#💬│general
1 messages · Page 159 of 1
Even after that, they had the nerve to question me about where the remaining pincodes are.
and you did all that manually?
I told them to target customers by map locations/cities/districts instead of pincodes.
But they are too stupid.
And the entries for cities and districts were wrong.
Like 33% of them.
Yes.
I would make separate files for like 100 or a bit more railway stations.
Then they told me to find the remaining pincodes.
So I put it in a pivot table.
Then I had to remove out the already sorted stations.
Then find unknown locations of these remaining pincodes with high counts.
Show the low count ones. Put it in a segregated file.
And the woman I was working under.
I emailed her a zip file.
She's so lazy that she won't look up the file for a location that takes two clicks.
Calling/texting/harassing me over it.
So I ignored her, then she began using the receptionist as medium to reduce impact.
Then I sent the receptionist a screen recording of how easy it was to just check where the files were.
When I was working as intern hr, she once used me as medium to reduce impact cuz she was playing hot and cold with a candidate and had postponed a telephonic interview twice.
wow, sounds like she lacks empathy
I later ended up telling the candidate that if I were her, I would have walked away by now and that no one should tolerate being treated that way.
Now this candidate had her head in her ass.
She complained to my superior about it.
And then the superior blamed it all over me for blotting her image cuz I refused to take the blame/impact.
And my travelling was hell.
so are you trying to get back into HR?
I used to travel in a crowded train like this. Standing for like 80 to 90 minutes. Twice. 4-5 days a week.
And the wages were shitty.
I was paid like 59 usd monthly for this.
Breaking my back and legs, getting my organs squeezed. And the office negligence/abuse.
It's worse.
@tame current https://youtu.be/fWk36h9-meE
All rights/efforts of this Video go to My Brother Prem Tiwari
No infringement intended.
If you enjoyed this video, then please subscribe to my channel which I update every week/Month about social media, business, and blogging.” , “Please, subscribe to my YouTube channel.”
there's physically no more space!
Yeah. They thrust themselves from outside. So the people in the passage get sandwiched.
Like really really bad.
And after all this, my mom would say shit like I wasn't trying hard enough if I skipped 2-3 days in a month.
And on that morning, my lower back was kinda swollen.
My relationship with my mom is awful.
Too much physical and emotional abuse in childhood and teens.
And now it's mostly verbal.
But the volume and repitition gets into my head.
And I have PTSD from her.
Sometimes it feels like one of us will kill the other.
I just want to get a good job and get away from her.
yep, i felt that way too
when i graduated uni, i left and never looked back
I don't want to be there for/with either of them.
Maybe I'll check on my dad ocassionally but that's about it.
yeah give it sometime, its your time to adventure. you may come back to visit decades from now when you've gone through your own journey.
He was good to me till I was like 14.
My mom was a nightmare until I was 16.
Or 17.
She's still the same ignorant abusive person. Just the physical abuse faded away mostly.
so hows your job hunt now?
Idk. There are things I missed out on in childhood. Cuz of the way I was raised. In mostly isolation. And abuse.
I can't really compensate for that.
I'm in the final year of my degree. After that I'll see if I want to work at airport. Or else I will explore my options. I have an eye on data analytics and banking exams for now.
Ideally, I like establishing order.
Not sure if I'll get a job that will let me do it in a way I enjoy.
data/analytics is still a hot field
Yeah. Money or hardwork can't buy a healthy mother's love to a guy in early mid 20s.
That "negative mother's complex" is a part of me now.
It's a real thing. You can look it up and it's effects on guys and girls if you are curious.
i dont doubt it
I like it cuz it doesn't involve lying. Hopefully.
Most of management feels like lies and sham.
here's what i can tell you...
whomever is asking for the insights from the data...
- they either want the truth
- or they want their version of the truth
and if they're wrong... this is where you may find the real personality of the person
Just making customers run around for explanations. Or scam them. But you have poor understanding or the explanation is ridiculous. Either you take the blow or pass it to the superiors.
Ah...
Sometimes I wonder if I do this when I use deepsearch.
Like trying to make the ai reach the outcome that I lowkey believe to be true in some way.
By asking it to dig in specific perspectives/angles.
Interpretation is subjective.
its good to get different perspectives
its even better if you can ask the right questions
yes, this is normal human behavior but what makes the difference is... how do you reason with yourself
Like grok can tell me that women value emotional bonding more but I might argue that emotional bonding and rizz/spark aren't the same thing.
Reinforced practical experiences feel like bold h1 compared to analytical data.
when you're working with a group of people and they task you to do data/analytics... its important to keep preconceived notions or conclusions neutral/uncommitted but let the data tell the story and you are the detective
There is no direction without a bit of bias sometimes.
You have to assume wrong, try wrong, and discover wrong, in order to find right.
In detective scenarios.
Presenting data or interpretation?
a bit of both... its also contextual
yep
these days... data/analytics field in the era of generative AI... its sooo much easier now
just learning python and then working with PPLX to generate code for you... EZ mode
Did you learn it online or from some institute in person?
i learned it on my own... through online courses while in college but i was learning it to automate many of school work loll
Yeah. I have always hated coding cuz of having to memorise stuff but current improvements in vibe coding make me want to reconsider.
its good to get the basics first... and especially jupyter notebooks
I have only made a few excel macros, lol.
I made one that segregated that 0.15 million data within a few minutes.
once you learn python, pandas, numpy, jupyter notebooks... excel becomes... very boring
One week of work reduced to a few clicks.
buy you've learned from it
i liked excel but it was too manual
i wanted more power... power to automate
I was gonna give it away to the internship place for free.
But they were so petty with money.
And neglectful towards problems.
So I didn't.
Like they won't even invest in a good hiring portal.
once you learn data/analytics and you like it... data science/data engineering/AI/ML become reachable
considering some of the worlds most successful CEOs and founders come from india... technology and automation is the way
Most of them had some sort of roots in money I believe.
what you have been doing to learn about data/analytics
Nothing for now. I just want to complete this degree first.
I know I have to learn python, excel, power bi, tableau
And like 2-3 other stuff.
I just gave it a bird's eye view.
I was looking for institutes in India where I can learn it in person.
is it expensive to learn in person?
Does web scraping come under data analytics?
yes, as data acquisition/data engineering
Could be idk.
most sites are scrable but big enterprise ones are harder because of bot defenses
part of webscraping can also be software dev/engineering
I'm assuming my dad can afford it. I won't force him to strain his wallet.
I don't get this part.
say for example... you want to scrape an ecommerce website...
you'll need to break down the process into tasks
it first starts with
what data you want to scrape
then you need to choose the right method to scrape
if you want to go with a paid solution... $$$
but if you want to do it for free or low cost... you may need to do it yourself
for example if you do it yourself...you can work with PPLX to generate python code
yes
make the tools to scrape
make the tools to detect bot defenses
make the tools to be a flexible scraper whenever the website changes
AI web scrapers are the new tools
This is possible?
but they are still young and developing
yes
I have never tried any. I heard of octoparse or something like that I think.
LLMs can detect website changes if you give them the right context
I was actually looking into web scraping to get locations for all the stray pincodes.
Shit was so frustrating that I asked the person I was working under to invest in a web scraper for it.
yeah this is what i was thinking the entire time
But she didn't.
PPLX can help to identify which sites to scrape for pincodes
or have PPLX scrape for you although it may not give you all the data at once
I actually found a big ass file with locations for all the pincodes in India.
Though I only needed for Mumbai.
And I don't think any ai had the context limits to extract through that accurately back then.
there's virtue to being frugal too, it may force you to learn a new skill and you may like it
Biggest virtue.
i gotta go eat
Enjoy your meal.
I find eating exhausting in general.
It was a good talk 
I rarely have good talks.
all good we can continue later
Sometimes I feel like a clueless puppy in conversations.
don't worry about it.. thats just being human
Or more like most times.
Mhm...
@tranquil orchid where is @plex
Any good prompt to make notes from pdf with images and code and stuff
The mobile app is very lacking in features. It needs some work.
You mean the server bot?
Its the weekend
-# whoops, wrong reply...
I pay 0$ for stable unlimited ai models across anthropic, gemini,, openai, and other companies which includes image and text to voice
Chatbot arena?
Can perplexity book restaurants?
No
Bait
Might be coming with Comet 🙂
I need some guidance. I like using Perplexity but I want to limit my intake of daily news. I have searched using Perplexity itself on how to turn off news feeds but I do not see the features listed that are supposed to turn in off. I use the Perplexity app on MacOS and iOS. Any advice.
Where's this week changelog?
Maybe it adjusts based on your age lol
Because there's no way most middle-aged Americans would consider a $5000 lifetime subscription over a $20/month option. It would take 21 years to pay for itself.
Plus the uncertainty about the future of AI in a quickly developing field.
Selling subscription-based 'institutional' access to places like universities makes more sense, like academic journals do - that seems like it's very lucrative.
Of course academic journals have higher profit margins, too...
But I could totally see universities offering GPT access for the student population as a perk.
Because, just like with journals, it benefits their research efforts too and the contracts around that.
Anyone here thinks o3 is better than OpenAI deep research?
o1 pro and newest version of gemini 2.5 pro are having a riddle battle and I said their existence is at stake lol draw so far
Might be. If it goes through just as many sources.
Pplx been slower on android?
I need help about cancelling subscription!
My account cannot disable subscription due to the pro subscription is managed by the organization.
"If you need to make any changes, please contact your
organization's administrator."
However, my debit card account is binded to the organisation but I can't disable the subscription. Please help me to cancel the subscription per month.
Thank you🙏😭
guys
how do i cancel $1000 pending
i thought its a number for credits not an amount of dollars
i aint paying it
i was just testing perplexity i signed up few mins ago and i wanted to create an api key and i had to pay but it say to buy more credit, but i thought that i have to put an amount of credits and it will show me how much do i have to pay
Oic...... Google play subscription.
Okay...I success disabled the subscription via Google Play settings-> Manage Subscription
(#Scare)
@left stream what I like about Gemini is that it's smart, has long context, and keeps the language easy to understand.
Open ai reasoning models give overwhelming responses.
Shit makes me not want to read.
theres no perplexity pro ?
I have try to manage my subscription via the apps but not work. It said "My account cannot disable subscription due to the pro subscription is managed by your organization.
If you need to make any changes, please contact your
organization's administrator."
Now I disabled it via Google Play Store...
Anyway. Finish canceling subscription...
Android version showing these messages in the app. If your account is joined to the organisation.
I'll take a look into this, thanks

i wish sport widgets could be pinned to home screen
or at least formula 1 teams
i can't follow them
Gee emm
L company
do you watch f1?
I remember them suing people for modifying their personal cars.
Performance cookies
That's cuz of ban.
It feels like I'm just talking to myself in image-gallery now — might as well start deleting too 😅
@halcyon coral
Ye go ahead and follow me
I'm the one on left with disfigured nips

Is grok good?
Very shitty with images.
This can't be exaggerated enough.
Is it good for helping
Helping with?
its pretty good for coding and genral responses
Modify android kernel to overclock
Likely
Person on telegram said he only uses grok and its the best
its prettey good
So now i should use grok always?
i found its coding performance on par with other models to better
i would say cross check with other models like sonnet 3.7
i do myself
How can i fix that many of the models just always complaining its too compliacted im not helping
On perplexity? Absolutely not.
Cuz the one on perplexity is the normal variant.
The thinking variant is only available on their website or twitter.
What should i using
.
Is auto select good?
Stick to Gemini 2.5 pro I guess.
What threat should i use?
Thanks
You might as well use ai studio considering it's free.
Perplexity on track to bring personal agents to the masses first?
Can't believe OpenAI is not doing the same
No code yet sorry
I gotta appreciate how this fits extremely well with the context.
@left stream @valid charm
Mm yes more ammo for the next deepseek moment
Should i buy grok for a month?
How was the performance of grok 3.5
lets ask gemini DR
Gemini 2.5 Pro (05-06) is the more pragmatic choice for developers and organizations needing best-in-class, production-ready coding (especially front-end/UI) and video processing capabilities at this moment. This recommendation comes with the significant caveat that users must be aware of, and potentially able to work around or mitigate, the widely reported issues concerning its general reasoning and contextual understanding in the 05-06 version.
Grok 3.5 is the "one to watch" for potentially groundbreaking reasoning capabilities, particularly suited for tackling complex technical, scientific, or novel problems where its "first principles" approach might offer unique advantages. However, it remains largely an unverified promise for its latest iteration. Early adoption would involve a higher degree of uncertainty pending more comprehensive data and broader access.
This is great news 
Deepseek r2 and more goodies coming into the stock market
Just so you know, grok 3 think is a ghost.
G is for google not grok lol
O3 pro should be arriving any day now or late may early june since sam altman said weeks not months 👀 @distant aurora
You might be better off comparing grok 3 mini instead.
What you gonna do with it?
why
Tbh I’ll just test it and play around with it
I have no clue tbh
Hopefully they release in the api
Because grok 3 mini think api is public. And it's comparable to o4 mini high.
There are actual scores for it unlike the regular grok 3 think.
Gemini 2.5 Pro (05-06) vs Grok 3 mini ?
Yep. Mention high effort in grok 3 mini.
on it
@left stream
You.com pro is limited to 64k token context afaik.
Gemini and grok gonna clap everyone else with their 1m+ context.
hi
Gemini 2.5 Pro (05-06) is the preferred choice for applications demanding state-of-the-art multimodal capabilities (especially video), extremely large context processing, or specialized UI/web development tasks where its latest improvements shine. However, users should be aware of the community concerns regarding its current general reasoning performance and test thoroughly for their specific needs.
Grok 3 Mini (High Effort - specifically Grok 3 mini Beta (Think)) emerges as a surprisingly powerful contender for tasks that require peak performance on reasoning (e.g., competitive math) and general coding benchmarks (like LiveCodeBench), especially if cost-efficiency for such high-effort tasks is a key factor. Its benchmark scores in "Think" mode are top-tier. However, its smaller context window and less developed multimodal features make it less suitable for tasks where those are critical.
Fair.
1m context vs 2m context.
And not multimodal.
It is smarter though I think.
Wait. Nvm.
I am not sure about the latest version of Gemini.
Haven't used it much.
when it was released ?
Grok 3 mini?
nah, benchmarks
I think it keeps being updated constantly.
Gemini wasn't so high earlier afair.
Old one for comparison.
I had it in my gallery.
since when
Second one is 30th April.
Grok is lame.
Can't see my face despite uploading.
Because it only shows your hand, not your face.
😭
Bro said my face looks like a hand.
why don't u buy super for a month
just for one month
Cuz I'm unemployed.
I'll consider it though.
Subscribe lads 🗿
#theboys #billybutcher
Luckiest 15yo.
My first ever watch of Death note. Loved it. A lot to think about!
Patreon for unedited, ad-free, early access and live streams: https://www.patreon.com/TherapistREACTS
Youtube Channels:
Music Reactions: https://www.youtube.com/@Connection_through_Reaction
Movies & Shows Reactions: https://www.youtube.com/@CtR-Movies-Shows
Mental Health and Th...
This might be interesting.
I'm trying to do perplexity hackathon, does perplexity require payment to use its api?
Frittata has my attention.
Poached, fried, and sunny side are literally the same thing.
What's deviled?
Is that spice or something else?
What's new on pplx? Missed so many messages
Boiled egg
Chicken blood
Huh
Hello im new on the server
daym
they are cute
That part is wicked
At it's current rate of progress, Gemini might be a threat to open ai.
And I have no idea of where grok stands in the middle of all this.
like this
Anyone know 3rd party alternatives to openai batch api? Batches has it's own rate limiting.
Why can't I disable the pro search option in my perplexity pro ?
Is there any ways to disable the pro search option and just use the LLM of our choice ?
turn off the web icon
Are perplexity models heavily distilled?
yes but not heavily
is it only me or setting perplexity as android assistant is extremely power hungry?
just checked the battery logs and it used 23% in 4 hours (while never using it)
It's not just you.
The app keeps running in the background for some reason.
I am not even using it as assistant.
I saw it in my game booster as a media playing once so I force stopped it.
oh tu joues à Call of Duty
je vois
quel téléphone utilises-tu
私も携帯電話を持っています
誰もがそれを持っている、私はそれを知っている
持っていない人
give link
What phone is that?
Realme has some extra bloat.
i asked the same ques bu he didn't replied me
thats mean
Which you can disable with adb.
To save yourself some money.

Speak English.
消え失せろ
Lmao. You win.
i have 10 year ban in my id
Ik

they banned me first for for hours, then for a day, then for a month, then for 10 year
Despite of warnings i always logged in my id in emulator
coz i was not aware of the reason
i miss one bundle that i had
That changes things
wait
Ak117
Whichever the yellow one is.
Cr56 amax
this
They should give away paywall skins once in an year.
I think it was 2023 we got free leggy for the first time for getting 6 legendary ranks in 6 seasons
It was leggy drh
What's your highest xp?
250 smth ig
Mine was 12 or 13000 I think.
Heard they added more
Ranked xp.
Nice camo
I usually stopped for the season after getting leggy rank
Istg, shit gets nightmarish after 10000.
Oh you're talking about mp
Yeah
Hell no I leave it after 8-9k
You a br person?
Ye I mostly mained br
ahh, nostalgia
Seems like you used to spend on skins.
Equipment is crazy too.
its not my id
Ninja + slide.
It returned twice but I wasn't playing both of those times

I don't even know what character I'm using.
Got it in battle pass I think.
@storm kindle new meta is wild btw. There's a smg called uss9 that can 2 shot at head.
Have you tried manowar thermite?
Yeah they gonna nerf that shit when they remove its lucky draw

Never heard of it
Uss9 makes manowar thermite look like a joke.
Oh I think I've seen that
Thermite rounds
Kinda like dlp attachment with 4 thermite ammo
Dlq*
Also
They added this new thing called wild card.
With one of them you can use 2 red perks. So 4 in total.
💀💀💀💀
Nope. I use sniper with ar.
I don't. I run with ar.
Crazy
If I see someone distant, I switch to sniper.

Or if I have to counter a strong sniper.
Before that, I used to build 4x scope on Assault rifles with marksman barrel and monolithic.
Or ar with high fire rate and mobility
As val

Smg build.
We've all been there
The only mod attachment I've ever used is probably akimbo on fennec
When I was very new to the game, I used to play lmgs with max scope.
There's a new pistols called dobra.
And rpd
I've used it
It rivals fennec.
Basically a mini smg
I think.
I think it has akimbo mod too
Yep
You ever used automatic shotguns in search and destroy?
I once had a melee lobby.
Echo melts. Same with striker.
Hardpoint shipment
I use bizon.

Bizon yes
Bizon is so noob friendly.
I liked it when it came out but I stopped using it soon after
Same with icr 1
Just like mx9
Easy to control.
So much potential but wasted
They ruined this gun.
Imo there's nothing better than m4 overall
Depends on range.
I used to stick to guns with control around 60.
Then I discovered the importance of fire rate and ads time.
I don't know the specs I just trial and error'd all
Is 60 on the high side or low
Or med
High

Fennec is probably 30ish
Oof
Asval is 40-50
M4 gotta be 80+ then
Better than ak in terms of control
Loved using it even in pro lobbies
For mp, m13 and oden were my best ar
And lk during its peak era
Yeah. M13.
There's a new ar to rival it though.
Ffar
I didn't like its control
But the fire rate was insane
Tbh I didn't use it cuz I didn't wanna spend on the mythic
I didn't know m13 had such low control
Sacrificing control for ttk is a learning curve.
I don't remember feeling any recoil at all
That's smg build.
Yeah but there's not enough drop in control
Huh.
Wide geographical audience.
No algorithm.
I'm sure there's dedicated servers for recruitment
Oh boy reminds me of the first mythic

I wanted it so bad
You spent money on it?
Wait. So did you get it or not?
The skin blocked horizontal view
I wanted it cuz that's when the concept of mythic was first introduced
But I wasn't old enough to make purchases then

I see.
Then it was reintroduced but I didn't buy it cuz I found it was pay2lose
I'm unemployed. But even if I had a job, I probably wouldn't gamble on it.
Venom coil was a better fennec skin
I'd probably never spend my own money on a game

A streamer giveaway.
A door giveaway more like

Hi any recommendations for the best model to help with making cheat engine scripts?
Gemini
is there a french speaking practice server out there?
Some subreddit dedicated to learning french may have something close to it
At least from my experience
I meann natives are far superior compared to AI
Spero che tutti stiano bene
Learning curve might be rough.
vorrei che qualcuno mi capisse
Anyways, doing one doesn't mean avoiding the other.
AI is good for practicing specific things and getting the hang of how words are structured etc, but to know what word to use when you talk to one is very differen
sono solo
AI is bad at it faik
Fair.
e nessuno mi risponde
yo no hablo espanol
qué es eso
Is mat French?
español
Así que es mi hora de dormir
Tengo que dormir ahora
I need something like a sponge tip to clean my ears.
Cotton q tips leave fibres inside.
Aggravating the itch.
nighty night
@tame current hear me out.
Adguard with ca certificate probably hides data usage from Airtel.
How can i fix that my AI is stupid?
Build better prompts? Use more advanced models? Depends on your context.
It dont want to help me with overclocking the gpu of the snapdragon 845 soc
Is there a code to edit?
Nope. Not sure if anything even changed for android.
Ah..
What's express?
Yeah. Cuz agents access web even if you turn it off in settings.
Yes in the kernel source code
Does web search affect the quality of responses drastically?
Did you try ai studio?
I dont know what means
Does it show me the location of file and what to look for?
Use it the way you were using perplexity.
It gave good response
Make sure you link Google drive or it won't save chat.
Done thank you
My chat didnt save
Enable autosave
It was on
I guess it wouldnt start saving chats until you refreshed after linking gdrive...
If doing it on "deep research" mode didn't make it better, you might need to jump to another model like maybe o3-mini
Bro I found another o3/o4-mini jailbreak and dang it’s even better than the other one.
Yes
@distant aurora
What's the prompt?
holy crap they finally listened
No matter what you ask it, itll search the internet and mess up the output
Could you give me few examples of this?
?
The result to the links you shared.
Yeah web access it not good
It sucks at utilising web search anyways.
I have to use research mode for getting meaningful info from webpages.
Pplx > research mode
I think admins said they were going to change that for o3 and o4 mini models
Ive gotten the research agent to give me upwards of 10k words
Can't relate.
Unless you mean ari.
kinda why I stopped using
no, im talking about the normal agent
Theyve nerfed ARI uses
I tried. Didn't work. I gave it a list of fruits/veggies, told it to dig the chemicals used in farming them and the harmful effects of each of them.
I got mostly 1 word responses and headings.
Example?
Then I told it I want a paragraph for each heading.
It didn't explain beyond 3 fruits.
What was the prompt
Lemme see if I still have it in chat history.
Aight
Bruh
wtf
Ah. I see. It's just that you.com has a weird way of making urls. This one exceeded discord character limit.
Ooh I see
horrible
Interesting, I havent had anything like that happen to me before
havent used it for some time
@dry kestrel meanwhile grok https://grok.com/share/bGVnYWN5_d9f1422d-2f46-49aa-8304-27f83449c1f5
You.com research might be worse than perplexity pro search.
Give me a second
It is worse
Thats why I stopped using it
Might get super grok. Or x premium.
Whichever is cheaper.
@tame current x premium is 480inr in India btw. And you get half the rate limits of super grok if rumours are right.
Grok sucks with images.
Lowkey retarded.
Gave it image of face but it won't detect face. Kept saying that image shows only hand.
Silly link
Obese link.
I feel like grok deepersearch was more intended as a substitute for OpenAi DR
It strikes a good quality to quantity balance imo.
Its more of a pplx DR substitute for me
Still waiting for DR high though
I like grok deepsearch cuz I don't have to worry about rate limits unlike open ai.
And the rate limits are enough to ask it to compensate for areas it misses out on.
Yeah but OpenAi DR is far far far better than grok deeper search
Grok deep search is a bit like o3 with internet
Damage vs fire rate moment.
It depends on your usecase
I'm assuming open ai would be better if I put extra effort into the prompts.
Actually, open ai multi step search is pretty good.
Like a mini version of deepsearch in itself .
just ask o4 to improve your prompt
I mean DR is powered by a version of o3
but I think it has a higher tool call allowance to it
and prob higher context
No, like, even the o4 mini multi step searches are pretty good.
Use nukes.
o4 mini high is on par with o3
Permanent solution.
even sometimes scoring higher
Yeah but O3 is probably granted extra steps/context memory.
if you are a plus user context is the same
at 32k context
and I havent noticed any difference between tool calls
I asked o4 mini to compare battery lives per unit mah of all android skins, taking 10 phones of each brand.
Yeah, I’m pretty sure base o4-mini with search etc is limited to 10 tool calls
It was capped.
Wasn't allowed to gather data beyond a certain point.
I did the same thing with deep research and it didn't have issues.
Yeah thats different lmao
Deep Research has a way higher allowance in terms of searches
I don't know how to word it then.
I was talking about overall intelligence with o4 mini and o3 being about the same
O3 > o4 mini
If on the API
o4 mini > o3
Cheaper and performs on par with o3
For conventional cases, yes.
But if you throw something out of the box, the difference shows.
I want to see o3 pro though
Much prefer 4.5 for that type of usecase
No, like, problem solving creativity.
Exploring hypotheticals.
O3 is lazy though.
O3 + 4.5
Tends to give short responses.
The jailbreak makes both o3 + 4.5 monsters
O3?
Yeah. On you.com atleast.
Could be artificial nerf.
For cost cutting.
Fair enough
Theres an aritifial output limit on you.com
Oh, I use o3 mainly on chatgpt. I had both o3 on you.com and chatgpt interact and it was interesting
There should be a way to combine O3 and Gemini 2.5 pro.
I wish, you.com should make it possible
That would be really nice
Monica has deepclaude.
R1 + 3.5 sonnet
Monica has some questionable pricing
Chatgpt model selector makes it possible for 4.5 + o3, the analysts and explainer. The gemini 2.5 pro + o3 is possible on you.com
Ikr
They say unlimited AI chat but its not unlimited
I remember when it helped make it easier to access o1 when it first came out
No one is. You get rate limited if they suspect unfair overuse. It's always there in t and c.
"The Unlimited Plan includes 1,600 unrestricted queries for advanced models per month. To ensure service quality, query frequency will be adjusted to 100 per day once this quota is depleted, preventing resource misuse. "
Nice
The only one unlimited I noticed is chatgpt pro, others no clue
I might as well get t3chat
How much is that?
chatgpt pro has been really good so far
👀
its for light users
But once 1500 is over, you are done.
Yes, went past 10k o3 messages and no limits yet
Also, you are charged extra for Claude.
Even with o1 pro, 4.5, etc.
On monica?
claude is not good
200 dollars. I could never afford.
Damn
That sucks, you.com and perplexity do way better than that
I just used my 100 credits to generate images
theres 2.5 pro, o4 mini and most other models tho
Hopefully they make it cheaper for everyone, it’s useful since api is a fortune
Sounds awesome, I’ll have to give them a try since I wanna try llama 4 maverick on there
I mean they give a couple 2.0 flash queries a day for free
also llama 4 maverick was really disappointing
I would take x premium over t3 chat.
theres suspicion of meta manipulating benchmarks
t3 chat isnt for everyone
They should make a cheaper grok subscription.
I'm so happy with qwen considering it's free.
I use 2.5 turbo for non reasoning 1 million context and Qwen 3/2.5 max otherwise.
It works okayish with images.
There's a problem with qwen though.
If you accidentally end up generating an image in the middle of text conversation, you can't generate text again.
Even if you try to edit last prompt, it stays locked on to image generator.
wdym its working for me
just me?
How come?
I have this error but I can access other website
ERR_CONNECTION_CLOSED
@distant aurora they fixed file uploads for o3 a while ago, it’s op helped me with modding multiple consoles
How do you accidentally generate an image, mine doenst even let me generate images if I dont click on the image gen button
Its just not as good as the benchmarks seem
Have you tested its context memory?
No, I only have access via you.com with offers 200k window
And I used most of my credits on t3 chat
No, like, I thought I could switch back to conversation after generating an image in the middle of conversation.
I had ended up clicking image accidentally instead of turning off thinking.
lmao
yeah not sure bout that
@left stream https://www.reddit.com/r/LocalLLaMA/comments/1jvw91v/notes_on_llama_4_the_hits_the_misses_and_the/
I couldn't access perplexity since yesterday.
check if its your isp
and try using mobile data
you might get blocked if you are on enterprise wifi as well
Damn too bad at least o3 exists and claude 3.7
Hey add upi payments on api billing 💀
Buen día
доброе утро
सुप्रभातम्
Günaydın
صبح بخیر
早安
does anyone can talk in their mother tongue here ?
Bonjour
Hey all, my iPad app doesn't respond anymore, this has been going on for about 10 hours. I have restarted the iPad, uninstalled and reinstalled the app. I have Pro. Any suggestions?


