#programming
1 messages · Page 444 of 1
they're trying to fix the render problems with ue being inefficient by just painting over it
exactly
i hated every form of DLSS but 5 is literally just a generative filter
not to mention expensive to run
so most of ur computation power will be put into FIXING an issue they made
which is insane
i also kinda just want to say. i get that a lot of AAA games use high quality models
theres nothing wrong with that
but there is NO NEED for 2 million polygons
ue's issues with poor rendering efficiency scales with the amount of vertices
and yes leon in the new res evil game does look good, but you dont need to model every pore on his skin
some ue5 games have FAR too many renderpasses
yew
so much so UE tries to do LOD everywhere
UE5 is built for screenshots not gameplay
they re-rendering the same mask 4 times per frame, once for each different process that uses it
instead of re-ussing it
which is why i always hate still visuals for comparisons
once u start to move TAA and DLSS destroy clarity
that is a setting you can turn on, but most dont since optimization is never thought about much
exactly
UE5 kind of makes game dev TOO easy
Sure doesn't seem like it destroys clarity for most people
it does in a lot of settings
im a gamedev, a 3D modeler, and a technical artist, im trained to see it
and i hate it
especially for any proper FPS players
(who will turn down the graphics to try to reduce the slop)
its also just not good by nature
there are better alternatives than temporal based AA
infact AA shouldnt be dealt with until later in a game's development
OK why are they not being used then
i remember when playing fortnite on shitty laptop was better cuz you could turn the foliage down, so you could see a lot more than ps4 players 
because engines and tools by these companies punish the users
alot of effects and stuff are made ONLY to work "properly" with TAA or similar
when they try to make a game every year, they dont even bother trying to fully optemize it
so turning it off will break it in different ways
DLSS exists to "fix" the problems created by TAA in the first place and poor raytracing
and TAA exists to fix problems with other things in the engine
TAA can be implemented better and there are alternative versions
but UE5 TAA which is the most known, is HORRIBLY added
again its optimized for screenshots
in general, i dont like temporally based rendering
the moment u move it goes to shit
OK so why did, say, CDPR move to UE5 if it's so shit
CDPR also did a shit job with cyberpunk and still has a ton of bugs
(from someone who played 2 months ago)
still have NPCs walking on the railings
and crashing cars into the side buildings in a loop
number of other issues
the problem with the game industry is that its still an industry where the main focus is money.
the games released will only be as good as the customers are willing to pay for
OK so CDPR are unskilled idiots, got it
every engine team is skilled but underclocked
i just tested it in satisfactory.... native and dlss look identical. on belts

actually... if you look at belts at a 45degree angle it does look different...
the devs working at the companies want to make good things, 99% of them got into the industry because they like games.
but making games you have to do a lot of things for management, for the stakeholders, for making profit for the company
its sadly not about making good games
it about making games that are just barely good enough
private companies ftw
Like, say, Epic Games?
you seem to be right.. i only looked at close-up belts. like the texture on the belts to compare because i though thats what you guys meant... i'll be honest. i have 200h in satisfactory always with dlss on. and never noticed this being a thing 
fortnite has UE5 optimized for it and companies keep buying the slop engine due to cost and production times
phantom items 
both earn them money with minimal loss despite the quality
it only happens at a distance for somereason.. if you stand right next to it its fine
odd
this is the same belt. at the same speed moving. but i stand right next to it
while they dont have shareholders they have companies paying them for engine use
so they might as well be public
there has to be some wonky rendering stuff going on with satisfactory
thats UE5 doing its thing
you can see it change 
some of the bugs are fundemental issues with TAA+DLSS and some are floating point bugs that they cant be fucked to fix in the UE5 code
when the CEO or whtever think of makign a game,t hey dont actually think about the product
they think about how they can go through this entire thing while makign as much money as possible
scratch maintenance unless its an online game these days lmao

somewhat realted
some larian devs are going to my college to show of some balders gate 3 modding tools
notice how modding isn't connected at all in this chart
any game that releases private server hosting tools is based in my book
Offical mod support in Blade & Sorcey is one of the main reasons I've played that game for so long
i agree depending on the game. like, you can host your own lethal company instance, but for smaller games you kinda expect to have to host it yourself
i dont miss having to use nexusmods for it
thats what i want though
i want to be given the files for self hosting
I'm at 50/50 for mods on mod.io
sometimes I still need nexus
sadly yeah
microsoft ruined excel 😔
i mean its cool if u just want a project to code
plus i dont want to buy office360
this is related to company management
I wanted to program a score counter, but I just went with libreoffice (whatever they call the excel mode)
idk how the libreoffice equivalent of excel works
might be better, idk, never used it
been a while but i usually just use sheets now
excel has some funky things you can do with it i never knew before
you can feed it a csv file with data, and you can make it do so many different data charts and shit
its actually crazy
excel hasd a GUI for its graphs
Some game I'm working on
vzeroupper
libreoffice supports this too and it's soooo nice for quickly validating some data from a program
i plan to make a neural network, is lua a good language for neural networks
not really iirc
is phython good?
most ai stuff is done in python ye
whether the language itself is good is controversial
so how do these things work
If you want a big NN, either use PyTorch or raw C++ and CUDA
torch unmaintained 
For a small NN you could just do NumPy
how straining would it be for making something simple like detecting numbers. I heard it was a simpler neural network somewhere
vzeroupper
so OCR?
ocr?
vzeroupper
Optical Character Recognition
oh yeah that
easiest would jsut be using prebuilt tools
like pytesseract
or google's vision api

hm alrighty, does the neural network I want to make take up allot of resources? I have a pretty weak computer, my lover worries that my computer would like die
to be honest I just want to like understand these things and potentially get into this topic more as I've gained an intrest in these sorts of things
depending on how well it’s formatted I’ve seen 2gb ram models for number reading
I was looking into yolo just now where did my search history go 
well I suppose if you’re training it yourself it’s more
any nixers here? 
https://github.com/s0me1newithhand7s/kube-parts
https://git.hand7s.org/s0me1newithhand7s/kube-parts


wtf do i do with this though
this shit is basically rtx 6000 ada
eh who knows i might actually be interested with 3d related stuff
if you dont want it

i dont mind taking it off your hands
ill trade you my ram for that 4090
but ye nah, the 48gb 4090s are worth 3.5K 
i gues sill post this here too
this is what my ram is worth now
it was only 750 in january
at this rate the ram alone will cost more than i paid for the entire pc
in last september i could have bought 2 3090s with 1K
wrr
idk about worth

multitasking vm
scheduling overhead
0.38ns per instruction
approx, via jank maths
stdout throughput is about 923mb/s
native was like, uh, 2gb/s
this is for printing a single char unbuffered to stdout using fwrite_unchecked
pretty good

https://vortex-downloader.duckdns.org/
can someone check if my site has cert
the last one was a mess that's for sure
1024 * 256 threads each with 3 instructions take 1360 microseconds to fully finish
how does that even work
or is it only because of this specific test case where you can just run them in any order
i hardcoded a short program and then spawn 1024 * 256 threads all running that program
it takes 1300 microseconds to complete, but a lot of that is the overhead from spawning
according to those values, the vm runs at about 0.58GHz on a 5.2GHz gpu
which is slower than expected
though maybe not unsurprising for multitasking since it has thread switching overhead
i would expect a longer running task to show up closer to 1GHz, with a hard cap around 1.6GHz

sketchy link
thats a lot of threads

one 4090 worth of threads

my linux machine has less than 1000 threads with librewolf, some nix dev shells, and a terminal emulator open
262000 threads is a stress testing number
how do you see that?
ps ax -o thcount | awk '{x+=$1}END{print x}'
easy to remember
yuh

hmmm
my vm fully pins a cpu core
wrr
wrrr
though
the moment i try to bake some physics that thread counter probably go 10x
i think i have over 100 hours in interpreter design
does that mean you are good at it or slow?
all i have to show for it is an hbvm impl, my cursed jit proof of concept, and my way-too-fast multitasking vm
ye
why complex when simple do?

different usecases
just like QOL features?
oh
JIT
ye
binary translation
polymorphism
proof info
i see
mhm
aware
basically think of the difference in writing
brainfuck interpreter
or
python 3.13 interpreter
its funny when the cpu starts predicting
nearly 100% runtime spent just uh
jumping to the pointer

down arrow: jump down in code
up arrow: jump up in code
right arrow: jump somewhere, we dont know where
no no
cpu smart
predicting what happens next and saving stuff in cache
red doesnt mean bad?
it means that the program spends a lot of time in those instructions
look where it says 50.00 and 49.85
that means those two instructions are using 99.85% of the program runtime
no
it means over the entire execution of the program, those were the instructions that had the most cumulative time
approximately
you can only measure it so well, this is using the perf utility
why does specifically the jmp instruction take long?
high execution time usually just means the cpu has a hard time predicting the effects of the instruction
also, the cpu does not execute instructions directly
it compiles them to its own language first
relative to other instructions
if the other instructions are costing nothing, then youre gonna see some big numbers on the ones that cost something
thats why you cant directly say "this instruction takes that much time", since the cpu's compiler may have changed the code up a bit to make it easy to execute
branch predictor is smart so everything else is cached
does the cpu need to compile it to its own language for recompiled programs too?
you mean microcode?
idk what i mean
mhm
happens every now and then
with librewolf
try disabling or enabling resistFingerprinting or some privacy settings
or just
wait
it'll work again eventually
assembly/machine code is the lowest level, everything below that (microcode) is a black box
i think you can write a forth interpreter in like 100 bytes... but it's a crappy language.
its a good language
its a language?
stack based 
as opposed to heap based?
yeah, it's like readable and understandable in an hour or 2
stack as in the data structure
you push your arguments and then a call pops and uses them
i've seen code for one before for an old microcontroller.
you overestimate my knowledge of data structures
stack = last in, first out
see my link 
oh i see
wasm works this way too btw

this website hates scrolling. it jumps around for me 
ye
its weird
scroll once to jump to next textbox
or just use tabs for navigation
tab and shift tab
"The machine-code kernel of the eForth Model is 171 x86 instructions in 399 bytes. "
https://github.com/hagna/eforth/blob/master/doc/EFORTH.ASM
i don't hate forth. i myself wont use it but it looks neat
if you like RPN calculators, you'll probably like forth.
remjnd me to continue talking about vms tomorrow
calculators like where you type 1 [enter] 2 [enter] + and get 3
reverse Polish notation

humans are taught the standard notation in schools 
just need to level up your 
i've used an RPN calculator. they were slower because I kept having to think about what I was wanting to do instead of just doing it.
mhm
as i said, its not how you were taught
rpn is literally "do this, then that, then that"
if you're used to it, it will be faster ye
standard algebraic notation is tree shaped, doing multiple things "at once"
practise definitely helps. my friend whose calculator it was could use it faster than i could use my normal one
(yes, we raced)
but prefix, postfix, and infix notation can all be represented with a tree shape
it's just where you "draw" the operation
mipi csi 
Whoa wtf Garry's mod creator gonna launch a game that's like a game engine? Using source 2?
dang the new vscode update changed the default theme
chat, is dis gud? :3
My pc specs:
- CPU: Intel i5 gen 2 2400
- GPU: not exist
- Storage: 300~ gb
- ram: 4 gb
- 64 bits
- barely running roblox
Revive my pc pls
lol I guess I'll drop these here since NN doesnt seem to care. I Renedred a couple neurosynth Demos with the latest build of it.
Where do you get neuro sama voice?
oh Superbox and I have been making it
he's really the public facing side if it though. I just build the training datasets
most folks dont even know I'm invivled
that wasn't a finshed project lmao
only about 1/3 of the song is done
Ironically I've been working on neuro cover tech since june 2023
silent helper 
those poor poor sad RVC models...
in more ways than one
i still got version 3
i actully uploaded a few early fan covers
to youtube
they suck
oh a few of these actully aren't that bad
I played around with a few things and I feel I improved things a bit. Training a new model with a cleaner data set.
This is a fan work. I'm in no way associated with Vedal or Neuro-sama. I used my own model trained from a dataset of Various Neuro VODs.
This is a fan work. I'm in no way associated with Vedal or Neuro-sama. I used my own model trained from a dataset of Various Neuro VODs.
these two aren't bad
i dont know what's worse on this one. the cover or the shitty art i made for it https://www.youtube.com/watch?v=BTqSEEXN3Wg
Well this was a fun one.
I'm now on Version 7 of This AI model, but I may soon be on Version 8.
This is a fan work. I'm in no way associated with Vedal or Neuro-sama. I used my own model trained from a dataset of Various Neuro VODs.
guys, i'm fine-tuning qwen to build a Neuro-like AI personality for an AI vTuber project (streaming on bilibili, the chinese youtube). using unsloth for training.
any tips on collecting and filtering high-quality conversational data? trying to get a consistent personality — deadpan, self-aware, a bit chaotic. would love advice from anyone who's done similar character fine-tuning 

you cand ownlaod your discord data
They can vouch
which is pretty much a collection of everything you've ever typed
in discord
i did that last year, i should see if it changed
Nah it's still the same
i mean the words i use
There was the wordmap we did
But yeah you get json which is very simple to parse into training data
i still have the old wordmap on my sata ssd somewhere
discord also have this feature? i didn't know that before
Go to settings
however some chinese social media don't have this feature
so how can i get the data from there 
I'm so sick of idiots parroting pure bullshit about turboquant and how it's 'bringing down ram prices'
No it isn't morons nobody cares and you got baited by a random shitty literally who from twitter who misunderstood what kv cache quantization even is
Probably depends on the platform and is likely against da rulez
im failign to find it
maybe i was still on windows then?
i found part of it
it was on the windows drive ye
Lmao I forgot about this but that immediately reminded me
i havent seen the stuff on my windows drive in a while

oh
i found it
i think
now i need to make a new one a year later
TODO: fix
ill store this python script on the discord servers so i remember where it is
what?
can i post link to post on forum on rust server?
you dont need jsutification for an opinion, not a suggestion. jsut because the oter person doesnt liek the idea doesnt mean its invlaid
probably not here no
i think that's against neurocord rules
i mean its funny
wait let me screen more
man fight against idea of adding alt tokens


no
is it still your birthday?
birthday role for birthday girl
sure
Birthday Konii???
LMAO
time is a blur
why having a word version of the && a problem?
and why is just using && a problem?
both sides 
&& and & are different operators
but are both called and
i think the actual names are logical and, and bitwise and
what are you gonna name &
yes

qrrr
drumsEvil
jsut call it and
and
and and
and and and and
and and and and
im having a stroke
bitwise and the short circuiting one
ye

chipwise meaning it evaluates ona chip by chip basis
ye
simd chip
can do more than 1 chip
1 chip and 1 chip is
2 chip

i have a chip, i have a chip
chipchip
i have a chip, i have a chip
chipchip chipchip
oh song starts with
cccc

how is it pronounced?
like cht
chipmunk does sound like bird
no that’s fork_knife_plate
google bord
no
too bad, here it is
this is very location based btw
is it?
yes
de, ru
i see



it looks so much dumber on your screen
two eggs with black yolks and a burnt sausage on a yellow plate

i just pasted a png into my text editor?
that's the emoji screenshot png
why does discord make these ones not big?
this one is a year old atp
https://github.com/jdecked/twemoji this is the twemoji fork discord uses, readme says it supports emoji 17.0 (sep 2025) which is the latest complete spec

Might be that discord supports it but the fontface doesn't
Oop misinterpretted that
Could also be that ttheyre just behind on uupdates
I know at one point they were like 2 years behind on electron
Also some work on mobile
emoji fall back to unicode => system ones if discord fails to parse them
and on ios they just are the system ones
always
I've noticed that in screenshots yeah
apple has a weird rule for app store
you can’t use custom emoji typefaces or your app will be rejected
so they kinda have to
when did this become a thing on discord
verbose.
isk who decided that having & and && is good idea
bitand and and
a
btw, in cpp
int const bitand foo(int and val){return val;}
is valid code

programming is arguing over int* val vs int *val
Interesting
I should learn to argue first then
Before the program language
yeah, thst kind of core skill 
A 3090ti popped up for 450
But "sometimes the red vga light on the mobo turn on and there is no image output"

val: &i32
everyone knows int *var is the objectively correct answer
neurocord int* varphobia situation is terrible
i only write correct code sorry
My dear front-end developers (and anyone who’s interested in the future of interfaces):
︀︀
︀︀I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept):
︀︀Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow
userland text measurement
i would hope that my typescript text measurement isnt happening in the kernel
can't wait to drain my battery in 5 minutes because someone decided to do 60fps text reflow in their blog
now that they have consumed all the ram and network transfer, the final frontier is cpu
smh not also using the entire gpu
totally forgor
i mean, its a cool demo but i hope nobody uses it
text reflow is not where we have problems
I hope it is
typescript in kernel


also cpp objectively wrong as always, thats why i never use it because i cant use a language thats wrong


how is podman of all things not working
helo
the sheer amount of random issue's ive had because I'm using podman instead of docker are incredible
podman itself is great, the whole world is just used to the way docker does it
well podman is different from docker of course, but its simpler in implementation so i would have thought itd be easier to make work in wsl
no daemon, user perms etc
i tried podman a while back because i wanted a rootless and daemonless runtime, but it didnt have a working compose at the time so i kinda forgot about it
its probably a skill issue on my part with my increasingly oldge ubuntu install
i think the success of using podman kind of also depends on whether you are making your own images on base or using prebuilt software ones
yeah i dont think thatll work in podman even without wsl lol
well yeah, thats kind of the fundamental problem lol
which is great if you're on the cli but I think I had vscode dev containers hit me with that prompt in a noninteractive terminal
it's sxm4
needs a baseboard/backplane
2 versions upgrayyd of the connection this thing i bought uses (sxm2)
damn, somehow i thought dcs just plug in the consumer gpus in hah
interesting
it definitely looks more efficient though, with flat mount has more and shorter traces prolly
it can deliver more power than pci (at least sxm4 does, it'll give 400w thru socket)
they also run nvlink mesh thru it so for a100 case it gets 600GB/s per gpu on that, pcie can only do 2 gpu pairs
whats your lockin
rocm on rdna4
surely amd is not that bad :v
i ran claude code /insights because id never done so before, was curious if i was right
i was right
in that
i use it almost entirely for desperately troubleshooting rocm
50% of my sessions of cc were just fixing rocm environment shit
wack, usually amd is known for reliable and open drivers
it's just pytorch and shit expect cude
Cude
cuda
so half the time even if you specify the rocm versions of stuff, random dependencies will go ahead and install a bunch of cuda shit and fuck the environment sideways anyways
write your own stack 4head :p
every single use of pytorch on my system i have to manually block it from using any venv or similar local build wheels of pytorch etc
i guarantee nvidia pays these people to keep it broken
tinygrad fixes this
-# 
i cant run anything that uses rocm on this gpu atm without manually forcing system rocm
and 0 prebuilt wheels work
i have to manually build
tbf its just python, ive never had a worse experience with deps than in python
even the ones that supposedly target my gpu (gfx1201) dont work
having no dep manager is better than whatever python does
problem is that amd really only guarantees anything will work on like
ubuntu
for this version of rocm
6.4 kinda works
7.12 nope
something wicked this way comes
python deps are fine, unless you have incompatable deps at latest versions or they require a specific python version, then its hell
when i was working with tf and torch with a friend the deps broke every other week, everything just had random version requirements
yeah I ended up doing that, I just got annoyed by it because vscode handles dev containers silly
we ended up working with a 2 year old python and tf/torch
so you have to go through debug logs to figure it out at all because otherwise vscode just sits there on a blank screen forever
the problem with just using like 6.4 is that 7.1/7.2 rocm are like
mightily faster than 6.4
for my gpu
and also it wasnt supported in 6.4 yet
so it's still fragile
soundcloud has become absolute garbage quality wise in the last year or so, I encounter massive issues with the app every week or so nowadays, didnt used to be a problem couple of years ago
there are 10 competing package/env managers, the repos are half broken, nobody knows how to write broadly compatible code, and its all slow as hell
uv is the fix for all of this trust (
)
didnt you hear uv is already legacy
well openai bought it
https://openai.com/index/openai-to-acquire-astral/ devastating blow to my life
i'm still wondering what openai gets for acquiring astral, other than hire the devs i guess?
what
its a meme, python changes them like socks
Why everyone doom posting ai
there are literally 10+ of them now
anthropic bought bun so they got jelly
no idea really
whenever these corps buy it its a death sentence
I open google I saw "AI replacing all the developers", tiktok another one, youtube same thing
mostly because it's a looming existential issue
uv will be on the chopping block soon enough
cuz they need the doom for more investor hype money
cuz the bubble is almost bursting
just like anthropic "leaked" their next gen model info
"were so close to upending the economy trust, just a few more bucks maannnnn"
Kinda annoying, when I'm doing something, there will always be "AI replacing you" whispering through my left ear and come out through my right ear
it already burst didnt u hear some 3 iq moron on twitter announced google's "new algorithm :oooo" turboquant reduced all memory requirements for llms by 69x and ram prices already cratered, which is why 59832749823749283749 people and outlets keep saying that without any basis in reality surely
even though that's not what it does at all
Can't do anything ig
I'll just ignore, because I wanna see what you guys think about ai
just keep honing your skills, ai will still exist after the correction, but it doesnt get much further that boilerplate. the ability to really solve customer problems is still rare and needed in droves
you will never hear the end of that until it's either confirmed that it wont actually replace people or that it will for sure
Yeah
accountability in general
accountability especially
idiot box computer can't be responsible for itself fucking up because it's too stupid to handle that
and the company will say fuck you
although i worry about it, because usa is notoriously unaccountable for anything
everyone there is looking to sink accountability somewhere people will stop asking about
we're generally alright at suing the everloving fuck out of each other so b2b/b2c accountability will be needed
thats just a show, the real damages to life and working class are never accounted for
It feels like someone just stab me everytimes I'm reading it, you can't block it because it's everywhere.
im talking more in the sense of day to day workplace stuff
itll be needed
for that reason
at least for a while
I'm still in human side
(I do not hate ai, I even like it but these doom posting just something else)
thats the point, but remember to ask the ai to count r in strawberry once in a while to remind you that its dumb autocomplete
i wonder what this conversation will look like were i to read it 20 y ears from now
if im alive
lol
in 20 years i hope i will have managed to have my business and change the landscape of development

i hope soon ill succeed in true artificial stupidity and will just be burned for practicing witchcraft
What yall thinks, next 4 years will look like
same as the last 4 unless something changes
dumb shit globally
shortages in the age of unbelievable excess due to lopsided distribution
lotta pain
cool developments in math/science probably
ai bubble will pop and america will be on foodstamps, europe will become independent, china will be the new global daddy
i was going to say chinese expansion
but also i'll be very curious to see what happens to europe if/when usa doesnt prove too stubborn/big to fail somehow
i think everything might be fundamentally fubar'd
not for the failing thing
but i mean just in general system seems to have the needle broken off on 11
whenever it does stop it's gonna be spectacular
europe is already moving shop at speed, we're kind of done answering to others, i just hope we do it in time
that's a good thing, i never understood why europe took such a weird positioning
really thought they'd go on a collective art student sabbatical
because it was beneficial to everyone for most of the time
you gotta also remember that eu formed out of a brutal war
short term sure
but yeah i mean i get the whole "let's maybe not build up like that" sort of thing
they just overcorrected a bit i think in the grand scheme
some interesting experiments came out of it
lot of good theory for like
human rights stuff
maybe, i think overall it was a net benefit. also we are far different from america and such, we are dozens of completely independent nations, languages and cultures, including ones that hate eachother and have warred for centuries. tough task making everyone work together
i really thought in like the 2010s that things would just continue to evolve positively
near universally
embarassing
maybe it was hope rather than thought
hate conflict for no reason
i mean, when things are good or maybe not close to home it all seems to be going up
yeah it's just naivete
its all coming and going in waves
though we are in a very unique time now, we have never in history had univeral global instant communication
yeah i really cant say i have any clue what's gonna happen next which is why i kinda just have to operate assuming it'll all stay roughly similar
even just a decade ago it wasnt as utterly ubiquitous as it is now
i feel like the now feels very very different depending on what you care about in general which is always true but like some folks are deep into what is basically i feel an alternate reality compared to the day to day of the opposite end
the alternate reality bit is indeed quite worrying, i think that is the biggest curse of the internet. actual monkeys paw stuff
sure you can talk to anyone in the world instantly regardless of language, but also nobody has shared context or values in your neighborhood now
and if you so choose you can hyperspecialize so deep into the things you care about you might as well live on another planet from people around you physically much of the time
it is quite a different world indeed
dont even need to leave home really
ubereats, remote work, vr etc
ready player 1 wasnt exactly prophetic considering how it was timed, i think by the time it was written it was pretty much underway to get us to where we are now
but it still feels funny to think about how i viewed that "future" when i read that like 15 years ago
basically true now minus the doomsday and meta horizons exploded so it's over for that
heh the metaverse is never not going to be funny
imagine dropping 80B for corpo vrchat
zuck will never nnot be a lolcow
real human bean
it's so real and so good for it
also
minipc just spun up randomly
so i ssh to it
????
time to reboot buddy
classic screen locker
most windows thing it could've done
modern standby be like
well it's probably my fault because i force it to stay awake cuz it runs echo's service that i decided needed to be a systemd service instead of anything else
it's just not something i need running all the time in truth
ah
i mean, it being a service process is fine for the reasons i chose but here i am havcing it do nothing
would wake-on-lan work? i never actually tried myself
so it only works when traffic comes into it, but i guess you then still need the discord loop
probably but it's the part that does the waking (it gets the incoming calls/messages)
it's supposed to then send a request to load the model on this pc and if that fails, load it on itself but i intentionally hamstrung it for that so it wouldn't for now
it's probably fuckin restarting over and over because of that now that i think about it
haah
man's barely conscious over there, but he keeps goin
quality slop must be delivered
LOL
it hasnt crashed since the other day but it went almost 40k before givingup
nice
ive seen this number before but i cleared the agent on a timer that checks on it and it still noted it again
really needs me to know
you gotta add restart backoff to your unit
😄
!! filling in last command is so good i need to use it more
i mean so does just pushing up which is hwat i normally do but it's nice for prefixing stuff (although i think arrow up might work for it im always scared it's gonna replace the whole line)
i pretty much only use it for saving 1 second for prefixing sudo lol
me irl
im trying to think of other uses other than sudo or prefixing like 1 thing
i think there's one that directly just sudos the last command as its function but i forgor
one that isnt !!
ah no it's just sudo !!
idk what i thought the other one was
lol, you are completely wrong
- belong to type, variable is pointer
int var; --var declaration
const int var; -- const declaration
int const& ref; -- constant reference
int* ptr; -- simple ptr of type int
const int* ptr; -- simple ptr to type const int
const int *const ptr; --constant ptr to type constant int
when you cast you do (whatevertype*), and not (whatevertype)*
what type is v2 in int* v1, v2;
v2 is indeed another pointer
P L A I N
cuz type for variable is int*, and it is not variable-ptr to type int
it is not int *v1, *v2;
it is int* v1, v2;
use ``
you're wrong
i forgot it would eat that
talking about programming on Discord is hard. so is discussing how to draw an emoji with text..
nvm i retreat
it just cursed thing to declare using ,
i never do it, it is cursed
int* val, val2;
return sizeof(val2);
}```
main:
mov eax, 4
ret
good thing there is no #programming or something
we never talk about programming here
if you want to do that, use a typedef. Should use a type for most ints really anyway way so you know what they actually are.
anyway
it is type_qualifier type pointer_declaration pointer_qualifier variable_name
but yeah the * binds to the var name not the type so
the comma in declarations is an obfuscated C technique though
no opportunity to be stupid
c++ try not to get eaten alive by c challenge: impossible
no it's not, but also * binds to variable not type, it's an objective fact
hence int *v
it is obsolete c notation
can't hear you over you being wrong
ptr should be logically bound to type for humans
what if i legally may not count as a human
than you may use int * val










my podman doesn't work in wsl
