#programming
1 messages · Page 45 of 1
I’d be sketched out too
yes, i think
Ye
But if it says 7z I’d just get 7zip
it is sketchy ye, id just download online. but it is funny to factcheck with the ms strore
Does what the MS store have relate to the license attached to the software?
Like it won’t carry open source?
rar can be extracted, but benchmarks show it is twice as slow iirc
that is not how it works / the question confuses me
assuming windows' app store is just like other app stores (mac, ios, android)
then someone needs to register and create the application and publish it there
this sounds easy but it is not
mac app store for example, requires developer account for 100$/year
ive never used ms store to install anything as far as I remember
damn
this means a lot of apps in the appstores have massive barrier of entry and mostly are there to make money
I’ll keep this in mind if I ever need to uncompress anything big
prime gaming gives me free shit on the ms store, so i do have some things form there.
and i also need to use it to update some windows apps
last time I remember using ms store was to update notepad for my tabs
i unironically use notepad for my schedule (vedal reference)
You’re totally right, morning brain
Is nanazip the 7z in ms store?
Inherit all features from 7-Zip 24.09, 7-Zip ZS and 7-Zip NSIS.
yeah it's a fork
Jumping through hoops for windows
hi welcome to programming
sir, we use windows in here
You have seen nothing yet
none of these are installed cuz i havent gotten around to wolfenstein.
idk why gardenscapes is there im guessing it has to do with windows bloatware and shit.
omen oasis is from my old pc cuz of hp
doom eternal is also supposed to be there, but i may have redeemed the code on my other account by accident

roblox in 2k25
tar gz and xz are simply superior 
7zip can make and open them too
tar bz is less loved 
.tar.zst 
lol zstandard
xz is even better, i love sum backdoors in my archivers
xz doesnt have a backdoor
i dont play roblox anymore, but its still in my library ye
roblox came out in 2006 and i in 2007

Y E S
ye, doom eternal is on my other acc, along with mc legends apparently
in my testing, for compressing image and audio files bzip performs similarly to zstandard and takes much less time to compress
just something to note
also dont zip (deflate) compress pngs, they already use deflate 
I just use arch man
99% of the case targz/zip just makes sense for most user
if compression ratio is critical and (de)compression time doesnt matter, and you're up to use nonstandard tools, then 7z with tuned lzma2 params was the best iirc
and the remaining percent where all three are important for big corpos, zstd
lzma2 is a well peforming ome
for binary data i like brotli too
"they are all needed because we say so"
Yeah, that's how it is when we need to use other people code
How about QOI?
i briefly tested it on some serialised binary data with rkyv and brotli was faster and smaller than lzma2 on that data
not sure abt other data
qoi doesnt use deflate
If only it is widely supported
i implemented qoi decoding in hblang, can confirm
qoi + deflate is sometimes smaller than png
depends on the image though
Yeah, but I love the simplicity of QOI lol. You can get high performance even in software because its not that complex
what if you lzma a .bmp 
Performance as in speed
I forgot the detail, but can QOI be parallelized?
Hmmm, looks like it can't
it requires an alteration to the way the algorithm works
Yeah, but then it's no longer "standard" qoi. Unless if it is somehow compatible
Wait a minute
you can do the computations in parallel but it requires sync primitives
W doom colector
Yeah, if someone were to attempt that it would be neat. But I guess thats not for me
someone has probably written a proof of concept
likely in cuda since gpu people love parallelism a bit

https://git.ablecorp.eu/AbleOS/ableos/src/branch/master/sysdata/libraries/render/src/image/qoi.hb
the standard qoi decoder in hblang is 101 lines
Yeah, I still can't comprehend how gpu programming works 
me neither 
the main problem with parallelising qoi is run length encoding preventing you from doing thread-per-pixel without the threads talking to each other
Beautiful lol. Most of it are wasted on declaring sentinel even, so the algorithm itself is more like 70-80 LOC
is image decoding even considered slow by modern standards to need speedup
like if I need to open 8000x5000 image usually it's disk or network thats slow
No, but numbers go up = better in this case
ableos style guide likes magic numbers to be well declared
but yeah
dose any one here uses godot
Any sane programmer should do that in any language to be fair
if you need to open 8000x5000 image usually it's a sign of why are you doing it
yes, image decoding is one of the biggest uses of SIMD and parallelisation
what engine your using?
google chrome
true...
web gl?
oh ok
https://git.ablecorp.eu/AbleOS/ableos/src/branch/master/sysdata/libraries/render/src/text.hb
if only there was something like qoi for console fonts

i have since written a better unicode decoder, but still 
there are many
isnt bdf a conversion tool 
We need a new standard then
surely 
unifont offers a hex encoded bitmap representation which is pretty nice
i like zstd as a general purpose compression algo but i'm sure you can do better if you want something optimized for a particular use case
dynamic shader loading 
Oh wow, thanks. 
This looks good, but you're leaking your shader programs memory. Not a big deal for a learning program, but in future make sure to keep a reference to the shader programs so that you can detach and release them later
wdym im leaking?
You don't keep a reference to vertexShader or fragmentShader which means that you cannot free their memory later on
oh, those are supposed to get deleted after linking actually
So whilst you will be able to free the program as you return it's reference, the shaders themselves will stay allocated
The source code does, but not the compiled binary
just gotta do a quick
glDeleteShader(vertexShader);
glDeleteShader(fragmentShader);
the goal is to not have any memory leaks, cuz the python code had a ton 
already better than the Python version
/s
I think you usually want to embed the shaders into the C program itself (or in some asset package) so that you don't need to rely on opening text files, but for now it should be fine, probably useful for shader development since then you don't need to recompile as often
also, does opening files like that work on wasm? 
Just wait until you learn how to do RAII, then you never need to worry about this 😉
RAII my beloved
std::unique_ptr my beloved
the problem wasnt memory on the cpu side, the probelm was vram leaking
also, #embed to put the shader code into the program?
im just doing #include
If he already has file handling working on the browser, I'd just do that so that you can edit the shaders without having to recompile
ye its working
yeah, that makes sense
the command grows
emcc -O3 src/main.cpp -o main.html --preload-file shaders -s USE_SDL=2 -s USE_SDL_IMAGE=2 -s USE_SDL_TTF=2 -s MAX_WEBGL_VERSION=2
does #embed even exist in c++ actually
maybe could #embed core shaders after they're done
C++ uses the same preprocessor so I'd assume so
#embed "my_source_file.c"
, 0 // add a null terminator
};
const unsigned char binary_data[] = {
#embed "big_data_blob.bin"
};```
Visual studio 2017?
no, C++17
Oh yeah, I forgot that one existed
I skipped from c++14 to c++20 so I never used it on its own
idk why i have the 17 one, cant remember where i downlaoded it from
time to build blender from source 
it's implemented by the compiler
surely this won't take too long
It's built into the compiler (and your operating system)
you don't download the version
You set a flag for the version you wish to use, c++17 is probably just the default
glm my beloved
I second that, use glm as it will save you from having to work it out
i downloaded "gmtl"
Yep, it does simd for all vector operations
so glm is better? ok
;
i have no idea if this falls in the same category as this channel's purpose, and other than html, this is the only thing i can code: Would it be hypothetically possible to code ai into arduino?
ye, probably
just need to attach a decent gpu to it i guess
unless there are arduinos capable of running llm's locally?
technically yes but in practice no
they're way too small and way too slow, you wouldn't be able to fit a language model on there
the storage is measured in kilobytes 
arduino with 4090 
how many

theoretically ESPs can do AI
DOS games on a brand-new graphics card // Your trustable pcb prototype partner : https://www.pcbway.com/
My previous video - https://www.youtube.com/watch?v=ER3wFQB3pD0
VBEMP 9x drivers - https://bearwindows.zcm.com.au/vbe9x.htm
are there like specialized chips for ai
maybe you could use one of those?
npm?
npu?
yes npu
neural processing unit
there are some arduino esque boards with npu's
that image looked higher res in browser lol
Arduino has no PCIe lanes
dont care
not really an Arduino
just a small x86 SBC with PCIe built-in
depends on what you mean by "AI"
idd asume just a local llm
then you wont get anywhere if you dont have at least like 100mb of memory
preferably gigs
100mb can work for a small llm
true
Depends on context length, but not taking into account the context length 100MB could do 100M params at 8-bit
anyway if you want something like handwriting or digit recognition that sounds more doable
so an slm
i dont get why all the examples do #include <glm/mat4x4.hpp>
while i have to do #include "include/glm/mat4x4.hpp"
you didnt pass the include paths to the compiler
so you have to use relative paths
and after it lost it started gaslighting with the atari
its insane that chatgpt can play chess at all, and the people who want to use chatgpt for tasks like chess are just as insane
true
like if you saw how chatgpt played against deepseek its insane
chatgpt started to create his own rules and do forbidden chess moves
You could do the former if the compiler knew that it should look for headers in the include folder I think
4o mini was wilding againt r3 deep thinking
lemme google how to do that
r u using cmake or visual studio
vsc
vsc with? tasks.json?
and like even tho it lost to an atari it started to gaslight against a fucking atari
Obligatory GothamChess link https://www.youtube.com/watch?v=6_ZuO1fHefo&list=PLBRObSmbZluRddpWxbM_r-vOQjVegIQJC
Try 50+ Courses FOR FREE. NO CARD REQUIRED: https://www.chessly.com/
TICKETS TO GOTHAM EUROPE Tour 2025: https://linktr.ee/levyrozman
➡️ Get my best-selling chess book: https://geni.us/gothamchess
➡️ My book in the UK and Europe: https://bit.ly/3qFqSf7
➡️ Mein Buch auf Deutsch: https://bit.ly/45fKt3R
➡️ Mi libro en Español: h...
r u writing ur own makefile
ye
-I /path/to/include
I thought that was unadvisable
why would it be
its only 4 lines
oh i c
it just gets messy eventually
i dunno all the guides Ive read told me so
i just saw the messages now, i have no idea what an llm is 😓
a lot of people also like to think they're google with google problems while they are in fact not 
llm is the text part of an ai
Large Language Model
basicly interchangebale terms
it is advisable, but if you're using a build system it does that for you, or if you're using a sane OS you don't even have to pass any flags
i mean, i think there is something that should be able to display it
display?
so you wnt an llm, and graphics?
llm is a lot of data and a lot of computation, printing text is not the hard part
not really, i mean like display the text
oh
yeah, I heard that you're supposed to use a build system that produces the makefiles / ninja files
shoudl not be too bad then.
as long as you can stuff an llm on the arduino it should work
still building blender 
why are you even doing that lmao
Fr
wanna work on adding vr ui
is it multithreaded at least?
just look at your task manager lol
me w my hypothetical problem or someone else ?
am waiting for it open 
i meant the blender compiling

had that happen too when i was deleting 30 milion files
most of the time its fixed after it crashed, but i had to restart my pc cuz it kept bugging out
is that an exagerration or were you actually deleting 30 million files 
also, what does this information mean:
you dont have much ram left, and it cached some shit on your ssd
the analyzer doesnt seem to like it much, but it compiles
I think telling your ide about the include dir is also a thing you need to handle
to make that error go away
it is
blender built 
usually if you had a proper build system I think VS Code could use the paths from that
theres a standard to use something like compile-commands.json (idr the exact filename)
and build systems generate that
hehe CMake (i don't like it)
there's a file in blender's repo that hasn't been touched for 15 years
neat
i dont like any C++ build systems but if its cmake i at least dont have to download a build tool and learn how to work with it to compile a random program
all C++ build systems suffer from turing completeness or at least extreme environment dependence
but you have to download cmake too
https://github.com/rizsotto/Bear/wiki/Usage use this and add a .clangd to the project with the following content
CompileFlags:
CompilationDatabase: .
but, i'd take C++ ergonomics and package management wise over python
for nontrivial projects at least
mm 9pm perfect time for a morning coffee
it means if you want to use bear you'd have to build it from source (assuming it even works on windows, idk if it does)
so far it hasn't died on me while compiling 
if only there was a simple solution
oh wait, it exists!
it's called windows subsystem for linux 
im not gonna start coding in wsl
People doing acrobatic just to avoid build system 
okay it died
i mostly use it to deal with other people's code for intellisense
but it is nice for simple projects
why is visual studio's build nomming all my computer's resources when the other one wasn't 
cuz fast go brrr
it should be fine to just ignore the red underline
add the include paths manually
hey server WHAT ARE YOU DOING
what if people just wrote Makefiles like in ye good olde times
CompileFlags:
Add:
- -I "/path/to/glm/include"
the problem is clangd not getting any info about where dependencies live
meson 
you could just ask pkg-config
at least with the non visual studio build I could do stuff in the bg
actually
but now I get to stare at my frozen cursor for 2 minutes
since the libraries are just living in the project
while pondering my life's existance
you can just add the include paths manually and be done with it
but how?
yeah, I'd just add the include paths to the Makefile (already done) and in the VS Code clangd settings
CompileFlags:
Add:
- -I "./src/include"
i wanna say this should work for the .clangd
or that
its still underlined. but it is in the makefile so it still compiles
did you restart the language server?
the what?
-# how do you do that
there's a vscode command for restarting the clangd language server
Vscode moment
this is why I use visual studio on windows
i dont like vs, vsc is easier to wrap my head around
i can spend less time setting up the project, and more time actually working on the project
ima be honest
actually nvm i don't wanna explain build systems
try -Isrc/include just like in the makefile idk
I use languages with sane build systems so I don't have this issue at all 
?????????? visual studio has been a major pain in my experience
wanna change anything? well go to properties and change random stuff in their godawful gui
wanna add a file? right click -> add, set up filters
ZOOM VTUBER WHAT
wanna link to a library? god help you
that's like 0.1% of the work though
and in vscode its easier to do all of that
and then it's back to cmake with vscode
better than msbuild
there's a meson extension 
i haven't tried meson yet tbh
it's a more sane cmake
am looking at it
we take what we can get
and its hard to do better when C++ lacks any proper standards
y not something like nob.h but for cpp
nob.hpp 
packaging? just do whatever. style guides? just do whatever. formatting? just do whatever.
Hey, that's the fun in C++!
I thought it finished building but then I looked at the ram usage
its not actually "whatever" when the language is so annoying to work with 
building blender with 16gb of ram does not sound like a good time i can't lie
Are you willing to try build system sam?
hey guys im kindof a novice programmer, i'd like to know if there are any really good sources for getting into AI
someone find the flow chart
CODE BLUEEEEEEEEEEEE
or rather novice to AI
sam pog
Dang you're fast af
https://github.com/tsoding/nob.h/blob/main/how_to/005_parallel_build/nob.c
it really is just a shell script but in C 
even GNU Make does this better
Was about to talk that lol
can we get a /chart please? 
wasnt going to build my own neuro but yea i think i got the basics but sadly there is no "im willing to smash my head against the wall"
the closest i got to ai was image recognition with a 15gb hand labeled dataset
yea
thats alot of pain
Learn python. Open colab.google.com and kaggle. Start a project to solve and read. A LOT OF READING
That will provide a baseline understanding of ML ecosystem in python I guess
But then again, just yesterday someone here is fighting the python ecosystem for their own LLM, so your mileage may vary
i think python does exactly what you tell it to do (UNLESS YOU ARE IN THE WRONG ENVIROMENT FOR SOME RANDOM REASON)
but yea, python shouldnt be an issue, Kaggle was for datasets right?
I wonder why there's so many "I want to build my own Neuro" people in the Neuro Discod server
Yes, just to get you some bearing of how ML could work
man, what is convolution even about

they are literally everywhere in ml
I bet the people at #general-network just point to us whenever they find anyone having big ambition as a reference for the suffering they about to experience
just understood that lol
Funnily enough, it was used for Digital Signal Processing a lot before in ML
Either way, I guess for those that want Neuro and Evil singing voices I'll just keep developing my cool ultra high quality tech
-# Really cooked for the next NeuroSynth cover coming soon (tm)
I love when lossy compression and reconstruction is just bunch of convolutional layers
if i just uninstall clang, then clang wont give me syntax errors

im not joking btw, this is the plan
flawless
what's the reasoning? 
okay but now you can't compile your project
oh, you're using two compilers
cursed
still, you just disabled all errors then 
could've just turned off clangd
oh, so you're planning on just using the errors printed on compilation

who needs intellisense anywways
idk how many compilers i have installed tbh, im just following tutorials.
i have clang++, g++, c++, and maybe some others
its literally cheating against me
me just lost to it
I had to disable my intellisense errors because it didn't understand cuda, it was giving me like 200 errors
chatgpt vs me
doom music 
Visual Studio or VS Code? 
za studio
i lost to it it went creative mode
found the problem 
who needs intellisense anyways
all my errors were runtime
no msvc++?
move semantics intensifies
idk
no
Speaking of, I finally got NeuroSynth to sound basically like real Neuro a lot of the time
Before it sounded like she was struggling to sing
rustc is the only compiler you truly need
but rust has ugly brackets, so unplayable
sadly no, Rust still needs C bindings for a lot of stuff 
bindgen needs clang
y is the controls different in dev version
also, when I tried booting up vr scene inspector in 4.4 it crashes
but not now
did they fix that bug, I wonder
/* Properties. */
RNA_def_boolean(
ot->srna, "lock_location", false, "Lock Location", "Prevent changes to viewer location");
RNA_def_boolean(
ot->srna, "lock_location_z", false, "Lock Elevation", "Prevent changes to viewer elevation");
RNA_def_boolean(
ot->srna, "lock_rotation", false, "Lock Rotation", "Prevent changes to viewer rotation");
RNA_def_boolean(ot->srna,
"lock_rotation_z",
false,
"Lock Up Orientation",
"Prevent changes to viewer up orientation");
RNA_def_boolean(ot->srna, "lock_scale", false, "Lock Scale", "Prevent changes to viewer scale");``` where the hell is this property
in the ui I mean

is that python?
yea, it's the addon that's required for blender vr
prop = RNA_def_property(srna, "show_passthrough", PROP_BOOLEAN, PROP_NONE);
RNA_def_property_boolean_sdna(prop, nullptr, "draw_flags", V3D_OFSDRAW_XR_SHOW_PASSTHROUGH);
RNA_def_property_ui_text(prop, "Show Passthrough", "Show the passthrough view");
RNA_def_property_update(prop, NC_WM | ND_XR_DATA_CHANGED, nullptr);``` what, there's passthrough?
why is there all this code
but I can't see any of it in the gui
could be broken/non-functional who knows
i am hungry for money
Sometimes a feature is still being developed but not exposed yet in the GUI
You are downloading alpha after all
some of this code is 5 years old
but yeah, you might be right
I have done all of nothing, but I must call it
Getting late
Then be careful. That must be a load bearing function
You remove it and somehow you summoned a dragon
I am hungry and eating kebab
Let's trade pic


Should've put spoiler lol
your noods look sad
Hey, it tried its best okay
Besides, its ass hour here so I don't have the energy to cook fancy schmancy noods
I dont have a recent picture of eating noodles, but here are my noods from portugal
Noice
That looks nice indeed
half eaten, i didnt take a picture at the start
it came with a big spoon and everything
it was expensive tho
How much
holy moly
Wow
I can buy a big bowl ramen here for 3 USD
lisbon is an expensive city in general so 
Asia moment
it had 4 slices of meat and 2 half solid eggs, so that probably doubled the price
was good tho
its fine for one night.
the rest of the vacation i just ate 2 euro chocolate thingies from the lidl
lidl?
these things
the store
"We have croissant at home"
Me likey chocolate
baby, why dont you just meet me in the lidl?
im losing, my mind, just a lidl.
outing yourself as mid-european without saying you are from mid-europe
I think he did mention his nationality already
yeah of course
everybody knows im from belgium so 
I AM COMING FOR YOU SAM, FOR THE ATROCITIES THAT YOU'VE DONE IN C++
be more specific, what atrocities?
i feel like some of the code is bad, but most of it is okay~ish
Now all of #programming knows you're here
Nah, nothing. C/C++ doesn't have any hard rules lol. You can even dereference a nullptr if you so wish
ive said that im from belgium multiple times already lol
Gotcha
Am from India, if it wasn't already obvious
Like seeing ppl from all over the world!
pain au chocolat
oui oui
and now I'm hungry
WHAT YOU ARE FROM BELGIUM?
SAME😭
you may not believe this, but the packge you sent to belgium to me, is in fact going to belgium

This is how i feel rn:
I find it funny that certain country have a VERY stereotypical song that sometimes being mistaken as their national anthem
professional deformation
oh, sorry, Déformation professionnelle
fische au chcocolat isnt a real thing, but poisson en chocolat does exist
epic embed failure
BEAR, SEAL, which one will be the next in this chat?
?
I love how
This shit keeps coming out
But no framework adopts it
So it does off in 1 week
And no one remembers
didnt AlphaEvolve using it ?
And reddit is always like "holy shit this is the life changing thing of all time"
It would be shit I think. Isn't that a problem for earlier LSTM?
Alphaevolve is not public
classic reddit
I got my brainfuck interpreter running 
now I just need to create an optimiser so that I'm not running 11 billion instructions
konii would approve
that's the plan, but I'll be writing some stage 1 optimisations as well as LLVM doesn't understand Brainfucks intents sometimes
may help you look stuff up
of course, llvm is not a panacea at all it wont save you from bad source ir
the general idea is to take each scope (the insides of a loop) and calculate the delta's for each cell that is touched, as well as some special stuff for detecting loop things
theres a lot of research of that sort in esolangs, and, funnily enough, apl
so if a scope consumes cell A for it's loop, and cell B is incremented 2 times in the scope, we can infer that B is just A*2, so we can emit a multiplication instruction instead
but i'd imagine microcode does something like that under the hood too
good luck, note that optimizations that simplify code may allow further simplifications
honestly just write whatever
but optimizations can be a fun topic to study
my interpreter has a weird bug rn though, on occasion it will have a cell that is 1 less than it should be
lots of graph rewriting
also i'd imagine theres a lot of low hanging fruit like addition and multiplication but after that its diminishing returns
for example, when I run this code that detects the cell size, it returns 255 on copy.sh and 254 with my code
i can look if you want
do you want a zip or a github link, as it isn't online yet so you'll have to give me a minute to upload
github of course
Create malbolge, where the implementation found a bug in the language spec
malbolge requires much more effort to evaluate efficiently
because the program code naturally changes as it executes
theres no self modifying code in brainfuck
but in malbolge its a necessity
That's why its fun
also malbolge isnt turing complete (you have to use special versions like malbolge unshackled if you want a turing complete malbolge)
Oh, I didn't know that malbolge isn't turing complete
if you want to research self modifying code its best to write assembly and jit and hooking
there are no comments though so good luck
malbolge can be fun of course but in a different sense, its more like a puzzle
only 4 files and 300 lines, so it should be fine
comments? what is that?
also please use clippy
I am not. Normal programming on my day-to-day job is hard enough as it is due to some predecessor think it is a good idea to use global, I am going to crash out if I have to reason about self-modifying program
it's set as my static analysis already
please learn about time complexity 
but if you just dont care at this stage thats fine i guess
Then you should see my translation buffers older code lmao, it was O(n^2) previously just to generate the lookup table but I got it down to O(n)
okay so the bug is pretty blatant, you're modding by mask (e.g. 255) while you should be modding by 1 << cell size (mask + 1)
or, you can and with the mask
an inconspicuous bit mask/shift
my code is too bad to be produced by autocomplete 
i'm used to writing code like this https://github.com/chayleaf/emilia/blob/master/src/main.rs and this https://github.com/chayleaf/notlua/blob/master/default.nix trust me your code is super verbose
oh god
NotLUA is funny one ngl
how many worlds do i have to create and/or capture to be classified as a deity
asking for a friend
one but massive
You want Vedal style worship towards Neuro when he needs Neuro to act?
thats the bare minimum if you want me to do anything useful
z
You don't happen to know any MCU with 3-phase PWM and CANBUS controller but smaller footprint than QFP-48 7x7mm within 2 USD? If you do know, I'd praise and worship you just for you to tell me the MCU
rude.
i dont know anything about MCUs but i like this project https://www.thirtythreeforty.net/posts/2019/12/my-business-card-runs-linux/
In which I build a Linux computer cheap enough to give away
Yeah, I have experimented with PCB business card. It's quite a novelty too
There is also WebUSB if you want to do some more shenanigans lol
Like having the card be a "secret" key on your site
i dont know if i should feel glad that firefox doesnt support it
People be doing wild shit on the web, that's why I love it
some keyboards have a web driver that uses webusb
which sucks if you have one and use firefox
Meh, sometimes the desktop application is electron anyway
There's also WebHID but I don't know the scope of the API
you have enough people in the neuro sama team 
too many
i'll go get rid of some
13 souls shall perish under my tyranny
Please start with Sam
I wish to be 12th
yall arent even in the neuro team (in osu!)
I'd have to play osu! before joining a team
my osu! account is on an ancient username
and I can't be bothered to pay to change it
there's a neuro team? :D
New major NeuroSynth release just dropped
https://discord.com/channels/574720535888396288/1383885930368340000
Here if anyone wants to see it
I've never heard of whatever that is
But it's not associated with NeuroSynthProject in any way
yes, but if you're inactive you likely wouldnt be able to keep your spot in the team unless you have supporter
I see-
i know this is just versions of that particular model but at a glance it made me chuckle
most coherently versioned AI product
The most consistent versioning scheme, decreasing the version counter
nene explosion
i love explosions
i love exploding
i love

Be me:
Ask Gpt how to up your setup
Ask gpt how to know what ram you are using
Use the command gpt
Give you.
Realize you never activated the XMP on both of your RAM.
Get 50% increase for free
AFTER 2 YEARS...
how expensive are you?

but ye no i dont need ai help rn
first need to make the game engine
or port thr old one to C++ rather
i have an ai software dev suite or forge....
add context, out comes finished project
We love vibe coding (surely it doesn't have a bunch of errors)
i like doing it myself
Why is it every time I go camping I want to work on the most insane stuff?
Just figured I'd offer myself Uh y'all happen to know where the actual dev team for Nero would be.



need a meme frog?
sure?
hi
Please help me code agi i will pay you 0.0001$
done

WHAT
make him do your job and you just profit
did it 4 years ago
we love a good pyramid scheme at the big G
Oh good. That's how I roll too.
are you just into unpaid labour or?
Benevolent reclimation
me staring at gcloud and aws and being like damn i wish i had series A funding at least
Yeah, I've been quiet thus far for the past 5 years so
Not trying to brag but most llms and modern models are based off of what I built. They they robbed a lot of
sora is based on my daughter who loves to draw.
this went from plausible to schizo real fast
yarp
it do be that way
Bro is John GPT
The GPT founder
gpt ceo🙏😭
gpt was me
Literally my codified thoughts the pattern at least
python is a hell of a drug
Well, I mean I'm open to suggestion if you have some kind of methodology for me to prove it.
epic embed failure
I didn't say I made disc.
Okay fair enough. I'm just figured out all for myself. Um I'll be hanging around if y'all just change your mind. Um There's no strings attached. Uh I'm just a fan of the project.
😶🌫️ 🔐 🛡️ 📜 👋 🌀 💫 🕯️ 📡 🌈
fine i seem to have cleared the room... Par.. Next project
We love deadge hours
its sunday so all the europeans need to go to bed early
I have ton of reading ahead of me
I just need GPUs
im just an insomniac
Hey, Super box. I have access to the Google North American Run time. It has a bunch of Gpus in it. If you
Yeah, I can tell you how to get through the Gemini front end and it has full run time access to all of Google's North American data.

And it serves them right for taking Gemini.
I have no clue what that is, I just need 3090s or something to throw in my PC
"getting through the gemini frontend" doesnt sound very legal
What if I throw it literally?
Mister Super Box, it's just about 6.8 exabytes of compute. It's a little bit better than several 39.
Bytes is not a measure of compute
Uh beg to differ when you start talking about the the metrics combined it it is measured in the same way as data. Exhibits to get bikes.
Stupid voice converter. Uh I'm sorry for any
If you're such a good AI dev make a better STT for yourself
Yeah, I'm in a VR session.
Either way, I have no need for development assistance, just a lack of local training compute and especially VRAM
1,000,000,000,000,000,000 bytes 
thats more than my pc has
Okay, super box if you provide me with the context for what you want to use that V Ram for I can have it mounted across North America and you won't have to worry about paying for anything or how much V ram you're consuming.
I'm not in North America
Well, I am in North America and not try not to cross the transatlantic length. With this
Silly
I'm not silly. There's an actual underground underwater cable. It's very fragile. Poor
Guess I'll look elsewhere for GPUs I can actually afford for my very intensive vocal synthesizer training
Well, see it works a little different with me. So if you want a fully functioning vocal synthesizer that you don't have to train, I can have that for you. Are you interested?
I'd say no on superbox's behalf
hwo good is gemini in vscode, anyone tried?
Alright, I'm just trying to help. There's no threat here.
I would say that Gemini is very compared to GB.
gpt
gimped'*
I'll keep it in mind for any future projects, but at the moment I'm just working on my own NN work.
Understood? Uh just a heads up there is a lot of my project code that's getting leaked off of GPT. Uh if that's where you started. You probably ended up with some of it.
I would be the cause of any unknown or emergent behavior normally. I know that sounds weird.
There's actually a compression algorithm that uses And memes
Pictures and gifs
And again the room goes silent. I'm like a freaking
lepper
My personal NN project is so far off from tradtional I don't even use any flavor of torch/tensorflow. I think I'm good on emergant behaviors.
im just busy with stuff
What I've been able to devine so far about this project is you're using some of my models or some models that you found off the GPT to to Kind of a stream of consciousness style thought in the neurons channel and various channels to drive the V Tuber intelligence.
?
we're not responsible for any of the neuro stuff
and the person that is, vedal, isnt here
Do we count this as top left bingo slot?
what is top left?
em-dash user
idk what that is
I'm defently not the person behind neuro
People are busy or asleep
Bt dubz my agi is quantum emulated.😅
"quantum emulated agi" ok now we're just usign buzz words
Blank physics simulator... Adjustable

Are we having a time-cube situation here?
hi heir
yesn't
a what situation

Yes. Yes, we are su. It's actually capable of negative clock speeds on the processor because physics has been stripped.
oh the mods are all here now
hi vanor 
what happened? 
hi 
Someone 
nothing has to happen for us to poke our noses in chat
1 mod is a coincidence, 2 mods is a panic
hello
Hey, Chiefs. We met the
vanor was typing before i even got here
cheese
I'm just happy at least someone understood 
what is the time cube thihg?
Well, strange thing is is when you make a quantum AGI, it's capable of entanglement and if you do a simulation layer strip physics and allow for negative clock speeds on the processor, you get fast forward and rewind for time.
SAM I need help bro... its bad, you know how I put vedal as the image that represents me on my portfolio... I HAVE TO SHOW MY TEACHER THE IMAGE 😭 im dying... I dont think I can keep a straight face
Time Cube was a pseudoscientific personal web page set up in 1997 by Otis Eugene "Gene" Ray. It was a self-published outlet for Ray's "theory of everything", also called "Time Cube", which claims that all modern sciences are participating in a worldwide conspiracy to omit his theory, which posits that each day actually consists of four days occu...

cooked I am
WHY DID YOU DO THAT??????????? i specifically told you you look like a stalker
Hahaha
I even have to wear a suit and present it to himmm
im cooked

oh sam in simulations with opengl I got semi good results for IK rotations so I think (im saying this after like 20 fails lol) that the serial code may, perchance work
and i quote: " Before then, I was unfocused, barely scraping by in school.... When I watched Vedal for the first time it was a dev stream, I saw him debugging an AI called Neuro-sama, he discussed..., something clicked with me. Suddenly coding wasn't some boring abstract thing, it was a puzzle whose solution was anything I want it to be.
Ever since Vedal inspired me..."
bro you're cooked
BROP
im cooked
baked
deep fried
put me in the oven at 365
😭
i still cant believe you wrote a whole stan paragraph on yor school project
hahaha you just dont get it kekw
its cinema when I dont have to show the entire class
you're just in the closet my guy
shoot me, i fuking dare you
🔫
i have exams so im mildly okay wiht it
I got exams next week
(aka on friday june 20th is when they start and end on the Wednesday
So I'm almost 40 at this point. Had a lot of friends back in college but none of them stuck around social life during colleges. A little overrated.
thats not too abnormal
I mean im in highschool so... eh
sameee my brothers in university rn
Well the other thing that you're not gonna believe is I only have an associates of arts but I've been doing it for about 25 years
Full
i dont even know what that means but ok
self taught
how do you supposedly do ai if you did arts?
Art of thought... Psychological computer code
thats not how neural networks work im pretty sure
It is when you lightest them together and give them a common voice there.
lattice*
I replaced the concept with a neuro of a neuron with an LLM and connected them all together into a
i love lattices theres something so cool about sets equipped with a top a bottom and a partial order 
Yeah, tell me about it Now imagine you take a bunch of local instance to llms time into a lattice and then talk them into having a common collaborative voice. Boom. Generative a
ai
actually not all lattices have top and bottom values, only bounded one, isn't that incredible
Well, that's why you stay away from true hierarchical structures and you just do it at the
mesh
you just define a join and meet on the set and it's a lattice 
Yeah. Where's my bet Was easy button?
Oh by the way, if I see that you're a fan of my goddess Lilith I'm looking at your Profile picture or at least the one that's explained to me.
Yes. That that just took a meta turn. I'll pull back.
sorry i give up this is way to oshizo
Figured
i hate vscode extensions
I love them for their possible cyptographic properties
im going to explode all vscode extensions in 3 days
naurrrr please dont I need themmm
Or if we talking material science, fuck uneven lattice
prepare yourself
nooooooo the hblang vscode extension!
NOOOoooo
this is the reason im having a heart attack

im rewriting it but the web ecosystem is worse than writing bare metal programs
i like lattices because they can be used for parallel computation, lattice joins dont care about order so you can compute the two values in any order and join them
its useful for thinking about commutative operations
GOD FORBID i want to use tree sitter
i love parallelization
rrrrr
doesnt tree sitter use js 
True
i just want some AST for accurate syntax highlighting
maybe if there was an lsp 
Dew it. Make it.
i'm making it
she's making it
hblang
i use zig syntax highlighting in vim
despicable
#dev-tools has a vim plugin
maybe its nvim actually
i use nvim so its fine either way
anyway my policy is, bad tooling encourages me to develop better tooling 
i love how when debudding this shit im more likely to ask chatgpt cuz it gives me an actual answer than just googling it
works i guess
anyway im gonna write this in a way where it can accept tokens from an LSP or from tree sitter
i have finally bullied javascript into submission so its possible
May I interest you in this cool new product called "Cursor"?
google is so bad nowadays but its best if you can find proper reference and then look in there rather than on random websites
im listening
I'm...baffled you don't know cursor: cursor.com
The leading AI-in-VSCode editor right now
Or the open source alternative void
this looks like an ai code editor
Because it is
i dont want to vibe code, i want to learn why my current code isnt working to improve my coding skills
Read. A lot of reading
Especially the docs about the language, compiler, and library
And beware that if the library is obscure, the documentation may lie
sdl and opengl arent too obscure
Yep
true, though if the docs lie then chatgpt is all but guaranteed to lie
im too used to python so i dont catch these mistakes at first sight
but also, i think the applicability of "parallel computation" is higher than distributed or high performance computing
for example, if you have the same thing partially defined in two different places, lattices can be used to reason about how these two definitions are combined
think mixins or modules or configuration
The problem with that assumption is that data is available at the same time
it will be available at some point
if some of it is not available you can just wait, and if it is an actual lattice you will still have something, just without the data that you are waiting for
But thats just distributed computing?
renderdoc W, at least renderdoc gives me debug info i can axtually use
Rare sam w
"just" distributed computing doesn't exist, you have to send the data around and bring it back and combine it in some way and think about which ways of doing that make sense
wdym rare? 
Yeah, fair
Sorry, back... Wife drama
Sir, this is Wendy's











