programming

[unkempt citrus]

It depends on how small you want to go

[olive sable]

get a mini itx mobo and new case probably. like a fractal terra probably

[unkempt citrus]

most mini ATX will be compatible with all ATX parts, just having like less NVMe slots

mini ITX gets a bit more complicated

[idle dune]

Main issue for downsizing its that GPU, 4090 is a big fucking boy, most minitix cases cannot fit it let alone cool it proerply.

[jovial remnant]

Tho do make sure the 4090 will fit the new case

[unkempt citrus]

And yeah cooling

[olive sable]

[idle dune]

It won't its a triple slot brick. Most mini itx cases won't fit it.

[amber fractal]

[real sierra]

is that the new razer toaster

[olive sable]

fractal terra with hole cut in the side

[idle dune]

Also power, Mini Itx cases have hard space requriements for PSus, high wattage PSus that can run a 4090 won't be able to do it.

[jovial remnant]

Yeah, the gpu will probably be the main limiting factor on how small you can go

[olive sable]

rizer

[unkempt citrus]

That seems like acutom mod

[idle dune]

Micro ATX is the way to go.

[olive sable]

shoudl be fine

psu's are a standardised size

[unkempt citrus]

https://www.youtube.com/watch?v=eIJUlIc07xM

[idle dune]

Yes but that systme is like 1000w power draw

[unkempt citrus]

This is a build with a list

mini itx with 4090

[amber fractal]

there are smaller ones for ITX, and you'd be limited to those

[olive sable]

[amber fractal]

also standard, just diffrent sized

[idle dune]

https://cdn.discordapp.com/attachments/1359530258344185876/1380218154730520607/GIF_Evil_Neuro_What_The_Fuck.gif

[olive sable]

i think those would still have decent wattage

[simple sky]

Idk, watching a video on lambda calculus, it just seems like a very basic programming language

[patent walrus]

that looks sick

[unkempt citrus]

https://www.youtube.com/watch?v=t20lyc1H_Ug Heres another with watercooled GPU

[patent walrus]

omw to damage my 4090 by trying to water cool it

[unkempt citrus]

Have fun

https://www.reddit.com/r/sffpc will have more resources

[real sierra]

https://cdn.discordapp.com/attachments/1072697081443131476/1337547882592473230/GIF_20241007_095836_385.gif

do this to ur pc

[idle dune]

Well fuck it works I just skipped through the video. I would still worry about heat. But the last time I had a mini itx gaming pc it was a 7970 GHZ Edition in a Lian Li cube case.

[patent walrus]

i can just sell my PC and get a steam machine

[olive sable]

[unkempt citrus]

https://www.sfftime.com/product-page/u-itx-black This one is designed for 3slot GPUs

but its more vertical

[real sierra]

steam machine makes me think of the gamecube

i still have my gamecube

[unkempt citrus]

Thats why people call it the gabecube

[quick condor]

If I end up getting the job I'm applying to, ima go ahead and buy 2 rtx pro 6000 Blackwells

[idle dune]

GabeCube

[patent walrus]

a small form factor tower is a neat idea

[real sierra]

very smart marketing on their part tbh

evoking the right vibes

for a malexample, see "XBox Series X"

[idle dune]

Lemme find the case there also mini itx pillar cases

[patent walrus]

finding a motherboard that'd work with my specs (also ideally with two m.2 slots) gonna be interesting

pcpartpicker time

[idle dune]

https://www.fractal-design.com/app/uploads/2024/06/Mood_Feature_06_Support_for_Cooling-1080x764.jpg

Fracal Design Mood

All the parts hang down

https://www.fractal-design.com/app/uploads/2024/06/Mood_Black_02-810x573.jpg

[true hemlock]

tf did y'all do to make vani want a sff build

[unkempt citrus]

https://pcpartpicker.com/product/9hqNnQ/asrock-b850i-lightning-wifi-mini-itx-am5-motherboard-b850i-lightning-wifi

[unkempt citrus]

The steam machine

[olive sable]

mini itx with 2 m.2 slots are kinda rare, goodluck 🙏

[unkempt citrus]

This one has 2

[idle dune]

Yeah in Mini Itx your lucky to have one

[quick condor]

Not enough cooling for me

[patent walrus]

okay not just the steam machine, im goin back and forth between family and relatives and i just want a pc that i can pick up a bit easier than my giant ass case

[true hemlock]

can i introduce you to the

uhhh

[idle dune]

Why not just get a SteamDeck

Or a Legion Go

[olive sable]

oh the other one is on the back

[opaque wharf]

Power

As in computing power

[patent walrus]

if i were primarily focused on portability, sure - i'd just main my macbook, which i have done before, i just need windows for the times i want to game or use some programs

[olive sable]

vani already has a 4090 and cpu and everything
the goal is just to make the volume smaller

[opaque wharf]

How demanding is the program and games?

[true hemlock]

i was about to pull out my briefcase build till i remembered my back hurts carrying that shit everyday

[patent walrus]

i could really just get a smaller case

[opaque wharf]

Because you may be able to get away with handheld and a dock

[quick condor]

I really need a 25k pay bump so I can not budget and still buy 16k of gpus

[patent walrus]

i already have an LCD steam deck, a handheld is not for me unfortunately

[simple sky]

https://youtu.be/ViPNHMSUcog?si=5Vm3peQG_5obU2Or

so lambda calculus is just math written in a fancy way to make it a programming language?

[olive sable]

man stop flexing how much you earn, this is uncalled for

[true hemlock]

i brought it to hotel, though mine is 15kg in weight but you probably could build a smaller one

[opaque wharf]

Yeah, but what I meant is not use it as a handheld but instead use it as portable PC by just bringing the dock and mouse and keyboard

[true hemlock]

inb4 bro paid $16k for a bunch of rtx 4000 ada sff

[idle dune]

Ah got it. Then yeah grabbing a mini itx or micro atx compact case would be better. Though imo as good as mini itx looks. Thermals are always an issue. I’d go for a Micro ATX system and fine an appropriate LAN style case.

[quick condor]

I mean that's more saying how underpaid I am rn

[opaque wharf]

Is a concept, not a programming language

[patent walrus]

i'd still be limited by processor, memory, and the graphics chip
i like playing games like metro exodus enhanced edition, or maybe minecraft that can meet my monitor's refresh rate

[quick condor]

I've only bought 1 rtx a4000, no adas

[olive sable]

if the first thing you want to spend your money on is gpus instead of a house or a car, then you make enough

[opaque wharf]

Ahh, yep. Not handheld then

[simple sky]

Yeah but it looks like a programming language

[true hemlock]

"underpaid" yeah sure bud

[quick condor]

I have an audi s6 and a tuned A6 paid off

[unkempt citrus]

You're assuming they're nott in debt

[idle dune]

Difference in motherboard sizes for comparison. Mini OTX is the smallest, Micro is the second smallest. You’d get more utility out of a Micro ATX.

[olive sable]

imagine having a car

[quick condor]

For 9 years in the industry it's legit that I'm underpaid

[opaque wharf]

Some programming languages indeed are designed with lambda calculus in mind

[true hemlock]

what exactly are you even doing?

[simple sky]

Like, it looks like I can add more friendly terms for different things, add some syntax to help with organizing, and make a programming language out of lambda calculus.

Holy shit is this how programming languages are made?

[patent walrus]

micro ATX does sound like a good possible route to go, i'll just look around for a motherboard that fits my specs

[unkempt citrus]

micro atx mobos are really easy to get

[opaque wharf]

Something to do with crack judging by the name

[unkempt citrus]

at that point you're just looking for a case you like

[opaque wharf]

Yep

[quick condor]

Datacenter operations, position I'm looking at moving to is an Ai compute dc

[true hemlock]

high end chipset exists for micro atx i think

[true hemlock]

mf didn't do much and expected more pay

💀

[patent walrus]

i guess a challenge is finding one (micro atx) that has two m.2 slots
i know the one rare possum linked earlier has two but that's mini itx...

[true hemlock]

most matx has at least 2 m.2 slot

[quick condor]

?

[unkempt citrus]

2 m.2 in micro atx isn't rare

[true hemlock]

some even has 3

[unkempt citrus]

my own which I picked out at random has 2

[true hemlock]

nah, literally every single matx has at least 2

one with direct pcie traces to the cpu

the other through southbridge/chipset

[patent walrus]

oh, neat https://pcpartpicker.com/product/Tdbypg/asrock-b850m-x-wifi-r20-micro-atx-am5-motherboard-b850m-x-wifi-r20

[true hemlock]

what cpu are you using again

[patent walrus]

9950x

[unkempt citrus]

_ _

[olive sable]


🔺

[true hemlock]

hm, wouldn't recommend

[patent walrus]

this guy putting OBJ in chat

[idle dune]

Not this exact case. But since you said you travel a lot. Some manufacturers make LAN Party focused cases either carry handles and shock absorbers to keep parts safe.

[true hemlock]

that board only has like 8 phases

[olive sable]

misinput, idk what happened there

[true hemlock]

with not that good of vrm heatsink

[true hemlock]

the top side doesn't even have any heatsink on it 💀

it could work still but

[opaque wharf]

So if I fight it, it can transform? Cool

[true hemlock]

vrm throttling was ass to deal with

@patent walrus do you need extra x16 slot

[patent walrus]

never used an extra slot, just need one for my GPU

[true hemlock]

aight

https://pcpartpicker.com/product/c3FCmG/asus-tuf-gaming-b850m-plus-wifi-micro-atx-am5-motherboard-tuf-gaming-b850m-plus-wifi

this one seem to have plenty of 60A phases and more adequate heatsink for them

3 m.2 slot i think

its about $195 in here rn

230w shouldn't vrm throttle at all with this

[unkempt citrus]

You could also go down from 850 to like 650

[true hemlock]

that, the chipset isn't much different

[true hemlock]

14+2+1 60A phases

that'll work nicely with the 9950x

[true hemlock]

couldn't find anything better than this for now

https://pcpartpicker.com/product/Wjwypg/asus-rog-crosshair-x670e-gene-micro-atx-am5-motherboard-rog-crosshair-x670e-gene
if you wanna ball but i think this one is pricy asf

[simple sky]

You know, part of me is screaming at myself to drop the course and take something easier, but another part of me is intrigued by lambda calculus now, and I kinda want to see how it turns into a functional programming language

Damn, I wanted my last semester to be easy

[olive sable]

just do it and see if its fine

and if its too hard change

[true hemlock]

though I don't see a point with high end chipset for a matx

[simple sky]

It’d be too late to change by that point, and I plan to graduate in the spring so no more semesters for me

[true hemlock]

but it might have T chain memory traces which means 2dpc would work better

ah, vani is using 2x32

wouldn't need high end board

just proper vrm phases

for that beefy cpu

[patent walrus]

so long as my vrm doesnt go through an emo phase im okay

[real sierra]

unity

why can you not start UI elements as inactive

[olive sable]

https://tenor.com/view/sne4kyunityhater-gif-9062254570736238262

[real sierra]

all UI elements MUST be visible at the start of the game

you CANNOT change this

the only workaround is to start the UI offscreen, and then at runtime make it inactive and move it onscreen

💔

[unkempt citrus]

What if we made an invisible ui element that makes the rest of the ui elemtns

[true hemlock]

if you're sharp you can probably find a case that can fit the matx board and also have vertical gpu mount

that'd probably be the best case to make it somewhat portable

[olive sable]

i jsut wanted to get gpu temp readings, but it seems like adding that to the nix config made it necesary to compile a lot of stuff??

its also using half my ram now

[patent walrus]

TEEEEAAAAAAAAAAAAAAAAAAAAAAAMS

[olive sable]

its compiling blender from source

[real sierra]

[jagged turtle]

Someone at OpenAI decided to let GPT write their docs (I mean of course they did)

[opaque wharf]

"ensured to be in valid JSON" and then proceed to write exception

[jagged turtle]

Might as well say "this toggle increases the chance you'll get valid JSON"

[jagged turtle]

oh also they aren't even saying when those edge cases happen, despite saying "see below"

[opaque wharf]

Indeed they don't lol

[olive sable]

Entire pc froze

[opaque wharf]

Recompiling a big software tends to do that, yeah

[olive sable]

I need it to not freeze tho

It stopped doing anything, includding the compiling

[glass jetty]

nix moment

[glass jetty]

linux ran-out-of-memory moment

[olive sable]

I dont think it was out of ram tho

[glass jetty]

imagine having 8gb

[olive sable]

I have 64gb of ram tho

[glass jetty]

99% of the time linux freeze because of memory

1% is gnome while i used it

for me

[glass jetty]

but you also run compilation of several units at the same time on 16 cores

i failed to compile qt6.4 in parallel with 16gb and 4 core btw

poor me

[olive sable]

it actually might have been ram ye. reattempting immediatly went to 51GB used

do i not have swap on here?

[glass jetty]

no, run with less jobs

try only 8

[glass jetty]

adding swap on already marked up disk... idk how even do that

[olive sable]

marked up disk?

[glass jetty]

how to call completed gpt table?

if you left some memory unused them you can for sure, might be even without restart

[glass jetty]

i mean i used whole disk on system back then, so i did not even thought about it

[opaque wharf]

Frozen system is almost always OOM. 100% CPU will maybe froze GUI, but not to the point of not resuming after a while

[glass jetty]

happy birthday

[flint dome]

one time my testing software for my python class somehow ran out of memory (literally just a safe browser) and put my fedora into a kernel panic

[opaque wharf]

Thanks a lot!

[flint dome]

how they effed up this bad I still do not know to this day

[opaque wharf]

OOM escape is hard

[flint dome]

the university testing platform for comp sci sucks anyways the thing can't handle files or libraries properly and is basically useless for anything beyond python, like for java

speaks volumes about how the code would be written in the back

"oh your internet sucks ass? you must be cheating, get fricked" immediately disconnects you and gives an F grade

[glass jetty]

wtf
kernel panic on memory outage? is it just program that use kernel space?

[flint dome]

like my bad bro I can't do anything if the university wifi is just plain dogshit

[glass jetty]

complain about it

for real

[flint dome]

idk I was just using the program they gave for testing and suddenly I couldn't type into it anymore cuz screen froze

I look at the caps lock key and BAM its blinking

I guessed it somehow ran out of my fucking ram (i have 16gb)

[glass jetty]

OOM cannot cause panic in userspace

panic is kernel space only (as far as i know)

no, it cannot cause panic in userspace

[flint dome]

then how the hell did it panic from just a safe exam browser application

[azure lynx]

did it require a kernel module to prevent cheating? ;/ coz that's a dumb idea. i bet it gave you a "potentially unwanted program".

[glass jetty]

are you really sure it was panic, and not just out of memory?

[flint dome]

the teacher was even worse, phd in bloody cybersecurity and didn't know what linux is

[azure lynx]

funny.

[flint dome]

my laptop has its caps lock light blink when it has a kernel panic when i tested it with a user induced kernel panic using the shortcut key

so im sure

[flint dome]

im guessing whoever wrote the browser electron shit probably did that

[glass jetty]

what?

user induced kernel panic using the shortcut key

[flint dome]

there was a shortcut key you can use to induce a kernel panic for debugging purposes

[glass jetty]

electron does not use any kernel thing

[flint dome]

I forgot what it was

[flint dome]

Oh ok

[azure lynx]

did you have to give it system privs to install the app? if not, it's probably not the app.

[glass jetty]

^ and therefore cannot cause panic of the system

[flint dome]

also has to be enabled in Linux personally using a flag (again,i forgot which one, I don't have the greatest memory)

[flint dome]

let me look it up again if I did

If it was "ran out of memory" then it probably could clear it because my fedora has done that plenty of times when I wrote dogshit python code that overflowed when writing some rendering program

[glass jetty]

no any guarantee

one time i left pc for 2 hours and had to reset with unsaved code

[opaque wharf]

That's the thing tho. It CAN be. Electron is just like any other native desktop program. When installing it, the installer could ask you to install some kernel module and then communicate with it via some IPC

[azure lynx]

it could also be something in the app tickled a bug in some driver you use.

[flint dome]

it was just standard sudo dnf install blahblahblah.rpm

[azure lynx]

if you hadn't used the app it wouldn't have crashed, but something else might have crashed it

sudo means it can install anything

[flint dome]

that sounds like the most reasonable explanation to me

[glass jetty]

js program cannot, electron itself can, but i still dont believe in panic, there no way electron use kernel module (i mean that they did something such bad)

[flint dome]

so yes, was possible kernel based as well

[olive sable]

i got sidetracked from the gpu temperature stuff and now im trying to make blender work wih cuda and optix instead of cpu rendering

[opaque wharf]

And we were talking about electron no? I am just saying that it could be the case that an electron app requiring a kernel module

[opaque wharf]

Classic sam

[azure lynx]

this are ways. but it's uncommon. but it has been used for anti-cheating in games so maybe it's being used to install a process monitor to make sure you aren't using chatgpt on the same computer or something

[opaque wharf]

Its probably sysreq keys (print screen on the keyboard)

[olive sable]

[azure lynx]

and maybe that process has a bug

[opaque wharf]

I usually use the sysreq + REISUB

But it could indeed be used to induce a panic IIRC

[flint dome]

understandable, would be nicer to do literally anything with cuda because it speeds my stuff up so much (I was shocked when I set my opengl rendered to cuda accelerated on my code, great performance gains)

[flint dome]

yes

[glass jetty]

no point (chatbot)
if teacher cannot detect gpt code himself, that is already bad

side topic

[olive sable]

opengl has cuda settings?

[opaque wharf]

Well, it could very well be a 3rd party vendor solution. The uni just buy stuff not develop their own

[flint dome]

the idea of the program was mainly to prevent alt tabbing into chatgpt during test but the stupid thing failed me instead

[flint dome]

opengl cuda interop

[glass jetty]

chat got

is
the

worst

thing

to use

for code

[olive sable]

[flint dome]

it is, its just poorly written code marketed very well towards administration of a lot of universities, including the best in the country to whatever mine is

[glass jetty]

i lost 90% of performance because i believed that a function was amortized, just used without checking docs at all.
but it actually use 3 syscall and read a couple of chunks of data on every call,
even when given args are exactly the same.
and that function was called 3 times every 'frame', up to 300 fps
and that wasn't only the case, it is what i will say to anyone - why i will never trust ai in anything

[olive sable]

🔺 🔥

[flint dome]

when you're a kid who refuses to think for even basic python code then even chatgpt is a god to you (even if a false one)

[flint dome]

truly one of the statements of all time

[olive sable]

me when i refuse to think

[flint dome]

the only think I ask claude is for "hey mr clanka how can I learn x thing recommend me resources" and its good for that

[glass jetty]

nah, that also bad idea, because of outdated or unrelated things

better to ask ppl for question of this kind

[azure lynx]

models love to hallucinate helpful features. you're lucky it ran at all.

[glass jetty]

chatbots useful only for old things, like last 5 times i used it was for modding civ6, just ask what i want, he give wrong answers, but at least i know what is names to look at in gamefiles

[olive sable]

motherfucker its rebuilding blender again

[flint dome]

can't really get old for basic electronics concepts and embedded stuff and math

[glass jetty]

nah, i coded myself and then asked for clarification

[olive sable]

i jsut wanted optix to work too
now i need to wait 10 min again

[glass jetty]

it said you not need to call if(single_call){func()}
so literally just ruined working code

[worthy niche]

i made a custom tts program in shortcuts to read out NN with each person having their own voice.
i have to manually give each user a voice, what have I done

[glass jetty]

let me find exact function

[flint dome]

https://tenor.com/view/sweet-liberty-helldivers-helldivers2-automaton-gaming-gif-16800556279554128270

[opaque wharf]

You've just created a masterpiece

[worthy niche]

the script is held together by duct tape and prayers but it works

[glass jetty]

btw it is industrial software and environment the same, so we do not check on is file exist or anything

[azure lynx]

the 2nd "loaded = false;" should be "loaded = true;"

[glass jetty]

i removed some and pasted it from above lul

removed static bool but not false->true

[amber fractal]

The only thing AI is used for here is "hey can I do X using the numpy libary"

[quick condor]

NGL just watched a bit more of the ellie Vedal stream and there is so much of it I relate to (for instance I remember the battery muel job for robotics competitions)

[covert bane]

#1428980640895733820 can you guys give your opinion on this?

[olive sable]

cool

[azure lynx]

the roadmap doesn't say what the module will be able to do.

functionality is important. knowing what it is too.

[quick condor]

Side note: robot c is cursed

[covert bane]

Which one? Or did you mean what the integration will do?

[azure lynx]

the thing you linked.

[covert bane]

What do you mean by modules

[azure lynx]

I don't know. what is the expected output of the project?

[covert bane]

Ohhhhh

[azure lynx]

it's not clear what the goal is or what problem you are solving,

"by the end of the project we will have XYZ which will make it easier to ABC".

[amber fractal]

I could use some tidying up of my readme for the NN project, but that is for after it remotely functions first.

[jagged turtle]

[covert bane]

The goal was to give neuro access to windows. With the actions preferrably be something like close window, open roblox, etc

There will be two modes, a low level mode, and a high abstraction mode

The low level mode is what we're doing atm, mouse and keyboard controls

[amber fractal]

I have wrote 3 lines of incomplete code in 30 minutes

[azure lynx]

and all of that will be for the milestone on the roadmap you linked?

[olive sable]

i worte 4 lines of nix config in the last 4 hours

[azure lynx]

coz putting what your milestones are is pretty helpful a roadmap so you can tell when you reach them

[jagged turtle]

I lost my mind at python for the last 2 hours

[covert bane]

Kinda yes, not all. But most

[olive sable]

(blender.override { 
    cudaSupport = true; 
})
cudaPackages.cudatoolkit

[covert bane]

I did put a milestone, the dev builds have what we'll add: #1428980640895733820 message

There was a roadmap higher up

[covert bane]

Idk how to ask this right, but are you okay?

[jagged turtle]

no

[covert bane]

Why is everyone changing their names to weird things?

[amber fractal]

You know what, I'd rather clean that up then attempt working on more singular lines of code

[jagged turtle]

Can I borrow the s in your name

[amber fractal]

my brain hurty from all of the thinking

[covert bane]

yea... but why..? Is this a new years resolution thing of the swarm?

Its been weird since the new years / October

[jagged turtle]

thanks for the s

[covert bane]

Started at october

[amber fractal]

[flint dome]

I'm stealing the c in your name

[covert bane]

Uhmmmmm, okay...

[jagged turtle]

@covert bane now remove one s from your name

[covert bane]

Sharing is caring

[flint dome]

now get rid of it from your name

[jagged turtle]

whenever someone asks, ALWAYS ask them for a letter from their name immediately

[amber fractal]

@covert bane I think you have your answer to why everyone's names are messed up

[jagged turtle]

very funny I assure you

[jagged turtle]

the more people that do it the funnier it is

[amber fractal]

public execution

[covert bane]

Like this?

[flint dome]

https://cdn.discordapp.com/attachments/1059569601144442911/1460182550864597209/weeklyhangings.gif

[azure lynx]

so it's not actually very funny. 😛

[jagged turtle]

exactly /j

[jagged turtle]

according to you

[covert bane]

Why are we doing this exactly...?

[amber fractal]

Just another thing #programming is doing

[flint dome]

has someone stolen from shiro yet

[covert bane]

But why...?

When will this end? I'm so confused

[flint dome]

because like an engineer says "because we bloody well just can"

[amber fractal]

Nope, I think @jagged turtle still has an i from Vani

[covert bane]

Who started it, and why

[jagged turtle]

I think so

[covert bane]

And can we make new letters back?

[jagged turtle]

no

[flint dome]

alright im going to rob one from shiro next time I spot

[jagged turtle]

[amber fractal]

good luck

[jagged turtle]

I tried asking nicely and got a no so good luck, guards up

[flint dome]

ill make an offer they simply can not refuse

[orchid crest]

England on the 18th be like

[amber fractal]

It sorta just spawned for what I understand

[flint dome]

nice pfp

[amber fractal]

someone just deleted letters from their name and the rest is history

[jagged turtle]

the reason why was because half of chayleaf was missing and someone gave letters

[orchid crest]

Buraña

[azure lynx]

ik ram (nau) stey <= "yak ruminates"

[jagged turtle]

I somehow never noticed that

[flint dome]

https://tenor.com/view/neco-arc-neco-arc-fate-cat-gif-1748062773671823114

[orchid crest]

https://tenor.com/view/neco-arc-gif-26529312

Oh I forgot that

well time for the slander

[opaque wharf]

The same reason we once have bime

[flint dome]

unfortunate embed fail, laugh at this user

FUCKING autocorrect i cant

[amber fractal]

minor typo

[orchid crest]

insert nuke meme

[opaque wharf]

https://mlntcandy.github.io/bime/

[flint dome]

https://tenor.com/view/minor-spelling-mistake-major-spelling-mistake-spelling-mistake-gif-25140477

[orchid crest]

I shall post gifts until you give me embel rights!!!

[flint dome]

is this the temporal equivalent of the library of babel

as in that it is in the same vein as it

[orchid crest]

context?

[opaque wharf]

Nah, one day we just declared that base 16 for time keeping is good

[orchid crest]

bruh

[flint dome]

the library of babel website is based off a short story by the same name which aims to have every book that can be written ever stored on it

thing is, it just combines any character anywhere in any combo so most of it is just randomly written things

[orchid crest]

I see

so a lot of bullshot and there in the middle of nowhere you find the bible or something

[flint dome]

yes

[opaque wharf]

Yes

There is also a search function

It works because it is generated algorithmically

[mighty bane]

Do we know tutel's hardware specs?

[opaque wharf]

At least whatever is he installed during last pc building stream

[orchid crest]

I just realize, this not the monky writters paradox but with the randomess being on the word and not the characters?

I guess the story was written before the paradox

but funny two folk created sort of the same paradox

[mighty bane]

I don't remember the details

[opaque wharf]

Nah, library of babel is not a paradox because it is created algorithmically for that exact purpose

[opaque wharf]

Neither do I. But try to check the youtube VODs. I am sure it was on their channel

[azure lynx]

most of the books which match your string will still be mostly nonsense, just with your string in the middle of it. so it wouldn't be a helpful search tool.

[orchid crest]

tbf the paradox itself arrives at sort of the same conclusion (thought that deals more with infinity than anything else)

[azure lynx]

and if the books are infinitely long, you'll get an infinite number of matches for any string.

[orchid crest]

probably you would need to know the entiry text to reliably find the book

as literally you can two books that their only difference is like, one uses inquiry instead of enquiry

or something like that

[opaque wharf]

Yes

[azure lynx]

there will be an infinite number of one letter changed versions.

[orchid crest]

the funniest one are really the inquiry-enquiry question where you in all practical terms have a copy of the same book

but because of being multiple ways to write a word you have like an enormous amount of books that are the same

[azure lynx]

or there is a different 12 letter nonsense word after "THE END."

[orchid crest]

"Your mother wrote this"

[potent birch]

is this normal?

i made a shortcut now after exporting .. so you cant really move the exe file from its folder, you can only make a shortcut of it

idk why it has the debug thing though

[orchid crest]

whut?

[jagged turtle]

oh god

[orchid crest]

regarding the impresition, I guess this accumulates over time

and its sort of the point of that page?

[opaque wharf]

It's basically another unix epoch style. We develop it to remove all timezone

So we can just make an appointment at day 200, bime A12B

[orchid crest]

an clock for everyone then

[potent birch]

what i said, theres (debug) on the title idk if thats normal. imtrying to export as is like every other game

[opaque wharf]

Yes

[left solstice]

Hi nerd chat I usually use mpv for videos but GPT said that I can check how the bitrate varies throughout the video using vlc. The video is normal h264 at ~2300kbps and plays fine in mpv but turns into a colorful mess in vlc. What am I doing wrong in vlc and also how do I check it's bitrate graph (preferably using a gui program)?

[obsidian mantle]

i once tried to watch some megahigh bitrate video through vlc and it couldnth handle it
had to use mpv

[left solstice]

Nevermind, vlc doesn't show bitrate info like that but it's still borked when playing videos so no idea about that still

[opaque wharf]

Maybe try ffprobe

[left solstice]

Wouldn't ffprobe give me a long wall of text as info + other unrelated info to read and scan through? I just wanted a visual graph of bitrate variance so I can compare videos and different qcomp configurations

[opaque wharf]

Hmmm, I have no idea then

[left solstice]

Happy birthday btw HyperCake

[proven merlin]

You can, you must copy the .p(Something) File aswell

[velvet crest]

You have unlocked new role

[opaque wharf]

I do find this tho
https://github.com/zeroepoch/plotbitrate

[potent birch]

the .p what

[opaque wharf]

Thanks

[proven merlin]

In your export folder, there should be atleast two files by default im pretty sure

[potent birch]

do you think its this?

[proven merlin]

PCK

[potent birch]

its set on debug only earlier

[proven merlin]

Embed PCK I think.

See if there is an pck file near ur current exe.

[opaque wharf]

Yeah, I think this is what you want no? @left solstice

[potent birch]

nope

[proven merlin]

There is

"Dodge_the_creeps_learn.pck"

[proven merlin]

Either Copy that + the exe to run ur game, or enable "embed pck"

[proven merlin]

Thats pck seperate

[potent birch]

lmao im blind

[left solstice]

Thanks, but I also found mediainfo which also seems to do the same thing and I reinstalled windows a couple days ago so I'll have to reinstall python and matplotlib before using that repo. I'll try mediainfo first and then try the repo if it doesn't work NeurOk

[proven merlin]

Here you see binary format -> embed pck
Then build/export as normal

[opaque wharf]

Alright

[potent birch]

idk where that is

[proven merlin]

Mood. I managed often to not find system32 even if im looking straight at it

[proven merlin]

In the image you sent

Lemme circle it

[potent birch]

oh

it was not on

[proven merlin]

Enable, Then just export as normal and then you should just have the .exe

[potent birch]

its still there

[proven merlin]

Try copying just the exe now

To idk, your desktop

And then it should still run

[potent birch]

lemme see

[proven merlin]

(also nice cursor)

[potent birch]

still debug

thanks

here https://ko-fi.com/s/95328bc5e7

[proven merlin]

Yeah, debug vs release is a different thing-

[potent birch]

ill get to it eventually

now time to try the godot 3d tutorial

[proven merlin]

I found a older reddit post, not sure if its accurate anymore

Apparantly when exporting and choosing where to place the exe files

There is a checkbox "export with debug"

https://www.reddit.com/r/godot/s/wUplT4tI5N

For a release build youd want to disable that

This may have changed tho soo

[potent birch]

ohhhh i read it thanks

[proven merlin]

I hope that helped-

[left solstice]

MediaInfo doesn't work with common formats like h264 for some reason so I'll use the repo instead. Thanks

[obsidian mantle]

what the hell? win 12? is this fake

they launched win 11 like a year ago

[left solstice]

Aintnoway they are working on win 12 while win 11 is still the worst os that has ever existed instead of fixing it

[unkempt citrus]

Its talking about bill gates with relation to modern microsoft so yes

[drifting raptor]

never underestimate the ability of Microslop to fuck up 2 OSes at the same time

[left solstice]

I signed up for extended win 10 updates recently and will keep using it as long as I can until I'm forced to switch to linux but I'm never installing a newer severely downgraded version of windows

[fickle rain]

Why would Bill Gates give a shit considering he hasn’t been Microsoft for a while now

Could bring up Ballmer too I guess

https://youtu.be/8fcSviC7cRM

[fleet jungle]

What's crazy to me is that the beautiful UI of vista was still essentially an extension of windows xP, the predecessor. Like, the difference is huge (and shiny)

oh I responded to the wrong message

[hard delta]

what i wonder about is

how and where will win12 push ai

they clearly can't use your regular text input as training data

because that would mean you can't input any personal sensitive data in anything that has text input, making the whole OS useless

if people did anyway it would eventually show up in replies and that would be their PR disaster of the century

so no training data

nobody wants most of the AI features either

but it costs a lotta money

[trim valve]

I'm so confused why are we talking about windows 12

[unkempt citrus]

Because a stupid click bait thumbnail was posted above

[trim valve]

ok but like why are chatters talking about it like it's real news

https://youtube.com/watch?v=LoQkTYA2tUk

the video is literally ai slop

I should make a channel like that just to ragebait #programming

[true hemlock]

this sums up the whole bubble rn

[hard delta]

windows 12 itself is obviously real news lmao

there will eventually be a new windows version

i haven't clicked the clickbait video if that's what you're worried about lmao

[trim valve]

why do I even come here any more

[hard delta]

bro what the fuck is your problem

there has been no official announcement of python 17

it doesn't take serious deduction skills to see that eventually there will be a python 17 anyway

there was no press release confirming chrome 150

yet there will likely be a chrome 150 at some point

[trim valve]

ok but you're talking about it like it's on the doorstep already being installed on your computer

"guys I can't believe python 17 is going to be only vibecodeable, I hate ai and I hate the bubble"

[hard delta]

ok you know what i agree

try nn

[trim valve]

[hard delta]

silly cat

[left solstice]

I reinstalled win10 and while reinstalling python, it added python to the path but not pip. I don't want to use 'python -m pip install' instead and searched the directory for an hour but couldn't find any file named pip.py or pip.exe. Does python 14 just not allow that or something?

[finite ledge]

[fleet jungle]

to tell me when I'm about to have a bad idea

[hard delta]

i've never seen a clause like this

[wary mauve]

anyone know good benchmarks for finding certain models response times (their tokens/s and total time took) for LLM and such

attempting to benchmark how a fully local setup would work for my fella as he takes a while using openrouter right now

can take anywhere from

1s - 20s -> Thought LLM
10s -> 30s+ -> Vision LLM

obviously model dependent so experimenting, if u have any decent models i should do especially for vision thats decent but fast works for me

[fervent jolt]

Not for agentic use? Just for benchmarking?

[olive sable]

Just had photoshop exam

[wary mauve]

Guess a bit of both, need good agentics with good benchmarking, dont just want fastest wins

[olive sable]

Feeling good

[wary mauve]

if thats what you're asking

[olive sable]

[wary mauve]

what does a photoshop exam entell?

whats th exam like

[obsidian mantle]

I use win11 and its fine

[olive sable]

You gotta recreate stuff by the specific instructions

[fervent jolt]

You might want to try the top sota for oss like glm 4.7 or minimax m21

[obsidian mantle]

Soon we will have 1b agi that will take over the world
Relax

[fervent jolt]

Or if you cant handle the requirements, then downgrade to qwen

[olive sable]

Like "install this specific font, use this image as a bg for the font.... and make it non-destructive" but you mostly have to deduce how stuff was done from the example image

[sage crag]

https://cdn.discordapp.com/attachments/1072697081443131476/1459353564764835902/A5D6E55C-09FB-4941-B002-2F28164FCAC3.gif

[frozen hollow]

computation isn't constant time for inference if that's what you were hoping for

[wary mauve]

eventually I want to fine-tune said model so as long as its open source and on the easier side for fine-tuning, but never finetuned in my life and only gave it like 5 mins of research before just realising need to improve a system prompt first, but also my goal for fine-tuning is creatign a better persona to match inputs and outputs of convos i like, primarily from discord and grabbing data is SO aids

[hard delta]

thank

[olive sable]

And there's also just a theory exam where you gotta know the difference between serif, sans-serif, monospace, ...
Some colour theory shit
...

Boring things

[obsidian mantle]

Minimax? Isnt it chess engine type

[frozen hollow]

It's used for programming I'm p sure

[wary mauve]

yeah my question was a bit bad, I dont just want a fast model but I need something thats decent response / latency between, stopped speaking -> thought -> tts

can't use streaming responses either because my output is json to perform actions and so i can expand the response i get with other info later, as i cant settle on a good way to do actions

[obsidian mantle]

Llm programming theough chess engine

[wary mauve]

fair enough, whats it for? certification or

[left solstice]

The msix python installer installed pip but didn't include pip.exe to include in the path so had to force reinstall pip to include it using
python -m pip install --upgrade --force-reinstall pip setuptools wheel

[sage crag]

@owobred

missing

[hard delta]

what the

man people keep shooting themselves with me lately

[stray dragon]

owobred is here

[fleet jungle]

Are we allowed to share rage bait?

I found some what looks like a linkedIn post of what I can only presume to be pure ragebaitium

[jagged turtle]

What's the ragebait...

[fleet jungle]

this

[olive sable]

My bachelor

[wary mauve]

i mean some valid points, microsoft is so riddled with bloat nowadays, plus dont get me started on azure, i dont blame him, but mac is also locked tf down but it does benefit from being unix

but leaving a place is the ragebait part for sure

[jagged turtle]

I mean the windows part is understandable (I don't know what the 20-step auth is about though)

the rest of it is kinda ragebait

[fleet jungle]

Oh you're seeing the 20-step auth on a company-decided-what-you-as-the-worker-can-do basis? Then, I guess it's not as ragebait-y

[rough bloom]

20-step auth sounds more like bad internal policies rather than Windows specifically (though MS Azure also absolutely sucks at implementing auth, infinite auth loops or having to auth 3 times for one action is not that uncommon)

[jagged turtle]

Oh yeah, authing with Azure takes 10 years away from my life

god it's annoying

half the time it just fails

[fleet jungle]

Yeah I've had trouble with sign-in "loops" from MS websites. That is a pain, I don't disagree.

[jagged turtle]

the other half it just redirects into itself

[hard delta]

oh look you didn't you do an escalated / privileged operation (git push, or looking at this repo, or looking at the ci/cd pipeline) in 10 minutes, better use biometric auth and enter this code on your phone oh sorry code expired try again

[fleet jungle]

I kind of saw it as the person whining about windows limiting their software options when no such limitation (other than architecture/OS compatibility) exists that I know of.

[hard delta]

yeah this basically

[jagged turtle]

I had a stroke reading this

[native flower]

I understand only the half of that..

[jagged turtle]

I still don't understand wym btw

[fleet jungle]

He's saying that's what MS says when trying to do certain things within a time limit

[jagged turtle]

I think I'm stupih

I would put evilDentge if it was still in this server

what happened to that emote btw

[olive sable]

[hard delta]

generally big corpos have insane timeouts and "security" in case there's a ninja lizard spy looming in your office waiting for the moment you leave your machine unattended for 30 seconds to steal corporate secrets

[rough bloom]

any actions that require elevated/admin access are commonly configured to require MFA because security™
you then get a token that your application can use to do whatever it wanted to do
that token may expire after 10 minutes, causing you to have to go through MFA again
and because MFA commonly uses time-based codes you can fail to enter a code in time, causing you to have to retry

[rough bloom]

was never in this server I'm pretty sure

[faint sandal]

Microsoft
repeated logins
@ vani

[olive sable]

[fleet jungle]

its ok, youre sleep deprived. I forgive you

[faint sandal]

I just had ghost pepper chicken tendies

my stomach is now complaining an hour later

[fleet jungle]

With ghost peppers? How'd you manage to eat anything with that?

[faint sandal]

spiced with ghost peppers

not even the hottest thing they have on their menu

[jagged turtle]

I try not to be

[faint sandal]

[jagged turtle]

I swear it was but I guess I'm going schizo

[fleet jungle]

Yeah but I imagine having the equivalent of needles punching your tongue and lips as you're eating doesn't let you eat in peace

[jagged turtle]

Ok well I havent really had that type of issue ngl

[faint sandal]

a bit dry tho

[fleet jungle]

Looks nice

[olive sable]

ghost peppers are already dead, imagine how spicey they were when they were alive

[fleet jungle]

but ghost peppers

[faint sandal]

they died from the pepper next level up

the highest is Carolina reaper

at this restaurant anyways

[fleet jungle]

Was this Nando's by any chance?

Tho I guess Nando's doesn't use ghost peppers

[faint sandal]

it's some placed called hotties

[olive sable]

[fleet jungle]

oh

Well, at least you know it stands up to its name

[faint sandal]

I would have liked some sauce tho

[nocturne olive]

Apparently it's 4070Ti level

[rigid snow]

ye or worse than that even

[hard delta]

any day now

[fleet jungle]

what are we waiting on

[hard delta]

my new machine getting shipped

[fleet jungle]

ooh

[hard delta]

so i can comment on the "5070ti a beast" stuff

[olive sable]

damn

[fleet jungle]

ahh

[olive sable]

with who?

[hard delta]

my bedroom

[wary mauve]

can i borrow it

[fleet jungle]

do you have a nuclear power plant to power that beast?

(I'm joking)

[hard delta]

yes it comes in the package

[wary mauve]

tip to becoming a better programmer, listen to this gas

[hard delta]

tips to becoming a calmer programmer, listen to this water
https://www.youtube.com/watch?v=LGC90fmf8gw

[opaque wharf]

Tips to be a calm programmer:

Take up farming, leave tech behind

[wary mauve]

or best of both worlds

https://store.steampowered.com/app/2060160/The_Farmer_Was_Replaced/

Steam

The Farmer Was Replaced

Program a drone using a simple python-like language to fully automate various farming tasks that would otherwise be very grindy. Feel the satisfaction of simply pressing "execute" and watching your drone do all the hard work.

Unlike most programming games the game isn't divided into distinct levels that you have to complete but features a cont…

Price

$9.99

Recommendations

6890

😏

[opaque wharf]

Abandon tech, embrace stone age

[hard delta]

abandon stone age, embrace bronze age (bronze sword cut down stone spear barbarian, strong empire)

[nocturne olive]

Someone make this Xorg compatible

[nocturne olive]

By the way I recommend passing "stream":true into your request so you can get tokens as they're generated rather than waiting the whole generation time, that will reduce latency to start of response by a ton

[wary mauve]

Issue is my response is pure JSON, so streaming wont help as i cant take any of that data and pipe it to tts

but im just going through open source "neuro" style clones and just researching their prompts and responses and how they handle things like

internal thoughts
mcp / tooling
speech

etc

and going to refactor a bit

[nocturne olive]

You can accumulate streamed tokens and send it to the TTS in chunks

[wary mauve]

but v1 of my son is working so happy with what ive got rn even if its h igh latency

[hard delta]

he's a little slow

[nocturne olive]

Split at sentence boundaries

[wary mauve]

issue is my output is currently

{ "speech": "what i want to say ", "actions": [] }

so i could wait until theres some speech and do some custom parsing, theres always a solution

[potent birch]

son? you making ur own ai?

[nocturne olive]

If you're using an OpenAI-compatible API for the LLM it'll also stream the actions

[wary mauve]

well making my own "neuro" sama like clone, but for a server im in on discord so we can all just have him there

[potent birch]

good luck!

[wary mauve]

wdym?

[nocturne olive]

If you use the tool call system of the OpenAI API it'll stream the tool calls

[wary mauve]

oh no i dont use the tool call system

[nocturne olive]

It'll even tell you when the streamed chunk contains a tool call

[nocturne olive]

You should, it's kinda great compared to manually handling tool calls

[wary mauve]

because either it responds with a tool call or it responds with speehc, and i didnt want to do another LLM call to get speech back based on an action, i only do another llm call if that action it requested requires data that is needed to perform another output to said request with the extra data / knowledge

[nocturne olive]

If you want latency you'll either have to parse streamed tool calls because full LLM completions being waited for will never have good latency, or use the OpenAI tool call system

[wary mauve]

well my goal is to possibly switch to mcp later as thats what ive seen used more

i just need to rethink my architecture, its a modular system, where every "turn" / thought, the "core" system asks every module for their prompt injection, so i can have dynamic prompts based on changing data

but the output system is my main gripe rn

[nocturne olive]

Funny thingy that does tool calls and uses streaming

[wary mauve]

my current brain module, the think function is the main thing that gets called when it recieves some USER_INPUT

[nocturne olive]

It sends events to the terminal when reasoning begins, when a tool call begins, when a tool call ends, and when reasoning ends, as they happen, rather than parsing afterwards

[nocturne olive]

Why have such a big thing for thinking rather than just letting the LLM use the reasoning system?

[wary mauve]

My only gripe with tools is I need to send another LLM always if theres a tool to run, where as my solution right now allows me to have something to say and run a tool with one LLM call, I only use another LLM call if that specific tool requires data to be parsed to respondn to the users

[wary mauve]

New to using LLMs as a whole so not got it all down yet, I assume the LLM reasoning system is built into certain models? and it just iterates its thoughts until it lands on the desired outcome

unsure

[nocturne olive]

Yes, and when using the chat mode of OpenAI API in streaming mode it'll tell you what tokens are part of reasoning and what the final answer

[wary mauve]

My issue with streaming is if its a specialized output I still need to wait until a parser i write can start interpretting speech annd sending it off to the TTS which does save me some latecy for sure, ill get around to doing that at some point as itll feel better, just seems a bit messy but i guess llms as a whole is messy

so

[nocturne olive]

The more of the built-in systems of the API you use, the less manual parsing you need to do

[wary mauve]

Well I m using openrouter right now, which i assume is "based" around openai api spec but unsure if theres many differences

[nocturne olive]

If it follows OpenAI's API conventions (like using the same endpoint naming and style) it's probably compatible

[sage crag]

https://tenor.com/view/borzoi-snoopa-lord-foog-dog-funny-gif-20125052

[olive sable]

hi konii

[sage crag]

wrr

[olive sable]

wrrrr

[sage crag]

https://cdn.discordapp.com/emojis/1154436961277313126.webp?size=48&animated=true&name=evilSlam

[olive sable]

[sage crag]

https://cdn.discordapp.com/emojis/1181502825440747530.webp?size=48&animated=true&name=hiyoriNya

[fleet jungle]

Fellas is it possible for one person to build up their own firmware for their modern hardware in a reasonable timeframe? I want to make Vista work with hardware from the era of Windows 10

[olive sable]

uh

[sage crag]

https://cdn.discordapp.com/emojis/1460031442292703556.webp?size=48&animated=true&name=autismu

[olive sable]

depends on what you mean with "reasonable timeframe" and "work"

[fleet jungle]

(I miss the UI )

[olive sable]

cuz some stuff wil work, some stuff will work

[fleet jungle]

In the sense tht it wouldn't require manhours of a small/medium team

I guess not since modern hardware engineering is pretty advanced

[olive sable]

i think your best bet would be to look at people already attemprting to do this online

instead of doing it yourself

[fleet jungle]

Yea

Good call. Thanks

This image contains unending sadness

[rough bloom]

not really I think
firmware is typically closed-source blobs, often also signed
so you may not be able to replace it at all
you should in theory be able to write drivers for Vista that just load the newer blobs
could base it on Linux drivers, otherwise you may have to reverse-engineer it yourself

[fleet jungle]

Ooh, thanks for that insight. I was overlooking the limitations of hardware signatures, and the thought never occurred to me about having a MiTM to interface the OS.

[hard delta]

mlem in the middle

https://tenor.com/view/neurosama-evil-neuro-melba-toast-licka-gif-10148052058494061158

i'm sorry i'll leave now

[olive sable]

[rough bloom]

if the issue is just that the drivers are too new for Vista then you wouldn't need to intercept/MITM anything I think
you'd have to either rewrite or patch them, idk how difficult either of those options are, I don't write Windows drivers

[hard delta]

also heavily depends on whether you've written one before or not i suppose

[fleet jungle]

Hmm

[fleet jungle]

or I could hexedit them then I die

[rough bloom]

ye that would be the "patching" option

but idk how much Windows device driver APIs have changed between Vista and Win10

[fleet jungle]

It'd be best to avoid that tho. Lots of headache involved, and I know this because I was trying to work with a DOS BIOS

[fleet jungle]

I'd imagine its gotten heaps more complicated

Finding out how exactly CPUs work was cool though, with register, instruction, and data referencing and all that it entails

[rough bloom]

oh it'll be a headache either way

[fleet jungle]

yep

[rough bloom]

I guess depending on the device you could do a user-space driver too
might be easier

[fleet jungle]

hmmmm

True, but with trying to install vista I need the drivers ready at install time 🙌

I'll just keep this idea in mind. For now I'll put it on my ideas backburner

[rough bloom]

can plan more if you know the exact devices you want to write drivers for

[olive sable]

im sure theres a windows vista skin you could put on a more recent version

[faint sandal]

I tried to vibecode a filter driver today

did not go well

[frozen hollow]

A filter?

[fleet jungle]

Idk about vista in particular, but I do recall seeing something about getting aero theme for windows 10 (or was it 11). Paid tho

[rough bloom]

yeah I wouldn't let an LLM anywhere near that
at least not near any design decisions

[faint sandal]

filter driver is a type of hw/fs driver on windows

to be fair I only needed it to spoof device io queries in a controlled vm

[fleet jungle]

Oh wow. This screenshot made me forget that Windows 10 follows a flat no-fancy-shmancy design language

Wait that just looks like windows 7 nvm

sourced it from here https://www.reddit.com/r/Windows10/comments/1i6qgjq/windows_10_with_aero_glass_makes_it_like_the/ lol

[hard delta]

so i was pondering on buying a new PC and narrowed it down to 2 options:
one of them cost 720 neuro coins, the other cost 900 neuro coins
i ordered, paid, it's now being shipped
the shop changed prices since, the same builds now cost 900 and 1100 neuro coins respectively

what a dodge

[fleet jungle]

nice

[karmic birch]

landscapes best wallpapers on win7

[hard delta]

anyone remember this masterpiece

[olive sable]

nope

[sage crag]

https://cdn.discordapp.com/emojis/1154436961277313126.webp?size=48&animated=true&name=evilSlam

[olive sable]

season 5 of the immortal king is releasing

[proven merlin]

Only the storage driver, snd nvme drivers exist..

[fleet jungle]

I don't have nvme

[proven merlin]

https://github.com/Maplespe/DWMBlurGlass
Here u go

[proven merlin]

Then nothing should Stop u from installing vista-

[fleet jungle]

Yea but Idk why the installer doesn't seem to recognize the hardware at installation time. Something odd about it maybe?

[proven merlin]

What Hardware tho

It should only care about storage

[fleet jungle]

Ah, right. It prompts for storage drivers. I have a 1TB Toshiba HDD, which I'm sure is roughly around the same age as the era of Vista, or maybe slightly newer than the OS, but maybe I'm missing something.

[proven merlin]

I assume its not connected via Sata?

You can install win98 on an ssd for all it cares.. the connector protocol (ide, sata, sas, nvme) Matters

[fleet jungle]

No, it is connected via SATA. If it's not the connection, then perhaps something else?

[proven merlin]

Otherwise grab your chipset Driver, or a compatible close one.. that should work.. and idk if vista supported GPT partition table

[fleet jungle]

Eh? I thought Vista was GPT ready, but maybe I assumed wrong. The HDD is labelled GPT

Then again, I think if that were the issue it should have been resolved when I ran FDISK on the installation boot's command prompt

[proven merlin]

Apparantly it can only boot from gpt if youre 64bit Vista and uefo booted

[fleet jungle]

hmm then it shouldn't be an issue, I think. 🤔 Did I format the installation USB wrong maybe?

[proven merlin]

try diskpart instead. Fdisk is old and only for formatting, not partitioning

[fleet jungle]

But then the installation wouldn't even boot

[proven merlin]

It may have booted in legacy bios mode, which could disable gpt Booting..

[fleet jungle]

Oh I knew I was using the wrong name. I was using DISKPART. Sorry for the confusion.

[fleet jungle]

hmmm I'll have to look into that when the PC is available since my bro is using it atm

but maybe that is why

[proven merlin]

Well try this in diskpart

lis dis
sel dis X
convert mbr

Replace X with the correct number ofc

(that is assuming you can wipe the hdd without consequences)

The bios does see the hdd though, right?

Some boards disable some sata ports when specific pcie ports are used

[fleet jungle]

It should be inconsequential. I'm happy to rewrite the entire disk if I need to

[fleet jungle]

Yeah. I assume it recognizes it since when I do LIST DISK it shows up as something capable of 9xxGB storage

close enough to a TB :p

[slender timber]

The guy fixing my hard drive just disappeared

[fleet jungle]

That's not good.

Were you paying them for it?

[slender timber]

Well i haven't paid

[fleet jungle]

Oh

But they have your HDD?

[slender timber]

Ye

My paperweight

[fleet jungle]

oh

They'll probably get back to you (hopefully)

[slender timber]

It is probably unfixable

I just gave it to him to get the data out

Maybe try some magic to fix it

Haven't heard back from him

[fleet jungle]

hmm

Well according to a video I saw for storage drive data recoveries, they use pretty specialized gear. Did this fella run a shop for specifically this?

[slender timber]

Yea

I asked around and most people said he is the best choice

I mean the place looks proper

[simple sky]

I’m enjoying Racket so far. This is a pretty nice language

[wary mauve]

guys am I overthinking the use of "tools" parameter for openai apis, my main gripe with it is that I have to do 2 runs of an api request if it requests a tool, because i have to do

initial llm request, of what the user(s) said in the discord call and just the default system prompt -> LLM respond if it needed to use a tool ( if not then thats just 1 llm ) -> but if it needs to use a tool it wont return any content of what he wanted to say whilst using the tool, so I have to send off another LLM request which essentially give the LLM the contexts of "hey you just this tool / its still running, reply like you're affirming the request to u se a tool or mention youve use d atool" or something similar ( but what if a tool is called because "he" wanted to use it out of the blue? anyway )

My main concern in latency but maybe im being a bit fat dumb idiot baby

[simple sky]

Holy shit, you can define functions locally in Racket

[wary mauve]

sounds like my boy Lua 😏

[hard delta]

maybe they travelled to the secret moon base for the secret world-government technology used there in hopes to fix the drive after failing by themselves

[floral wraith]

keep us updated

[fleet jungle]

Fair enough. I was wondering in case it was just an ordinary PC sales/troubleshooting store, which I don't imagine typically has specialized technology for data recovery.

[sage crag]

https://cdn.discordapp.com/emojis/1154436961277313126.webp?size=48&animated=true&name=evilSlam

[fleet jungle]

I need a sticker of that animation of evil asking to be picked up by vedal

[wary mauve]

I know this isnt answerable to a T and i feel bad asking but i understand Neuros voice is Azure Ashley but would that mean Neuro still uses it as a service or would it be better to train a different tts model but use the voice from azure as data to host one locally?

Its not like its an advanced voice so unsure if thats a good idea, i do love elevennlabs but its os expensive and I want to start locally hosting my own tts but I know i wont get the same quality at all, and all the abilities like sound effects and things like laughs and such

[sage crag]

https://tenor.com/view/30000-usd-gif-13674633028161831364

[fleet jungle]

https://tenor.com/rD5McydU9YJ.gif

[sick owl]

So DLSS 4.5 right

Really cool, except

Why did they not launch it with ray reconstruction support

Its meant for DLSS Performance mode, most people using performance mode are gonna be people with 4k displays

Most people with 4k displays are gonna have high end cards

Ergo most people using DLSS 4.5 are probably going to want to crank settings to max ray tracing and path tracing

[proven merlin]

Yeah then its some partition nonsense. Worst case you use dism to deploy manually.

You can also use a modern windows installer, modded by replacing the install.wim if you dont wanna install windows via the command line

[fleet jungle]

here, take some tunz

[fleet jungle]

Will that work? They're swappable like that?

[proven merlin]

Yeah surprisingly. All that the installer does is partition, and dism deploy

[fleet jungle]

Ah.

[proven merlin]

There is a nice video from enderman "installing windows 10 manually" it essentially works for any modern (NT6+/Vista) windows, you only need to follow till dism finished

He goes and skips OOBE and such aswell, but thats not needed

[fickle rain]

Big fan of autounattend TBH

[proven merlin]

Thats also a solution, unattended iso's

Im personally not a fan of them unless i have the same machine over and over. Otherwise i want to have some control

[patent shard]

it has a lot of issues
and the sharpening filter is pretty strong
it can be really good... but also not so good

look at the shadows from the car

[wary mauve]

is my ass seeing 87 fps on GTA V with a 5090??

WITH DLSS

[patent shard]

it's DLSS running at 2160p and outputting at 2160p
but, yes

[rigid snow]

hello

[burnt marsh]

hello

[tender river]

hello mlntcan

[burnt marsh]

still no response from vedal

expected lol

[rigid snow]

respond

[burnt marsh]

i mean, hes in a different timezone so it makes sense he doesnt check his emails during the night.

[wary mauve]

is he not uk timezone?

gmt

8pm rn

[tender river]

vedal timezone

[burnt marsh]

im in est, and i sent it at like 7 this morning

[fickle rain]

I can only assume there’s a drinking binge involved

[rigid snow]

bee doll

[hard delta]

bro is on holiday after a subathon

he probably won't open emails till february

[burnt marsh]

yknow... i just realized that

makes sense, sorry for my stupidity

[frozen hollow]

Just curious what text to speech models you have looked into so far?

[unkempt citrus]

There are higher quality locally hosted TTS models available if thats what you're asking. Laughs are a bit harder but there are a good few around

Also Neuro is being moved off Ashley once V3 is done, whenever that happens

https://huggingface.co/models?pipeline_tag=text-to-speech&sort=downloads

Have a squiz for models you like

kokoro was pretty popular a while back

[frozen hollow]

Tbh I thought it was kept the way it is because it’s “her” voice

[burnt marsh]

v3 was the friends we made along the way

[unkempt citrus]

It is, V3's biggest problem is that Vedal thinks the audience wants it to sound like Ashley while being as expresive as Evil's

Because Evil's was supposed to be Neuro's after all

[sick owl]

Its only intended for use with DLSS performance mode, he's using it wrong

That's DLAA

[unkempt citrus]

So while he can make a good voice, hes pushing it to be more similar to Ahsley and thats cuasing the delays

[sick owl]

Model K AKA DLSS 4 is still the one you're meant to use for that

[patent shard]

yeah, it might actually hurt the quality by using dlaa
not even just performance

[sick owl]

Its straight up broken in that configuration because the model isn't trained for it

That's why it looks oversharpened too

[patent shard]

nah, it looks oversharpened at any resolution

[sick owl]

DLSS Performance mode from 1080p -> 4K it doesn't look oversharpened, it looks great on a 4k display in that configuration

[patent shard]

it depends on the game/settings, though, if the game is more neutral, it looks okay
but if the game already has a good amount of sharpening.. oof

[sick owl]

Oh yeah if you add a sharpening pass on top, that's not DLSS's fault though

I always disable post process sharpening when I enable DLSS because the model takes care of that

[patent shard]

well, DLSS 4.5 is also adding a pretty significant sharpening filter

[sick owl]

You're just blanket sharpening an already ML sharpened image

[hard delta]

gotta be as sharp as possible

dlss, integrated gpu driver, and screen built-in

[patent shard]

it does fool most people
"it looks sharper"
uh, yeah, it's using a sharpening filter >.>
I'd rather them just make it neutral like dlss 4 was

[nocturne olive]

From what I've heard, he even has something that sounds really similar already, but he's stuck in the "it has to be perfect" trap so V3 is never coming out
Nothing in audio ML is perfect

[unkempt citrus]

Yeha he mentioned showing it to people and many said it was close but people who were more addicted could tell it apart or something

[burnt marsh]

exactly, v3 was the friends we made along the way, its not happening for at least another year

[nocturne olive]

Yeah and doesn't seem like he understands being able to tell it apart isn't necessarily a bad thing

[unkempt citrus]

tbf v2 could be distinguished

and got pushback

which is why its evil's and not enuros

Classic Evil, always getting Neuro's hand me downs

[nocturne olive]

Well V2 was completely just a different voice

[rigid snow]

i hear the similarity

so not completely different

it’s a bit like neuro’s v1 singing but made into an actual real sounding voice

[nocturne olive]

Never thought of it like that
Though probably it'd be based on V2 if that existed back then
It's entirely possible they went the "distil the singing voice" route

[sick owl]

MODS

[stone flint]

@cobalt vessel

wait... he's playing mcsr.... should i ping shiro instead

[sick owl]

I submitted a modmail ticket

[unkempt citrus]

<@&574931772781887488>

[real sierra]

they're dead

just cleanup

[potent geyser]

This again? I guess people actually fall for these

[unkempt citrus]

I have seen those exact 4 images like 5 times

[real sierra]

[real sierra]

yes they've been trying to send them here for 2 years

[patent walrus]

https://tenor.com/view/skeleton-skyrim-gif-12158785412394332359

[unkempt citrus]

https://tenor.com/view/rip-bozo-gif-22294771

[real sierra]

they finally figured out to embed them instead of linking

[rare bridge]

https://tenor.com/view/fnaf-ucn-camera-five-nights-at-freddys-runwithluna-gif-25298485

[nocturne olive]

Dumb scammers being classically stupid

[potent geyser]

Is it possible to setup automod for those image url

[sick owl]

I got a DM from one the other day but it was Elon Musk, this time it was Kai Cenat

[rare bridge]

they grow up so fast

[real sierra]

for image url yes, that's why you don't see them literally every day

but they're stealing accounts with embed perms and uploading them instead

which bypasses our filters

[unkempt citrus]

With image hashing and a custom mod bot you could do it for images too

But might not be as practica

[rigid snow]

1.jpg 2.jpg 3.jpg 4.jpg

[frozen hollow]

I was looking at f5-tts model apparently you can optimise it quite a bit after you’re happy with it

[rigid snow]

why not me ban

[sick owl]

Remember chat, always use a 512 bit password

Its not overkill if it works

[azure lynx]

perhaps require 2fa before embed permissions?

[nocturne olive]

Should totally train a CNN model specifically to classify if an image is or isn't one of those images

[real sierra]

people get compromised even with 2fa

[unkempt citrus]

CNN is way overkkill

[real sierra]

token stealers don't care about your auth

[unkempt citrus]

Token logins bypass all authentication checks

[nocturne olive]

It would also be really funny

[unkempt citrus]

Genius design from discord

[nocturne olive]

Unfortunately tokens aren't per-IP

[sick owl]

A simple similarity score would work way better tbh

[nocturne olive]

Though I guess with regular dynamic IPs those could expire often

[sick owl]

I doubt they're smart enough to figure out what's going on and adapt, its probably automated

[unkempt citrus]

It 100% is

[frozen hollow]

But what if it’s actually Elon musk

[real sierra]

even tokens being per country would fix a lot of issues

[frozen hollow]

I don’t know who Kai cenat is ngl

[unkempt citrus]

Hes the biggest streamer atm

well he hasn't streamed in 3 months but otherwise

[real sierra]

just like our streamer

[unkempt citrus]

Its proof vedal made it

[sick owl]

As always, the best solution is to not be gullible

[frozen hollow]

2 things - apparently he’s in good burger 2 and also there’s a good burger 2 movie wtf

[rigid snow]

good burger 2 sounds like a movie gpt 3.5 turbo would hallucinate

what a name

[sick owl]

You know, I've found a silver lining in the recent hardware shortage

We're probably not going to have to spend money on a system upgrade for a lot longer now

For those of us that already have good PCs bought at pre-shortage prices

[unkempt citrus]

I was talking to a friend who was after an upgrade

at this point, prebuillts are really good value compared to doing it yourself

cos some places haven't raised prices yet to match

[hard delta]

are tokens invalidated by logout tho 🤔

[unkempt citrus]

No

Tokens are only invalidated by password change

[hard delta]

(that's kind of stupid on discord's part lmao)

[potent geyser]

Isn't the latest Nvidia GPU 3060

Nvidia may bring back the RTX 3060 by restarting production of the GPU

[sick owl]

They are bringing back the 3060 it looks like yeah

How much you want to bet it's the 8 GB version too

[unkempt citrus]

With the price of VRAM

I mean 8GB is still fine enough for most things

[potent geyser]

Older GDDR6 used 20-16nm node

[unkempt citrus]

Yeah but even DDR3 is spiking in price

[quasi furnace]

For the steam game Kovaak's FPS Aim Trainer, does anyone know how the chosen JSON file containing a theme is parsed into the actual game?

I am trying to make an application to change the theme based on the scenario. The various themes are stored within JSON files in a theme folder, but I am not sure how the game actually updates for a specific theme to be chosen. As far as I have seen, it's at least not through file explorer.

Please let me know if you have any ideas!

[nocturne olive]

8GB is absolutely useless in some cases

[unkempt citrus]

Sure, but if you're just gaming at 1080p

It'll do

[sick owl]

Barely

There's already some games where you exceed it at 1080p max settings

[unkempt citrus]

Yeah that's fair

But you also don't have to run at max settings

[quasi furnace]

8gb is high key goated

5060 ti 8gb is overhated

❤️

[unkempt citrus]

If you're after max settings, I don't think you'd be considering a 3060 these days anywaay

[nocturne olive]

The 8GB on the 5060 is so low that DLSS makes the FPS lower because it uses more VRAM

[quasi furnace]

Frame gen?

[nocturne olive]

Who knows

[quasi furnace]

wait what, im asking you 😭

[nocturne olive]

Either way 12GB is definitely a minimum for reasonable new GPUs

[unkempt citrus]

I think that's fair

But its a 3060

[nocturne olive]

3060 12GB is great for budget ML

[mighty thorn]

My PSU has magically been bricked overnight while not even connected to a system or power