#programming
1 messages Ā· Page 271 of 1
why not
you can turn pi (or any other number) into any number rational or irrational by multiplying
noo i mean specifically here
no, for the same reason

contradiction is when you "assume its true then .. blabla... 1 = 0 => then its false"
no?
ok then it really is extraneous

i dont understand my brain at all
its dumb but sometimes its smart
the simplest proof uses cantor's diagonal argument
You can add one or something at every decimal point to get a different number right
non-formally yes
Good enough
well, this is actually just required as a proof that irrational numbers exist
and that they are uncountable
the proof that they are dense amongst the rationals is the simple proof by contradiction
but it requires that you have at least one example of an irrational number first
or have axioms governing of operations between rational and irrational numbers

I'm glad I don't have to do theoretical math in uni anymore 

so it unwraps like this
((lambda (f) (lambda (x) x)) 5)
(lambda (x) x)
((zero 5) 6)
(((lambda (f) (lambda (x) x)) 5) 6)
((lambda (x) x) 6)
6```
I officially signed up for my thesis today btw 
the implementation is basically done
i dont know any maths
or programming

i dont know any maths or programming either
i know how to make clay pots
and how to filtrate water

i dont
i can filtrate water with a filter thoughh
How do you do that as a tv
Be careful, don't want you to short circuit 
True 
no nitro 
@opaque sigil
i think this clarified it.
i managed to unwrap (add-1 zero) using that
you can have this leaf instead š

oh and solution from another website features example where we add functions what print "hello"
and if we add 1 to this function it prints it 2 times
now i get it when you said its not about numbers
now im confused again
so if i want these church guys to work on actual normal integers
it looks like i need to define "one"?
whatever
didnt need this to make a chess bot
do you guys think the universe is deterministic in the absolute, even though its impossible to determine it
did this text im typing have 100% chance of appearing right now since the big bang
this dracula theme is pretty 
its deterministic in the sense that its largely governed by laws that we can understand and use to our advantage, i don't think "universal non-determinism" even makes sense to think about outside of multiverse theory
even something as "simple" as weather patterns is incredibly hard to predict, statistics is very useful even for 100% deterministic phenomena
god how i wish i could do {[index] = value, ...} in c++
yet this stupid rain i had for 2 weeks was determined to appear since the beginning of the universe 
HEY that's not nice
or wasnt it 

sorry i mean weather 
i'm gonna cry now
i actually didnt see your nickname when i typed that 
deterministic laws do not mean we cant change anything, they mean we can study these laws and predict what will happen based on our actions and plan our actions accordingly
so determinism = good imo
doesnt mean everything is deterministic but either way we cant use laws we dont know about so they may as well be non-deterministic for us until we discover them 
yeah subjectively to humans it doesnt seem to be deterministic, at least for now
human agency doesnt in any way conflict with determinism, in fact its supported by determinism
it is a part of it
"freedom is the recognition of necessity" and all that

to clarify, what im saying is - its not guaranteed that universe is 100% deterministic, but we should try to make sense of the apparent chaos and hopefully find some order in it
the more of it we find the better for us
its very interesting but useless topic

because to see it deterministic you probably need all the information existing in the universe
to predict everything
which is basically entire universe within itself
or model which is 1:1
and that model has to contain this model too
probably dont need that much within any useful scope
it doesnt matter what happens 50 godzilion light years away from us
or does it
RVC3-mini is actually going surprisingly well as far as my attempts as AI projects go
Nowhere near complete, hell training loop hasnāt even started
But
Neither of the datasets have given me the middle fingers yet
and they were my first picks for their respective types
You gonna make your own pretraineds?
that word looks so odd 
"pretraineds"?
its supposed to be read "pre-trainds" but you want to spell it pretrainets
How silly
Rvc actually isn't even that bad to pretrain
I guess
Still definitely more intensive than just doing the usual training
Yeah
Yup
Hope you have some GPU power
My 3090 is property of NeuroSynthProject
-# Though technically I would lend it for cheap
this mini one is just a scaled down test capable of being run on my gpu. if it shows promise i will rent the compute to train the full scale model
kaine u can always use my 3090s for free
That's crazy
If just I could afford to have more than just one 3090
Or even better, a 48GB GPU like a modded 4090D
Modded 4090d would be so goated to have
i want one so badly
its like not even that bad of a price for the mod
Ew
It will be slow but still
It has same compute as a 3090
i saw sources saying they were like $3500 (in china where people arent scalping like they are here)
Way too little for that memory
Why not, put it in the case
in my humble opinion as long as its moderately faster than straight cpu training its more than worth it
True but it depends on what you are training
Well I'm using GPUs already
So I would just like bigger and beter GPUs
do you think pewdiepie will let me rent his 10x 4090 rig
NeuroSynth vocal synthesizer
Like the bandwidth for training a transformer is actually not bad
Since it's a single step at big batch size
It doesn't need to autoregressively generate
Like if ur gonna run a transformer yeah dgx spark is trash
I think NeuroSynth is already compute limited on the 3090
Have you tried checking how bandwidth bottlenecked you are
Well as I increase batch size the time it takes to train goes up, so I would kinda assume it's a compute limitation
3060 12gb š
kaine u got the power of friendship
the issue is i dont have multiple gpus in my server so i couldnt test distributed training until it was already on your machine, and remote debugging sucks
Huggingface accelerate my beloved
Build for 1 gpu
Run on many
I love huggingface ecosystem sm
Either way, right now I'm mainly money limited more than anything
If I had the money I would just go and get a 4090D with a 48GB mod
But I'm still broke from getting this 3090
Dude we all would get oneššš„
bruh i am gonna be halfway through today's college class by the time this finishes loading data
and thats only dataset 1 of 2
Very samples
single
Damn goated
i have another one for multi and binaural
CHiME-6, which has 3d coords for every speaker and recording device in each sample
if the main RVC3 works im gonna be able to do asmr
like moving around and speaking into specific ears type stuff
Kaine isn't rvc diffusion?
*RVC1 and RVC2 do
this is new ground
Are you gonna add a breathiness embed?
plus asmr is an area of interest for me so the final model will likely be specifically tuned to enable it
idk how ill do it yet tbh
since most asmr has sfx
the final audio generation part is
Sound effects are the bane of RVC's existence

Asmr

Not even vocal removal models catch them
Imagine knowing how to edit the mel spectograms yourself to clean up audio
Have you seen NeuroSynth yet by the way? It's kinda cool
also i need a better name than RVC3. as much as this is a direct spiritual/architectural successor and they don't plan on making it themselves, i dont wanna directly steal their name
Invent something new so it isn't purely retrieval based any more
i mean im stealing vibevoice's method for zero shot voice cloning
they finetuned qwen3 1.5b to extract speaker embedding stuff for sample clips
im gonna use gemma3-270m
i dont expect it to be as good, but the point of rvc is finetuning so it doesnt really matter
how hard is it to get a grasp on ai to become capable of doing some projects on it
like you do
Depends how deep you want to go and your programming experience
lets say i want to make my own neurosynth
(i dont but just for example)
and i have entry level programming experience
NeuroSynth doesn't need much of any deep programming experience, all the hard parts are essentially done already by others
I just need to patch it so it works

For many kinds of architectures it's mostly already done by others, you at most just need to put the pieces together
The most low level putting pieces together you'll ever need is PyTorch
Unless you truly need a purely compiled engine
They've really made machine learning easy these days
The hardest part is getting enough GPUs
they are not selling modded 4090ds in my city but i could afford it if they did
thats why im asking
i can probably order a delivery anyway
Woa
That's some crazy money you got there
If just NeuroSynthProject had that kinda budget
thats what office job for 3 years gives you
That's actually crazy
i hate my job i want to work from home
How cool would it be if I could figure out money from programming or vocal synthesis
Truly would be insane
Wouldn't that be crazy
But I have no clue how to turn programming into money, nobody wants programming
i assume you need a bunch of projects to convince someone that you can work
or have connections
I don't know, I have no clue how any of this works
actually i randomly got this job when someone invited me so i have no idea how to get a job
if i had to do it myself
As far as I can tell getting jobs is just impossible these days
fluffy rabbit
its okay if you are still a student i think
Getting a grasp on the concepts? Fairly easy for CS literate people
Getting a grasp on the models? Impossible
Kinda true
Machine learning models have become black boxes
Magic stuff goes in, magically magic stuff comes out
how does it work?
you choose a model, then train it on your data?
I normally visualize LLMs as something close to a pinball machine but I lack the skill to make an animation demonstrating how I imagine it
Well you choose a base architecture, adapt it to work on your inputs and outputs, then train
what did you mean by "model" here
Well thatās finetuning of a preexisting model.
You can also make a small one from scratch on your pc or a reasonable one if you rent cloud compute or are rich
Sort of
Depends on the arch you're working with if you can do it like that
My RVC successor project is a bunch of finetuning of existing models and some translation layers to link them together as one system
and you do it all without understanding how models themselves work
sounds like induction
it scares me
Eeexaaaactllyyyy
I mean itās like an opaque bottle
You filled it with your custom little smoothie
Then later when you go to drink it
You bank that the thing going down your throat is the thing made out of your ingredients and not a veil of poison
Full RVC3 model is likely gonna just barely fit into an RTX Pro 6000 Blackwell 96GB
Wuh
So inefficient
Inference should only be 2-5gb tho
NeuroSynth at training time fits into around 22GB
NeuroSynth at inference time needs 300MB
what is the maximum memory customer available gpu right now
Rookie numbers
*Efficient numbers
what do you mean with "customer available"
uhhh
If you count only gaming GPUs, either the 5090 at 32GB or modded 4090D at 48GB
If you count enterprise probably like 192GB

Rare Heir spotted
gulp
below $10k
Yes

Is best
though
I love RTX Pro 6000 Blackwell
its just 96
there's the dgx spark too
for Nvidia gpus the biggest vram card they have is 141gb from the H200

AMD has some 192gb and 288gb ones
You could distill your new 12b model into 3b and 1b variants for Paige mobile app 
they sell it for $36k 
actually more wait
but yeah, gpu cost goes crazy for the high end enterprise ones
even a100s are still a lot 
so wait if 6000 blackwell is 10k
and 4090d modded is 3k something
is the math right
yeah
consumer gpus are wayyyyy better value, despite the state of the market
keep in mind the modded 4090d has half the vram and is one gen older
that too
two gpus does not equal 1 beefier one
And is also unofficial hardware
Made by scamming people with fake 3090s
i meant get two 4090d's
But you know
for 96gb
A 48 4090 is a 48 4090
and blackwell is 96 = 48x2
while costing 10k
i mean its more cheap to get 2x4090
Well the thing is
separate cards run slower than 1 big card i'm pretty sure, the high-end AI cards also have nvidia's special link stuff
Multi GPU stuff is less easy and stable than single GPU stuff

For running models, not training them, a single high vram GPU is almost always a better
nvlink does not exist in consumer gpus past the 3090
Cause overhead
oh wtf
didnt know that
wait but
Not a real GPU
Not if itās hxm5
pretty sure nvlink works differently, i don't know much about it though
Also nvlink is just not fast enough for distributed training
I mean
Technically it is
Since you can do it without
But itās not that much better
surely they have some other tech for the big cluster gpus
Given the throughput of modern GPUs compared to the ones it was built for
it works yes but you have to go through the pcie bus which is slower
Training too because each separate GPU needs to load the model weights
granted for most things it legitimately does not matter
Well no you can split it if you do it right
It'll not be as fast though since it'll not be parallel processing then
when we talking "not fast enough", how much time we talking about exactly?
is it like 1 vs 2 hours or 1 vs 2 days? or 1 day vs 1 month
Well yeah but if you arenāt a company it donāt really matter if it isnāt 10 it/s training, all that matters is that you eventually get the model
Kinda
Either way for me 4090D 48GB would be more than good enough
Like it would probably give me 8192+ batch size on NeuroSynth no problem
Just the price is too insane
See now
How about instead
You just make Neuro synth 8192 times larger

No that'd be stupid
More batch size at the current model size scales way better than making a bigger model with lower batch size that overfits more
-# wouldnāt it underfit since you donāt have 819200% more data lying around?
-# No overfitting means fitting the data too well, underfitting would mean not fitting the data sufficiently well
Basically, if the model is way larger, it has way more parameters so it can just memorize all the data
That's overfitting
Underfitting would mean the model has too few parameters to accurately represent the data
just like me 
I kinda want to mansplain the backpack contraption cause like talking about it but Iām on school bus so I wonāt
common sense

Wa
š¤
I fixed the bad apple glitch so Iāll post some evidence of its existence before I save the patch

RIP
Bad Apple Glitch
11/2/25-11/4/25
you will be missed
RIP
holy fuckballs my implementation is almost 4x as fast in release mode

i'm actually gonna crash out
i am here
In "I am here," "here" modifies the verb "am," specifying the location of the subject ("I").
hi here
no
read this
i think this is wrong and it is actually specifiying the name of the subject ("I")
i think you're wrong
i'm always right though
you're only a letter, not even a whole ass word
so that couldn't be the case
The training is not cheap and it's less stable
You canāt fit in 3090s?
Kaine
It's over half a million of training examples
That shit is NOT fitting on 2 3090s
Technically NeuroSynth is a distillation
With the data source being Neuro's singing voice
Of course there's a lot of augmentation data that makes things better
But it's basically a distillation
which channel do I go to for lava lamp tech support lol
No I meant distillation
Not repeating the entire training pipeline for 3b
Iām not picky
I canāt even use the damn app
No android phone
emulate it
Well yeah but
For every day usage
Get an old phone
Google pixel 6
Could get for $120
I mean I donāt have $120
But I could
everyone in #programming is either fully broke, or has enough money to but 100 gpus
I should get a pixel watch too so I can put Paigeās stupid face on my wrist and talk to her like Jarvis

i have a galxy watch 4
so in theory it should also work
1.4 inch 450x450 display
Beautiful
gpu: Mali-G68```
not much compute but it should work if you cut down on the animations
How-
How much ram
Does the watch have
1.5GB
so more than enough to be running the model
not the llm
but for the llm just use the phone sync stuff
That was the plan
But
It would be really funny to try and run the LLM on the watch

rememebr that the 1.5GB of ram is also used for system tasks
The system consumes about 0.9GB (900MB) to 1GB of the total 1.5GB. damn thats actually a lot
Nah itās fine we donāt need the system
Build your own os which takes almost no ram

idk, galaxy watch drivers
We are working with hardware that runs on 64kb ram
are we?
I mean at my job
oh
There are linux controllers on raspberry pi too but its something new
the galaxy watch core system probably only needs a couple megabytes, but it has so much shit running in the background like heart-rate monitors and sleep tracking and blood pressure and gps and fitness tracking and shit
it even has an oxygen sensor iirc
is that core enough to run the llm
Donāt need it
And can you actually make such os
I mean is there something that wont let you
Like you need source of actual os for it
Tizen OS is open-source for parts supposedly
wear-os isnt tho
and the watch 4 is the first to use wear os
Can you attach more ram to the watch somehow
its still based on android in some ways i think, so its based on an open-source foundation 
idk duct tape ddr5 stick to it
new ship on it i guess?
1tb ram watch when
Gonna run DeepseekV3 on this watch
the newest galaxy watch 8 only has 2GB, they havent even tried to improve ram capacity
Tbh you donāt really need more
it is Penta-Core tho instead of dual
wear os lets programs stay open when you go to the home screen, and so every 3 weeks i remember to go to the hidden "open apps" screen to close them all
1tb watch
watch with a 9950x3d and a 5090 
all to run the fluid simulations to track your health
Some "innovations" really are gimmicks after all.
https://good.store/products/dishwasher-powder
2nd channel video:
https://youtu.be/WnBb3DLlVPw
Other stuff:
Technology Connections on Bluesky:
https://bsky.app/profile/techconnectify.bsky.social
Technology Connections on Mastodon:
https://mas.to/@TechConnectify
Have you ever noticed that I've ...
babe wake up, we have youtuber approved dishwasher powder now
Every day I wake up and wonder how Samsung thought they would get away with the moon thing
they just thought we wouldnt notice
"omg the s22 ultra takes such good pictures of the moon"

Have they never seen a desktop background before
They seriously never anticipated someone looking at a blurry picture of the moon
[
]
it even has a headphone jack 
ye im sold, ill be putting this thing under the christmass tree

man, i really wish more companies would use under display cameras
it's kind of just zte and the couple foldables, yes i understand why they don't but still 
yes it's a sub-brand of zte
i can't lie i'm very tempted to get one myself
now that you get 5 years of updates in the eu

a bit annoyed they took the china version of the phone and lowered the battery capacity and charging speed for the global version though
my old phone is still hanging on, but after nearly 5 years its not happy
crazy how 5 years is old for a phone
the chinese version has more than 7500mah and 80W charging?
ill buy one from aliexpress then maybe
there's two, either 8000 with 80W or 7500 with 120W
this is the one on the local retailer website
they took the worst of both worlds for the global version
ooohhhh
the red 11 pro doesnt seem to be on aliexpress yet
i don't care about fastcharging, i just want the bigger battery
are they even being sold yet?
idk
We'll notify you as soon as it's available!
Plus, get updates on the REDMAGIC 11 Pro Early Bird Perks starting Nov 13.
welp, i have signed up for the early bird perks
https://keepandroidopen.org/
pls spread awareness
Advocating for Android as a free, open platform for everyone to build apps on.
Oh is this Google trying to destroying sideloading
The only part of android making it worth using
if they remove sideloading i might go to linux on phone 
I donāt think theyāll be able to stop people from sideloading tbh
honestly i doubt it aswell
put the apk hidden in a powerpoint file or something idk
surely there are arbitrary code execution things still available
uh, that's not something I would hope for 
you'll be able to use adb still
i mean intentional arbritary code execution
like how you can have a different os on the xbox 360
Yeah I have CFW on my 3DS and Switch 1. I'm okay with that because
- they're not attractive targets for black-hats and companies like celebrite
- they're only exploitable through bootrom defects and need physical access
An exploit that breaks security in Android is 1000x more dangerous.
pinkponk
Not that I agree with google's stance on sideloading, I think that's stupid. I just don't want gaping holes in my phone's security.
ye but thats not what i mean. i mean
you install something that the phone doesn't want you to install
with intent
not a hole in the security
what i mean by "arbitrary code execution" in this context is just all the code i want my phone to execute as a verified owner and user of the device
Yeah ADB will still work afaik
it's called gaining root access (also a massive security risk in itself)
ye
well that
its only a security risk if you make it a security risk 
user error
i dislike the need for adb for this cuz i dont want to have to connect my phone to my pc to install angry birds without ads
Eventually more and more apps will be unusable on rooted phones because they'll require play protect with fully-secure boot chains.

yes
but tbf mostly only money related apps check this(like bank apps, wallets)
not really worth it
i have the dev options enabled but i don't think my phone is rooted cuz i need my bank stuff
Only if they leave holes in it forever.
It is possible to have a full cryptographic chain of trust for every piece of software that runs on the device.
Switch 2 is probably impenetrable since they fixed the blatant cryptography fuckup in the Switch 1.
so its a hardware thing too?
I have my phone rooted(and running a lxc container in background 24/7)
barely notice the difference, even most of bank apps I use works with some tweaks
the only thing not working is google wallet
Yeah, basically each part of the boot process checks that the next stage of the process was signed or trusted by the manufacturer. The public keys are fused into the chip.
actually a few of the 3ds titles were susceptible to RCE through online multiplayer
you could literally trigger a9lh on someone else's 3ds if you wanted at the time
Ah yeah the 3DS was exploitable from userland. Not sure if it still is.
I used soundhax to put A9LH on mine
it still is, but since Nintendo network for 3ds has shut down it's not an issue anymore
most if not all of them are patched through pretendo if you go online these days
The boot9strap paper is a fun read.
leave bootrom backdoor in and hope no one notices
screw up the signature checking code so the signature only needs to be mostly correct
protect sensitive memory during bootup but accidentally expose DMA registers which have unrestricted access to all memory

ban shiro
9
9
1
ye
ā

Mfw I realize each of pewdiepieās 4090s have 48gb
@real sierra oh no
nice knowing you

conveniently i became a moderator for the abandoned archive discord today
so its more of a relocation
lmao
Congratulations!
damn
congratulations for modding a discord?
š
The days of getting pinged every hour because another 10-year-old tried asking how to pirate pokemon are behind me
I'm sure someday Shiro will be free and happy
I wonder if anyone here knows what could possibly be so wrong about my network that the lava lamp outright refuses it
Nobody can figure out why it doesn't work
All that's been found out that either the lamp or the WiFi is cursed
how is it failing?
not connecting to WiFi at all?
Yeah it just entirely refuses to connect to the main WiFi
dns server?
nah, WiFi doesn't need DNS to work
i see
It connects fine if using the same WiFi trough a laptop hotspot too
Like it only fails connecting directly to the main WiFi
If I can't figure out what going on I'm gonna have to get some cursed Linux machine just for being a WiFi relay
You mentioned yesterday that you set your extender's IP address and now all of your devices are under 192.168.129.0/24?
That does sound really suspicious, and if the DHCP settings are your router are borked that could be a problem?
ah woops
confused blue usernames that start with the letter S and have grey profile pictures
That's crazy
If anyone wants to research, you can try to figure out if the Inteno EG400 has any specifics that would break the lamp compatibility
How do you give it the password
Oh so you first connect to it with phone through bluetooth
Do you have some weird wifi name
Non english letters
It even connects to a laptop hotspot that has the same password and SSID as the main WiFi
But then doesn't connect to the main WiFi
Nor numbers
Oh

It just doesnt connect at all or it has no internet after connecting?
had a
moment
Doesn't even show up in the connected devices graph
So definitely doesn't connect
I wonder what chip the lava lamp uses
Hmmm
Tuya chip
The router is broadcasting on 2.4 GHz right?
Want to say its DHCP but it probably should connect even if dhcp fails
Or shouldnt it
That SSID I'm giving it is 2.4GHz-only yes
I don't even know if you can configure DHCP stuff on the router
And other devices can DHCP just fine too
Is lavalamp the only device connected to wifi with this subnet
No it's the 2.4GHz SSID of the router that everyone else also uses
Can you try disconnecting everyone but lavalamp temporarily
WiFi just acts as the physical layer so it doesn't technically need an IP to stay connected, but the lavalamp firmware could detect that something is wrong and reject the WiFi if it can't obtain an IP
Maybe your dhcp giving out some weird ip
And lavalamp has some shit hardcoded check for it to be in 192.168 for example
The WiFi should be just normal 192.168...
Even if my desktop has a weird IP since it's bridged
Are all clients of wifi under 192.168
Yes I believe so
Doesnt router show it explicitly
I don't know
But the devices show in their IP configurations as being such so I believe them to be
Ok ok then it is
I'm not home at the moment so I can't directly check
When i connect my laptop to some linux based controllers they get dhcp'd into some weird shit like 169.147 idk
Maybe lavalamp wants that instead
Shouldnt be the case though
well that means DHCP didn't work and it's just the IP for the direct link
Oh
What ip does lavalamp have when you connect it to laptop hotspot?
Also can you connect to lavalamp somehow

Does it have some linux inside with vnc enabled
yeah I doubt it's running linux
You should check router logs
I don't think it gives logs
What is it
I feel like this page would still be unhelpful even if I could read Swedish
https://www.telia.no/internett/wifi/rutere/inteno-eg400
Her finner du informasjon om innstillinger og annet for Inteno EG400 ruter.
Silly
yeah all manuals i found are in swedish
Well would be good to know how to access them if so
why is the compressor turned to 11 on neuro-sama vids
it makes me not want to watch it, tbh
Hard to tell without seing available settings
I would just try different settings to see what happens
it does have a serial port
can poke at the router OS but I have no clue where the logs would actually go
https://openwrt.org/toh/inteno/eg400
I don't have hardware to poke at serial ports
And the router is also stuck in a corner where I can't get to it
Thats some extraordinary setup...
Doesn't seem that unusual, the router has to be in one corner of the house and it just so happens we've kinda built a wall of stuff around it
went to local shop to buy wd blue sn580 1tb
guy had the audacity to price it at 100$ when even amazon is selling it for 70$
You need logs and just toggle some settings..
Logs could clarify which settings you need exactly
also I have a question
I need to use an old router as network switch for four raspberry pis should I put openwrt on it or will default firmware be fine
last firmware update was 2019
Would be kinda great to know where the logs are
You will have to log in and search there
I may have to port it to openwrt though ngl
Guess I'lll try once I'm home again
i think
just for simple network switch it should be fine right
Should be yes
$70 is still robbery for one of the lowest-end 1 tb drives
bet I could get it for 40 or 50 at the pc market
or even lower if I haggled like a lowlife
default should be fine unless you need VLANs and such (assuming that this is a potato consumer router)
there's a very high chance that there's an internal switch anyway
nah I just have to plug them into the router with ethernet and occasionally ssh into them remotely when im not home
To SSH into stuff on your own network it's better to VPN into your network
You don't want that directly port forwarded
I know, that would be genuinely stupid
and yeah if it's just acting as a switch that wouldn't matter anyway.
the main router would handle it
that shit only worth $30 at best
i could get a second hand 1tb for even $19 and it'll be even more reliable
(i actually did last time, still running well. it only got 20-30TB written into it when i got it but should last for few more hundreds)
SSH isn't inheritly less secure. It's a matter of whether you need to have easy access to the internet network routing and how much you need to expose
Dafuq???? WD Blue 4 and 6 TB, NEW are less than $150. How can a 1 TB of that, especially SMR models they use now, run $70??
sn580 SSD, not the hard drives
okay that's actually not as bad as I expected
https://www.techpowerup.com/ssd-specs/western-digital-sn580-1-tb.d1541
still wouldn't buy one for $70 after being spoiled by SSD pricing from 2 years ago though
Oh, sorry, my bad.
Still not good, though.
WD Red NVMe is like $115 for 1 TB... So...
And that's like... WAY better writes
arent they supposed to be 60 bucks per terabyte?
isnt 115 for 2TB?
its the sn580 nvme ssd right? 123 for 2tb
when i try ot seartch for the 580 on belgian amzon it changes my search to 850 so idk what prices are here 
idk wtf this weird changing of search words is, but its kinda creepy in a way
oh erm
i think that doxxed me slightly
eh, just imagine the video of me typing sn580 and it changing to sn850
So stupid
who does that
Nobody saw it 
i mean, if you delete the video attached to the message, people cant see it anymore right?
cuz i know there are things to see edited message history
I actually tried to find 1tb version of this ssd in my local store and they were redirecting me to 250gb variant
But its because 1gb is out of stock
1gb 
Quantum ssd
must be a very fast gigabyte
If it was a link, mods see it in their mod-log channel
just an mp4
But maybe discord uploads everything on their server
And its still a link
Depends on if discord deletes it from server if you deleted it
No but mods will see it too
I mean we can surely trust mods
But still
I'll check in some other server where im a mod hold up
surely shiro wouldnt blackmail me 
If you suddenly disappear we know what happened 


Idk if @Twins role has access to mod log
They dont have access to any channels i think
Did i misspell it
imagine getting doxxed on stream cuz of a video you posted in the discord for 5 seconds 
Why is it not converted to fancy @
However im pretty sure there is some extention for discord
unpingable role
Allowing anyone to see deleted messages
now thats scary
<@&1061758877609234622> i meant this
Looks like desktop client converts text properly to that but android client doesn't
If its unpingable of course
For pingable doesnt matter
wrr

iiiii

typical projects page. I have to scroll down to find new projects post 

sort by new not an option?
it is
theres date posted
but that sorts ny dates and the most active on the current date
not by newly posted
isnt date posted exactly what newly posted means?
pinkponk
^
discord has a different meaning for it, from using that sorting
sorthing by dates is what you want tho?
i wanted newly posted. And not sorted by whars most active on the current date
Sorting by date posted
i.e. what post has the most recent messages bt ranking
wait-
i just realized i was stupid for not checking the time it was posted
i just already saw most of them

so i thought-
Windows 11 you piece of crap, I'm trying not to install a MS account here so I can make correct install media so I don't need to enter accounts
rurururururururururu
have you tried making it from scratch?
Windows 11 tends to work more cooperatively if you just.., borrow the source code (TM)
nomboard
also it's very funny that 3 of those are non-game integrations and then neurobot1k
Of course, I should just remake windows 
alright, the vm is installed and time to give back your network permissions
and open source it! :D
If microsoft is vibecoding the os, so can I
These usb drives are so slow I think they took the usb 2.0 speed limit as something they could reach in perfect conditions.
this code is so bad 
Still working on that game?
yep
you'd literally die if you saw my gltf handling code
Silly
i doubt it
you see
i program in vulkan
this code is meant to read a midi file and turn the drum track into animation
opengl 4 is nothing
i'm not masochistic enough to do vulkan
š
Very silly

numbers
these are indices for each limb
2 for the feet, 2 for the hands
so since its jsut a pedal for each foot, those jsut go from 0 to 1 if active, and the hands go from 0 to 7 if active
with fractional part encoding the velocity
from 0.0 to 0.1 cuz i quite literally just did * 0.1 to put it in the fractional part
now i need to make the actual animation keyframes these will be indexing 
actually lemme eat first
hmmm, not yet
the problrm isnt the looks
its the functionality
when you have a midi file that jsut says what note is being played, how do you decide which arm is used to play which note?
i program in 

Do the lava lamps have an open API for changing the color / toggling the power or do I need to spin up wireshark?
Ah thank you!
Wa
Filament runout sensor, new mainboard, and dual z-axis
Why does this need to be warned 
"Oh no, we made the product too waterproof
We better warn them"
clearly this means that if it gets dirty you have to dry clean it
no i'm joking
t
hi konii
not its not
youre the letter t
cant be blamed for your actions
as you are the letter t
buh

i see
Oh isn't that the same kind of AIO I ordered like 3 months ago and still hasn't been sent?
probably, it's a popular one
arctic liquid freezer III pro 360
Yep sounds pretty similar
I have no clue why it's taking so insanely long to be sent
I got mine on discount ages ago and it's still stuck in limbo
They gave me a metal toothbrush
Wtf am i supposed to do with this??? 
why are you dissasembling a printer
What are you having issues with on it?
it just doesnt speak
And does it give errors?
ahh
are you installing steamOS on your printer or something š \
Have you tried stripping down everything around it so all you have is a terminal and anything you type to it will be immediately fed into the TTS?
it's a 3d printer
hold on let me try that
alright the test is working
but why not in my original code rah
Then you just need to find what it does differently in the test case and real case

i mean i get it man, you have some experimental new update out there. why do i need to see this icon all the time 

-# 

is anything being researched on combining neural networks
like a bunch of NNs that do different things
and then you make additional NNs to combine results of those first NNs
and do something with that
because it feels like every NN is just doing one task and thats it, we use that
uhh duh

first thing you do after making two bad things is shoving them together to make them better
what if we take two good things
what about those decision making AIs
how do they work
idk what if we wanted a NN that can play any videogame
so we give it visual NN to read the screen
it forms some kind of output
then another NN distributes the result to other NNs, each with some specialty
which all do something according to their specialty
then these results get shoved into some decision maker NN which outputs to the keyboard control NN
the modular NN
is it a thing?
Why does this link need to exist now 

not like there can be any problematic windows updates or anything 
they test each update thoroughly after all, and do not use llms to build critical software 
then you give it a real life simulation
then you give it robot body and tell it to play life
so google says modular ai is a thing
How does one configure 2 routers in a master-slave configuration? Anyone here know something more than the lava lamp thread people? (yes this is for the lava lamp)
what protocol
look up mixture of experts
I have no clue, I have the second router's WAN port connected to the main one's LAN port
And why is that?
then you can configure "master" router to do something with that port
Off of the top of my head, bridge mode or AP mode depending on what your trying to do
for example forward it to the wan
The WAN port is also the LAN4 port anyway so it's technically both WAN and LAN
the wan port usually has special set of settings
actually yeah you probably can do that thing
So how should I configure each of the routers?
so you need slaves' wan configured as some kind of no-password-nonothing
depends on the routers i guess
One is an Inteno EG400 and the other is a ZyXEL VMG1312-B10D
so this is what my router gives me for example
lets say i wanted this wan to be slave's wan that is connected to other routers lan
actually now i look at it idk what to do
if i had to do it, it would be trial and error
i would leave it as is, dynamic ip. this way - master router dhcps slave router to some ip
then if i had to access something connected to slave router through master router i would uuhh
i can only think of port forwarding
i access slave router ip from master router lan
slave router port forwards it to the corresponding lan
or ip
it can also be done with vlans but im not sure if consumer routers usually have it
Well at least the slave supports something called VLAN groups
yeah mine has it too
what are you trying to do actually
you found some router that works with lavalamp?
Yes
you probably already have two vlan groups like wan and lan
modern java is actually based
I want the lava lamp to connect to its WiFi, which gets its internet from the main router getting its internet from the fiber
jep 454 is nuts
wait have you tried to just run it
connect lavalamp to wifi
connect slave router to master router (100% lan on master side, try both wan and lan on slave side)












