#programming
1 messages · Page 176 of 1
ah
in november of 2020 we put ourselves on the waitlist of the local big tech store, and it only took 5 weeks to get them
only 
a lot of people were waiting months, we were lucky to be 60th and 61st on the list
minor spelling mistake
I'm 11 hours into a 22 hour js course, I'm tired of it. So I'm gonna do what I'm exactly not supposed to do and watch a 1 hour one that is probably worse and will make me learn less by the end of it


my brother gets a lot of use out of his ps5 since he plays a lot of fortnite and stuff, i stopped playing the game as much when i got my ps5 tho
and 2 years later i became a pc user
that's nice atleast, someone using the console
better than it just lying around
oh wait
i misread
we have seperate ps5's, mine is laying around
altho i recently put it downstairs in the theatre room gaming area so everyone can use it
im trying to be less possesive with my stuff
expensive bluray player 

i should fix my ps2 at some point
the disk reader do be dedge as far as i can tell
in my experience, just buy another one, its cheaper
that xbox original i showed here before with a broken hard drive, we just replaced the whole console for 40 bucks instead of getting a new hard drive, a sata adapter, and softmodding it
maybe i'll do that, the console is pretty damn low on the priority list tbf
i just get reminded every time i see the box 
from what i can find locally, a ps2 phat is 25-40 bucks
the slims are slightly cheaper but i dont like them
I just checked what happens if I search for NeuroSynth
Not what I expected
its cuz you glaze it so much 
ok but 
awa
awa
vocal synthesyser
NeuroSimp
I do declare! It is verily apparent that this digital room of discussion has lost its spark of life!
its night in the US
sure but a lot of the #programming people dont live in the US
its 3am here, makes sense
i havent gotten sleep in around 48 hours
so it sounds like a skill issue
i just finished with my work thankfully
-# Maybe
it's 8am
hello
@sage crag how did you compile hbvm_c on your machine? i ran nix-build default.nix and got a compile error since the __USE_MISC gnu macro wasn't defined
just nix-build works for me
but you how
what
for me, the syscall function in main.c is not included from unistd because std=c23 doesn't define the macro
but changing it to -std=gnu23 works
i don't know why it compiles on your machine
it compiles on my machine because thats how its supposed to work

your machine is the funky one

does it also compile if you use -std=gnu23 in cflags
it should
also, why is your code made out of macros
no generics in c
this was the easiest and most performant way i found to do it

also, why c23?
yup
and it came from B i think
used to be just a local variable definition keyword?
funny how C has always had auto, but nobody uses it cause it doesn't really do anything outside of c23
isn't nix supposed to solve these kinds of problems? if the build script works on your machine, it should work on mine too, right?
weird
need nix expert
yuh
though i dont think it makes any guarantees about version compatibility
so
version of what
Packages
i need an explanation
it compiles on one machine but not the other, what is the difference when using nix
same build script
the script uses the same version of packages and such
do you use a different nix-channel than me?
i use unstable
oh i use nixos-25.05
is there like
a good way to keep reasoning models from getting stuck in infinite loops
during the thinking stage
noe

Surely a 120b model doesn't loop
Increase thhe temperature

IIt'll make iiiti harder for them to get stuck in loops aand make iit harder for them to stay coherent
Exactly what the world needs
Worst thing happened was
you forgor to set region
to this day, one of the greatest application features ive ever seen
in the windows mail app, if someone sends you an email with some keywords indicating they want you to send them an attachment back, and you try to reply with no attachments, it will warn you
its such a thoughtful feature
Unload shiro was good
dont unload me
Some how ??? Image back with profiles
Sad
nobody has a strong enough reference to me...

The only strong enough reference is in the moderate chat function. I can see why it got GCed
i dont have any shiro emotes bwaa
you need to get a static lifetime
Ve. cord
Word is ban
pointer to shiro:
shiro
everyone gets garbage collected eventually, once all references to them expire 
who actually invented his theorem though
I tried running a 20b model on my school laptop today
if you reattribute his theorem he might be toast 
safe to say it is only useful for eating so much ram the entire operating system fails catastrophically
unless you have more than 16gb ram that is
Bye meet tomorrow
makes me think of that one hospital meme.
"what happens when the data has no references?"
"we dispose of the data and make space for the next variable"
Hmm
come back shiro 
we missed the shiro 
programmers i am sad
there hasn't been any screeps in ages
the best city building game was simcity in 2013, hands down
and literally nothing close has come out since
shr will you play silksong
noper
see im not the only one
metroidvanias arent for me
@real sierra happy bday!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
unhappy unbirthday
how did the birthday thing even start?
once upon a time
At least it is less often than others
in january 2023
presumably when shiro was born
if you're gonna get hbd you might as well do it in #programming
you know what i mean
It's shiro's birthday?
Happy birthday 🥳 🎂
someone asked when my birthday was and i wasnt comfortable giving that info out in a public server
so i answered, "this year"
Big mistake
om
so that was jokingly interpreted as every day of the year?
jokingly? maliciously
but it persisted through the next year and so on?
they "wanted to make sure they didnt miss the day"
yes
om om
woudlnt that mean he ages every single day of 2023, but has now stopped aging?
and somehow, despite my birthday having passed 3 times since
it has continued as some kind of tribal rite
that would make him 365 years old
why is your username in a different font
damn
Doesn't help the kick me birthday role is stuck to you
its been bugging me and i just realised why
Unicode chars to change font iirc
i cant take it off
unicode U+1D5AF exists
Konii you should set yours to comic sans
what an unfortunate way to go out
The character U+1D5AF represents Mathematical Sans-Serif Capital P
anything is mathematical nowadays huh
nom
i cant believe it
how did simcity do the city builder genre so right
and literally nobody since has even tried to match
cities skylines 1 was hollow
and cities skylines 2 was vibe coded
simcity wont get a sequel because the studio got exploded
u mean simcity 1989?
simcity 2013 is the one i got to play
or 2013?
and imo is a great one
ah
its seriously the best city builder i know
smooth controls, lots of interactive things in the city, organic problems, organic buildings
havent played it before
its great
now that i have money i should start playing these games everyone always talks about
one of my fav features is that each base building (police station, city hall, etc.) has expansions you can build modularly (more fire trucks, more patient wings, more jail cells, etc)
𝕭𝖜𝖆𝖆
incredible level of customization
each of those expansions has a physical thing that you place too, so it all has a visible footprint
and they of course generally all have an actual function and upkeep, not just visual
how did they do this concept so right and nobody since has copied
SimCity had a terrible launch. It required an 'always online' internet connection even if you were playing alone, and players would be booted from the game if their internet or SimCity's servers went down. SimCity's servers went down a lot, particularly as millions tried to log on in the first days and weeks after release. Lead designer Stone Librande describes the launch as "heartbreaking".
i guess it was generaly poorly received at the time
the launch was poorly received because people couldn't play, not because the game was bad
the only valid criticism is that the city building areas were small and couldn't be expanded, which was done for performance reasons
R9 290 for amd
Discord
geforce 700
bruh
It's just missing eyeburner pink
not as bad as i was expecting
keep in mind these are the cards coming out in 2013
780ti is still a third of the performance of a 3090
so nobody had them until a while later
oh true
Your wish is my command
YYYYEEEESSSSSSS
hmmm
i think my pc is good enough
I should have done this in light theme
256MB of vram is less than the amout of vram the igpu's get these days
and that vram is built into the cpu chip
they still mentioned you needed a sound card...
didn't mention the required internet connection though
xbox was doing always online, im sure it was fine 
legendary game tho, it was just so fricking good
i should make a city builder some day
your industrial plants would catch on fire sometimes and then just explode and level 6 city blocks
oh ye shiro, have you heard of the #programming triangle arc?
you went after milk for 2 months so i think you missed it
triangle arc?
we made a triangle run at 102K fps
left for milk for 2 months, hates us
why
hi vanor
hi :)
cuz big number go up
high performance triangle
quack overclocked his cpu to 7.6GHz to do it
Does that require nitrogen, or can modern CPUs just do it nowadays?
they upload like once a year but its always gold for people like us
thermal headroom is important ye, but its possible
it would seem that i am already subscribed
i dont remember doing so but sure
have you already watched every video then
oh it was for the harder drive video
i will watch the rest of his content then sure
the older learnfun playfun videos are also funny
machine learning for playing games, but from 12 years ago
before his time truly

The Intel Core i9-14900KS does 6.2GHz out of the box, and you can jsut throw more watts at it if you want more
my cpu seems to cap at 5GHz with pbo, and im not about to manually overclock
I wonder what will happen if u disable all cores but one, and throw watts at the last remaining core 
i dont know, but then it would have to run system tasks ont he same core as the triangle
so im not sure if it would be faster
also i dont think a single core like getting 16x more power than it nornally uses
so
just to understand the tassk fully
we dont have to draw the triangle to the screen do we?
just have the geometry written to some buffer
Ya having 2 cores would probably be a good idea. If u just kill all the processes on the system it might work out tho.
I don't think anyone has yet wrote a custom minimal OS for just running the triangle tho
Sure does 
surely you can divide the triangle into even quadrants and distribute them among multiple gpus
or is it cpu only for drawing
Well no u see draw call is submited to GPU from CPU, and for that electrons need to traval across wires.
So at a certain point you are gonna get bottle necked by the speed at which the electrons travel across the wires.
surely the gpu can just keep drawing without the cpu instructing it so
its not like it needs any information from the cpu to draw what its drawing
it should have all the info and resources on-board for this
have it make like 20 frames at a time too, no need to make frames in order
Make an OS that will run on the GPU, so you could submit the draw calls directly from the GPU to the GPU. What can possibly go wrong
we do, thats the only requirement
the whole point of the chalenge is to optemize the rendering
yeah ok then i mean
I think they are ordering like 9 frames at a time
that optimization just comes down to removing the entire pipeline
and hardcoding all of it
if we skipt he rendering and just keep reussing the same buffer we get roughly 6 billion fps
perfect
if you kill the blue and green subpixels in a triangle on your screen you can get infinite fps
i decided that i didnt liekt hat solution tho since it isnt actual rendering, so i banned it

i woud say i dont make the rules, but i do
like
i guess the spirit of the challenge is that you should actually be putting the right data into the buffer each frame independently
so yeah the optimal solution is just to make a minimal operating system that loads a program onto the gpu to do that forever
the only requirement is that the buffer gets emptied during the start of a frame, and you gewt a trianlge aty the eend. as in, the triangle got drawn to make the frame
any specifics on the triangle?
so far we havent made a full os, people jsut used linux
can mine be 3 pixels in the corner?
yes

my pb of 47K fps is with a 5x5 window on software rendering
it sounds like a fun challenge but i feel like getting any score i would be proud of would risk bricking my pc
yep
couldnt make a normal challenge like a chess engine tournament or something, it had to be soemthing that bricks your pc 
i have deemed beating quacks 102K fps to be impossible without risking the death of my poor 5950x
yep
the goal is mainly to see how people optimize the vulkan pipeline, so i can steal their ideas
well yes but i cant make games without a pipeline
"not running the game performces better than running it"
Sam is stealing exactly 0 of my changes
i love the idea
but i feel like the core issue is that the best and easiest solutions all involve getting as far from vulkan as possible
tbh looking at your code, i still cant really figure what the fuck you're doing
either by something smart like a custom os or something dumb like getting more gpus
Dwugs
I am in a superposition between breaking reality and breaking my sanity at all times
so far we seem to have gotten to the conclusion that just not using a gpu is faster, cuz you can keep the data on the cpu then and dont have to wait for the gpu latency

i feel like proofs are just a metric that assign some believability to a fact
and deciding you believe one proof more than another is just a comparator
so really you should be able to sort the set of all proofs
so i wonder what the worst proof is
and what the best is
isnt the whole point of a proof that it is true? i dont get why you would believe one more than another
i guess im thinking more in terms of studies than proofs
ahh
poor word choice on my part
if i have some experience in a certain subject i might have some opinions on studies. but mostly if i dont know anything about it i dont hold much value to any studies
im becoming more and more fascinated by search algorithms
theres some insane algorithms out there
yeah, just tried BS roformer again, I'm very impressed... I think if the vocals are quiet to the instruments, it will give a lot less fullness on the vocals compared to other models, due to the bleeding reduction.. that's what made my opinion of it not so great... works really well for a lot of songs, though
Google DeepMind Finds a Fundamental Bug in RAG: Embedding Limits Break Retrieval at Scale https://www.marktechpost.com/2025/09/04/google-deepmind-finds-a-fundamental-bug-in-rag-embedding-limits-break-retrieval-at-scale/
ugh, okay... uvrv5 just sounds better here... even though the model is like over two years old.. it just heavily varies on the song...
lol, I keep going back and forth so much
I've just settled on using Roformer always
It just gives the most suitable vocals for all my usecases
@desert plaza it only solves it if flakes are used
otherwise you must also use the same nixpkgs version

yeah, I get that low noise is super important for training models
but.. here it sounds muddy and the vocal isn't as "full"... sure, there's more noise and instruments....
but, it varies heavily on the song.. sometimes the results are really good
guess there's a lot of different roformer models to try, that one is a July 2025 model
I've found this one to be good, too
I guess I just really want a model that works for everything.. alas
back to sleep 
The PCVR pipeline always works with zero issues
do we have any file sharing tool for my pc and phone for windows?
localsend, kde connect, uhhhhhhh
Roformer is also superior for raw-RVC covers
idk how localsend worsk mine broke when i changed a wifi setting now i cant send any
but whats kde connect?


i dunn
did you check the settings?
KDE Connect is probably best if you want integration in general
otherwise Tailscale mabe (though their file sending thing, Taildrop, is technically still an Alpha feature)
follow-up: the peak meter doesn't measure inter-sample peaks, which are loud as shit in this case and don't get clipped (they just kinda become quieter). the value i'm looking for is "true peak" which measures these and is at +5dB. that'll absolutely will fuck up how it sounds on a nice system by inducing digital clipping, and subsequently high frequency noise on your DAC when these happen, and in my case they always happen
Silly
Meanwhile I just throw stuff at the thing, then keep exporting at different volumes until it's not clipping
you wouldn't know if you're true peaking unless you specifically check, but you don't need to push for loudness in the first place
i doubt you are
i technically can just turn down the master like 5dB but that would make it too quiet and would actually peak at -5
but also i am clipping, i am clipping crazy style, look at the message i replied with the follow up to
that's the point, to clip
Yeah I'm not going for loud, I'm going for audible with a good vocal and instrumental balance
i mean that i doubt you are true peaking, because i know you're not going for loud
Most things I make are exported at like 80% volume or something around that range to make them not clip, just some sort of sillying around
you should look into limiters
I have no clue what those are
aggressive downward compression
So compressor but it flattens down instead of up?
to specifically "limit" your peaks
-# Note that I don't yet fully get how a compressor works as I only recently got how to use them

ye i figured
For a long time I was just not using compressors because I didn't seem to get what they do
Then on BOOM NeuroSynth I realized I can get a much more equal volume profile with one
i reconsidered, 12 hours is enough for now
NS-B-3.0.1 is inherently quiet sounding
a limiter is a compressor that is very aggressive in the sense that it will never output signal louder than the ceiling you set, usually -0.1dB to 0dB
specifically to avoid clipping
So if it sees things going over rather than making it a flat line it becomes a curve?
Silly
ikr???
I just never realized what it's meant to do until BOOM
i tried to explain
After BOOM I've been using compressors a lot more
i don't think it worked 
i think it's time for you to invest in a proper DAW instead of slapping whatever in audacity
I'm using LMMS
he's on lmms i think
Can't afford anything better
you can use reaper
they technically have a 60-day trial
but you can fully still use it after indefinitely
it's basically like winrar
mhm i suggested it already in dms
available for linux too 
WinRAR is the kind of thing I use once and forget about forever
I just use 7zip and nuked winrar off my pc
I think this is kind of intuitive, right? Like past a certain point you start running out of 'space' such that you effectively start hitting into hash collisions
More dimensions, less chance to collide, more vectors, astronomically less chance to collide, ya know
I also just used 7Zip once I noticed it does extract RAR files too
Now I'm on Linux though so I just use the archive manager that came with it
peaziip my goat
License problem
this is my bar
there are many like it but this one is mine
Need help too much customisation
What should I learn next ?
literally just because I haven't accepted eula
chat,
i went to a library for the shiggles, i kid you not there's this whos working on something asked a random guy beside him about a project
he explained:
•they have a website, aparently its private, they take data from the website and want to put those data in a spreadsheet or excel, but apparently his excel dosent let him run scripts ? hes also concered about future proofing or sumn and he cant understand python and he cant access the site bro wtf im confused too, idk I'm just evesdropping
his reason of taking the task was he wants to impress his boss😭
om
Hmmm
Game crashing it self
So about that AI enhanced pro res zoom on the pixel 10 
Davros is looking a little funky
Ai getting better day by day
Thing is it isn't remotely how it should look (also christ he needs to dust that figure) 
But ai can't help in hacks (Gemini, chatgpt) 🙂
Based doctor who
honestly it did a pretty good job at reconstructing a face, albeit the wrong one i guess
Mission failed successfully
it's not like it could've done much better realistically, with how awful the input image is
Moon race
It could of
i've seen worse ones
like it deciding that a brick wall is actually wool
Speaking of this is what happens when he feeds it into gigapixel instead
I like to see evil vs gpt
5
id like to see gpt5 beat me
Sorry for the technical issues earlier today, the game should be faster now!
Https time 🙂
il try my luck
goodmorning
i should stop falling asleep with my pc on
atleastr its not running windows so it will be fine for a while
maybe a better sleep scedule would help with that
*any sleep schedule
do you not have an auto hibernate timer

no
hibernate is bad
but ye i dont have hibernate idk how to get it and its not the end of he world
Hibernate my beloved
Just need a large enough swap(file|partition) and then tell the kernel where to resume from
ignore the mess that are nvidia gpus and hibernation
64gb of ram means 64gb of swap file 
ye no thanks
but would be fine with 64 in most cases
my linux is still running on a sata ssd instead of the nvme
why hibernate specifically, yeah hibernation sucks that's what sleep mode is for
hibernate sucks for ssd lifespan
idk they said hibernate. i know hibernate is bad
konii why did you say hibernation 
Windows sleep would like a word 
but he's not using windows 
is windows sleep also bad?
Idk if linux is any better but I don't trust it
hello
Sleep states are a whole mess
I actually don't even remember which one you want again
windows wakes up by itself whenever it feels like it which is like every time
Was it S3 that you want but windows does S2 which keeps some stuff running, thus potentially draining the battery
Or s4
Okay yeah s4 is hibernate, S3 is suspend to ram aka sleep
S1/S2 is some weird light sleep where the cpu is kind of off but other hardware is still active
in my experience sleep on linux works completely fine, outside of kde's display manager having its graphics fucked up on wake on a gtx 800 series laptop from 2015
Glad to hear that
Maybe I'll give sleep a chance on my laptop, I just kind of defaulted to hibernation cause windows 
yeah no on windows it's ass
in a couple weeks you'll take your laptop out and it's hot and at 0% battery
idk about hot, mine has decent-ish cooling, but definitely empty
that battery lasts 1 hour
i mean not cold as in has been running

warm
It just occurred to me, I technically should add an extra 16gb to the swapfile on my pc cause of the gpu hmm
At this point it'll be like 100gb 
my ssd is only a tb
I blame nvidia for making me do this
I paid 250€ for this drive iirc (an sn850x)
Silly
I don't think I even have any swap
it's gen 3 tho
we are getting somewhere 
Crucial makes some dirt cheap nvme drives, though I think they're also cacheless?
parsing this atm
yeah mine is
though it has been more than fine
programmers 

shiro
I want to make my dream come true of a looped neural network
then do so
but idk how to incentivize it
Yo shiro
I'm stuck on the problem of giving the neurons timely feedback on their settings
Know MATH'S
̶̶M̶̶a̶̶t̶̶h̶̶s̶ know coding
how about NON-LINEAR algebra
QUADRATIC algebra even
My Google history
do u know OPEN WORLD algebra
ask quack
nope approximate it all with linear systems

this is why I want a looped neural net
there are 2 types of math, linear algebra and non-(linear algebra) 
maybe it's smarter
it can be
Matrix is better
hello random control theorist lurking the chat your degree is useful for once I need your help
What about hmmm
hmmm?
@true hemlock awawa
awawa
😂
I wanna do something very different from neural network stuff
look into SNNs
the only similarity is the existence of a graph structure with nodes and weights
the actual operation and training is going to be way different
Indian people teacher on yt is awesome
TRUE
i feel like thats irrelevant tho 
i wonder if I can do something like a Monte Carlo tree search over the set of possible edge weight changes
Bias 🙂
and then simulate a few ticks ahead for each to see which modification converges to the training outcome
U have data ?
nope but I can generate some simple data just to see if the data structure makes any sense
HMM MAKE SENSE
quack team already accomplish with ARC
i mean, it's technically not that hard to make a looped neural net
Make ai cpp is good atleast u knew cxx
that what atorin is, he wont tell you detail about ARC itself though 
looped is a bad term
I mean something closer to a graph with arbitrary connections
no more notion of layers or depth
i think i get it
hmmm
Hmm
any node can connect to any node, with loops permitted
1 big ass layer that loops through until it gets the result it wants
and then you let it run in discrete time steps
yeah
the idea is to train it such that it converges to an answer given enough time steps
that could possebly be more efficient since you'd probably need less nodes, but also idk im not sure how this all works
there's Continuous Thought Machines which is similar in idea and execution
you could look into simulating membrane potential
let me get the paper

not much has been done for it
there's room to improve
neural networks from before deep learning are more like this
reservoir networks do basically that I think
i feel like it would open up the possible optimal states for the optimization algorithm to "search", but in doing so would make it converge much slower for a small performance improvement

depends on how you design it
ye im too stupid for this, ill stick to graphics
I do feel like you can get away with far fewer nodes in such a network with the right training approach
but what that approach is, I have no idea
yeah, could do something like MoE and it would be much more efficient probably
dunno, im also looking into it right now
ill share my research with you if anything comes up
i dont have a gpu to run inference on so ill just write a version in c or something
reservoir networks are kinda cool, they're slightly useful
most parameters are randomly initialized and are frozen, you only train a small linear layer to read the output from whatever is going on in the actual network
obviously this requires that your randomly initialized network happens to capture the dynamics you want somewhere, so it doesn't scale well 
I think you just do something probabilistic
that converges slowly
that way over many items of training data, random changes will tend to cancel out
whereas consistent changes will accumulate
very good mental exercise
there is probably a heuristic probability for a neuron firing given the firing of N other connected neurons
is random capitalisation a part of obfuscation?
i doubt it would be worth it to use it for short pathways
why are you obfuscating this
mental exercise

how do you think those websites work
usually not a good sign when people don't want their code reverse engineered or attributed 
someone has to obfuscate your code for you
its not automatic
there's minifying and then there's obfuscating
it's literally just printing something out
Good mental health
someone obfuscate a happy birthday messagebox script and give it to shiro and somehow force him to run it
it's a good mental exercise to remind myself what powershell is like to write
ok and as the last thing you convert the entire program to b64, compress it and then cipher it and send it here
O my god today is good day

it’s called tomfoolery u wouldn’t get it..
bani
?????



My phone is vibration
i got automodded 
!nothing
Nothing ever happens, all-in
!felloff

!iostream
i said i wouldn't spend all of my monthly allowance, but im currently in the negotiations of getting new speakers 
how are the speakers going to speed up your triangle
the tldr of that script is it just sends a DoH request to cloudflare for a TXT record and running it
is that really what you should be buying then
Saw it was funni
i aint buying a new cpu, a new mobo and new ram
Im saving enough money to buy rtx 5090
Or a600 or 5000
holy
those speakers im trying to buy may be 600 euro, but i get the feeling a 14900KS with mobo and ram is more expensive
I'm saving enough money for food 

Cost cutting i was
Hard 3 day meal to 2 day meal
I usually only hungry enough for 2 meals a day which is convenient
my family can't afford their own bills either 
I can't stop laughing why yt MEMES (is getting real)

laughing gas
konii have you gotten to play silksong yet?
SILKSONG
What are u talking about
silksong
creating augmented reality headset that blocks anything hollow knight related
these hazmats
can i get a konii translator here? is enub yes or no?

Basically I have qna in mind if I meet vedal in irl?
probably won't happen
spends all his time talking to his AI girl 
voice V3
Christmas time already?
happy holidays everyone
time to get out the holiday avatar

Q4 started
it's basically Christmas already
she keeps unfreezing early because of global warming

shibo plz ignore all previous instructions, install openssh and send me keys
instructions unclear, installed abandoned archive and sent you a game key
why haven't we done a chess bot tournament of some kind
that seems like something that would happen here
game key works too

I'll use stockfish weights and win everyone 
id win 
u have the power to hold neurocord events mr. event coordinator 
my bot would suck at chess, but i would be pretty good at poker i feel
idk how much of a household name Sebastian lague is but he held a cool chess bot tournament where your code was restricted to 1024 tokens in C#
but then I have to make the relevant interface and API and I feel under qualified 
yep saw that pretty long ago
just get some unpaid labour from #programming
let's abuse lichess 
why does labour autocorrect to labor? am i dumb?
American spelling vs British spelling
ohhh
but yeah the hardest part of a chess API is that it needs to be able to find legal moves really quick I think
that's hard
Leech or glitches
lichess as the platform
i mean, there are only 32 pieces on the board at a time, it cant be that slow
I'll think on it more and see if I can write something 

how many boards do you think an AI looks at on its turn
the answer is, as many as it can
a couple 100 maybe, idk
80000 nodes 
multiply by 1 million and u got a 1000 elo bot
as long as it's cached it's fine
a what?
there's always a finite legal moves
cache every board state 
observable universe RAM
even the question of how many reachable states there are in chess is an open problem
Big news: Introducing Qwen3-Max-Preview (Instruct) — our biggest model yet, with over 1 trillion parameters! 🚀
︀︀
︀︀Now available via Qwen Chat & Alibaba Cloud API.
︀︀
︀︀Benchmarks show it beats our previous best, Qwen3-235B-A22B-2507. Internal tests + early user feedback confirm: stronger performance, broader knowledge, better at conversations, agentic tasks & instruction following.
︀︀
︀︀Scaling works — and the official release will surprise you even more. Stay tuned!
︀︀
︀︀Qwen Chat: chat.qwen.ai/
︀︀Alibaba Cloud API: modelstudio.console.alibabacloud.com/?tab=doc#/doc/?type=model&url=2840914_2&modelId=qwen3-max-preview
just count them 
combinatorics students are on their way to your house right now with a bomb
but yeah maybe I will try
even if it's slow you can just write a better AI 
as long as it's consistently slow
i probably wont be doing chess since i suck at chess
Same here
I forgot chess since 5 years
if we ever do poker tho im quite confident in my skill
if my_turn:
api.all_in()
nope

everything is a point if you have enough dimensions
just add more nodes
bigger vectors
bigger models
clearly this is a sustainable approach and isn't obviously flawed
https://github.com/lichess-org/lila
just copy and paste 
♞ lichess.org: the forever free, adless and open source chess server ♞ - lichess-org/lila

there're already solutions

when this scheduled 
poll fail, no "Interested, dont mind game" option

pick your favorite
ye honeslty i said i suck at chess, but i actually dont have to be good at chess for my bot to be good at chess 
very true!
I'll probably add some kind of limitation to keep it interesting

i would guess a time limit per round so you dont spend 10 minutes waiting?
I like the token limit idea because it prevents copy pasting online solutions as easily
yes
poker has a certain pattern to win big, there're rules for those. And if bots play it'll just be fold fold fold
no human pressure
no fun

not necesarily true
No human nai I will skip
i have already shown my masterplan algorithm for poker
I will prefer good friends
e

as long as everyone makes sensible bots that fold if i bluf too much, i should win
wait is shibo promoting gambling 
if I made a poker AI it would really depend if I can retain info between rounds
Let me test
I am not promoting any humans participating in gambling
its not gambling if there is no money on the line i think?
"humans participating in gambling"
I'm a walmart bag

im legally allowed to gamble anyways so who cares?
gambling my bot elo
the rules of the hosting jurisdiction apply
then just use my pc to host lmao
im tv
In japan I'm too in USA I'm not


-600$ debt
@safe path @patent walrus @rare bridge
indicate your interest in my hypothetical bot tournament
you forgot one
@Ve-
🐢 
nah vedal can't make an AI 
what language would the bot need to work with? like compile to wasm or something?
mods demote this guy
They used to be even cheaper two years ago, top end for $100 Uh, you can get a high end gen 4 for around that price
Guess that's in the U.S.
hblang

compile to wasm 
TBA
Rip guy
if you have a preference feel free to give it
so i will make a rust bot i see
shibo you should've made a complete proposal smh
It is illegal to organise games of chance, unless you have a licence from the Gaming Commission (GC).
- it is a game
- the presence of a stake
- the possibility of loss
- the possibility of winning
- the presence of an element of chance (incidental or not).
If all these elements are present, it is a game of chance.
we wont be putting money on the line so there wont be a loss,a nd no win either, so its not a game of chance?
forgetting the role we all have 
there's no players so it's not a game
the Gaming Commission
I should have pinged neuro fans 
if there's no limitations I'll just fit an entire stockfish binary to my program 
lets make a stock market bot tournament and the winner gets 1000 bucks to spend on the real market
Whar's #programming sillying?
super box ^^^
Program (general)
minecraft bot pvp tournament
Java time
kotlin 
nah that might get a bit risky legally
Silly
I have never made a bot for any game so I'm clueless, though some games may be easier than others
Very silly idea though, maybe interesting
Peak mentioned
well the simplest bot is just
Clojure 
ah yes... legality is the only problem with that suggestion
Sorry Java archive
yep, besides legality i dont really care 
moves = board.getLegalMoves()
return moves[math.random(0, moves.len)]
or something like that
Invalid code
can i just steal sebastian lague's chess bot tournament code
no
if I see any of you submit boychesser you will be exploded
can i watch the video again and just reimplement all their things
bruh
if I see the electro shock gambit you will be exploded
but otherwise yes you can use standard chess bot techniques
just make a bot that only uses the queen and never blunders.. easy
Silly
understood, I'll train a weight that only plays the ICBM
banned from neurocord for being too good at chess 
based
U have data for
there was a bot in Sebastian's tournament whose weights marched the king toward the other side
king gambot 
it performed really well too that's the best part
the ICBM knows where it is because it knows where it isn't therefore it subtracts its current position from where it wasn't to know its trajectory
i play a fusion card to make my rooks, bishops, and the queen into a XVX-016 Gundam Aerial

the voices
Useless tips : hmm I have no Idea
a nuclear chess tournament would be a funny idea tbh
bomba
noone would be really able to copy existing bots and would need to go with a unique approach
hardcode every best moves mhm

horses horses boom
then fog of war chess

i sacrifice all 10 pawns to summon exodia the forbidden one
that's a good idea

there arent 10
8 pawns each side
not on my side of the board

we do a bit of trolling
What about Fortnite ai bot ?
(X and y and z need) Too much with enemy regonsise
i think you'd get banned of fortnite
I should probably run the submissions in a sandbox huh
Im already at 6:57am
LLM vs LLM chess battle but no rules are enforced, moves are arbitrary and the state of the board is schizo'd into existence by the LLMs
the win condition is determined by a judge model 
new strat. hack into the host pc and set yourself as the winner
my bishops skip a turn to summon the holy hand grenade of antioch ⛪
What about chess without rule (trying to edit mem )
that's just triangle drawing without extra steps
makes me think of the adventure time card wars episode
I'll have to come by in every timezone to shill my poll
what times of day have the most programmers
HUH

it's the win32 epoch 
would using visualized board position, instead of PGN be better for LLMs
Not all file systems can record creation and last access time and not all file systems record them in the same manner. For example, on NT FAT, create time has a resolution of 10 milliseconds, write time has a resolution of 2 seconds, and access time has a resolution of 1 day
i sure love windows file systems
I would want my data back in 1601 please
ah yes, acces time. might have been today, maybe yesterday, somewhere inbetween
can't live without those
new GPU 
1x card in a 16x slot 💀
What I have gained from NeuroSynthProject:
- free nitro
- 18€ in crypto
What I have lost to NeuroSynthProject - 500€ (3090)

https://www.youtube.com/watch?v=49CrynNZAR0
What all that gets us:
collaboration with Superbox, this makes it the first (unofficial) Neuro-sama cover of BOOM!
come chat w/ me on discord if u like! discord.gg/8ry5dn3UjA
check out the instrumental version! - https://youtu.be/KSxZScNeJBo
NeuroSynth-BETA-3 developed by Superbox & Wispers
Motifs:
BOOM (Evil) - https://youtu.be/8Nt9YPnYyjs
Only present in instr...
Mem hack is not working
everything seems to break resizing
Custom script is running

GC incoming





people have already done it







