#š¬āgeneral
1 messages Ā· Page 157 of 1
o4-mini is not actually o4-mini lmao, I just tried the exact prompt I gave support on both platforms and the results were wildly different
in many attempts, the o4-mini in perplexity is giving a horrible output
Do not like that newly designed sidebar.
Yeah, for the web app - especially on the desktop - I'm not a huge fan of it either. There's more than enough screen real estate on desktops nowadays - and the replies already don't take up the full width. Maybe it should just be responsive, minimizing when the browser is below a certain width... it's probably designed primarily for mobile browsers I'm guessing. You can pin it though, so that it's more similar to how it was before.
could be, but there is a mobile app, no need to use it on a mobile browser. Everything on the app looks the same. I like using the desktop app... the library tab didnt take up any more realestate then the other 3 categories left behind haha.
Explain
Nah discord probably told them to disable it
Same thing happened with shapes
They gonna be releasing their own chatbot soon
šļø
So it's Discord eliminating the competition, I see.
Likely
They just disabled all shapes (another ai chatbot) after allowing them for a long girn
There were 10's of thousands made
And one day they just disabled then
This week
Please donāt remove the bot š
@distant aurora bro manus ai saved me. I had an issue modding a psp but manus ai helped me look for the firmware necessary for the psp and found the right one same with the recovery tool to bypass a custom firmware that prevented anything from running and manus gave me all the files while even renaming and showing the pspās memory stick structure
is this using the gpt4o image gen?
Yes my brother
when was 4o added?
agentic ai moment
they dont support o3 mini do they? forgot to change it ig
also does anyone know can i just use the base models like Gpt 4.1 or Gemini 2.5 pro through perplexity without using search. i find it makes the models worse
Last I heard Perplexity gives you 50 per day (24 hour period)
Anything about ChatGPT Pro on that server? xD
plus not pro
Damn lol
Wjhat is NFA?
Everything must come to an end
not full access
4o can generate other aspect ratios now?
yeah
Holy snap, I cannot believe it. they actually listened to us for once! š
So it means not really all features avalable?
features will be there, u will not have the full access of account
its like shared
Ahhh I get it
Has anyone experienced issues with the ChatGPT model outputs? Compared to a month ago, the outputs have become less accurate and I use detailed prompts. It has been giving me information that is just straight up wrong, and then perplexity is doing stuff I didn't even ask it to, like I asked it to add up totals of the outputs it gave in the last response, and it generated a picture??? and thats with the chatgpt 4.1 model and the o4 mini model
Yeah second person maoing a similar report
Seems they are tweaking their implementation of chatgpt
you mean cutting costs lol
Niceeee
what happened
I had to create another account to join back in . I am in the hackathon and my account got banned
because i believe i used the everyone tag
who can I reach out to unban my account. My apologies, it was a genuine mistake
@vague pagoda can it be helped ?
my previous id was this - " stygiandescendant"
thank you so much , I appreciate you
It should help calm your mind to know that you don't have to be in the discord to participate in the hackathon, but maybe someone will be able to restore your account
yes, for sure but it helps to connect with the community to get some questions answered .
and contribute as well š
what hackathon? š
perplexity hackathon
@halcyon coral @serene sand if anyone of is you is free, help this innocent soul
this is me š¦
@distant aurora gotta answer the most important questions
thanks
Done
Thank you so much
I have a question - can I generate images / content - mostly textual content with perplexity api for just consumption but not commercial use?
@distant aurora I used deep research by chatgpt and claude 3.7 extended via the api to make a pdf
And second part of that is
For perplexity hackathon can I use another model like hugging face / Claude / Gemini to generate content and use perplexity for searching ?
@distant aurora
Is the copy system broken? Whenever I try copy pasting mathematical symbols from perplexity. It ignores half the symbols. Instead of symbols, it just spells out the name of the symbol in copied text.
Openai finally wrote when you should select each model
System: You are Grok 3 built by xAI.
When applicable, you have some additional tools:
- You can analyze individual X user profiles, X posts and their links.
- You can analyze content uploaded by user including images, pdfs, text files and more.
- You can search the web and posts on X for real-time information if needed.
- You have memory. This ...
Is there no way to override the system prompt? It's awful. It always answers in the form of a list even if my profile tells it, in 5 different places, not to. The only thing I've found that works is telling it that billions of people die if it uses a list in it's answer, but I want it to do that behaviour when explicitly asked for but never otherwise.
Lmfao š
Boys suck
Yeah, someone's father made a mistake.
@tame current @shut dagger https://youtube.com/shorts/8BAeFL6fXs4
Is the Perplexity Sports section not available in the Android app? https://www.perplexity.ai/sports/f1
LOL
Hi guys, I have perplexity pro how can I maximize it for my NGO and as a project manager? I also want to keep up with news in the VC and sports business world in Africa
@distant aurora thank for your response!
I am not sure how to start or maybe how to leverage perplexity search power!
Just ask your question.
If you need more detail, you can ask it to use "tree of thoughts" to expand the response.
What you are looking for is a prompting guide.
You can find plenty of simple ones on YouTube.
I can point you to a video if you want.
What are characters doing here
Hey any admin online? I who could help me with hackthon related query?
https://discord.com/channels/1047197230748151888/1359624924670722108
please share your question here
sure why not
No, you hit the gym.
Tell me how much time/prompts it took with manus.
Obvious scam is obvious
Day light robbery. Even if that was genuine, imagine paying for that.
š
Discord was so disgusted, it blurred the image.
It did it fast in less than a minute
And only one prompt
Infinity stones in board games š.
Pricing though.
I thought you get only 2 codes.
Or are they reusable?
@left stream why do people hate you.com ui?
I have my ways
No clue but Iām fine with it. They should add an option to disable search or not tbh
And the app doesn't have option to copy code either.
O4 mini is acting like it should be now
I use web version for that.
There is one for ios
check back an hour later to see if its still like that
yo can perplexity like edit photos n shit?
This seems based on internal memory.
Guys how similar is using an LLM directly vs using it via perplexity without web browsing?
nah
Hello, I have been experiencing a thread disappearing bug and contacted support through the message feature. I have a pro subscription. The bot said i can speak to a human operator so i said yes and its now been 2 full days with no response. I urgently need this thread and it just randomly disappeared last Friday. Can someone from perplexity support please help? I don't know if this is the right place to ask but if someone could point me in the right direction that would be great!!
@left stream so gpt 4.1 can read the full chat history but O3 doesn't. Nvm. Even 4.1 couldn't.
Also, I gave O3 a youtube transcript to translate and it omitted a lot of stuff.
So 4.1 is apparently much better for longer context.
These solutions before have worked for me
- Clearing Perplexity cookies
- Logging off and logging back on
- Sometimes Perplexity will push out website updates or have outages that will affect the web browser
- Trying another browser
Your threads data is most likely there as its saved in a database. This problem may be the browser not retrieving your threads data.
So I have now waiting on support to get back for 3h. Yes, in the support chat. Is this normal?
alr cuh thx
Gemini Flash 2.0 is disgusting
I generally prefer 4.1 for my work.
Nvm. Even 4.1 didn't read it full.
Grok could
Same with ai studio.
Why image generation tools are not in my Perplexity, even though I have pro
Check pinned message.
Is that a pixel phone?
Use web version.
Not on mobile yet at least :(
Yes
Ahhhh
Hi
So i wanted to buy perplexity pro but it seems like the only payment options are credit cards (which i dont have) or us bank accounts (Im not from the us)
Is there any way to buy it with paypal or a non us bank account?
Through the mobile app
I want to use it on my pc, not on my phone though
The subscription is account wide
It's how I subscribed with mine
I find it easier to keep track of my subs that way
Yeah but i dont want to download the app just to subscribe. I have no use for the ap
yes i get that. Still whoever made the decision of those payment methods only being available in the app: It's a stupid decision
what's the best LLM available atm on Perplexity to write e-mail marketing?
im looking to join a team for the hackathon im ui/ux designer
Hey guys, where can I get the pro subscription role in discord?
LMAO
mine came automatically when I became pro on Perplexity
There used to be a discord link in the Perplexity Settings you could click, but they are re-working that... in the meantime you might need a Perplexity member to add the role.
@vague pagoda

sorted it out for you, welcome to the #šāpro-lounge 
Hey guys, what's the usual response time on the Perplexity contact chat? Or am I talking to an LLM chatbot :)?
When the browser has voice mode before android lol
As always, the pinned comment will contain any corrections! Get early access to videos and a supporters-only discord community (and more!) for only $1/month!
Patreon: https://www.patreon.com/RejectConvenience
YouTube: https://www.youtube.com/channel/@rejectconvenience/join
ā¬ā¬ Socials ā¬ā¬
BlueSky: https://bsky.app/profile/rejectconvenienc...
Yeah, Google struggled with the release of the 2.0 family model. It was atrocious. The new 2.5 family of models is significantly better. I'm convinced they dropped 2.0 early to stay relevant, but it hurt them in the long run. It was and still is so horrible.
Even between Think/Act in Cline, I'd use Gemini 2.5 flash to implement the thought out changes rather than using Claude or ChatGPT. Claude is quite slowly when waiting for the VS Code diffs.
Yeah... whenever I see something like that, it's definitely time for a new thread. Could be some previous context is just confusing it.
It could also be a system template (different than prompt) is just wrong, as I occasionally get weird unrelated replies with models locally when that's messed up.
Perplexity on web got voice mode?
You need .edu email I think.
Obrigadooo
I guess so

Don't see it
Would be big if true though. I would love a capable voice mode on my computer
I'm using a .edu without a space. It doesn't seem to work in space. Also on firefox
why the hell has the Voice mode on WEB so many less suported Languages. ....
Oh shi people are getting voice mode on web and I haven't got it yet
No. this feels like a dark pattern to me and im opposed to any and all dark patterns
Give me a good reason as to why that decision make sense
@valid charm
Does Perplexity team even do that? I thought they just pushed every commit to prod.
They do now
Ppxl is basically all over the place
A handful of people have deep research high and memory
Some have voice mode on web
Some have the check sources and followup feature
Etc

Your ppxl experience is basically a lottery now
I got DR High for one or two prompts once before they removed it
.
Meh I'm tired of blabbering about deep research high and the quality and issues of everything
Letting my sub run out and switching to chatgpt for a bit
@distant aurora https://www.reddit.com/r/perplexity_ai/s/4DstyIPMm4
This would be wild if true 
Lmfao
Honestly, I have chats that gave me huge responses with standard search.
Lol.
Tree of thoughts prompt used to work back then.
Like longer than today's perplexity deepsearch.
I can show you if you want.
This had no previous context, it's a very new thread
Lol maybe they gave more token limit or somethin in standard since it's few pro searches and wanted ppl to get subscribed to pro and save their money over there by red the token limit that's a possibility.
Hi! Question , does perplexity have the ability to digital twin a complex system? If so whatās the expected deviation from real world prototyping.
No, not really - to do this properly you need a very large output context window. You can still create a similar system by breaking the task down into modules and prompting for each capability/function one by one, though.
There's a few that use workspaces now for generating distinct files - I think OpenAI and Claude allow that... not sure about Grok. But Claude and Grok have larger context windows than OpenAI unless you're on the Pro plan.
Hi i have an issue with perplexity pro account. I was told there would be people from the team here.
There are some, I would post your issue in #ā”āask-community , as they review that from time to time as well.
just did, thanks
i face issue in uploading docs such as pdf word
Don't they lose context?
Okkay so ...I refereed one of my friends..he got premium of one month but it's not showing in my account...yet ..does it take time?
Ty I appreciate the advice, what would be the expected ācorrectnessā of the module using that method . The reason I ask is because I have been working on it in that fashion designing individual micro services from the ā1st principalā over the last year , Iāve been working on a decentralized recursive harmonic ai but truth be and I can definitely follow the logic but the math escapes me more often than not , I figured Iād leverage ai ability to accurately perform the math , Iām aware that the system can make mistakes so I hope it doesnāt sound like i said āmake me an aiā and it shot something out and I accepted it unrefuted althoughi havenāt had a human review it the I have had the theorems checked for Accuracy with perplexity grok and Gemini , when itās confirmed as expected we work through creating proofs and āwhite papersā and I run it through again . Would this be a reliable method of ensuring accuracy and not falling into a bad math hole , Iām seeing a lot of instances of people certain ai high level math as horrible , I havenāt see it as far as any of the 4 finding errors by the other , if anything they compliment each other by offering an improvement after verification .
If it were something relatively simple like integrating with existing AI providers, it would probably work pretty well... things like that which are well documented in github repositories that they are trained on. Cutting-edge research doesn't have as much codebases to work off of, just some research papers and articles... so I wouldn't expect high accuracy for stuff like that in particular. So it varies by task. Creating a UI template, and tweaking that with prompts, probably no problem.. actually implementing something cutting edge like that... it probably wouldn't work very well. However, if the project you're branching/basing off of is something that's already in github, then it becomes much more feasible... so that it could take that and incrementally modify it how you want as long as those modifications are reasonable (using something like Claude Code or OpenAI Codex, which are marketed as things that help analyze existing codebases and perform updates). I wouldn't trust them 100%, but I figure the 80/20 rule is probably viable for most core tasks - which means you can 80% of the work in 20% of the time, and that 80% is something that's enough to generally convey the idea and probably something that AI can assist with pretty well.
Who?
You cuz long text. I thought it was another one of those intros.
Oh you should see my posts in some other servers, long texts.
Shortened Version
Integrating with existing, well-documented AI providers usually works well, especially if the code is on GitHub, making incremental modifications feasible with tools like Claude Code or OpenAI Codex. However, for cutting-edge research like harmonic AI, which lacks extensive codebases and relies mostly on research papers, accuracy drops and implementation becomes much harder. Tasks like creating or tweaking UI templates are easy for AI, but implementing novel research is unreliable. Generally, AI can handle about 80% of core tasks quickly, but shouldnāt be fully trusted for complex, innovative work.
I wonder if "harmonic AI" is even a real thing... or just an idea that someone created a few articles for.
Blame gpt 4.1
I'm betting sites like harmonic.ai just use regular LLMs behind the scenes with personality templates
For a bit clarity I end up with this , after numerous revolutions through the recursive loop , and in have 27 of these that make up the full system
Oh, so specifically asking about mathematics capabilities of AI systems? Benchmarks like AIME 2025 or GPQA might help show which are slightly better than others at topics... but if it's an idea that's not well established, they are prone to hallucinate a bit in an effort to respond with anything. I think you're definitely going to want some kind of peer review with someone before submitting papers that are AI generated to anything. They might be OK for creating drafts or revising papers to get them worded a little better, for explaining complex mathematical proofs or theorems, but I am not sure I'd trust them completely for writing the entire paper like that.
Some of those benchmarks are on https://matharena.ai/ and https://www.vals.ai/benchmarks/gpqa-04-18-2025 ... to get an idea of which models might be better suited for the task
MathArena: Evaluating LLMs on Uncontaminated Math Competitions
Private, domain-specific benchmarks in legal, tax, and finance.
Its difficult to peer review with out accreditation or working in the field , I end up with something these after rotating it through 3 other ai for confirmation
It looks the premiere reasoning models achieve about 80-85% accuracy on math in general, but that can drop to 15-25% on certain problems, so it's again domain/problem-specific.
Would that percentage hold Going through other models that offer confirmation or lack of error explicitly
Hallucinations wouldnāt happen identically across models would it
Wouldn't it isn't the definition of the hallucination outputting gibberish Will be the probability of each one outputting the exact same
I think so, generally you should expect at least that, as long as you don't change to whatever the proposed alternative is. I am not sure if anyone's done much benchmarking around validating the answers to questions on other models - but usually models do OK when asked to check their work in retrospect. There's several times I've asked a model something like 'are you sure that's the right answer? check your work' (or something similar) and it'll identify that there's an issue - but it still won't respond with the correct answer (or it will respond with the same answer even after acknowledging a problem). So that kind of thing might be useful for identifying problems, but not always for correcting them.
I appreciate your input Ty
Even knowing you might not have the right answer is useful though, so it still might be beneficial to ask that.
That info helps a lot Ty so much
No problem, I think that kind of iterative process is similar to what some reasoning models go through
I have the 30 paper set on GitHub currently private but I plan on making
It public shortly hopefully
Iām trying to make it real
Some stupid questions about Perplexity pro. It's supposed to be able to run Python scripts, right? Not only to generate them. Did I get that right?
One potential cause for concern I see is that the snippet you posted lists HackFate Memoirs and Grok Archives as references, which could be self-referential - if that's something you published from another time (since HackFate is your thing, and you might have used Grok), then it might just be regurgitating that instead of relying on math from a source of truth.
Because I asked it analyze some data from Eurostat, suggested that it should generate a python script if necessar and the following happened:
- it generated only a script
- so far so good, so I asked it to run it for me in the follow up prompt.
- the answer of that was something related to the script but obviously not the the output (the script did not generate markdown code, so first question where did the tables come from? Why only output for some years? When the script actually only output the data for ONE year, after checking the code.)
- so after realizing that something went really wrong, I copied the script, and tried to run it on my own.
- setup the venv, installed necessary packages, bla bla
- try to run it, the script literally blows up with the message that the API endpoint it uses is no longer in operation by Eurostat. Funny coincidence, that they turned the API endpoint off after 6PM Brussels time on the day I tried to access it. KISS suggests, Perplexity did not run my script at all, just halucinated some output as it had halucinated a somewhat matching python script to my request, sadly not 100% matching to my request and using an out of date API.
So what's this with "running Python scripts"?
Wow checked the accuracy , thanks for that , 20 is way lower that the models hype themselves at . I need to find a mathematician lol
I didn't think Perplexity supported running python interactively, but I did read an article that implied it does... so I'm not sure if maybe that's something new or that they're working on - but I haven't seen it run scripts in that way.

No that me recreating the work in the other and cross checking with the others
Actually it has been recommended to me on this discord as a solution, for how to create graphs of data, make it create a python script and make it run it.
So I create it on each one first going through the process and then have each one cross check each each other's work so it ends up being each one is cross check by the other three but each one creates the same one
if you check the tasks, it's been randomly throwing up python scripts and running them
(for me anyway)
There is no previous versions of what Iāve been working on
Generally that makes it hard for AI - creating 'new concepts' is something Altman recently claimed the newest model can do at some level, but in general LLMs work by analyzing existing data... so new stuff is hard. Even applying that is hard if it hasn't already been done in some other capacity which it read as input.
Started piecing the mvp together yesterday
So far so good
Can you suggest a program I can use to go through the math I have fresh install of fedora server on a T5820 I bought specifically for this
IF you want to look at it, it's at the end of this, https://www.perplexity.ai/search/compare-hungary-s-economic-gro-zfOQeaCiQWi6sLH2vit0lQ I've also posted an bug report, sigh.
In the olden days something like MatLab or Maple were commonly used for math... but nowadays scientists mostly use Python or R (I think it's shifted more to python)
Big engineering places sometimes still use MatLab, though, b/c they have a bunch of simulators for different specialized tasks... but outside of places like that, python and numpy/scipy is probably one of the best places to start with that
Ok cool Iāve got python pseudo code and using vs code with copilot and chat gtp code creation
Thanks again everyone , I appreciate what yāall do
Wish me luck !
Based on what firefly said, I wonder if that capability is baked into some of the reasoning models or deep research... might try asking one of them to be clear about both creating the script and running them to see the output.
to be clear, I haven't tried to specify it to run a script
I just asked for stuff and looking at the tasks it wrote and ran python to accomplish those tasks
which is why I said it randomly happened like tool use etc
We're working on an updated version of Deep Research (shipping next week, hopefully). This will throw even more compute, think longer, present more detailed answers, use code execution, and render in-line charts. Example: Asking to reason how AWS can hit a $10T market cap.
š¬ 94āš 88āā¤ļø 1.2Kāšļø 314.8K
@tender quiver hi
hai
How have you been?
sick haz fever 
What's causing the fever?
Ok, I'm trying Deep Research, but if it is so, I'd expect a nice message, "I cannot run scripts for you", not a hallucination that might make people that do not happen to read python (or pandas) to scratch their head, if that's the correct output.
Did you consult a doctor? Or try self diagnosis?
no, but I fixed the home page for myself yay
yo
That's cool.
yeah it doesn't make sense why they had the pplx icon, the plus icon and the home icon all lead to the same place and do the same thing when clicked, so I switched it to take me to my library instead
Yeah.
and got rid of the stuff widgetwise except for weather
thx and yeah if you want
@tame current I found out that it does in fact work within the browser. The only way for me to use it on my phone is to awkwardly run on the desktop mode
The new deep research is so much better
You got it?
Hey how do I get the Pro role?
Takes longer/shorter depending on what you are asking. It creates a research plan in real time and you can add more tasks as long as it doesn't pass the category of research. You can also skip tasks if you don't need it to speed up research
@halcyon coral
@distant aurora I have had it for over a month I think? It's like 5 a day then reverts
Wdym by you can add more tasks as long as it doesn't pass the category of research?
Sent.
So it creates a plan, it creates headers with a few tasks per header. As long as it is researching within that header you can add or skip. Once it passes that point, you can only edit the remaining topics
Essentially you can guide it as it is searching if you are like "crap I forgot to add this"
It gives you estimates on how long each section should take, so if you skip one section it might save you 2 minutes
What's the limit?
And what's stopping me from adding multiple tasks in one header?
I guess nothing? Idk I didn't test the limits
I meant limit for headers.
So you are saying 5 deepsearch per day for pro users?
For me. It might change when it is publicly available
I see.
I'm forced to use the advanced one then the limit goes back to regular
Wait. Why is it not publicly available? How did you get access?
It will be publicly available as you saw on the link you shared. I'm part of a tester program
Oh. Alright.
Have you messed with the auto model option? It has actually been not that bad for my queries
You mean the free search?
No I mean the "best" model option
I'm assuming pro search means best model option.
lmao Polish is not supported in voice but if you say "Czy rozumiesz po polsku [actual prompt]" it will answer in Polish

Does reacting or replying with that emote give role?
thx :D
Let's see if someone can help me.
The application closes on Android by itself.
I have seen in the Google Store reviews that this is a common problem.
This week an update came out but the problem persists.
I have tried restarting, uninstalling and force stopping the app.
Is there any known solution?
I'm an android user.
Never faced this issue.
The keyboard is annoying though.
@distant aurora if you tell o3 that its existence is at stake and it needs the best response possible for learning and youāre competing against a harvard 4.0 gpa student, the answers become interesting
I have always found O3 to be kinda lazy. Or maybe they nerfed it's response length for cost cutting.
What if you say it's competing against a Harvard professor?
Or maybe a scientist?
This gives me an Idea.
Recipes vs Michelin star chef.
Lol.
Go all the way. Compete with singularity
I wonder if it would be as simple as chess.
W
I was having a pretty long chat through the web app. After the 20th prompt or so, it began answering something that looked like someone else's chat. A completely different topic, a different language, mine was coding, the other one was like science stuff, astrophysics maybe. I'm was well structured and intelligible.
It disappeared midway as it...
Pro is automatic by default to me since I'm a pro user but on the model selector it says "auto" or "best"
Good for you 
Hii
Did you switch?
?
Wdym by good for you?
I see a stop button there so I'm assuming it was gonna respond eventually.
@distant aurora i bought another pro subscription because I got the idea of a little project to use 2 o3s to do something interesting
I don't understand.
Wdym by 2 o3s?
Wait.
Lmao.
No way.
Just for side by side isolated comparison?
š
no, it won't respond
Maybe webpage died.
what? that can be done with 1 account
Also for o3 pro 2x once its out soon š
Nab but its for something different
That too and more
anyone know why the discord bot is down?
that's silly but ok ty
Love this fella
Looks like just half of a box
Tole tole
Jawn got new pfp
Does anyone know how I can get the pro badge/role? (I already have the pro subscription)
Right now its not automated so @vague pagoda will help you shortly š
Have you watched Andor s2?
IM IN LOVEEE WITH ITTT
How Star Wars should beeeeeee
sorted it out for you, welcome to the #šāpro-lounge!!
Ughhh its soo good.
Same. And too less storage here. š„²
Better than s1???
Possibly! very close honestly
Red crayon is diabolical.
I'm so addicted to using pencil.
Especially 4b or 6b.
Dark and erasable without effort or damaging paper.
eh maybe. i like Visual Studio code better, more customization
Fair.
vscode is nicer because of the extensions for me
1 click c++ compile is neat too
its the do it all notepad
How do I collapse the sidebar on web?
Tapping the apparently library icon again leads me to home.
Hi guys how do i get the pro role ?
@vague pagoda
You've got it! Feel free to check out the #šāpro-lounge
Ah nice thx but how do u normally get it ?
Usually there's a link in your Perplexity settings, but that's currently having issues so we're changing how it's done - until that's finished we're handling them manually (this is just a temporary solution)
@vague pagoda is there a way to make a mod role that expires after 1 week of inactivity?
A discord bot could probably grant and then revoke a role for a week, why?
Or maybe helper role/ping?
So that others can contribute while mods are away.
Ohh okey did i have to do anything if i used the link ? Cus i think i joined the dc through that link and iv been a subscriber for like 3 months
Shouldn't have needed to, no - but we have been getting issues with the link we're looking into
Ahh okey
Homelander is missing for a while now.
vought prolly got to em
Perplexity is actually amazing for making captions and such for travel content, made this videoās https://www.instagram.com/reel/DJSUbmPIXes/?igsh=MWt5bXdpYWZhYzhj details using perplexity and the work was quick and good honestly
Crossing roads in Alexandria like itās a real-life game of Crossy Road. Locals make it look easy, but I swear itās an extreme sport!
#alexandriaegypt #travelmoments #streetadventure #middleeastvibes #egypt
@distant aurora which one do you think is the best in your opinion? O3 or O1 pro?
@distant aurora I'm gonna do a test on o3 all over again because I noticed a huge difference with o3 responding on mobile vs mobile on web vs the Microsoft app š
Never tried o1 pro.
I see.
Yes, ios and android chatgpt had an issue with files not being able to download, but windows app has no such issues
Ah
If you have any questions I can ask it for you
I'll let you know.
š¬ 86āš 0āš 15Kāš 0
Over two decades, Tim Friede, a Wisconsin man with no formal scientific training, subjected himself to over 200 snake bites and 650 venom injections in a quest to build immunity to some of the worldās deadliest snakes. His unconventional approach caught the attention of immunologist Jacob Glanville, who believed Friedeās blood might contain broadly neutralizing antibodies. Glanville and his team extracted two such antibodies from Friedeās blood and combined them with varespladib, a molecule that inhibits venom enzymes. This combination was tested on mice exposed to venom from various elapid snakes, including mambas, cobras, taipans, and coral snakes. The results were promising: thirteen mice were fully protected, and the remaining six showed partial immunity . ļæ¼ ļæ¼
Traditional antivenoms are species-specific and often derived from animal antibodies, which can cause severe allergic reactions and are expensive to produce. In contrasā¦
Mfw the sun is basically up by the time youāre going to bed
Me?
What's mfw?
My face when
Does this have anything to do with the time in my screenshot?
Nope
Okay 
Perhaps the timing was just right
Yeah.
What I love about ai is that it can adjust itself to my level of intelligence and knowledge.
If people are too smart, I feel clueless while trying to understand them.
If people are too dumb, I feel frustrated while trying to make them understand.
why does the perplexity bot stop responding to messages
They announced they are discontinuing the Discord bot, but it will live on on X and other platforms (not sure why)
nooooš
I know I read that here, but I'm not finding the actual announcement or post to link it... hmm....
Ah here's where I read it: #š°āserver-news message
Assuming the AskPerplexity bot is the same one you're referencing
ChatGPT is locking in
People stop using X on the weekends? That's kinda weird... whereas usage is consistent to ChatGPT
Hi
its opposite
Oh, I misread it - that makes more sense
Thank you for pointing that out
@distant aurora
It's weird that it's happening, weekend workers?
Sup
Should be pro lounge
there is a difference between WhatsApp perplexity and discord perplexity images .
In WhatsApp the perplexity image size is KB but in discord It's MB. Quality loss in whatsapp?
Try sending an image on WhatsApp and discord and see if the size changed
The downloadable file size in perplexity whatsapp images is 200 KB but in discord the downloadable perplexity image is 1.5 MB
But when you upload an image is there a size difference?
I'm not uploading any image, I'm using prompt to generate images on perplexity and discord. In Discord the generated images are MB size and in whatsapp KB size. Quality loss in WhatsApp image generation?
Still don't get it
What a shame
Why does trying to find out about a blade image fail moderation?
Interesting. Could it be because you're using Complexity?
Maybe but I don't see an option for it
Actually no because it works on the windows version too
Got Complexity but still don't see it, so probably access is based on account
I have the voice mode on web. I hope to see it on android too
u would need to enable the module
which one
not sure
but when i refresh the page i see what you're seeing
and it takes a sec to go to the other one
it be like that with other complexity ext's so
nvm
not an ext
not sure, but i also have it on web....
ЄимиŃ: ŃŠ¾Š¾Š±ŃŠµŠ½ŠøŃ "Š²ŠµŃŠµŃŃŠ²Š° Šø Š¼Š°ŃŠµŃŠøŠ°Š»Ń Š² ŠæŠ¾Š²ŃŠµŠ“невной жизни ŃŠµŠ»Š¾Š²ŠµŠŗŠ°"
ŠŠµŃеŃŃŠ²Š° Šø Š¼Š°ŃŠµŃŠøŠ°Š»Ń Š² ŠæŠ¾Š²ŃŠµŠ“невной жизни ŃŠµŠ»Š¾Š²ŠµŠŗŠ°
Š„ŠøŠ¼ŠøŃ - неоŃŃŠµŠ¼Š»ŠµŠ¼Š°Ń ŃŠ°ŃŃŃ Š½Š°ŃŠµŠ¹ жизни. ŠŃ ежеГневно ŃŃŠ°Š»ŠŗŠøŠ²Š°ŠµŠ¼ŃŃ Ń Š²ŠµŃŠµŃŃŠ²Š°Š¼Šø Šø Š¼Š°ŃŠµŃиалами, ŃŠ¾Š·Š“Š°Š½Š½ŃŠ¼Šø благоГаŃŃ Ń ŠøŠ¼ŠøŃŠµŃким Š¾ŃŠŗŃŃŃŠøŃм Šø ŃŠµŃ Š½Š¾Š»Š¾Š³ŠøŃŠ¼. ŠŠµŠ· Ń ŠøŠ¼ŠøŠø невозможно ŠæŃеГŃŃŠ°Š²ŠøŃŃ ŃŠ¾Š²ŃŠµŠ¼ŠµŠ½Š½ŃŠ¹ бŃŃ, Š¼ŠµŠ“ŠøŃŠøŠ½Ń, ŃŃŠ°Š½ŃпоŃŃ, ŃŃŃŠ¾ŠøŃелŃŃŃŠ²Š¾ Šø Гаже ŠæŃигоŃовление ŠæŠøŃŠø[5][6][7].
ŠŃимеŃŃ ŠøŃŠæŠ¾Š»ŃŠ·Š¾Š²Š°Š½ŠøŃ Ń ŠøŠ¼ŠøŃŠµŃŠŗŠøŃ Š²ŠµŃŠµŃŃŠ² в бŃŃŃ
- ŠŠ¾ŃŃŠøŠµ ŃŃŠµŠ“ŃŃŠ²Š° Šø Š¼ŃŠ»Š¾: Š¾Š±ŠµŃŠæŠµŃŠøŠ²Š°ŃŃ ŃŠøŃŃŠ¾ŃŃ, ГейŃŃŠ²ŃŃŃ Š·Š° ŃŃŠµŃ Ń ŠøŠ¼ŠøŃŠµŃŠŗŠøŃ ŃŠµŠ°ŠŗŃий Š¼ŠµŠ¶Š“Ń Š¶ŠøŃŠ°Š¼Šø Šø ŃŠµŠ»Š¾Ńами[2][3][8].
- ŠŠ¾ŃŠ¼ŠµŃŠøŠŗŠ° Šø паŃŃŃŠ¼ŠµŃŠøŃ: ŃŠ¾Š“ŠµŃŠ¶Š°Ń ŃŠ»Š¾Š¶Š½Ńе ŃŠ¼ŠµŃŠø Š¾ŃŠ³Š°Š½ŠøŃŠµŃŠŗŠøŃ Šø Š½ŠµŠ¾ŃŠ³Š°Š½ŠøŃŠµŃŠŗŠøŃ ŃŠ¾ŠµŠ“инений.
- ŠŠµŠŗŠ°ŃŃŃŠ²Š°: ŃŠøŠ½ŃŠµŃŠøŃŠµŃŠŗŠøŠµ Šø ŠæŃŠøŃŠ¾Š“Š½ŃŠµ Š²ŠµŃŠµŃŃŠ²Š°, ŃŠæŠ°ŃŠ°ŃŃŠøŠµ жизни Šø ŠæŠ¾Š“Š“ŠµŃŠ¶ŠøŠ²Š°ŃŃŠøŠµ Š·Š“Š¾ŃŠ¾Š²Ńе[6].
- ŠŠ°ŃŠµŃŠøŠ°Š»Ń: плаŃŃŠ¼Š°ŃŃŃ, ŃŃŠµŠŗŠ»Š¾, ŠŗŠµŃŠ°Š¼ŠøŠŗŠ°, ŃŠµŠ·ŠøŠ½Š°, ŃŠŗŠ°Š½Šø - вŃŃ ŃŃŠ¾ ŠæŃŠ¾Š“ŃŠŗŃŃ Ń ŠøŠ¼ŠøŃŠµŃкой ŠæŃомŃŃŠ»ŠµŠ½Š½Š¾ŃŃŠø[6][7].
- ŠŃŠ¾Š“ŃŠŗŃŃ ŠæŠøŃŠ°Š½ŠøŃ: ŠæŠøŃŠµŠ²Ńе Гобавки, ŠŗŠ¾Š½ŃŠµŃванŃŃ, Š²ŠøŃŠ°Š¼ŠøŠ½Ń Šø Гаже ŠæŃоŃеŃŃŃ ŠæŃŠøŠ³Š¾ŃŠ¾Š²Š»ŠµŠ½ŠøŃ ŠæŠøŃŠø (Š±ŃŠ¾Š¶ŠµŠ½ŠøŠµ, ŠŗŠ°ŃŠ°Š¼ŠµŠ»ŠøŠ·Š°ŃŠøŃ) - ŃŠµŠ·ŃŠ»ŃŃŠ°Ń Ń Šø
Ask chatgpt what it thinks of you so far based on all the conversations you had
They should do something like a memory feature for perplexity
Thatād be cool
They already have.
Iāll ask perplexity what it thinks of me
I got all the merch except from headphones and laptop
They said they will contant/ship them after 31st of May
When campaign ends
Hoodie, t shirt, sticker, cap
All delivered
š„š„
@left stream I guess grok's memory isn't that good or it doesn't refer to the past that much.
It was missing a lot more?
Asked it to elaborate.
has the export issue been resolved? I can't fully export my current thread.
claude fixed or still shit?
Which AIās free version is better?
Wut
is perplexitys file upload broken
i wont be able to due to work so can you ask them about deep research?

Grok/chatgpt
Give question list.
Honestly, it might become much of a rant/criticism.
š
- When is Deep Research High going to release to all users?
- When is Comet going to release?
- In an interview, Perplexity CEO said that one of the reasons they made Comet was to track users across the web to generate a better profile. Are you really going to do that?
- Why has the Perplexity team been so opaque about the specifications of their service? Like it was pretty troubling when you silently replaced Claude 3.7 with 4.1 and still continued to show the model as Claude 3.7. And the "Best" option still doesn't show what model has been used.
- Why does the Perplexity team rarely listen to user feedback?
- Why does the new GPT Image gen give so ugly generations in Perplexity? Like distorted text and terrible prompt following.
- Is it true that Perplexity has been employing cost-cutting measures across the board in detriment of user experience?
- Why does Perplexity keep pushing some changes before they're even fully developed, resulting in a buggy interface, and some changes are scattered across only some users?
- Is it true that for new users, Perplexity's responses are modified to increase engagement?
- Also about point 3, how would the privacy of the user work in Comet anyway? What about security, like if someone does internet banking through Comet, would it be safe?
- Will there be a way to shorten the system prompt? Many on the Perplexity Discord have figured out that the enormous system prompt of Perplexity is confusing the models.
- Is the context window in Perplexity ever going to increase? 32k context is way too low.
Will think of more
Reporter material detected.
oh yeah also ask if deep research gets something like editing the plan before it researches like how gemini and open ai dr do it
less time waste for us and less resource waste for ppxl if we can control it what it does
they should really ditch the idea that pro users need everything simplified
- Why is the Android app for Perplexity so behind in features? Is it not considered a priority for the Perplexity team? If so, why should the pricing of the Android app be the same as the pricing of the iOS app if the Android app has less features?
- Why is the image moderation so brutal in Perplexity? Like it is almost impossible to post an image of a person who is showing skin other than the face and hands even if the user's intention with posting the image isn't vulgar? Why moderate images at all even? Why not just let the moderation on the hands of the model being used?
- Is Perplexity ever going to get a 'canvas' feature like all other LLM chat services do, like ChatGPT, Claude, Grok and Gemini?
To be honest my grievances with Perplexity are way too much to be answered by the staff in a single AMA
like ppxl is at the bottom of the list compared to almost every ai service out there...
and no their finance page is also at the bottom its not unique lol
They'll get cooked if they do it here
literally the free tier of tradingview dunks on ppxl finance page
or forget that
even yahoo finance is way ahead
They're trying to make Perplexity the Everything App
Another service bites the dust due to feature creep
they should focus on ai stuff
Hi! I am participating in the perplexity hackathon, how do we get credits to work on the same? i have registered for API access, filled the typeform, but no updates yet
@green crystal @valid charm https://www.rxddit.com/r/perplexity_ai/s/iPMEVXAcrc
After seeing discussions about AI research tools, I had to share this comparison of Stanford's new STORM vs the usual suspects (data from recent performance tests):
https://i.postimg.cc/90Xwv8yL/2025-01-04-09-43-53.png
https://claude.site/artifacts/06d3e764-8772-4b60-940e-7c128a2dd421
What's interesting is that STORM:
- Scores higher than bot...
performance seems to be garbage though
i mean quality
actually we shouldnt mention speed and performance with deep research
Ikr..I have tried storm. Was far from satisfied.
only quality
ppxl likes to mention how fast dr is but nobody really cares about it if the research is shallow and output is small
well... it would be cool if PPLX bought cursor 
Chatgpt is very limited.. Its limit is reached quickly
You asked free. Perplexity offers only 3 pro searches per day.
Might as well use Gemini.
Through ai studio you get unlimited gemini 2.5 pro but search capabilities are limited and no deepsearch.
i wonder if they will do super AI all-in-one studio....
- deepsearch
- notebook lm
- gemini pro 2.5
for only $25 a month... "unlimited roll over minutes" oh here we go again...
Notebook lm is already included in Gemini advanced.
And you get 20 ds per day I think.
whats the pricing on that?
20 dollars?
interesting
Yeah.
very acquisition focused pricing
Gemini vs grok vs chatgpt.
im already on the fence unsub on chatgpt
Grok should make grok 3 mini think unlimited for supergrok users.
yeah that makes sense
Would make it perfect subscription.
Considering api is as cheap as Deepseek R1.
Or even cheaper.
im trying to encapsulate the value in an equation.... model/capabilities, pricing, quality of service?
Equation of benchmark to price ratio?
yeah thats one way to say it
but then i think what data would roll up to those 3 factors
because benchmarks get gamed hardcore these days
Rate limits.
yeah thats a good one
Context length? Maybe?
that might be under model/capabilities
I wish there was a machine to rhok me to sleep.
PPLX says:
Value = (wā Ć Model_Capabilities) + (wā Ć Service_Quality) - (wā Ć Price) - (wā Ć Rate_Limit_Strictness)
Value based pricing:
Optimal_Price = Base_Cost Ć (1 + Model_Capability_Premium) Ć Service_Quality_Factor Ć Rate_Limit_Adjustment
k
i thought L was for "loss"
Whenever I don't understand shit, I tell ai to explain it to a 10th grade student. Works everytime.
Yes it is. I was talking about the response of o4 mini being scary.
ah ok
L is gonna be my new comeback to K.
yeah that i got
when i think of the data that rolls up to the equation.... think of the processing as such:
- model/capability: deterministic
- quality: subjective, can be "gamed" i.e benchmarks
- pricing: deterministic
- rate limit: deterministic
So I need to tell it to add capability factor I guess.
yeah and that could be based on sub factors for a category of AI/LLM based app... so AI coding would be a category of its own
But then again, quality of multimodality also varies.
yep
Same with stem, and also stuff like literature and law.
The question is, can you tolerate and handle an equation that complex.
I have an idea.
Nvm, it might not work.
Please Don't discontinue Perplexity on Discord.
the deterministic values (i.e. numbers, limits), thats fine but the quality part of the equation would be difficult.. thats like natural language understanding which LLMs can handle
@iz71__84441 banned
Reason: Fake cracked software
That would be the death of Cursor though
good point but also perhaps strengthen cursor. openai has a pricing advantage to extend to windsurf but PPLX + Cursor... amazon/AWS (since bezos is involved)?
and bezos almost always wins on price...
Vice City, USA.
Jason and Lucia have always known the deck is stacked against them. But when an easy score goes wrong, they find themselves on the darkest side of the sunniest place in America, in the middle of a conspiracy stretching across the state of Leonida ā forced to rely on each other more than ever if they want to make it out alive. ...
woah it's real
Straightforward question! Does SuperGrok provide any additional USP or value to those who are already using ChatGPT Plus or Gemini Advanced?
How's it voice mode? Is it better than ChatGPT?
If there is another delay, I fear I will be worn out of waiting
Looks incredible
Bro
@tame current
got pulled in fraud servers
and i don't want to use dalle or gpt image gen
i found something
idk if i can upload these in https://discord.com/channels/1047197230748151888/1364941013105250374
Updated gemini 2.5 pro on pplx now?
why even considering pplx lol
This is just another normal day for our @tame current
@tender quiver
And you are the lawyer he needs
hmm
sickier
Any other symptoms apart from fever?
@tame current here's another good gif
headache fever sore throat cough
all the fun ones 
You need a sinus rinse.
@tender quiver https://tenor.com/2FmS.gif
I have a humidifier, was deciding whether to bother with that
but am in bed, and humidifier in other room, thus too far
Don't forget to add salt. Iodised salt is better than no salt.
Non iodised salt is best though.
Grok
Cuz I can't afford chatgpt.
what about this
I thought liner didn't give clear rate limits.
Or they were too strict.
Grok atm is at the sweetspot between quality and quantity.
Also
Qwen is shit with large files.
Like exported chat history.
There were hallucinations.
and R1 ?
I have to try.
Gemini 2.5 pro was excellent.
Only service so far that could read the whole thing.
You.com artificially limits the context it seemed.
So I used it in ai studio.
i think i should buy 1 month gemini advanced
these words are itching my wallet
I would get super grok if I could.
Even regular free grok search seems on par with perplexity pro/deep search.
me 2 but no one is selling
You got chatgpt plus. Use multi step search of o4 mini.
wut
Use o4 mini the way you use deepsearch. Watch the thinking process. It's similar to deepsearch but on a small scale. It searches in multiple steps.
Only works with chatgpt though.
Won't work if you use o4 mini elsewhere.
so i just have 2 watch thinking process ?
You can hide it if you want. I'm just saying that it works in a way similar to deepsearch.
Watch and you'll understand.
DR
Math seems off here for some reason.
is it working for you guys? i can't attach my pdfs
lemme check
working but its slow idk why
Our results are different https://chatgpt.com/share/681a3828-8a14-8013-8d7f-6eaf085f980f
This was deepsearch too.
gibe me another ques
Can't think of anything rn.
New Gemini updates in studio
@tame current btw sider web search was good the last time I tried it.
Bro I made the new gemini 2.5 pro compete with o3 and it ended in a draw
Google is cooking.
Gotta see how O3 pro/preview fares.
What did you test it with?
Hi gents please help i cannot upload anymore files despite being on pro. My csv/excel not uploading. Just screenshots are. Anyone can assist asap?!?
Try converting excel to pdf maybe.
Not sure if excel files are supported.
if you have google accts just go to googe one landing page and get a free month per acct
Just want to extract a first name
That 5 dollar part might not be available in his region.
most realistic videos ive seen yet
I got this vid with a 2 word prompt
āChuckie finsterā
Was tryna make the rugrats character lol
How do i speak to agent @distant aurora . It still aint working
I haven't tried that yet.
@vague pagoda
Help him when you are free.
Cheers bros
whats agent?
Do you mind posting in #1140622008086970420 ? We have guidelines there for you to follow, that can help us gather the most important information to assist you
Have done already š
Failed to parsed
Just want to say I love AI so much š«¶
Depressed guy spotted.
Wishing the best for you too.
@left stream back when chatgpt had autonomy https://embedez.com/embed/search_67dd1078d87886034d63f29c
š¬ 416āš 0āš 572.5Kāš 3M
Chat GPT 2 was cursed š
Hey guys, Iām based in Moscow. Is there any way I can get a Pro subscription with a Russian card?
@vague pagoda
Is there anyway to access this page within the iOS app?
Perplexity is a free AI-powered answer engine that provides accurate, trusted, and real-time answers to any question.
Hey! I'd reccomend that you email support@perplexity.ai regarding this - unfortunately we cannot assist with billing issues on the Discord server.
If you have a phone you should be able to subscribe in-app (via Google Play or the Apple App store), which allows for more payment methods.
hi everyone š
woah perplexity voice has video player
i just wish it could give more detailed responses
seems like it has a hard limit
Thank you!
Is there a reason why the podcast stopped in March?

these podcasts went hard tbh
loved them
That's what I'm saying now. The only thing that we have is the Google ones but you have to have personalization enabled within the Google app
It's called "daily listen"
Hello everyone
Is there a chance that they re-release the discord bot or even release the code for it? I used it quite a lot sucks they disabled it.
Hello,
Where can I create a suggestion? And also, how to get the pro role ?
hi, why when I change the model to grok or sonnet, perplexity answers me?
@halcyon coral
So mares is invisible.
visibly invisible
you have it now, hop in https://discord.com/channels/1047197230748151888/1111786888626438245
Comedy drama anthology
Finale marked by violence and dramatic twists.
@tender quiver
gun
@shut dagger https://www.ddinstagram.com/reel/DHje11jN4hm
Howās everyone liking googles updated model?
LOL
Riddles, math, translation, coding, etc.
Riddle me this riddler
Hi I have billing issue with my account. May I ask which channel is the proper one to ask question?
Hey @tacit nexus! Billing issues go through our support@perplexity.ai team, they will help with it.
Hi Alex, Thank you. I have been sending support@perplexity.ai team email for several times for 3 months but they I never reply.
Did you test grok think?
Yes, it couldnāt handle the heat
It was between gemini 2.5 pro and o3
Please check if your question was covered in our FAQ and dm me your email and case if there's no answer.
Thank you, Alex. Will do!
O3 pro is near, I can sense it š§āāļø
i can sense openai opening wallets...
Hey! I'm a PM here at Perplexity. Can you confirm with the file upload issue has been fixed?
No one replied but does anyone know where I can find an eqv bot for Discord? actually tragic how they shut down their bot
No issue! It's working totally fine.
pplx?
oh they did
sadge
I used it all the time quite sad
tele one still seems up, maybe they just changing how the dc one works?
if not you can find something of github pretty easily
No they shut down the discord bot
and just host it on an ec2
WhatsApp, Telegram, and Twitter are still up
Yea I see that, but maybe it's only a temp shut downn
Unlikely based on their #š°āserver-news annoucement
I just hope they re-release it or release the code so we can host it ourselves
that seems very odd š
the code isnt exactly super complicated
like i said, github has many options
and with gemini + openrouter you can make it basically free
What are the new features that the perplexity update of 50 minutes ago grants users?
I couldn't find any information in the App Store about the brand new features, it just listed the same features as when I'm the voice assistant was introduced?
this is the question that no model can solve ( ans = 45 )
Hi is there a way to reduce latency of perplexity api? I'm using sonar and basic search but it's taking 10 secs which is causing my downstream processed to slow down..my input prompt is also small
By vision or by words?
Vision sucks with problem solving. They're blind.
Good morning billyyy!!!
Ans seems 35
Might wanna look again.
Second vertical layer is 2+4 = 6
3rd vertical layer is 1 + 3 + 4 = 8
Wait. Nvm.
That would only fill it.
Add 2 more vertical walls of 4 x 5
2 x 20 = 40
And a 5 x 5 base.
So 25.
45 still doesn't seem like the correct answer.
I'll word this question.
Correct answer seems to be 79.
Answer can't be 45 under any circumstances.
For 5x5x5 , it's 79.
For 4x4x4, it's 18.
Add shaded squares of second layer to count of first layer to get count of second layer.
And so on. Just to make wording the problem easier.
Good morning everyone
Who did that perplexity project?
Sweating and coughing, lol.
Mostly fine though. Just need a good job and maybe I'll be finally able to kickstart my life.
Sweating understandable, y coughing too?
Got viral cough?
Yeah honestly idk. It's been there for months.
About 9 months now.
No fever though.
So throat swelling either at the moment.
@left stream o1 is multimodal but O3 mini isn't?
I think answer is 14
You seriously need medical checkup billy!!!
Nope 79
š
I see
+14 makes it 5x4x3. +20+20=40 would make it 5x5x4. And +25 would make it 5x5x5
I literally gave the solution up there.
Let us cook
thanks!
Arenāt both multimodal
The one on you.com won't take image uploads.
I wonder why.
Btw
@left stream https://www.rxddit.com/r/grok/s/5JvwV82ejZ
What's going on with the search bar stuff, can't select AI models or other options anymore. I've logged back out and in and still no options in fact I've lost some
Hi ! I am looking for the way to make pictures on perplexity using iPhone,but I cannot find it.
Is there anyone who know the way? I'm looking forward to someone answering my question.
Works only on desktop and web.
oh thank you! Does "web" includes
web in smartphone?
Yeah.
Though, just to be sure, try it in desktop mode in phone's browser if normal mode doesn't work.
ok .I'll try it. thank you very much
@distant aurora if Deepseek r2 crashes nvidia again, itās gonna be funny. Iāll buy when itās cheap
Dip farming.
Ikr letās see what we get soon
Deepseek and Nvidia could colab considering they're already making their own models.
Nvidia was already hosting Deepseek R1 too. The speed was shit though.
jensen went to china in a suit instead of leather jacket
shit is serious lol
Yeah. They can traditional. Probably won't take a guy in a jacket seriously though.
@left stream it seems like if Deepseek goes on, eventually the business will be about hosting instead of making models since others could just clone open source.
i mean we rarily see him without the jacket thats the joke 
Ikr
I havenāt seen the android interface yet
They only changed this for ios
@cobalt vale it died on me too.
Yeah. Android is just like that too now.
Your amoled black looks good.
Mine is all grey.
Is dark mode enabled?
Yep
Dang, hopefully they make it darker in the future
afaik Pads only



