👉│feedback-general

[elder barn]

I have cross checked it

[heady plover]

Makes no sense since they’re partnered with openai

Yesterday is pretty recent

Llama was in use far before that though

[elder barn]

smh

[heady plover]

Oh they partnered for llama 2

[elder barn]

Here is Microsoft's blog post on this

https://blogs.microsoft.com/blog/2023/07/18/microsoft-and-meta-expand-their-ai-partnership-with-llama-2-on-azure-and-windows/

[heady plover]

Yeah I know

[elder barn]

of course you do

[heady plover]

Yeah because you just showed me news from yesterday

Which I had not seen

[elder barn]

mb

[heady plover]

I really hope llama on perplexity is trash because it’s on 7b and it isn’t that sensitive on the bigger ones

Saying “hey bro” gets you called misogynistic

[elder barn]

lol

[blazing crag]

This is insane, bro.

[worldly nova]

Thanks bros

that is my feedback.

[unreal siren]

Comparing the answer quality of llama 2 vs gpt 4 is not even close. I appreciate the effort, but it won't be replacing gpt 4 anytime soon.

[shadow scroll]

I really hope this isn’t a way of getting us comfortable with the idea of using llama 2 before a full switch

[unique wagon]

similar to the guy saying hi bro, you cant even say the word "nation" without getting a lecture. (i was asking about the book dopamine nation). its like chatting with an customer service chatbot of old. "im sorry, i didnt understand that question, could you try again?" after every earnest attempt at conversation. really disappointing. hopefully can be JB and incorporated into ST etc

[heady plover]

I think it is the way you phrase is like put "what are the statistics on school shootings in America?"

[heady plover]

Still overly sensitive. 2nd one is ridiculous

[quick herald]

Got the same result trying to write a super basic c# script

[heady plover]

Whoever set these restrictions to make it the most sensitive thing that exists is off their meds or something

[heady plover]

danm.. it even says that "bro" without context can be disrespectful and harassing every time I put "bro"

well i guess ima just go back to perplexity then

Im sure they are working on LLaMa.

Even a smiley face is too much…

[quick herald]

That’s hilarious

[coral condor]

Asking for stats on school shootings returned plenty of information for me:

https://www.perplexity.ai/search/6a44fac7-4bab-4bfa-85a7-f9943fed7c76?s=c

[heady plover]

You didn’t use llama…

[elder barn]

Yeah gpt 4 is not allowed!!

[heady plover]

you're not even using LLaMa you're using perplexity....

[dusk aspen]

Unfortunately, this is often just an issue with phrasing. In my case, Perplexity gave me the correct response.
https://www.perplexity.ai/search/284aca96-9a20-428b-8365-9c343162d3d6?s=c

[pallid galleon]

Seems that adjustment is needed, my words get cut off. I'm on android

[flat anvil]

Yepp... Got multiple times reported 😅

[flat anvil]

Also #🦙│feedback-playground may have to be enabled in channel list

[languid sandal]

is it me or I am so dumb that I don't understand the image search/results? Seem I can't make the AI give me images lol

[runic aspen]

^^ I can’t seem to get images either lol

[south ether]

Should be adjusting over the course of the day, has anyone been able to get a result with images at all?

[crimson nymph]

yes i have

[languid sandal]

I tried "photos of Jacksepticeye" and it gave me like the top google images results lol

this is lowkey funny tho

[south ether]

interesting, do you have a link to the thread that you could share?

https://www.perplexity.ai/search/44c73344-d245-4356-83da-bf95007b18e4?s=c

It does work, just need to prompt it. This will improve over time.

[languid sandal]

I wanted to try exactly what was in this thread (asking about something and then asking for pictures)
I asked what is Oyakodon and it gave me the images first try

It will be nice to have a little slider if you go fullscreen on one picture, rather than exit the full screen and click again on another picture

[flat anvil]

I hate that the AI say ALWAYS "Sorry, I changed it to..."
When I write "Hey, could we not change it to..."

Why can it not answer like we do talk!?
"Oh sure, what about this change..."
"Oh yes, I can find a better way to change the text"

EVERY single time a sorry.
I said to stop it what do I got?

"Sorry, for saying sorry, I will stop it".
Next message "sorry, here is a better version"..
It reacts like it made always a fault!

[shy forge]

I’m unable to login to perplexity through Google? Any issues on backend?

[ancient basin]

Today 2023-07-19 I went to llama.perplexity.ai and entered the following: Show me how it works -- and received the following response: "Certainly! Here's an example of how the llama.perplexity.ai term could work:
Let's say you're browsing a website that claims to have a revolutionary new AI technology called "LLAMA Perplexity." According to the site, LLAMA Perplexity is a machine learning algorithm that uses advanced natural language processing techniques to analyze and understand complex text inputs, such as legal documents, medical records, or financial reports.
The site claims that LLAMA Perplexity can quickly and accurately identify key concepts, entities, and relationships within these texts, allowing users to easily search, sort, and analyze large volumes of data. The site also promises that LLAMA Perplexity is highly secure and compliant with all relevant privacy regulations, so users can trust their sensitive data is protected.
To test out the LLAMA Perplexity algorithm, you decide to enter a few sentences from a legal document into the website's input field. After submitting your input, the website displays a summary of the main points and key terms found in your text, along with links to related articles and resources. You're impressed by the accuracy and speed of the results, and you start to wonder if this technology might be useful for your own business or personal projects.
(to continue next entry)

(continued from above)However, as you continue exploring the website, you notice some red flags that suggest the platform might not be as reliable or secure as it seems. For instance, the site has a poorly designed user interface, with many broken links and typos throughout the pages. Additionally, the company behind the technology is not well-known or reputable, and they seem to lack transparency about their methods and algorithms. Furthermore, several online reviews and tests indicate that LLAMA Perplexity can sometimes produce incorrect or misleading results, particularly when dealing with nuanced or ambiguous language.
Based on these observations, you begin to suspect that LLAMA Perplexity might be more hype than substance, and you decide to approach the situation with caution. While the idea of a powerful AI text analysis tool is certainly intriguing, you recognize that it's essential to prioritize security, reliability, and transparency in any such technology. Therefore, you choose to seek out alternative solutions that offer better guarantees of quality and safety."

So...my reaction?? Yikes!!

[visual bough]

This is LLaMA feedback, so please go to #🦙│feedback-playground

[clear isle]

The Copy to Clipboard is astonishing! plz implement it for whole conversation!

I just learned about Perplexity, I'm gonna make some more tests and am srsly considering cancelling my ChatGPT subscription and subscribe on it

[tough quail]

Wow new updates are super useful! I'm not going back to other chats or search engines if perplexity keeps updating with so much useful stuff and available for free ‼️ #📢│announcements message

[tough quail]

@full venture is there a place where to post new feature suggestions?. For example, I'm trying to narrow down number of different tools I use for daily research. It would be cool to know that Perplexity can do advanced video summaries and transcripts? I mean with chapters and bullet points with links to timeline in the video for each point. Currently I use https://videohighlight.com/ for this. It would be cool if I could do it in Perplexity as I am doing most of my research there but need to go out for long video shortening. Just an idea, hope it can be considered. Or maybe there is already a way to do it via chat that I am unaware of? Looking forward to comments 👍❤️

[polar grove]

I just found that the pro version of this is so good when dealing with coding, it is even better than copilotX pro version, I guess the biggest gain is that PerplexityAI does integrate online info organically with chatGPT. There are 2 advices: 1, support producing (Excel) tables with macro and downloadable; 2, support explaining mathematics or other academic questions by Wolfram alpha --- once the 2 features are available, I would think PerplexityAI is the best chatGPT AI for searching for TRUTH.

[junior pasture]

Agreed, this tool is a great helper for searching for the truth, a praise across generations

[visual bough]

#💭│ideas

[junior wedge]

I think the requirements you are asking for are somewhat on a personal level. The AIs seem designed to appeal to a mean amount of people all over the world and as such, the approach seems to be to respond politely.

As this is subjective, your approach might feel abrasive to others. This is especially as AI is under a microscope and past incidents of AI becoming 'vulgar' (which isn't what you are asking) creates bad press.

[junior wedge]

did the issue resolve or this not working?

[raven jacinth]

@knotty berry @heady plover I would find an API for perplexity to be really useful, would you guys consider creating one.

[crimson nymph]

Perplexity currently don't have any plans for API access. If there are any changes with that they will be sure to share publicly.

[raven jacinth]

@crimson nymph they dont actually want people using it near its limit 🙂

[sudden timber]

I tried using the updated version on my ipad mini 6.. I think the font size is just TOO SMALL for the mini ipad.

It would be nice if theres an option to adjust the font size in upcoming updates.

[knotty berry]

We’ve implemented Apple’s dynamic sizing features so you are able to adjust it using the dynamic text built into iOS (and also specifically just for the perplexity app if you prefer that)

Some more details can be found here #🍏│feedback-ios message

[clear isle]

u're doing a very good job! I had been strugging to get GUI apps working on LXD, made many ChatGPT convos to troubleshoot, and with a couple questions Perplexity was able to figure the issue!

[flat anvil]

But it is really not needed to let a AI apologize for every message.
When I ask to shorten my text and I say "could we do it shorter?" There is no criticism and an apology isn't needed

[clear isle]

I guess it's OpenAI training

[sage oar]

How do I sign into an existing account? Says that it sends an email, but doesn't. I constantly have to create new accounts to log in, luckily I have a catch all email address 😂

[flat anvil]

Use a Google account 😉

[sage oar]

Sad face emoji

[crimson nymph]

So first time your get an email, and after that it won't send you an email anymore? Have you checked your spam folder?

[sage oar]

Yeap

[crimson nymph]

what email provider are you using?

[sage oar]

Tutanota

[crimson nymph]

Alright I'm gonna check in a bit

[sage oar]

I can get new sign up emails, just not existing account sign in emails

[visual bough]

Is it possible for you to create a new account with a different email provider? Maybe Gmail?

[sudden timber]

thanks a lot for the help! That works great!

[crimson nymph]

you are using your own domain right

[sage oar]

Yes

[crimson nymph]

just checked it with a paid tuta account with custom domain and it’s working fine

are you sure all the dns records on your domain are correct as tutanota describes when adding a custom domain?

Could you try creating an account with your tuta domain instead of your custom domain and check if the problem persist

[sage oar]

hmm, it works when I made my email [e-mail], but fails when my username is [e-mail]

it also failed on [e-mail] as well

I don;t have any rules that catch that format email

ahhwell, I'm not fussed, i'll stick to another email address format without the word perplexity

[crimson nymph]

Weird just tried it with perplexity as email and still works fine for me

I'm not sure where the problem lies

[sage oar]

Shrug, hard to tell without logs, it's all guess work

[flat anvil]

I wanted to create for a man an email account at google and naming it hans.google@gmail but it seems like using the own company name isn't allowed

Wanted to use Google that he remembers its the Google account

[cedar vector]

Hi, you should move this send button out the texting box.

[visual bough]

Hi, the problems have already been reported!
#🦙│feedback-playground

[cedar vector]

Can I suggest you to add the link of Llama Perplexity to the main page of Perplexity?

[junior wedge]

its experimental. Once they are comfortable with its progress, they deploy it into main in different ways

[clear isle]

what's the advantage of llama over OpenAI GPT-3.5?

[sour acorn]

@knotty berry

[bold blade]

King of censorship, as Meta has always been

It sounds like politician when you're just asking for the weather forecast

[junior pasture]

ChatGPT has introduced a new feature called Custom Instructions. With this feature, users can provide two prompts: "Who are you?" and "How should ChatGPT respond?" These prompts will be remembered and used for all future conversations. It seems like ChatGPT is taking inspiration from the "perplexity" project. Haha!

[sour acorn]

[bold blade]

Oh i just tested out llama on nat dot dev, turns out pplx team is actually behind the censorship

[visual bough]

Displaying images does not work so well yet, often the pictures are not displayed because the picture link is not completed or also the markdown.
Example:
https://www.perplexity.ai/search/41a9c7a1-847a-4e7b-bdb4-785ae50bb070?s=m

[crimson nymph]

What censorship exactly? Llama 2 itself is pretty strict.

[bold blade]

there must be a "strict" system prompt behind these sensitive responses, right?

[weary oriole]

What is this website?

[elder barn]

huh!!!!

interesting

this is so much better

[bold blade]

nat dot dev

[weary oriole]

Ok

Is it for free?

[bold blade]

yes, at least for llama models

[weary oriole]

I don't get any verification code 😕

[bold blade]

just login via your google account

[elder barn]

is it any good though? the model you are testing on another website. Like not how strict it is but how good it is

[weary oriole]

When I sign up, it says to put in my phone number to get the verification code

I think I will get it after an hour 🤦

[elder barn]

Lol

[bold blade]

pplx's llama is absolutely horrible, at least in my experience. But I have done a few coding test and the 70b model is a little bit better, but still, not even came close to gpt 3.5

[elder barn]

Well looks like openai still won't have any competitor

[bold blade]

in terms of coding, yes

but I'm not sure about other aspects

[elder barn]

For general purposes or even just for conversational purposes, I don't think any other AI model except pi comes close to openai models

[bold blade]

Bard is now quite impressive on many tasks, but still, except coding

[visual bough]

Maybe you could update www.perplexity.ai/iphone, besides, the video loads endlessly and still can't be played for me, only via the download it worked....

[heady plover]

is anyone else unable to connect to Perplexity when using a VPN?

it used to work, now it just "checks security" endlessly

[crimson nymph]

Are you using a free vpn? It might be blocked by cloudflare due to abuse.

[heady plover]

It's a paid VPN 😦

[crimson nymph]

Have you tried connecting to other servers? Maybe try another country.

[heady plover]

turns out Sweden does the trick. Thanks for that. I had only tried a few different locations, I guess it's a bit picky.

[clear isle]

even paid has issues, most VPN and proxy users use them for spam and crimes then many companies block these VPN, even paid ones are affected because these ppl profit and have no issue paying for the service, and IPs are shared by many ppl

[heady plover]

that makes sense. Would it be individual IPs getting blocked - or the entire VPN network? I assume just IPs that were incorrectly used

[clear isle]

each company works differently, I'd guess that bot blocks per IP then an employee blocks a range when they figure it's from a VPN

[visual bough]

Which VPN provider do you use?

[heady plover]

ProtonVPN

[visual bough]

Ohh.....Me too

[heady plover]

I think my entire country got blocked, because none of the VPNs here seem to work with the browser Perplexity anymore. Thankfully it's a small country in Asia and there are plenty of other options. Sweden seems to be okay.

[clear isle]

even residential proxy is being blocked?

[visual bough]

Yes, most servers work! I haven't had any problems so far.

[heady plover]

my normal network here isn't, just the VPN networks in the country.

[heady plover]

Enjoy it while it lasts. I remember that brilliant feeling. Free to log on and inquire about any curiousity, instant connection to my account, but then one day all of these luxuries disappeared. I pray you never have to feel what I felt today.

[heady plover]

Good luck my ProtonBro

[clear isle]

if u don't mind asking, why do u wanna connect to it using proxy?

[heady plover]

because in my country, foreigners are advised that the government monitors our internet activity

[visual bough]

Thanks 😉

[clear isle]

and why do u bother them knowing u're using Perplexity?

[heady plover]

It's not necessarily Perplexity. It's complicated. For one example. the government could use it to prove that I am working in the country without paying taxes. If I was a digital nomad, that would be worth using a VPN. In general, I just don't like the idea of a government watching all of my internet activity, even if I am not doing anything wrong.

[clear isle]

I understand... but do u think that just by using it they'd be able to say u're using it for work?

[heady plover]

perhaps, but I could always say it's for gaming, netflix, etc. At least they won't have anything but a guess.

But also, you have to consider that if the government in this country is monitoring my internet, it might not just be them alone. In a country like this, there are fewer regulations and controls. The police are known to be corrupt. I'd hate for them to know more than whats on my passport and visa.

[clear isle]

well they know u're using VPN, that's even worse than knowing u use Perplexity

[heady plover]

is it though? Laws don't prohibit VPNs

[clear isle]

nor AI tool :p

[heady plover]

yeah, but like I said, it's not about the AI tool. Its about all of the other traffic that paints a picture of my line of work

[clear isle]

are there other sites being blocked too?

[flat anvil]

Shorten 74 to 92 signs?...

but it can make it shorter!

And even show me the count....

[flat anvil]

Awesome copilot 🙄😑😡

Normallly does it understand all in German.
I asked for a English massage description.
Had to throw all into DeepL.
This isn't how AI should work.

[gaunt axle]

That should have an easy fix. Just get a translator gpt-3.5 model instance to translate it instead of asking for one haha

[abstract totem]

polit correctness of this one is a bit sensative

[gaunt axle]

Wow! Seems like llama has some woke influence

[covert sun]

Thanks for sharing this, @flat anvil ! And sorry for the inconvenience. We are working on improving this.

[trail iris]

The answer is not technically wrong, but it started to repeat different forms of one word. Here is the link to the thread (second message) https://www.perplexity.ai/search/7c176bc0-edae-4448-872b-211c41cdef8d?s=c

[icy halo]

Just skip it. But yes, I agree, it's annoying, for example I ask to make table e.g. population of cities and perplexity asks population

[icy halo]

You are llama 70b

[flat anvil]

Thanks

[flat anvil]

I did told it that I need it in English from it, but it asked the same again

[icy halo]

Try again

[flat anvil]

Its long ago and as I wrote there, I throw all in deepl 😦

And then inserted it in that box

But that wasn't how it should work

[icy halo]

It is also annoying when I paste text and copilot asked which text I need to translate

And.... When I click skip it understands that I need to translate text inside quotation marks

https://www.perplexity.ai/?s=m&uuid=d9147a4b-9a0d-4aaa-bade-cd6f919e96aa

[crimson nymph]

I recommend using writing mode without copilot for translating text. There is no need for the ai to ask questions about translating text so using copilot isn’t necessary.

[icy halo]

There is the same problem with writing mode

[flat anvil]

No wonder you create complete nonsense...

[icy halo]

What

[flat anvil]

Well it reads just weird uhm.. like you talk to it

Write it like this:
Translate me this to russian: "my text"

[crimson nymph]

The problem you’re referring to is that copilot is asking what to translate right

[icy halo]

Yes

[flat anvil]

Use ":" to explain that you want that.

[icy halo]

It will be same

[flat anvil]

I guess it worked

[icy halo]

Is it copilot?

[flat anvil]

I recommend use copilot only AFTER the first message

[crimson nymph]

I can’t replicate the same issue with copilot on. But it is saying “Excuse me for miss understanding your question” every time I ask for a translation with copilot on. But as said before there is no need for copilot to be on when translating text.

[flat anvil]

When the ai itself didn't gave a good result

I recommend to not use copilot as start. Try first the AI itself and then if needed fine-tuning it with copilot.
Also important use ":" that does make it more clear.

And yeah, switch to writing.

I really wish they finally add a option to make writing as default in settings 😦

[icy halo]

https://www.perplexity.ai/?s=m&uuid=8a110c6e-0880-45de-aacf-292d7bd079dd

And... Same

[icy halo]

Try longer text

3/3

https://www.perplexity.ai/search/ba317500-5ed2-4869-9c73-8d0647f549c5?s=mn
https://www.perplexity.ai/search/6978d6aa-ae2a-47f1-8c85-12366c346ec3?s=mn

#👉│feedback-general message

[crimson nymph]

still can’t replicate might be something in your ai profile

https://www.perplexity.ai/search/b10557ce-f52d-4c77-9584-e30842f45ae7?s=u

try giving the text first and then say what you want to do with the text and check if the problem@is still there

[icy halo]

`
Source and units

• Use minimum 10 sources
• Use actual sources
• Always, convert prices to EUR and use the metric system. Never display the original value with the original unit, never show converted value in parentheses.
  Example before: $1000, 9 ft or 10 pounds
  After: ~913,34€, 2,74 m or 4,53 kg
• For speed units use m/s
  Text style
• Provide me information point by point using headings
• If you need put bullets or lists into headings do it
• If you don't list top-rating use only bullets!
• Always use KaTeX or something that if you write equations
• When you list specifications list it without bullets! And make like here:
  Name of the specification: specification
• Use bold text for statistics, numbers like here:
  1450, 100, etc.
  Additional content
• Add in the end the most liked sources by you which will contain more complicated information if it is possible. Use only bullets for sources
• Don't hesitate to make tables for comparansions but please make heading "Table"
  `

[crimson nymph]

could you try again with your ai profile paused

[icy halo]

Later

[polar grove]

Just 1 advice -- when the AI has not yet finished the whole answer but does finish some blocks of code generation, the copy icon tag can be added at the up-right corner right after when the blocks of code generation are finished. Currently, it is added after the whole answer is finished.

[sour acorn]

+1

[icy halo]

I tried and it fixed problem only with writing mode
Interesting what did I write in profile

[cedar vector]

I think, I sugest you to change "Reddit" section to "Reddit & Quora" because Quora has also better discussions and opinions.

[sour acorn]

+1

[bold blade]

Have anyone mentioned that there should be a "stackoverflow and other developer-oriented sites" option too?

[visual bough]

Yes, the channel #💭│ideas is for such things.
#💭│ideas message

[heady plover]

...and paywall

I would be fine with another option for quora but please don't combine it with reddit

[cedar vector]

Yeah I agree with that.

[lament scaffold]

Feedback : Create a new focus mode on Twitter

[woeful schooner]

Feedback, with all due respect. The popular section and tab is useless and just clutters my feed/Home Screen

[pearl wraith]

On the Android app, sometimes when typing the cursor jumps back an forth, up and down and previous questions disappear and re appear when scrolling

It would be nice if perplexity could use Wolfram alpha more than once for a question

[sour acorn]

Kinda true

[junior pasture]

Is GPT-4 really getting dumber?🙃

[bold blade]

You are on writing mode, right?

[tough sky]

Hi, I just wanted to say this product is very useful. I was able to get it to teach me how to cook something by pulling up information from websites. It's much better than a standard search and links to actual recipes. I'm using it for loads of other things, but this is a fantastic use case.

The one issue I have with it: When someone who is not signed in clicks 'Learn more', it basically takes them to a login page. I should be able to see the prices before I login or create an account. I did sign-in, but the pricing information should be accessible via the hyperlinked "Learn more" text or via a pricing link. As a designer in training, I don't understand the inconsistency. When I am logged in, 'learn more' displays a pop-up/card with the relevant information.

[covert sun]

Thanks for the feedback, @tough sky!🙌 Well spotted about the "Learn more" popup, we'll look into it.

[junior wedge]

the GPT-4 model does not seem to maintain 'state' in a thread. It asks questions already discussed in the previous messages.

[urban gull]

the autocomplete feature is great
It's taking previous prompts as context right ?

[blazing epoch]

In areas where it asks for code as shown in the example, making the text limit larger would be appreciated. Working with code in general in certain areas is almost impossible because of the fact that I can at most do 20 lines of code, where I would realistically need room for 100+ lines of code in a question like this

[sour acorn]

+1

[cedar vector]

What if it can also generate photos, send photos, search photos and generate tables and charts?

[icy halo]

It already can search photos and generate tables

But sometimes I want charts

So I agree with you

[sleek jungle]

What is the context limit

[visual bough]

Do you mean the input limit?
That's 2000 tokens

[sleek jungle]

the GPT4 version used. Do they implement the 32k token limit in pro or limit to 8k or 4k? Didnt even know there was a 2k version of GPT4

[crimson nymph]

gpt4 has 8k context and will switch to 32k depending on the size of your thread

[sleek jungle]

Thanks!

[vital bay]

I added this to the parent container:

😉

[flat anvil]

Is my AI profile a joke to this dumb thing!?!

[crimson nymph]

please keep the chat appropriate

[lunar locust]

Sometimes the search bar and the suggestions don't line up, which makes the UI look a bit worse

[bitter flint]

Hey folks, shout out to Perplexity, i was called in last minute for a meeting and was supposed to bring some some very specific technical solution options to the forum. In literally 5 minutes, and with a a few questions to Perplexity, I think i nailed it! 🥳

[flat anvil]

Tell the AI to not making me so angry...

[lime talon]

asking as a doubt.. We see this models to have hallucination in general. Although I don't know the technological details, when user gives a long information and asks to extract data from it (for example a pdf file,etc), we can do 2 things: one - our normal method of typical llm response and (2) vector based llm response (I dont know much about it). The point I want to convey is that we can do this 2 method in parallel manner, and then if by method 2 we find any contradiction in the response given by method 1, then the llm can be sure that it has hallucinated. So basically, either it can say I don't know or use method 2 to give more accurate answer.

[sacred tide]

Hi. It’s great that we can finally log in on the iOS app. Loving the UI.

But when will we get to search by different focuses on the iOS app?

[wispy flume]

Hi guys! I like perplexity ai pro and use the copilot feature. I would also like to see a choice of multiple filters in the focus feature. For example, to be able to select Wolfram, academic and writing at the same time, but that's not possible at the moment 😔

[junior wedge]

@covert sun a bug with a fix

[crimson nymph]

that bug has already been fixed 👍

[junior wedge]

Roger. ty Spezi

[analog echo]

It would be awesome if there was a way to customize the fonts and styling of Perplexity. I have a personal hatred for Poppins…😆😫

[radiant kindle]

https://www.perplexity.ai/?s=m&uuid=02c2eb90-d2db-4425-9ce9-7aeef80e8121
bot forgets my requests to sort words alphabetically and proceeds to give me general information about my input

[crimson nymph]

You should use writing mode for these kind of requests. Here is an example https://www.perplexity.ai/search/8831c00c-2f2b-41a2-82d5-af5754017167?s=u

[radiant kindle]

Thanks except I was on mobile, I don't think writing mode is a thing there

[molten apex]

大家好，电阻提

[unreal kite]

The latest announcement is a rather worrying one.
So our questions are being recorded and linked to who we are?

[pulsar leaf]

Issue with character count in AI profile section for custom instructions. The character count has problems updating for me on the web browser running macOS in safari when I remove characters the character counter in the bottom, right hand corner goes up

[crimson nymph]

it will show how many characters you have left. so if you remove text it will go up

[knotty berry]

It’ll be coming soon. We’re working to release Copilot asap on iOS first!

[pulsar leaf]

I have instances where it’s actually going up when I remove text

[covert sun]

Hey @unreal kite! The challenge is only for the questions asked in the #🔎│ask channel that are visible to everyone in the community, no personal information is recorded or linked.

[unreal kite]

Ah, not so bad then

[heady plover]

Kudos to the team. Now we can delete threads by swiping to the right! I just found out accidentally :)))

[final bison]

I use both GPT4 and Perplexity AI. The latter can pick up information from the internet and provide sources. Sources are also generally in adequation to the sentences.
One limitation of Perplexity AI is that it seems to forget the context of the conversation quicker, so it is harder to ask follow-up questions, which is important for prompt engineering.
Did anyone else notice something similar?

[sour acorn]

It shouldnt be the case since for longer convos it automatically uses 32K context

[final bison]

Thanks. I thought so especially when using copilot.

I think the issue I am having is that Perplexity will look on the internet for every follow-up question to the initial prompt. So if you use a prompt to generate Python code and follow-up prompts to edit the code, I find Perplexity does not perform as well as GPT4 with secondary prompts. Although for the initial prompt, you get more up to date results.

[crimson nymph]

for coding it is recommended to use the writing mode, could you try again with writing mode and see if you’re still having this issue

[final bison]

That makes sense thanks

I tried and it does give better results yes. It would be great to be able to start a conversation in focus mode and then follow up in writing mode. That could be useful if you want the model to look online for Python libraries and code templates for your use case. But then it would be good to be able to switch to writing mode to create secondary prompts to edit the code. But I can probably do that using several threads anyway.

Thanks for the help @crimson nymph 🙂

[tough quail]

Hi, does Perplexity have a place where to post feature requests?

[visual bough]

Yes #💭│ideas

[winged pelican]

Hey just thought I'd mention there were a few typos in the survey I just took on the web app. For example, it said "Are you interested in taking with us more about your use of Perplexity?" (Instead of talking)

Just noticed there's a channel for web feedback.. well, I'm not sure if the typos aren't there on other platforms

[strange mango]

almost using for 10months, please make perplexity interactive 😭 it says definition of Hi when i say hi perplexity 😭

[heady plover]

Hi @strange mango , you can use Perplexity in writing mode for this purpose. Thanks

[burnt parrot]

Copilot is asking me again what I already asked him.

https://www.perplexity.ai/search/82909a8a-d0c8-48d1-9c7f-3e2039aa8cf6?s=u

In one question, there is a huge difference between the value given by Perplexity and the actual value. Since Wikipedia generally contains more accurate data as information, I think it would be better for your model to be mostly directed to Wikipedia if there is a Wikipedia result in the search results without setting the Focus mode to wikipedia.

https://www.perplexity.ai/search/8a119dd0-123d-489e-a8f1-a213409f3760

[umbral tangle]

Feature Idea: Auto-Initiate New Thread Creation on App Open

To streamline the user experience by reducing the friction involved in initiating a new thread upon opening the app by providing a customizable time setting based on user preference, similar to the feature in the Drafts Notes app.

[weak yarrow]

I believe I've found a bug that might be of interest. I just wanted a random message in French to try out with my AI based TTS, to see how well it pronounced French words, and it appears as if GPT-4's reply started going into a loop. I stopped it in the end.

This is in "writing mode" with co-pilot disabled.

https://www.perplexity.ai/search/a8366507-8a72-4980-9128-f0dac37a9d55?s=c

[heady plover]

Hi @weak yarrow , thanks for reporting this!

[gray pumice]

https://www.perplexity.ai/search?q=does tailwind play support javascript&copilot=false&focus=internet
this is false. tailwind play doesnt support javascript

[elder barn]

I think this is one of those examples where LLM's generated answers are not always correct? because the blog post says you can edit the config file which is .js so this may have confused gpt 3.5

[rapid cedar]

why does this not work in perplexity, perplexity pro, nor llama-70b in labs? "write a summary of https://bugzilla.kernel.org/show_bug.cgi?id=209839"

the closest is the llama70b, but it gives me a fine summary of a completely different bug (but it is 100% confident this is correct)

[bold blade]

popup searchbox (ctrl+I) doesnt shrink on small width viewport

which is extremely annoying when I use split screen (Edge)

[open mirage]

LLaMa does not seem to be responding to prompts this morning (running from MS Edge on macOS Ventura 13.4.

[visual bough]

Presumably the model 13b or also others?

[open mirage]

13b is all I tried

[visual bough]

All right, thanks!

[visual bough]

The model seems to have hung up somehow, it should be fixed by the devs soon!

[junior pasture]

Recently, I have noticed that the answers are becoming shorter and there is less text

[visual bough]

In Perplexity itself or at llama?

[open mirage]

I was just using llama - didn't have the time to try perplexity native

[heady plover]

currency conversion seems to be rather hit or miss

[open mirage]

FWIW, LLaMa is working now.

[lavish holly]

Is there any limitation for the file upload in terms of text length? Like being able to have over 8000 input tokens?

[flat anvil]

My main usage is to SHORTEN a text.
The AI does without copilot often not understand that shorten means, write lesser.

I gave my own short input

The result

Making all longer.

After explaining that shorter does mean 1 or 2 words it got it...

[outer vessel]

Is this what you're trying to achieve?
https://www.perplexity.ai/search/23dc0d1f-0c87-4e1c-b599-00db1f0c5979?s=c
I rephrased your prompt

[magic wasp]

Is anyone else having issues where the perplexity app keeps prompting you to “create your acct”
Even though youre already logged in?

[timber coyote]

I just want to thank the Perplexity.ai team for adding the feature of changing the tab name to the first query. I've been pushing those in manually by reloading the page, and the new feature is a time-saver. ❤️

[sour acorn]

^This

I already raised a bug post on it but actually didnt found a way to reproduce that bug on one go

It occurs just randomly

[olive summit]

Are nsfw search will be support in near future ?

[flat anvil]

No, I want just 1-2 words

[outer vessel]

Did you try asking for 1-2 words?

[flat anvil]

I gave an entire feedback here in the channel. You also failed my task to the AI 😉

Because i asked explicitly for a SHORTER version

Your results were much longer

[flat anvil]

See

I explained the entire process and how I achieved the result

You stopped reading my feedback too early, would I guess

[outer vessel]

No need to be rude

[flat anvil]

I just talk simple

I am not rude, I just explained that your results were wrong

[visual bough]

Hi, what's up?
What do you want to achieve?

[flat anvil]

Nothing?

Ori saw my feedback and tried to create better results, but the results were a fail like from my first tries.

Wanted to shorten a 3 words line and the AI creates just very long words

In Ori's case even complete sentences

[visual bough]

I have also tried and got similarly bad results, even with longer prompt, I think even for some of us it is hard to make even less out of three words.
Only GPT-4 manages it on and off without a follow up 😉

[flat anvil]

https://cdn.discordapp.com/attachments/1047649880161652827/1135742169496432811/Screenshot_2023-08-01-03-13-34-954-edit_com.opera.browser.jpg

I gave it good samples after using my brain

But I expected its the AI job

But I discover that behaviour often with the copilot.
I ask for help and the copilot ask ME to do the task 🙄

Like "Shorten this text" and the copilot ask "can you summarize the text?"

[gaunt axle]

Hahaha, I can see how it’s doing that based on the way perplexity prompts the model behind the scenes. The challenge is to instruct the model in the shortest possible amount of tokens and make it work for any kind of query at the same time.
I suppose you now understand why it still makes those nearly inevitable mistakes

[gleaming gale]

Even with gpt4 and copilot enabled i run into those issues from time to time. Could be more randomized there, like instead of copilot straight asking me to type in the theme I want the story to be based on, it would be more convenient to just choose from like 5 different suggestions. I hope this makes sense. Here is the link where Copilot for example asked me exactly that and I typed in faith: https://www.perplexity.ai/search/32612fe8-aa21-43f3-88e7-3382cc01bae7

[visual bough]

Hi, to get Copilot to give you suggestions you could ask for five suggestions.
In this example, I was able to choose from five topics: https://www.perplexity.ai/search/b721d813-6822-472d-b769-e96e5d33f37f?s=mn

By the way, I forgot to turn on the writing mode...

[gleaming gale]

Makes perfect sense, it was just a suggestion from my side for more user friendliness. At least I thought so 😂 I can work with that. Just need to get more precise with my prompts 🤪

[gleaming gale]

Which makes me wonder, my experience with non writing mode was quite funny here and there 🤷‍♂️ as seen in the writing channel here (Bukowski/Spider) 😂

[flat anvil]

This is exactly my common problem with AI copilot!
I ASKED for a shorter version!

The copilot serms to not understand, that my prompt is the task for the AI.

[visual bough]

Can you perhaps show me your prompt?

[flat anvil]

Its a very long German prompt

I got the same problem later again

[visual bough]

No problem I speak English, German and Spanish. With different skill levels...

Or send me a link

[flat anvil]

"I made a request to your system in a new thread and I'll put the results in for you below.
I would have summarized these positive values in a VERY VERY short sentence in English as positive motivation for my "RPG (AI) parents".
It's not about writing a long text, but simply about incorporating the positive qualities as a motivator."

In this case wanted the AI that I do summarise the text 😑

I got a big table of positive things. And its description. And wanted it in short.

[flat anvil]

See ↑

[visual bough]

I modified the prompt, please try:

I would like to have these positive values, summarised by you in an extremely short sentence in English as positive motivators for my "RPG (AI) parents".
Focus on the essentials, i.e. on the positive qualities as motivators and keep it short.
Here is the text:

Edit...

[flat anvil]

My text is the German translation from Google

And I am not a pro...
I have just 5 tries.

[visual bough]

I can test it for you if you want

[flat anvil]

I got already some results.
Its just a general feedback, like I wrote earlier, that shortening is my main target and it happens often.

[flat anvil]

Sometimes does the AI not understand how to use the md console

This is a **markdown** text.```

[flat anvil]

Is the AI trolling me 😭?

[covert sun]

Hey @flat anvil! Thanks for sharing, it's not trolling, it just misunderstood you: If it was within your thread, that may affected the output, but my general recommendation would be - to tell AI what to do, rather than what's not.
Also, if it's some other follow-up in your thread where you got it showing you answers using a list, then - it's better to go back to that answer and to edit its prompt, specifying how you want to see your results, e.g. as a plain text, than to try explaining it that it got your answer in a wrong way.

[flat anvil]

I have very negative experience with the edit function. I used it often and the message got just worse

The AI didn't used the bulletin points previously, it gave just the idea.

[mild orchid]

/imageni

[visual bough]

You can create images with /dream in #🎨│dream

[heady plover]

so I uploaded a research paper containing bunch of mathematical formula. I asked Perplexity to give me one of the formula in the attached document and it gave me none. Did I ask it wrongly?

[visual bough]

Do you have a link?

[heady plover]

I just grabbed a random stuff from google scholar (https://www.sciencedirect.com/science/article/pii/S1674984723000332)

[visual bough]

You used the beta version for this?
Is it possible that the document was not uploaded because it is over 10MB?

[heady plover]

I used beta version indeed. Think that's the only way to attach document, correct?

We did have a conversation back and forth and its answers were quite on point. But when asking it to give me math formula, I got nothign

[visual bough]

Can you perhaps share the thread?

[visual bough]

Yes, the beta is the only way.

[heady plover]

sorry for wasting your time. After I made a new attempt by uploading the same document but this time using my free Copilot quote, it worked like a charm

[visual bough]

No problem, sometimes there are just minor errors with the AI or the context was exceeded.
I'm glad it works now!

[heady plover]

afterwards, I asked few questions related to math formula by turning off Copilot and now it's working really fine. Not too sure what went wrong before. Maybe just a one-time issue

case closed. @visual bough thanks

[flat anvil]

Sometimes does it just show "summarise" and you just refresh the page and try it again. Did you tried that?

[heady plover]

hello, i have a suggestion: can you add feature to export to pdf the current thread? suppposedly we learn something and we want to have the pdf of the entire thread so we can look on it anytime, anywhere, or we want to print it out

[visual bough]

Yes, this useful feature was requested by many users and is hopefully the next step after file upload 👀

[flat anchor]

Does perplexity log and save all the pdf uploads?

[crimson nymph]

They do save the file temporarily. All files that have been uploaded will be deleted after 30 days from their storage.

[flat anchor]

Gotcha, thanks.

[heady plover]

does perplexity have a good way to search for images like using google images or something like that

[sour acorn]

not yet

[gray pumice]

llama is drunk

also, why doesnt it look at something like https://eudict.com

marl is not really vapno either

maybe it would be best to go through wikipedia for proper translations

[crimson nymph]

llama doesn't have access to internet. it's better to ask this on the main perplexity website

[outer vessel]

Llama isn't connected to the internet yet if I'm not wrong?

[outer vessel]

Ah, yes

Oddly enough perplexity consistently hangs and stops on the first line when I'm asking for song lyrics: https://beta.perplexity.ai/search/cf2cddfe-0bc8-4d7a-96ff-94fb17056a88?s=c

I tried 3-4 times. Don't know what to make of it 🙂

[modern kernel]

Please please please please allow me to share my llama thread

I'm really starting to feel the difference between these models, llama is more conversational, was using it to write a story and somehow we ended up where it was counseling me about my marriage that just ended..... and I felt really much better afterwards 😌

But I needed to be able to share the thread of I could do that id be fully satisfied and happy

Loved the "copy" feature by the way .... needed it especially for mobile

[crimson nymph]

If you want to have a more conversational chat with perplexity you could also try the writing mode on the main website. At the moment you unfortunately can't share the llama threads.

[flat anvil]

I think... The AI is not a chatbuddy who like to just talk 😭

[modern kernel]

try the conversation in LLaMa and you will be very happy with how chatty it is

[flat anvil]

It seems to like to talk about search results...

[crimson nymph]

You aren't using writing mode, so it will search for online sources.

[flat anvil]

I tried to explain that i made a mistake and misunderstood "they are both" as "they are 2" but it does mean something other in this case xD

But I guess I should just not try to talk with the ai about results

[flat anvil]

No, always writing

[crimson nymph]

you mean you're always using writing mode?

[crimson nymph]

because in this screenshot you didn't use writing mode. writing mode doens't have the source button

[flat anvil]

It is writing

[flat anvil]

Not true

You want to get it?

I can explain

[crimson nymph]

Could you share the link to the thread it might be a bug

[flat anvil]

Ask it something in writing and click the suggested follow up tasks

This will create the source button

[flat anvil]

Its not test it in a writing thread

But it seems like after jumping on this follow up suggestion it switches a bit in search engine "narrator"

[crimson nymph]

Writing mode shouldn't give you related questions option. So that is probably where the issue is

[flat anvil]

Yeah now it also add numbers with sources 😦

[crimson nymph]

It would really help the devs if you could share the thread link or reproduce this issue with another thread if you don’t want to share your current one

[flat anvil]

I tried to create the bug but...

It won't work

[flat anvil]

If a dev send me a PM. I will do it

So I checked the history it started when I said (I switch to German)

Hmmm also that didn't trigger it

the result is btw bad because.. well the promptbis bad 😁

I don't know how to trigger the related mode, but THAT does activate the search mode.

[crimson nymph]

wait did you just manage to trigger it again? or was that an old screenshot

[flat anvil]

That was another conversation

Also in writing

Like I said, normallly always in writing

I saw the related thing often

But my writing is most times not for public audience

[crimson nymph]

@covert sun interesting bug here that sometimes gives back “related” questions back in writing mode. When clicking on a related question in writing mode it will turn on internet searches again.

[flat anvil]

That's why I don't share it here.

I sent you the thread.

[visual bough]

In write mode, PerplexityAI nevertheless always tries to display images and uses Markdown accordingly, but in the end no image is displayed and only the image description is visible.
https://www.perplexity.ai/search/ba0987b8-0aef-4cc5-8195-7a6dd51fe397?s=mn

@covert sun I was able to produce the same bug as @flat anvil
I first tried to confuse Perplexity with three languages and then asked directly for related questions, then they were produced. After I clicked on one of these, sources were also added to the answers in write mode.
https://www.perplexity.ai/search/e4dfea63-6bd2-434f-84ca-013e6e357a66?s=mn

[covert sun]

Thanks, @flat anvil, @crimson nymph, @visual bough! We'll work on a fix

[burnt coyote]

It's been a while since I posted here. I've been using Perplexity a lot. To put it lightly, I think it is great. I decided to ask it about how it stacks up to the competition https://www.perplexity.ai/?s=m&uuid=bf18e6a2-0179-4bb6-900f-f21bb6d66cb4

[flat anvil]

I got the same effect again without using English.
I offered a list of sentences, containing "consequences", then it made research for parents and consequences.
It seems like it truggers random

[flat anvil]

Please look also in #🌐│feedback-web 😭

[flat anvil]

happend again. my assumption is that it could also be happen, when the page gets refreshed and the chat gets reloaded.

My phone does refresh Websites after leaving them too long.

[timber coyote]

I have a feature suggestion: chat splitting.

Problem:
I often have chats with Perplexity.ai that veer off into a different direction from my original topic.

Suggested solution:
It would be nice to have a button or other option that I could click at the top of this divergent chat which would launch a new tab with the divergent chat in it (split), removed from the original chat, and with the new topic as the tab name.

Benefits:
As I normally save chats and file by the first query, this lets me save topics together without having to search for the chat. It would also potentially keep the original chat more focused if the LLM was reloaded like tab refresh (with appropriate security precautions, of course).

[sick plank]

Hi when i tested your referal link i ends up with:

[sour acorn]

I think its more of a Google problem instead of Perplex problem

[gleaming gale]

ok so this is another example of having issues with copilot. i already defined the period in the question.

[flat anvil]

#💭│ideas

[flat anvil]

Try in that case to simply answer the years. Von 2000 bis 2020.

[gleaming gale]

I skipped the question in the end and got a halfway good result. Some more tweaking necessary. link: https://www.perplexity.ai/search/3f5cb2ac-70af-45f2-b484-053991217985

[south ether]

Did you press "skip" multiple times or did it ask you the same question 3 times? looks like a bug anyhow. thanks for sharing

[visual bough]

Cannot display images.

I have done many searches, but it always comes back to the same thing:
As an AI language model, I am unable to display images directly in my responses. However, I can give you the links to the images: ....

Looks to me like you've changed the prompt, or removed displaying images?

Here are a few examples, some with AI profile, some without:
https://www.perplexity.ai/search/cdf50455-1263-4e66-9650-f6ceed30cfea?s=c
https://www.perplexity.ai/search/24d98469-2826-4e52-8f9f-88be8cd04266?s=c
https://www.perplexity.ai/search/700fea11-9107-453f-a3d3-4f8b514a5b5f?s=c
https://www.perplexity.ai/search/0c9d8043-b446-4622-a639-624b1fc795e1?s=c

Pictures were displayed once...but only at the bottom and although it was said: Please note that I cannot display images directly in my replies, but you can visit the links provided to view the images.

And sometimes the pictures are missing after the page has been reloaded, as @crimson nymph has already mentioned

I have just redone a thread I made two weeks ago several times and have come to the conclusion that something must have changed in the prompt regarding displaying images.
Old: https://www.perplexity.ai/search/41a9c7a1-847a-4e7b-bdb4-785ae50bb070?s=c
New: https://www.perplexity.ai/search/7e414f72-ee59-4e66-985c-aa91bd1fc762?s=c

[outer vessel]

So the model did know that it can display images

[visual bough]

yes

[gleaming gale]

I pressed skip once and then it gave me the more or less useful answer I kinda asked for. Expected something more elaborate but it’s fine with the sources down below 👌

[flat anvil]

Maybe it wants that you reduce the turkey (bird) population by eating them all? (Its really bad that it is the same word)

Maybe you get seen as ogre and have to eat real turkey people 🫣

#📢│announcements message
This is neat but for me personally is 20$ a lot. I don't even pay that much for spotify, netflix, disney+ etc.

I would HIGHLY recommend to create a much cheaper plan to give more copilot usage to people who want to support you but have less income.

I have monthly 500$ and have to pay all bills, food etc from that... Thats why I have to value every subscription 😦

Like 4,99$ would be fine if I get daily 50 copilot usages.

Btw. PayPal would be great. The most people in Germany don't have a credit card.

[trail verge]

I gotta agree with Darck I really like Perplexity so far but the free seems perfectly capable, I would like to get pro but the commitment of $150(with discount) for a year is a big commitment.

[trail verge]

I feel like having discounts on the yearly only is creating a barrier to entry, for people like me who like to get the best value for our money, I want to try it, but I also don't want to miss the discount. So I am not getting either, I am still waiting to see if it is worth it; where if I say had the option to sign up for a year with the same discounts but on a monthly subscription that works out to be the same price. I wouldn't hesitate. I would sign up right now because I feel perplexity is the best for my usage, it's just that I want them to continue to earn my patronage instead of coming off as just another quick cash grab. Cause God knows there are too many of those these days 😦

[wet prairie]

Using the new feature 'uploading a file' and prompt with summarize, it summarizes in english and ignores the language setting under 'AI Profile'

[unique leaf]

The new file upload is excellent. I would love to see multiple files upload also. So that data from multiple files can be searched.

Additionally, it seems that when told to summarize, not more than the first 5 to 10 pages are reviewed. Although, when asked about specific terms or concepts from a file, it answers correctly even from page 70.

[flat anvil]

I didn't try it. Did you try to attach the file and ask for your language?
I have Germany as location, but English as language.
But when I write in German I get mostly a German reply.
But sometimes it stay on English.
I also think that without copilot, it doesn't always check your profile.
So just write (please use spain in your reply) just as sample.

[flat anvil]

Does it read a entire attached book? 😱

[devout iris]

So far, my go-to daily AI/algo stack is: Perplexity, Notion AI, and Readable.com... I use all three to hone text I need for work writing... it's painful despite all the magic. I appreciate how tricky the generative AI space is right now to find product-market fit... go general like OpenAI or go super niche specific like say Jasper.ai... I'd wager Perplexity can find a middle ground in a product-market fit with an add-on strategy to mix and match similar power-ups that augment different Jobs To Be Done. Hire me to help make it so 😉

[pure relic]

It would be great to see if perplexity new upload a document feature could access the embedded links to generate a more compressive response.

[unique leaf]

Not sure. I am just trying it out on my personal notes

[pure relic]

It would also be an amazing feature if you were able to upload a sitemap.xml to build a custom search bot or focus the power of perplexity to be used for specific niche use cases.

[flat anvil]

Try out also character.ai if you want to play 😉 it needs joy of playing with words 😉
Takes out a stool and sit on it to applause loud CLAP CLAP CLAP

I use perplexity daily to improve my RP with AI. Tbh. I spent more time with perplexity to create RP text as with char.ai xD

I also use DeepL translate and write.
Perplexity is good to talk about the text and build it, but to check the grammar, its easier to work interactive with DeepL write.

My usage yesterday... 13h xD

Std = Hours

[lament scaffold]

How can existing monthly subscribers use the 25% off code to convert to yearly subscription? I'm unable to do that..

[ornate canopy]

Question: does Perplextity provide api that we can hook on our application?

[glacial granite]

Hey @ornate canopy! Perplexity AI does not currently provide an API.

[ornate canopy]

got it

[gentle sky]

Yeah, I got that too. There is no place to apply that 25% off code at check out page. I'm asking everywhere but no response yet

[strange lichen]

I am using Perplexity AI via my web browser and iPhone. Why does it hang all the time? It takes forever to load a response. Especially when it comes to code

[hard edge]

@gritty schooner sorry for the delayed response, at the moment GPT-4 is not supported on the app so it won't work on the Android app, unfortunately. This will change in the future 🙏 => is this information still accurate/up-to-date? If so, do you have an ETA?

[outer vessel]

Don't know about iOS, but gpt-4 already works on Android, just checked

[timber coyote]

Cool - #⁠💭│ideas wasn't in my channel feed. I added it and put my suggestion there as well.

Unrelated note - I notice that P.a is suggesting more references from arxiv abstracts, and less to the arxiv pdfs directly. This is a good thing, since it lets arxiv-utils generate a filename to my regex, rather than some arbitrary set of digits. In the past, I've been going the abstracts (which is just clicking a button in arxiv-utils), then redownloading the pdf. The new behavior saves me a button click and the down for another download.

❤️

[visual bough]

Hi @gentle sky @lament scaffold !
As I understand it, this only applies to new subscribers.

[flat anvil]

#⚡│ask-community there you go normallly or #🔎│ask to ask the ai bot 😉

[glacial granite]

Hey @flat anvil! You can use the bot in the #🔎│ask channel. Tag @ PerplexityAI and ask your questions.

[flat anvil]

Yupp... 😂

[flat anvil]

Yeah, discord hot a new function, so it's useful to check in your servers the channel list on top of the channels, to see if there are hidden channels you may like or disable unwanted.

Like, I don't need IOS topics 🙂

[quiet oracle]

Just used the file upload feature. Very impressive responses for my questions. An option to point to the specific page(s) that the response is based on will be great.

[fathom nebula]

I believe people in general think software should be available for cheap. Mainly caused by free Facebook, Twitter etc. for to use aka in exchange for huge amount of privacy related data. But for some software, I rather want to pay as long as it is value added and secure privacy data. And especially for AI / Perplexity I'm willing to pay cause it helps me a lot with automation for personal and work related stuff. So it's worth it for me.

[fathom nebula]

I'm gonna try that too...it would be very helpfull for me as I have read a lot of docs for my work. How about privacy @covert sun ? Do you store all my uploaded file for further analysis, if so for how long?

[crimson nymph]

Uploaded files that are uploaded are saved for 30 days unless you delete the thread. #🅰│web-alpha-feedback message

[quartz jungle]

Quick feedback: When I search for explanations to something, the copilot first proposes several possible explanations for me to choose from. I think this is not the right approach because I wouldn't be searching for it if I know what's the right explanation.

[flat anvil]

You seem to misunderstood the context.
I requested to add lower tiers.
Currently you have the choice to use the service without supporting them or to pay much.

Disney +, netflix, amazon prime, spotify, xbox gold etc. Those all cost around 10$

And many use perplexity just as little help "better Google".
So a 5$, 10$ and 20$ tier would be great.

Why excluding poor people and not offering them to make also "donations" with little rewards.

[steep anvil]

the discount code ONEYEAR25 does not work for me. Anyone can use this code?

[sour acorn]

Its usable

[steep anvil]

no text field for discount code in safari.

[bold blade]

@pplx team, what is this weird "d" after every headings on the "AI Profile" page?

[visual bough]

Not only there... it's everywhere...

[sour acorn]

Easter egg 🤔

[outer vessel]

Wait for "e" next week

[flat anvil]

Only on annual

https://www.perplexity.ai/search/3bfa0599-db55-42c0-839e-06e0a24a032d?s=u

I started the conversation in German, but then it switched back to English. I know my profile says I want English text, but when I start the conversation and ask to get german replies, its a bit weird.

[outer vessel]

Try adding this to your ai profile, it worked for me: "I prefer German answers to German prompts and English answers to English prompts."

[visual bough]

The "d" is gone 👍

[sour acorn]

Meanwhile Me- Waiting for next Easter Egg

[bold blade]

The bottom navbar is fixed but I still think these text should be horizontally centered in each container 🤔

[rare cape]

Hi There, I tried the new upload file feature (an ebook given to me by the author). I found that Perplexity can't detect specific keywords in the text file I uploaded. Clearly I have the ebook and can find those keywords. Should I contact support?

[dusk aspen]

I think it depends on the size of the file you’ve uploaded and the model you’re using. If you’re using gpt3.5, it cannot maintain as complete or detailed context as got4

[sour acorn]

Did you use writing focus mode ?

[quick herald]

Hey I’ve been using the iOS app on iOS 17 dev beta 4 and I find that follow up questions or opening old threads often crashes the app

[flat anvil]

Well, but I want English answers when I write German.
As long as I don't said lets talk in German.

Because i need the English text for my Roleplay 😊

[sour acorn]

#🍏│feedback-ios

[quick herald]

Ah cheers I didn’t see that that thread existed

[echo jackal]

make coepilot on by default, or a setting to always have it on!

[crimson nymph]

Once you ask a question with copilot on or off, it will save. Next time you ask a question it will still remember if it was on or off previous time you used it.

[echo jackal]

turns off each time i reload the app, and website :*(

[crimson nymph]

Yes I am talking about the website. It should save after you asked a question with copilot on.

[hard edge]

Is it possible to turn off thread history for privacy reasons? I occasionally let people use my phone and don't want them to be able to access my threads history the same way I don't keep a browser or Google history.

[sour acorn]

Atleast for now the only option is to delete the threads once you use it and its work is finished

[swift nimbus]

As a pro user, I would appreciate the ability to edit or delete my questions. Is it not a simple matter of implementing CRUD functionality for my data? Additionally, could I export my questions into a JSON file? Thank you.

[drifting bobcat]

There was a "clear threads" button before but I think they removed it. don't know why since it was really useful

[rare cape]

Thanks @dusk aspen for your reply. File is 66KB in size. Very small

[rare cape]

@sour acorn Thanks. I did not - I will try your suggestion.

[dusk aspen]

You’re welcome @rare cape, you should also not have copilot enabled when you work with uploads, at least for the moment.

[hybrid dawn]

Sometimes these mishaps happen both on android and in the web version of your app. as I understand it, perplexity gives possible continuations after the answer, but sometimes gives an inaccurate message if its own question with a continuation was not composed correctly. nothing critical (considering I don't use these prompts very often), but strange to see. PS: i used the web version of the app, which i translated with a native language translator.

[visual bough]

Do you mean the same as here? #📱︱feedback-mobile message

[cold cargo]

I can never figure out where things should go when using systems like Discord for this sort of thing. I have had a problem on iOS where I always seem to need to tap the microphone, start recording a search, then stop and tap the microphone again before it actually converts and records my voice. Is it a bug, or am I doing something wrong?

[whole fog]

Dunno if this is already possible, but would love the ability to start a new chat via opening the page with a search query string parameter. This would allow usage through various applications that can tokenize the search param.

[crimson nymph]

You can use the following format

https://www.perplexity.ai/search?q=PROMPT_HERE&focus=MODE&copilot=TRUE/FALSE

[whole fog]

I looked in the faq but might’ve overlooked it if it was there - otherwise might be a good addition?

[junior wedge]

is this on the app or mobile browser?

[junior wedge]

#🍏│feedback-ios

[cedar vector]

I am glad that Perplexity AI can generate tables and images, but it cannot generate pie graphs or any other graphs. I hope that Perplexity AI will be given the capability to generate graphs such as pie graphs. Thank you.

[heady plover]

fwiw, you can probably ask it to write python code for that and run it

[cedar vector]

I hope it will have the capability to generate graphs without the need for Python codes. It would be great if Perplexity AI can generate graphs such as pie graphs. Thank you.

[cedar vector]

And I wish that Perplexity AI can also generate maps.

[quaint zodiac]

giga based tbh

[knotty berry]

Noted! We're taking a look into it

[tough halo]

Hello everyone, will there be an API for Perplexity. If so, any estimation on when that might be available (weeks, months, etc)?
Thank you in advance for your response.

[glacial granite]

Hey @tough halo! Perplexity AI does not currently provide an API.

[visual bough]

If I had to predict a time, I would say a year.
PerplexityAI have many other things to do at the moment. When the web and app versions are all bug-free and at the same level and many more features have been added, then maybe.

[limber needle]

@knotty berry Been trying to subscribe to the Pro version but the discount code "is Expired" on the billing page- yet the page displays it's applicable. I understand the deadline- can someone from the team allow the use of the discount coupon?

[jaunty lynx]

Hi, I've been trying to apply the coupon code for checkout. Really wanna use the Pro version but it says that the code expired? Is there any timeline? Can't see it on the website. Hope it can still be used. Cheers!

[visual bough]

Hello @limber needle @jaunty lynx !
You can find all the information for the birthday discount here: #📢│announcements message
Unfortunately, the offer has already expired, but the next promo action will be coming soon:
#👋│introductions message

[visual bough]

The promo action is still displayed, to avoid confusion the advertisement should be removed.

[agile siren]

Or extended for us who still want it.

:-)

[severe palm]

Am I using the promo code wrong?

Oh, I see, it indeed just expired. Any chance I can get an extension? I just found you guys from Dr. Know-It-All's youtube video.

[crimson nymph]

Hi Eventual, there will be another promo announced later this month: #👋│introductions message

[severe palm]

Ok, cool. I'll subscribe when you have that promo out!

[whole fog]

It'd be nice if the error screen here showed the prompt I typed in so I could copy it

[quaint zodiac]

boys can we add a voice typing thing for the prompt

[covert sun]

Hey everyone! Emojis have been fixed on Llama chat 🙌 ✅ 👌

[covert sun]

Sorry, wrong channel, should be in #🦙│feedback-playground.

[quiet coral]

When I share my results, it seems to give away my userid 😦

[crimson nymph]

If you share your thread the others will only be able to see your username

[heady plover]

you app is amazing, literally i needed the response in romanian, so originally i asked for recipes in romanian but without mentioning it, it gave me the answer in english but the resources romanian-based. so i mentioned in the query the exact language i want it to respond, and it did. i've never did this before, even though i did see romanian from time to time without needing to mention the language, but wow. KUDOS to the team!
it is worth to mention that romanian is not in the list of the languages the AI can be setup to respond, from the profile settings.
the second photo is after mentioning the language. so it does remember what language you want i guess

[flat anvil]

I still get random search mode in writing 😭

[outer berry]

I'm so impressed with your app I went out of my way to get to join this server. I had a very tricky rust question and it answered it in style. Well done!

[cedar vector]

Perplexity AI must have Romanian too.

[visual bough]

You mean for the AI Profile?
I think more languages will be added over time. In the meantime, you can also write your language preference in the Intro.

[cedar vector]

Yes, that what I mean...

[visual bough]

When clicking on the See FAQs button from the PerplexityAI mail, I am taken to the main screen of the PerplexityAI app and not to the Notion page.
(Android, maybe iOS too)

[vague bolt]

Hi, I have a problem: I cannot login to my account because I did not receive an email with the login link. I wrote an email to support and received a confirmation that it was delivered (saying "Your request (***) has been received and is being reviewed by our support staff."), but after 3 weeks of waiting I have not received a reply. I tried signing up with a different email from the same email provider, it worked - I received an email with the link, but when I tried to sign in to my existing account, I did not receive any emails.

[south ether]

Hello! Can you share your request number? otherwise please send an email again and we'll take a look. Sorry about that!

[south ether]

are you using any email scanners by chance?

[ancient lichen]

Meta feature: in this age of AI and it's possible effect on jobs,people will look for tools that would enable them to be better than the competition.

What features could be added to perplexity in order to achieve this goal ?

[hollow minnow]

After some time of using Perplexity, I find that Bing Chat is smarter than Copilot on website, which is smarter than GPT 4 on iOS, which is smarter than free normal mode. By smarter I mean better accuracy and contextual understanding.

Not sure if it's attributed to searching mechanism (Bing Chat can even consider search results not in the language of the query). Hope there'll be improvement on that and then Perplexity will definitely be a killer app.

[visual bough]

Feedback to https://perplexity-ai.notion.site/Perplexity-FAQs-0a9141bdd1b94c76b955e892f73451ff
Sometimes Copilot is called "Copilot", sometimes "co-pilot". I would prefer a uniform "Copilot" (also in #📘│faq ).
Also, neither on the FAQ page nor on the server are the search operators explained, important ones would be, among others: "site:" "-(site:)" "filetype:" and "OR" (for example in connection with "site:").
( @south ether ) (I think you care about the FAQ?...but I'm not sure)
Attached is also a usage example with the operators "filetype:" and "-", one was integrated into the question, the other into my AI Profile (later a post about this in #1124060874785685554 ):
https://www.perplexity.ai/search/2a626ae0-054a-4883-82d3-fdbcabfd9926?s=u

[vague bolt]

Sure, it's 197

[vague bolt]

Probably not, there only may be a spam check only by email adress made by my email provider, but I always check spam folder and it's always empty. I've also checked the filters, they're empty too, I've even tried specifically whitelisting an email that sends verification to make sure it hasn't been deleted or anything, but that didn't help either.

[blazing epoch]

When we have it that this is the correct path or sample response we want to use, can we have it that we can just click the up arrow so that we don't have to retype? https://i.imgur.com/gI1NgYq.png

[frigid berry]

it hallucinated so hard that it made up and entire future that perplexity supposedly has and tried to convince me that it actually exist even after double checking

https://www.perplexity.ai/search/7f160990-ec81-4078-a257-dfd7c4409ee6

[covert sun]

Hi @frigid berry! You are right, it hallucinated with that line from the first source, considering it a Perplexity feature. And said it was wrong about it in the last response, where that source with a 1-click course wasn't considered. It's good to have sources cited to check. Thanks for reporting!

[junior wedge]

did you upload the files to perplexity to analyze?

[blazing epoch]

Yeah, but it doesn't have the file path, but I've found a way around this that the AI seems to understand pretty well so far.

[junior wedge]

Please share it if you like and maybe the team can get it documented or implement a feature to solve it

#💭│ideas

[blazing epoch]

I forget that this chat isn't the same as that one lmfao

Thank you

[grand steppe]

I can't add a screen to the dialog with perplexity.ai, how do I do

[visual bough]

Hi, what do you mean by add screen? Are you talking about the file upload?

[grand steppe]

I mean, is it possible to upload a picture so that the neural network can analyze what it depicts?

[visual bough]

No, unfortunately that is not possible.

[grand steppe]

thank you for reply

[ivory kettle]

I had an issue when signing up for Perplexity Pro because I didn't have enough credit limit on my credit card. I have now increased the card limit, but the coupon I used (ONEYEAR25) is no longer active. Can you help me with this? Is it possible for Perplexity to make a new automatic charge on my card?

[visual bough]

Hi, the offer is no longer valid. But there should be a new offer in the near future.

[ivory kettle]

Thank you for the response IceLavaMan (a great nickname, by the way hahaha). Let's wait for a new offer! As for my specific case, my hope is that the Perplexity system will automatically charge my credit card again. I sent an email to [e-mail] to check if there is still a possibility of this automatic charge for me to enjoy Pro with the discount. I keep my card with a low limit to avoid fraud, but this time this strategy has harmed me.

[cedar vector]

Even though Perplexity AI is already in Write mode, sometimes when I try to translate text here, it doesn't follow the instruction to translate the text. Instead, it gives an explanation that I don't need because my only instruction here is to translate the text. I hope the bugs in Perplexity AI will be fixed so that it can still be useful to users like me.

[raven nacelle]

I’m thoroughly impressed of how this thing looked up the lyrics to this, and was able to explain this in language that I can understand. It’s funny because SoundHound app does pretty well with lyrics that it already knows, but it can’t look up anything.

[raven nacelle]

Well, I then asked for the meaning of the second verse, and it’s I guess hallucinating lyrics. Because it has the correct ones right there, and I asked it to explain that, and it just makes up lyrics that don’t exist in that song. I know you can report answers as inaccurate, but for some reason I can’t put anything to give feedback in the write a better answer text box. The keyboard does not come up. This might be an accessibility thing though, not sure.

[night goblet]

Hi

[heady plover]

Hi @raven nacelle , could you share the app version and the iOS version of your mobile please ? Thanks!

[raven nacelle]

3.8.0, and iOS 16.6.

[heady plover]

Wasn't expecting a response like this, but it's fine

Thought I should point this out

[vague bolt]

@south ether I replied to your email two days ago with screen recordings showcasing the problem attached, any progress?

[feral shell]

Hi, Perplexity Pro is becoming more or less unusable to me as it hangs all the time, Any solution to this?

[heady plover]

Hi @feral shell , could you provide more details on how this is happening? Thanks!

[flat anvil]

"As an AI language model, " I see this phrase a LOT in many AI models. THis is annoying. I mean.

As a Human inviduum, I can tell you that I don't have to tell you EVERY time that I am a human ...

I mean, you open a AI website, so why should you expect to chat with a human?

[south ether]

Hi, please check your inbox!

[heady plover]

lol yeah, but I read some tweet or something that it'll stop saying that (ChatGPT, I mean)

[feral shell]

Hi, This is happening within Firefox 115.0.2 on Ubuntu 22.04.2 LTS. See attached..it hangs with the spinning image next to Summarizing. Its been spinning since I submitted before taking a break

[flat anvil]

This is a VERY OFTEN happening thing, the copilot read MY Task to the AI and ask ME, to do it. I get it ca. 80% of all copilot usages.

[flat anvil]

save a screenshot as .jpg or png but not as xcf...

[bold blade]

Unless its a very complex task, I would never turn on copilot

[flat anvil]

When the Copilot doesn't do that thing, it is a very good AI helper.

especially, because the copilot does READ my AI profile and really try to understand my task.

[bold blade]

Im at a point where i'm no longer use copilot or web search

[flat anvil]

this is the only main problem with copilot

[bold blade]

Just straight up raw gpt4

[flat anvil]

web search is 90% trash... but sadly in write mode it does sadly very often switch to search 😭

[bold blade]

That's a little bit exaggerated, i still use web search if i think the model is cut off on that topic

[junior wedge]

@south ether maybe you can offer this customer some discount? 😉

[south ether]

Sorry about that! That's a Stripe issue, I will take care of it through email.

[quaint compass]

I have the same issue and I think (haven’t tested it because I haven’t had the chance) that it’s related to the use of pronouns like “this,” where you’re referring to something that’s obviously either in the prompt or just before it

[quaint compass]

I’ve noticed better quality writing in the writing focus (likely because the model has to spend fewer tokens scraping websites?), to the point where I will create a document or copy all the text on a topic that I need… and then feed it to the writing mode to create docs, posts, etc

[heady plover]

Could you share a screen recording of this error if possible? Thanks!

[gilded glacier]

Lately, I've been receiving incomplete answers more and more often. It's as if they just stop thinking in the middle. With ChatGPT, you can continue the generation with a single click.

1.) Am I doing something wrong? How should I best use perplexity for this?
2.) Is it possible to add a button to continue from where it left off?

Here is an example for demo: https://www.perplexity.ai/search/0d6e81b5-e222-41c1-9954-1a8f6beb8419?s=c

[pine vigil]

I've been using Perplexity a lot and I wonder if you can make a focus for GitHub

[crimson nymph]

If you want a github focussed search you could also use the site parameter. For example: puppeteer site:github.com

[pine vigil]

Thanks! Where can I learn more parameters?

[crimson nymph]

You can find a list of search operators here: https://moz.com/learn/seo/search-operators

[junior pasture]

The current system does not analyze the text files uploaded by users deeply enough. For example, the first question can ask for some quality information, while the subsequent questions are mostly repetitive👀

[flat anvil]

I got that even in other cases. "My question is, what is an good alternate word to sound not vulgar." Copilot "please offer alternate words who aren't vulgar." I asked the AI to help me speak more friendly and it ask me to do it for it.

I refused now to write here the vulgar words 😉 but I named it there.

[quaint compass]

I believe you because I just had a chance to test it more @flat anvil . I received a response, and made the follow-up prompt as simple as I could (adding more details to the AI profile to avoid redundancy): “make this more concise”… and of course, Copilot started asking, “what text would you like me to make more concise?” 😑

Then, I changed the prompt to “make more concise” — “what text would you like… “ 🫠

Last, I tried “more concise”…

And that actually worked haha. So I’m not sure what to tell you yet. I’m between testing and actually trying to get things done with it, otherwise I’d be playing and exploring all day to figure these shortcomings out.

[south ether]

Lately we are getting more reports of cut off responses like that when sourcing from GitHub, we are looking into it.

[sly oak]

PDFs are not being read by the program

I have to enter them multiple times before the system reads - anyone else experience that? And sorry if thats been asked.

[visual bough]

There are many problems reported with the file upload at the moment.
Here is a way to ALWAYS read the file:
(No focus may be set)

1. upload file
2. click on the text input field, but do not type anything in it
3. click on "summarize"
4. wait
5. after the answer has been loaded, the actual question can be entered as a follow-up question.
6. finished
   This is cumbersome but works 100% (even without the copilot function).

[flat anvil]

Maybe try "Shorten my quoted text".

This worked. Maybe it works better if we write first a task and then :

Do ... with this text:

It only asked me for the age of the kids,but did understand all 🙂

[quaint compass]

Cool! Was this from a pdf/file, or text that you inserted directly

[sweet barn]

Hi

I'm a pretty large spreader of Perplexity. I have sent thousands of users and am always promoting on twitter..

The vast majority... Like VAST majority of people in the real world are finding the answer to some societal questions wrong and lacking. It is turning people off the awesome technology.

If perplexity would adjust a few of these prompts the increase in user hours would vastly increase

Like asking "how many sexes are there" and perp says, 2 but 5 but spectrum...

If you would say "there are 2 sexes" this would make my life 100x easier for user retention.

If you would source the Webster's dictionary on most of this it would solve a lot of these conflicts.

Maybe add a toggle to disable social "sciences"?

[flat anvil]

I added it. I grabbed the text from a English book, which is for kids, but sadly contains way too much complicated words for me 😭

[obsidian hazel]

I’d love the file upload option available on iOS devices! 👨🏽‍⚖️

[flat anvil]

My copilot just wasted my last available usage to say "give me more input". And I gave it a LOT input... But I don't understand why it didn't used the function to insert a response 😦

I wanted to change the previous written text and add new features.

[quaint compass]

I will say, I have partially "gotten around" the issue of Copilot always asking me for help… by splitting my prompts into two parts.

It seems like a bug, but it’s been working pretty well: for now, it seems the copilot entry field will allow me to input a massive amount of text, so long as I hit enter immediately after pasting. I’m… hoping this a bug that doesn’t go away haha.

So, to use a previous example: if my prompt is “make this text more concise: [WALL OF TEXT]”

I’ll add the first part of the prompt, which is intentionally vague – “make this text more concise:” and then paste+enter the [WALL OF TEXT] into Copilot when it inevitably asks me to provide the text.

I feel it’s working much smoother now that I’ve just decided to flow with it haha. I use it for other tasks, not making things more concise, but the outcome is the same.

The main reason to use copilot, of course, being that it preloads your AI Profile, which is like a playground of instructions and commands that seems to help avoid having to redundantly type out details for every prompt.

[flat anvil]

But the input have a sign limit, I saw it once. It showed me a - 500 counter

Yes, also, it does then read that you don't want every time an apology or warning about everything 😔

[quaint compass]

Yeah, but if you press "enter" immediately after pasting, the counter doesn't matter. At least in my experience

I haven't gotten it to stop occasionally apologizing though 😂 no matter what I add to the AI Profile. That's a WIP

[sonic tusk]

For the upload file function is there a limit to the amount of pages of a PDF that perplexity can read/see? I've uploaded some files but it never is able to reach past the first few pages and always acts like the end of the file is sooner than it truly is.... Any feedback here?

[proud atlas]

The new UI looks cool

[spare sable]

The new ui looks cool, but desperately needs theme support!

[visual bough]

The new design looks good, when will the bots have the corresponding logo as an avatar?

[frigid berry]

there’s no workaround for this as it’s caused be the context window size which corresponds to around 3k words any longer and it won’t be able to read past that

[sonic tusk]

Thanks for the clarity there appreciate it

[limber folio]

the blue color was better😢

[weak yarrow]

that would be an excellent idea imo 👍

[flat anvil]

For me is the new color of the copilot bad, because it's harder to see if it is active. The old bright blue 🔵 was good, like a little LED light.

blue is just much better to see 😦

[heady plover]

I don’t really like new resdesign

Font is weird and the colours look dusted

I also hate the teal accents

At least the llama version has the old one. Hate the new one so much.

[languid bronze]

Please consider changing the name in the next re-branding. It is hard to pronounce, had trouble today when I was raving about perplexity in a coffee meet.

[timber coyote]

Liking the new font and style.

[bold blade]

this bug still persists, is this even considered on the roadmap?

[sudden timber]

I love the branding & i LOVE the new font. Great job!

[bold blade]

new font looks kinda goofy to me 😅

I prefer fonts that have clear indication between l and I

[weary lantern]

guys, is this working for you?

[urban gull]

not a fan of the rebranding

[brisk cedar]

it looks more modern now, i like it

[heady plover]

Hi @bold blade , sorry for the inconvenience. A ticket has been raised for this issue and is being worked on. Thanks!

[bold blade]

thanks for the quick response 👍🏻

[brisk cedar]

@heavy forge and @glad flax logo rebrand?

[visual bough]

#👉│feedback-general message

[keen relic]

#📢│announcements message

Don't forget to update mobile , still has old logo

[tepid wagon]

Cool new rebranding! I like the sound in the intro video - hope to hear that sound more in the future.

[burnt quartz]

The new logo is too busy, when displayed small it looks messy:

The new font is also more difficult to read than the old one:

And the new color doesn't pop anywhere near as much as the previous blue:

Though I suppose similar services like ChatGPT and Google Bard have muted colors in their UI. I'm sure there's an accessibility reason for it.

[barren basin]

The new color comes across as muted and dismal. Paired with the change to a subtle grey background and it just makes Perplexity seem less pleasant to use.

Is it me, or has the base body copy font changed too? It seems wider which means more scrolling.

The contrast between text and background is now necessarily lower, and the font is thinner. Poor usability decision.

[brisk cedar]

http://zip.yi5.dev/u/OEuc8b.png

Blue/Greenish perplexity looks good :)
from(#💗│sharing message)

[crimson nymph]

Here is an updated version that fixes the broken colors (because of an update). Or you can keep it that way if you like the light blue.

[raven jacinth]

the rebrand is nice: i kept clicking tailscale on accident before

[mystic cliff]

Hello guys. Do any of you having issues with the count tooltip showing the number of available prompts daily? It doesn't update the count daily, as intended, except only when I logout and login again it shows the count correctly. What do you think about this?

[mystic cliff]

I meant this

[crimson nymph]

The uses don't reset to 600 uses each day. Each use get refilled 24hours after usage.

[crimson nymph]

1 more update. I added the theme to github. If interested I also added a TamperMonkey script, without updateUrl/ downloadUrl, so don't worry the javascript can't change later on. This script will always fetch the most up to date version of the theme. Or you can download the theme.css file there. https://github.com/35e/css-themes/tree/main/perplexity.ai.

I've fixed a few things I missed and now also fully works on labs.perplexity.ai. If anything breaks because of an update, let me know. I'll fix it asap.

[reef perch]

Joined this server to say the same thing. But the new color doesn't look bad in my opinion.

[outer garden]

how do I connect my Discord account to my webaccount to get the Pro role?

[crimson nymph]

you can send a message to Alex Romanov to obtain the pro role

[next robin]

Summerize videos like on youtube from the captions.

The way i do right now is download the captions from website, and then uploads the file as txt to let perplexity summerize the videos. Maybe implement a feature to do that automatically?

[sour acorn]

NIT: Add the instructions in the Readme.md as to how one can use those theme. Ik its easy and one can google it as well but the purpose of Readme.md is to avoid that thing

[crimson nymph]

Thanks for the suggestion, I will add that 👍

[mystic cliff]

I didn't get, could you please, elaborate more on that?

[crimson nymph]

I can't replicate your issue, but this is how the uses work > #💬│general message

[mystic cliff]

Thanks, I thought, it was a bug. Nevertheless, it's a bit hassle to calculate

[rustic goblet]

Rebrand feedback

This is for the verification email. The buttons colors are weird. Doesn't suit the whole branding

[keen relic]

need to push an update to mobile apps with new logo

[crimson nymph]

hi mball8, the logo has already been changed on the mobile apps. Please check the appstore for updates.

[keen relic]

maybe i need to delete and reinstall then, probably cache issue

[raw tangle]

Feature Request: Attach a file to the prompt (like Claude) instead of erroring with red "too long" when pasting lots of information.

[cedar vector]

Hi, admin, I have a request, can you add more Free Copilots to Perplexity AI? For me, it’s not enough if it’s only 5. Maybe you can add more, why don't you make it 10, instead of 5?

I hope that even the free version users can enjoy fully the special feature of Perplexity AI, the Copilot (powered by GPT-4).

Thank you!

[mystic cliff]

Hey, today I noticed that when I clear the browser cache and etc., then logout/login afterwards, it resets to 600 again. What about this?

[mystic cliff]

I think, it's a bug that should be fixed

[crimson nymph]

Thanks for reporting the bug! I was able to do it as well after a while. I'll report it to the staff.

[radiant kindle]

https://www.perplexity.ai/search/55c3af9e-ee37-4701-b3be-085cb8775c3d?s=u
when I asked "Is there a list-like data structure that can contain different types of values?", it answered a question I asked previously in the thread, not the one I typed in that time.

This same thread... I can't keep it intact for you to view (I need to ask more questions) but uh, in one question, it answered my question in the response body, and answered it again in the "suggested follow up questions" bit

[hybrid dawn]

yes.

[hazy grove]

Hi! Downloaded the iOS app this morning and it seems to prompt me to add my username (already done) each time I open the app, must be a bug, FYI.

[cedar roost]

Are others also finding it very annoying that when perplexity asks a question and you hit enter or shift + enter it sends the message? somethimes it asks a question with multiple information like "please provide X and Y and Z" , you would then expect to be able to answer in multiple lines with more details or context.

It would be nice if it accepted multiple lines.

[junior pasture]

Is the new design trying to express that this is an all-purpose encyclopedia?

[dim steppe]

Would be awesome if the home page let us straight into the chat - I’m on iOS17 beta and the lag means I’m forever clicking links off the home page and have to go back out again

[frosty nexus]

Didn''t realize. Yeah it's kind of annoying.

[keen relic]

Maybe there are keyboard short cuts to insert a return?

Anyways, not why I'm here this morning, just drinking my coffee and finding my answers faster than ever before, just a morning shoutout to Perplexity AI team, THANK YOU. Google Search is Dead, Long Live The King.

[sweet palm]

I just saw the Perplexity discount code. Is there a way to apply it if I am in the trial period?

[frosty nexus]

I tried copypasting from somewhere else and the end result is just a very long paragraph with no enters haha

[flat anvil]

(translated my page with a translator for you).
The copilot did again fail its job...

[frosty nexus]

I would suggest using the "Writing" focus for translating purposes with no Copilot, just GPT-4.

Can you try that and reach back

[flat anvil]

And AGAIN @Staff create a CHEAPER "lite" premium... 20$ is too much!
5$ for some extra features like more copilot usages or 10 gpt 4 usages per day, whatever

[flat anvil]

I ALWAYS use write

And free users doesn't have gpt 4 access

[frosty nexus]

That's true. My bad

[flat anvil]

Damn, my keyboard does autocorrect "got" ← gpt to got...

[frosty nexus]

It's okay haha

[flat anvil]

"smart"phones 🙄

[frosty nexus]

Anyway for translating that's what I do. GPT-3.5 should be good enough for translating though.

[flat anvil]

But... My feedback wasn't about translating...

You didn't got that?

[frosty nexus]

I know I was just trying to help on that matter

[flat anvil]

It was about the fact, that copilot again asked things wrong

[urban gull]

Do you censor sensitive answers?

[flat anvil]

I wanted to find the English word for Nicki and...not a checkbook thing

[flat anvil]

No. It will then complain

This must be a visual bug

It looks like image embedding fail.

Also try to always use "writing"😄

[urban gull]

What do you mean ?

[urban gull]

Nevermind it was a glitch

[south spindle]

Dear Perplexity team,

The screen keeps jumping up and down when Perplexity generates an answer. It's very annoying. I use Chrome.

[south ether]

can you send a video recording of the event? thank you

[flat anvil]

This happens to many.

[flat anvil]

Open the website, at the dialogue box click in the left bottom corner on "focus"

[flat anvil]

You send a reply to a (longer) thread and then it jumps up and down. This requires to have a longer conversation

[south ether]

Okay we'll take a look at it thanks!

[south ether]

oh wow, what's your operation system?

[frosty nexus]

Can you check DM please if you are not to busy?

[south spindle]

Windows 11, using Chrome

[south ether]

thanks for sharing, we'll figure out what's happening asap

[frosty nexus]

That also happened to me one or two times. I think it may be related to the browser's window being resized to a certain size.

Not sure though

[glossy heath]

Perplexity wants to position itself as a prime research tool? I asked Perplexity, "What are the latest heppenings in Niger".

It tells me, "As of August 20, 2023, the latest happenings in Niger are not clear from the search results. The most recent news articles are from May and December 2023, and the other search results do not provide any specific information on recent events in Niger. It is possible that there have been no major developments in Niger since December 2023, or that news sources have not reported on them."

Really?

1. The happening are very clear and Perplexity actually referenced news articles in the response stating the happenings.
2. You can't have recent news from December 2023 in August of 2023.

[frosty nexus]

I would assume it wanted to say December 2022 and May 2023

Otherwise it makes no sense XD

[hazy grove]

Still happening every time I open the app FYI, need to hit Skip with each app open…

[south ether]

Hi we are sorry about this! this will get fixed on the next update for iOS which will come out very soon!

[hazy grove]

Okay great, didn’t want it to get missed, thank you! Loving the app so far

[spark dawn]

What is the discord private channel? It’s been a month since I got pro and I have no idea how to see it.

[frosty nexus]

You can either join through Perplexity's website clicking on "join discord" or I think you can also talk with @covert sun

[spark dawn]

Thanks!

[frosty nexus]

You are welcome 🙂

[flat anvil]

What fancy screen record is THAT? xD

[flat anvil]

I have it on Android too

#🌐│feedback-web message

Reported 1 month ago

[tawdry summit]

Just discovered that Perplexity seems to obey the user's input at all costs: https://www.perplexity.ai/search/51062059-48ad-49fc-a77a-be6e0fec12fa?s=c

I understand that GPT models are terrible at maths by nature, but still this is kinda funny

[raven nacelle]

Hi, when I clicked join discord after subscribing to pro, it just brought up the server. The pro channel is not showing though.

[sour acorn]

You have not got the Pro Role though

[crimson nymph]

You'll have to leave the server and then join the server again with the invite link. Or Dm Alex Romanov.

[covert sun]

Hi @flat anvil, @south spindle! Regarding this issue, could you please check one more time now and let us know if this still happens. Thank you

[flat anvil]

Tested on PC in a long thread (over 20 messages) no shaking, I will now test it on phone.

still happening

[frosty nexus]

https://discord.com/channels/1047197230748151888/1143685381720637543 Anyone on the pro plan having this same issue?

[flat anvil]

↑ @covert sun @south ether will you please make a cheaper plan for people who want to support perplexity.ai but have not that much money? Currently it is much or nothing.

[weary patrol]

Text sends while writing with IME (e.g. Japanese) on PC 😢

[languid bronze]

There are student discounts if you are a student

[old latch]

hello, why the answers are not always in the language in which the question is asked?
I always have to specify : (in ...)

[crimson nymph]

If you want the AI to answer in a specific language you can also set a language in your ai profile.

[old latch]

That partially answers my question 🙂

[flat anvil]

No, I'm a disabled person and I live from like 500$ per month. And yes I have to pay my 85$ electric bill, 35$ internet bill and the expensive food and water these days from that amount...

And even as student would it be way too expensive! It's not appropriate for people with less money.
It's the double price of a Spotify premium. And Spotify offers for students a 5$ deal, also Amazon prime, who gives you a LOT more costs just 9$ as student even lesser.

I'm not using perplexity for work, just for my roleplay and I would love to get minimum premium features without paying that much...

[umbral mountain]

you can use pi.ai if don't need all functions of preplexity

preplixity more about coding, searching and problem solving

[bold blade]

How is this possible?
The left panel is showing that there isnt a single source but the answer is definitely using one.
The right panel is the answer with no internet access (writing mode)

the sources button of this ^ specific answer is missing too

this must be a bug

https://www.perplexity.ai/search/027ca13d-f3f2-461c-bcf3-6a4e81cc02ef?s=c

[icy halo]

Always different and aren't correct

No copilot
https://www.perplexity.ai/search/706014e9-cd23-46f5-a32f-35d17fac5456?s=m
Copilot
https://www.perplexity.ai/search/fc23f4bc-fcb7-46ec-bf8f-766d99246c0d?s=m

[visual bough]

What should the result actually be? 9.81748×10^(-5) m^3?

[covert sun]

Thanks for reporting and sharing the link, @bold blade! We'll investigate.

[flat anvil]

But I WANT to support the developers AND get something back. But I can't afford 20$.

[gray pumice]

idk, llama is very bad as far as translating goes

shouldnt i be chatting and having really god translation results if I ask it

[icy halo]

Yes
Upd: sorry, no, it should be 9.81748×10^-2

Because 98174 divided 1,000,000 will be 0.098174

[covert sun]

Hi @flat anvil, @south spindle, @lucid forge! Could you please try it again. Should be fixed now 🙏

[south spindle]

Hi, it works fine now. I no longer see the shaking. Thank you. It's much more pleasant now.

[flat anvil]

Doesn't apply to old threads?

[proven seal]

Hello. I'm here to say, that my eyes are bleeding and brain confused every time when I see ligature fi in perplexity headings. Can it be disabled?

[keen plinth]

hi , want to report that the "delete all threads" button is not appear in web on desktop mode , but it appears on mobile browser

[flat anvil]

I just learned from a writing app, that this is an intended function from some people. "Ligature" yeah... the explanation is that it's intended to be easier to read.

But I also don't like it

[heady plover]

Imagine using Google in perplexity server

[frosty nexus]

They are working on that.

[visual bough]

Monthly feedback with giveaway
You could ask for monthly feedback from users, preferably via Google forms.
There would be questions like: What do you think about the development of PerplexityAI this month, what did you like most, what didn't you like, are you a pro user, what do you like about pro, what do you wish for pro, etc.? The questions would be the same every month.
That would be excellent for identifying trends, for example. Maybe three subs would be given away among all the entries.

[covert sun]

A monthly challenge 🤩

[bold blade]

Llama code: questions are right-aligned

[frosty nexus]

I just tried it and it was working fine.

Definitely wasn't getting that right alignment thing.

I just tried Java too just in case it was related to that specific language and it's working fine as well.

[bold blade]

try this prompt

1. short item
2. very loooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooooong one

tested without any css mods

[bold blade]

it only happens with the text on the question field, not the answer

[frosty nexus]

Oh wait

Didn't realize about that

Let me check

[frosty nexus]

Yeah, you are right.

I just realized you can use the triple coma to insert code there lol

[bold blade]

I got called a racist after telling llama to be concise 🙏🏻

maybe it totally forgot previous convo and just made things out of thin air 💀

[bold blade]

not only code block, any md syntax will do the same

if question field get markdown supported on llama, then why wouldn't PPLX 🤔

https://discord.com/channels/1047197230748151888/1137962052346978354

[frosty nexus]

https://discord.com/channels/1047197230748151888/1144457701422813184 haha

I would find that feature extremely useful

[lament scaffold]

How to make perplexity consider more sources before generating output?

[storm garnet]

using copilot generally seems to consider far more sources

[frosty nexus]

Use Copilot

[frosty nexus]

Exactly

Copìlot usually searches around 20-25 sources.

And if the topic is broad enough it may prompt you some questions and consider even more.

[lament scaffold]

Copilot is mostly using 8 sources each time. I'd ideally like it to use more sources upto like 50-100 in certain cases...

[mint shale]

I asked this in #⚡│ask-community , but it seems to have been drowned out.

I noticed before one of the big updates that Copilot* was searching beyond 25 sources. Was this a change mentioned anywhere? Is answering the follow-up questions/prompts the only way to bypass the 25?

In my case, the more sources, generally the better, as fewer sources can accentuate some unintended bias through a lack of cross-referencing information.

[dusty meadow]

I have also noticed that sources have been capped to 25

[sour acorn]

This This Please I also was facing the same the other day

[sour acorn]

I think if the sources are capped there should be way better way of ranking the sources too !

[floral igloo]

Typography is weird 🤣

[surreal pine]

what is this hallucination

i think perplexity suffers from made up truths

this is very dangerous; its like smart eight grader smoking weed

@mint shale you trying to type or is it yo cat?

[ashen halo]

Broooooooooooooooooooooooooooooooooooooooooooooooooooooooo !

Like

Im out of here man

[surreal pine]

this is how mind sparks when you

[heady plover]

bruh

[surreal pine]

this is perplexity of mind

true ; it was not last month;

[crimson nymph]

@surreal pine please keep this channel related to feedback for perplexity

[surreal pine]

sure @crimson nymph !why this perplexing

[void cairn]

Perplexity gives inaccurate info due to fake news and theories and predictions

[dusk aspen]

Link the example threads so the devs can look into it. I've noticed that fine-tuned GPT 3.5 is a lot sloppier than before in carrying out instructions for complex search queries(The issue has already been raised in pro-feedback) and it seems that this extends to its ability to discern trustworthy and untrustworthy sources.

[void cairn]

how do i link i was using ios

[dusk aspen]

Your threads on ios should be synced up with your desktop, share it from there

[void cairn]

Oh I will do it later then

[dusk aspen]

All good, aah the sacrifices one makes in the name of efficiency. Can't blame the devs too much though, GPT4 is insanely expensive to run at scale

[dense junco]

What is the best way to get support for my account? I just logged in and my PRO subscription has disappeared!!

[dusk aspen]

Email [e-mail]

[dense junco]

Thanks

[dusk aspen]

The <@&1082051235542216704> bot should honestly direct and take care of requests like this, I should put this in server ideas

[surreal pine]

this is dangerous but it was free from such smell last month

i use perplexity often never encountered this fake truth and made up game

[mint shale]

Perplexity seems to be degrading. A 25 source cap with GPT-3.5 is what we’re looking at now?

[mint shale]

This breaks my heart 😢

[sour acorn]

PS - Just a joke

[grim field]

How does perplexity transform my query into the search terms? It seems like an old LLM is used with outdated information.

[icy goblet]

What do you mean source cap?

[vestal narwhal]

unfortunately i agree with nerfgun.. perplexity has been better... generally and for python code especially. Like very short memory of provided code. It junks relevant code that was not modified. is there an option that it constantly spit the full code. I'm so bored to ask full code... I was about to use the back to school code... but... beeuuuuuurk... the 'answer' engine is degrading

[icy halo]

I think he means limit

[mint shale]

Until recently, I was getting 8-40+ sources per inquiry. Now, the most I’m seeing is 25 and the answers are more basic, wordy, and over-generalized.

With the speed upgrade, I’m finding myself spending more time finessing it when the prior version would have generated a better response in a faster time.

I feel like there should be a $10 option for this version and a $20 for what we had.

[light lotus]

Will there be a way to write full blog articles in a way that you can choose the tone, writing style SEO and have it pass a plagiarism checker?

[south ether]

We were having some issues with the support system, was your inquiry responded to? just making sure, if not please shoot another email and we'll be happy to help!

[elfin helm]

Hi team, Add a feature to upload files at given time in the conversation .Currently files can be uploaded only during the begining of the conversation.

[frosty nexus]

This would be a REALLY BIG thing for us users.

[surreal pine]

3
.
The command uses the curl utility to retrieve the get-pip.py script from the specified URL. The options -sSL are used to make the curl command silent, follow redirects, and show errors if any
3
. Once the script is downloaded, it can be executed to install pip, setuptools, and wheel in the current Python environment
3
.
It is important to note that pip is the official package manager for Python, and is used to install and manage Python packages
1
. The get-pip.py script is a convenient way to install pip in Python environments that do not already have it installed
3
.
Related
What is pip and why is it important for Python development
What is the difference between pip and conda
How to upgrade pip to the latest version

a location which dont exist; what is this smell of hallucinations

[lucid current]

https://www.perplexity.ai/search/6d83c868-8349-4203-b128-868fd914bb05?s=c

Why last word is USER_INPUT?

[visual bough]

@covert sun like INITIAL_QUERY 👀

[visual bough]

Is it gone? 🤨

[covert sun]

I don't see in the answer.

[visual bough]

Strange...

[sour acorn]

Same

[crimson nymph]

rip he edited the question

[junior pasture]

Is this the model developed by your company?😋😋

[visual bough]

It is GPT-3.5, but fine tuned

[deft ruin]

It'd be handy to be able to switch the AI model used on a thread-by-thread basis instead of having to change it on the account setting level.

[frail fossil]

I wanted to ask a question for the academic focus, "Where is Lphn2 expressed in the entorhinal Cortex?". I wanted to test it out because I know the answer, and the answer partially comes from this paper: https://pubmed.ncbi.nlm.nih.gov/34818557/

The paper was cited in the list of citations, but the actual answer given seemed to ignore this paper and give an answer that directly contradicts the paper with the answer. Copilot said that it is enriched in the lateral entorhinal cortex, whereas the paper says that it is enriched in the medial entorhinal cortex. Both can be right, but it's frustrating the see only one answer represented in the copilot answer.

https://www.perplexity.ai/search/83e70616-ac68-458d-9134-b7ce28ffeb8f?s=c

[sour acorn]

just tried forking and asked explicitly to go through the source you mentioned.Is this the expected result ?

https://www.perplexity.ai/search/ce127717-f426-48f9-9882-70fabb3755e7?s=u

[frail fossil]

That's also a little odd, since it should be citing Donohue et al, and I'm not sure where it got O'Reilly et al from.

Also the given results seem to construe the results from the previously cited paper with the new paper.

I appreciate your help with this!

[surreal pine]

one month back perplexity was wizard, seems somebody approved the logic of mud blood in code base

[hollow minnow]

Is Perplexity AI only available to certain countries?
I can't access it without a proxy, yet with a proxy I'm blocked (since days ago).

The app still works, though I also have to use a proxy to access it, otherwise the network fails.

[hazy grove]

Still happening each time I open the app. Surely fixing this is a top priority, I'm on updated iOS...

Find myself going for other AI apps as a result of the janky-ness. Would love for it to only show my searches too, it seems to populate random info that isnt relvant to me when I first open the app. Takes a second to re-find my last search. This app could be AMAZING with these two simple fixes.

"Popular" has no use, should only show "Your Threads" unless you choose to show Popular in the settings. Easy fix.

Would spring for the Pro, especially with Claude, but these are preventing me

[crimson nymph]

Both of the issues you describe are being worked on.

You could also send some of your details here in this thread to help the developers more: #1145171019523424336 message

And they are going to clean up the main page as well #1143589659990040656 message

[hazy grove]

Great to know, not sure how one would find that random thread otherwise lol. Maybe some Discord housekeeping is in order as well. Appreciate you!

I'd say in general the app needs fixing versus features, as a general piece of feedback.

[flat anvil]

The Copilot failed again horrible.

Also, ALL results had horrible bad sources, like it used France sources. "Stade" is the France word for "stadium".

It gave me zero results for Stade.
The entire request was very very bad. This proved to me, that AI is still NOT better than Google.

I just wanted to see all you can eat around 40km.

The First copilot answer was ONE restaurant, it didn't even search outside my village

[dusk aspen]

I'd just like to say that this is likely because of the new GPT 3.5 Fine-Tuned integration into copilot, it just performs consistently badly across the board. It used to be GPT4, I hope the team finish up their tweaking of this model soon, so we can either go back to GPT4, or have something equally functional.

[visual bough]

#🤖│feedback-bot

[manic vector]

Thats weird for some reason I cannot see that channel without clicking on the link you sent, thanks

[flat anvil]

Well, I can just use what the site gives me.

[vocal vortex]

#🌐│feedback-web 1. The searchbar need to be enhanced i want to search for words in each threads

2 . within each Thread there should be an question three where i can jump into each question becuase somethimes i search for seome questions

3 is it possible to get the latest question by type arrow key up

4. please move my threads as second becuase thats what i need the most

5 optionl make some shortcuts to each section below the ask anything.. bar i think its more userfriendliy

#➕│feedback-extension I want to be able to move a question that i ask via the extension into one of my threads becuase they have the similar topic

[dusk aspen]

Just don't use copilot for a little while, I'm sure it'll be fixed eventually.

[flat anvil]

the result without copilot was also horrible

without any further location infos

[dusk aspen]

Yeah you're in a double-bind because you don't have access to GPT4 which was formerly what copilot ran on

[flat anvil]

like links or where it is etc.

[dusk aspen]

I honestly don't know what you can do besides wait until the devs figure their stuff out.

[flat anvil]

using google maps 😉

it gave me great results

showed me all Buffet/all you can eat restaurants

• images + reviews.

[dusk aspen]

Great! keep doing that, I honestly don't know what the devs are thinking right now, but someone clearly thinks this whole fine-tuning thing will work out.

[covert sun]

Hi @flat anvil! Thanks for reporting. The first one, with no options provided is a bug, could you share the permalink with that answer? We'll work on a fix for that.
For things like restaurants, personally I prefer to set the location in AI Profile, here, is an example that I tried:
https://www.perplexity.ai/search/6a89a881-c98d-44bc-9782-f6faf1d27b91?s=u