#programming
1 messages · Page 183 of 1

music related stuff seems to work
its not specifically that it needs to be an artist channel
yeah
cuz otherwise neuro would qualify
and i tried it
and it didnt work
you cant go from short to music, you need to go from short to youtube notation to yt music
https://music.youtube.com/watch?v=ENhfIeZF_AY podcast-y stuff also seems to work
Can confirm the not a music channel, as I grabbed a skit from someone who does music and only their music worked.
i think it's ai allowing content that's music related, my youtube music recommended is literally all production content
not music

downloaded the UCI spec
filename has a (1) after it
wtf is this notepad formatting
Didn't get tricked with a music based meme so either it is good or we need another theory.
i have a music based meme in recommended but can't share because rule 3
That checks out
I have one thing I can test
On the music side, I have this https://music.youtube.com/watch?v=jXFsBapXZ0g
chat what do you think of this article as a regular end-user
I wanna see what most normal users think
nyaaaaaaa
first of all i think their formatting looks a bit odd
havign the 3rd author on a newline for no reason bothers me
that aside 
the image is higher than the title
I'm talking about the content
I'm not affiliated with this company
I'm more so asking what do you think about the data presented here assuming your company installed the EDR on your endpoint at work
host your own instance, this was always a risk with third-party appliances
we're talking strictly work devices here and not your personal one
how many data spills have happened now that are attributable to unwanted third-party data access
how many more need to happen for people to learn this lesson
on-prem access is not necessarily always available for all EDR providers but assume it is in this case
this is a lot of reading
gimme amin
i like how to article doesn't shy away from mentioning competing products
like malwarebytes and bitdefender
this is the best source i can find for what i need https://youtu.be/D4DBE4UQoAY
Learn how to easily program the UCI protocol. Learn how to set up the basics of the industry standard protocol that communicates with Arena, WinBoard, XBoard, or any other UCI-enabled GUI. Its another step in creating a professional chess engine.
Official UCI Protocol: http://wbec-ridderkerk.nl/html/UCIProtocol.html
Code: https://onedrive.liv...
this microphone

the start of the sebastion lague video also shows some chess logic
its nothing bot related im looking for right now
i need to make boilerplate for the engine to communicate to the chess server
and i want to make a wrapper for it that people can use
but theres nothing other than this cursed spec to go off
what chess server u using?
arena 
seemed a good option and i got it working
it has the features im looking for
now i can load up any bot that can communicate with UCI
so i just need to make a wrapper around UCI that abstracts the protocol away a bit and then we're good to go
mandatory data collection: programming proficiency
2
8
I'm surprised I'm the only one who voted for that one
honesltly i cant be bothered to read the whole thing. the google translate stuff and browser activity is wild tho
I finished reading and it had conflicting tones between a diary and a research paper.
Made it hard to focus on the actual content
it is weirdly unprofessional from my point of view
it felt like they couldn't decide if this was a personal blog or a corporate one
oh that was supposed to be corporate?
it felt like a guy making an article while trying to be professional
or well, 3 guys i guess, 3 authers
phrrr
phrrr? 
ive learned an unfortunate quirk about UCI
single threading
im thinking if i do have to multi-thread it
i guess async is possible too?
it may make more sense for the entry point to be a wrapper i write
and then have the wrapper call the chess bot
and supply an interface to it
then the wrapper can run it in a child thread
ill prob still require chess bots be single-threaded, but it means i need to make sure the environment is ok with multi-threading
i dont always fail
Discord bot failed
Pacman -Syu aways failed
My network FAILED has ####
python users, are there particular versions that you prefer using, or is the latest stable release always fine
3.12 because some people don't update deps
My personal projects are on 3.13 as I can handle that myself
yeah I'm also currently using 3.12
3.12 
ive had a time the min version for a rep was 3.13 and the max for another was 3.12 
i ended up manually modifying the one that needed 3.13 to work correctly on 3.12
it wasnt a full incompatebility, there was just a bug that they fixed in the 3.13 version and i needed it gone in 3.12
I know exactly what project this is
That turn of events caused me to end up joining you

im trying to type attrib but it constantly changes to VkPhysicalDeviceVertexAttributeDivisorFeatures 
This is one of the reasons I don't have any autocomplete
fair
It legit has a 100% chance of getting in my way instead of being useful
Wiping system because I muderized nixpkgs by letting the battery die while updating. Let's hope the config I switches to actually functions.
Windows allows itself to be installed on 16GB
ideally i only need to remake the system drive
I should post the image I have so someone can try and do a ramfs large enough to hold it.
surely 64gb is enough. altho i have no clue how to do an os install on ram only
I'd want to do some more disk cloning as the used storage is a lot smaller than what windows allowed itself to be installed on.
time to install labubu again
I'd think VM with the image in RAM
After I find where this win11 image I'll let you know after I remove some apps.
dont bother, i myself wont be trying it out
I needed to do it anyways
I know the FS that is one is getting wiped so I need to move it off
turn all of ram into a ramdisk for the OS and then use a pagefile for all your ram
what is an ssd but jsut more ram?
well wait
if your os is on a ramdisk
then the pagefile is on ram too
so its probably wicked fast

me wehn i put my puter in sleepmode and everything gets moved from ram to the pagefile, that is also in ram
so when i turn it back on the os wil be gone
That fits on a ramdisk 💯
I so badly want to shrink that C drive some more
7GB C: would be funny as frick
(this has edge removed and firefox installed, it was never opened as I commandline installed firefox)
im trying to do dynamic objectlaoding now in vulkan instead of only 1 tutel, and the code is not pretty.
its not done yet but give it some time
Is windows installing updates to the D drive? I have no files here
i guess look for hidden files?
True
nope still nothing
well then i dont know
windows just kinda does random bs under the hood and i really admire that
it's the pinnacle of spaghetti
classic automod

This isn't tiny11 I debloated this myself


guess we're really installing everything huh
forget gcc my c is gg
too late
garuda arch has a "nix subsystem" that i havent bothered to try
tfw nix got wsl'd from linux
Any time I do nixos in a vm it is slow as balls with the UI
why is it so slow
god only knows
is that a repo issue
probably
577mb shoudl be downloaded in less than a minute
KB 
for some reason when i get an update to one of my packages i think it's like a google font package or something, it downloads SOOOOO slowly but only that one
mirror held together by gorilla glue]
]
"nix subsystem", probably just installing nix on arch as well as configuring for opengl/vulkan
probably
i am too afraid of it ruining everything to try it
could do a snap but eh
well, it'd auto do a snap
I regretted installing the google-fonts package on my system
but that's so much work redoing id have to do literally 2 things
It was fast, but also it was heavy
funyun
what's poppin
have you heard what im putting together
i have not unless it's the chess thing
it is
i take it you're not interested
ive been making a stupid man page browser thing so i can load them in my browser instead of terminal 
no i mean it sounds cool i just know almost nothing about it
other than that it requires c bindings
lmfao
i figured putting myself in a position where that knowledge is load-bearing is a great way to learn
now i get to pretend to be an expert on something i know nothing about

and everyone else gets to learn too 
Besides the others plotting to get Elyvn to be a viable candidate I got no idea.
what is the concept of this chess thing
load bearing bones are the bones you need to not become jelly
concept:
chess
participants each make a chess bot
i will take them
chess bots fight

ye if elvyn want to participate she have to write her own bot
some limitations added to keep things interesting
probably with quack
and these bots i imagine aren't meant to be hooked up to stockfish
no network connections, and max token limit on the source
so stockfish should be off the table
the bar is not to make an engine that can beat stockfish anyway
imma fine tune the shittest 1B llm on only chess moves
you only have to beat the other submissions
If I join I'll have to probably break a couple rules due to not having direct conversions of some metrics
can it run like https://github.com/mlc-ai/web-llm with an llm as its "chess engine" (bad)
(still waiting btw)
im trying to make the old vulkan system work with the new model so i have to pass bufferManager.testmodel.primitiveDataList[0].indices.size() to my framemanager 
you probably can't squeeze an LLM in
the token limit is intended to be quite low
probably below 10k
if it loads the weights from a separate file does the .safetensors or whatever it would be count towards the length limit 
If someone does a text input of a 8x8 grid with 7 states, then you got 22 turns
submissions must be one source file 
i like how the latest man2html uses http://www.oac.uci.edu/indiv/ehood/man2html.html every time it renders a manpage that it links like this at the bottom of every one
and that link is so ancient (pre-2003)
This is assuming ASCII chess with a delim
damn, well it sounds fun so i'd need to see the interface and do some 
i haven't ever considered how i'd do anything liek a chess engine before so it'd be something way new
i dont think anyone here has any experience with this, so doing even a little better than random moves will make you a strong contender
ill have a proper github page for all this ready when i have it all put together
-# ignoring the fact you can just use play by play to completely break the limit.
i have no life so ill probably see it but ping me when you have the specs, in the meantime i'll drool on my keyboard and thonk
came to find out thru rendering this page with every manpage that's able to be found in any directory on my root drive
i have over 25,000 man pages on system apparently?
that's a lotta manual
it just finished
and now a 42MB package 

my system has been up for almost 3 days now because this php import is STILL going
it has to be so close
but i just ran my first update
70 packages w/ updates
honestly not as bad as i expected
If an LLM is stateless already and caching is useless at this scale, why not just do a play by play and keep resetting the context window.
yeah the llm would be tiny anyways so likely unable to handle more than a couple moves regardless of whether or not you give it them in the context
it'd have to work off the current board state only i think
maybe a move or 2 back
because it would be so dumb
With 10k tokens, you got more than enough
even with 1k you have 15 past moves. (I misdid the other one)
And that is assuming char level tokenizing
8x8 board + delim
when we're talking tokens here i'm trying to comprehend if you're talking like, 10k tokens as in the length of the source file
or
10k tokens for teh chess game
in text for moves
Fair question
10k token source file
as in, after the lexer turns your source code into tokens, there must be at most 10k tokens
thats what i thought but iggly made me doubt
though i will probably aim toward the lower end of that bound
1k might be a more reasonable working amount
but i may bump that up a little
still mulling it over
so like a keyword is a token, a literal is a token etc after that yeah
in general
10k is a long boi
int myvar = 1;
==>
["int", "myvar", "=", "1", ";"]
ye that's what i expected ok cool
Python has the advantage due to no semicolons or brackets, noted
im getting an error
If the parameters define an export operation and the handle type is not VK_EXTERNAL_MEMORY_HANDLE_TYPE_ANDROID_HARDWARE_BUFFER_BIT_ANDROID , allocationSize must be greater than 0
but now im wonderimg why android gets a free pass on allocating 0 memory 
keyword/identifier/operator/literal/separator and/or punctuation all 1 token each
coo coo coo
but it's slower than C as well, so it's up to you to decide if the tradeoff is worthwhile
the answer is probably going to end up being "use the language you know best"
in c++ the brackets are more like a suggestion these days i think
but ye there is more syntax spaghetti
The only one source file is going to be tough, Surely I can port to not numpy.
would a single scratch block count as a token?
i was gonna ask that lmfao
but then i figured it probably is like
60289753329184798 somehow
scratch block still need some typed in numbers here and there so i guess not
well they have to be converted into whatever the intermediate is for scratch before compiler
and that's where the token count would happen i imagine
one for block and one for each value, this is balance because scratch doesn't implement enough basic math
compile to JS IIRC
oh
that makes sense and is boring
i have crafted my magnum opus
svg vitruvian man with weird proportions as an icon
If I could manage to actually finish my lib, I could see myself going for the train on stockfish approach.
honestly?
if you somehow
fit a scratch vm into the token limit
i would be super impressed
regardless of actual chess performance
load scratch vm
As I could probably get it into the 10k token requirement with ease.
instantly just surrender
ggs wp
if you think about it, you can go through the entire tournament without losing a single piece with this strategy
?
there's actually no shot an llm is fitting in this 10k limit at least not using something like llamafile
1b llama's llama file is like 250 million tokens since it's a gigabyte
well i mean the actual cpp part of it wouldnt be that long
but like
you have to fit the llm in the single file
which is like a binary
?
im all e-

how is that multi language
thats single language
buildscript++
i cant wait to get insightful submissions like apt-get install stockfish; stockfish and sudo rm -rf
lmfao
I was going with pasting strings into compilation software but that is more funny
i will submit a single file and it'll be dont_look_at_my_size.cpp with a massive 1gb llm's binary embedded in the file
and since the binary isn't technically tokens it doesnt count :^)
not as far as a lexer is concerned
submission name: zipbomb.tar.gz
submission comment: my chess bot is inside its just zipped to keep it small
actually is a .rar with the recent path traversal cve
and thats why its all being run in a vm
"hey guys that vm was set to autoscale why is the instance costing $20/hr?"
Another approach to building and having source code in one place is that the bash string is only one token.

well one day i'll be able to say this import finished and then ill be able to do anything else that might require me to ever restart my pc, update anything, or use all the resources i have
but that's not quite yet apparently
it's finally larger than the .xml export itself was
idk how much "compression" the xml dump achieves but it cant be THAT much
almost 3 days 
ram use continuing to increase like a boss
surely it'll be done soon
i just want to reboot
this is my first time building python from source
i gave it two threads and its in a vm
should i let it run overnight or will it be done within an hour
Does the python interpreter count for the token limit for chess source files?
i would expect CPython to build in like 15 mins
but idk about how long with 2 threads
i have the vertex and indexbuffers working on the new system 
by "tokens", i mean the tokenized source code that you have submitted
now i need to get the textureloading on there too
damn
you will be given an interface of some kind that lets you interact with the chess board, and then your bot just uses that to decide and make moves
when u hit a lick
that part is what you write, and only that part counts toward your token limit
the language's standard library and the chess board interface are free
if my language has an interpreter to run does that count
you kinda implied that with scratch vm thing
i said the scratch vm thing because i dont intend to add support for scratch
That was my reason for asking yep
i'll try to support a few common languages at least, but i have to implement the chess api myself for each language i plan to support
point being i wont support all of them
i welcome requests for language support if it seems easy enough, atm i plan to at least support c and python
I'm sure the scratch folks will build out their own API

python can just use the c api probably, but stuff like rust will be more difficult ye
doesnt rust have a C api
yeah the C api will be public, so if you can create bindings from your language to that, i'm happy to make those bindings official and remove them from having to be included in the submission
Turbowarp and some other forks have web requests IIRC
sorry python users
i built python before installing zlib or lzma
no compression in stdlib for you 
is it already done?
i just started building it because i was curious lol
to see how long it would run for
ez
that was quick
im honestly surprised it worked
i had no idea what i was doing
sweet
not having zlib and lzma might be like
unusable
try using pip
that would kill gzip and zipfile without deflate, no
i am, i need it to install my cryptominers
nope
dammit
also rule 2 is no network connection so unless you're donating that crypto
as long as it never needs to read a png or use any sort of web traffic that gets compressed for the interaction with the game ig
no tar.gz or anything either which makes obfuscating infostealers harder this is impossible
...though.
i dont know if python pickles anything when interacting with C
does the pickle protocol have compression
its weird as hell
"linux tip no.1 - remove french language pack" sudo rm -fr
as something of a .pt myself
what is so bad about pip
that wasnt the question
does it compress objects before (or after) it serializes them
you can compress it after which is very common
no it just takes the object and serializes the raw object
ok thatrs fine then
Comp rules, std only and no network
Alright, time to build Hy-
We love relying on the filesystem when it changes for paths
Once again, time to build Hyprland
After I install 13GB of other files
I'm sure I'll be installing more as my system and home packages are split
well i didn't disable tests in this python build so it's over
the tests are taking longer than the actual build did
I see that building xwayland
at least they seem to disable the tests that would take time by default but doesnt that defeat the purpose of running iwth the whole test suite which im pretty sure i did lol
we love making u still use xorg without saying anything
I probably have apps that need it
i have it, idk what runs it and dont care enough to find out as long as stuff works
i like the name of the extensions
Also man every time I look at the gnome installer I am reminded on why I didn't choose it
DAMAGE

ive tried to gnome a few times
i just cant do it
i end up installing 4189374519847129837198 extensions to make it work like kde or something
YOUR-MOM.DOUBLE-BUFFER
which at that point might as well just use kde or anything else
she's a BIG-REQUEST alright
yo momma so fat i have to call BIG-REQUESTS to have a chance of her loading
emotional DAMAGE or something
that's a big cursor
cursor caught in 4K
it says the entire screen #0 is the same dimensions
so in case i ever wanted my cursor to fill the entire screen ig
idk what that is trying to refer to necessarily
Your mom has to be double buffered,
a composite if you will
she is a big request-
that's a lot of damage!
now i want to see what other xwayland/xorg extensions are called
ok python how long u need to test sockets
Alrighty, hyprland actually complied this time
So that is an improvement
now to get the rest working
yeah i mean that's the best running hyprland ever if it managed to compile
even if u havent ran it yet
as an nvidia boi ive not had hyprland cooperate very well for more than a short amount of time
I don't have the knowledge to fix whatever the build process on nix is smoking
I'm just glad it works
sometimes
it's smoking that pack (nixpkgs)
python at the end of the build said it can't compile ssl because i dont have at least openSSL 1.1.1 installed
scared me a bit
k python whatever brah
so i guess even if there was zlib and lzma installed there's no guarantee their crappy build script will use it
I should have transfered some of the botch fixes I did
Need to remount a few of these file systems
i finally broke down and put a repo up with a shell script that remounts my drives and adds it to fstab so i dont have to do it every time i reinstall which can be frequent if im feeling extra stupid
who needs nixos when you can have a fragile bash script that makes a bunch of assumptions
not me
have to come up with all the functions I want in the chess API
and then figure out how that interface is gonna work
i wrote a whole example of using pickle with compression on a base64 encoded server uri, as a joke, and then didnt post it because it seemed too close to something actual malware might use
and it's all your fault
honestly the setup for this tournament is going quicker than I thought
it doesn't actually seem too hard
concept is easy enough
i guess you only really need a getBoardState() and move(endposition)
well I'm gonna provide a little more than that
getLegalMoves() for example
there's lots of boiler plate like that where it's basically always done the same way and is essential to a bot
so I don't plan to make everyone implement all that as part of their token limit
token savings
Would a getPastMoves() exist?

I think it could exist, given how board objects will work
you'll be able to make and undo moves on your local copy of the board
getPastMoves() but for the entire tournament so i can quickly train a small nn from scratch on all the moves previous and use that each turn
so I think they'll inherently have a move history
theorethically, would we be able to see the sourcecode for stuff like getLegalMoves() or do we just go in blind?
I'd hope that it is in chess notation
sure, I'll happily make that part of the backend public
it would save tokencount but if there is a timelimit doing it yourself might be faster
idk yet tho
there will be some documentation for it too, to make it easier for people to write similar interfaces for their languages of choice
use 9000 of your 10000 tokens rewriting all of the api functions in rust because you literally cannot fathom not rewriting everything in rust
I've thought more about how it'll work tho, and it may be more work to write a custom language implementation than just wrapping the c implementation with some bindings
btw how do you see how many tokens you're at?
good question
I haven't figured that part out yet but I'm sure there's ways
find a lexer for your language
i figured this was the plan
every lang has c interaction of some kind anyways
just write a C api and every language can write for it
it's awkward because I think for C, it would be easier for the interface to just import your code and call some run function in it
it's very awkward where the interface sits
there's kinda two parts to it
there's the part that provides definition and convenience functions for working with boards and moves
and there's the part that communicates with the chess server via UCI and mediates those commands with the chess bot
If I were to do this, I'd probably be building out the full stack myself. Just because I'd need to for training.
the latter needs its own thread which complicated things
i tried to google it but i didnt really find much
ill leave the tokenizing to someone else
I'll provide some info on how I'll check your token count per-language
I don't expect people to submit blindly
will be in the detailed info
logos
case 'a':
case 'b':
case 'c':
case 'd':
case 'e':
case 'f':
case 'g':
case 'h':
case 'i':
case 'j':
case 'k':
case 'l':
case 'm':
case 'n':
case 'o':
case 'p':
case 'q':
case 'r':
case 's':
case 't':
case 'u':
case 'v':
case 'w':
case 'x':
so each charachter of a string or a number is a seperate token?
the number 800 is 3 tokens?
no the int 800 would be a token
a lexer doing character level tokenization would be goofy for a lexer in a compiler since the point is that the parser doesnt have to deal with every character in the source
it's like the prep chef that cuts all the shit up so the actual chef doesnt have to start from the raw ingredients because that would take forever
well then the lexer code i found is shit
it's meant to abstract things a bit not make them worse
unless its only checks the firrst digit of the token
that thing you posted i could only imagine is some sort of extremely verbose and poorly conceived check to see if a character is a letter so they can further see if it's like a variable name or something idk
are you sure that's not yandev
yandev is the master of the ultra switch statement
no, the problem with yandev iirc was a lack of switch statements
that must be recent code
idk what i just posted is literally from yandere simulator from 5 years ago
also wtf are these attributes that the studentWitnessType has??????????????
human readable 
the updateLabel seems to constantly use the same numbers so that should probably just get built in as standart setting
i didnt actually care about it when it was a big thing and you are right the switches were a replacement for the infinitely nested else if statements i guess 
wew
yep
from what i heard a publisher wanted to fix the game up but yandev wouldnt let them fix his code
too much pride i guess
the first thing i wrote myself was a python blackjack that did something of the sort just using only if/else cases
in like 2010

i have nightmares about it
bro was playing python balckjack while i was 3
it's actually the reason i stopped programming entirely for years
and only restarted in the last few years
i was like how can people do thjis
this is blackjack and it was getting too hard to mentally conceive of what was happening in all those cases
i would've been like 13-14 whenever this occurred so i use the excuse that i was literally a baby toddler
I locked myself out of the stable version, yikes
Back to iso I go

blessed file length
Okay, at least nofail is doing me favors
Back to iso, I need to regerate this. I forgor to mount the home partion
you dont need that it's fine
Well none of my user partions are loading, so preferably I'd be fixing that
At least it isn't wiped
I'm lazy so I just use disko to mount everything in one go when I need it
I found setting up disko to be a pain
looks like a fancy fstab
Well yeah, make it a normal function lol
How do you pass arguments for something that runs automatically
why woud you use parameters for a destroyer
it only knows how to use lazers and rockets
but it doesnt need u for that
you make a lot of sense, but vulkan needs me to pass the device to free the memory
I'll just not mount those two, as they really should not be btfrs because I didn't bother to subvolume them. (I was very new to it when I made them)
hardcode the device 
no
i do find it funny how im basicy passing VkDevice device to almost every single function in this engine
the vulakn tutorial has everything in a single class so not a problem for them since its a class variable, but i dont
Invest
including the 2 volk files i have 19 files of code
wdym?
I'd just make it a class var at this point, it seems like it belongs there
the primitiveData needs a destructor defined the class then when you call destroyAll on it it should already have the device reference handled because it's in the class
i cant just make it a class variable of every single class tho? or should i?
PrimitiveData looks liek this btw
struct PrimitiveData
{
std::vector<Vertex> vertices;
VkBuffer vertexBuffer;
VkDeviceMemory vertexBufferMemory;
std::vector<uint32_t> indices;
VkBuffer indexBuffer;
VkDeviceMemory indexBufferMemory;
VkImage textureImage;
VkDeviceMemory textureImageMemory;
VkImageView textureImageView;
VkSampler textureSampler;
};
struct AttribDatta
{
const unsigned char* dataPtr;
int stride;
};
maybe i'm not gonna backread that
its mostly chess stuff and whatever else happened
ya so you just move the destructor into that under
VkSampler textureSampler;
and put all the cleanups there in the class

i did anyway wasnt even hard
struct PrimitiveData
{
...
VkImageView textureImageView;
VkSampler textureSampler;
void destroyTexture(VkDevice device)
{
vkDestroySampler(device, textureSampler, nullptr);
vkDestroyImageView(device, textureImageView, nullptr);
vkDestroyImage(device, textureImage, nullptr);
vkFreeMemory(device, textureImageMemory, nullptr);
}
void destroyVertexBuffers(VkDevice device)
{
vkDestroyBuffer(device, indexBuffer, nullptr);
vkFreeMemory(device, indexBufferMemory, nullptr);
vkDestroyBuffer(device, vertexBuffer, nullptr);
vkFreeMemory(device, vertexBufferMemory, nullptr);
}
};
like this?
it was a whole lotta speculation of how best to destroy shiro's computer
i fucking hate formatting a block of shit in discord
option 1: make texture/vbuf object contain device pointer
option 2: just dont use raii
option 3: make a custom smart pointer that wraps a texture/vbuf with the owning device (similar to 1 but different indirection wise)
option 4: singleton 
struct PrimitiveData
{
...
~PrimitiveData() {
vkDestroySampler(device, textureSampler, nullptr);
vkDestroyImageView(device, textureImageView, nullptr);
vkDestroyImage(device, textureImage, nullptr);
vkFreeMemory(device, textureImageMemory, nullptr);
vkDestroyBuffer(device, indexBuffer, nullptr);
vkFreeMemory(device, indexBufferMemory, nullptr);
vkDestroyBuffer(device, vertexBuffer, nullptr);
vkFreeMemory(device, vertexBufferMemory, nullptr);
}
};
whatever this comes out looking like i copied the classes from ur codeblock and it hated me
it's the same as any destructor using the ~className(); syntax in the definition https://www.geeksforgeeks.org/cpp/destructors-c/
then it's just tied to the object lifetime
and blows it up when the object is out of scope
automatically
i already dont do raii mostly, i thought it might be handy for game objects tho
do what i do and format it like total ass, let markdownlint underline the entire document
and then slightly add some parts of the md to it so it knows what you're going for
autoformat
its too late its basically done
which is terrible
because it means i have to do the coding part now

is so cute lol

have you considered making it shorter 
lett gemini write it and just edit what needs to be changed
nah actually, make gemini edit it

minmaxing
that's how you get a shitload of emojis added to your md
"remove the emoji's" boom fixed
every time ive given up on life and had an llm write a readme or something i feel like i've spent more time un-slopifying it than i would've just writing it myself
can't stand how they write them
jsut dont unslopify it then
if the README of all things is turboslop then i dont know how anyone could trust the content of the project
"write it like a 15th century esteemed noble" 
you might be on to something
refactor this codebase like a 15th century esteemed noble
laughing like a 15th century esteemed noble


shiro how late is it for you rn?
made the github, made the readme, set up the vm, installed arena, built python, installed c build tools
2am

chess engine like a 15th century esteemed noble
worklife balance is not balanced, its almost vertical
this is how i am with projects
i like how they just hadnt invented perspective techniques yet back in the day so they just didnt give a shit and totally warped objects to fully display them
i will invest every waking hour for a week or two straight
and then never touch it again
im here working on my vulkan at 8am so icant say shit, but still
im going after my new speakers in half an hour 
shiro do you accept submissions by mail 
tomorrow: the chess api 
instead of lord farquaad it's just me inside the chess trojan horse
you may struggle with the time limit
i'm not going to your house
i am
i'm not going to your house
Won't fix
the more I think about the details the more bleak the future becomes
how do I even multithread in C...
pthreads
we'll get there when we get there (never)
or on windows uh CriticalSection and CreateThread
multithread in a different language instead and expose C bindings from it...
ah yess
yes yes multithread in python now that the GIL is gone
calling my Python bindings from C
pthreads isnt really hard
will I do it correctly first try
the only problem with c multithreading is it being platform specific
do you trust yourself to do threading correctly
in any language
no

ye
I still can't figure out if the chess bot should start the API thread or vice versa
I guess the bot probably should
something like
pthreads is pretty comfy you got this 
how long would these api calls need to run it cant be that long can it
if theyre super fast could get away with it from the main thread and pretend nothing else exists
:^)
surely nothing woul ever block it
#include "chess.h"
public void main(int argc, char *argv) {
ChessAPI *api = start_chess_api();
int lenLegalMoves;
Move *legalMoves = api->getLegalMoves(&lenLegalMoves);
api->push(legalMoves[lenLegalMoves-1]);
}
idk something like that
don't judge I'm on mobile
you get the idea
hey can someone help?
depends on what
the UCI spec requires that the chess bot
always be listening for UCI commands from the server
this has caused an epiphany, c was built for chess, the right hand side of the board is called the h file
chess.h is real 

but yeah does that look like a useable API?
does that seem like a reasonable way of exposing its functions
someone told me to test out a bot on the website "JanitorAI". i normally use tavern so i don't use the online one much. i would like to know what good llm models are out there, since jllm is trash at handling role plays. any suggestions?
ah, i see.
i dont think public void is a thing in c
sorry it's been ages since I've written C
might be some java bleed over
the vertexbuffer and indexbuffer are now stored inside the objectclass instead of only being created there and sent outside
int main(int argc, char *argv[])
Cava 
wake up babe new hybrid frankenlanguage just dropped
should probably also be Move** for the array
any suggestions on how to garbage collect any of this

i relaize now that eventho a lot changed under the hood, this does not look different at all, and i shouldn't really send pictures
well the legalMoves array needs to be freed
shiro!
hello!
that should be the responsibility of whoever is using it tbh, not of the API itself
well all i need to do know is make the textures object class variables instead of having one global texture. that is for later tho since i need to get ready to go after my new speakers
true but my concern is
i dunno if it memoryleaks if some moron like me forgets to free it properly that's a bit rough
well if it leaks memory then your bot will run out of memory and crash
I guess in practice it's not important, the bot just goes until it runs out of time
why even implement the whole spec I'm sure I can half ass it
but yeah the thing I'm wondering about right now is, if the chess server decides the chess bot has gone over time, how does it terminate the bot
16 files changed, 1092 insertions(+), 488 deletions(-) 
send SIGKILL
kill the process?
I didn't make the chess server
it might ask it politely to die via UCI
which would complicate things
exit() if you get politely asked to die?
then it's the caller's responsibility to check if it should die
which seems like a weird requirement to put on participants
i mean the bot should be able to figure out if the game ends
who would spend the tokens on that
and die
well no, since you're the one writing the UCI part
it should yeah but the API will have to handle that
yeah the UCI handler should do it I agree
or I completely misunderstood the architecture
"hewwo plz destroy urself thank"
How are we even communicating? Web requests or some other thing
as a participant, you just get a nice API that lets you do calls like getLegalMoves and push to read the board state and submit moves
but to make that work I have to puzzle out an interface between that nice system and the universal chess interface protocol
that's the part I'm dying on rn
i dont think it needs to be overthought it just needs to send a message that the bot should expect and then it should use kill() or equivalent to end itself
not like it'd burn all the tokens in the world for that
If the neuro API does it, then we should as well
Well that one is more like prepare to die then they close it

Apoptosis (Programmed Cell Death) of T685A human melanoma cancer cell.
The movie was taken using Nanolive's technology: the 3D Cell Explorer. A label-free holo-tomographic microscope able to reconstruct living cells in 3D based on their optical density: www.nanolive.ch
Movie description:
During real-time monitoring, we can observe the typical...
my chess bot
it'd be a cool name for a bot honestly
https://www.youtube.com/shorts/3SCkjbnSpgY why is this the soundtrack chosen lmfao
😭
Im on the road to my speakers 
soon you will be crawling around hooking them up and malding
also this 3 day uptime while running this fucking php import script has also lead to my plasmashell running this whole time essentially and i think that has revealed some sort of memory leak so that's great
ram isn't being used too much but plasmashell is currently using 31 gb of swap LMAO
wtf
31GB 
plasmashell 31493180 kB
is what i get when looking at the swap rn

restarted it and freed all 31gb instantly nice but wtf was that
better ignore it forever
Shiro is written in c good
Good morning humans
i think the culprit is unironically the giant folder of wallpapers i have that just cycles in a slideshow every 5 mins
somehow not being freed
and i dont usually have the thing running for 3 days

i wanted to have anime grills on my wallpaper
but i didnt want it to be all the time
so i downloaded like an entire wallpaper website
slapped it in a folder
and that just cycles
Weird how it isn't only preloading the next few
well, i'm not positive that's what it is
but i was searching around and found some people complaining about plasmashell using a lot of memory
and multiple times ive seen it asked "well do you have a slideshow wallpaper" or similar
idk why it was just going into swap (so it's not actively in use) instead of just being freed
didnt even look at it before because i just figured it was some bullshit being caused by this php script
I've only had one anime gorl wallpaper since forever now. I should spice it up when I get the chance.
but then when i saw swap at 30+gb i was like surely not
how would it use 30 gigs swap when the db is only
sudo du -sh /var/lib/mysql/archwiki
14G /var/lib/mysql/archwiki
but also how do the wallpapers do it when the folder is smaller than that
did you get your usr folder back
Nope, no luck yet
javascript good? npm good? in this video we see why maybe yes sometimes no
https://medium.com/@_jaydeepkarale/largest-npm-hack-in-history-f953acf82b76
https://github.com/Qix- (go show his repos some love)
https://gist.githubusercontent.com/sindresorhus/2b7466b1ec36376b8742dc711c24db20/raw/1eb4379d73f00a967bfa7bc1edc0afbd3c6ff494/chalk-source-in...
that's crazy
they stole literally like 5 dollars
they put the shittest payload all time in there
mfw pull off a million billion zillion gillion user supply chain attack and forgor the payload that does literally anything
I have them 
Ignore my shadow lmao
They're a lot bigger than i was expecting
They came out of an actual cinema apparently

they are smart to be able to pwn an npm maintainer
but not smart enough to be able to make it more hiden and useful
That place had a lot of cool old tech
that is cool as hell
wallpaper engine is so good for wallpapers, cant recommend it enough, probably the best 5€ i have ever spent
No wallpaper engine on linux 
Bwaa
I still need to change my wallpaper in linux
Yo what are u trying to do swap memory
I heard of a project for it, can't say how well it works
Daily?
i saw an option to have an animated wallpaper at some point
Not daily buf i have customized it weelky or so
we have wallpaper engine at home (running mpv in the background)
sam it's very important for you confirm if i have identified the MPV in the pic
i need to know to ensure my autism is calibrated
the swap memory @hollow spruce was just my KDE w/ a memory leak lmao
The what?
vehicle
I have accepted the reality that I will only ever see the wallpaper on startup before covering it with browser windows 
My car?
the class of vehicle the opel zafira A belongs to is commonly known as an MPV (at least to me)
Yes it is an opel zafira
Idk if its A but i think so
Here we just call it "den otto"
The car?
Ye
mpv type vehicles should be more popular here
about 10 years ago a girlfriend had a ford cmax energi and it was pretty good as far as being a useful vehicle
Luckily battery is simple enough to replace
It's not older than me so it can't be that old 
My buddy had a dog shit (he would call it that, he hated the thing) old Hyundai, it was like a 2007 but it was a Hyundai before ~2012 so it aged in Hyundai years which are a bit like dog years but for cars
Once when I was visiting the exhaust cracked
And the thing sounded like a NASCAR every time he wanted to accelerate
Just the loudest Hyundai maybe ever in history
I followed him to IHOP one morning before it was fixed and I don't think I've ever laughed that hard in my life for so long
Crying the entire way there from laughing every time he hit the gas
So that's what I'm imagining it does

They have safely arrived home
They're just a lot bigger than i was expecting so ill have to think about where to put them
in your stomach
Breaker tripped o7
Bot is offline mode should I start using npm run ?
Speaker 🔊
I havent even connected them yet 
Set up must be pain
(Wire management
Speaker setup
Connect to dolby support equator
)
Respect
3090'd too hard?
meow meow meow meow
Yo 🙂
No, this is the theatre room
Then I have no idea
@tender river what game are u playing
osu! - Rhythm is just a click away! With Ouendan/EBA, Taiko and original gameplay modes, as well as a fully functional level editor.
Either a bad breaker or something is broken or became shorted for a moment
Peak game spotted

Best luck for ure theatre room
Since thid place doesnt have actual groundwires im using a splitterbox with built in fuse. And it seems like it doesnt want to keep being on even with nothing connected
Well, kinda
Which one bass one & which one is 5.1
5.1
But i have some other speaker dor height
So when its all installed ill have 5.1.4
Hmm
Hello @olive sable do u have projector 📽️?
Does this still count as sucking at the game?
What is that ?
Nah, just old 4k tv
Its 3d tho
Awesome
Numbers from Osu













bye 

fixed it tho