#💬│general
1 messages · Page 70 of 1
hmmm weird
"(schnell?) " No, Dev
It would be good if the new Pro could do it. It’s agentic. It shouldn’t be difficult to implement something like this
And it finally would be possible to generate images in the app.
That explains differences between text handling in Perplexity and in Grok 2. In Grok 2 text is much better generated. But dev is still good ofc
Hi there, Does anyone know how the rewrite feature with the different LLM works in Perplexity, does the same search's results being summarized using the different model or the search is done again to come up with the answer?
is the opus limit gonna be increased anytime soon? i just became a pro user today and ran through my opus messages rather quickly lol
OPUS 50. How is Sonnet 3.5 treating you? Should have a way higher allotment.
pretty good! there’s no limits for sonnet, right? though i have an unhealthy attachment to opus, i shall say… 💔
sorry. server meme from before 3.5 when we had unlimited OPUS and it was reduced to 50. so people would just make up AI art showing OPUS 50. it's died down since 3.5 came out.
pffft. i’m now very sad i was not present then. unlimited opus sounds like a dream come true
We had an entire art galleries worth of art that had no business being as good as it was given it was just a inside meme. I wanted to get it published somewhere. lol.
Or get a book made up I can present future generations. Sonny....back in my day..
Recently, for me perplexity has been doing regular web searches on wolfram for some reason, obviously getting no results....
Read it. Did it.
Got 7 fingers and 3 arms.
is anyone else struggling with the pplx-api response missing some punctuation marks? like no full stops etc
i'm using llama-3.1-sonar-small-128k-chat
they have so much compute tho, dev takes ~6x compute than schnell
it appears like gemma 2 27b in playground is dead
Hello, does Perplexity Pro support large context input?
Hi. Around 32k tokens
Cheers
To be fair the context window is the same regardless of model or Pro being on/off. 32k is the max size. unless Devs say something different.
Q: how do you get it to follow simple instructions after providing the instructions? I have to REPEATELY write "follow instructions!" for it to actually do it, it's such a waste of time and effort.
not really... uploaded files don't get added to the context window (which is more than 32k tokens - though i've given up trying to communicate this...)
Instead they get subject to a RAG process
that's nice, thanks for mentioning it. though I don't understand what pplx is doing with the prompt (I assume?), so schnell on free sites without registration is significantly better at text, than a more advanced model from the same family on pplx 😞
Perplexity - Platform B
#👉│feedback-general message
well, to be fair, some RAGs can be pretty good. it all depends on how "the most relevant information" is implemented. I think some implementations of knowledge graphs for source code worked quite well and inference cost a lot less. but it was not cheap nor fast to preprocess
hello. i just want to ask that how to change the aspect ratio when create image in perplexity?.
100%
there's nothing inherently wrong with using a RAG system (indeed, it is superior for various use cases) - it's just that it's not the same as directly inserting it into the context window
like for an entire repo, RAG would make sense - if you just wanted to be directed to the relevant files etc
but for working on a specific code file, generally the model being able to see that code in full works better
just for example (and imo/fwiw)
even the simpler RAGs have few knobs to turn which may drastically improve quality (chunk size, number of chunks). or allowing RAG to be used multiple times during the agentic search. not sure how exactly pplx handles it, but a lot of times it looks like even simple most naive RAG should work (like mentioning a unique identifier) ... and for some reason it doesn't. maybe sometimes "using RAG" fails (stuff around RAG itself) or something like that?
When will there be support for structured api?
would pplx be supporting structured api anytime soon?
if you mean JSON outputs, it was on their (now-removed) roadmap.. not sure whether they still plan to introduce it
aw man, yea this is what i meant
yeah i'm not too sure where it's at tbh (gosh it'd be nice though!)
the features that are in beta are still outlined in the Request documentation https://docs.perplexity.ai/reference/post_chat_completions
the fact that there isn't anything about enforced JSON outputs there makes kinda doubtful that they are still working on it..
but who knows.. do llama3.1 models support enforced JSON outputs natively?
https://www.reddit.com/r/LocalLLaMA/comments/1eqayuq/how_to_force_llama31_to_respond_with_json_only/
seems like no, but there are some prompting techniques
ahhh i see... that may be why it's seemingly been quietly scrapped.. (it would be way eaiser if the base models supported it natively i suspect at least...)
fwiw i assume it would be possible to try to make it work with prompting (in the system prompt)
but yeah.. it's sub-optimal...
yea, just not sure about reliability
one malformed JSON and everything breaks
yeah
also, the more prompting used to try enforce it, like more and more examples, the more of the model's attention is devoted to following the JSON output/schema requirements
rather than actually giving a good response based on the query/prompt and search results
kinda sucks that open source libraries like this https://github.com/instructor-ai/instructor-go dont seem to support pplx-api
hmm yeah i'm not sure.. but the pplx API is fairly similar to openAI's.. they're not identical or completely compatible, but kinda close (i think anyway).. If it's OS, you could perhaps try tinkering with it to work for pplx API
good point
btw, has anyone faced any issues with pplx-api generating responses that seem to be missing punctuation?
its a weird one. not all the punctuation are missing, only some
Haven’t used perplexity in almost half a year and thinking about coming back to pro. what has changed?
Is there any chance that there'll be a perplexity desktop APP for mac intel and arm or for windows. That supports Chatgpt4o features
desktop as in?
desktop app probably
Yes an app ✅️
Like perplexity desktop app that can read your screen and your face as well. Just like gpt4o
ehhh, seems a little out of scope tbh
Like this post if you want a Desktop App for Perplexity
i should add that im using isStreaming = true. is this a known bug with streaming a response from pplx-api?
What is the best setting for perplexity.ai related to a Software Engineer to use in daily tasks ??
Related to these give options the best customization anyone know about this ?
If you are asking about which AI model to use for software engineering, I would personally say to test it yourself, GPT-4 works best for me but some say GPT-4o or other models works better for them.
If I create login by email, and will buy Pro version on iOS. I can use it on desktop when logged in with this email?
Try between Sonnet 3.5, Opus, GPT-4o, and maybe Sonar Huge (since it's based on llama 3.1 405b). Make sure to use Pro Search since this is where the model will actually run/use python. If you already have any development specific frameworks/prompts, you can load those into Collections to inject custom prompts. (this doesn't require Pro)
Thanks for replying
Hey @zenith solar!
If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.
correct, no limit for number of devices
Can I fine tune the llama3.1 sonar models?
Wait why is there pro discord but links to this one?
You gotta leave the discord and then rejoin via the link in your PPLX settings (with an active Pro sub)
Still the same discord server
It's a channel on this same server
Ohh I see now it's the pro section in here
John Mccarthy, in 1956, coined the term artificial intelligence
What does it mean? Artificial and intelligence
The notion that developing and training our information onto machines to be able to learn from it is a form of learning method called deep learning
This was coined by a Canadian computer scientist in 1966
The extensive power of artificial intelligence means the definition alters all the time, leading to an enormous wealth of knowledge to digest and elements to pinpoint
AI is forever
Artificial intelligence is inevitable
Our ancestors cherished such an invention, disrupting existing industries paving ways for newcomers to dominate
The biggest companies in the world are using AI
You can’t defeat AI, you can only work with it
Defeating the systems that power artificial intelligence is the destined downfall of humanity
The very development that we created
Only to be branded as “AI bubble”
Albert Einstein once branded artificial intelligence as the soul creator of all things humans would thus rely on
He wasn’t wrong
The success of Artificial Intelligence during its early days are primarily thanks to these three talented minds:
- Alan Turing: Often considered the father of computer science and AI, Turing developed the concept of the Turing Machine and introduced the idea of machine intelligence, including the famous Turing Test for evaluating a machine’s ability to exhibit intelligent behavior.
- Marvin Minsky: A co-founder of the MIT AI Laboratory, Minsky made significant contributions to AI, robotics, and cognitive science. He worked on early neural networks and symbolic AI.
- Herbert A. Simon and Allen Newell: Developed the Logic Theorist and General Problem Solver, which were some of the earliest AI programs. They contributed to the foundation of cognitive psychology and AI as a scientific discipline.
Artificial intelligence is inevitable
Our ancestors cherished such an invention, disrupting existing industries paving ways for newcomers to dominate
The biggest companies in the world are using AI
You can’t defeat AI, you can only work with it
Defeating the systems that power artificial intelligence is the destined downfall of humanity
The very development that we created
Only to be branded as “AI bubble”
Albert Einstein once branded artificial intelligence as the soul creator of all things humans would thus rely on
He wasn’t wrong
I sent an email to the perplexity support email address and the bot responded back. However, the Discord invitation LINK seems to have expired.
If I tell this here on the Discord, will it be passed on to management?
OK we're on week two and the team at Perplexity have yet to acknowledge this is an issue and provide us with an ETA. We've been sitting on our hands, postponing deadlines and swatting internal emails. What's going on? What's the plan? Is there ANY plan? When will these issues be resolved?
Isn't the most basic function of Perplexity --> to source online sources? If not, why are any of us here?
company is a literal joke
ooof yeah..seriously what is going on
when i see mares name i read and think of it as calamary
I want to get trial period too 😦
this is getting quite serious. I know I may look like I am too soft on pplx, but it was usually advanced features or use cases not that directly connected to their main focus - ai web search.
but last few days, the search is a tragedy. like here - I explicitly told it, it was lying, plain wrong, and it still used wolframalpha and programming step to look for LLM model licences? 😩 then it just blindly parroted what I wrote without checking if I am correct (its only sources were useless wolframalpha searches, utterly irelevant).
that is not what I want nor how I imagine an ai search...
#1272961960765292625 message patience, end result will be 🚀
hmm, this (agentic search refuses to search web except wolframalpha) looks more like a backend bug though. the linked thread seems to be entirely front-end focused
This is quite exciting because one of the inspirations for App Router & RSC was to democratize the extremely dynamic rendering power of systems like Google Search and Facebook.
When you search for “weather in SF” or “Olympics medal leaderboard”, Google is able to dynamically
well, that is from almost exclusively front-end thread while my issue seems to be on backend...
are they changing pro search on backend at a same time as major changes on frontend? that ... doesn't sound smart to me and I wouldn't want to do such a thing at work...
Hi, all! For a simple question about how Perplexity works, which channel is best?
deep dive ➤ #💬│general message
people use #⚡│ask-community for simple question, though I think it wasn't supposed for that originally
Oh! #⚡│ask-community it is then. Thanks!
general overhaul, obviously
well, maybe I am blind (wouldnt be the first time 😅), but I only saw bigger changes on front-end. is there anything new backend-related? stuff like new search engines, support for bigger context or more LLM options like temperature, features like memory or persistent files?
did gpt bug fixes and slight update automatically get updated in preplexity too?
never mind preplexity gpt buggy lol, it replied to a thanks with the previous answer😅
Got a simple tip for you guys. Want a more thoughtful response? Start with "Go slow and methodical."
Can perplexity pro deal with coding? Specifically turning an HTML webste with css into a wordpress php theme. Or does perplexity pro deal with research?
"Could anyone provide information on the most recent publication date for academic research papers that are commonly cited in discussions of perplexity metrics? I've noticed that in my literature reviews, the majority of relevant studies seem to be from 2019. Are there more current papers on this topic that I may be overlooking?
it depends, is it a really simple (short) website? then maybe. I think prompt limit is around 8k characters, once it starts turning input into files, it starts using RAG and results from that may vary a lot. for html+js in one project I think it roughly corresponded to 300LoC, but it is a long time, I might be wrong. also there is an output limit, though I believe that's in tokens and was considerably longer then direct input limit.
for very short html stuff, it is quite good. for small web apps, simple utilities, I would say it is even practical https://monnef.gitlab.io/by-ai (majority of those mini projects wrote sonnet 3.5 on pplx; not all there I would consider practical).
tried to get a wordpress theme, but it probably won't be that easy 😅 https://www.perplexity.ai/search/see-uploaded-files-then-write-ZpLJtgqzRa6oO2byfwnY1Q it seems to know wordpress. I can't easily try it, but I don't think it would work as is
ps: you really want to use "pro" model for programming (sonnet or gpt4o). "pro mode" (the agentic search) is usually not recommended (context gets easily polluted with search results, so you might be better off using "writing focus", good model and "pro mode" off)
btw this looks interesting - are we getting file output from programming step? 🥹
Hey guys , How to generate AI images using perlexcity pro?
That misspelling 'perplexcity' could actually be a creative name that could be used in future.
"writing focus", good model and "pro mode" off....
So you say the pro model subscription is not THAT good for blog posts? Recently last week I use the pro switch to create blog posts to get high accuracy and make sure it doesn't hallucinate by using pro.
"you really want to use "pro" model for programming (sonnet or gpt4o). "pro mode" (the agentic search) is usually not recommended (context gets easily polluted with search results, so you might be better off using "writing focus", good model and "pro mode" off)"
Then I might have to now reconsider my current subscription to the pro version if it keeps on relying on the serps alone.
Perplexity is currently 100% useless.
Every single model is coming back to me with cached data... and well I tell the AI to go source new data, it tells me it cannot access the web. So basically, Perplexity is giving me ZERO online sources for each response.
When will this be fixed???
opensource >>.
Just subscribed to pro after like a year. What’s different?
Newer (better) LLMs and image generation models, better Pro Search. I recommend this link:
https://www.perplexity.ai/hub/blog/pro-search-upgraded-for-more-advanced-problem-solving
This is so much better oh my god
Am I the only one that is unable to get Perplexity to generate answers from any model other than GPT-4o today?
It's ignoring my setting and even when I do rewrite and select another model explicitly, it's still just using GPT-4o.
I've already done a full cookies/cache flush and restarted my browser and still experiencing the same behavior.
Thanks for this link specifically. I think I went wrong with asking for deep search for just creating simple blog posts. I don't need THAT much power now that I read the post carefully.
Works for me 🤷♂️
Did you try logging out, clearing cache, and logging back in? In that order
Hey @sudden wasp!
If you find the original message helpful, please consider reacting to it with the :star: emoji. If the post is appreciated by the community and receives 5 stars, it will go to the https://discord.com/channels/1047197230748151888/1082806833938436228 channel and the post author will get the <@&1082034222778302614> role on Perplexity.
Yes, I said that in my original message. Tried everything.
@blissful pendant I just figured out that it's unique to when I start a chat from within a collection
if I just start a new thread outside a collection it works fine. But from in the collection it will only give me GPT-4o
Yesterday it was working fine from within the collection as well, so this is new behavior today
Sounds like a new bug to create, although I'm not able to recreate it from Collections
https://discord.com/channels/1047197230748151888/1140622008086970420
Thanks, yeah I just created a new collection to test and I'm having the same problem still in that one too
Hello! As an experienced Senior Software Engineer with expertise in Full Stack development, AI, and blockchain, I am here to offer my support. Feel free to contact me for any collaboration opportunities or assistance you may need. Thank you!
buzzword after buzzword
Did ChatGPT write that?
If that’s what you’d need ChatGPT for..
perplexity runs on gpt 3 or 4?
Does anyone have trouble with Perplexity today?
Been having some since, whats your cause?
- When I type in question, the answer doesn’t show up until I refresh the page. If I upload a pdf or image, that image uploaded was gone
- When I upload a pdf and ask to research something outside of the pdf, it cannot do it
- The pdf that uploaded kept into next conversation and I have to restart Arc
Im not sure about 1.) , but im pretty sure 2, and 3 are something need to be fixed.
Hope it fix soon! Even it has its flaw, I do love Perplexity a lot.
I am thinking of subscribing to perplexity and give it a try. Previously, I was using you.com.
Can I use Paypal to pay for perplexity? I can't find the option when I tried to subscribe. Has PayPal option been removed? On the web, it says PayPal is supported.
I saw a field called "Promotion Code". Where can I get promotion code? Anyone here can share? I heard it's a win-win. you get $10, I get $10
I hope so too. It's been a while since I've noticed these 'bugs,' if I might say, or maybe they are not fully implemented features.
Is perplexity campus strategist also open to Canadian students? I’m at Uwaterloo and thought it would be fun to take on this role
hai guys i need to ask.
are the perplexity with claude 3.5 will have same result with claude 3.5 on claude.ai?
I accidentally recharged $1,000 in API points. Can I apply for a refund?
What do you think? Is it too late and no major differentiation or is it worth it ? What would you like to see in a Perplexity browser?
is the sir guy still here
I'm getting different answers when using Web vs API. The web answers are generally correct but the response from using the API is sometimes false. After digging deeper I noticed that the answers provided by the API are generally the same with GPT4's response. What's the solution to this? How can I get the same answer I get in web in API?
Hi everyone! what's the main difference between perplexity pro and perplexity labs playground
Hey @stark hedge!
Perplexity.ai and the Perplexity API work independently of each other. Perplexity.ai is the web interface with a wide range of features. The Perplexity API offers the Sonar models in the online and chat variants as well as open source models.
Hey @winged wave!
Perplexity Pro is the premium version of Perplexity, you can find more information here: https://www.perplexity.ai/hub/faq/what-is-perplexity-pro
Perplexity Labs is a free playground where you can try out different LLMs.
I know they are different. My question is how do I get same result I get in web in API? Should I change the model? If yes to which model?
The most advanced model is llama-3.1-sonar-huge-128k-online, but it is not possible to get the same response as from the normal Perplexity Free version, as different techniques and LLMs are used 🙂
You haven't still answered my question sir. Yes I have used the 405B but still not the same as the Web. So does it mean there's no way to get accurate answers using the API? Because from my test only the Web give accurate answers.
hi
most likely not. sonnet on pplx has lower temp, more limitations on output and input (eg RAG), but much more daily uses. depends on a use case, it may be similar enough for shorter inputs (till 8k characters) and outputs, and when you don't want it to be creative
Explain what you want and tell it to describe it in his own words, then after the answer press generate image
Thank you , I did that but I get only One image . I was expecting a bunch of images
hi
Everyone gets one imagine, you can generate I'm your own style by pressing on generate image again or in a different style
Can students 1 year out of college apply for hte campus strategic program??? @frank rivet @inland dagger @grand mauve @toxic moss. I discovered perplexity like 4 months ago from a Stanley Druckenmiller CNBC interview and it has been an amazing tool in my portfolio management. I would love to help grow perplexity and work with the team
You'll have to use an actual image generator for that if you want more than one image
what is the maximum device loging in one single perplexity account?
May I ask you why the Strategic Campus is available only for US based students?
I’m Italian and live in Italy, I have no problem with the English language, I’m a Psychologist specialised in CyberPsychology.
Any chance to see the Program expanded worldwide?
Im using perplexity on different devices, sometimes my account on random device just auto log out without a reason. I wonder if this is because it reach the maximum device allowed for 1 single account so it logged out to ensure proper performance right?
If it is the case, then what is the maximum devices to be allowed logging in by 1 single account?
what's the image limit?
there was no indication that a limit was in place until I got this and refreshed to see I have "0 left today"
Hey @weary crescent!
The limit is 100/24h
That should be displayed somewhere. People should know the limits lol.
any plans to expand the Campus Strategist Program for canadian students?
@halcyon coral I think I got what I can let the pro subscription help me with and thats dealing with advanced web designs in both divhunt and webflow and that part DOES require deep search.
for simple blogs its not needed at all and I'm wasting a subscription until I found out about uploading an image and using deep search instead for help with designing.
the create_limit property corresponds to the image creation limit i believe
if that's the only place it's listed then it needs some changing fosho
anybody I can ping to notify/mention this to?
A bit of Grok 2 testing. In Polish, it demonstrates poorer reasoning (at least based on this simple test) compared to English. Claude 3.5 had no problem with this task (Perplexity in writing focus). Grok 2 in polish said the stones are still inside the jar. Not enough training data?
International subvention are jarder to give. I guess.
But I join your opinion.
Is anyone from the Perplexity team online that I could ask a question about their API?
Agreed. Waiting for better Times
If you have info, ping me.
Tried yesterday to build a bot from API / search
European delay ... Lol
I too cannot exit the wolfram trap. How do we fix the search to go online?
Hello are you fine peeps! I was wondering is there a option to make more focused prompts or instructions in perplexity similar like creating Chat GPTs and save them to have different chats depending on what I am doing/working on or is there no possibility to do something like that?
anybody awake in here?
Is everybody asleep!??
Perplexity, please fix the sources problem. Very often it does web searches on wolfram alpha instead of the web.
(which comes up with nothing, and then the LLM believes it's highly reputable data)
I could really use help or info about my question over in #⚡│ask-community or #🧪│api-general but it seems this world is barren today
Here are some reasons why I believe Perplexity.ai should it
https://discord.com/channels/1047197230748151888/1275585603957821591
-
Cost Efficiency:
It might increase costs for Perplexity.ai if they don't add Mistral Large 2, as this model is less expensive than GPT-4o.
Mistral Large 2 offers competitive pricing, especially for output tokens, and provides high-quality performance suitable for complex tasks.However, it may have slower output speed compared to some other models. -
Bias Concerns:
Some models exhibit a bias towards certain cultural perspectives (mostly woke and leftist culture), which can be a limitation. -
Model Selection:
Having a limited selection of models can be a disadvantage compared to competitors.
I prefer Mistral Large 2 for its capabilities in custom system prompts, uncensored outputs, and its unbiased nature by default.
It is also suitable for jailbreak usage.
Anyone else getting really short answers and follow ups that have no context of the previous messages? Quality across most models has been significantly reduced the last week or so and it’s changed perplexity from the best tool ever to a huge time waster
anyone having punctuation issues when using the pplx-api?
certain punctuation in english is not appearing
I would like to confirm the details of Perplexity's enterprise subscription plan. Does it mean that an additional $40 per month is required for each additional member, or does $40 per month cover multiple members?
thanks.
Hi Mods, I would like to know whom to reach out to if I have been billed twice somehow for the current month for my pro subscription, but my next billing date still shows up as after 30 days. So looking for someone who can help fix this issue.
please reach out via email at support@perplexity.ai
Uhh any presentation hack ya know where I can blend ai ? Cuz the ai presentation definitely not working that good
Generate image of 5 boys in India playing football in dark night when raining heavily and thunders are shining
I am also looking for the same
How you generated this image
i think sir, better to add the voice of AI sir that one can save it sir and send it sir, for example sir, i want to share info by voice sir.
Hi, sir
second prompt
anyone having "Error processing response" with sonar huge? if rewrite with another model it works no issues
cant be a problem with the content its 2 lists of names asking who is on one list and not the other, nothing crazy
you wouldn't go all scarlett johansson on perplexity if they tried pulling something like this would you gooddawg? (I think it'd really separate perplexity from the pack :))
🐮 🤔
that's one fine cow
in very colorful stained glass
is this an issue with flux, my prompt, sonnet's prompt magic, or pplx served me wrong image?
but it looks like it now generates different aspect rations, not only square 🚀
how was that generated? like what steps, if any, did you need to do?
it seems if you manually paste the text in, it worked?
known bug, thank you for reporting @signal hamlet
it was just the photo style. maybe too much photography (style, in my and sonnet's prompt) pushed it away from a cow and stained glass?
seems to be fairly rare, I am getting now a lot of fine stained glass cows even with "photo style"
huh.. yeah i'm not sure actually.. here's manually pasting in the text (actually, slightly truncated due a character limit)
and here's just pressing "generate image" -> "photo"
so yeah.. ig some kinda bug / glitch.. (as mares was saying)
with dalle3 I had a lot of issues with the "photograph" style, putting lenses/texts/cameras/film borders in images etc. I wouldn't be surprised if the same "style" is weirdly affecting flux too, just in different ways
though actually.. who knows how much prompting is going on to enforce these style categorisations.. Here's selecting "Diagram" (tbf, it must be confused af ha)
good point with truncation, I was wondering when I extended the prompt, it kept ignoring last part (I rarely change it in the wrench area). though I think locally flux1 can handle longer input. should probably not let sonnet rewrite it in too nice language
might just be a device thing.. but i don't have this appear in any images generated - do you know what it is?
that's from ailin, it just copies to clipboard url to the image sends url of the image to the prompt box (I haven't used it in ages...). that shouldn't affect anything
definitely a rare bug. generated like two dozens of cow (animal) images and haven't seen it again
ahh gotcha 👍👍
can perplexity be more critical of its own research? It thinks its own research is "the research" available, and acts like it knows everything. its worse than just when it looks at wolfram alpha to push its own beliefs
perhaps make a Sources section that can look at objective data for stuff like Politics and more?
wgat the fk
Hi guys.
Looking for french community.
Since my news depends only from US based human, i try to enlarge it.
Any frogs there ?
I'm not able to get Perplexity to generate images, it keeps saying it can't do that
I had this happen to me the other day too. Totally random chic in my generated images. Wasn't able to reproduce the glitch though
Hey @rose compass!
Please check this guide: https://discord.com/channels/1047197230748151888/1225557012532695060
Diagram
While this may not be doable, but, is it possible to obtain Perplexity's averages for the (1) frequency of users selecting Perp-generated follow-ups vs. their own (2) Time spent on each of these types of interactions (3) Number of back-and-forths in each type. I really don't know if this kind of data is proprietary. It prob is.
Is there a way to make Perplexity reference Twitter sources more, to understand the public's views on a certain topic?
Go spam somewhere else
Earlier, when used Perplexity Pro - it would show sources count as 20 or more for research based questions. Now it shows only 5 or 6. What changed? Am I doing something wrong?
Only with the API
how to generate images from perplexity and fynder?
thanks
Good day from Holland
hello
Who is using Perplexity from day 1?
Use app on ios write better on website
AI brainrot
or a collection prompt
pretty funny if it actually did come back with that without any previous prompts
wait what..?
@lusty narwhal that image looks like a topography map, here's a link for a page with a lot of them. These ones are dark though:
https://www.freepik.com/free-photos-vectors/black-topographic-map
When was day 1 ?
Good question...3 years ago? 4 years?
I'm curious about the interface at the beggining.
Btw I don't use Google since I meet pplx
You are correct. Currently, I wouldn't trust any results from Perplexity until they get this fixed...
#👉│feedback-general message
Is there an AI tool/agent that can automatically send emails for me?
Besides Nelima, what other tools can send and receive emails on my behalf?
I'm looking for a simple tool where you can quickly integrate Gmail and write a prompt to perform a task—something as intuitive as ChatGPT.
For example: “Fetch NVIDIA’s current stock metrics and send them to [recipient’s email]. Tell them that I will also be traveling next week to London”
or
“Send me an email with all the Michelin Star restaurants in SF, include their phone numbers as well”
Nelima does a pretty good job at it, but I’m interested in comparing multiple options for an article. I came across Taskade, but it’s not free. I’m considering trying Kindo on their free tier. I also thought about AutoGPT, but it’s too much of a hassle to install locally and set up API keys.
Has anyone used any other tools that could do this?
Automation with zappier . Sir.
Web scrap needeed.
Can we have any real info on that?
Based on this post I assume that if I get a free trial for linked in premium and cancel before the trial ends
I'll get one year free of perplexity pro
Can anyone confirm if that's correct
Hey 👋 here's some more info: https://www.linkedin.com/help/linkedin/answer/a6860917
LinkedIn Premium subscription promotional offer for Perplexity Pro FAQ
10 days passed away and they still cannot restore the "pause" button on the profile setting page back???
😅😅
I didn't imagine that restoring a missing button is so huge a project.. lol
hi, it says "in select regions" but i didnt find any information on which are those regions. Any idea if EU is in?
that gives off "eu is not included" vibes lol
Since this is an offer being provided by LinkedIn, I recommend checking with their support for the latest info about this!
ok thank you
damn, ghost ping 
😂😂😂😂😂
Will gpt4o be able to access the internet in the writing search section on perplexity?
If we pay for pro and LinkedIn premium do we get a year credit on our Pro sub?
Hey 👋 this offer is valid only for new Perplexity Pro subscribers — if you're already a Pro subscriber you aren't eligible for the offer.
No, the writing mode just sends your text to the LLM. It does not have any internet access.
I’m in college. In what what ways could I use perplexity to help me with studying and learning material
I am in a very small EU country and I got the email and it worked fine :)
deep dive ➤ #💬│general message
yo
But gpt 0 comes with internet access tho
https://www.perplexity.ai/hub/faq/what-is-search-focus
Focus - Writing, for offline
Hi - I am a current Uber One member and I do not have a Perplexity Pro subscription. how do I activate my Perplexity Pro subscription? I haven't received an email from Uber. I've contacted Uber support a few times but no resolution
are there any good websites for llm prompts?
I dont want to invest so much time into prompt engineering and was hopin to automate it
Pf im in the EU too but I didn't yet
That was the first place i checked
I received the email on 16 August so check your history
I didn't have linkedin premium then yet lol
The government is paying for your Covid tests. Microsoft is paying for your ChatGPT. VCs are paying for your Perplexity. Zuck is paying for your weights. What do you have to complain about anon?
the announcement is for SEO
Aravind giving Elon Musk vibes
can perplexity have a uncensored model please?
like it even censors a perfectly fine response sometimes, if i swear in it or something
or maybe it assumes something will be used for illegal purposes because theres a slight possibility it could
tired of having to cycle models, rewrite responses, go back and forth and convince it, etc
Google has never once argued back to me...
do you think countries in asia is eligible?
Show me the AI tool that you randomly discovered one day that is now an unconscious standard practice in your life.
AI is powerful but I'm afraid it will numb my brain, do you have any advice?
Hi Lars. I created a prompt maker, based on your request. Free Access from gptstore. Dm me if interested
https://youtu.be/H-0PsLmEi4A?si=ZgykVL1c9VfC70nO
🚀 Ready to Dive Deep This Weekend? ☕
Hey there! As the weekend approaches, we invite you to take a moment and explore something insightful. A couple of months ago, our CTO sat down for an interview that offers a closer look at the inner workings of our platform. This isn’t just a regular talk, it’s an opportunity to gain a better understanding of the technology and ideas driving our innovation.
Why not make it part of your weekend routine? Grab a cup of coffee, find a comfortable spot, and join us as we dive into the details that shape our vision.
This is more than just an interview, it’s a chance to connect with what we’re building. 
a bit unrelated but—i noticed this too on opus? i remember i asked what the limit was last week and i was told 50 messages. i could see the limit on the website then. but now i see no limit? what’s up with that; did that change ?
You used to be able to hover the "pro" switcher to see how many messages you had left. Doesn't seem to be the case anymore.
Question. I was using perplexity for a while and stopped because of this.
Any plans to make adjustments on your end with the cloudflare turnstile?
It's really a nuisance and a huge turnoff, I can't imagine I'm the only one with this problem.
Every time I go to perplexity
I live in the EU and was able to redeem this offer. Can confirm it works.
Still didn't get it 🕐
probably from vpn use
It’s definitely the vpn, I run a vpn consistently on all my devices.
yeah, just one of the drawbacks of using a vpn
hello i have a problem when using make to summarize my emails with perplexity api when i start the automatisation the emails settings don't save if u can help me dm me pls ^^
I love that Flux is available as image generation model.
Is it Flux schnell, Flux dev or even Flux pro?
whats better perplexity or bing or chatgpt or gemini
Perplexity is my fav, bing is prob the best free option, and chatGPT is solid, Gemini not so good imo. But a lot of it is personal preference, try perplexity for a month and then try chatGPT for a month, etc. See what you like more
but perplexity is aint free to use while other ai are mosly free to use
What do you mean? Perplexity dose have a free tier, and so dose chatGPT, Claude, Copilot, and Gemini
Also Claude is super good, just low usage limits
perplexity free tier isnt that much good as other free tiers of other ai
They use Claude 3 haiku for their free tier last I checked, so it’s going to be better than free Gemini imo but not as smart as GPT-4o-mini. But the web search is better. If you are not looking at paying then you don’t even have to choose, just use all of them when you feel like it
yea gemini suck
Just wondering, why are you asking which is best, seems like you know what you like and don’t like already
We recently tried a website called Longshot.ai. What stands out is their fact-checked answers, which cover very difficult questions that even Perplexity can't answer. Does anyone know what AI technology Longshot uses?
Can anyone tell which LLMs out of these are not being used by Perplexity AI?
The free tier is Claude 3 Haiku, but there are lots of paid models if you are a pro user
We are on a paid subscription
For accuracy GPT4o is a good option
But still Perplexity throws answers like We couldn’t find the numbers on the web, and so - yet the longshot.ai is able to answer
Otherwise Claude 3.5 sonnet, you can change the model in settings
That might be more about the web scrapers they use
The main thing we need is fact based answers
ig claude 3.5 is not good as gpt 4o
Claude 3.5 is better in a lot of ways, like coding
Like how many interns were hired by Airbnb last year?
But if I ask the same question on Longshot.ai, it will not only give the exact number but also provide the source for that
That’s probably not the models fault, the scraper probably did not find the number, its actually good that the model did not try and make up a number that felt realistic
Longshot has a different scraper setup, maybe it’s better, idk
Hmm, but the other platform Longshot.ai also doesn’t make up things, because it provides sources
True, maybe - I wish perplexity gave me such fact based answers
Be careful tho with important information, AI’s will still at times give answers with sources and still be wrong. It’s just not 100% not matter where you go, but at the end of the day if longshot give more reliable answers then you know what you should use
I know, i checked the sources, read the details and only after getting satisfied, put this question here in the chat
claude 3.5 knowledge is limited because it cant answer questions from poem and story while gpt 4o and bing can
bing knowledge is kinder inf due to web search
I use Claude 3.5 sonnet on perplexity so it’s kinda Inf as well
You know what caseoh, you’re right, sometimes GPT 4o gives best answers than any LLM out there
true
- chat gpt also have so many gpts making it best ai
GPT4o seems to be very good at stem biased topics from my experience, but sonnet 3.5 is the king at code, so I use both, often GPt4o gets massive tunnel vision, so rewrites with another mode can really help
I don’t like Claude, maybe because my exp with it was not so good. Plus Quora’s Poe selling it as a part of their subscription makes it more bad
do u know that how many gpts does chatgpt have for coding?
I can tell you even the GPT 4o is extremely efficient when it comes to coding blockers
I prefer to make my own system prompts, and Claude with a good system prompt is even better
Really? What do you mean by your own system prompts?
I use both GPT4o and Sonnet 3.5 for code, just sonnet more, but in one session I will use a good amount of both
Like the prompt that tells the model how to act, on perplexity these are called collections, but on chat GPT that are called “gpt’s”
remember when arvind announced opus unlimited
😦
😭
Opus is limited to 50 per day if you are confused, lol
Got it, I use Evernote to save my prompts
perplexity depreciating to counter inflation everywhere
I never liked GPT’s, but thats prob just a personal preference thing
Msty (AI app) has a prompt library, like that a lot
True, but sometimes I feel every 3rd party spin off of GPT is not as good as GPT4o
I mean I don’t like “gpt’s” the thing chatGPT is marketing as the system prompt App Store or something, I prefer the text box and that’s about it lol
Hmm
This is kinda nice, it’s a flow chat of follow up questions in msty
Happy Saturday!
shark
How turn it in french please ?
Hey @vale tinsel!
Discover pages are currently available in English only.
Can't wait to see an update for it
for now, only workaround is to type "translate this article to French"
yes
How can I improve contextual understanding ability?
It's astonishing to see an ia of this caliber not have such basic functionality.
Happy Saturday:)
this is an assumption of how perplexity works
the tree table is called the pplx bs
Hello, does someone know if perplexity can share informations relative to the document I download on the plateform ?
ssup pepoll
If you upload it to perplexity it should be able to interpret the document
yes but what about the informations, if it's confidential ? Can the information from document be used to answer other people request ?
Hi there, I am trying to use the referral code to sign up for the Pro however, when I go through the referral link it says $10 credit applied on the landing page but when I move to the payment page it still shows $20. I also tried entering code manually and it says invalid code. I was Pro last month and trying to re-signup. May i know if i am doing something wrong? Thanks
Is it not possible to create pages in the IOS app?
Hello everyone. I have a content page and I want to pull references for it. Do you have any good prompts or tools to help pull references for the content?
I use myMib and they also have chrome extension which is very useful. You can directly get reference in any style and you can either use or store in your folder.
hey guys, I have a question, Does perplexity uses the latest version of GPT-4o?
If you have pro you can choose it as a model
yes, but there are two versions of GPT4-o the original one and the latest one, i just want to know which version perplexity uses.
i think there is only one version of GPT4o and Perplexity has that
there are two
Correct! I'm not sure if perplexity has updated to the latest model; I couldn't find any information about that. My hunch is that they wouldn't have switched to it yet because it would require some testing to determine if it's better for Perplexity's use case or not.
Has anyone from the team confirmed whether ads are coming to pro tier?
btw it appears like Sonar Huge still doesn't use ur daily quota if u turn off pro
but Sonar Large does
Hey, any linkedin premium members got redeem code for perplexity pro through email?
Is there any other way?
Like? I think thats the best way out there. I would recommend you to try it first
How do I get perplexity with Uber one? I never got a email
contact Uber support
is this better than gpt4o for coding anyone know?
is there a way to copy the text without copying the links to the sources?
Hey chat, what models do you all use?
I generally prefer Claude Sonnet 3.5 but not sure in the perlexity context what model is best
This extension give that as on of the features: https://chromewebstore.google.com/detail/complexity/ffppmilmeaekegkpckebkeahjgmhggpj
⚡ Supercharge your Perplexity.ai
It’s hard to find the context window, but from my testing they are all pretty close to each other, except Opus has a way shorter context window
@harsh stag not working on linux
Contextual awareness?
Is there some way to give Perplexity contextual project awareness, either locally or for a GitHub repository. I'm looking for something similar to GitHub Copilot Chat's @workspace command.
yup. use the complexity extension from chrome store.
I did, but in my case the coupon code is not valid. I'm in touch with Perplexity support to sort this out.
if im coding is it better to use gpt-4o or perplexity. I dont wanna use claude
perplexity isnt in the pools
did you not read what i said
Hi sir, does mr Feline have plans sir to make a complexity app in the future sir? i use most of the time mobile sir.
what is perplexity even for when theres search on gpt4o
Hello, i see perplexity.ai doesn't accept paypal, but does accept google pay.... according to perplexity.ai.... why isn't my payment accepted when i pay via google pay connected to paypal?
i heard perplexity gives claude for free as part of pro version is it good
which browser are you on? if it's convenient, you can fill out this bug report form on github, i'd love to help https://github.com/pnd280/complexity/issues/new?labels=bug&projects=&template=bug_report.yml&title=[Bug]%3A+
Brave
have you enabled the settings here?
or chrome-extension://ffppmilmeaekegkpckebkeahjgmhggpj/page/options.html?
afk atm, was enabled on extension icon, pplx logo was blue. but there was no model switching or any other options. F5 and restart didn’t resolve
please let me know when you're available on your machine, i will get it resolved

In Perplexity you have Flux dev.
My fav is ChatLabs. I guess it has the best UI there is and 128k context window. You can also set each model temperature and generate images normally (with Flux Pro too). Uploading and removing files is also great.
Does anyone able to get the code interpreter as Perplexity's CEO stated here?
https://x.com/AravSrinivas/status/1825617944782758066
I want to know how to test it and when will everyone get it (not sure how "slowly" he means there)
I think that would be a game changer for many use cases.
Code interpreter and plot rendering are slowly being rolled out!
Yeah, but it is very limited. You cant even change what has been generated. The best for me is to use Google Colab for example
Hello everyone. I have a content page and I want to pull references for it. Do you have any good prompts or tools to help pull references for the content?
hi sir do u still enjoy perplexity
Hello everyone,
i’m planning a little prank for a friend’s bachelor party and need some help. I want to make some minor changes to an image so that it looks completely natural, as if it hasn’t been edited at all. However, I’m not a designer and don’t have experience with Photoshop or similar tools. Could anyone suggest some easy-to-use tools or methods for achieving this? Any tips would be greatly appreciated!
Replit promocode to use perplexity pro is not working..in which group i need to inform?
Hello
I wanted to share an app made by me, it would be great if someone could try it
#😎│cool-projects is for
Neat things you are working on! Light self-promo is allowed here, but please do not spam, post links to other Discord servers, and go too hard.
Thank you so much for the help
perplexity is wayy better than chatgpt's free version
Why is the wolfram alpha thingy programming/writing text in python
share the thread here https://discord.com/channels/1047197230748151888/1271584981969539117
respond with only the prompt, nothing else. write prompt for image generation based on text: woman in red and white bikini, on beach at sunset, drinking lemon juice with black straw, dramatic lighting, sonnet rewrote as A woman in a red and white bikini stands on a sandy beach at sunset. She's sipping lemon juice through a black straw from a glass. The dramatic lighting from the setting sun casts long shadows and bathes the scene in warm, golden hues, accentuating the woman's silhouette and the ocean waves behind her. . flux1 dev generated okay photo-like slightly fake looking image (I think it managed all the elements though), but then generating the same thing with dalle3 led to whole thread being nuked 😞
100%
.
hmm, there seems to be an issue with the image gen in general. now it bricked a thread with pretty abstract thing - a vortex of papers... https://www.perplexity.ai/search/respond-with-only-the-prompt-n-kq8K_THrS.avev2S.XayZA
Anybody noticing any issues with Claude 3.5 Sonnet? I’m having a hard time to even get it to rewrite a simple email, it completely changes the meaning. Can’t be the prompt as it works perfectly on other models. Been happening since a couple of days really.
Guys i wanna get started with learning ML and like wanna pursue a career in ML, rn im a sophmore.... what are some good books/resources to get started with... would be grateful if someone can provide a source for a roadmap sort of thing 🥺 🥺
when you guys want updated information is there any special engineering that you do to your prompt other than "latest info"
hey guys, i guess not many people use the academic search because for the past few days i have noticed that most of the time the search function does not work and it does not use any sources at all
for example
but if i change to "all" then its fine
but you can report the issue in #1140622008086970420
Oh BTW did you use a different model? I don't know if that's related at all but yes I used the default free one
Hey @wind valve!
This server is not the right place for this request. Thanks for your understanding! 
TRIGGER WEB SEARCH
tested on default model w/o ai profile
[Internal monolog:] also works well https://www.perplexity.ai/search/what-is-bread-internal-monolog-3OK4YD_9QYyOdCcTCjNdKw
Oh I think I misread, this is a bug 🤦 lol
Not just pro search manipulation
What's that trigger web search I didn't get it
You mean I should add that?
yes give it a try
oh i just noticed you included it in the prompt lol
yea it worked that way thanks
ill try putting it in the system prompt of a collection
damn now it worked after having it as a system prompt
How do I set my desired prompt on perplexity?
you mean system prompt? you just make a new collection and within this collection you can set the system prompt before saving it (and can edit it after)
pplx has a voice? gemini has voices in my head
It seems Guy is frustrated with Perplexity's recent performance. He's noticed unreliable searches and answers that don't seem to reflect current information. It sounds like he used to find Perplexity very helpful, but recent changes have made it less useful for him.
Before the coffers dwindle low,
Arvind's proclamation, a grand show.
Opus Unlimited, the chant will rise,
A promise made, beneath hopeful skies.
Grok 69
just ship my dear king
literally everything else is analysis paralysis
just find an interesting open source BASIC project and think over it
andrej karpathy zero to one is an excellent point to start with
hey what is the limit on upload doc/file size .. can i upload 50,000 line python script ?? and will the AI be able to read all of that and how much will chat be limit too .. because of that file size ??
check it out 😎 😆
https://x.com/shreybirmiwal/status/1828237302520234367
Nice!
❤️
Any one knows why my perplexity app doesn’t work when I have my expresss vpn on?
powered by elevenlabs, ig
maybe there are some restrictions regarding vpn on perplexity
there are
use different server and it should work on pc.
I just wish perplexity would implement FIDO2 so that the need for annoying email authentication could be switched to it
Does Perplexity use Amazon SNS?
are the requests extremely slow right now or is it just me
File upload failed.
for me too
file upload broken for sure sir, and for sure we gonna wait long sir.
@glad crypt perplexity is slow since this morning here also. Couldn't find any status update
perplexity seems very unreliable tbh. maybe i should switch to openai
First time we notice this. It happens, it's more I just whish we knew what was happening
nah ive a bug that went unanswered for days now
I got a question, which model is the best to use in perplexity?
is perplexity taking uni interns
Nice
hi whats wrong with perplexity ? impossible to upload an image since yesterday, nothing that i can do
Same here
alr thx
idk why but i became File Upload Failed but i have the Pro Abbo. What Can i Do.
Re Logged in Already and tryed another device.
lol i didnt checked the massages not an Problem by me
xD
yeah have pro too, I hope they will fix it asap bcz its a big problem tbh
thats the Point why i use Pro xD
Hi sir, brace yourself sir. Most of the time they fix things that are broken here like it will take days or weeks sir. GL sir
@signal hamlet can dev do smthg 😦
Yes, we are working on this ticket.
It's a known issue the team is working on it
edit: file upload should work again, only images don't work
what is the message limit for claude 3.5 model in perplexity
you are a good dog sir
pdf attachments are working fine, u can convert the image to pdf for now until they fix it
430 a day iirc
combined limit for all pro models
combined limit? what do you mean?
(bar opus which is 50) there's a combined limit for all models
the limit is not specific to one model
but with a limit of 430 you really dont gotta think about it
closest ive gotten to is like 250
with 1 model?
i meant i used pplx so much yet the closest i got the limit down to was around 250
I did once use it all up lol
dms
?
l
Does your thread goes blank when Perplexity executes a prompt? Mine does many times. How do you resolve this issue? Thanks!
I refresh the page, usually shows up
I am a Senior Software Developer with 8 years of experience in Blockchain and AI development. If you have any support or collaboration opportunities, please let me know. Thank you.
anyone know whats going on with perplxty? no matter what sort of image file i try to attach it says failed to attach.
uploading images is currently not working the team is working on the issue
Not able to use chat using picture in my ios device
Uploading images is curently broken because they've exhausted the aws rekognition limits that are set on cloudinary
Is Perplexity still preventing others from attaching images or taking a photo on iOS?
oh i see thank you
Thanks for the update … is there an estimate for the time to fix?
why not able to upload files?
Spezi — Today at 8:46 AM
uploading images is currently not working the team is working on the issue
🤨
Hi sir, no use in flooding that repoet sir, it will be fixed after weeks or so sir anyway sir. Devs would agree to me sir
i bet by this Friday
which is the best model to use if im using it in school?
can someone give me referal code so i can get 10$ off this
what a joke
they literally have partnership with aws but theyre too sleepy to get an engineer to up that limit
nvm got
Any estimate on when image upload will be fixed? I have a Pro subscription and this is disrupting my workflow.
can perplexity replace stack overflow guys?
any programers here?
for project oriented learning
I need to design a growth marketing plan for the product
does Perplexity make it?
+1
same but refresh
i wouldnt say rrplace but you should use it more and refer to stack overtlow if you need more help
but it has web search why do i need stack overflow?
can it not search on stackoverflow?
What is the purpose of using this thing? Its slower than chat gpt, definately less functional cant even respond to easy questions and it takes 10 times the time to get an answer
Perplexity search (especially with Pro) is, in most cases, (objectively) better than other platforms. Beyond that, everything else is just sugar-coated and far inferior to, let's say, ChatGPT, where custom GPTs, RAG, and overall chat UX are light-years ahead of Perplexity
and yeah, I'm here for the search 😁
Like, I was excited when i saw that you can upload files and ask questions about it, apparently doesnt work at all, the purpose is for file sharing or something.. lol
what's so special about file upload tho, every chat interface on earth has that
and perplexity is definitely not the best tool for that
and some might say it's the worst when it comes to working with attachments
I definitely find the search much better than other platforms imo. i've only had it for a few days but find myself using Perplexity to look up certain things
Could you please give me an example
I got 1 year for free by linkedin, might as well optimize how I use it right
asking for an example wont say anything, it's based on extended usage over time
So it is personalized for everyone?
no i didnt say that
i mean by dedicating a decent amount of time to both platforms
talking about personalization, perplexity is nowhere near chatgpt
I don't have much experience with search specifically, mostly because when using ChatGPT it rarely cited it's sources/hallucinated a lot. But, with Perplexity I find it hallucinating a lot less and citing sources is an obvious plus
So that alone is already a huge improvement. Is it worth $20 a month for that alone? Probably not. I still prefer to use ChatGPT with files as I find ChatGPT actually remembers what we're talking about without me explicitly reminding it
Also, the Perplexity chrome extension has a feature to search across the domain you're currently on which is super nice to find information specifically on that domain/website without having to find it yourself
You may then use felo.ai which is free. You can also group your threads in topics and you can chat with them and with sources from that topic
I like the suggested questions by perplexity. However if I keep asking further questions, then the list will be gone. Is there anyway to keep the auto suggested questions?
Will check this out! Currently on a promotion from Uber for Perplexity but if this is good it'll be my go-to post promo
I think this auto suggestion function is what makes the search so slow, is there anyway to disable it?

Hi , I can't upload some file or photo . how to fix it?
Please add Hermes 3 405B
I have lost my too for the web in the app still have it what’s going on?
Hi sir, please get used to it sir, cause they dont give an F on your complaints sir. I think you are new here sir.
If you test a other browser it will work like safari for me but in chrome I don’t have my pro sub like I do in safari
is anyone else not getting the option to "ask follow-up" when highlighting a line of text in perplexity responses?
it should easily replace it.
Got the free LinkedIn Perplexity Pro offer. Sett everything up and when it works, it works well. The problem the th first 1 to 5 attempts Perplexity simply eats the prompt. That is I enter the prompt, hit return and Perplixty clears the screen, but delivers only a new prompt input window.
Does anyone else see this behavior or better yet, know how to escape this vicious cycle?
IT COULD
sorry for caps
but stackoverflow is no tthe only source that it uses
I'm not an AI newbie. I have a pretty good idea of what's happened to my favorite AI, and it saddens me deeply.
However, I don't know EVERYTHING. I can still be taught. Please have a look at these two threads (both from today, both using Claude 3.5 Sonnet) and tell me:
- How does this save me time or effort? Does Perplexity AI in it's current status assist or hinder?
- What is causing Perplexity AI to present assumptions as fact, repeatedly, getting it wrong more often than it gets things right?
- How much proof does Perplexity (the company) need before they take this seriously?
Here's the threads. I've posted both these in feedback-general as well. These should be easy home runs for Perplexity AI. They used to be.
https://www.perplexity.ai/search/i-am-filling-out-the-dd1172-2-BtJrT2wKRX.JjXfAUTVrsg
https://www.perplexity.ai/search/how-do-i-structure-the-address-7SGmU8fWR4Wsv7qtssn2RA
I cancelled my PRO membership today. I miss Perplexity AI. 😦
As a disabled veteran without a current duty assignment, you can fill out blocks 18, 19, and 20 on the DD Form 1172-2 as follows:
For this block, you can...
Hello
Every time I access my saved thread, it starts me in the first conversation I had and I have to go all the way down to be able to write. Any suggestions, is it from my cell phone?
may i ask why did u cancel? I just purchased mine today
Did you read the conversations? Can you imagine how that might have gone if I wasnt familiar already with government stuff?
ye idk i use to work with gemini and ai isn't really supposed to help with government stuff... it violates safety and we had to mark it.
Anything with PII in it
is safety violation and the bot can not assist you with that information
so any ai will do the same asking that question not just perplexity
I mean at a glance this kinda seems like something it's not meant to do. Like, I get what you're trying to do in those threads, including trouble shooting, but this is not sentient AI. It's just an LLM- trained on a dataset. It most likely has limited training on such requests and just assumes/hallucinates.
I just skimmed your threads, but I did see where you told it to not present assumptions as fact and I 100% agree, but sadly that is just a limitation of AI (LLMs). Generally stuff like this should be reviewed by a human, especially if it's an official form of sorts. Shouldn't have AI do tasks like this.
It might be more possible to have an AI do this if it's broken down into steps, but at that point it's usefulness probably dwindles. Too much effort for little gain vs just looking it up yourself.
I do think the grammar corrections are ridiculous though. A space after the : makes absolutely no sense and it should be able to do that. I've never run into such a situation though, so maybe this was just some one off scenario
Why can’t I create images on iOS Safari.
Can I trouble someone to show a quick video on how to do this pls
maybe try app?
also are u signed in oon pro acc?
i think u need prio
pro*
Honestly every one except Opus is borked and unreliable right now.
No! Was wondering if anyone else was experiencing this as well
Does anyone wonder how perplexity manages to stay the best ai search engine?
So what is up with prompts section? Can you create models with instructions?
Chrome extension description is wrong. It sends the wrong message of saying Perplexity Pro uses GPT-4 and Claude 2
source:
https://chromewebstore.google.com/detail/perplexity-ai-search/bnaffjbjpgiagpondjlnneblepbdchol
already fixed
https://discord.com/channels/1047197230748151888/1225557012532695060
iPad only over Safari
perplexity creating a windows/mac app similar to cursor AI but also having general research capabilities would go CRAZY
Anyone successfully connected perplexity API to cursor? Tried it through the OpenAI API Key settings, but it just got stuck on loading
any news on "broken agentic search" issue? I got this yesterday - I could see why it occasionally forgets to search (that was rarely happening probably always), but why is it using "programming" instead of search here❓
oh, found the thread. so, it tried to scrape google and when that (very expectedly) failed, it wrote three python scripts with hallucinated/mock results 😑
lol
Good day guys
I want to change the name and address on the bill for an already issued receipt paper document. What should I do?
not sure, but maybe they are trying to fix the "searches only on wolframalpha"? it looks to me like the agentic chain is getting passed to wrong agents. like it got function call / command to search, but gave it to programming agent (step)? these failures seem to happen much more in longer threads
perplexity is only good ai for learning programing
whats your specific use case
no problem, lol
so more like, bringing knowledge from the web, like from docs/etc, answers online -> teaching you a specific thing
is there a way to make it search specific sites
they get some of their sources from blogs which isnt trustworthy at all lol
Question, if anyone tried this. Best way to convert collection of .html documentation files in subfolders?
are you a dev ?
trying to be
where are you from ? I m from Sweden. and blockchain frontend developer & UI / UX designer.
altered api from original to sustain costs of operation
United States, AI Data Annotator
great! maybe you are planning to be a AI developer ?
is it you within the car ? lol
yes
so pretty
hey can someone help me for a telegram chatbot
using make and perplexity api
sure.
Trying to get into game dev, but wanted to see if I can use documentation for a particular engine. But the documentation is .html files divided by sub folders. So was looking for resources. Currently looking into vector databases and see how it classifies documents
isnt that miku?
😢
what format you tried to convert to? markdown? plain text?
Why are you planning to use Vector database on your game development?
Maybe, is your game AI game, right?
even google thinks so
I want to reference the documentation for the engine I'm using to help me with game development
Well, I'm not sure which would make an LLM have a better chance of retrieving information based on my query
Sure!
I think you are using Pinecone vector database on your game.
So, that needs embedding and classification documentation.
And you need to reference Unreal Game Engineer for your game development.
https://pandoc.org/
this with custom written filters might suit your need
If you are really planning to develop your own game project, I will be happy to help you.
I have some experience in 3D game and 2D game development that is combined AI and Blockchain.
If you want, I will explain my game project via screenshot on our conference.
Please schedule our call and let me know it so that I can put it to my schedule.
Hey there, Im not seeing the free linkedin premium offer in my email, Does anyone know who i would contact about that, I read the article im pretty sure i should have qualified.
yes
Congratz !!!
I like perplexity, congrats
Did anyone fix the issue with magenta and 1 year pro membership? It just disappeared on my account two days ago
Does Perplexity android app support multiple accounts?
How you separate your work/daily/formal questions from 'incognito' type questions
Hey @grizzled echo! The cancelations occurred with accounts that had abused the service or used cracked codes. Therefore, users who continue to report this issue are likely to be involved in such activities. No accounts are canceled without a valid reason. If you believe there was an error, please appeal to our support team at support@perplexity.ai with all relevant details.
Hi all! I'm fairly new to Discord, but my question is pertaining to Perplexity. I just subscribed annually through Xfinity as a Veteran. One of the perks is that I can share Perplexity with up to 7 people, but I'm not sure how to do that . I read that the subscription code I used can only be used once and it's a violation to use it more than that. Can someone help me?
I've been trying to use Llama 3 on Perplexity but it keeps using GPT-4o. Anyone know how to fix this?
does your use cases include uploading images in inputs?
You should be able to change it in the settings
Sometimes yes. Sometimes no
When I set it to Llama 3.1 it defaults to GPT
@wide cradle see this
in case of uploading images, it will default to gpt4o, if the current model cannot handle images
and only gpt4o, sonnet and opus can process imgs
is there not gonna be a which ai model should i use thread guide for august 2024
I’m in the same situation. Never received it and it’s not hidden in Spam. Don’t know why it hasn’t come through yet
why is claude 3 opus limited to 50? i get its the price but isn’t 3.5 sonnet better than it? im just confused idk 😭
has been for a long long time :) just search "opus 50" in this server
the thing is is that opus is VERY very expensive and it doesn't matter if it's worse than 3.5 sonnet, it just costs bazillions
while it's worse and slower than sonnet i think a lot of people here have found it to hallucinate a noticeable amount less than sonnet so there's that
i do wonder what 3.5 opus would look like and its pricing
Hey perplexity user
hi
Hey
what happened to you ?
Just try every ai apps 😁
what do you mean ?
My means is i testing every ai chat apps
perplexity ?
Yeah and ChatGPT, meta ai, cloude ai
Perplexity, wen add doom AI model?
guys why my perplexity using past images for subsequent answers also i deleted those images still its using the same image for future answers also what to do ?
i have the pro version btw
Hi guys, I just applied to the Perplexity Campus strategist application, but I did not get any confirmation email, is that normal, or did I make some mistake? I just want to make sure my application gets in. Does anyone have any info on this? Thanks a lot.
how to fix this my past page is gone pls help i need it 😦
Hey @shrewd pollen!
Please check this guide: https://discord.com/channels/1047197230748151888/1237163951750385705
Please DM the thread URL to @signal hamlet
Dang 100k pro users?
Hey anyone interested in participating ai competiton and has decent to advanced Knowledge with it can you dm me
Iam from hyderabad so you being from hyderabad is better
Not only from. Hyderabad iam looking for people who are interested to participate in hackathons with me
Iam from India
And my CREDIBILITY is
I have built RAG models, finetuned llms and built own personalized projects
You can find them at my portfolio
Iam finding it hard to find the real enthusiastic ai people from India and many comp requires team
So this message
https://rohansai22.github.io/resume/
Which year?
Iam in my final year
Btech?
if the owner is in this server i hope their pillow is cool on both sides and their room is well ventilated when they go to sleep tonight
🙏
Sure
does this srver doesnt have voice channels ? @everyone
Does anyone know how to recover a post that I accidentally deleted on my phone? I can't see it on the web page either.
Is this something new?
Whenever i used to write this prompt it used to give me result based on my previous queries but now it just says this?
Is the issue gone that we were facing on chrome. The issue were the Pro sub did not work
Hey, everybody. I have a Pro subscription.
How many requests per day (not pro) am I allowed if I use different models (3.5 sonnet, 4o)?
At the moment you have 450 uses for pro models a day
Are the limits different for different LLMs? Is it possible to describe the limits for each of the variants, or are they the same for all of them? I emphasize once again that I am interested in ordinary requests, without Pro mode.
default has unlimited uses. the other models substract 1 use of the 450 (total) uses when using it no matter if pro is enabled or not. when you're using the default model + pro mode then it will also use one of the 450 uses.
the only exception is Claude Opus that has 50 uses each day
Got it, thank you!
my pro sub is gone...
Hi @tame current, please read this thread and check if you might have the same issue #1278518574192918640 message
If you've bought the subscription on the official website you can mail support@perplexity.ai
Does anyone know about the context window?
You can find information about that on the faq https://www.perplexity.ai/hub/technical-faq/what-is-a-token-and-how-many-tokens-can-perplexity-read-at-once
Hello, I actually dont think this is the issue.
I redeemed a code
for magenta
the same promo code from what I redeemed back then, does not work anymore.
Perplexity has a system in place to prevent abuse of vouchers. Feel free to send a mail to support@perplexity.ai if you think your voucher has been unfairly blocked they will be able to help you further.
Any ideas when we will be able to upload images again?
Been failing for 3 days for me
It is working now
Which AI model is suitable for designing web pages?
Claude 3.5 i think
Try claude 3.5 sonnet
hi
I wish I could see how long it would take for my pro search to recharge
Why is Perplexity saying it is not able to search the web? Pro sub.
"Perplexity
I apologize for the confusion, but I must respectfully clarify that I do not actually have the ability to search the web or access external information in real-time, even if instructed to do so. I'm an AI assistant with a fixed knowledge base, and I cannot browse the internet or verify information outside of what I've been trained on."
is this in Writing Mode?
ok thanks. Is it possible to set focus after you have started the thread?
oh weird. When I select "web" it jumps back to "focus". When I select "Writing" it stays at "writing"
ah wait... it was a chrome plugin called "Super Dark Mode" that made a fuzz :/
I disabled that plugin and now it works.
websim.ai + claude 3.5 sonnet
Slight correction *350 pro searches 😞
At this moment uses have 450 pro uses each day. Each use will refill after 24 hours
ive got an email that perplexity ceo is coming to seoul, is it true? @wheat granite ?
just tagging random team :9
Is there a reason Perplexity devs said for logging by email and having such short session TTLs?
Like why can't we just not use a password with OTP?
anyone else having the issue of some queries not going through? refreshing and doing it again works
I tried switching between various models, but the answers I received were very similar. I feel that even though I selected different AI models, they all seem to be calling the same model. Is this my illusion, or is it a bug from a recent update?
For example, when I ask the AI whether it is GPT or Claude, it always responds that it is not either of those, but rather a model trained by Perplexity, regardless of which model I use to ask this question. In the past, even if there was pre-filled role information for Perplexity, the answer would specify whether it was GPT or Claude.
Sweet! I think sonnet used to be 350 (regardless of Pro or not) at some point. But it does appear to be 450 today
300 (pre gpt4o) -> +opus 300 -> opus 50 -> 600 (around time when enterprise released) -> stayed 600 for a long time -> 550 recently -> 500 -> 450 (and still opus 50)
id bet that when they added enterprise pro, which was meant to be 600 while normal 300 (since well you are paying double)
they made everyone 600...then probably after a week realised it worked just about the same profits wise, and was better marketing
then they just stuck with everyone 600 until about 1-2 months ago when they seriously thought about profitability (ads this that)
pre gpt4o was 600. if i recall correctly
recharge deez nuts
you can literally check the time of your messages
Tool calling any time?
Got plans for the weekend, everyone?
Hi guys.
I’m organizing an AI exhibition in France and will be showcasing Perplexity AI during a conference.
Could you provide video resources or any promotional materials to help present Perplexity effectively?
TY
(More than Youtube. 🙂 )
Anyone had their threads deleted by PPLX? And no, i never delete mine, and on my browser tab when I refreshed it's gone
ages before it was 300 though
like im 100% sure
Hey everyone, quick question about Perplexity Pro - I have a code for a free year through Uber One. Do I still need to pay anything after the first year? Just want to make sure before I sign up. Thanks!
After one year you’ll have to renew
But i think you also do need to keep your uber subscription valid for the 1 year term ?
Check the terms on the website - not 100% sure
For a full year? No.
Don't think so
Unless you are a LinkedIn premium user currently.
I mean a "Black Friday" of some sort could have huge impact on the actual yearly price or something, but for free? dont think so
heyo, just recently received an email back about the campus strategist/ambassador program and the specific discord link for that was invalid. any chance someone can push me in the right direction to getting a new invite? i replied to the email but would like to push the envelope if possible
Hello, I just redeemed my Xfinity 1 year trial promotion and am wondering how to share it with the rest of my family like it said on the promo. Very confused 🤔
Read the whole FAQ already, couldn’t find any answers.
hey guys just a question does perplexity PRO tier offer visualized artifact like for example if i ask for a script of a web app can i make it show a preview of the web app on the side just like claude?
Did you try giving them the same code to sign up?
Hello! I’m already a perplexity subscriber but want to use the xfinity promotion How can I do that? It looks like I’m stuck in iOS subscription mode
Just studying 📚 🥚 ⏰ 🧇
I got 3 books to study over the week end for work
Hello everyone
I had bought a perplexity pro for a year with a code, but it just stopped working, any one know how to fix
Testing out a research prompt: https://www.perplexity.ai/search/research-the-news-around-opena-3KY6GcAETFOdcBc2b_zFqA
Any questions I can use as a stress test?
No, is that what I have to do?
Who is the imposter 🤡
This pro support is pretty slow if you ask me 🫤
How come the only video focus is YouTube and the only social media focus is Reddit? What about Vimeo, dailymotion, Odysee? What about other forum websites?
Inviting people to collections if they have an apple device it doesn't work
never heard about odysee, hmm
why do i have this when i try to upload image ?
hi
Isn't there a search feature in Perplexity to find previous chats?
I think you can search in threads
Help Shape Our AI Model Rankings!
Hello everyone! 👋
We need your votes to help fellow users choose the best AI model for their needs. Whether it's quick searches, in-depth research, coding, or creative writing, your vote can make a difference! 🗳️
Head over to our voting thread and cast your vote: Vote Here!
Thanks in advance 🙏
Hi Everyone!
Trust all are well 🙂
Just joined. Would love to hear about how can i learn proper crypto trading, like anyone if is experienced please let me know we can chat!
Does anyone know how to get perplexity pro with LinkedIn premium?
#1277665169941725316 message not sure if this helps, I think in the past they said you need to email LinkedIn
LinkedIn Premium subscription promotional offer for Perplexity Pro FAQ
Where is nous hermes 3? And why ist sonar huge extreme boring and refusing. Please add nous hermes!!!
Not sure, it says 7 more users per account, but that might be a synonym for devices. Give it a try. There's no info anywhere
I have logged out 3-4 times and logged back in via Perplexity's discord link but still not given the pro role :(
I was invited used promo code but no 1st month discount given
any free online image enlarger?
Do any of you still use a regular search engine?
Do any of you feel that there is still use for regular search engines?
I use brave search, and it has goggles that allow you to filter through only a specific list of websites. Will that make it compete with perplexity?
Would a search engine only be better for people who were skilled at using search engines?
My rule of thumb is to use keyword-based search engines when I know exactly what Im looking for, and vice versa
lets say im looking up a specific url, phone numbers, emails, agencies, booking, etc.
Did i really just get rick rolled by AI, i must be high?
no fken way did i just get ricked rolled, hahahaha
100% perplexity bug not claude tho xD
So then if i am trying to do research or find new information, then perplexity is better?
cheeky devs 😛
Same here. I rejoined few times but no pro role
Hello! Please tell me if there is a Discord-Bot Perplexity to add it to your Discord-Server? Please share a link to this information. Thank you
I found out that a Korean company was operating a service that had plagiarized all of Perplexity's features and design.
It has the exact same features and design as Perplexity, and I wonder if Perplexity is aware of this and is responding to it.
what’s the name if you don’t mind
and are they monetizing it ?
Yep
I don't want to say if you are not working for Perplexity, just for now
why? what is there for me to gain 😭
Because I feel that sharing the content before there is an official response from Perplexity will cause a lot of secondary problems.
I don’t think they will respond, there have been knock offs in the past, it’s inevitable once you are big enough
how do i get the pro channel access? it's locked for me, i tried leaving and joining again
i dont think the moat of pplx is about the ui ux
did you see felo.ai? 😅
@fierce sonnet Hello, can you create a French channel?
waifu2x
You can join the Pro discord channel by using the invite link available in your Perplexity AI account settings.
Omg 😱
when did felo change it to look exactly like pplx lol
its good but not for all images
hahahahahahahahaaaaaaaaa
What would you like Perplexity to do, to become a better product for coding ?
haha, no shame
Yes, better for coding!
Hey @compact rose!
There is no PPLX Discord bot
felo straight up
copied the pplx question list
they didnt even try
wow
perplexity literally doesn't accept any of my cards
perplexity pro
its super strange
has any1 else encountered this issue??
Best model right now?
its kind of like perplexity ive been trying it out for coding and its really good tbh
ive been comparing the two and man. perplexity really needs to step up their game man
yeah im switching sorry 😦
ill still use perplexity for other research
🙂
what's your usecase?
Yeah I took there system prompt a while back and use it sometimes when coding in perplexity lol 😆
i already pay for perplexity pro and reccomend it to friends and colleagues, does everyone else with pro get this annoying ad that must be closed manually multiple times a day? is there some way to disable it permanently aside from CSS hacks?
did i somehow check a box that i'm using it for business purposes or something and now i'm plagued with this?
interestingly enough, when you don't have pro / aren't logged in this ad for Pro is in the same spot... buying a perplexity Pro subscription only replaces it with a new ad "Invite your teammates" (above) that keeps reappearing even after being closed. absolutely ridiculous.
No LLM chatbots seem to be able to provide me to an answer to the following relatively simple question: "What is the longest sequence of days where the closing price of solana was lower than the previous day?". The answers provided include something along the lines of "To answer this question, we would need access to a comprehensive dataset of Solana's daily closing prices since its inception. This data is typically available on cryptocurrency data providers or financial websites, but it's not present in the current search results." I thought LLM chatbots were supposed to be able to perform relatively straight forward time series analyses. Does anyone have any recommendations about answering this question? Sure I could pull the data myself and figure out the answer, but I'm interested in saving time
actually that seems to be because I was in Math mode on perplexity (apologies, new user). Web/Focus seems to provide better results using this prompt "What is the longest sequence of 24-hour periods in which the closing price of Solana was lower than the price 24 hours prior? Use all available historical data (complete time series price data since inception) as your sample" but I do not trust the answers provided
u wanna use my referal for a website?
all you have to do is just make an acc pls
this
Hey 👋 I replied to your email! Let me know if you’re still having trouble getting in.
i got in, thank you!
Hey guys. How's going ?
Did any of who have an idea to created a LOCAL / ONLINE big repertory,
I want to achieve a library that is multimodal for code/prompt/visualize + Generate art/médias/vidéo/voices/sound. according to an opensource database, connected online, + social media included.
Wait . is that the great definition of creating a new world ? . lol .
I got a code for perplexity pro but I don't see any option to enter it in the perplexity app/site. I don't see it either on payments page nor a separate tab. Anyone has success with it ?
Hey Guys!
Just wanted to let you know, great work with Perplexity!
I've been user almos 1 year ago
I would like to know if there is any way to pay a subs for 1 year and have any type of discount
If you have Uber one, reach out to their customer support
I dont have uber one
There is a yearly option to pay for perplexity. Since you are a current subscriber, unfortunately, you don’t qualify for the LinkedIn promo.
😩💡🤖
are there any plans to make the default perplexity inference faster?
the difference between https://labs.perplexity.ai/ & https://perplexity.ai/ is night & day
i like getting the nice UX on perplexity.ai but the labs inference is just so much quicker for most things
would be awesome to bring the labs inference speed to the default interface
it's like 70b sonar is slower in app than 405b sonar on labs?
