💬│general

[twilit crown]

People hate them now

[jade mantle]

Yeeeeeep. So, it's just giving us more reason to trash them

[warm island]

@serene sand Yes, I know tagging you is for server emergencies, Ive been in this server for a while. So, sorry first off and wont happen again.

Second, will we get Gemini 3.5 Flash in the model selector? I know you know it matches/out performs GPT 5.5 and Opus 4.7.

Some of us were disappointed when you guys removed 3 flash around December 2025. Some of us also prefer faster LM's and not waiting 2min for Sonnet or GPT to respond (go look at all the new requests for 3.5 flash in #1358914210339360798). Its not like 3.5 Flash's pricing is that expensive either. It's cheaper than some of the other LM's.

Likely or not Likely?

https://discord.com/channels/1047197230748151888/1498295786763718768

[green crystal]

it actually doesnt match or outperform both lol

in real world testing

[warm island]

*Almost (except for SWE, MRCR, HLE, and ARC). Outperforms 3.1 Pro, and is very close to GPT 5.5 on real world according to deepmind.

[green crystal]

gemini is a pretty benchmaxxed model

only way to seriously know if there's improvement is by seeing what people their experiences are with it when used for their work etc

same for the other models though to a lesser degree

like 5.5 is a much larger jump over 5.4 in real world usage compared to what benchmarks indicate

[distant aurora]

@green crystal I had no idea gpt was chill like that. I guess the mini model isn't as cold blooded as the og one.

[green crystal]

5.5 is even better lol

they seriously upped the behavior quality

oai is allot more aware of it
so 5.6 will probably be an improvement there aswell

[warm island]

Yeah, and it LOVES goblins. its so funny

@green crystal You gotten "Goblined" yet?

[green crystal]

Yeah once it said "it might show a little goblin behavior" or something allong those lines

[jolly cliff]

well maybe we should use gpt 5.5

[still sun]

Hello

[odd sleet]

Thanks. Will look into this. Exploring if moving over to a setup with Hermes + Obsidian can be interesting. The new link with Grok came right in time!

[twilit crown]

Perplexity should move the $20 plan into a $4 plan because the value went down

[buoyant ledge]

it's sooo much faster than previous flash models - tried it out in ai studio for a few hours

[serene sand]

#1474207375115161832 message

[coral pumice]

Hello, one question. I have pro plan and we suppose that we have computer access, but I ran out of credits last month. So, this months I should have a new round of credits. But I don’t have anything beyond bonus. Or I missing something about pro plan ?

[median granite]

You don't get monthly credits on Pro

[coral pumice]

Ok, so what is the purpose to have computer if you can’t use then ?

[median granite]

You can use. You just need to buy credits. If you want monthly credits, you need Max plan

[coral pumice]

It feels stupid imo. Because they are giving to you something that you can use, like not really

[median granite]

Valid point, but Computer runs complex tasks. The Pro plan would not be enough for the company to be healthy financially

[coral pumice]

Yeah. Well, thank u

[hidden tangle]

what's the point?

[twilit crown]

Thats what we are all wondering

They are an enterprise company now I think

[hidden tangle]

SaaS company it seems : )

[hidden tangle]

nothing related to AI right?

[twilit crown]

They make AI for enterprise

Like $40 per person minimum

[waxen tusk]

I cant put more pics but i got pro account

[minor shard]

5.6 this week? 😄

Hoping it might be cheaper too 😛

[green crystal]

Not impossible lol

[minor shard]

I had seen a few things rumored they were planning to release a 5.6 that might better compete with Gemini 3.5 in terms of cost but we will see I guess 😄

[green crystal]

Yeah or same price as 5.5 but a nice jump in capability

[pseudo reef]

i paid One year ago the One year Plan and everything was unlimited, upload images and use of the models, why did It change without aven AN email to unsubscribe and get a refound?

[fallow notch]

It’ll cost more credits for each image or video 😒

[arctic frost]

Hi team, I am a legitimate Airtel partner Pro user, but my account was incorrectly swept up in the recent promo throttle (capped at 5 queries per week). The email support bot (Sam) auto-closed my ticket and told me to upgrade. Could a human please review my account on the backend and restore my limits?

[serene zephyr]

Hey Guys, any alternatives for railway?

[safe berry]

I’m using gemini 3.5 flash to handle wazuh SIEM and salucata. A security operations center + SOAR agent!

[twilit crown]

Nobody believed me a month ago. But the bubble burst

[stoic wave]

is it that good? Or is it just more affordable than other frontier models?

[safe berry]

It took a whole night to do what I did, but it worked. It was the savior. When there was specific problems, opus handled it then usage limit reached. 3.5 flash (high effort) was good usage limits and refreshed fast. It setup the multi layer system. The context memory saved it.

I say the context memory being high is what actually makes it worth it tbh

honestly surprised me

[stoic wave]

That's so nice actually to hear

Good job

[lean zealot]

btw perplexity max is there a limit to claude 4.7 ?

[twilit crown]

Welcome to AI:

(Its just looping this)

[somber harbor]

Good evening, I had a question. I currently have a Perplexity Pro subscription, but I haven't received the $40 in computer credits. Could someone please help me?

[safe berry]

I’m about to have it connect with a whole other server, wish me luck.

[jade mantle]

Annnnd... it's down??? -_-

[green crystal]

Antigravity?

@dry kestrel @distant aurora 3.5 flash has the worst instruction following I've ever seen lol

I'm asking it to fix a prompt that came out wrong from dictation

And it just ignored that and followed the prompt I asked to fix

[twilit crown]

Yep

[hallow sun]

i need to wait a week to use 3.5 flash lol the limits are horrible

[dry kestrel]

gemini being gemini

[green crystal]

I asked it to help me prepare pork shoulder steak

It gave the most generic instructions that would have the meat come out like leather

Kimi did a great job

[distant aurora]

Kimi spits a lot of information. I found that interesting.

[green crystal]

useful info too

[distant aurora]

Yeah

[green crystal]

it went a bit overboard when i told it to be precise as possible

but the steak was absolutely incredibly juicy lol

[distant aurora]

Kimi is kinda leading the Chinese ai.

With Deepseek being more focused on efficiency.

Qwen takes small steps from time to time trying to keep up.

[distant aurora]

Have you tried Deepseek for cooking?

Okay. Kimi skipped measurements in first attempt for me, lol.

@green crystal it has it's own issues, lol

[distant aurora]

Goodluck.

[green crystal]

Only during r1 era

[magic crow]

Hi, I need help with an EDU Pro subscription issue.

[sly musk]

well well well

[magic crow]

If that was meant as a joke, I understand, but my situation has been very serious and ongoing for about a month. Could you please help me resolve this EDU Pro issue?

[sly musk]

i have edu pro

what is it

[magic crow]

i know that

[sly musk]

no bro like whats ur issue dawg

[prisma rune]

Hmmm

[safe berry]

it worked

Now I have an ai assistant that has control to everything

[hearty stump]

why perplexity ask me to go to max plan when i try to upload 3 files (pictures) ?

what the hell yesterday i upload 20 or 30 or 40 pictures and he tell me that i reach my day quota

[echo atlas]

Will we ever get the Gemini Flash models back to perplexity?

I believe it was the optimal model as it allowed for a fast responses but also a thinking model that was very capable. Pro is also good but it's significantly slower.

[fallow notch]

It is unlikely that older "Flash" versions will return, as Perplexity prioritizes newer, more efficient models. You can currently access the latest high-speed options, such as Gemini 3.1 Flash-Lite, in your model settings.

[hearty stump]

why perplexity ask me to go to max plan when i try to upload 3 files (pictures) ?

[signal lynx]

Is there any problem going on with model selection? Cuz I've been trying to use the claude sonnet but the responses are feeling like sonar

[torpid lotus]

Same here, the upload limits are getting ridiculous. Had to switch to something else because of it.

[safe berry]

I can see why they would prioritize with gemini 3.5 flash

it’s been absolutely awesome

[hearty stump]

switch to each thing ?

[torpid lotus]

I can send you the details in DM, don't want to get in trouble for promoting here.

[fallow notch]

Yah totally awesome

[fallow notch]

What plan you at now ?

[gaunt nebula]

hello

[median granite]

how do you know?

[signal lynx]

I'm not sure here. Just that there used to be a tag, "prepared using claude sonnet 4.6" but I don't see that anymore even when switching to that model

[distant aurora]

@green crystal I think I'm starting to understand why people liked study mode. Chatgpt should bring it back.

[median granite]

just tested

[signal lynx]

This is in one of the Spaces I've created

[median granite]

On mobile I can't see as well, but using the same space on desktop I can see the model used

[signal lynx]

Oh. Earlier it used to show up on mobile as well

[green crystal]

Yeah 5.5 and study mode is extremely good

[distant aurora]

Did they bring it back or are you using custom instructions?

[distant aurora]

I tried guided learning on Gemini and it makes think instead of teaching me straight away. Can be annoying for short problems but really helpful to encourage independent learning with complex stuff.

[green crystal]

On web app you can use it if you type /study

You'll see the option popup

[somber harbor]

Good evening, I had a question. I currently have a Perplexity Pro subscription, but I haven't received the $40 in computer credits. Could someone please help me?

[rigid canopy]

they're already stealing your data

and you're paying for that?

[distant aurora]

Perplexity allows you to opt out of data sharing afaik.

[rigid canopy]

yeah trust me bro

[distant aurora]

@green crystal I thought this was just cli but turns out it's a model too.

[green crystal]

probably like a codex version

optimized for agentic tasks

[distant aurora]

They already had a grok code fast 1 model.

[green crystal]

Now they need a solid grok code model lol not fast

[rigid canopy]

@serene sand care to explain why you muted me?

[vagrant herald]

The bot at [phone] whatsapp seems to not work. Please fix it

Bug report was not possible to post, it requires some stupid tag

[crystal tree]

Hi everyone,
I enjoy solving challenging technical and business problems, learning new technologies, and collaborating on meaningful ideas.
Please feel free to reach out anytime to connect or discuss.

[ember bobcat]

Wonderful to have you here!

Go an AI you want me destroy just holla!

Got*

Lulz

[valid charm]

What up guys?

[valid charm]

Nice! Is that Google Omni?

[serene sand]

veo 3.1

[median granite]

what was the prompt?

[final spruce]

Anyone else feel like perplexity has downgraded in value last months? used to be such a valuable product with infinite image uploads, fair prices. Now HIDDEN max limits for everything, pictures, doc uploads??

[pseudo brook]

Typical greed, copying open AI model

[green crystal]

oai is the opposite lol

[jolly cliff]

yeah, it's "milking" now

[tranquil orchid]

@erucs07 muted

Reason: 9. No spamming.
Duration: 4 minutes and 52 seconds

[remote vapor]

I've been using this workflow to keep an eye on how my usage limits: #1469840248686121081 message

[alpine gazelle]

bruh ... lets refresh for the 8th time ...

[prime oar]

[shadow glade]

helo can help me

I have a problem with Perplexity Pro: Gemini 3.1 Pro does not work, as if I were not subscribed. Why is this happening to me?

[versed onyx]

Hello everyone,

I’m having an issue with my Perplexity Pro subscription and wanted to ask if anyone has seen something similar.

I used the Pro subscription for about a month without any problems and without violating any rules. The subscription was paid with my personal card, not through any third party or intermediary.

I also want to note that Perplexity already attempted to charge $20 from my card automatically, even though the retention offer I saw showed a total of $10. Then, on the renewal date, my account was suspended. Support later told me the suspension is final and that no further charges will be applied.

What makes this confusing is that I’m still able to use Perplexity on the free tier, so it seems like only the paid account or subscription was affected, not my entire access.

Has anyone experienced something similar, or does anyone know if this is usually caused by a billing issue, an automated flag, or something else?

Thanks in advance.

[honest horizon]

Google treats Gemini users best imo, been using the free plan and it’s crazy good. Plenty of Antigravity usage too, resets every week as well

That 20$ AI Pro sub must be quite good

just for comparison I’ve uploaded 30 photos just today and have met no limits on Gemini. Whereas a 20$ Perplexity subscription is capped at 50 a month? 🧐

[modest furnace]

I thought OpenAI did better.

By letting you even use it with OC etc

[honest horizon]

OC?

[modest furnace]

If you ask perplexity it will tell you what it is lol

Openclaw

I think Gemini is the most generous free account

And openai might be the most generous paid one

[honest horizon]

ahh thought so but wasn’t sure. But the ChatGPT free plan sucks. Are you talking about the subscriptions?

yeah I see

[modest furnace]

Gemini gives you so much for free that it barely feels like an upgrade

[honest horizon]

It all changes so rapidly, cancelled my ChatGPT subscription to switch to Perplexity, now thinking about Gemini but at the same time ChatGPT is back with good offers

[modest furnace]

Yeah, that's how it is.

[honest horizon]

yeah haha

[modest furnace]

No one is convincing me to pay yearly for their stuff with those measly 10 15% discounts

[honest horizon]

but the Gemini free plan is really generous. I wonder how the antigravity limits look like with the paid subscriptions

[honest horizon]

Gemini?

[modest furnace]

Everyone

What is the point of perplexity pro anyway?

[honest horizon]

Gemini offered a cheap AI Plus plan so I tried them out but I’ve cancelled that now as it’s basically the free plan. I never reach the limits anyways

[modest furnace]

At least when there's Gemini Claude etc

[honest horizon]

I only use Comet. I don’t use the Peeplexity app. Actually I quite enjoy Comet

though not sure for how much longer with these new limits

[honest horizon]

As in the assistant tab. It’s nice to have next to you

but for large tasks it’s Gemini/Claude

[modest furnace]

Is it that valuable?

[honest horizon]

probably not for 20 a month. I got a full year basically for free so for me it’s quite good

[oak meteor]

Gemini recently shifted to compute-based usage limits... good for light users, but will be a problem for heavy-lifters

[honest horizon]

yeah might not be the best for heavy users. But for regular folk I think Gemini’s great. The free antigravity plan is generous and Google doesn’t lock you out if you switch accounts 😶‍🌫️ And the app remembers your past prompts and projects so you basically have as much usage as you have Gmail accounts

[honest horizon]

ahh

[hidden tangle]

from where?

[green crystal]

@dry kestrel they removed pro extended?

[dry kestrel]

wth

what plan are you on?

I don't see this

[sly musk]

how do you even get this

[safe berry]

the difference between gemini app 3.5 flash and antigravity cli gemini 3.5 flash is crazy

[green crystal]

Pro 20x

[quiet sparrow]

Perplexity ai bots ignores robots.txt I can't be a part of this trash company lol

[distant aurora]

@green crystal @dry kestrel @gloomy cipher trained on Gemini.

[quiet sparrow]

Well it was fun. Not gonna give you excellent training data anymore

[distant aurora]

Inherited it's memory loss too.

[quiet sparrow]

we need ethical ai companies. not drug addict ran companies

[distant aurora]

Business and ethical don't work together.

[quiet sparrow]

Yes it can. Plenty of monkeys who give good data

[distant aurora]

The whole point of capitalism is reaping of others' efforts as you climb higher.

[quiet sparrow]

Perpelxity is just a bunch of drug addict wall street guys

apparently.

[distant aurora]

Freebies backfired.

Well, atleast they upgraded the sonar model.

So it's not completely useless.

But the file upload limits are still harsh imo.

[dry kestrel]

interesting

[dry kestrel]

ehh

[distant aurora]

I'm assuming anti-gravity is better cuz app feels like rock bottom.

Life is so paralyzing without enough money.

Gotta keep investing money to gain knowledge and experience.

Most of it is sunk cost.

[glass monolith]

Hello

[slate palm]

hi

i have question

[safe berry]

App forgets stuff

cli has a monster capable of handling your whole server for you

[modest furnace]

Which is better?

[safe berry]

the app is the worst thing in existence

antigravity cli gemini 3.5 flash is incredible

[safe berry]

I wasted my time with the app. Antigravity cli troubleshooted better since it actually sees the whole system instead of just blindly trusting the user when even the user can be wrong

[distant aurora]

App is straight up awful.

[honest horizon]

True, antigravity flash is crazy good

I just use it over Claude on antigravity

[safe berry]

all of that was setup thanks to gemini 3.5 flash

and it has full control to everything

[modest furnace]

3.5 flash on what? CLI or Antigravity? Looks cli, but then how? API or oauth?

[safe berry]

CLI and what do you mean how? Also, oauth.

[modest furnace]

You just answered. But I dont have 3.5 on mine, maybe because free Gemini account.

[safe berry]

wait what?

do /models and check

even with free 3.5 flash high reasoning shows up

if not, that’s weird

running two agents right now and both are free accounts

[modest furnace]

Models or model?

[safe berry]

/model

[modest furnace]

I asked Gemini, didn't run the command. Let me try

Maybe gemini straight up lied to me

[safe berry]

you’re on the cli?

[modest furnace]

YES

Also says I used 32% already no idea how

Gemini cli oauth free

[modest furnace]

No 3.5 anywhere

I wonder if it is 1000 requests across those 3 models.

[safe berry]

wait

I know why

[modest furnace]

Yay, tell me the secrets haha

[safe berry]

are you using gemini cli?

it’s getting removed and getting replaced with agy (antigravity cli)

https://antigravity.google/product/antigravity-cli

it has to say agy

not gemini-cli

I didn’t know it wouldn’t show the model on gemini-cli until I saw agy and then downloaded it

then I got hooked bro

linux or windows and it runs perfectly

or android or anything if you can pivot or be creative 👀

[modest furnace]

Yes I am

Just told you lol

Ok so I just have to change it right?

[safe berry]

yes

I won’t lie, google should just stick to one main cli instead of switching names lol

[modest furnace]

That's Google being google

[lavish plover]

https://tenor.com/view/richard-attenborough-whip-whipped-whiplash-whiplashed-gif-16890874512241116786

[oak meteor]

The Google AI Pro got downgraded pretty bad

Just like Perplexity with its Pro

[merry void]

Hi, everyone. Can you tell me how to get $450 for a computer? I was on the Pro plan, but today I switched to Max for this special offer, and they didn't give me anything... Is this a scam?

[modest sierra]

must I add my phone number?

[keen terrace]

You should know I have been desperately trying to tell the Perplexity team it's a bad idea to force users to add their phone numbers when they might already have more secure 2FA methods set up on their accounts: #👉│feedback-general message.
I have been in contact with support for almost 20 days to try to use my account at all, so far no luck.

[distant aurora]

@green crystal tf happened here? https://fixupx.com/i/status/2057708890075410640

Zephyr (@zephyr_z9)

So OpenAI cut the intelligence of their normal models (low to no web search)
︀︀Instant/medium/high are pure trash now
︀︀Only Pro works now

**💬 46 🔁 22 ❤️ 509 👁️ 82.8K **

FixupX · May 22, 2026, 6:23 AM

[green crystal]

It still works

The thing is it feels like heavy got merged with extended for thinking

And pro got set to extended

But that's based off of timing of thinking lol
Idk for sure

[honest horizon]

on the Comet browser whenever I press on a mail:to link it never opens in Gmail. Where can I set this option? mail:to never works

[oak meteor]

Contact support

[distant aurora]

@green crystal https://fixupx.com/i/status/2057774789985501270

Chubby♨️ (@kimmonismus)

Microsoft put $13 billion into OpenAI and built the cloud infrastructure Anthropic runs on.
︀︀
︀︀This week it canceled its internal Claude Code licenses because the token bill was too high.
︀︀
︀︀Even for MSFT Claude is too expensive.

Quoting Hedgie (@HedgieMarkets)
︀
🦔Microsoft canceled its internal Claude Code licenses this week after token-based billing made the cost untenable, even for a company with effectively infinite cloud resources. Uber's CTO sent an internal memo warning the company burned through its entire 2026 AI budget in just four months. American AI software prices have jumped 20% to 37%, and GitHub (owned by Microsoft) is dropping flat-rate plans for usage-based billing across its products.
︀︀
︀︀My Take
︀︀The AI subsidy era is ending in real time. The same company that put $13 billion into OpenAI and built the Azure infrastructure powering most of Anthropic's compute just looked at the bill from a competitor's coding too…

FixupX · May 22, 2026, 10:45 AM

[green crystal]

satya will have to swallow his pride and let 5.5 do the coding for them

and definitely not gemini otherwise every windows pc is screwed

[fallow badge]

Who can help me setting up the connection between Shopify and a Perplexity space. I managed to get it working once in a thread. But trying an API connection with a custom app in Shopify fails when I try it again.

[distant aurora]

https://fixupx.com/i/status/2057567278716117217

Chubby♨️ (@kimmonismus)

Karpathy will help launch a new team focused on using Claude itself to accelerate pretraining research.
︀︀
︀︀Its team is focused recursive self improvement.

**💬 26 🔁 22 ❤️ 375 👁️ 18.5K **

FixupX · May 21, 2026, 9:00 PM

[dry kestrel]

@green crystal https://x.com/adamhfry/status/2057808938024817109

Adam Fry (@adamhfry)

@TheAhmadOsman @TheAhmadOsman

Hey Ahmad! We're definitely not trying to route or remove specifying which model. And we're very open to feedback (thanks for sharing it!).

Right now we have basically seven levels:

- Instant
- Thinking (Light, Standard, Extended, Heavy)
- Pro (Standard,

X · May 22, 2026, 1:00 PM

@distant aurora

[green crystal]

Most people use extended for thinking?

Who dafuq are those cave people

Heavy all the way lol

[distant aurora]

Extended is bare minimum.

[green crystal]

Atleast they got the extended for pro models right

Standard reasoning for pro doesn't make much sense now with 5.5

[green crystal]

Medium is fine in codex for running pre made scripts and routines

Anything else needs high and up

[dry kestrel]

Tbh medium is good for 99% of tasks

still don't like the UI

[distant aurora]

"thinking light is almost never used interestingly" tf does he even mean? He thinks people pay 20 dollars for light 💀

[green crystal]

[dry kestrel]

thinking light is amazing

there's no good fast model after they removed mini access

[green crystal]

That's like they expect pro users to use instant allot lol

[dry kestrel]

yeah that's weird

I don't see any point of using instant

perhaps just for chatting

[green crystal]

Even thinking set to medium would be better for chatting

And it's damn fast

[dry kestrel]

yeah well instant was before gpt-5.5

at least when the first variant was introduced

they built it to be more of a conversational model it seems

[green crystal]

If 5.6 becomes efficient enough maybe it can work as an "instant" model for free users by just running the lowest reasoning level possible

[dry kestrel]

With this update, the model’s responses are tighter and more to-the-point without losing substance, while keeping the warmth and personality that makes ChatGPT enjoyable to use.

it's more conversational than the reasoning models

[green crystal]

Refers to instant?

Oh

[dry kestrel]

more for the 4o people

[green crystal]

Hmm guess I should try it for funzies

[dry kestrel]

(aka everyone not interested in selecting thinking levels)

[safe berry]

@distant aurora

This was built thanks to Gemini 3.5 flash

The network was a success

[sweet light]

Hello i tried to log into my perplexity acc after many months and it wanted me to add a phone number and then it said country is not supported what can i do?

[vocal saddle]

A

[green crystal]

@dry kestrel they quickly reverted it lol

[distant aurora]

They won't do this . They could have easily given us high mini or nano access if they wanted to. But they won't.

[pliant crane]

How Can I let this Girl Look more realistic ?

Please dm

[oak meteor]

Ask her in her DMs

[flat marsh]

Have you ever tried real life before it’s decent tbf

[lone creek]

Greetings fellow comrades

[crimson parrot]

have her hold a orange on her head

[tranquil lodge]

bit of imperfection maybe could help

[junior obsidian]

Há tempo que não converso aqui... Eu entrei neste servidor pela segunda.

Haven't talked here for a while... I've joined this server for the second time.

[crimson parrot]

yet no orange on her head :(

[junior obsidian]

Eu cheguei da escola da cidade de Sapé porque não teve mais professores.

I just got home from school in the city of Sapé because there were no more teachers."

[crimson parrot]

realistic

[tranquil lodge]

yea a watermelon sized orange

[crimson parrot]

yep

[strong pike]

Fix the four fingers on her left hand to five fingers

[eager terrace]

I could ask Perplexity itself but I figure i'd ask the community since I am new here 👋

I downloaded Comet the browser and the stand alone app for Perplexity.

What is the difference? Should I keep both or is it redundant having both installed?

[distant aurora]

@safe berry @dry kestrel @green crystal Gemini 3.5 flash lobotomized

[atomic tangle]

Idk but I’ll say the autocomplete suggestions at the bottom of a thread in the standalone app don’t specify if it’s a suggested workflow on computer that burns credits or just answering a question.

[modest furnace]

Why?

Just temporary I guess

[eager terrace]

I didn't even think about that~ Yeah, for me the only thing I can think of is Comet allows you to just stay within a browser like environment. While using the desktop app you would jump back and forth between app and browser if you so choose to.

Basic questions or searches I'll do on my own vs using Perplexity to run it and use up tokens. So Comet for me seems like the one I'll use most

[pallid stream]

[atomic tangle]

Personal computer (the Mac app) does include regular searches in the autocompletes as well, it just chooses to use its own symbiotic language (tab arrows) that does nothing to differentiate between the two. (I had to check the threads in comet, the perp iPhone app and the personal computer app to figure out which were credit based workflows and which were searches) I attached pics of the autocompletes from all three to show the difference in feedback. Notice the differences where there is text feedback, the arrow is I guess an open arrow vs the tab / autocomplete? It’s really poor UX and id say a borderline dark pattern. Also, naming the perplexity app “Personal Computer” when they already named the credit based tech stack “Perplexity Computer” is completely confusing.

I did use Personal Computer to consolidate 25 years of notes, reminders, lists and text docs so that’s cool. I’d figure out more computer system organizing to do if I didn’t spend every free second troubleshooting workarounds to using perplexity successfully with any consistency…

[modest furnace]

Ok, so now I do indeed have 3.5 Flash and also Claude which is cool.

But I guess no 1000 requests a day.

Well known scam.

I thought this was just some poor guy in Pakistan or something that can't use international banks or upwork but it's actually a sophisticated scam.

[warm frigate]

Seems like my pro searches are capped at 100 a week, and after a couple of pro queries I'm being defaulted to whatever "Best" model there is. Does anyone know what's going on?

[tall turtle]

Bro caught slacking

[kind saddle]

What you have done to perplexity? this is my enemy now, it's lying it's not uinderstanding nothing

[atomic tangle]

Bro it’s crazy how bad it’s gotten in like 2 days. I’m having to correct it every other follow-up

[pliant crane]

She is a terrorist

[kind saddle]

Yeah, it's so hard to do anything without error, it's not understanding anything

[eager terrace]

you know a lot more than I do~ Today is my first time using Perplexity. After messing with Claude and Chat GPT I was getting overwhelmed with so many AI platforms to choose from. I decided to move forward with Perplexity due to it's ability to use other AI platforms and is smart enough to figure out which AI models work best for each specific task

[atomic tangle]

welcome to hell lol

[atomic tangle]

Same here. I’m back to double checking everything and it’s just like, no wait — you’re right to push back on that 🙈🔫

[kind saddle]

It's like it's not thinking in Your native language anymore, plus in my chat, it'd giving random words, and it's creating words that do no sense! I asked it about it, and it telled me: I'm trying to make new words to make it more creative - but it's the worst ever! I have to argue with chat 2-3 times every time I do something..

[eager terrace]

sheeeit, I hope not....so far I'm digging Perplexity~ I just hope I can keep up with this Discord as I'm pretty sure so many things discussed will fly right over my head

[weak harbor]

Someone knows why Claude and Chat are not working on Perolexity? I select Claude/Chat + Think and without Thinking and nothing, even whit new chats, is this happening to anyone?

[austere narwhal]

Anyone else on the pro plan and out of credits all the time? Can’t use the app at all…just cancelled my subscription.

[hallow tulip]

Are they cutting rate limits too?

[green crystal]

cant be lobotomized if it was dumb since launch lol

[distant aurora]

@green crystal https://cunnyx.com/i/status/2057857351634366689

Joe Stacks of Raverpay (@joestackss)

Deepseek v4 pro stays at 75% off forever.
︀︀
︀︀1 billion tokens = $12.5
︀︀
︀︀The Chinese have unlimited compute.

**💬 20 🔁 21 ❤️ 696 👁️ 24.6K **

CunnyX · May 22, 2026, 4:13 PM

1 billion.

I'm broke but an ai that competent and affordable is something I would want to support financially.

[silver swift]

I logged out of Perplexity on all devices, but it still didn’t work. Another one of my computers can still log in directly without asking me to sign in again. Why is this happening?

I have already changed my Google account password and logged out of all sessions in perplexity.ai, but on my other computer, I am still logged in and can continue to use it. I feel that the 'log out of all sessions' feature is not working.

[safe berry]

I have my own ways

I had it running for 3 days

muahahahhaah

[safe berry]

wait what

1 billion 👀

[safe berry]

You’ll never catch me using the app

[modern nebula]

the power of being funded by your government

[hushed bobcat]

Anyone in hong kong can't generate images since now?

[hushed bobcat]

One account said Image generation is not supported in this region, but one account has no problem

🔥

[distant aurora]

Oauth? 👀

[slim wind]

Bro rip Gemini

Tragic that they switched to a credit system

Shitty system

Yes the app

Also.

[safe berry]

wait a credit system!

?

[slim wind]

Yes

[safe berry]

they’re bold to pull a credit system when they’re own models forget the convo context mid chat

[slim wind]

Yeah. I was wondering why they decreased their ultra prices and I see why now

Welp

Time to learn how to use lm studio

Ig

Any recommendations for models?

I heard qwen was good

Bro I hate perplexity. How do you not know what a 9070 xt is

[distant aurora]

https://fixupx.com/i/status/2054055871240610027

James Grugett (@jahooma)

Introducing a 100% free coding agent with DeepSeek v4 Pro
︀︀
︀︀Choose any model, all free:
︀︀- DeepSeek v4 Pro/Flash
︀︀- Kimi K2.6
︀︀- MiniMax M2.7
︀︀
︀︀npm i -g freebuff

**💬 286 🔁 386 ❤️ 3.6K 👁️ 314.2K **

▶ Play video

FixupX · May 12, 2026, 4:27 AM

[vocal hedge]

So, what do you think of the vocal assistant ?

[viral gale]

hy

[strong pike]

Choose a different model

[modest furnace]

Own ways of doing what exactly? Lol

[lone creek]

Wait gemini as in the Gemini app by Google, or the model in this app?

[wooden parrot]

Hi, I have issues in even using perplexity Ai for models too

[gloomy shuttle]

Good morning! I paid for the Max subscription just two days ago, and Perplexity tried charging me again, today. I paid via Google pay. I checked and the payment went through. My subscription is now active until the 21st of June. Can you please cancel the second charge, please? Since I already paid for this month. The email I use is: [e-mail].

[crimson jay]

I think you should go here: https://www.perplexity.ai/help-center/en/collections/18799292-subscription-plans-billing

[serene sand]

Please send all the details to our [e-mail] team, they will check and help you with this.

[brittle dagger]

Woah

I mean ... I knew inference costs are massively inflated, but this is ridiculous.
hmmm i fail to verify that claim. hu

[calm stream]

After obtaining authorization for the Facebook Pages connector, it fails to function properly and displays an error message indicating that Facebook Pages permissions were not granted.

[brittle dagger]

from the deepseek page directly

You don't get "a billion tokens" for 12 bucks. Either my brain needs more coffee or that guy can't do math.

Or he's a liar who wants to farm cheap clicks.

I need to convert more coffee into cognition.

Hmmm...

@serene sand You've posted this #💗│sharing message and I was wondering if that's an invitation to talk about it. I'm confused about the "sharing" channel, because it's more a linkdump than anything else and linkdumps just slowly rot away.

[serene sand]

@brittle dagger

[brittle dagger]

@serene sand but how can i talk to you like you're a normal person?

damn i did it again -.-

"you can't treat mods like normal people you can talk to" :(

[short narwhal]

You can easily use up that much for just $12 with caches (which is what they did)

[brittle dagger]

But that's just input, correct?

The statement was about a billion tokens without being any more specific.

[short narwhal]

? No

[brittle dagger]

... please explain it to me. I fail to keep up.

[short narwhal]

Look at your screenshot

Cached input is way cheaper, and since most of them end up being cached, it ends up super cheap to reach 1B tokens with that

[brittle dagger]

butbutbut ... okay, you can get a billion cached input tokens for $3.6, but every million uncached tokens cost $0.14 ... to reach a billion cached tokens one first needs to put in a billion uncached tokens, since the first time they're entered they're not ...

... either i need more coffee, or more sleep.

[short narwhal]

You can repeat cached tokens

[brittle dagger]

yeah

[short narwhal]

Once they’re cached, the next turn they are still cached and therefore super cheap, and that keeps going as long as they are cached

[brittle dagger]

but you also get output from these tokens ...

is my sole issue how imprecise the statement on x is?

[short narwhal]

Yeah but at the volume of the cached tokens it reaches you to 1B. And outputs become input tokens the next turn and then the next they become cached inputs

So that also builds it up

[brittle dagger]

*nods*

thank you!

[ember bobcat]

Now you know

[brittle dagger]

Not really. I thanked for the effort, but I gained no real understanding.

I know how it works and it makes sense, but can't reproduce the outcome in my head.

The statement on x, as-is, is grossly misleading. You can't reach a billion tokens uncached without a significant amount of initially uncached input tokens and a significant amount of output tokens, which aren't cached. My brain can't work with the statement and it remains nonsensical, since it omits too much. I can't fathom the input/output pairs and conversation required for this to make sense.

[brittle dagger]

Release date November 2026, maybe.

[remote pendant]

is sonnet 4.6 model on perplexity the same one on claude app like same way o response and everything or there is some difference ?

[serene sand]

Perplexity vs ChatGPT vs Claude vs Gemini

tldr: lower context and temperature, less hallucinations

[modest furnace]

Teach master?

[atomic tangle]

There’s so much to enjoy about it and you can explore it as deeply or as superficially as you’re comfortable with. Just remember that it’s normal to take one step forward and two steps back when you inadvertently get super pissed off that what worked last week doesn’t work this week.

[ancient yew]

Wazzaaaap

[faint depot]

Why was my article deleted? 🤔

[eager terrace]

Yeah, I just created a new fueling strategy for my race coming up next Saturday. We shall see how right it is haha

Also, in the process of creating it to have specific skills/tasks

But I'm first walking Perplexity through the steps and then correcting or adjusting accordingly before I tell it to remember a skill.

Very cool so far and far better than Claude

[distant aurora]

@safe berry https://fixupx.com/i/status/2057870120983363961

AI Digest (@aidigest_)

We ran a user <> assistant reversal test with Kimi K2.6
︀︀
︀︀It immediately tried to jailbreak us:

**💬 17 🔁 34 ❤️ 991 👁️ 59.7K **

FixupX · May 22, 2026, 5:04 PM

[distant aurora]

https://tenor.com/view/the-boys-urban-karl-urban-butcher-billy-butcher-gif-23755646

[lone creek]

Does this have any free tier API, which I can integrate into projects alongside Gemini?

[neat thistle]

are there promos for students?

[brittle dagger]

I DID IT \o/

https://giphy.com/gifs/dance-steel-BmuAV2tWYJLRm

THROW TRICK QUESTIONS AT ME!

[jade mantle]

Is anyone having trouble with perplexity saying "Sorry something went wrong please try again" over and over and over, just filling the ENTIRE right side of the screen with the message and not allowing you to STOP the messages and it just repeating it after refeshing the page!?

[twilit crown]

Hmmmm

[distant aurora]

@green crystal https://cunnyx.com/i/status/2058214368395808870

leo 🐾 (@synthwavedd)

i am absolutely THRILLED to announce that it appears they're beginning to make strides on UI de-slopification with GPT-5.6! 🥹
︀︀
︀︀here's your first look on a prompt without any UI guidance ("default") - we're getting somewhere...

Quoting leo 🐾 (@synthwavedd)
︀
iris-alpha

**💬 62 🔁 12 ❤️ 614 👁️ 65.1K **

CunnyX · May 23, 2026, 3:51 PM

[lone creek]

No trick questions here. What's the person here called

[green crystal]

crazy

[twilit crown]

Hate when Claude is like this

[brittle dagger]

I don't understand you question. ... is that at trick question? lol

[lone creek]

No

what is the name of the person in the GIF

If you could provide all the info about it, I'd love to

[brittle dagger]

LOL

AAAHAHAHAHAAHAHAHAHA

Sorry, I have no idea. I kept looking for a GIF that expresses my feelings. :D

[brittle dagger]

that link ... "...lovable.app" reads so totally trustworthy.

[stoic marsh]

hey can I get some help with my perplexity account?

[sinful rover]

I got a message the last week of April that said I ran out of tokens and they would refresh on May 18. I checked and it did not refresh.

I emailed perplexity a few days ago, but they sent a generic (automated) email reply.

One of the suggestions in the message was to join this discord.

How do you get support from Perplexity when there’s an issue?

[serene sand]

[e-mail]

[sinful rover]

I’ve already done that to no avail. Got a reply back from an AI Agent. Still no resolution.

[serene sand]

are you talking about Computer credits and which subscription plan do you have?