#programming
1 messages · Page 466 of 1
My handwriting is dog water
mines worse
Mine needs a translation
by MVT there must exist some value c \in [0, 365] such that
f’(c) = \frac{1}{365} \int_{0}^{365} hbd(x) \,dx = \frac{1}{366}
however hbd(x) is only defined to have a range of {0, 1}
by proof by contradiction hbd(x) must always be 1
so it is always shiro’s birthday
Mine gave a teacher a mental breakdown
okay im done
I KNOW ur not using the MVT 😭
My teachers let me write everything on the computer
also nobody question the proof by contradiction it is definitely rigorous
bro things they allen turing bruh
I still remember when my teacher said "I would have a better chance of understanding hieroglyphics than understanding your handwriting". I mean she wasn't wrong tho lol. I don't even know what my own writing says
I dont have time for writing classes to try to fix it
And thankfully coding isn't a job with much writing
Yeah I mean it doesn't matter much
wow what a token waster
it generated the file and then realized it needs to remove the old file first and then regenerated the entire thing again
and the code doesn't even work
wow it only took 5 attempts good job

@faint sandalai literally cannot generate working code
just why even bother
code it yourself
https://www.youtube.com/watch?v=mHj6IvBmlpU
in other news: fork found in kitchen
Taking a dive into "Browsergate".
🎵MUSIC!
1080 Snowboarding OST- White Out
Goodmorning 
that is 99% of my experience with AI. its better to just let it create a basic skeleton of a script and then fix it manually.
water found in ocean
At the moment I'm just having Codex build a small app, defently not just a run now button as I have things called user requirements
I reread this and somehow it's way funnier
I'd not trust for anything that doesn't already exist, but this is simple enough and should mostly function
it's actually faster to not use AI at all
you can rip shit from open source projects
it's not as easy as it is to rip from oss projects
you will have a guarantee that it works if you have the ubderstanding if what you're reading
like I haven't touched http file streaming in unity in ages
tried AI
useless fucking clanker made up unconceivable shit
quick google, open some github project, I have everything in a clean form I can copy and adjust
in certain languages that takes longer than just asking ai to do it
english and the most popular game engine
idk man seems pretty basic
You have unlocked new role
True, if I wanted to rip features out of gimp just to make my prefered workflow then it'd take longer than just trial and error with AI
it's a matter of knowledge
It is a matter of time as well
if you know what you're doing you will only waste time with AI
if you don't know what you're doing then you won't catch all issues AI produced
and the public models just keep getting worse in my experience
nah it is faster to do yourself
if you not know some funcs ask ai what is used for task and then search for examples of func usage
Agree to disagree
👆 especially the really basic utils stuff should be done yourself instead of relying on a package
I won't have AI touch some projects, but I'll be doing others
the day oxlint has full compat with js plugins is the day I will really rejoice
except asking AI didn't work because instead of using unity basics like coroutines it tried some wacky c# tasks
if it gave me wrongly implemented code using correct functions I wouldn't complain as much
but it's just wrong at every step
first open source projects popping up on google are at least highly likely to be of acceptable quality
and more often than not working
ai is tool for initial search
and yes sometimes it cant even make use of dataset
as I said the models just keep getting worse
I was only using it to know what to google
now it's just misleading
my last successful usage of AI was it writing me box blur in hlsl
except when I asked for a different type of blur it gave me a different implementation of also box blur
I was only using it to know what to google
i mean exactly that
it is only use of ai for me now
yeah but it sucks for that too
LLM tech has long plateau'd, it now started deteriorating
internet getting increasingly swamped by AI generated text that gets fed back into training datasets
i disagree, it's been getting mostly steadily better
It is getting better. But just not as fast, that's why people are saying it's plateaued. I don't agree with that tho. Improvement is improvement, but I don't like when they overhype it
the outputs are getting longer
that's it from "improvements"
the quality is going down tho
Chat TIL at my new job the ci cd they have is they have pycharm installed on a windows remote desktop on which they manually clone the repo and have a Windows scheduled task that runs it as a cron job
This can't be real
- why not, basically any other devops stuff but
- so fucking fragile i cant believe
Imagine manually restoring all the stuff in 2026 while all ready to use solutions exist
chat as a hypothetical how hard would it be to make the fart gun from minions

Not that hard
Getting the right chemical is the hard part
After that, it's basically spray gun or you could make it like vape by heating up the chemical first
easy
but not very practical
Well, chemical weapons are not suggested in the geneva suggestion after all 

im just thinking outside the box
if something isn't suggested that doesn't meme you cant do it
Well, yes. Teargas is a prime example of it

vrchat is multi threaded apparently but damn
people that uses ai and claimed "it worked" are just either stuck with JS slop or some simple "saas project"
istg llm coding never worked for me
it just always waste like 7 hours of my time
Honestly, I prefer if app is using the full resources that is available to them if they are indeed compute intensive. FreeCAD for example still struggle to use more RAM and CPU
i had to learn java all by myself because i was so fed up of clankers being incompetent statistics generator machine
finally someone sane 🙏
it's a loop for me
I try using it -> it's shit and I get reminded why I'm not using it regularly -> time passes and I forget -> repeat
i've been against llm coding since the beginning because i just don't think trusting a model that is meant to emulate text based "intelligence" is a good idea and they should NOT be given such constructive task
it won't and will never work
just see what the fuck happened to win 11
the idea of a time saving tool is great
but at best it moves the time spent to fixing badly written code while risking bad architecture for no reason
exactly
it has no internal reasoning, it's just a massive inefficient text completion algorithm
and it operates on human languages (programming languages included) which aren't good at direct and consistent communcation of the truth, they are just an abstract representation from natively inconsistent humans
i understand if people use them for like basic js backend and front end but no, DO NOT use them on system critical things
they're also ass at writing compute related code
I mean, it's just a technology without future
like shader as example
finally someone said it
like
it ain't the future.
you're the crunchiest of fellow toasts
heart
LLMs are inefficient and inconsistent
they have limitations in datasets and computation power
and we managed to hit both those barriers
the dataset is increasingly self-fed which makes it less consistently correct over time
and we all know what happened to hardware prices and supply
LLMs are maxed out and still some idiots keep throwing money at them instead of new stuff
I mean, it is a money making machine for the companies
as i said
for the company, its for cash
for the ceo though
they're actually dumb
sam altman doesn't even know how ai works
get investor money, burn it all, pocket license fees from customers
you technically don't profit so you don't pay taxes
investors are coping they will get to fire all employees for a rich get richer scenario so they let you burn that money
and adjacent parts of the industry are loving it
hardware production and hosting services
just spin that money in circles to pay less taxes
iirc not even the researchers know how AI works
other than math
NNs are massive complex algorithms
that get built by pushing data through it until you achieve expected inputs and outputs
researchers don't understand it in a way that they can't precisely adjust specific things inside of it
they do know how outside changes affect its functioning
ehh that one, its more like they're attention seeking iirc with exaggerating that they "don't know how ai works"
i mean, they don't keep track of every single computation happening nor any of the datapoints, but they build the architecture to inference such compute
this
its same thing as we don't control what the CPU do in hardware level
yea sorry I was being a little broad
but yeah sam altman is actually genuinely an idiot
the shit we get rn is because people put such idiot in charge
just a reminder that the guy that caused all the hardware crisis rn AND actually wasted it collecting dust in the warehouse
is a fucking first year drop out
Samuel Harris Altman (born April 22, 1985) is an American businessman and entrepreneur who has been the chief executive officer (CEO) of the artificial intelligence research organization OpenAI since 2019.
Altman attended Stanford University in Stanford, California for two years before he dropped out and co-founded Loopt, a smartphone geosocial ...
doesn't even have a degree.
Sam Altman == 
I think he's doing a good job
so many years of milking investors without the company creating anything of value
Thiel like the taste of his blood though
idk what you're expecting, most ceos are like this
if he was technologically competent he wouldn't be the ceo of a large corporation
I am just starting my own company, so I am a CEO 
sorry i don't make the rules 
I can't tell who is toast 
It's ultra fun when several Toast are typing
Or when you hover to see who reacted and it's all toast
oh cool
Job opening?
yummy
always has been
Vrchat should not be using that much cpu tho. I wonder what happens under the hood
Cryptominer

When everything else doesn't make sense, it's always cryptominer 
yeah i don’t get how people don’t get this
ceo is the ultimate management role of them all
people acted surprised when they heard sama is an idiot
i've been saying most ceos are idiot
he’s not an idiot, neither is he “technologically competent” whatever the fuck that means
Was gonna say, he's not an idiot
He's just not a tech guy
The mans still very savvy at business and PR
Most ventures like OpenAI would not be securing the investment they are/have
The RAM was a fuck up though for sure
was it though
ordered EB worth of ram every month only for it to waste in a warehouse
Yeah since walking the deal back was a big embarrassment
and he cancelled most of the data center building shit
what repercussions are they going to face for failing to fulfill the ram orders
Shaken investor confidence
i even did the calculation. its nonsense
did they though
Less expectations of reliability in future business deals
it could be you
you need shit loads of nodes
It was defo a fuck up
that much of distribution of nodes is going to fuck up your utilization, nor its even realistic to maintain that many nodes.
Reputation
don't underestimate how much is a fucking exabyte
Or you can also call it trust
i see this as his stupidity in making the calculation and by extension, its a questionable management
all while his company is in major debt
that's why i said
he's a fucking idiot
on top of all that now you have investor making it an issue
yeah ig, if they announce a deal in the future it might affect how it’s perceived by investors / their ability to sign deals
good job, sama. good fucking job
fuckin idiot
all that to fuck up the whole economy
its a big fuck up that could've been avoided
like
its NOT even that hard to calculate all the shit
i am also pretty sure he has bunch of advisors and a team for management telling him its a horrible idea
like
i genuinely don't get his thought process.
vibe businessing
in what world would you fucked up with something that obvious, really?
obvious to you, not to him
especially when as a ceo, you need to consider A LOT of the variables
A LOT of the consequences
but no
this moron just went with his childlish idea
the fact that he ordered normal dram and not even hbm
If they had the additional investment to follow through the idea actually would have worked in their favour
it won't
its still absurd numbers of nodes
Its not about practical utility, its about starving out competitors
They didn't buy the RAM to use it
you're just wasting your own capital
Its not a waste if it reinforces your position as the "Google" of AI in the eyes of the common consumer and causes B2B rivals to falter
LOL you think they actually paid
doesn't openai have a board, wouldn't they just block it if it was just a bad idea
how is that not fraud 😭
No binding commitment was made
it’s only fraud if proven by a court of law
he went in person to sign the deal no?
how is that relevant
the whole thing is just absurd
There was no actual legal clause saying they couldn't back out though
insane
He effectively signed a statement of intent
so where the fuck did all the ram go
Well AI hardware demand is actually high still, some of it was anticipatory of the OAI deal
Some of it was tariffs and global instability
1 eb per fucking month?
Some of it was the fact the RAM market is totally corrupt and proven in court to price fix
"ai hardware demand"
literally as much as dram as the whole TOP500 list combined
let me repeat this
its dram wafers
fucking dram
not even hbm
its nonsensical to say that its for the ai hardware demand
That's where a lot of RAM stock went
The rise of MoE models accelerated demand for it exponentially
where the fuck is the demand even when current datacenters are already doing fine and still losing out profit from maintenance and operating cost
Larger models that are cheaper and faster to serve than smaller ones as long as you have the memory
Go figure
perform somewhat ass ngl
for its size
At smaller sizes yes, but I wouldn't call the Deepseeks and GLM 5s of the world ass
yeah regardless, 1eb is still fucking overkill
when you can literally batch the models aswell
its just absurd
pretty sure i've talked about the node count
The exabyte was storage and not specifically RAM btw
Oh you're referring to the wafers
for 16Gb die on average yielf rate
Is it really a surprise
SUPPORT US: We launched our brand new WIREFRAME V2 MOUSE MATS https://store.gamersnexus.net/products/gn-wireframe-v2-mouse-mat or check out our e-waste Inductor tabletop gaming dice set! https://store.gamersnexus.net/products/inductor-full-tabletop-mtg-dnd-premium-dice-set-7-piece-dice-wooden-box-token-card
THE DRAM CARTEL: With absurd RAM price...
Because AI demand isn't the sole reason
when companies are already known to back out
i don't buy it when people said its really ai demands
i say
I thoight you were talking about general stock exhaustion then not the OAI deal
its fucking greed and rugpull
now dram companies realized they can charge whatever for ram
people tell me i should be happy with how its all going up in price because i have a lot of ram
dude its like
i have a house and housing price goes up
but i live there so like, i can't just sell it??
They already realised that, it was climbing even before all this
Like I said, the memory industry is proven to price fix
There's only a few major players and they collude with each other
now they're going to do the diamond industry bullshit
limiting the supplies
to keep price high
while diamonds themself are actually dirt cheap
Yup, no use getting angry about it though
Is what it is
Just makes you miserable over something out of your hands
onion flavoured toast
I will get angry tho
I'm angry and i'm all out of ideas
404 logic not found
it do be the most basic example of supply-demand situation
now that im done venting and dooming
ram be like stonks bro
also discord may soon be using cloudflare turnstile for captchas instead of hcaptcha
imma get the s7200ap actually working
I sure hope we don't treat RAM like investment
New speculative bubble just dropped 
i have 768GB of ddr5 ecc but what good of the stonks when i literally still need it
playing 3d chess while we all playing connect 4
yeah but the magic portfolio
have you thought about how number only go up
-# until it doesn't
i just hope that when the world economy detonates in an uncontrolled demolition when openAI implodes that we at least get a sora-generated video of sam going "GUH"
hopefully someone generated one of those before sora got shot out back
wouldnt be the same if veo generates it
god i fucking hope openai goes down
Why do you need it 
Detailed analysis of Gemma 4 31B quants from Oobabooga just dropped
Confirmed my suspicions about lmstudio
Lmstudios quants do be kinda ass
Same for ggml-org
Unsloth and Bartowski trade blows at different sizes
why not
i thought this was known
I mean, do you have it running?
What is it filled with
i don't have as much ddr5 in service as he does (320gb) but mine is all in use
It was, but a lot of discussion about it was based on vibes
it's all vibes all the way down at the end of it all
ai = vibes 
it never ceases to amaze me that a shortened form of "vibrations" (mysticism) made its way into common vernacular as vibes
do you feel the motion of the ocean
no
damn, your 5g antenna is miscalibrated, please see your local verizon technician
i only get like 0.00001% lighter when the moon passes overhead
this LLM model is peak overthinking
lookin like qwen
Just write the damn code cuh
Wait,
it is
its one of the thinking models
im self hosting rn
i am in tune with the slop
thinking hard or hardly thinking
its thinking hard AND hardly thinking
All the Qwen 3.5 models are reasoning
welcome to qwen in thinking mode
If you get annoyed by looping try Gemma 4
because it is a simpleton
im just gonna set it to nothink mode
Much less prone to it on the new llama.cpp releases
imma try it
(Thought for 16m, 57s)
No.
(and 31B is smarter imo)
Qwopus is ass bruh
Most of the claude opus tunes of foundation models just degrade performance
ive used it like one(1) time and it didnt loop where qwen (regular) did so that's the full extent of my testing
wow that solved the problem(!?)
qwopus sounds fun so it must be a great model 
either way if i want something that can think about anything it's not a qwen problem
using c++ 
explicit return 0 
no endl 
qwen thinking mode has always been an unmitigated disaster since they gave it the option
never seen anything useful from it
I wouldn't go that far, its unstable but smart
Gemma 4 is way nicer to use though
unstable = bad
I wonder how carbon is doing hmm
one day theyll be on .2

had to roadmap that one out
memory safety AND cpp compatibility

tbh i dont want this for code i just want it to play minecraft

Cabron
dios mio
minecraft qwen have decision paralysis worse than neuro
rust propaganda 
at least neuro will enter an action once in a while
its like rust but instead of Fe(III)2O3 its C (not c)

Carbon: graduating from the experiment, NDC {Toronto} (May 5-8)
soon
surely
i remember the name
carbon: graduating from junior experiment to senior experiment
0.1 isnt out lmao
road to 0.0: we have done nothing
how to get google to pay you for 4 years while working with the pace of a personal concept of a plan
adding a design for memory safety to our 0.1 milestone
very memory safe: there are currently no publically used programs with unsafe memory behaviour written in carbon
LMAO
so their whole point is memory safety
and they want to start thinking about how to do that this year?
yes the whole point is mem safety however yes it's a roadmap item

i think the thing they mainly dealt with so was is cpp interop
we released carbon 1.0.0 a decade ago, here's our roadmap to memory safety
which is an absolute clusterfuck to achieve
2028
safety strategy 
beyond 2028, ship 1.0 language
very optimistic
i like how it goes from
its very true
2040 is beyond 2028
google wont let them hope for more
0.3 and 1.0 same thing basically
being a google project, 1.0 is insanely optimistic
unless it releases as 1.0
in which case
1.2 is insanely optimistic
those are the 2 options
the repo has 333k lines of carbon 
dent
10k lines of bazel, average bazel project 
it's all just one guy manually typing out
"I will not illegally access memory"
like bart simpson
for 333k lines
with the second main ingredient being
may include carbon
in large quantities for mass production
dont cut down swathes of land for plantations
we're all toast now pog
Tiktok having Ai warfare in comments
You have unlocked new role
become Toast NOW
the original toast is here too
there is an original toast?
What is going on
right but like how does that scale and/or is there a known producer that doesnt do this
@opaque sigil 
omg
What do yall thinks about gemini cli and claude cli
the answer is
uhh
palm oil ethically sourced from palms
I think it's very good tool
our lard who art in heaven, oiled be thy name,
I kinda don't like claude cli because it's a lot of money
i used like $600 of claude code making it rebuild rocm for like 2 days
i dont like claude because its kinda like making an 8ball named einstein
(didnt pay $600)
I have to write hunrdeds lines of detailed .md just to give gemini more brain to do project for me
it does feel a little disrespectful
it is like being project manager for the most eager of morons
Also do yall use llm wiki for your project
they should have named chatgpt after sam altman
robert
(llm wiki I mean it's like docs folder that Ai can read, so it doesn't wasting time searching on internet and RAG)
robert mc stealyourmoney
i forgot that's his full legal name
you shold pitch this to garry tan of ycombinator, he might pay $300 million if he forgets about his epic gstack in a drug fueled haze for long enough
that requires an existing docs folder 
the fact that it's not generated automatically is a joke
instead of eating raw text they should embed it first and save permanent memory so they can later match the new context with saved data
ycombinator is that one website that comes up if theres not already an argument about the topic you were searching for on stackoverflow or reddit
Ai likes to forgetting things, so I have to write a bunch of .md for Ai to make it smart
haxx0r news
on the other hand if we want efficient and consistent technology we're not using LLMs
15th wayland argument in the last minute
Honestly I don't even write code anymore lmao, I write .md and reviewing pull requests
sorry i forgot you have to check phoronix first too
the world needs more discussion of the minutia of implementing wayland
there's not been enough human suffering caused yet
well i think wayland bad and we're all gonna die if we use it and x11 is the tech jesus of the 21st century
i was told this in a dream
viva la xlibre
Clearly they both suck and we need a third standard
o captain my captain
one per distro, ideally
yes lets name it x12 and then eventually rename it wayland
i think canonical should carve a brave new path once again
toasts, translate this discussion from linux language to english
imagine for a moment if people had been locked in a debate as to whether or not explorer.exe was better than explorer.exe for the last 20 years
Using cli gives me aura in public lmao
terribly leaky boat that works everywhere vs sturdy boat that everyone has a different version of but all universally sink if using nvidia components
certified linux user moment
new proposal
relying on LLMs gives you crazy negative aura ngl
run everything as root and give access to framebuffer + hardware
no display server or protocol needed

everyone draws to the same buffer
lets share all the stack memory of programs too
next we can make it so all programs use the same register space
I did not rely on LLM 100% I don't believe what Ai gives me it's actually good or correct until I review it
you don't need malware when all ware is made to be mal

in fact lets turn off the preemptible scheduler and make a yield syscall instead
it's guaranteed to be bad
"I don't know if literal shit tastes bad until I put it in my mouth"

dont actually need syscalls, just make them function calls into the kernel
terry??

it gives me as I planned in hunrded lines of .md prompt, I would not say it's shit but not always good, it do as I expected and code as .md said but still needs human to fix or review some results(If bad prompt)
that reminds me, my program menu logo got inadvertently set back to not-temple
phew that was a rough few moments of realizing that
i wonder if ZealOS or whatever it's called ever touched the whole "we consider user space and kernel space the same thing"
32-bit color VBE graphics
Fully-functional AHCI support
Network card drivers and a networking stack
UEFI booting via BSD2-licensed Limine bootloader and Public Domain ZealBooter prekernel
Changes include:
60 FPS
VBE graphics with variable resolutions
Reformatted code for readability
Added comments and documentation
HolyC -> ZealC
System-wide renaming for clarity```
dont think so lmao
program safety should happen at compile time and verification time
runtime safety is for people who fear like
exploits
and stuff
the macos model, cant get exploited if nobody targets your shit
until they do
nee, must manually jump to another program
run your shit like a 1970s coroutine
what if OS but forth interpreted
i need a VPS that isnt shit for a thing
need sudo 
maybe i just self host on my residential ass connection and then pretend i didnt forsee complications later
cloudflare tunnel means anyone can be a colo
would have 5 9s but my cat jumped on the main host box's not-enclosed motherboard
surely this search panned out
they got 8 answers
all presumably pointing them to the ID IOT vps provider
the best free vps is the one you pay for and forget about
Safety shouldn't happen to begin with
i had a lifetime use VPN for 6 years that i paid a one time like $5 fee for when they started
6 years because the company went under lmfao
I have a vps at strato. To no ones surprise, they suck.
But its cheap
And i have a proxmox running on it.. sooo
I would just use ultra.cc but they do not give root in the cheapo tier I'm on and whatever it is to be a full vps type service is >$5 month that I'm paying for that
And I'm pretty sure something like that is $5 elsewhere
hetzner my beloved
hetzner is on the list
vultr was another that came up
and ovh
but vultr is murican and ovh is french
both of which are not fantastic signs for wanting to not get spied on tbh
What about Hostinger?
that popped up and honestly i thought the name sounded murican so kinda skipped it
but now i see that i am mistaken
maybe i should just use spaceship
i already have a domain through them
but theyre super merican
I think hetzner tends to be quite uhh, skeptical if you're not from germany or at least EU
so I've heard at least, I wouldn't know 
hello it is me, latvian funyun, here to purchase a vps
Vpn through my netherlands box 
I love getting a call from some recruiter right as I need to leave for an appointment
couldn't have chosen a worse time now could you
yes but now you are the asshole for rushing them off the phone
oh well
he said he'd call back tomorrow I think
I was planning on sleeping after coming home from uni

Recruiters have been a rabbit hole of garbage lately
lovely
I need one of these circle bois but right now is the one time I can't find any of the loose ones I know are all over the place in here
Immediately lost my thumbscrew and a slot cover
Nice

Found it after I already found a replacement
Finnish?
I cannot confirm or deny
That website is Finnish
The problem is that most of the hardware is worthless
I have no need of 290Xs
Or a 4930K
4960X? That's an entirely different question
900d is a
case tho
i wanted one real bad back when i ended up with a cm 690 II
My best Tori find has been a Caselabs STH10 for a fair price of 100€
Same seller guy ended up helping me with thesis :D
Finnish used market somewhat sucks unfortunately
I would know
900D
STH10
full tower gigachad detected
Meh, I have the humble Asus AP201 now
It's tiny and good enough
Ended up selling STH10
Finnish 3090s so expensive 
i have one of these for a sort of test bench type setup but nothing permanent because it doesnt fit a full graphics card and its hard drive cage
How much should a 3090 cost these days?
brother the address is right there in the image wdym
Old GPU anyway, stinky
I guess good if you need le VRAM
cuda + 24gb vram premium basically
it'd be more reasonable if it were 32
24 is too smol
Not an alltogether terrible price - but still very high for lastgen I think
It even comes with le GPU block
1650 euro?
Yes
about 2k is what ive seen 4090s going for 
Keeping in mind this is w*tercooled
Which is fun thing to do but very much not for everybody
if you had that STH10 you could've slammed a whole phase change unit in that son
To the best of my memory it was like dual 560s + 1 420
i havent ever done a full loop because scared of leaks and never cared enough
500€ is a reasonable 3090 price by my standards
your standards are reliant on someone giving you friend pricing
4090 definitely not worth it for that much
i think skipping a full loop wc setup straight to phase changer [air conditioner]
is the move
depending on your tastes
it might be in an industrial sort of way
calyos thought about doing a consumer unit at one point but idk if they ever did
they do rack phase changers now tthough
I'm all for performance-first
Aesthetics are secondary, because I cannot afford any
And that is how I ended up with this thing
that's what i was thinking of as soon as you mentioned STH10
We will soon relase controler software for - Windows, MAX and Linux where users can control fan speed, LCD light, brightness....etc
Yeah bro sure """soon"""
I don't think the company is alive anymore :D
they had this unit
i wonder if any of these ever got into production and actually sold
lmfao
peak
-#
Also soon that will increase by 10TB on storage
LOOOOOL there was a kickstarter for this case
and it was funded
back in 2017
and i just checked it
most delivered kickstarter
send me the drive so i can fill the remaining void in my heart without buying anything else in my state of the art managed box of horrors that i keep in the closet
Nah I keep it
Send me that tower instead
theyre all sas which is why one of my sata drives isnt in there
Drives
cmon man i just want you to search for a pci device without having to do it manually dawg i just wanna be lazy cmon man
Would a fakeroot/chroot suffice for whatever you wanna do?
it probably would work tbh, i really should just explore using that box first
You could also try running LXC-
That'd do it id think.. that is if lxc can be ran without root
i am so fucking dumb man lmao
now that i actually look at it
the thing im trying to run on a vps is one of the already pre-configured installable applications on this box
at ultra
so even if it needs root it's fine because they already have a preset install for it
derp
welp thanks @proven merlin you made me take a 2nd look at it before writing it off again and turns out im just 
Wtf 30TB
What do u store man😭 😭 😭 😭
Models and datasets and streams and audio and stuff
NeuroSynth models are like half of my /home
Rich person mode
fast is fucking expensive bruh
they gave me that $50 in credit and i used /fast to see how truly fast it was on something relatively par for the course (probably building rocm ngl) and
i think i burned that $50 in under half an hour
you'll enjoy the current storage situation here:
32 tb in that tower of drives (after 2 of them are non-storage but sacrificed for redundancy reasons) + my other enclosure on the mini pc which is 24+ 8 so another 32 tb, + 4tb nvme (2drives) in mini + 2tb usb(3.2) external on that one for another 6tb -> 70 tb, + 11tb in main pc as of writing for 81 tb, plus 2+2 for 85 in 4th pc, + 4 tb in laptop
89 tb
just a couple more to get to 100 if i count 0 of my externals except the one permanently attached to the mini
No problem, we all sometimes overcomplicate stuff-
Meanwhile me chilling at ~9,75TB in my desktop- and 3TB in my main laptop and well a buncha laptops with like 500gb .. i have storage, just not central

honestly if i werent a moron with what i store ~10tb would probably be fine
(my desktop has more storage than my server lol.. actually my laptop has the same amount.. huh)
I'm coding from phone
These alone would overfill 10TB
Using Acode running cli is funny
Honestly same tho. Tho i could easly shave off 2TB ofy my current dataset if id ever go and clear it out and sort it lol
Are caveman extenison really work?
The extension that make your agent use less tokens
Reducing things
2TB dataset
NeuroSynth is only in the 1-2GB range
speaking of used storage i desperately need to let Czkawka/krokiet finish running on everything but holy shit does it take forever
Ive not kept so many things to just regret it later.. site going down, or simmilar.. since then everything ZFS.. apropos i should see if that experimental zfs windows driver works as a windows root disk..
i have only seen people chatting about it but i assume it's like most fotm sort of things and is better/worse highly depending on what you're doing
ZFS 
exactly
the stack of drives i posted a pic of earlier is currently hosted by trueNAS (zfs only i think?)
1 PB of storage 
My bad, not that kind of dataset- meant dataset as in "data i copy to every device i own" and thus every time i have to wipe it due to no space (my phone has 3gb free atm.. its next) i backup everything blindly without sorting it and thus my dead data pile keeps adding up.. zfs dedup would help but urgh.
little less than 1/10th of the way
Looks interesting, but no time to try, cus I'm gonna sleep
I litterally dont know how i survived without zfs.. i wanna see a hardware raid zfs card.. perhaps an fpga based one..
i think im gonna keep zfs but move off truenas when i wipe the server

I'm gonna learn chinese bro
i think zfs specifically says dont run it over a hardware raid setup doesnt it
Gl
Caveman speak chinese
10TB HDD, filesystem suggestions? 
I mean, i use zfs on proxmox.. it works.
it works yeah but i thought i saw it works best if zfs can manage individual drives or something
Yes. Thats why i want a card thats "its own zfs hosting thingy" that removes the ram and cpu load out of my systems
Zfs does not work with hardware raid
It's software raid
Why do you want hardware raid?
Friends don't let friends use hardware raid
A zfs-accelerator is what i want
this is fun but keep in mind this doesn't actually get you much since the vast majority of tokens are thinking/tools which are unaffected 
I know it's software, i want a hardware version of it-
OpenZFS
But why ?
To reduce my cpu load and ram load
And what features does it have that make it better than the alternatives?
You can get an optane drive for l2arc
All the compression, all the modes, ram read cache, ssd write cache abilites, snapshots, virtual disks, quotas, stability, error checking..
What am i missing?
Self healing, raid, snapshots, datasets, etc
FAT32, 5 times 
Yes but im mostly cpu limited in many systems, and id just rather have a seperate card for it.
Something that can be updated ofc
I don't think FAT32 is gonna allow for files big enough for my purposes
good
If just I had enough drives for a RAID
I mean i already trade a pcie slot for a sata adapter to connect my disks.. id just want the zfs calcs and logic and caches to also be run over it at that point. I don't think it'll happen but yea.
That's the right way
And how would one make a ZFS partition correctly with all the features then?
IT? Is that like HBA?
Yea
Ah yea
IT mode stands for initiator target
Zfs doesnt mind that, aslong smartctl can read data iirc
you take a course in zfs 
When you flash a raid controller to it mode it basically becomes a hba and then you can do software raid via zfs

Ah.
this
Its fairly easy, but i always consult the decent documentation
https://docs.oracle.com/cd/E19253-01/820-2313/gaynr/index.html
Why is the thing so ass
I just want a big partition
a pool is just a group of drives
What about BTRFS?
i like btrfs
You have Pools, vdevs, and logs iirc.
i use it for my system drives on my main pc
i like the poor man's zfs (btrfs)
Me no need groups me only have one 10TB drive
Nice, but some modes are flaky (raid5/6 iirc)
mostly same features, with some caveats that you have to keep in mind but usually don't really matter (namely, defective RAID)
Me no have RAID because expensive
you can set up a partition on a pool
it's kinda weird
you start with disks/vdevs
combine one or more to a poool (zpool) which is just your storage blob
I think ZFS is too advanced for my purposes
it's just a giant blob of space
and then you can set up datasets within the pool which is kinda like volumes
BTRFS more manageable?
Iirc, for zfs all you do is
zpool create MyZFSDiskName sda
As example... You can also add -o options like compression=lz4 and such, but most options you can set after creating it and new data will be added with those settings. The only thing you should do from the get go is ashift=12 for 4k blocks.
you can make zvol that show up as a block devices
it is 
it's a good fs but you're probably gonna spend a day studying it before deploying it anywhere
use bcachefs 
there you go
Not the drama FS
live for 7.0 surely
It is very different from normal fs's but very flexible and able to be expanded.. and you can do "fake raid" on your one disk, via copies=2 and such.
How is zfs complicated to setup?
It's super easy
compared to any other file system that's in the kernel it is

granted it also does a lot more
Any other fs you do mkfs.ext4 or something and good is
can I run mkfs.zfs /dev/sdX and get a usable partition?
zfs is actually insanely easy to set up because usually it'll just eat all the available drives into a pool which then you divvy up into datasets or zvol from there
Zfs IS different and needs a bit of work
Thats true. Zvols and zfs datasets are so damn good
(do set a 80% capacity quota on your main tank though.)
(or 500GB, whatever is smaller)
zpool create -o ashift=12 -O compression=lz4 -O atime=off <Poolname> <drive>
Its a bit of getting used to its oddness, but it its so worth post-zfs-trauma
sudo zpool create <poolname> /dev/sdX without any other options
Tbh id keep atime on/default and just use a slog
For ssds i usually disable it
does this immediately allow me to store files on it
yeah itll just show up mounted
Not have had issues on my ssds with atime, been using em fairly heavly and theyre still at 99%, but idk, prolly depends on workload
You need to mount it via zfs set mountpoint then the cache file will mount it on bootup
first use yeah true
Zfs remembers on boot
Also you get to enjoy zfs send/recv
And vms just get a zvol
sudo zpool create -o ashift=12 -O compression=lz4 -O atime=off tank /dev/disk/by-id/<id> to be based /disk/by-id/<foo> enjoyer
Do a system update? Do a zfs snapshot first.
Ye do use diskid
And should you ever need a fat32 partition you just format the zvol, no wierd loop mounts and such
I now run zfs everywhere lmao
On my laptop, server, storage server
Im gonna put zfs as my root for my arctix install.. i hope it goes well

I Used it on arch before once, but arctix looks like more control.. and im not willing to go with nix.. at that point id just go past lfs and deal with it, but thats because i hate myself
Arch but not bound by init
artix is a non systemd arch
literally exactly what i was about to type lmao
Youd just need to get if from aur and write your own service files
(or from source)
i am finding that
Artix Linux officially supports "BSD rc.d" as an init system as well
Lol is Claus forcing ID verification now?
Damn
On a Linux system? That would be kinda funny
Its all just C, not really os dependant
it supports rc.d system
officially
lol
Best part is they offer per package a init file
So like openssh-r6
Is what youd install for r6 service file
Yeah but still I'm a bit surprised
Kinda stupid that its manual instead of a config with auto fetch, but one day, surely.
Fair
I think it might work with windows autostart if someone ports it
Lmfao

what is macos init system
Hmmmm actually lemma add that to the todolist: port reactos autostart to artix
launchd 
launchd is an init and operating system service management daemon created by Apple Inc. as part of macOS, iOS, iPadOS, watchOS, tvOS, and visionOS to replace its BSD-style init and SystemStarter. There have been efforts to port launchd to FreeBSD and derived systems.
SystemStarter kek
A port to freebsd of Launchd lmao
Granted a bunch of Mac os stuff should work on freebsd since Mac os is basically bsd
I mean people are trying to make whats a reactos to windows for macos-
hackintosh standalone
I forgot what their name was..

OpenBSD wins due to vulnerability count
"helloSystem, ravynOS, airyxOS".. hmm
OpenBaSeD
Also the art is cool yea
There was a doctor who puffy as well
They do have a devil iirc
The chad-ahh face lmao
technically the devil has been around for some openbsd
Tho honestly, idk if i should go Artix or opensuse.. maybe openmandriva?
mandriva is a crazy pull
oo it isnt freebsd mascot
in 2026
I wouibe surprised if i have a cd for that somewhere..
just the bsd mascot
The BSD Daemon, nicknamed Beastie, is the generic mascot of BSD operating systems. The BSD Daemon is named after software daemons, a class of long-running computer programs in Unix-like operating systems—which, through a play on words, takes the cartoon shape of a demon. The BSD Daemon's nickname Beastie is a slurred phonetic (in American Engl...
No not mandriva, openmandriva.
My netbook ran mandriva by default lmao
Ngl I'm super happy with FreeBSD on my server now
Runs very well and stable
Actually.. it was called mandrake by then still iirc
im finna run exclusively PCLinuxOS
while distrohopping a while back i briefly slapped ghostBSD on my main pc
ghostBSD worked great with a 3080
Oh i remember GhostBSD as well
But it's basically FreeBSD with a gui pre installed no ?
(thanks to nvidia-open being so portable
Lol
Freebsd'nt.. Schrödinger'sBSD
Apropos DE's and such.. any recommendations for minimal ram useage?
Im currently failing to go slimmer than MATE and it pisses me off
debian with lxde
Im sitting at 300mb with it
i dunno i watched https://www.youtube.com/watch?v=sxeRCpg9mfc once and ill have you know that makes me an expert
Get 20% off DeleteMe US consumer plans when you go to https://joindeleteme.com/actionretro and use promo code ACTIONRETRO at checkout.
DeleteMe International Plans: https://international.joindeleteme.com/
Tiny Core Linux is a magical modern Linux distro that can run on literally a potato.
⭐️ I auction old tech from my videos on Whatnot! ht...
(juicy 2.something kernel)
Openbox 
It doesn't run on my first laptop, due to 16mb ram + 1mb vram.. not happy (lol)
Or Fvwm if you like retro stuff
the humanity
I actually tried running it, but i dont really save anything and the lack of good panels drive me mad.. what panel is minimal, has auto app-detection and is "windows like"-?
Yes
ace of base kekw
What do you mean X only stuff is great
Xlibre hell yea












