#programming
1 messages · Page 165 of 1
the company, maxon, makes a 3d software product called cinema4d, it's analogous to blender and autodesk maya. they decided to make a free thing called cinebench so you can see how fast a computer can render in c4d without buying the software. i started 3d on c4d in like 2014, it's not obscure by any means. c4d is pretty popular for CGI work
new javascript builtin just dropped


should i use ai to make me game a 3d character
like I once tried to use ai to generate me a character for a fps game
but the rigging sucks
topology is also dogshit
nothing? i may have suggestions
Chatting "generate a highly detailed anime woman 800k polygons"
also parenting and shape keys
Yknow never mind💀
Particularly interesting
you meant to say silly
imagine using c4d in 2025
That's crazy
i'm on blender now of course
Is there even a Blender equivalent to Cinabench?
just dowbload blender xd
I suck at it
That's actually crazy
you would need a chunky pc to run blender
and plus if you don’t use it its just pure bloatware
what arer you saying
what

I mean it is kinda bloated
too many features yk
thank god it's opensource and I can just delete a module
you don't have to use any of them
I think blender's benchmark is a dedicated binary
so no worries on that
it downloads blender before starting the benchmark
oh so they do have a benchmark thing
probably a reduced version then
it's up in the link I sent
i think its the full version and its just running in headless
It doesn't run on the background though, so it's not really bloat
it also needs to download the benchmark scene's so about 1gb of things

I see
Kinda SMOL compared to my 29TB of space
seems to run full blender in the background
"This benchmark ranks in the top 0% of all benchmarks. " 
I forgot i have a passage exam in python
I hate
(Too many requests)
And this error
Microsoft
Sign in
Sorry, but we’re having trouble signing you in.
AADSTS900561: The endpoint only accepts POST requests. Received a GET request
"Spacemacs: Emacs advanced Kit focused on Evil"
?
write your own
emacs clone
torvalds019
the gif's konii uses seem to change every 2 weeks or so 
neurorbhd
what gifs were i using before
we are on chromatic aberation neuroorb now
that one too ye
no, wrong
i used this one
i clearly remmeber you saying "type awawawawa in the gif finder and sen the 3rd one" or something like that
yes, but i didnt use it after that
ok fair enouhg
hi
hi
sup
hi #programming! I'm coming here from #1403977329289728091, because we have a problem. Basically, we are organizing our art using Trello (https://trello.com/b/uIy1Iv6C/swarm-wplace) because, well, it was the first thing i thought of- and I didnt expect this project to become so large. The problem now is that to continue using this board would require either downscaling to 10 members (which defeats the point) or buying a paid plan, which is billed per user. Since the trello has about 130 members, that would be insta-debt for me. Recently someone in the wplace thread found this open-source self-hosted project (https://github.com/plankanban/planka?tab=readme-ov-file) that could replace the Trello board, so I wanted to ask here if anyone would be able to work with me to set this up- ideally in some sort of always-on system, as while i often forget to turn my computer off, I don't want to always have it on.
give us a sec to do the context switch
we may actually have this situated
they need help self hosting a trello clone
oh
im sure someone can do that but thats in the least of my interests
set up a server i guess?
it needs to be cheapish
Silly
oracle cloud is free
Meanwhile I have this very powerful PC/server just sitting around
bragging or being helpful?

superbox why do you always feel the need to interject the convo to boast? 
@peak ingot since you are working on this already (if i understand correctly), I wanted to bring u here
I don't know, what would I get for it
I kinda have a billion servers already
I'm offering a server spot in exchange for something useful to me
so either use the oracle stuff chay mentioned, or self-host it i guess?

My old port 443 stuff was NodeJS
Now I'm replacing it with Kotlin
the only times ive had the displeasure of using nodejs was when we were playing screeps
everytime i stopped playing screeps discord still said i was playing it due to nodejs still running in the background
or something like that idk, it went away after killing the task
Screeps server in Kotlin when?
misread that as discord showed you were playing nodejs lol
ask the devs not us
Well, I guess my server is not needed then
"server" 
Very server, it has static IP and is always online
hey, it serves something, technically a server 
i dont like the if statements in my main loop but there is no way to replace them 
besides that it has gotten pretty clean 
while (running)
{
const VkFence fence = syncManager.inFlightFences[currentFrame];
const VkSemaphore imgAvailable = syncManager.imageAvailableSemaphores[currentFrame];
vkWaitForFences(deviceManager.device, 1, &fence, VK_TRUE, UINT64_MAX);
VkResult result = vkAcquireNextImageKHR(deviceManager.device, frameManager.swapChain, UINT64_MAX, imgAvailable, VK_NULL_HANDLE, &imageIndex);
if (result == VK_ERROR_OUT_OF_DATE_KHR)
{
recreateSwapChain();
continue;
}
else if (result != VK_SUCCESS && result != VK_SUBOPTIMAL_KHR)
{
throw std::runtime_error("failed to acquire swap chain image!");
}
vkResetFences(deviceManager.device, 1, &fence);
submitQueue(fence, imgAvailable);
presentImg();
currentFrame = ++currentFrame % MAX_FRAMES_IN_FLIGHT;
frameCount++;
}
server
As long as the ifs are not on the GPU it's fine
this is not shadercode
Then the branch predictor most likely has you covered
i know youre not supposed to do branching on gpu
ye but i still dont like it
its the only thing making the code look bad
actually ifs are fine if they evaluate to the same result on all execution units within the group
taking a different branch should be fine as long as it happens on a different CU 
i prefer just avoiding it all toghether, but i think i have a singular if on my shadowmapping code maybe
aaaaaaaa yesssssss creating a freaking color object causes the undocumented functions to actually take effect

ok here is the context, hear me out
im trying to make my own compositor
compositor 
so i've been setting the input shape of my window to an empty region right
wlroots or smithay or something else
let me guess, you are just realizing that there is no documentation whatsoever for that bullshit that is wayland
but it doesn't take effect until i do another x11 protocol call 
no i don't even want to touch wayland atm
wait what
well akschually then you are not making a compositor but a window manager
no im making a compositor
vanor and wayland 
it seems like my shaders didnt get saved to my repo and i cant be bothered to start up windows rn
https://wingolog.org/archives/2008/07/26/so-you-want-to-build-a-compositor
im basically doing this but in rust
wingolog: article: so you want to build a compositor
so the issue i faced yesterday was that x11rb does this:
CreateRegion
DeleteRegion
SetRegion
DeleteRegion
why? because let region = CreateReigon...
falls out of scope before I do SetRegion
in C++ it doesn't do that because you manage memory on your own
so after fixing that, i used xtrace to figure out wtf else im doing differently from that c++ code
and then i found out a single call to XCreateColor after setting the input region was required to properly set the input shape down to an empty region
but for some reason flushing doesn't do the same thing so i've been ??? this entire time
wait so what does this as oppose to a window manager
draws pretty shadows probably 
i'm making a compositor for another reason tho
in x11 + i3 u usually use compton for this
or whatever fork of compton
but i need a compositor for a certain hobby of mine
turns out compositors are just x11 windows that you can't interact with at all, so what people like to call XOverlays

good codebase
good thing we decided to replace it with a billion xml files
now there's no bad code, or any code at all for that matter
the concept of compositor within x11 is wild to me and somehow I have never heard of it before
Hey, I think I missed like all of the conversation, but I think there are no compositors in X11? They are only Wayland, no?
surely not the wildest thing u've seen
I'm just not a ricer so it never crossed my mind I guess
also I'm using my own window manager so I would just add any features in there instead of wrapping it in another software
i remember having to install one for stuff like window shadows and blur in a simple wm
It's just confusing because compositor in wayland means something different than in x11
might be just a naming convention then
also just a side note i find it really funny that all of the compositor code basically lives in here: https://github.com/chjj/compton/blob/master/src/compton.c
it uses so many undocumented APIs i can't even
yeah it doesn't do the same compositing they do in wayland
surely there is a fork of compton that is.... readable
currently extending my lisp to get rid of the remaining bit that cant be reified into a continuation
amogus osu
toll free
a
a was found in other channels a long while back lol
no, a
fuck april is 4 months ago
-# wait til this guy realizes 2019 was 6 years ago
nah that im aware off
i cant believe 2007 was 23 years ago
vc chat has been discontinued
u mean in this server?
my bad
\
bad design choice
theyd idnt do it because it looks good
\o/
TwT
cry about it
so rude
won't this mix up concurrent chats? too?
i guess
<insert crying gif and use your imagination for extra sound quality>
classic embed failure
srsly who manages this server.
its cuz you dont have embed perms, jsut chat a bit more and you'l receive them
this is just to combat bots and such

hi vanor 
:c
vanor is my assumption correct or did yall close vc chat for other reasons?
good. I can;t even share files too
did you read the FAQ yet? you need to get embed perms for that
mb :c
no problem, everyone is new at some point! here's the FAQ: #1326303411041206303 message
ty!!
```By being active in text channels. Specifically, you must first obtain 1000 server points: you receive 10-20 points on every message with a 2 minute cooldown. To check your progress, go to bot-channel and type-in t!rank. Note that messages in bot-channel , voice-text and swarm do not give any points.`````
hmm
so I have to
send 100 (or less ) txt msg
idk what konii is doing in VC. but it sure is addictive
osu
they colsed voice chat?
Well I just started doing two stuff at same time, learning how to html css cause i suck and my professor sounds like dead fish in his tutorial videos so im checking another from YouTube and learning now about Mistral to learn how to make AI 
no no no better link https://osu.ppy.sh/teams/2303
osu! - Rhythm is just a click away! With Ouendan/EBA, Taiko and original gameplay modes, as well as a fully functional level editor.
not voice chat, the text chat of voice chat
I think what he meant for new people they cant chat there? Hol up lemme check by rejoining server

for everyone no?
they closed it
oooooH
what is osu about?
Oh i understood incorrectly
yeah i dont think i could
You have unlocked new role
anyways i gotta eat, bye
i think fnf better
ooo
aha
Sorry i understood incorrectly 
started playing osu few weeks ago, can i join?
send application and dm ingame
from my understanding... its a game about muscle mind connection
Anyways if anyone have any good tutorials for mistral or on creating AI, im interested in them 
most of osu is about reading, not about muscle memory
yeah never mind
fnf better for me
you basically have to learn the specific angles in which you have move your cursor
this is something you get through experience, but its not muscle memory
you can just switch from mouse to tablet (or vice versa) and it wont take that long
or change your tablet to a different one
you also have to learn ways of focusing on particular screen spots
I thought he's using somesort of a stylus bc those movements look unnatural for a mouse
main advantage of tablet is no mouse drift
iC
main disadvantage is nib wear and potential grip issues
hmm
and of course osu! is a rhythm game so you also have to learn following the rhythm
and finger control for tapping the circles with your keyboard
though i personally dont use fingers for tapping

If I were a player, I wouldn't as well

also this is how the game looks at the highest level https://www.youtube.com/watch?v=oDDEhCodb3g
Thanks for watching.
We did it
But I'm not done.
Pen slipping? Grip not secure? Try Korelax. (Use Code "MREKK10" for 10% off!) https://korelax.io?sca_ref=8544884.shBG1Up4f0
Profile: http://osu.ppy.sh/users/7562902
Discord server: https://discord.gg/mrekk
Twitch: https://www.twitch.tv/mrekk
Twitter: https://twitter.com/mrekkosu
Skin: https...
the patterns themselves are something you can play just fine like 1k hours into the game, but executing it at that speed takes much much much more practice
damn crazy
yea... it already feels overwhelming js by watching it...
there are many different skillsets too
but anyway wouldnt wanna flood #programming with osu!

what are you guys' thoughts on LangChain? I can't tell if it's actually useful or just a trap for Enterprise™ users
is shiro a guy
or a girl
i am confused
Shiro Cube
ok i think i have the avr sorted out now, just gotta do a quick 100km train ride to the guys house and back. buying 2ndhand is always a fun experience 🥀
at least its only 450 while newprice is 1800
oh they asked if i wanted to go tomorrow
guess il buy my tickets now
tickets have been bought 
im getting my Denon AVR-X6200W tomorrow 
the setup im gonna need to use will be really cursed 
the current subwoofer uses a proprietary jvc connector cuz it also has built-in amps for the full 5.1.
since the new avr doesnt support that i will need to daisy-chain the avr's 
new avr's subwoofer out will go into the old avr's audio in, which will then need to go to the sub
holy shit she's literally me
linkedin is somehow even worse than i expected

vk docs:
We went for the first approach, which ensures that the mapped memory always matches the contents of the allocated memory. Do keep in mind that this may lead to slightly worse performance than explicit flushing, but we'll see why that doesn't matter in the next chapter.
nooo worse performance 
conveniently ignoring
, but we'll see why that doesn't matter in the next chapter.
ye but im not there yet
also to make the triangle stuff faster i made it so each framebuffer had a seperate commandbuffer, but the vulkan tutorial doesnt have that so now i need to figure out how to modify their code
would be easy if i knew what i was actually doing, but this is the first time ive made a vertexbuffer in vulkan
yo any tips for me who's about to go to a programing school?
from my near-zero experience teachers like it when you place some comments so they know what you're trying to do even if the code is shit
lol
will take note while learning
so puts comments on the code im writhing so the teacher actually understands what im doing
understood🫡
beisdes that im not too sure, i havent had many programming classes yet
yeah im going to full on programming class like actual school like grades to decide my future

hopefully my worst fears don't come coding that is too much for my brain to handle
the triangle 🔺, now featuring a vertexbuffer 
the code has rgb values for each vertex too but im jsut not uing that in the shader
maybe i should

is it just me or is the red not very red?
oh good
i thought "i know im colourblind but was it always this bad?"
it looks more red on my screen, i guess it has to do with how hyprland screenshots work?
Looks like the vertex buffer maps the entire screen instead of just triangle
wdym?
oh i get what you mean
nah that shouldnt be an issue
it should be mapping full red to the vertex positions
i think it has to do with what colourprofile my screenshots get saved as, cuz it does look red on my screen
Yeah that could be it
Also I'm surprised this has not been brought up yet
The madlads actually did it
https://www.youtube.com/watch?v=OZRG7Og61mw
The Framework Laptop 16 is now available with AMD Ryzen™ AI 300 Series processors and NVIDIA® GeForce RTX™ 5070 graphics. The new Graphics Module is fully backwards compatible, enabling existing Framework Laptop 16 owners to upgrade to the latest performance. Other upgrades include a 240W USB-C Power Adapter, improved thermals, G-SYNC disp...
so can you just get a new mb or gpu without upgrading the actual laptop?

while i dont really care about the brand of my zapped mineral cube, it is nice to have options
i wish they'd make a battery pack too 
the batteries on these will die so fast if this is actually something they support on battery
im geussing mostly meant for plugged in usage
obviously
I always consider them portable desktops

I've love to see a battery pack as well for these
i wonder how that works with regulation
if it'd count separate from the laptop itself
since it'd technically be nothing more than a battery bank
no officer im not taking a laptop with more than 20K mah onto the plane, im actually taking 4 10K mah battery exstension onboard

140K ?
ye
Wait hmm
14K is the number, my brain got confused by K
That does mean you battery pack needs to be 4K mah
if its classified as a seperate device coulnt it jsut be 10K?
Seems like I made that up, werid
16Kmah in a single device needs approval, but only 2 batteries of that size.
As long as it is separate I see no reason
i woudl like to mention that i hate using nano and i open all my files in vscode
why does nano not use ctrl+c ctrl+v for copy pasting??? its so bad
because nano was made before that convention existed
@midnight sigil do you have a spec for the awa5.0 bytecode that isnt trash
try switching nano for micro

better than what vim uses still probably.. y to yank and p to paste. and yy to yank a line
It's an interesting thing
delete would be d or c or x. all of those delete something 
I'm personally pulling up the emacs tutorial again
im learning vim in vscode with the neovim extension
im getting better at it now a few days in. i can atleast navigate basic and generally edit things now
but im not "vim" levels of fast yet
is it jsut me or is chatgpt5 worse?
liek it actually sucks
i find myslef going to gemini 5/10 times
well.. what are you trying to do. could also be that their auto-router thing is still ass
anything from bugfixing my code to changing setting on my linux
like the firefox hardware accelaration problem i was using gpt5 for 45 min and it wasnt working, and gemini fixed it in 2 minutes
gpt5 is apparently really good at following orders, and not doing anything beyond that. so could be that your prompt is not good for the way gpt5 wants it to be
seems like changing my monitor setting in hyprland.conf turns off numlock for some reason
if gemini can do what i need then thats gpt5's problem
imma keep using gemini
i actually found gpt5 assertive which is good
or.. do it the old fashion way and google your problem.
for me it has been confidently incorect a lot tho
nah
ai for me is like the last resort when i can't find a good solution to a problem and not the first avenue
ai is good for wehn im too lazy to put effort into something
please don't use it like that, speaking from experience
you'll end up questioning why you even bothered asking it in the first place
but i've been doing that for 5 years now and its working great
then your problems aren't complex enough, you're good
i start by googling finding solutions and then if i still didn't find the solution i use all the info i gathered to that point to ask ai for whatever i need
gemini instantly fixed my hdr not being on, while gpt was telling me to change my vulkan code 
ai is useful when i have something like 70% worked out in my head and i know i can figure the rest of it out but just want to do the thing i'm thinking of and see if it works out 
what would classify a problem complex enough?
for me they just become time wasting hallucination machines
they don't know but they can't say they don't know
because they've been trained to doubt themselves less because benchmark number go up
well.. you should never ask ai for cold facts. but to combine informations and spit out a theory they are pretty good at
in this instance them not being able to solve them i guess...
every problem requires some set of facts
and they make them up
well thats why im googling before hand to give them the facts in the prompt 
at that point why not solve the thing yourself
why not give the ai the task for it to solve in 30 seconds instead of me trying it in 30 minutes and getting to the same result?
like i rarely rarely need to use AI for anything. lately the only useful use for AI is asking qwen3-coder to rewrite c++ code to rust 
i made it worse somehow 
i thought it was cuz my mmonitor didnt have hdr on while hyprland was set to 10 bit, but turning on hdr looks even more washed out
well, for me it's 30 seconds but it takes me up to 2 minutes to realize it's wrong, then 30 seconds to fix, something else wrong, and again and again to only realize it's all wrong and jank and it's already been an hour and i could've done the thing by thinking myself and writing it myself in 30 minutes
Toblerone
yeah i don't trust ai with any kind of implementation stuff. but i rather ask for concepts and ideas for implementations rather than straight up code.
the only ai that touches my code is qwen2.5-coder for my autocomplete (
i hope there will be better ones someday...)
that's what i'm saying, it's great for refactoring/boilerplate but usually if you need a "last resort" at a problem while using an actual human brain and google and everything then the ai definitely doesn't know
and will make shit up and waste your time
well i guess i kinda use it like a brainstorming thing. i throw everything i know at it and just let it guess to get new avenues for ideas
even if its wrong in that case, it might throw me a bone to the right direction
"oh it's easy you're so dumb user! just use this non-existent api with this non-existent package and a package from another ecosystem"

i've got a weird and silly case of electrical whine, where doing some tasks on my PC (booting windows, starting servers, switching view in solidworks) cause the hardware to briefly vibrate at an audible pitch. it seems to be directly tied to power usage or processing of some sort within the gpu (RX 6600 XT), and seems to have nothing to do with the fans (the sound starts instantly when i start a hard task for the gpu and stops as soon as that task is done, so clearly not caused by temperature)
i'm not looking for help or anything, i don't think it's harmful, but it's interesting that i can directly hear the gpu struggling
it also changes pitch depending on how hard the task it's working on is, so each view angle in solidworks has its own little sound :)
classic gpu whine probably
used to ge tthat on my 3060ti
back when i did a lot of stable diffusion it was always making noise exaclty when i started the diffusion
my hyperland install is 12 days older than a supposed fix that the hyprland devs put in place already 
go update it then 
cachyos mirror stuff doesnt have it yet for some reason
im trying to figure it out tho
install hyprland-git
(you don't want this)
apparently it got added on 2nd of agust, but the last official release is 19th of july
so ill need to install from git ye
its not working 
what are you trying to do
so the most recent version is failing to build

...
...
...
../src/main.cpp:32:61: error: ‘XDPH_VERSION’ was not declared in this scope
32 | std::cout << "xdg-desktop-portal-hyprland v" << XDPH_VERSION << "\n";
| ^~~~~~~~~~~~
ninja: build stopped: subcommand failed.
huh, that's defined by cmake
Hyprland fails for me so often that I wait months in between updates.
unless it can't read the version file cause it's not there for some reason
cachyos problem maybe?
idk
Could be a cmake version problem for all I know
Invalid syntax that silently fails to declare the macro
bro what
yes
Lol I left it on overnight to see how far it would get and I forgot about it
The tab crashed so I guess it is a real memory leak
guess where in here I closed the evil dog tab
also, its quite fitting for evil's first site to be a memory leak
also wait a sec, evil's site fixed something putting all my mem on standby somehow
6.4% gpu usage from spinning dog. thats impressive
What kind of training data had this hack 

waht do yall use for taking screenshots on linux?
cuz this grimblast shit looks ass in hdr
i personally use flameshot cause i havent found anything better
especially one where i can upload images to a server
spectacle (will only work on kde)

ok so while it would be possible to make a screenshot app myself with vulkan, i would need to account for the wayland weirdness
welcome to the club
we welcome you with open arms

literally line by line the xtrace is the same except for some querying
it just doesn't like the xwindow created by the rust app for whatever the hell reason
i can't deal with this, im gonna just pull some random person's gtk implementation

at least with gtk it'll work on wayland too albiet maybe positioning will be slightly off depending on how things are scaled
lmao
don't tell chayleaf i've forsaken my rust implementation thanks
i will also be performing an unholy combination between c++ and js because i don't feel like implementing widgets in c++, so i'll layer a huge webkit over it 
i would say caught in 4K but this is barely readable
wtf happened there
my hdr

we should forsake guis
i should make an hdr version of this gif so it burns the eyes of people with hdr screens
ive been trying a bunch of post proccesing shit, but i dont think i can fix this shit
if we didn't have desktop environments we'd be in the year of linux frfr
im pretty sure im losing a lot of datta cuz its clipping
glueless
i hate to say this but it just works with kde or gnome
this is in fact hyprland
consider exwm
ye but kde or gnome dont activly hate you
just use the shell instead 
exams no thanks
vanor did you see my hbvm interpreter
explode hdr
idk why but i kinda like hyprland, eventho hyprland seems to be doing everything in its power to make me dislike it
average hyprland experience
hyprland isnt bad, its just not got nearly as much support as kde or gnome
this is an abusive relationship fr fr
what is that two wall, xbox looking thing on top right
ask it
i have no clue, either the file system or the framebuffer???
its unfortunately hard for passion projects to be competitive even in an open source space
its mostly a fault of the complex and bloated ecosystem of modern computing
this seems like a gamut issue
are you mapping sRGB to P3 or smth?

more like the opposite
i have vesa hdr 400 something something 10bit on my displays,a nd hyprland does actually render that from what i can tell correctly.
but the screenshotting tools all suck
wtf is the problem with gui shenanigans and fixing problems with sleep
because they are trying not to blow up your cpu
and sometimes succeeding but mostly just freezing
that's pre main loop

it needs access to my accessibility bus, why
to take control of your cursor
screw it, it gets sudo into my non-sandbox user 
and steal your passwords

works 
i have the source code and i read it it should be fine

if i post scammed links will u ban me chat
demodded
i will ping @safe path and they will ban you


im startig to think taking hdr screenshots is jsut one of those rules of the universe
wherever i go it cant be done
im hitting drm issues now
tell me if this screenshot sucks

it looks fine
hdr screenshot
should do
guys i need some help. i must have changed a setting on vscode or smt cos now this thing shows up where my mouse is and ive got edge scrolling
i have no idea what setting it is
me neither
fuck
who decided this was a good idea

just a few branches
editor.hover.scrollbar or editor.minimap.showHover or editor.scrollbar.verticalScrollbarSize or editor.scrollbar.verticalHasArrows or some combo of these
in the settings json
made it inconsistent
what kind of behavior is that lmfaoooo
why would it be half fixed that's crazy
"editor.stickyScroll.enabled": false,
"editor.hover.sticky": false,
"editor.hover.enabled": false
try setting all these false ig
did that
shoot
that 4 dot line unit is your mouse cursor or the text cursor?
the hover stuff did nothing
its my mouse
it goes to the line that my mouse is hovering
should i just stream it?
wait a minute
this thread seems to suggest that it might get fixed by dragging a file into vscode while this is happening lol
its only in that one window too
it's an open issue i think
are you on linux
olooooll
yeah from the issue thread it's some linux specific interaction with dragging and dropping
you're welcome lol glad to help
weird ass issue
just keep an emergency "wtf.txt" on desktop to drop in when it happens xd
no i just needed to grab and drop anyu existing file
basically just click and hold a file
who tf added a literal tree into github, twigs and all
it's the AUR XD
well a backup of it
plucked it right from the soil and sent it into the digital realm
every single package has its own branch
ok i wont
i will
@tender river look at this
why do epubs compress so badly
un
but
regardless
arent they already compressed

yes in zip
i thought epubs are unironically just zip files
that's more than i would have expected tbh
all three
delete all images from ur epubs (covers etc)
the images are quintessential
extract -> specifically recursively swap the images into tiny webps or something -> recompress into an epub
extract -> compress with zstd/xz
STOP LOOKING AT ME
yeah but then it's not an epub anymore
tell the person on the other end to decompress
i use .cab files myself
to store my cabinet worthy items
hit it wit da NNCP or CMIX
uses transformers lule
supports cuda thank goodness i wouldnt be able to go on if my compression algo wasn't cuda enabled
take it further and extract all your epubs and then train your own bespoke transformer model to achieve the highest compression you can on specifically YOUR epubs
does this look good-ish?
very washed out
it looks a lot better than before but ye comparig it to my sdr one its bad
the reds just look brown
awawa
the gods have decided that your code is special
on clang its the same speed
292 with -Os, march&mtune native
on gcc
whereas only 245 with the same args on clang
which is strange
because on o3 clang is better than gcc o3
turning on lto makes it go down to 270
idfk any more
compilers are magic
ive been on this hdr bullshit for 4 hours now 
gemini said "you can build a program to get the raw stream from PipeWire using the org.freedesktop.portal.ScreenCast API"
and im now jsut like "cant you do tthat tho? please?"
we love vibe coding
according to my quick le google which is always helpful and not at all irrelevant, clang has a more conservative o2 than gcc and specifically it does not use loop vectorization and has less aggressive inlining on o2 which is where llvm gains speed over gcc
nope nevermind on the loop vectorization thing because that was changed in clang3.4+ (unless you're using an old one for the lulz)
basically: 
Have you looked at skiv?
Or if windows you can enable hdr corrector in snipping tool
idk what that is but it probably also doesnt get around the wayland drm protection
Oh
Yeah
thge problem isnt viewing the image, its taking it
raw stream from PipeWire
i was just looking at options for screen capture and going mad because this was really the best way to get a continuous way to see the whole screen with it
it being wayland
Left is HDR, Right is SDR tonemapped
Left will show up properly only on mobile when opened
Or if you have a viewer
ok but like, which api does this use to get the imagedata? cuz if it uses org.freedesktop.portal.Screenshot it will look like this
and most screenshot apps for wayland do apparently
fuck this hdr
im going to bed
Made an SVG that has a little animation in CSS to make a lava lamp that can change color. Maybe if I can be bothered to buy and maintain hosting (I probably can't) I'll make it so it syncs to the stream (if the lava lamp color API is public and such)
What does HDR have to do with PipeWire? PipeWire is the audio backend
So apparently Wayland sucks if you didn't know this before, isn't that crazy?
try doing hdr on x11

I don't even know if my displays support HDR
At least on Windows I was not given the option to HDR
So I probably don't have HDR in my stuff
minecraft for new nintendo 3ds came out the same year as hollow knight
Mission is
how to make dumb mobile Phone to ai smart phone
Sadly phone has not (npu)
☁️
There was Minecraft for the DS what
Holy
minecraft is on every console
You just gave me unrelated nostalgia for the old phone minecraft
I do have that on the apple tv
and some amazon fire tv minecraft
x11 has i3
and a billion other wms because x11 is easy to make wms for
also she meant hdr on x11 doesnt work, has never worked, and will probably never work
none?
For me it's remote desktop capability
I wish i could just turn off hdr, then take the screenshot, and then turn hdr back in 
Like a macro to change the hyprland config file
Wayland would just not cover my needs
You might notice i only slept 3 hour, i need to go after my new avr 
Its going yo be cursed, yet beautiful
I remember doing something called "Sourcing a file that I replace during runtime"
If I were to redo it, I'd keep overwriting a symlink location
Surely hyprctl can do this easily right
Surely
Nah i dont seem to be ablr to turn of hdr with hypctl
I don't think there is an command that you can use with dispatch to change config options, no
just have to replace the config and reload
you probably want to trigger a reload manually anyway because you might have a race-condition otherwise
Alternatively just tone map the hdr screenshot 
Nah i tried, tone mapping didnt work
but putting a sleep in the script probably works too, your screenshot will just take longer
wdym didn't work
Gamma 2.2 and shit
It just didnt work, idk what to tell you
Lemme get the chatgpt logs
It's a lot more complicated than just doing gamma correction unfortunately
Hmmm apparently ffmpeg has a tone mapping filter
I might have to play a bit with that
also can you make games with raw c++ i mean aside from vulkan but
why unreal needs a 4070 to run the engine
like i think unreal is shitly unoptimized
Unreal does a lot more than just games yknow
yeah i know
I was doing stuff like this
you can create realistic shit
-# Insert Sam summoning
plus
where do you find these rtx 3090
can it even work for a 1100$ build?
Used market is always a gamba
for me the used market is dead
i wish jawa existed for me
dose jawa only works in america?
My pc was about 1200 euro so i guess
like wat more than 1000$?
For Sam they just spawn at decent prices 
I had
2ndhand:
5950x + tomahawk x570 + 4x16gb 3600 €300
Case €70
3090 €675
Nh-d15 €50
New:
Ssd €100
But i can find cheaper 3090's here, i just wasnt Lucky
now there is a 3080 ti for 370$
its on facebook
but idk
the fact the cheapest one is a founder edition is crazy
550 euro is about 640.usd i think
Ye its just luck.
Superbox had one for 526 euro through my network
Which country you live in?
probably a lot
i mean super dose live in america right?
yeaaaah
dose the founder edition have most cores
BUT
oh thats not much
I live in belgium and superbox lives in Finland
oooh
now idk if they going to ask me taxes
thats the problem
wait let me show you smt
Performance-Focused Design Vga Msi Ventus 3x | Rtx 3090 24gb GDDR6X | OC Edition Used Very Clean With Box Price 555$ High Fidelity Gaming 24gb Gddr6x Dedicated Memory Nvidia Ampere Arch Ray Tracing Nvidia Dlss Nvidia G-Sync Vr Ready Pci Express Gen 4 Free Fast Safe Delivery Usdt Currency Accepted Order Yours Now 713301
look at this one
now idk about msi
Not bad
but i think the brand makes difference
i think gigabyte or founde edition have most cores?
Msi has its quircks but i prefer it over asus
That's not a thing btw
Cooler/warranty/some minor power delivery things/some are overclocked from the factory
oh i see
The actual GPU is identical since it's just bought from nvidia
Then you need to make the board around it
wait you mean the things around it
i know manufacturs just makes custom stuff to the chipset
like add fans or what ever features
like cooling or something or rgb or idk edition
wait i found someone sells 3090 ti
and around my location
The gpu chip is the same, the board, fans, other cooling stuff is all custom per brand and stuff
ooh
he has a zotac 3090 ti
but i need to call for price
what cpu do i pair it?
i think a r5 9600x
Ye something like that should be fine
Either current r5 or an older r7 if they're similairly priced
All 3090 chips have the same cores, just some may have a bit different clocks
Im back at the gent sint pieters train station 
Last time i was here was for superbox's 3090
same model same core count
clockspeed is gamble (only matters to overclocking anyways, and all third party brand are random. its called silicon lottery) but they have minimum guaranteed specs for it

Damn my train got undelayed, just to get delayed a 2nd time 
Baited 
Silly
You could call it The train station of used hardware
Nah its just cuz i live in the west, and all the good stuff is in the centre
Gent-sint-pieters is like a connection between the west and the rest
I wonder how many 3090s I'll actually end up with, if I can afford it I'll get two more so I can keep one in my main machine
well do i just get it?
i mean i would ask the guy to test it and the guy selling it is a local store for me
i think that 550 3090 is my best option
The thing woth the 550 one is that the owner seems to be non-responsive
The article has been on there since june 3rd
And im pretty sure i tried to contact them before but im not sure
idk if this counts but goddamn, i thhink i want one of these in the newr future
Shoutout to Framework for sending me this Press Kit and forcing me to put my skepticism to bed. This is by far the most impressive piece of tech I've ever reviewed.
0:00-0:48 - Specs and Price
0:48-2:13 - What's the Point?
2:13-3:48 - Soldered RAM + CPU
3:48-5:05 - Who's it For?
5:05-6:27 - Unboxing
6:27-7:43 - Initial Assembly
7:43-14:03 - Har...
like
3w idle is insane
Its nice for if you have dirty power u guess, but i myself dont really care about idke usage
mm well moreof saving energy for electricity ig
As long as it isnt doing 200W while watching youtube, i had that on windows 11
but if theres a fixed bill for it i wouldnt care as well either way
Oh we have solar so its mostly not an issue
man thats so good
oh i see well soo i could try myself go to the shop
The bus app connect to the bus with bluetooth to see your stops now, thats neat 
well then
why is the visual studio autocomplete thing so smart it scares me
it predicted exactly what i wanted to write 2 times in a row
Copilot?
no this funny grey text thing idk if it runs on copilot, microsoft visual studio confuses me
(yes i know there are errors, but at least they are errors i would make)
that's probably copilot then
it was there before i logged in to copilot, maybe it is intellicode? idk im new to this
who knows 
but it also predicted that i wanted to add a variable for arrowLifeTime and as soon as i wrote "public float" it suggested that. there was nothing in the code that mentioned i wanted to do that....
pretty convenient tho
okay it already suggested something smarter than what i was gonna do and added a comment explaining it holy
at this point its reading my mind faster than i can
i don't mind ai taking over the world if it's this convenient
code completions are actually great, agentic stuff is meh compared by how useful those can be
what is that
the tab autocomplete is called "code completions" in vscode and copilot (that's what copilot was originally) and "tab completions" in cursor
the thing you're showing you have in vs
i find copilot kind of nice even though its code is kind of braindead sometimes since it explains stuff without me having to pray to seven divine sigils and win a 0.0000000001% gacha. (looking at you google i always need to ask the exact specific question that everyone else did)
it's really nice for boilerplate
but any other popular llm can do that, i just like that this one is in the window and (presumably) optimized for programming
How much io do you want?
yes
i would use them but i use openrouter for ai stuff now and they don't support the code infill api even for models that do, and it's probably not cheap anyway because of how many requests you're making (basically every time you stop typing and move your cursor). i'm thinking of running a tiny model locally for completions, like a 3b or something, is that even worth it? better than nothing? can anyone vouch
i'm fine with 2022 public beta copilot level performance
that's basically every receiver ever
is that a multi hdmi manager-
Nice 👍
That too










i have given up






