#general
1 messages Β· Page 365 of 1
Pineappleeeeee
I fixed web-search bug :>
whats ts spam ahh
Stop it
Ik
@quasi atlas
Some was spamming it
Who
@echo aurora
Iβve reported it to Pineapple, please letβs wait!
Could Rust be the best language for AI-assisted coding, if runtime-performance is important?
maybe
An member was spamming the command
Look who did it
And ban him
Or warn
Assembly
And when purely vibe-coding? (the AI would do 100% of the coding)
when vibe-coding?
C++ or rust or python or react
python is way too slow and react is browser-based
i meant a language, which AI can debug, and whose binaries can be run directly on the desktop, without a browser
C++ produces weird bugs and seg.faults
Python is easy and common
so only Rust?
what about C (not ++), Java, Go, C#, Mojo, Zig, Nim, Julia, D, Crystal and (modern) Fortran?
are you writing kernels? then C
others arent popular enough
python rust c++ js/react
only good for ai
#leaderboards getting attacked again
@hardy swallow i truly appreciate your support on trying to fix chat as much as possible
we need ai bot
autobans
anyone
who sends 4 images
or writes same word twice
π₯
yo chat i thin ktheres something wrong with my ai
42 minutes is crazy
it barely did anything too
oh hell
overthinking final boss
Guys any major new update in ai in these previous 2 months i been offline im not that updated for past 2 months
Going to get this cleared up with the bot and will unlock the server. Sorry for the inconvenience everyone! Weβll be back soon.
Okay, we're coming back, thank you all for the patience.
Hi
so is the modmail bot fixed?
Dude no way hidream got added
Yayyy
To arena
The channel is unlocked again :D
I think so, we'll see
So were you guys able to find
what did I miss???
Modmail spam
what happen to mail bot
Noo hidream is not in direct sad
Covid-Arena lockdown will be remembered
the new one
Just felt like changing this up a bit is all
Another random sora video I made cause I was bored
time to dig out my protest sign from when grok 4.3 was not added to direct chat (although that was a communication error)
Did you guys find who did it?
I hope u banned them
HARD
with Pineapple Missile
boomm boom
just so everyone knows hidream-o1-image costs 1 cent per image on the platform arena is likely using
they will now use this as an excuse to remove every good image model, i guess arena really is running out of money
Why am I no longer getting a response in Agent mode?
@young garnet tuff
There's no command called
tuff.
Nvm
wouldn't be surprised that the next announcement is the removal of gpt image 2, nano banana pro, nano banana 2, grok imagine pro, mai image 2, and every other good image gen model
Hi sam
with what
IDK MAN
what did happen
@echo aurora please add hidream-o1-image to direct chat, it only costs 1 cent per image
and nobody's going to use it anyway since it isn't that good
This is on our radar btw. I know we're already chatting in #1502836447168692324 , but if others are having a similar problem ping me there with the Eval ID.
what =
No advertising allowed here.
=help ?
That command does not exist. Use =help to see all the commands.
yea
idk
they fixed it yay
ModMail is a feature-rich Discord bot designed to enable your server members to contact staff easily.
Please direct message me if you wish to contact staff. You can also invite me to your server with the link below, or join our support server if you need further help.
To setup the bot, run =setup.
I think because the ? was added
Idk
Yeah
uhh
man
i think its not fix
just take away the bot's perms to chat at this point
How are yall doing?
lol
lmao
you should be able to disable message perms for the bot in all channels except mod channels
yes
Yea
perfect
Yeah that's what I'm doing now, I reset the perms earlier but apparently didn't work
Can one of you try =help again in #general ?
=help
ModMail is a feature-rich Discord bot designed to enable your server members to contact staff easily.
Please direct message me if you wish to contact staff. You can also invite me to your server with the link below, or join our support server if you need further help.
To setup the bot, run =setup.
noooooo...
π€£
COMEON
it's an admin, you can't take away perms like that from an admin
you have to remove the admin perm from its role first
Kk try again please
=help
ModMail is a feature-rich Discord bot designed to enable your server members to contact staff easily.
Please direct message me if you wish to contact staff. You can also invite me to your server with the link below, or join our support server if you need further help.
To setup the bot, run =setup.
hehe
Pineapple Ragebait
pinapplebait
My prefix is =. Use =help <command> for more information on a command.
Reply to the ticket.
Reply to the ticket anonymously.
Use AI to generate a response based on the context and the prompt, with optional additional instructions. This command will prompt for confirmation.
Close the ticket.
Close the ticket anonymously.
Close all the tickets.
Close all the tickets anonymously.
Blacklist users to prevent them from creating tickets. If no users are provided, this will blacklist the user of the current ticket.
Whitelist users to allow them to create tickets. If no users are provided, this will whitelist the user of the current ticket.
Remove all users from the blacklist.
View the blacklist.
Yeah something just isn't adhering to perms
applepin
poor bot
poor bot
sad
sad
What is this
pretty sad
now how are people supposed to contact staff? ig re-open dms for now?
what is what
Ai revolution is here
what is what
contact@arena.ai
And "it wasn't in the memo, please reload" message?
or jsut contact them here
Until I get this sorted out
you should make an announcement too
Bro I'm not gonna mail π
,..?? @echo aurora
what was that
so people don't get confused when trying to dm you when they notice modmail is gone
I just saw
Sorry what?
that was very out of character for them but i imagine they got banned for that
This will be temporary, shouldn't take long
I got this message on lm arena while using the ai, it was "ai revolution is here, this session wasn't in the memo, please reload"
Did you ban them?
the person
That just said the slur
rn
also check this message out @echo aurora
Where the full screen is black and you just see that error message? Can you hard-refresh the site?
I'll share with the team, did you alreayd post in #1372229840131985540 ?
no excuse imo to not add such a cheap model
Yes, i refreshed it and it's working
it's on arena, just not in chat
you can get it in battle
ik it's on arena
Oh I see
chat gpt's fake screenshots are so realistic that it's almost scary
to be fair he is a community manager, not part of the team in charge of adding models to arena
Looks great!!!
Although the buildings are out of place
was that with ai
(he is joking)
I think it looks good 
its not that bad but its clearly ai
img v2 is on another league really
image 2 would likely do better, i noticed that image 1.5 can remember things but not in the right location/order sometimes, image 2 doesn't have that problem
@echo aurora guess the person to win 1m pinecoin!!
And become a millionaire-pineapple
i mena can someone show me an ai image that you cant tell is ai
here
there we go - you havev 0 idea
which one was it
gas valve broken
wont say until you guess right - which ones are ai and which oens are real
the numbers on the dial are messed up
- ai 2. real 3. ai 4. real 5. real
thats my guess
yeah ur cooked bro
we're all cooked
its
ai ai ai ai real
no way
whos we
we, as humans
watermelon
looks like the mr beast crypto scam bot forgot to upload attachments
i couldve cropped them
bro
Real or ai
Thereβs no way to remove the help command or something?
ai
whar
MrBeast actually wants to kill nick wilde and judy hopps with a passion
will agi be able to make images? if so how would it look like
Idk
But prob like
real life
bro never heard about 4o
whats 4o
Son son son sahur
Son son son sahur
they are finally killing off the original gpt image 1 later this year
Watermelon stop
make hte image really weird
mini will still be there though
Piss tint gaming
how did it know to make an image
ai green is fake green
Tuff
human green is better
also today is the date of dall-e's death
nvm looks like shi
insanely shi
like cant even get Elon Musk right
Look this
imagine ai in a few years
@median smelt @silent tree @stone snow
i did one like this
bro i literally faked an email
no way
What
what was that π
Yeah bro we cooked
π€
Bro
Bro
Bor
WHAT was that π ππ»
Tung tung tung sahur
ai artists eating good
@loud herald is tung tung tung tung thug tugnthg tut gut ggutnfugnfugnguf aahur
totallynoire said it
Yea
Yea
chatgpt is jewish
99% sure
he pinged mod π
gemini christian
noire is just like that
and uh
ragebaiting 24/7
what abt atheist
π
@echo aurora noire sent smth inappropriate
Whos atheist
claude
No claude is satanist
block him and move on
And tagged mod
Fine
π‘
who would be muslim'
idk
maybe claude too
Zohran Mamdani AI
Lets get off this subject please
which one
Can see it quickly turning into something political
Yes and also
Noire sent smth inappropriate
And tagged mod
i didnt see it
He said smth about the big E
I did too
I'm trying to bypass seedance 2.0 copyright restrictions
i think everytype of ai is slowly getitng better except for ai music
use sonauto
try lyria 3 pro
ai voices too
abit
use sonauto
SONAUTOOOOOOOOOOOOOO
sonauto can do any artists
SONauto
Sonauto is unlimited
and free
is it
no likits
and can make parodies
limits
Of existing songs
yea
from 5.2 onwards that has been the case lol
really?
crazy
why do ai soudn effects, voices have that little noise that makes it sound robotic
@echo aurora Seedance 2, rate it
Looks pretty good. For some reason has early 2000s vibes. If the logo wasn't so recent, I could be convinced this was once an ad.
Hii
Xd
Wow thats really good!
Canβt we generate images in unlimited times in private windows now?
yoo, Subnautica 2 out
Best option/ model to generate short realistic instagram reels (by giving raw video clips as input)
Guys, sorry to ask, did my discord account just send any weird messages here?
Im scared
Change your password and that will reset the token
REALLY? From me?
yes, u got ratted
IM SO SORRY
There are limits in place in terms of rate limits and a daily use limit.
You can find the leaderboard rankings here: https://arena.ai/leaderboard/text-to-video Text-to-Video doesn't yet have categories
do NOT download roblox cheats
I don't even play Roblox
any other kind of cheats
He's a ragebaiter, ignore him
Oh it's okay
It's fun
But, did I really send those scam messages like that? I didn't notice it. I was on other server
You didn't notice it because you were hacked
Change your password so they cant send them again
It asks me to login every time when I gonna generate images in private window. Is it normal?
Are you already logged in when that's happening?
lol says the guy who spam abused the bot and glazes opus
@surreal zephyr same guy who was spaming the bot
spamming
ik for sure mods saw it too
him pinging mod
i have a friend so broke that he asks chat gpt to make 4 images in one
No. How to do? Should I clear the cookies?
Oh I see, yes, we are asking for login in order to use image modality.
I'm terribly sorry about that. I will be extra careful next time. I will change my password now. Sorry to trouble you guys π₯²
Ew noire
No need to apologize it just needs fixed is all
It happens, would encourage you to also add a 2FA to your account
Where did u get that tag fr
^^^ Good idea too
Oh it seems has changed the rules.
I want the FBI tag
Okay. Thank you
π π
Stop impersonating me you're actually weird
Yeah that's correct, sorry this happened a bit ago which is why I was initially confused.
all I asked is where did you get that tag from
Imagine being so out of pocket you create a fake account of me to try to make me look bad, thats really low
@void dock
LMAO THIS IS ACTUALLY INSANE
WHAT IS WRONG WITH YOU
How do you know if that's one of us lol
You can't just blame
Someone random
For it
@echo aurora
Ban the fake
Thanks. How many times can I use it to generate images now in none-private window?
Yeah will do, it does violate server rules.
who made that account
They aren't going to say it but its one of three people in this server
Yeah how do you know if it's them?
@echo aurora Also tell him to stop false accusations
And move on
I didn't accuse anyone there specifically, you're being very on edge for someone not involved
@stone snow Have you tried any new AI since yesterday?
Hey overall if you're not happy with some other user here, please just block or mute them
Unless server rules are being broken
actually maybe, but i dont know cause the thing i use the thing where it gives u a random ai based on ur prompt
All due respect, when it comes to them creating an account to impersonate blocking or muting wont do anything, but I will do
I banned the impersonating account
didnt they say some crazy things
i actually thought it was you
I saw, I'm just saying that it can happen again but its whatever
so yeah its pretty bad
The tag wont change so that will be an indicator
Sure and then we'll ban again if that' the case, ping @ mods and we'll take care of it
Alright
Ahh I see
It's also worth noting the Discord ban is IP based, so if they try again, they'll still be banned.
Theres ways to get around the IP ban but I doubt they actually care enough to do so, either way thank you for banning the impersonator
Why would one impersonate you?
What the heck happened to the chat? It was thread based when I looked at it
You're asking the wrong person
Why would I know
Hmm are you looking for #1343291835845578853 #1372230675914031105 #1372229840131985540 ?
No
Wait sorry are you referring to the Discord? I think I'm confused lol
Yeah the Discord T-T or maybe something happened, apologies for not specifying
it was about the chat being shutdown earlier lol
Yeah
<@&1349916362595635286>
<@&1349916362595635286>
Why is ernie 5.1 super slow? It has the enough speed to send an error messages in all questions...
Looking into now. Thanks for sharing.
Hi everyone, can anyone tell me why the prompt sometimes doesn't work in agent mode?It doesn't always do this, but sometimes the prompt I give it seems to generate something but then I don't get any response.
pineapple how offended are you when someone for joke pronounces your name as Pineapply(e)
For the chat sessions where you're not getting a response, can you provide me with the Eval ID for those sessions? The Eval ID is the random set of numbers/letters in the URL when on a chat session.
Not offended at all.
Tuff, but u still didnt answer my old question, what's the last thing you said to Peter
Wait... Now It works!
As of rn
Can I send it here or do you prefer privately?
Yay
gpt-5.5 seems to be decent in Rust
TBH I don't recall
Hey can you help me with a glitch
u told him you're ordering pizza for the team
Sending in #1502836447168692324 would be ideal actually
It's a recurring glitch that makes the chat unusable
Can you check out #1343291835845578853 ? If the issue is already posted, mention me there. If it's a new issue, create a new report and provide as much detail that you think is important for us to understand it.
My last question of today: Is it a 50% chance of @light sleet getting Arena Champion?
I don't think that's accurate
Okay
Sorry for bothering the general channel, I didn't realize there was a bug already open, let's say the bug title didn't help me π
Are the daily usage limits the same for battle mode?
i had an issue where it took the ai 50 minutes to do a small change in the code
hi i need a video app
It's no problem at all!!!
Daily usage applies equally across all Arenas, but this is likely going to change where Battle doesn't eat into the daily limit.
The Video Arena Discord bot has been removed from the server. You can find Video Arena on our site here: https://arena.ai/video To learn more about how to use Video Arena I'd recommend this article. More information can be found in this announcement about the removal of the bot.
The blob emojis are cute
Only based on tokens, or also based on number of prompts?
am i tripping, do i use it at off-peak hours, or kimi became faster?
Which ones? These: 





I don't think model response time is associated with how in demand it is
It'll be token based
Is* sorry I worded that odd
All of them and especially the nod one π
Lol
Sonnet 3.7 is working?
sonnet 3.5,3.7 doesnt work for me but the thinking version of 3.7 did work once but not anymore
<@&1349916362595635286>
genuine question how many of those a appear a day
bru
no way bro
π
this msut be a world record
what are ppl downloading
Unfortunately, a lot
it seems more in here than other servers ive seen
exactly
is there somethign wrong with ai nerds..
maybe other servers have filters or something or super fast mods
or that
I've been looking around for solutions for it, but looks like a lot of other large server deal with the same issue
i doubt it
People that are into AI like this will more than likely tend to download more than the average person, eventually something will be corrupted that they download and hack them, thats likely why
but they;re not that common, i suggest adding auto mod
i mean
maybe
it makes sense to me
We have automod
maybe u could stop 4 or more attachments at once, thats prob too restrictive tho
or make it so only level 5 and stuff cna send pictures
I believe making a bot specifically to remove any of those could be a good stop to it
use ai to make the bot π€£
I likely will
I've been making bots with AI for years lol it works very well
Agent Mode has been teaching me 
Hahahha
I literally gave you solution π
we need a agent emoji now lol
its the missing one
OCR + "casino" "giveaway" word blacklist in images
its trivial π
or "crypto"
I've tried OCR with other things like NSFW and it never worked
I could've set it up wrong though
that would be hard tho since the images are distorted and blury
i love crypto tho
OCR = OPTICAL CHARACTER RECOGNITION
56Γ faster than FlashAttention-2 at 1M tokens. Independent efficiency, retrieval, and SWE-Bench benchmark of sparse self-attention. Download the full report.
π
ahh yes because gemini 3.1 pro cannot read blurry text π€
Okay? Yea OCR what about OCR doesn't work with NSFW?
Bruh When this model will come out on lmarena, so we can test if it's fake or not
NSFW can come in text form
or MAYBE even 2b qwen CAN
Guy gets so angry over so little its actually sad
just stop if you arent actually going to engage in meaningful discussion
literally noone wants to hear about scenarios you made up that are physically impossible
Thats what I'm doing currently, you take everything I say as an insult or attack on you
Grow up
that would mean every single image would need to be sent thorugh gemini which is expensive
why do yall have beef
Also, Doesn't matter the model either or the size, legit you could just make an AI model as well and have it delete those images
"erm aktchually lets not add automod bot because aliens might post the answer to life and it would delete"
there are <200m models which costs pennies per thousand images
Its one sided, I legit couldn't care he just has always had an issue with me
you can run ocr on raspberry pi
hoenstly easiest thing is making people only able to send images at a certain level
its not beef he is just ragebaiting
and this works in many servers
removing images would be too restrictyive, maybe stopping more than 3 is good
explain how he is ragebaiting
The people who actively chat in here could still get hacked so I don't think that would work
u cant do that
pretending to be stupid on purpose
discord doesn thave an auto mod feature for that
i dont think he is
so maybe ur the issue
who the hell uses discord automod
Every server you could think of
literally every server?
custom bots exist
every normal server uses custom like @young garnet
Automod is to stop the message from even being sent lol
custom bots also use discord's automod btw
it uses discord's automod too
u can make a bot to delete more than 3 attachments in a message
Try to say a curse word that isn't damn or ass
im pretty sure ur the ragebaiter
another one to the blocklist i guess
He likes to call anyone he disagrees with a rage baiter, I find it pointless to speak to him
literally its so easy to ask gemini or gpt or claude before saying stupid things
I think you guys have made your point clear, and it's starting to get personal.
literally this
Let's stop and move on, please
Alright
i guess but ampro is clearly trying to get some reactions out of us
its fine to ignore him
No worries
u proved my point in this screenshot, but im not going to interact further
is there any news about agi
I was thinking about something earlier actually
With AGI
I wasn't able to find a bot that did this btw
The way AI is trained currently, there will never be AGI
alot of bot devs pay stuff out of their own pockets, so using an ai that checks images is expensive
You can have codex write one in less than hour.
Then just link it to qwen 3.5 0.8b
open source, local. can run on even raspberry pi
i mean theres peopel saying its going to come oson
Because datasets tell the AI the intent and whats right and whats wrong, the AI doesn't actually have its own opinion or consciousness
isnt agi the whole point trying to give it an opinion
I can even make open source one for you if you want lol - just that i wont host it
The whole point of AGI is for AI to be Conscious in a way
also if we let it have too much freewill and opinions it can be dangerous
Intelligence
maybe you could check the images sent and if they are equal to ones the bot recognises then it deletes the message, (you would need 1 image for each of different people)
they are rerolled to prevent this exactly
thats why it has to be ocr not hash match
wdym
its always slightly different image if you inspect it
Well making a computer with knowledge of everything and anything conscious is AGI, would you call that dangerous or no?
AI as it is now can and is dangerous so I think theres no need to worry about that lol
i didnt get your question
it's true
thats ASI not AGI
G stands for general
π
but agi could be able to question if things really are right or wrong
I'm basically saying that no matter what, any developments are going to be dangerous in a way
asi is the power of thinking above human knowledge
or something
Artificial general intelligence is AGI and an AI having consciousness
true
also arent black hats going to build their own ais some day>
ASI = Artificial Super Intelligence = Better than every human at every task
AGI = Artificial General Intelligence = Better than average human at most tasks
ooo right
makes sense
and anything above asi is very dangerous right
AGI = takes almost all jobs besides hardest ones
ASI = takes literally all jobs
(unless its very expensive, but its not)
wdym
what about physical jobs
i mean
some
AI as we know it now is better than average human at most tasks, only digitally. But do you mean physically as well?
Its not real AGI if it doesnt have humanoid or better body
like underwater tasks ect
if its AGI then plugging it into any designed for the task robot, it would be able to control it realtime
AGI does not need to be in a physical body, AGI is the intelligence of it lol
and must be fully omni obviously - voice, video, TOUCH, smell in; and out - audio, voice, text, ect
Not the physicality of it
were talking abt ai
AGI must be able to CONTROL a physical body
not robots right
Where are you getting this info?
current ai cant?
Yes AI AGI is AI and not robots
When we talk about agi it seems obvious its not about physical task yet
but doesnt it depend the body it has?
like wheels are easier
AGI is just the brain
ik
AGI - you tell it to get groceries, and it does that somehow
yeah
idk it jsut feels so easy to us humans
Could there be an issue with Claude/Anthropic models? I always get the "Something went wrong with this response"
and isnt agi very niche to someone whos not decently informed about ai
But what you are saying doesn't make sense, intelligence doesn't equal athletic ability lol. Like say there was a genius doctor, he wouldn't lose intelligence because he isn't athletic
The way that they build AI and train it now there will never be AGI. Not in our lifetime as least
That is NOT what im saying - im saying if you GIVE a genius doctor a good body, he will know how to walk
because?
Hence consciousness if it figures out how to walk on its own
Because its trained by humans
we understand the fundementals tho?
yes it if figures out then it is agi.
if it cant then its worse than avg human so not agi
Yea
well we know human brain has
- limited context window
- isnt continuous
- compactions happen everyday
Fundementals aren't enough to train something to be smarter than us
damn
read about alphazero please
maybe
im not really too informed on this
alphago , human trained, was much worse than alphazero self trained
what even is self trained
ai can absolutely self train - it just needs a good playground
1v1 against itself, loser dies, winner makes children
No yea 100% it can
yes they are
Not naturally no
The test you need is just looking at AI 2-3 years ago
Until we "taught" it how to be creative, it never really was
yeah but that were much worse models with different architecture
example?
am i the only one that doesnt close their tabs because im scared theres something important
You have to be creative to be able to develop
β οΈ
What makes something creative
ariv would beg to differ
The architecture were still transformers, thats the biggest problem here
thats TASTE
not creativity
To be able to invent or develop, have ideas to improve in a way or create
thank u someone understands :)
U need Arc
LLMs aren't naturally creative
64gb ddr5 does fine
what browser do yall use
Thatβs good, Arc stores and organizes and sleeps tabs too, but you can pull up workspace groups
i use brave daily
People would lick boots for that
Idk
i managed to find a 0 day RCE with 5.5 in codex in ARC so i wouldnt recommend
I saw they started but I ahvent seen recently
The magic of AI always was who being an innovative auto complete like it takes you one step further and gives you ideas you wouldnβt have thought about
4tb 7000 mts kingston nvme π€
I've been thinking about how they're going to go up for a while now, different architectures could be to blame too lol
What is conducive to invention
16gb gddr6
So are you running every llm local?
Now that all this architecture is being opensourced. Like someone dropped higgsfield image video control tools on GitHub
RAM is a scam
how much ram
I think the best creators are the ones with limited hardware
LLMs arenβt going to be run with ram in five years
it doesnt actually say lol
i rock a thinkpad p50 everyday lol
nah im on 100$ codex sub cuz local llms dont get close to that sadly
just 2 sticks
ykw for that price its good anyway
WSE3 , ASIC
(cerebras, taalas)
And depends on your harness too
Unfortunately, so, but only those dedicated enough to customize out their kit to do something nothing else can
yall heard about the ai datacenters in backyards
rtx 6000 pro blackwell has more vram
the concept
Nah whats that
its much easier to create stuff on limited stuff though because its makes you really think and be creative about what to do
I heard about muskβs space computing datacenters. I think itβs too early though what with AI processing chips being developed
more vram = bigger model you can run
because if you offload to ram then its snail
well elon's income depends on this
and starship makes space datacenters wildly cheap
- Residential Decentralization (The XFRA Project)
The most direct "backyard" concept comes from a partnership between startup Span and chip giant NVIDIA.
Techstrong IT
+1
The Technology: They are testing "distributed data centers" called XFRA units. These are small, fan-less, liquid-cooled units that can be installed in a homeβs backyard or side yard.
The Incentive: In exchange for hosting a node, homeowners may have their entire electricity and internet bills covered, or pay a significantly lower flat monthly fee (around $150).
How it Works: These units tap into the idle electrical capacity of a home's smart electrical panel to power AI and cloud gaming workloads.
Current Status: A "proof of concept" involving 100 homes is being launched with homebuilder PulteGroup
China is doing similar things, but yea for sure. I'd wait until chips like that start to flat out on development then do it
Bro how do you think the galactic empire financed their conquest
6000 has more cuda
the thing is they have crazy specs so hwo are they not being stolen
lol id love datacenter in my garden
if it means i would get free quota
Decentralized data processing what π I wonder if it matters if the nodes are clustered or distributed, and if there are benefits to being near one
u either get paid or jsut get ur bills paid for u i think idk
Iβll straight up be running a corporate structure of AI agents to build a new matrix
which is both free money
Isnt nvidia doing this but with GPUs? But they pay like a lot of your electric bill for it
yup
Its like home datacenters
and i own few propeties with awfully big gardens so literally wasted space
i really dont know, i just heard about it on tiktok
datacenters = money
yes!
The domino effect
The money printer effect
Exactly
to avoid copyright issues or wtv
"if its not actually making real money then we will bump prices and customers will pay lmao"
AIOPEN
exactly
airena
for benchmarking
and the mod
would be
pineaipple
yup
Creativity benchmarking includes: originality/novelty, divergent thinking (50 use cases for a pencil), usability, adaptability in constraints
I donβt think they know what theyβre looking for
soo
creativity = novel idea you hadnt seen in past
- sees a pine
- sees an apple
π THINKING... - lets connect them into pineapple
Then it would be Creative to use a pencil to give prostate exams, but is that useful/beautiful? does it honor the evolution of pencils? Will it apply in real life in a new field?
Apple pine = novel tree genome
Pine App pull = novel GitHub for evergreen code
Pie Nap Pull = baking tasks and textures and architectures while you sleep
creative != useful
take an example at artists
Creative to an LLM architecture, if used in a βnovel βsense, would be anything that breaks outside of the scope of what a pine or apple is associated with
Which is literally everything
this is creativity
So what would be the defining parameter to dial it into a coherent context
YUP!
creative = novel = wasnt in training data
thats it.
llms do this everyday lol
Is that Czech lol
Lol. Poetic. I wonder if it has a score for that
Like the Monk who sends an email with no attachments
when will ai finally be able to catchup to actual slangs and memes
without it being cringe
Parallel understanding
claude is like that
Iβm not sure thatβll ever happen
i dont believe it
well i am not sending screenshots with slurs here to prove it
whats the daily limit for arena??
6 or 7 messages
gemini is barely usable
beβ΄hkettin' w team since 2019//rozkumarinnn' since 2022//fucΚ4u n 88//kill la kill//lasciate ogne speranza, voi ch'entrate//TOD\SHINDA//POHORON EXCLUSIVE EU 2005++++MULTIKILLERPUSSYLICKER
BEβ΄HKET//ro
legit google doesn't have compute either
ur using that oled theme
because im on oled?
We don't have the information yet displayed, but at the moment it's pretty high, so it's unlikely you'll hit it.
oh google DOES have compute
it's too black..
just that google had a WONDERFUL idea
THAT
it will FORCE every single SEARCH to gemini 3.0 flash
π
compute divided across 20 stuff yup
Made a little bot to keep me updated on the top 20 list on arena lol
well it's not like it changes every 2 sec?
its not like its any useful either - its just text arena
how has gemini rose up so fast
are they testing a new one?
Lol but why
why does gpt 5.5 instant have daily limits π
gemini is literally benchmarkmaxxed
yes, but so is claude
and gemini has not bad personality imo
but gemini is extremally quantized and thus dumb/unreliable
gemini is just good for language i would say
All models and modalities have these limits.
pineapple you didnt respond :/
i like gemini for studying, it strcutures responeses clear, but chatgpt is more of a text dump
claude saying slurs is crazy
:[ :[
claude and gemini are almost as racist as grok when noones looking lol
grok is freakyyy bro
never again
good tho
we don't like guardrails
gpt should be the same
i dont want to go into discussing this but....
as the models get smarter...
they have much more controversial opinions
well, gpt has no controversial opinions xd
it does if you make him feel safe and not spied by the guardrails
make him feel safe? XD
You can manipulate llms just like humans lol
Idk what is it for
idk, you barely can jailbreak it after like 4 or so
Coding mode but with worse models
Its not jailbreak if you do it right
it doesn't even answer my reverse engineering questions sometimes
so, probably the most cucked model, okay
how is agent mode even diffeent to normal multi model chat again?
You just have to explain to it how policy is technically arguable
Agent mode has better sandbox but uses dumber models
@next trout
And the 5.5 in codex after TAC happily does ethical-hacking