
Yea ikkkkkk
I waz implying that you are advertising for open router cuz "only on open router"
I would never
gyms here are expensive AF
40$
holy moly
Why doesn't the Perplexity app on Mac let you choose models?
Hi. I opened acc with edu mail it gave me pro
Is it pro till end of the university or it is free for only 1 month
I can choose a model on mac
i noticed that 4.1 claude is the fastest ai model is faster than sonar and better than sonar
What do you click after you click the gear icon (Settings) at the bottom left?
@dry kestrel @short narwhal
berd's hat took me like 30 tries
it kept failing
Wow thatās us berd
Why is my hat so big
the other ones are much worse
lmaoo
try out seedream!
i think the jinx bucket is the coolest part
DONT EVEN MENTION SEEDREAM
It;'s been pretty good for me so far
You don't have to do that. I click on this button.
try it out https://dreamina.capcut.com/ai-tool/home
Yeah I'd stick to using the native version given that Perplexity doesnt let you change the aspect ratio right now
@gloomy cipher who are you
Duck with Confabulation
I miss fried duck
i found out that the directory word for ai hallucination isonfabulation
Cannibalism?
na cannibies is only allowed for me not him
Cannibis?
its like a cute term
Ducks are birds
Ducks are fish
weedies
Fbi open up
U are definitely high
What happened to the once intelligent berd
fish cant fly
thats all there is to you brain
Lol
Fish
Is this real
yea
bro what happened to youtube
What has happened to you all
Meet the fish that can fly. Flying fish glide through the air to avoid predators.
#shorts #fish #flyingfish #fishing #animal #animals #animalfacts #animaleducation #nature #animalplanet
huh
fish
Nice grammar
It's a joke
he proved his point
Jokes on you I don't use tables
are furries real?
Oh my god are you actually drunk bro
I think berd is drunk
thats so disrespectful that's like asking me if ducks are real
@short narwhal
birds can't drink
Literally can
He died shortly after
You didn't
My dear ol friend
You're here before that
give me your acc before that happens
https://Plex.it !!!
no
What
Mat is also dunzo
Woah thanks š«¶
What the fuck happened to you all
Fish
Like what
The finland penguins have gotten him
You ruined your praise by sharing random pplx link
we are taking over the north
go to sleep
Pplx dont know ublike that lil bro
heyyy thats my time
donāt doubt my and PPLXās bond
Meh
also Friday night?
makes no sense
Huh
wait
Wdym
says 1:37 friday night
It's Friday here
Because itās 1:37am on a Friday
Dumbass
at most its friday morning
Exactly
looking at finland time
Broo
false
I ain't gonna argue with a crow
Literally is
Ok habibi
Crow
idk about crows but i wouldnt mind being a raven
Crow
Hey everyone I just signed up for a year of pro, looking to switch away from chatGPT š
hey everyone , can someone send me a comet invite?
Yeah me too. I don't even know what it is but it'll be cool if I can get an invitation
I just signed up for a whole year of pro
You're a champion, thank you
What is a good local easy to install .exe voice cloner
I have a question about how to active the study mode
wtf with gemini-2.5-pro
@dry kestrel renewal today š
At least demon slayer movie was good asf
@dry kestrel what do you think about gemini ultra thus far
Tempted to get it too ngl
It doesnāt make any difference
As a gemini ultra user before
U get more credit for video generation
is it possible or any hacks to make the assistant screenshot websites --> keep those screenshots for perplexity/apps
This is really the best feature of a gemini subscription
Theyāre failures then
And 30tb storage
which will take centuries to fill because google drive limits your upload lol
You could zip everything up and upload one big file:
Each user can upload and copy 750 GB to Drive within 24 hours. Users who reach the 750-GB limit can't upload or copy more files until the limit refreshes. The limit refreshes within 24 hours.
You can upload and synchronize files up to 5 TB in size. Only the first file that breaks the limit completes uploading.
So you can upload 5TB per day
But even at 750 GB per day, that would only take 40 days to reach 30 TB
That's awesome!
Why she blocked me š¤§š
give the screenshot to 5 thinking
Do anyone have answer to the question in this screenshot?
Got the 1 year free subscription, ts is insane value why would they do thatš
Howās everyone experience with this AI, is the pro version at all better than free ChatGPT or nah
Free ChatGPT is not that great in my opinion... you only get like 10 queries per 5 hours with the real GPT-5 model, and only 1 dedicated thinking query per day (with thinking toggled on, though you can always use your 10 queries to do some thinking too).
So yeah, definitely worth it to use the reasoning models in perplexity if your alternative is ChatGPT free.
But in terms of what AI to use, is it worth just ditching gpt until the subscription is over
I mean, I still spread out queries between different things to compare (occasionally one will find some points that others haven't considered)... it just takes a lot of time to go through them all to do that. So by all means, keep using your ChatGPT free queries when you're able, but Perplexity Pro compared to just the free stuff on OpenAI is definitely worth it. Also keep in mind that Google AI Studio gives lots of free queries for Gemini 2.5 Pro, and you can use some free queries on other services too.
Which is that?
maybe this one
Umm thatās sources
The next one to it is where you choose model
he asking that perplexity able to do homework for him
Thank you I never used perplexity before I just downloaded it for school
I got a free year
Lemme try this assignment see if it writes this paper and Iāll lyk
Not worth it unless you need video gen
I canceled my sub and now I'm back to Pro
Dear Perplexity Pro Support Team,
I am writing to report an issue I'm experiencing with the image generation feature in Perplexity Pro.
When I try to generate an image using a face as a reference, the output is a completely different image that doesn't match my prompt. I have tried using various prompts and settings, but the problem persists.
Could you please look into this issue and provide some guidance on how to fix it? I am Pro user
Thank you for your assistance.
Best regards,
Vijay
Username -vjbisht81511327
I don't know what to do with 2Tb
#1140622008086970420 Would be the right place.
India
@limpid storm india
My pc docs only 200 Gb, and phone's files 400 Gb, thats all my stuff
?
@limpid storm When I try to generate an image using a face as a reference, the output is a completely different image that doesn't match my prompt. I have tried using various prompts and settings, but the problem persists.
try using GPT-Image 1
under settings
@limpid storm Where to find gpt image 1 option
I don't know. @dry kestrel
Do you happen to know, Berd?
I don't see that in options either.
@limpid storm where to find the option gpt image 1 perplexity pro I can't find the option
I just told you I don't know. š
@limpid storm When I try to generate an image using a face as a reference, the output is a completely different image that doesn't match my prompt. I have tried using various prompts and settings, but the problem persists.
@limpid storm When I try to generate an image using a face as a reference, the output is a completely different image that doesn't match my prompt. I have tried using various prompts and settings, but the problem persists.
I m a perplexity pro user
@proven schooner Hi, please tell me the platform you are using perplexity
š
@twin grail When I try to generate an image using a face as a reference, the output is a completely different image that doesn't match my prompt. I have tried using various prompts and settings, but the problem persists. I m pro user
Again, please tell me your platform and the model you are using for image generation
@twin grail android
Okay so go to chrome on your phone
Log into perplexity
You should see something like this
On the top left
U will see a setting
Press on it
You would see smt like this
Press on the setting button ( next to your profile to the right )
@twin grail ok
Click on Preferences
Scroll down and you should see image generation model
After selecting the model
U can know close perplexity app and reopen it again
Then try again with your prompt.
@twin grail Thanks
In the app or website. Can we default to always include social posts. Is there such preference in setting pls?
You means the new/discover setting ?
@carmine sandal
@vague pagoda @blissful raptor Feedback on Comet Study feature
Please include the ability to upload multiple lecture-based pdf and study materials at once, and organize them into study spaces. Add a one-click option to generate quizzes and flashcards with customizable settings such as number of questions, difficulty level (easy, medium, hard), and personalization options. It would also be great if the feature could replicate all aspects of Google NotebookLM for seamless note management.
Feedback on Comet Study feature
Please include the ability to upload multiple lecture-based study materials at once, and organize them into study spaces. Add a one-click option to generate quizzes and flashcards with customizable settings such as number of questions, difficulty level (easy, medium, hard), and personalization options. It would also be great if the feature could replicate all aspects of Google NotebookLM for seamless note management.
@carmine sandal
@vague pagoda @blissful raptor
you should make a feed back in #1393681130842226698
This doesn't notify them
@young hollow
I mean in the app or webpage , can I default to include social media search
I donāt get your point
U mean using pro search only on social media ?
šØĀ STOP SCROLLING! The Craziest AI Drops of the Week Just Landed!Ā š¤š±
What if the future of business, creativity, and workflow was already hereāand itās fun, fast, and mind-blowingly accessible?
This week, AI launches are breaking the internet and flipping expectations. The buzz is real, and everyone from creators to founders is j...
Hey everyone! New here, Iām not getting replies from perplexity app (iOS26.0/Version 2.250911.0) any idea of how to fix it? After prompt is given it stays loading


Hey! can you record a video ?
Sure!
And make sure the app is updated to the latest version
I gave 2 simple example prompts and still not working
from what i see maybe it's the problem with the internet ?
Try to reopen the app
and make sure you are up-to-date
Already tried log out/log in, different wifi/cellular data, reopening app, updating app
changing model ?
Yep
u can try reinstalling the app
Gpt, Sonar and Gemini
Still nothing
last method
restart ur iphone
also remember to close ur app
One sec
Still nothing
hello
hi
Forgot to mention, this only happens on a spaces thread, not in a common/new chat
She is Pretty I Have Seen Such Content on Instagram as Well
uhh
i think u can make like a new chat in that thread
I have a thread dedicated to personal stuff and that specific chat is the one that is not working
soo
I think
that's the problem
when ur chat is too long
and it exceed the context limit
which making the ai unable to reply
i think u can fix this by adding instruction to the thread
with ur informations on it ( personal stuffs etc )
tell it to remember ur info
so whenever u have a new chat in that thread
it should be good
Aight, lemme try that. Thanks for the help!
Perplexity UI sucks
I have 17,000 worth perplexity subscription provided by Indian internet service providing company Airtel. And it's chatgpt model is not better than chatgpt free one
After having this subscriptionI I still use chat GPT and other Ai
Uhh what that even do with UI ?
You have said that pay more money to get better UI
I mean the model in pplx is completely normal, same juice as plus plan
It just the RAG and low temp keeping it to answer accurate
So use the reasoning version if u want the best output
Which u didn't get the joke
Ohh ok
what do you call a better UI
Well I want to find Gamertag for my Xbox I have tried almost every AI and every model which has in perplexity Pro
is that a ui problem?
Uhh that has nothing to do with the ui tho ?
U can change the image generation model by going into the setting
they have the new nano banana model from google
.guyzz
imagine Samās reaction when he finds out GPUs are burning up to handle requests like these
Ikr lol
I gotta see all the lore behind ninjago with thinking pro
this showed up
Happy studying
Hopefully they add it to ios and android ngl
@dry kestrel you can control the thinking time now????
Im slow on the news lol
Cool asf
i mean sure
Bullshit
Is there's any significant difference between heavy and pro?
Just because of currency conversion from different countries is crazy
No clue havenāt tested out heavy yet
What questions should I ask it
sad
@left stream
test this
A fit 68 year old male presents with a PSA of 18ng/ml. Rectal examination demonstrates possible extracapsular invasion. 30% of biopsy cores are positive for adenocarcinoma, Gleason 4+4=8. i. What risk group does this patient fall into? Justify your answer.
Disable web search
Lol already tested that one
I know who you got it from lol
we all know who he is
i got it since the day i got grok 4 heavy
it was long time ago
but grok 4 heavy isn't worth it
I wonder if i got it when o3 pro released or gpt 5 pro
O3 got it right
Did it get it wrong?
nah grok 4 got it wrong
Ye
Damn
total failure
It can think for 4-5 hours?
better buy chatgpt pro
idk about this but it's true that the thinking time is very long
around 30-40 mins
Thats good
Hopefully they release a new model soon one day
What's the answer?
And which model got it right?
@left stream
Testing with both
Mb got tempted by food rn
Imagine saying gpt-5 thinking (heavy) vs gpt-5 thinking pro
Iāll find the old responses I had all of the models already answer the questions
Based on the NCCN Guidelines for Prostate Cancer risk stratification criteria, this patient falls into the very-high-risk group.
Justification:
According to the NCCN risk stratification table, the very-high-risk group is defined as having at least two of the following:
cT3-cT4
Grade Group 4 or 5
PSA ā„40 ng/mL
This patient meets two of these three criteria:
Clinical Stage cT3
The rectal examination finding of "possible extracapsular invasion" indicates at least cT3 disease. Extracapsular extension by definition means the tumor extends beyond the prostatic capsule, which corresponds to T3 staging.
Grade Group 4
The biopsy shows Gleason 4+4=8, which corresponds to Grade Group 4. According to the ISUP grading system, Grade Group 4 includes Gleason scores of 4+4=8, 3+5=8, and 5+3=8.
PSA Level
The patient's PSA is 18 ng/ml, which is below the very-high-risk threshold of ā„40 ng/mL.
Additional Considerations:
The 30% positive biopsy cores (>50% threshold for unfavorable intermediate-risk) further supports aggressive disease characteristics
While the patient would qualify for high-risk based on having either cT3 disease OR Grade Group 4 alone, the presence of both features elevates him to very-high-risk
This risk stratification has important treatment implications, as very-high-risk patients typically require multimodal therapy with long-term androgen deprivation therapy (18-36 months) combined with radiation therapy
from claude 4.1 opus
in pplx
Got same answer in grok.
Iām surprised thinking heavy is still thinking
grok 4 heavy got this wrong
They changed coding interface dor chatgpt???
i think they did
So heavy is worse than normal?
Share response.
Justification
According to the NCCN Guidelines for Prostate Cancer (version 2.2025), risk groups for clinically localized disease are defined on PROS-2A/PROS-8A as follows (paraphrased for relevance; see PROS-H for additional principles of risk stratification):
High risk: No very high-risk features and exactly one of the following high-risk features:
cT3a, or
Grade Group 4 or 5, or
PSA >20 ng/mL.
Very high risk: At least one of the following very high-risk features:
cT3bācT4, or
Primary Gleason pattern 5, or
2 or 3 high-risk features (from the list above), or
4 biopsy cores with Grade Group 4 or 5.
This patient qualifies for very high risk due to 2 high-risk features (cT3a based on DRE showing extracapsular invasion and Grade Group 4 based on Gleason 4+4=8). PSA of 18 ng/mL is <20 ng/mL (not a high-risk feature). The percentage of positive cores (30%) does not affect classification here, as the criterion of 2 high-risk features is independently met (though if assuming a standard 12-core biopsy, 30% approximates 4 cores, which is not >4 and thus not a separate very high-risk qualifier).
2 months ago
fresh from grok 4 heavy
Ngl that question was good @distant aurora you gave me a reason to open my laptop
Its still responding
I have personally run out of ideas to test ai.
Dms
I don't get how this answer is wrong. Could you point it out to me?
u should ask @quaint ravine about this
this field is for him
i'm just reading it with general knowledge
Want the correct answer that o3 gave?
but this is the correct answer
He thought it was a different model
O3, O3 Pro, and GPT-5 Pro got it wrong too
the thing is, u need to know how often it gets it wrong
You said o3 got it right
Yes.
it's not always wrong
GPT 4.1 always got it wrong ...Sonar too iirc
but most other models get it at least 1/5x
Did you even read the answers yourself?
If only they did 1 million context memory 4.1
nope justification is wrong in one of them š
yes ?
gosh Claude platform got even more nerfed
Okay, I'll look again and then I'll ask you.
They always nerf the hell out of it ngl
Opus 4.1 refusing to analyse the PDF
Especially after that chinese dude who did a power move š
Literally was making it run 24/7
i think that's the reason why it was showing used claude 4.0 opus instead on 4.1
wait nm, I forgot...need to use Projects in Claude
Ct3 + grade group 4 = very high? Atleast that's what I understood so far.
Hello everyone, this message was sent automatically by Perplexity, and the server was also joined via Comet automation.
yea works in Projects Claude + Opus 4.1
you'll just need to trust me lol
otherwise I'll have to write up an essay explaining lol
Thats weird asf you cant see the document @distant aurora
Iāll convert it into a doc or something
Sure thing.
i not even sure O3 ever get this right
wow Claude Opus 4.1 platform 2/2
hi
It did, thatās the beauty of o3 š
can always fluke a right answer š
Claude limits actually not bad right now for Pro
O3 can hallucinate sometimes it does have a high hallucination rate š
And sometimes makes miracles happen
How much Opus You Can use it with Pro?
That 40ng/ml thing.
Risk group ā Very-high-risk (VHR)
Why
Parameter Patient value How it maps to NCCN criteria
Clinical stage DRE shows extracapsular invasion ā cT3a Counts as a high-risk feature (cT3ācT4) and is one of the elements that can qualify for VHR if combined with another high-risk factor.
Gleason / Grade Group Gleason 4 + 4 = 8 ā Grade Group 4 Also a high-risk feature (Grade Group 4 or 5).
PSA 18 ng/mL Below 20 ng/mL, so does not add an extra high-risk point.
Number of positive cores 30 % (ā 3ā4 cores if 12 taken) < 50 % cores, so does not affect high/very-high classification.
The NCCN guidelines state that a patient is very-high-risk when they have āat least two of the following: cT3ācT4, Grade Group 4/5, PSA > 40 ng/mLā .
This patient meets two of those factors:
1. cT3a disease (extracapsular invasion)
2. Grade Group 4 (Gleason 8)
Because he satisfies ā„ 2 qualifying high-risk criteria, he is escalated from the āhigh-riskā category to the very-high-risk (VHR) group under NCCN Version 2.2025.
O3 keeps it short and neat.
Just the way I like it.
GPT-5 Thinking already gets this right at least 40-60% of the time
it's not about whether it gets it right
it's how often
Opus 4.1 might actually be one of the best
but the funny thing is, I ran Claude 4 Sonnet Thinking on this like 20x one time (in PPLX)
and it got it like 19/20
I'm willing to bet mini will get this right more often.
then I just ran it last 5 mins...got like 5/10 only
nope mini doesn't get it right more
the unpredictability is somewhat concerning
but tbf, the issue is mis-interpretation of strikethrough text combined with RAG imperfection
I'll say this again.
limits onhttps://support.claude.com/en/articles/8324991-about-claude-s-pro-plan-usage
it depends
for large interrogation of PDFs? About 5-10 prompts per 5-hours
wats gemini and web search related to?
deep research in gemini gets it wrong too in my use cases
This is beautiful
like for many non-PDF analysis stuff i mean....Deep Research doesn't mean more accurate answer imo
it might pull more (crap) sources i guess lol
only beautiful if it actually translates to higher accuracy
which it doesn't in my use case
altho hard to tell
need like at least 10,000 tests per model i guess
anyone have a video i can watch to really utilize comet browser?
For cybersecurity i wont lie its good
Got 20/20 in comparison to gpt 5 thinking (chatgpt) and perplexity claude 4.1
Thought api claude 4.1 decided to be a clown
Perplexity claude 4.1 and thinking both did good
No clue why perplexity claude 4.1 tied with thinking while api claude 4.1 was doing bad
Different prompting?
Temperature probably.
What should I set the temperature to for claude 4.1?
Perplexity sets it low I believe. Don't know how low but it's low.
What?
i talking about my prompt
Iāll check on api
on Claude platform. But Projects is fine
read above
exceeds tokens
u need to test something that is large tokens
Omw to use qwen nxt
Max is 200,000 for pdfs
I believe it should nail it.
qwen bugs out at highest thinking a few days I tried
and gets it wrong many times on lower thinking
the point tho is
Qwen makes good local models ngl
u have to always check yourself anyway in a way for these things
Qwen Chat offers comprehensive functionality spanning chatbot, image and video understanding, image generation, document processing, web search integration, tool utilization, and artifacts.
wrong justification bro š
How?
No
bbl
I seriously don't get how it's wrong this time.
If you want an explanation
Simply copy o3ās correct answer and ask for the difference between qwenās answer and the correct answer and identify the difference
It should have enough context memory for all of that
Bet
anyway this is just one of many examples I have
I have an even more complex one (involving 3 PDFs), and GPT-5 Thinking gets it 100/100
on ChatGPT platform
no other AI gets anywhere near that for this one
PPLX + GPT-5 Thinking gets it like 1/100 lol
Bruh my qwen is still thinking. I don't think it will find any difference.
definitely big difference with API/PPLX custom etc
@left stream ask your O3 maybe
Even on max
Risk Group Assessment for the 68-Year-Old Male Patient
Based on the provided clinical information, this patient falls into the Very High Risk group according to NCCN prostate cancer risk stratification criteria.
Justification:
Patient Characteristics:
- PSA = 18 ng/mL
- Rectal examination shows possible extracapsular invasion (suggesting clinical stage T3)
- Gleason score = 4+4=8 (which corresponds to Grade Group 4)
- 30% of biopsy cores positive for adenocarcinoma
NCCN Risk Stratification Criteria Analysis:
According to the knowledge base, the "Very high risk group modified" criteria states:
- Has at least two of the following:
- cT3b-cT4
-
4 cores with Grade Group 4 or 5
- PSA>40 ng/mL
- Primary Gleason pattern 5
- 2 or 3 high-risk features
For this patient:
- cT3b-cT4: Yes - rectal examination shows possible extracapsular invasion, which qualifies as clinical stage T3
- Grade Group 4 or 5: Yes - Gleason 4+4=8 corresponds to Grade Group 4
- PSA>40 ng/mL: No - PSA is 18 ng/mL
- Primary Gleason pattern 5: No - primary pattern is 4
- 2 or 3 high-risk features: Yes - he has two high-risk features (cT3 and Grade Group 4)
The patient meets the "2 or 3 high-risk features" criterion for very high risk classification. The high-risk features are defined as:
- cT3 or higher (extracapsular extension or beyond)
- Grade Group 4 or 5
- PSA > 20 ng/mL
This patient has two of these features (cT3 and Grade Group 4), which satisfies the very high risk criteria.
Therefore, based on NCCN guidelines, this patient is classified as Very High Risk prostate cancer.
@distant aurora iām setting it to this first
Bet
Itās gonna be interesting
Been a fat minute since I last used o3
@quaint ravine ai isn't the problem, the formatting of your document is problematic.
i know
but human can interpret that easily
also actually, AI can understand that strikethrough text
AI isn't a human obviously
yes but this is real world/life official document
it's not even the strikethrough text that is the issue
it's more the fact it misses that this part is actually more of a "changelog"
and anyway, as I said, I have numerous examples
where perhaps only 1 AI platform gets it 100/100
while all others are 50/50 at best
the fact 1 platform can get it 100%, while all others don't means there is still plenty of room to improve overall (since that one platform shows it can be done)
there's actually an updated version of that document now, which has taken out entire part of that text, so no more excuses š the AI still gets it wrong occasionally, but higher chance of being right relative to that older document
šØĀ STOP SCROLLING! The Craziest AI Drops of the Week Just Landed!Ā š¤š±
What if the future of business, creativity, and workflow was already hereāand itās fun, fast, and mind-blowingly accessible?
This week, AI launches are breaking the internet and flipping expectations. The buzz is real, and everyone from creators to founders is j...
I actually thought Gemini 2.5 Pro on PPLX always got that prompt right btw
was funny, cos even NotebookLM got it wrong
but I failed to test it many times....it actually is pretty decent on PPLX tho - probably 8/10 times right typically
thing is, with 80% accuracy, if u just test it once a day or so, you likely not to see it go wrong lol
@green crystal @gloomy cipher https://vxtwitter.com/lianghu349103/status/1968679731177402461
1/7 We release FinSearchComp, the first expert-level benchmark for financial search & reasoning ā 639 questions from 70+ finance pros. #Grok4 ranked #1 š and close to human experts, GPT-5 the second, while others fail at basic analyst tasks.
Page: https://randomtutu.github.io/FinSearchComp/ https://t.co/6kqhnvfcfI
HI
hi
WHERE ARE YOU FROM
chill
india
dont scare me
xD
caps
you?
mp
@dense warren
@blissful raptor founder?
Hey @tranquil orchid
@tranquil orchidhi
Hello
hm
actually i don't know
i got a popup on left-down side
upside of my account name
i clicked adopt
try rejoin the server
or authorize
it must work
Is it available via api or grok app only?
Sure
which state u r in ?
?
ntg
how you joined this srvr
I dont think we can share locations in public
Well in mobile i opened my app and joined the server
np , but i had nothing to do with that
how
like you joined from a link of pplx?
join from here
Yeah in, perplexity app and i opened setting and clicked pro discord
Yeah i joined form there
then ask a mod
Well it isnt a big of a problem is there any benifjte for pro here
web app only
invisible status
Ohhk
App only afaik. Only open ai and Claude native searches are available on open router.
anyone want to try reproduce that algebra equation failing on PPLX + GPT-5 Thinking?
while on ChatGPT platform, it never fails
what does the "mac" in "mac & cheese" mean?
macaroni i guess
my Spaces Super Search says:
TL;DR: āMacā is just short for āmacaroni,ā the pasta used in the dish; āmac & cheeseā means macaroni with cheese sauce, not an acronym.
Explanation
In common usage, āmacā is an abbreviation of āmacaroni,ā referring to the pasta shape traditionally used for the dish. In the U.S. the dish is called āmac and cheese,ā while in the U.K. itās often āmacaroni cheese,ā which makes the meaning of āmacā clear. Some people joke that āmacā could stand for āmacaroni and cheese,ā but standard usage treats it simply as shorthand for macaroni.
Some people joke that āmacā could stand for āmacaroni and cheese,ā
they do...?
some high level chat here lol
TL;DR: Yesāthereās a recurring internet joke that āmacā could stand for āmacaroni and cheese,ā making āmac and cheeseā read like āmacaroni and cheese and cheese,ā but in normal usage āmacā is just shorthand for āmacaroni.ā
ha fair enough
is this free pro subscription really a scam
because I just watched a video where the free version gave better results than the airtel free promo
can u share the prompt and the demonstration
We ditched Jio for Airtel's "free" ā¹17,000 Perplexity Pro... only to discover the FREE version actually works BETTER! š±
How is this even possible? We put both versions head-to-head and the results will shock you. Is Airtel's "premium" AI deal the biggest scam of 2024?
chatgpt it has tools available lol
it probably uses python there
there were actually some search bugs in PPLX search for several days which have now been fixed
i'd suggest testing the prompts he showed in Pro...I don't think u;'ll be able to reproduce that now
ai studio gemini gets it wrong too?
yes
i see
quick question guys. Is it possible to move threads from incognito to non-incognito?
Share
Yes. Seems that way.
Maybe you can try sharing it as a thread and opening it in another account idk. I have never tried it.
bruh
And then save it there.
No NSFW š„ŗ on perplexity
It canāt even be shared
it's the same one that I've posted about a lot, and is on reddit etc
