programming

[olive sable]

Im not at my pc rn, testing new tv

Optical works

https://tenor.com/view/subwoofer-speakers-subs-bass-woofers-gif-9734151

I found some of those 3D glasses 2ndhand for 10 bucks, im going after them wednesday

All i kneed besides that is an RCA switch. Old tv had more rca

[hoary lion]

10$ for those red and blue 3d glasses?

[olive sable]

No

The one with polarizers

[hoary lion]

Or is it something else that I am not aware of

[olive sable]

4 with mixed polirization, 2 for dual play.

So 6 glasses for 10 bucks

[hoary lion]

Ah more expensive looking ones, I see

Haven't used these for years

[fleet dust]

okay so in the final chapter of my "GTA V Killed my ethernet" saga, it did actually kill the ethernet on my motherboard

I booted a Linux Live USB and it cant connect via ethernet either

[hoary lion]

Classic high level antichwat behavior

Istg if riot vanguard messes something up

[noble zodiac]

and how can you link the two?

[olive sable]

You dont

Its passive

[trim valve]

tbf I've had some pretty bad experiences with vanguard that have made me never want to touch it again

but those were due to faulty memory I think

[noble zodiac]

no, I mean how do you link the two events in cause and effect

[nocturne olive]

Tip: don't play games that have kernel malware (anticheat) built-in
Instead play games that aren't lazy and implement anticheat on the serverside and actually do it properly

[trim valve]

having a clientside anticheat is not mutually exclusive with having a serverside anticheat

[noble zodiac]

you need both if you want anything usable

[trim valve]

I mean it wholy depends on the game

but for shooters I think I mostly agree

[nocturne olive]

You don't need an anticheat on both sides
A proper client just needs to on its end know what the server allows

[noble zodiac]

ok, so how does the server anti cheat know that you were the one controlling the cursor with your mouse

[nocturne olive]

If it's smart enough, it can detect bot patterns in the data, potentially with some kind of machine learning

Either way, I'm not playing games that come with client side kernel anticheat

[noble zodiac]

so if my aim assist makes me play like a pro thats dandy

also, how does the server anti cheat know what kind of visual assistent the client has? just highlight the enemy, all that good jazz

[nocturne olive]

The thing is, making a client side kernel anticheat is being lazy
If you need to touch the kernel, you're doing something wrong

[noble zodiac]

explain why

[trim valve]

tbf even microsoft agrees (citation needed, current source is my iffy memory) that anticheats being a driver is nonideal

[nocturne olive]

Do you KNOW what happens if something goes wrong in the kernel driver?

[noble zodiac]

yes, I am aware

[nocturne olive]

Making a game require a kernel driver is very, very bad practice
A game shouldn't need kernel access just to operate
Kernel access should be reserved for things that actually need it, such as hardware drivers

[trim valve]

iirc the source was https://blogs.windows.com/windowsexperience/2024/09/12/taking-steps-that-drive-resiliency-and-security-for-windows-customers/ in relation to the crowdstike incident

[opaque sigil]

yeah they talked about wanting to move some of the security related stuff into userspace

[noble zodiac]

The point is that you need kernel space anti-cheat to be able to catch kernel space cheats

its not a matter of being lazy or doing it wrong. its a matter of do you want to catch all cheats or not

[nocturne olive]

Well, it's kinda stupid

Gamers should stop accepting games having access to the kernel as all

[opaque sigil]

blame the people making cheats

[trim valve]

blame people for using computers

[noble zodiac]

stupid or not its what reality is

[trim valve]

go back to frolicking outside

[nocturne olive]

Everyone move to Linux so game companies can't make kernel anticheats

[trim valve]

[nocturne olive]

And because modern Windows sucks

[opaque sigil]

we both know that's not true, people will just write kernel modules for the cheats

and you end up at the same point

[nocturne olive]

Well
Definitely never playing any game with kernel level access either way
Because those things are very stupid
Almost as stupid as people that think transformers can do "sentience"

[trim valve]

how is that last line related

[olive sable]

Tv is now fully working

[opaque sigil]

[trim valve]

that sound in the recording is something lol

[olive sable]

my phone's mic can't handle the bass

[fleet dust]

My ethernet was fine for 5 years and then GTA V crashed my PC and upon reboot my ethernet no longer worked

[trim valve]

what level of "not working" are we on

[pulsar sentinel]

I’d like to know how vedal made neuro

[noble zodiac]

dont we all

and yet no one knows

[olive sable]

GET THE CHART

[pulsar sentinel]

What chart

[noble zodiac]

quick copy pasta

[olive sable]

EEEEEEEEE

[fleet dust]

My ethernet just does not connect to the network anymore

[noble zodiac]

get the flow chart as well

[fleet dust]

No activity lights

nothing

its dead dead

[trim valve]

does it show up in device manager

[olive sable]

where is it?

[noble zodiac]

no clue

[fleet dust]

Yes

But it doesnt connect

[noble zodiac]

we need a mod to pin those at some point

[fleet dust]

It doesnt work after a fresh Windows install, or under a Linux Live environment

[trim valve]

if you use a different cable / plug into a different bit of hardware does it still not work

[fleet dust]

Which means its not an OS problem

[fleet dust]

I plugged the cable into 3 different computers and they all connect to it fine

[trim valve]

try plugging your pc into another pc via eithernet, you won't get network but if there's still life in your nic then you should get link lights I think?

[fleet dust]

Ill try that but Im not optimistic

[trim valve]

I'd be very suprised if it straight up died

[fleet dust]

Me too but the fact it doesnt work under a fresh install or under Linux seems to suggest its a hardware problem

I even reflashed the BIOS and its still just not working

No more games with kernel anti cheat I guess

Cant risk shit crashing while accessing the OS kernel and breaking shit

[trim valve]

I mean there's always a chance its not your nic but whatever you're plugged into

[nocturne olive]

Still the version with the typo

[olive sable]

i dont have the new one

[fleet dust]

Yeah but my modem connects fine to everything else I plug into it with the same cable/port

[trim valve]

hm yeah

[olive sable]

can someone explain to me why lambda is uesful?

i have not used it a single time so far

[trim valve]

like in python?

[olive sable]

ye

[trim valve]

I usually use it when I'm using map or filter

also occasionally useful for making partial functions

e.g.

def do_stuff(a: int, b: int):
  print(f"{a} + {b} = {a + b}")

do_stuff_hardcoded_one = lambda a: do_stuff(1, a)

(obviously this isn't the best example because I'm assigning it to a variable but I'd use it if a function took another function as a callback or something)

[olive sable]

so its single-use?

[trim valve]

wdym single use

[olive sable]

you hace to def the function each time

or well, the lambda part

[scarlet arch]

Chat my ant colony has just born its third worker ant

[noble zodiac]

you usually pass the lambda as an argument to another function

lst = [1, 2, 3, 4]
list(map(lambda x: x * 2, lst))

[trim valve]

yeah this isn't the cleanest example

[olive sable]

so whats the point of using lambda instead of just doing * 2 the normal way?

[trim valve]

wdym doing * 2 the normal way

you mean like list comprehensions?

[olive sable]

oh wait, ye *2 on a list only works in numpy

[trim valve]

it also works on regular lists it just does something different

[olive sable]

true

[trim valve]

>>> [1, 2, 3] * 2
[1, 2, 3, 1, 2, 3]

[olive sable]

so lambda is just a faster way to do list comprehemsion?

[trim valve]

not really faster

[noble zodiac]

lambda is not limited to one usecase and can do whatever

map doesnt care what the lambda does, it can apply anything to the list entries

[trim valve]

tbf a list comprehension can too its just sillier

[noble zodiac]

isnt list comprehension just syntax sugar around lambdas

[olive sable]

[rough bloom]

lambdas aren't limited to being used with lists, they can operate on whatever you want because they're just functions

[opaque sigil]

nah i think they have their own case in the language, could be wrong though it's been a while

[trim valve]

[opaque sigil]

now i'm actually curious

time to open the cpython repo

[trim valve]

I'm pretty sure they're different

and there's also a decent performance difference iirc

[opaque sigil]

okay yeah they have their own ast node, that's good enough for me

[noble zodiac]

I just love the functional style of map, filter, reduce, etc pp with lambdas

[opaque sigil]

python thankfully does this lazily

cough cough js

[olive sable]

idk man, still doesnt seem too useful to me

[rough bloom]

they seem to have different bytecode too, no function calls here

[trim valve]

again the major usecase is function callbacks

[olive sable]

ok

[opaque sigil]

neato website

[olive sable]

i guess im just not doing anything that fits the usecase

[noble zodiac]

iono, I use it all the time when I deal with lists

like uhhh....

a.iter().zip(b.iter()).map(|(&x,&y)| (x - y) * (x - y)).fold(0.0, |acc, x| acc + x).sqrt()

[trim valve]

lambdas mentioned, must post https://gist.githubusercontent.com/Zomatree/375316777961d0db2f0f9ed3f06a35a0/raw/fcca7261e95870415dfe77f2bd6293540a631fc9/onelinebot.py

[olive sable]

404

https://tenor.com/view/404-not-found-error-20th-century-fox-gif-24907780

[trim valve]

classic

I love copying the wrong thing

[rough bloom]

I would love to do this in Python too, but all the iterator functions are global functions instead of methods and they don't have a pipe operator or anything

[noble zodiac]

iono, I dont use python often so I have no clue how to do it there

I just assumed its possible

[opaque sigil]

it is you'll just have like 5 nested functions

[rough bloom]

it's possible, just incredibly ugly and hard to read because operations are listed in reverse and are deeply nested

[noble zodiac]

unfortunate

[trim valve]

I love how easy hdr images are to spot on windows

they just subtly mess up everything in their radius

[olive sable]

i dont, windows hdr sucks ass

[trim valve]

its not unusably bad

moreso most software sucks

imo windows is actually pretty sane about it

[olive sable]

idk

[noble zodiac]

filter (== x) $ map sum $ foldr (\u v -> let z = v ++ map (u:) v in filter (\w -> x >= sum w) z) [[]] $ reverse $ takeWhile (<= x) $ map (^y) $ [1..x]

haskell is the king tho

[olive sable]

when i play a hdr game, it automatically does hdr.
but when i close the game, all my monitors have lower brightness and i need to update them by clicking their hdr brightness sliders

[opaque sigil]

have you considered turning off auto hdr

[trim valve]

😭

[olive sable]

its not even auto hdr, minecraft has hdr built in

[rough bloom]

someday I will be able to read this
I know it's not actually that complicated, but the syntax still looks insane to me

[trim valve]

I've only had that issue with some games

[opaque sigil]

actually, are you on w11 or w10 i don't remember

[trim valve]

iirc tf2's experimental vulkan branch was doing that

[olive sable]

w11

[trim valve]

but like ow2 handles it fine

[opaque sigil]

are you running your monitor with hdr on all the time or sth

[olive sable]

its not like i have too much problems with it, i dont get flashbanged each time

yes

are you not supposed to?

[trim valve]

I do

[opaque sigil]

i don't

i probably would though if i had an oled screen or one with way more dimming zones

[olive sable]

i just do this the whole time

not even oled nor dimming zones

just ips

27gp850-b or smthn

[opaque sigil]

i just turn hdr on when i wanna use it via win alt b

[trim valve]

yeah that's kinda fair

I've gotten used to the very specific look of hdr on my monitor

mostly the screen being bright enough to consider using dark themes

[opaque sigil]

tbf for me it's not hdr that's the issue it's the local dimming

[olive sable]

its jsut the same as normal windows but brighter in my experience

[opaque sigil]

do be distracting

[trim valve]

honestly if I have that I haven't noticed it

[opaque sigil]

what monitor do you have

[olive sable]

the only local dimming problem ive had is my monitor automatically lowers backlight brightness in dark scenes, and it makes the bottom row of pixels freak out

[trim valve]

https://www.amazon.co.uk/dp/B09444FNW9

I have no idea if its "proper" hdr or not

but it seems to be

[olive sable]

just VESA DisplayHDR 400, like mine

so a max of 400 nits

[opaque sigil]

400 is only really hdr in name tbh

[trim valve]

I mean its not bright bright

but I can tell when stuff is hdr

[olive sable]

nowhere near the 5000 nits of those big ass 150" tv's

[trim valve]

(which is rare tbh, got bless firefox)

[olive sable]

yeo

same here

[trim valve]

I mostly like it because if I end up with hdr content (be it blender or something else) I have a chance of seeing it properly

[opaque sigil]

but yeah, looks like a standard fully backlit monitor so there shouldn't be too many differences in terms of brightness

local dimming is great for watching stuff for the contrast since you have more control of how bright different parts are but for websites it kinda sucks cause random stuff is now way brighter than other

so you need a lot of dimming zones or go straight to oled where each pixel becomes a dimming zone basically

[olive sable]

[trim valve]

unfortunately my experiences with oled and hdr have been a bit meh

I should check if its improved yet

[rare bramble]

i dont know if list comprehension is technically a lambda though I would count it as one

example_list = [1,2,3,4,5,6,7]
result_list = [x*2 for x in example_list]

or doing a map with a lambda

example_list = [1,2,3,4,5,6,7]
result_list = list(map(lambda x: x * 2, example_list))

you usually wouldnt use actual lambdas in python since list comprehension is so much more useful for any list operation. the only real use i can think of is for giving a function as an argument to another function, but i dont like design patterns like that so I dont use that myself

[stark needle]

that is very of you

[opaque sigil]

in what way

[olive sable]

tbh, ive never use map either

[opaque sigil]

brightness?

[trim valve]

I have a display-p3 colour space laptop

[rare bramble]

ye, i dont know if there is any reason to use maps over list comprehension in python

[opaque sigil]

ah right

[trim valve]

this doesn't play nicely with windows hdr so all my colours get washed out

or well

I think they're washed out

it could just be correct and the normal display is borked 😭

[opaque sigil]

we'll never know

[opaque sigil]

laziness mostly

as in, lazy evalutation

[trim valve]

ah yeah the wide gamut test page doesn't work when I have hdr on 😔

[olive sable]

you know how you can use global x to use something outside a function, can you do something similair for functions inside functions?

like, get data from the function outside?

[opaque sigil]

can't you just access them directly

[trim valve]

I want to say lambdas can do that

[olive sable]

nope

[rough bloom]

wdym "get data outside"? as in, returning it?

[opaque sigil]

i might be misunderstanding something here

[trim valve]

nevermind it just isn't working now ??

[olive sable]

as in, change a variable outside the function

[rare bramble]

i'm pretty sure python variables bleed outside of scope, though that is a bad pattern to use, just return the data you want to use

[trim valve]

ok so after restarting edge

it now work on both hdr and non hdr

I love colour

[rough bloom]

nothing special going on with functions there AFAIK, you just mutate any variable you want
you already described how you can declare a global variable wherever you want

[olive sable]

the probelm with lambda is the /=
test1 = lambda: frustumcorners /= frustumcorners[:,3:None]

invalid syntax

[trim valve]

yeah that's not valid

[olive sable]

im missing something here

[trim valve]

I think you're doing something really wacky though

[frosty seal]

i think i havent touched python in such a long time that i forgot it

its all been just typescript

[opaque sigil]

i'll take typescript over python any day of the week tbh

[frosty seal]

bunjs is extremely neat

just nodejs but quicker

and built in typescript

[opaque sigil]

bun my beloved

[rough bloom]

right, because it's supposed to have an expression there, and assignments are statements
you usually wouldn't write a lambda like this anyway, it's usually a simple "input -> processing -> output" scheme without side effects like assigning some variable

[trim valve]

god bless the myasus app

[frosty seal]

not so splendid now

[trim valve]

I love the dedicated myasus button on the keyboard

[olive sable]

i need it to be a function cuz thats how timeit works
lemme try some things

[rough bloom]

you should be able to just use a normal function then

[rare bramble]

what is supposed to be the lambda variable here? frustumcorners?

[trim valve]

I love semi pointless statements like this though

[rough bloom]

def test1():
    frustumcorners /= frustumcorners[:,3:None]

[frosty seal]

i tell everyone that a two monitor setup is very productive, then i go sit and code on the main monitor and watch house md clips on the other

[trim valve]

I recently realised my samsung phone might actually be wide gamut the default settings are just cursed

[opaque sigil]

alternatively

test = lambda f: f / f[:,3:None]
frustumcorners = test(frustumcorners)

[rough bloom]

just not doing the assignment in the lambda would work too, yeah

[olive sable]

same.
left monitor is my yt distractions, right one is discord and possebly an article. middle is my vs code.
so i only work for 20% of the actual time

[rough bloom]

should be frustumcorners = test(frustumcorners) then

[trim valve]

also the main reason I daily drive hdr is because I like seeing random stuff break

the saga with my screenshots is always fun

[frosty seal]

i think my monitor has fake hdr, it just saturates colors then tells windows it is hdr but then when i activate it everything is desarutated even harder than on a non hdr monitor

[rare bramble]

now that my main monitor died on thursday, I have to suffer with my secondary monitor for a while, I'll try to revive my monitor tomorrow, will send some interesting images from that. if that doesnt work, i'll be shopping for a new ultrawide 💸

[olive sable]

i just have 3x the same one cuz of ocd or smthn

[frosty seal]

but i dont care about hdr, for this price 1440p240 is a steal

[trim valve]

I really want to try out 240hz or maybe 360hz because it sounds funny

[frosty seal]

plus i upgraded from 1080p75 so its absolutely bonkers

[olive sable]

figured out how to do it

solution was using both function and lambda

def test1(frustumcorners):
    frustumcorners /= frustumcorners[:,3:None]

t = timeit.Timer(lambda: test1(frustumcorners))
print(t.timeit())

cewrtainly one of the things ive coded of all time

[sour lotus]

https://tenor.com/view/anime-dance-dancer-girl-anime-girl-gif-26428864

*Flips table, bloody },spent 10 minutes searching where it broke and why

[trim valve]

microbenchmarking to the extreme

[olive sable]

i need every ms i can get

[trim valve]

unfortunately: python

[olive sable]

yepp

which is exactly why im doing htis

if this was a good compiled language the compiler would figure it out, and idd be gpu bound anyways

[trim valve]

ngl I think you're only gonna get so far with python unfortunately

its just kinda hard to do micro performance optimizations ime

[olive sable]

the point is to push python to the limit

[trim valve]

yeah

is your project oss I kinda want to see how it runs on my machine

[olive sable]

ill eventuialy switch to c++ or crab to make actual game

[olive sable]

ye it is

gimme 2 min to push the most recent version

does not work on tbalet anymore sadly btw

[sour lotus]

Wait, we can play this??

https://tenor.com/view/shut-up-take-my-money-money-gif-12362351

[olive sable]

its not as much a game, more jsut the engine

[sour lotus]

yeah yeah semantics

Show me the goooods

please ^^

Keen to see this 3000fps youve been talking about

[olive sable]

its 3000fps if you turn off shadowmapping

[sour lotus]

lol I remember

[olive sable]

and im using a 3090

[sour lotus]

runs dxdiag

RTX 4060 👍

32gb
Ill crank it, turn on all the things

[trim valve]

gtx 1660 👍

[sour lotus]

Raytracing yet?

[olive sable]

https://github.com/samvanmaele/game-engine
code is updated, the web embed link thingy not

[olive sable]

i mean, its possible, but RT core are not accesible on opengl afaik

[trim valve]

[olive sable]

is that my thing?

[trim valve]

yeah lol

[olive sable]

i do have some decently sized files

128MB is fine maybe i think

[olive sable]

lemme see what eating space

[trim valve]

that size was at 7% of the objects downloaded lol

all the objects are different sizes but still lol

[olive sable]

that might be open-cv

[trim valve]

the 600MB one is git

is there a requirements.txt

[olive sable]

yes

[trim valve]

I may be blind

[opaque sigil]

look at the main.py file

[olive sable]

well, kinda

[trim valve]

o

[olive sable]

im pretty sure cv2 is using most of that filespace

[opaque sigil]

btw pls don't put struct in there

[trim valve]

yeah struct is a builtin lib

[olive sable]

ah

[opaque sigil]

it's not an actual package so if you run this it tries to download it and fails

[olive sable]

i stopped using struct anyways lol

i forgor to remove it

[opaque sigil]

[trim valve]

damn

[olive sable]

i need to find a lib that supports 16 bit images without being as bloated as vc2

[trim valve]

imagemagick might

otherwise just diy a format ngl

[olive sable]

ill figure something out

[olive sable]

what are you running htis on rn?

browser or python?

[opaque sigil]

numpy can handle images technically

[trim valve]

python

[olive sable]

aight

what gpu?

[trim valve]

1660

[olive sable]

true

[trim valve]

spike wall my beloved

ew wait I'm on python 3.10

[olive sable]

ye... my homemade blender heightmap has some issues on the edges

idk why, ill just make another one eventually

i need to have a row of alpha no height pixels on the side anyways, which this one doesnt have

[olive sable]

do you have integrated graphics?
40fps is slow even for a 1660

[trim valve]

I do but they're not getting touched

[olive sable]

you sure its not using the igpu?

[olive sable]

ah

88% so cpu limited anywyas

[trim valve]

it runs fine in some places

[trim valve]

if only

[olive sable]

ah

[trim valve]

nice pause menu

[olive sable]

then its just on the python side

[trim valve]

unironically I wonder how much perf you're losing by setting the window title every frame

[olive sable]

never bothered to fix it after changing some image things

[trim valve]

or is it not every frame and its just stupid fast in the menu

[sour lotus]

If Anny, Neuro or Evil isnt up there ill be dissapointed lol

[olive sable]

def calculate_framerate(self):
        
  self.time = time.perf_counter_ns() * 0.000001
  self.frametime = (self.time - self.last_time)
  framerate = 1000/self.frametime
  self.last_time = self.time

  self.savedFramerate += framerate
  self.savedFrames += 1

  if self.time - self.savedtime > 250:

      pygame.display.set_caption(f"Running at {int(self.savedFramerate/self.savedFrames)} fps.")
      self.savedtime = self.time
      self.savedFramerate = 0
      self.savedFrames = 0

[trim valve]

icic

[olive sable]

and i never updated the menu code so ther it is

def calculate_framerate(self):

  self.time = time.perf_counter_ns() * 0.000001
  self.frametime = (self.time - self.last_time)
  framerate = 1000/self.frametime
  
  pygame.display.set_caption(f"Running at {int(framerate)} fps.")
  self.last_time = self.time

[sour lotus]

Vedal's pretty smooth, even the parkour 360 jump

Nice one bud,looks epic

[trim valve]

do not go outside the spikes

IndexError: index 5000 is out of bounds for axis 0 with size 5000 the game will not be happy

[sour lotus]

Well now I have to

Oooh the gravity is pretty smooth, nice touch (Assuming thats part of your setup)

[olive sable]

i used the formula from my physics lessons

its semi based on reaal physics

[sour lotus]

Education works? Pshaw

Its butter smooth man

-1 for no double jump though

[trim valve]

peak

[olive sable]

oh ye, everytime you go into the pause meny and bak, you effectivle have created a new copy of everything in vram without deleting thew old versions

i need to cache that

[trim valve]

it freezes pretty hard when I unpause

[olive sable]

thats why yes

its loading the enetire game's assets in again

[trim valve]

I like the slightly different gpu usage metrics, they're funny

[olive sable]

i did a test once, my 3090's vram was full after pausing and unpauseing about 140 times

so 24gb

[sour lotus]

Prepping for Monday battles, night and fijne week!

[stark needle]

bwaa

hope everyone's having a nice day

/weekend

[olive sable]

i am

[stark needle]

/day before work

[olive sable]

def jump(self, jump):
        
    if not self.jumpTime:
        self.jumpTime = jump
        self.jumpStartHeight = self.height
    
    t = (jump - self.jumpTime)
    jumpheight = 0.005*t - 0.0000049 * (t**2)
    
    if jumpheight < (self.height - self.jumpStartHeight):
        self.player.position[1] = self.height
        self.jumpTime = 0
        return False
    else:
        self.player.position[1] = self.jumpStartHeight + jumpheight
        return True

jump do be working

i just modified the actual physics formula to be in ms instead of s

[olive sable]

the cord of the actual map go from 0 to 5K, you get this issue if you go above 5k, or below -5k.
so if you went out of bounds in the other direction you would have been fine for a bit

but ye, needs to be fixed

[trim valve]

yeah that's fair

[trim valve]

I mean it could be many things

id say most of it is gpu limited but then the last little bit will be waiting for the GPU to do stuff where you don't need to ig

[olive sable]

its either python or the gpu bus i think

[olive sable]

possible ye

[trim valve]

if you want a profiler output I can get you one later

[olive sable]

i already changed some functions

center = np.sum(frustumcorners)/8 -> center = np.mean(frustumcorners, axis=0)

its quite possible the github one has something diffrent, ive changed this part of the code 50 times in the last week

[trim valve]

funny graph

[olive sable]

it do be graphing yes

[trim valve]

I love pylance strict

[olive sable]

it doesnt have any problems on my screen

[trim valve]

I have type checking on strict, so it complains about like any type problems

[olive sable]

oh

ye i didnt know that existed

oh boy

Type of ...... is unknown for most of these

[trim valve]

yeah

happens

[hoary lion]

jokes on you monday's off for canada

[trim valve]

ngl I wouldn't really worry about making your types check nicely unless your project is well suited to them

[olive sable]

wdym with well suited?

[hoary lion]

the 'pythonic' way

[trim valve]

tbh most stuff is kinda fine now, but if you're using a lot of libs without types it's a pain

[hoary lion]

just take care of custom types defined

don't be when you code in python

[olive sable]

what does google do?

[hoary lion]

make shit ton of types defined, almost unnecessary and barely reusing them. If possible, most of the time dict is okay.
You really don't have to encapsulate values that are not received by http with pydantic models

And worst of all, they know that they would use dict so it is a union type

[olive sable]

i do have a types thing in my precompiler

ELEMENT_SIZES = {'SCALAR': 1,
                 'VEC2': 2,
                 'VEC3': 3,
                 'VEC4': 4,
                 'MAT2': 4,
                 'MAT3': 9, 
                 'MAT4': 16}

STRUCT_TYPE = {5120: 'b',
               5121: 'B',
               5122: 'h',
               5123:'H',
               5125:'I',
               5126:'f'}

VALUE_SIZE = {'b': 1,
              'h': 2,
              'i': 4,
              'f': 4}

[hoary lion]

it... it's just a dict

[olive sable]

yep

it is

[hoary lion]

how does it used as a type

[olive sable]

jason, my good old pal

we do a bit of

def readAccesor(gltf, accessor):
    
    bufferView = gltf.bufferViews[accessor.bufferView]
    buffer = gltf.buffers[bufferView.buffer]
    data = gltf.get_data_from_buffer_uri(buffer.uri)
    
    count = accessor.count * ELEMENT_SIZES[accessor.type]
    struct_type = STRUCT_TYPE[accessor.componentType]
    value_size = VALUE_SIZE[struct_type.lower()]
    data = struct.unpack(f'<{count}{struct_type}', data[bufferView.byteOffset + accessor.byteOffset:bufferView.byteOffset + accessor.byteOffset + count * value_size])
    data = np.array(data, dtype=struct_type)
    
    return data

[hoary lion]

oh, this is just like predefined configs

it's fine yeah

but not 'types' if I had to say

[stark needle]

this is not necessairly true

[hoary lion]

well maybe it is just me or im doing a rant on how bad it is documented

I do not want to use genai package bc of this

[olive sable]

bwaadow catching strays here

[hoary lion]

I never had a good time with google packages 🥲

[scarlet arch]

Don't get me started on their language Go

Also good night #programming

[hoary lion]

golang better tbh

[hoary lion]

gn byter

https://github.com/astral-sh/ty

due to astral team python is becoming like rust but easier

[olive sable]

[trim valve]

I do wonder if there's any chance of making "native" wasm modules with sam's setup

i.e. standard pyo3 but compiles into the wasm build

[olive sable]

what im usign now is based of Emscripten

[stark needle]

u need to understand

this is how my inbox looks like

i'm a maintainer for a google repo

we got a lot to do

[olive sable]

damn, a maintainer and STILL paid like a slave

[stark needle]

OSS maintainers dont get paid shit anyways

[hoary lion]

are you perchance a google-genai package manager

[stark needle]

no

[trim valve]

it mildly seems like this is possible which is cool

[hoary lion]

you are omitted from my op list

[olive sable]

it should be

[stark needle]

but

u need to imagine

[olive sable]

limited to vsync tho so cant really compare performance, depending on the browser some are stuck at 60

[trim valve]

😔

[stark needle]

the genai maintainers need to catch up with what AI studio wants, google cloud wants and requested features at the same time

[olive sable]

fair

[rocky stratus]

I'm learning programing just to create an ai

[trim valve]

tbh if its possible to make a rust module that works on both native and wasm easily that would be ideal

but idk how much that'd imrpove your perf

[hoary lion]

fair, but they should reall refrain from removing already good functions and not documenting them well

[safe path]

we need one of the "thing" vs "thing in japan" meme

but instead its "thing" vs "thing in rust"

[olive sable]

web performance isnt really an issue in my experience
as long as i get more than 60 fps im happy

[stark needle]

it's 10 people carrying everything

[warped narwhal]

use rust they said, it will be better they said:

pub fn get_cooldowns<T, T2, F>(&self, objects: &[T], pred: F) -> Vec<T>
where
    T: Eq + Clone,
    F: Fn(usize, &T) -> T2,
{
    objects
        .iter()
        .rev()
        .take(self.max_consecutive + self.cooldown - 1)
        .dedup_with_count()
        .enumerate()
        .map(|(index, (streak, obj))| (index + streak, streak, pred(index, obj), index))
        .filter(|(_, streak, _, _)| *streak >= self.max_consecutive)
        .filter(|(dist, _, _, _)| *dist < self.cooldown)
        .map(|(dist, _, obj, index)| (self.cooldown - dist, obj, index))
        .filter(|(cd, _, _)| *cd > 0)
        .map(|(_, obj, index)| objects.get(index).unwrap().clone())
        .collect()
}

[trim valve]

yeah tbh I'm mostly wondering about native perf increases

[trim valve]

not enough generics

[safe path]

generics

[warped narwhal]

[trim valve]

that reminds me I need to get back to planning my cursed image library

[stark needle]

and out of those 10 people 2-3 did most stuff

[trim valve]

ngl I don't think its that hard to beat the image crate in terms of having loads of generics

[olive sable]

lemme compile, and push to git.
ill test performance on my phone then

ah wait i fucked up, gimme 2 sec to fix

[hoary lion]

thing in rust is infinitely better
unlike japan memes

[olive sable]

idk, half of those thing in japan memes are true

[hoary lion]

rust memes would have 100% rate of truth rate

[stark needle]

"thing" vs "thing replaced by ai"

[hoary lion]

[olive sable]

[trim valve]

ai powered clippy lints

[hoary lion]

shadow what google repo are you maintaining

[prime ridge]

[olive sable]

ai powered ai code writer for ai image generationa and ai tts

aiaiaia

[hoary lion]

only the last two are usable

[prime ridge]

Rust AND Japan mentioned?

I gotta move back to Japan mann

And make more Rust projects

[olive sable]

back?

[stark needle]

i ain't doxing

[hoary lion]

back?

[stark needle]

but it's likely ur using it in one way or another

[hoary lion]

i only use genai but okay

[stark needle]

Indirectly

[hoary lion]

yeah probably somewhere in the hole of dependencies

[stark needle]

Glueless

No more glue for you

[hoary lion]

"less"

[stark needle]

glue

[olive sable]

english is hard ok

i do the double s in less too all the time

pygame.error: Could not create EGL context (call to eglCreateContext failed, reporting an error of EGL_BAD_MATCH)

this is a "it works on my machine, but not in github" situation

[olive sable]

it is with double ss in clueless?

[stark needle]

Ye

[olive sable]

what the

[stark needle]

bro was wrong about being wrong

Bruh

[olive sable]

so less doesnt lose the double s, but full does lose the fouble l?

english fix your shit i swear to god

[hoary lion]

uhhh

[stark needle]

English is the language where

[hoary lion]

i mean, full goes to the front while less goes to the back

[olive sable]

useful?

useless

[stark needle]

"i read something" and "i read something" can be spelled differently and have distinct meanings

[hoary lion]

oh

[stark needle]

hi konii how are you

[prime ridge]

Yeah hometown is in the mountains

[hoary lion]

nvm

[olive sable]

@sage crag hi @sage crag

[hoary lion]

[olive sable]

yes it is finished

yes

you asked me yesterday too lol

did you drink too much?

10% -> 5%

then you must have been really tired cuz i feel like im having the same conversation

for you ye

[stark needle]

It was midnight

Wait true ur -1h

[trim valve]

midnight enjoyer

[stark needle]

yes waying

[hoary lion]

possibly unhealthy behavior

[stark needle]

i still wanna some day just lay outside on the grass and look at the stars at night but I'm not allowed to

[hoary lion]

sleep well

and enough

[olive sable]

if i have the choise, like in vacations, i go to bed at 10am and sleep till 8pm

[stark needle]

i think konii really likes

[olive sable]

chipi chipi chapa chapa

[hoary lion]

24/7 better

[trim valve]

that is diabolical

[hoary lion]

well tbh i dont even follow the principle but

[olive sable]

sorry

[trim valve]

I usually have a very good sleep schedule of 10pm -> 7am

unfortunately I have messed that up a bit recently

[stark needle]

my sleep schedule is 3am->7am

[hoary lion]

look at these insomnia #programming

[olive sable]

no, im just extremely non-functional

in weekdays i either sleep 8 hours, or 2 hours depending on how much work i have

[hoary lion]

diabolical distribution 😭

[olive sable]

so mostly 2

i do look like one

[stark needle]

What if he's night shift nurse

[olive sable]

[olive sable]

am i tho?

[hoary lion]

[stark needle]

true

[olive sable]

just let me by dysfunctional in peace

i eat bad, i sleep bad, and i code in python

https://tenor.com/view/kermit-fall-and-back-gif-25193110

[stark needle]

At least you shower

[trim valve]

surely not

[olive sable]

or do i?

[hoary lion]

this can't be real
I am only 33% dysfunctional than you based on this

[olive sable]

(i do, just not every single day)

i suffer

in school

[stark needle]

Me for the next n years

I struggled today with helping my sis with some math questions

[olive sable]

shadow...

do you need help with math?

we can

[stark needle]

Nah

I need to figure out

Also

It's 2am ish

[olive sable]

sure is

[hoary lion]

time to go go sleep

anti-insomnia #programming

[olive sable]

nah gotta debug this shit first

im having a "who programmed this shit? oh it was me" moment

[hoary lion]

What are you working on?

[olive sable]

so my wasm will run in localhost, but not on github

and i need it to work

[hoary lion]

hmm

good luck

[olive sable]

github workflow wont stop running

they aint gonna bill for this are they?

[amber fractal]

I know they don't charge for me despite running firefox as part of my build system.

[olive sable]

The `set-output` command is deprecated and will be disabled soon. Please upgrade to using Environment Files. nah

im so confused why github doesnt do the opengl context the same way as chrome

[olive sable]

my brother in christ, this does not need top tier grade security

lemme in

https://tenor.com/view/let-me-in-gif-13767303

[amber fractal]

This is why I opted for ssh keys

which same thing tbh

why do I need this much

[olive sable]

its complaining about the opengl context im creating, but the things its complaining about are in a
if sys.platform == "win32", and i know for a fact that github is running on linux and going to the else

[amber fractal]

I can't say I'm suprised

compiling to different oses is always fun

[olive sable]

honestly i have no clue what is even breaking it anymore at this point

pygame.display.init() was the problem, we cant init apprently

apparnelty not, its sttill breaking

goddamnit

oh, its breaking diffrently now

[opaque sigil]

are you running this via a github runner or what

or github pages via wasm

[olive sable]

its an action ye

pages

[opaque sigil]

[olive sable]

i havent tested this in a few weeks so i cant seem to figure out which change broke ir

imma try this one last fix, and then ill call it a day

[hoary lion]

pages??

you adding webgl stuffs for github pages?

[olive sable]

yes

[hoary lion]

and that is compiled with python wasm stuffs

[olive sable]

its only complicated when it WONT WORK

[hoary lion]

well if it works it is much better

better to have 500 lines of code that works rather than 50 that breaks

[olive sable]

code that used to work now does

Traceback (most recent call last):
  File "/data/data/org.python/assets/site-packages/platform_wasm/pygame/__init__.py", line 82, in patch_pygame_display_set_mode
    return __pygame_display_set_mode__(size, flags, depth, display, vsync)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
pygame.error: Could not create EGL context (call to eglCreateContext failed, reporting an error of EGL_BAD_MATCH)

[amber fractal]

[hoary lion]

maybe you are asking too much for a github page

[olive sable]

it literally worked 2 weeks ago

i aint asking for much

[opaque sigil]

didn't you start specifying the opengl version explicitly recently

maybe that caused some issues?

[olive sable]

ye, but it even errors without it

[hoary lion]

man

this python really messed me up

No hashmap use dict lmao
no wonder I suck at data structures

[olive sable]

i think it might be due to the --git command arg used in the yml, but cv2 wont work without it so

[amber fractal]

My honest reaction when failing to build

So many commits lost

[hoary lion]

most optimal commit message

[olive sable]

after a while you justt give up on it

you want to know what changed? look at the fucking changelog

[opaque sigil]

run a git bisect

[olive sable]

i need to calm down lol

[opaque sigil]

also TIL gcc 15 defaults to c23 which breaks older versions of the linux kernel, very dope

[olive sable]

its 3am so maybe tomorrow.
ill go commit by commit and see where it stopped working

[hoary lion]

gn sam

[olive sable]

goodnight

[prime ridge]

@amber fractal r u still working on ML stuff?

[amber fractal]

Not at the moment

Not abandoned, just need time to recollect my thoughts and do something fun

[prime ridge]

nice

[amber fractal]

Ideas are still flowing for those neurons, I'll have to see how well I can handle that once I get back to it.

[prime ridge]

wait handle what?

[amber fractal]

Plenty of competing thoughts to where to go next with the project

[prime ridge]

Well if it helps the only things you need are:

• A nonlinearity of some form
• Input
• Output
• A way to alter the value based on a different value

It will converge so long as you meet those criteria so go nuts

but yeah it can be pretty overwelming

took me ~1 year to come up with my arch

[amber fractal]

It doesn't help that I'm optimizing for speed. Which means I have to weigh my options carefully

[prime ridge]

I mean ur not gunna outperform modern neural networks in speed

but u can def create a specific type of NN that showcases certain concepts

[amber fractal]

I doubt I will be able to outpace big tech, regardless I'm making something and I at least want it to be fast enough.

At least so I can use it

[prime ridge]

When u come off your break you can look a little bit into reservoir computing

It's similar to what you are doing but a bit different

very fascinating but underdeveloped

[amber fractal]

I'll have to look into it

[olive sable]

i am not asleep

[tired hinge]

I might be a little under inform but I assume the way Neuro speak and her model react is through agents with one gives out respond a long with another rigging her model...

[hoary lion]

Go to sleep

[olive sable]

im trying

[hoary lion]

well if things goes sideways just solve that github problem

[prime ridge]

it's late asf for u rn

[tired hinge]

I am stupid, please inform me if you can

[hoary lion]

it will make you extremely tired

[olive sable]

agents? like people?
nah its all automated

LLM

[prime ridge]

agents is an AI term

[olive sable]

ye...

[prime ridge]

doesn't mean actual people

[olive sable]

oh

well

[hoary lion]

stooopid

[grizzled viper]

The short answer is Not necessarily

[prime ridge]

and yes Neuro is an ensemble of agents

[olive sable]

i dont do ai

[prime ridge]

but not the way you described it

[hoary lion]

well it could be but due to latency issues I don't think there would be enough time to do correct agent stuffs

too heavy time-wise

[olive sable]

wrong react, chat goes brrr

[prime ridge]

Neuro has an LLM which passes the text to a TTS model

[olive sable]

ye...

[hoary lion]

you litearlly called him stupid

bad sam

[tired hinge]

I am currently having interest in giving an AI an avatar, especially a realtime responding one, which is why Neuro is really fascinating to me.

[olive sable]

i didnt mean to

meant to reply to apollo

[hoary lion]

just try it out via workflow first

don't chain llms

bad behavior

[paper nova]

Pretty sure agents is like memory call

[hoary lion]

in that context probably

[grizzled viper]

From what i understand her avatar is mostly an API bound to keys

[hoary lion]

but just deploy qdrant locally cause it is more faster

[prime ridge]

I think neuro's movements are probably a cached list of actions that another agent controls

cuz the question look is always triggered when there is a question

[tired hinge]

alright, thank for the advice

[prime ridge]

Do you intend of making something like this yourself or just curious?

[hoary lion]

just run it when query is processing to pass latency

[tired hinge]

I am currently study and really want to make one. However my current approach is through generative models, which is heavy.

[prime ridge]

I doubt the movements are decided based on an LLM

[grizzled viper]

it would be part of her 'wait for the user to stop speaking' most likely connected to the tts model

[tired hinge]

I am thinking of NeRF method for real time talking avatar. However I am still not very well understand of NeRF itself so I can make use of it.

[prime ridge]

no no no

Don't control movements with a NN

that's horribly slow

make a cached list of movements

and give them tags and call them periodically

otherwise neuro will start geeking out

[tired hinge]

I see

[prime ridge]

would look pretty cool tho 💀

[tired hinge]

So my approach is fault to begin with when it comes to real time application

[hoary lion]

yes

NN is slow asf

[prime ridge]

Most things can be cached

[hoary lion]

still slow tho

preds and shi

[prime ridge]

like movements should just be precomputed and called with a tag like <Question_face>

just like how neuro can decide to spin by doing <spin> or something

it's not manually controlled by an AI

[grizzled viper]

If you want true real time as far as i know theres no current working examples

[prime ridge]

too slow and expensive

I mean it's def possible just not worth it

[grizzled viper]

i mean theres always baking it into the runtime custom

[tired hinge]

The expensive case is really hurt

Anyway, my thanks to you all

[hoary lion]

if I become rich there would be signs

[prime ridge]

People gotta donate more to Blender fr

[hoary lion]

brokie ass

[grizzled viper]

threw together a quick mockup if you like??\

[prime ridge]

of what?

[tired hinge]

My apologies, can you explain the sentence

Because me dum dum

[grizzled viper]

wanted a kind of real time face right? that talks via text?

[prime ridge]

how did bro make it that fast

[tired hinge]

wait what?

[grizzled viper]

its horrible but hey isnt everything when its a starter?

[prime ridge]

There is no way it's actually a text to facial movement model

[tired hinge]

It is just a simple show of rigging it when words happens. Not a model.

But it helps

[grizzled viper]

Your welcome! 🙂

[tulip rivet]

sorry for jumping in ^^
is the goal towards an avatar expressing emotions matching a text or more like syncing lips/mouth to speech?

[grizzled viper]

From my understanding it was the matching text rather than lip sync

[tulip rivet]

So wouldn't it be the best aproach to let the llm generate the text-->"Agent/any other system" use a "Emotion Recognition"and add a <tag> -->tts & Avatar control (wo has a set of emotional "displaying" options matching diffrent tags)? Than you youst had to sync the timeing of the streamed text with the start/duration/end of the displaing?!

[grizzled viper]

pretty much yeah

[tulip rivet]

I have no real code for this atm since this is still a backlog task with no timeline, but if anyone wats a runndown on the topic "Emotion Recognition", shoot me a DM I just got my text summarized and translated to english just in case.

[tired hinge]

I am interesting in lips/mouth syncing too

So if you have any insight, please inform me

Why tf am I speaking like a LLM that trying to sound like a 18th centuries rich man speaking and ask for direction

[tulip rivet]

Sry on that front I have nothing realy, since its nothing I have considered to be relevant as of now in my work/project.
The "Emotion Recognition" topic on the other hand is already pre-prepared in the basics.

[tired hinge]

Still much appreciated

[tulip rivet]

but in any case you need a "Middleware" handling all actions related to the in and output for the llm if you want a more complex system

[hoary lion]

middleware

[tulip rivet]

Smartass 😝

[hoary lion]

middleware is certified my opp

[tulip rivet]

what do you mean with certified?

[hoary lion]

oh look at my terrible grammar

*middleware is certified to be my opp including those who enjoy

it is my problem that I cannot comprehend middlewares but still

[tulip rivet]

In computer programs:

You have the frontend (what users see—like a game, a website, or an app).

You have the backend (where the data lives—like a server or a database).

But these two don’t always speak the same language.
That’s where middleware comes in. It’s like the helper in the middle that:

Translates messages

Makes sure data gets to the right place

Keeps everything running smoothly

[hoary lion]

yeah that couldn't be easy

[tulip rivet]

an API Gateway for example could be a middleware

[hoary lion]

oh, that's something I could comprehend

but im doing unity rn so that's story for the future

[tulip rivet]

😅

[ruby timber]

G'morning

[frosty seal]

morning

[maiden geyser]

morbing

[charred ice]

I made a little something https://dev-stream-counter.netlify.app/

[amber fractal]

@prime ridge Yeah seems like the next step is overhauling and almost entirely redoing the neuron design. That'll be fun to tackle in the future

[nocturne olive]

What kind of model are you even building? Are you making one that works like most other models behaving essentially like a complex mathematical function or a more interesting one that works with some completely different system?

[dense cosmos]

I'd like to know too 👀

[stark needle]

Good morning

[ruby timber]

Hi Shadow

[olive sable]

morn

[stark needle]

Brother https://www.hola.com/us/celebrities/20250516832502/bill-gates-predicts-jobs-safe-from-ai/

Bro is clueless

100% some company gonna try to ai generate baseball

Ai generated sports

[tight tinsel]

ai not gonna replace factory jobs thats good to know

i can still move to china

[stark needle]

It is doing that

With robots

U can do most shit with a robot arm these days

[tight tinsel]

damn it guess it cant replace living under a bridge

[tired hinge]

I remember watching some ai generated football

The horror

[stark needle]

But yeah if "making things snd moving things and growing food etc" will be solved problems, how to earn money??

[tight tinsel]

i cant wait for the startup that replaces pilots with ai

[stark needle]

Like sure replace everyone but replacing everyone doesn't mean we will somehow enter a state where you dont have to pay for anything anymore

[tired hinge]

In an Utopia, we don't need to since people love and lving free of worry

But this is not utopia

[stark needle]

https://www.google.com/amp/s/curlytales.com/plans-for-the-worlds-1st-ai-pilotless-passenger-plane-unveiled-are-we-ready-to-take-flight/amp/

You are in for a ride

[tight tinsel]

it neurover

[stark needle]

Ye cause rich people will keep wanting to support their lifestyle so they need customers, which tbf understandable, but like this with replacing everyone it will become self sabotage at some point

[amber fractal]

@nocturne olive @dense cosmos
Considering the current iteration doesn't contain a single mathematical function (it's all binary opts, it is different in that regard)
Still the furthest thing in existence from functional, but I'm slowly working stuff out. That's all I got at the moment.

[dense cosmos]

My network rn is SNN

Spiking Neural Network

I'm considering either using the classic weights * inputs, or just take the outputs of the neurons when they fire as 1, meaning it's just sum of weights

[stark needle]

"everyone is equal and everything is free for everyone" is something extremely unlikely to happen cause someone needs to pay for it and in any sorts of society someone will always wanna have more so yea

[nocturne olive]

Well, is it a model where you can go model(args) and it gives a value or set of values, or is it one that runs continuously and doesn't depend on if an input is given to run?

[maiden geyser]

of course, what's the point of watching sports if no genetics is involved

[dense cosmos]

mine is the latter

here's how I do it currently

[nocturne olive]

Rare instance of someone actually doing something new rather than just boring old Ann

So what are you trying to achieve with it?

[amber fractal]

I'm not at the model stage yet, still in development.

[dense cosmos]

:D I wanted to see if there was any performance benefits to a directed network

performance as in how good the network is

[nocturne olive]

I'm just trying to figure out what classification your model falls into, continuous operation models or input-output models

[dense cosmos]

not how well it runs

[amber fractal]

I'm taking a million years to figure out a way to sum outputs

[nocturne olive]

In what kind of task?

[amber fractal]

Ask me in 4 months when I have a model that exists

[dense cosmos]

though, SNNs have the benefits of sparseness of compute in time

mine is also sparse in connections tho

[nocturne olive]

I thought you'd already know at least if you're going to be making a brain-like or function-like model

[dense cosmos]

I'm trying for a generalized model, but I want a model that can do problem solving, so like planning, reasoning, etc

[amber fractal]

I tend to leave my options open, but also I have a lot of things I want to try out

[dense cosmos]

rn I'm just using evolution because a custom learning algo is a bit much to tackle, and I want to test if it's even worth it

[nocturne olive]

Well, you're gonna need to figure out a way to get that information into the model somehow
You may want to consider pairing the SNN with an embedded transformer of some kind, I've heard that can allow for better language control

[tired hinge]

https://tenor.com/view/ian-malcolm-jurassic-park-preoccupied-gif-14303484

nvm, I was trying being funny with the Jurassic park meme

[dense cosmos]

I was planning on just feeding in binary inputs

[stark needle]

[dense cosmos]

but that's more because I haven't put in much thought on the I/O yet

[stark needle]

https://tenor.com/view/neurosama-evil-neuro-embed-failure-gif-15456327638408323033

[nocturne olive]

Silly

[dense cosmos]

my current plan is to just force some of the neuron's values for input, and read some neurons as output

[amber fractal]

One of the things I'm looking into is AI images (plz mods do not flag this), However I'm am not stepping into the whole AI art thing because I'm creating all of the training data for it.

[tired hinge]

My favorite gif

[nocturne olive]

That's crazy
Maybe try out some new ways of doing it, like how made a sketching model just because he hates diffusion AI generated images stuff

[maiden geyser]

<<<blocks, threads>>>

[nocturne olive]

That looks so silly

[dense cosmos]

currently, how my network does its forward pass is: For every firing neuron, add the weights of its outgoing connections to their destinations. Then a simulation step on the neurons (Leaky Fire-and-forget) where active neurons writes a bitmask. Lastly, a stream compaction, where the bitmask is turned into a set of active indices needed for the first step using a prefix sum scan and a gather.

[amber fractal]

My hatred lies in stealing art, which is why I'll be doing zero of it and instead using my own past works.

[dense cosmos]

that's just how you fire a cuda kernel

[nocturne olive]

It looks like you're defining a type of a nonexistent class and that type is nonexistent class of type blocks, threads

Or maybe there's one more layer of nonexistent class, not sure

[dense cosmos]

it's not a template or anything

<<<>>> are used to launch kernels

you pass in the structure of the kernel

[stark needle]

Tbf the sketching model was pre-existing already

[dense cosmos]

and if you're using dynamic __shared__ memory, then also the amount of memory to allocate

[nocturne olive]

And also like this to define types of things

That's what I'm used to seeing

[dense cosmos]

yeah fair

I didn't invent this notation

[dense cosmos]

I could optimize the first step further by using reduction sums to integrate the synapses

however

it's a pain

thus I will just use atomic operations

outputs are sparse anyways

[nocturne olive]

Just some sillying around I guess

Sillying around is fun though, that's what I've been doing with this NeuroSynth tech

[hoary lion]

chat

what should I use for open world 3d game
unreal or unity

[maiden geyser]

self-written engine

dunia engine

[nocturne olive]

Both kinda suck in their own ways

[hoary lion]

[nocturne olive]

Maybe wait for Godot 3D to become mature

[rough bloom]

Godot
(idk I'm not a game dev)

[nocturne olive]

Godot 3D is not fully mature at the moment
Godot is great though, so I'm sure it'll be good once it's mature

[amber fractal]

bevy

[hoary lion]

well technically it is more of a technical implementation, not much of an actual game-for-release
so if that's the case ima just go for unity at the moment

core is not about 'performance' or 'graphics' tbh

[maiden geyser]

godot is open source unity wannabe

[scarlet arch]

Goooooood morning #programming. I wake up and am immediately in a meeting

[maiden geyser]

morbing

[amber fractal]

Good morning

[nocturne olive]

And also the engine of choice for the Neuro Undertale fangame

[shrewd ember]

Well...

It's the Undertale community.

[nocturne olive]

No, it's the Swarm

Undertale community would use GameMaker

[scarlet arch]

I just set up my out of office notification for my 3 weeks of vacation starting next week, and according to outlook that sent out 102 notification emails. what in the fuck

[nocturne olive]

(me being the lead programmer)

[maiden geyser]

what is that thing you have been playing for the last 443 hours???

[nocturne olive]

Oh, I've had Comfy studio running for a while
It's a Project DIVA charting tool

I have a chart I'm close to finishing, but just have been distracted with a lot of other stuff

[amber fractal]

I've been a bit bored

[livid hatch]

Crazy Effin Robot body

[stark needle]

Sam's game engine

[ruby timber]

pygame

[frosty seal]

i still have scars from such long errors on typescript

[scarlet arch]

very big boys :D
The company I work for has multiple customers that monitor the network traffic of these, so we got a tour once. Very scary if they don't have a cage around them

smallest C++ error :P (I have no clue if it's C++)

[dense cosmos]

it is XD

[livid hatch]

We use Safe move Zones and they have collision sensors the cage is only for finger safety lmao

[dense cosmos]

just realized I accidently doxxed myself

[scarlet arch]

oh yeah I figured. I was around the ones that move whole car chassis

fucking scary lol

if I recall correctly they told us that the robots move the things they're holding just below the denting speed. as in, if they moved any faster the car chassis would deform or get pressure spots from the thing that grabs onto it. absolute insane strength in these

[livid hatch]

I visited a Mercedes production line in Bremen Germany and they have 5 Robots for 1 human they have 2500 workers

[scarlet arch]

yeah :D

the one I was in was Porsche, forgot the city

[livid hatch]

everything synced and in perfect harmony

[scarlet arch]

and then you see their network traffic and you wanna cry

[livid hatch]

We just use profinet cards from PLC to Controller

[stark needle]

Why is this an ad from the official government institution for teacher education

brother

Who asked for this

I already have teachers that do the assignments with chatgpt

[trim valve]

I had a lecturer who cited copilot 💔

[stark needle]

My homeroom openly uses chatgpt

And tells us to use it

We were supposed to read a book

But he said "in the current age you can just ask summaries to chatgpt so there's no point in me giving you this assignment so I'll cancel it"

Like bruh

[trim valve]

yikes

[dense cosmos]

I guess at that point it becomes a matter of whether you want to do it

you always have the choice of chatgpting your essays

but if you wanna learn

you know what to do

except the system doesn't really incentivize learning as much as it does memorization

frick the education system (ours)

idk about your guys' education systems