#programming
1 messages ยท Page 381 of 1
hmm
It would be faster that a single drive as well
What a steal
Ooooo, nice
time to buy these for myself
that's my current roadblock i need to produce a large box for a billion drives
he has 5 lots and this is not even the best deal son
damn it
24 300giggies lmao
this is overprice
but 24 make laugh
and also theyre 15k rpm
C H E E T A H
10k rpm, 5.4tb, 45 buxx
god damn
I know I bought like 20 of the 150 GB VelociRapter drives for cheap and redundant caching for sone future projects and also just in case.
300 MB/s, $10 each
They came as lots of 10 on eBay
yeah i didnt know about this meta until literally like 3 hours ago
and i have obviously been
b r o w s i n g
have to have sas controller that's the only consoomer issue here
I won't buy Seagate, though
I don't trust their quality.
While I have a few laptop HDDs that are Seagate, I will never outright purchase their drives.
Almost all my NEW drives are WD or Kioxia
fucking rip samsung
o7

seagate ate them and enshittified
i have multiple 15 year old seagate 1tb still goin
erre
sammy
Samsung used to make the Seagate drives?
naw, samsung made banger drives on their own
spinpoint f3 etc
GOATED
seagate bought their consumer hdd division in like
2011 or 12
Oh, I see.
I didn't even know Samsung made HDD's, at all
Knew about Seagate, WD, Toshiba, HGST
my only scurrent seagate
i have a 12 year old hitachi deskstar 1tb still going strong as well
i recognize that This cannot continue., see you soon
Ooooo, that's interesting
hahaha
i yoinked this out of a 24tb enclosure
black friday last year 240 bux
10/tb made me happy
Well, hopefully it lasts
Seagate has the nasty tendency of being an INVERSE Bell Curve. They either last forever, or they don't even come close to reaching MTBF
and is mechanically basically an exos x24
When I get office or enterprise grade HDD's, I get either Red Pros or Golds (WD)
the model number is the same as this https://www.amazon.com/Seagate-Enterprise-Helium-Sealed-ST24000NM000H/dp/B0CNSZ1B5B
Maximum Storage Capacity for Highest Rack Space Efficiency. Seagate manufactures hard drives that specifically address the demand for hyperscale cloud scalability. The Exos enterprise hard drives are the highest-capacity hard drives in the fleet. Market-leading HDD offering the highest capacity a...
24+ on their desk expansion units you can find all over the place are apparently extremely likely to have this
ddue to hedt
helium filled and shit
im using wrong acronmy
I do have several WD Blues, though, in a drawer...
Need to figure out what to use them for
6x 6 TB
2x 4 TB
acronym
that's the question isn't it
the drive above the seagate is an 8tb toe sheeba
fucking dusty in here apparently if i actually look wtf
the 8 is like
empty
i moved everything to seagate and only keep a snapshot of specific shit on there
then i have too many nvmes
which is great because i just bought more of those

efficient layout with this new windows install especially when i immediately crammed c (nvme) full of a bunch of shit that does NOT need to be on this fast of a drive, but must reinstall to either rocky or almalinux on H
c f and g are nvme
err
c f and h

futureproofing for the inevitable nand shortage 
i think i got the technology
now how the hell do i target it at something else and not the quest...
how do i stop it from wiggling when its close to the target point 
it sounds easy but for some reason nothing works or it gives walk error
whatever i'll fix that later 
https://youtu.be/WQKSyPYF0-Y have you guys ever tried just recursing again?
I believe Iโve set a world record for the deepest recursive function call: over 17 trillion nested invocations. Itโs literally the most profound thing I've ever done with a computer.
05:06 The stack grows up, not down!
10:52 FUSE
15:41 NULL pointer? No problem!
20:54 Kernel patching
21:22 Far call to 32-bit mode
26:28 Ideas for how to recu...
no you don't get it.. the 7 years refers to when RIOT games bought it, everything RIOT has done has been rolled back
Silly AI butt kisser spotted!
aawwa
just think about these poor immutable JS variables
unsafe 
yeah... to unlock the .set() and .remove() methods
I believe Iโve set a world record for the deepest recursive function call: over 17 trillion nested invocations. Itโs literally the most profound thing I've ever done with a computer.
05:06 The stack grows up, not down!
10:52 FUSE
15:41 NULL pointer? No problem!
20:54 Kernel patching
21:22 Far call to 32-bit mode
26:28 Ideas for how to recu...
recurse
i have been going through a time
I give up, Reaper is completely unusable
I'll just stick to LMMS unless I specifically need something Reaper has
lmao why
I don't know, I just couldn't get it to sound good in Reaper the same way I do in LMMS
oh that reaper
i was thinking of something else
i mean that's probably a lot down to familiarity ig
๐ Introducing Qwen3-Max-Thinking, our most capable reasoning model yet. Trained with massive scale and advanced RL, it delivers strong performance across reasoning, knowledge, tool use, and agent capabilities.
๏ธ๏ธโจ Key innovations:
๏ธ๏ธโ
Adaptive tool-use: intelligently leverages Search, Memory & Code Interpreter without manual selection
๏ธ๏ธโ
Test-time scaling: multi-round self-reflection beats Gemini 3 Pro on reasoning
๏ธ๏ธโ
From complex math (98.0 on HMMT Feb) to agentic search (49.8 on HLE)โit just thinks better.
๏ธ๏ธ
๏ธ๏ธ๐ง Think deeper. Solve harder.
๏ธ๏ธTry the adaptive reasoning experience now: chat.qwen.ai/
๏ธ๏ธ
๏ธ๏ธCompletions API:ย modelstudio.console.alibabacloud.com/ap-southeast-1/?tab=doc#/doc/?type=model&url=2840914_2&modelId=qwen3-max-2026-01-23
๏ธ๏ธResponsesย API:ย www.alibabacloud.com/help/en/model-studio/compatibility-with-openai-responses-api
๏ธ๏ธ
๏ธ๏ธblog:ย qwen.ai/blog?id=qwen3-max-thinking
๐ฌโฆ
98 fk
highest november human score on hmmt feb was 70.68 ;_:
man im still fighting qwen image and glm imge i dont have vram
not that that thing is gonna be local
Noice
wonder if theyll do a distil to the current crop of locally runnable qwens
Searching for technical drawing/3d model
But I've searched orange pi 5 and it doesn't have any detailed model lol
They have the schematics on their website
Doesn't give me a 3d model tho
https://grabcad.com/library/orange-pi-5-max-simple-model-1
This one seems to be good
Yeah but that's simple model
I was talking about more detailed model with all the caps and whatnot
caps = capacitors
rip in peace samsung hdd division
worst picture but im not taking it out of the system it's in rn
"DO NOT COVER ANY DRIVE HOLES"
The bolt/screw:
that was a factree option
this one is actually in a sliding mount thing that does not screw in the to the side holes at all
im a good noodle
Lmao
managed to cast this spell without it being equipped 
but targeting still sounds like a challenge
for me biggest thing is that you need to figure out an objective to pursue
you should also have a general baseline understanding like, what is a programming language, how does an editor work and interact with what i'm writing (or how doesn't it), why are these words getting highlighted in funny colours and what do they mean, do i understand a bit about how a computer functions & processes information, make sure you're good at troubleshooting or at least don't instantly give up if errors appear (you are going to have to troubleshoot constantly so if that makes you give up you're not gonna get far)
choose a language that doesn't look terrifying to start i think, classic python or something (you can start on any language though and some would recommend not starting on an interpreted language like python because it abstracts a lot away, but that can help people learn and definitely helped me) follow some courses and introductions, purchase a book on it (chayleaf approved method) and just follow it, book learning is a bit underrated these days probably and it has worked well for people for years
once you get through like "ok i can write a hello world and follow an example program" or something then it's just
and try to get something you want to make going, imo
im sure some learn differently, i took courses and stuff from middle school through college and knew a baseline but i never LEARNED learned until i was in a position where i knew what i wanted to make just not how exactly and i pretty much had to set out to make it and learn along the way through trial and error
might be inefficient and im bad at explaining but yeah
idk
i do be typin
@maiden geyser illegal react while lurking
penalty awarded
you should start by making operating system in assembly from scratch
terry o7
literal 1894

after all the yappin above also make sure u are comfortable with math because you gon be more comfortable with it if you arent, it's kinda not required i guess but it helps with logical thinking
the other reacter is 15d sprout, state of that
15 days is too fresh off the vine to roast
in college some of my mates quit after a year because of the math, they hated that part even though they loved computers.
yeah math really helps in a lot of places although it does kinda depend on what you're doing as far as criticality
at very least you should be prepared to learn more if not highly ๐ค enthusiastic about it, because youll likely get forcefed the concepts necessarily to get through certain problems
but there is a good reason that compsci makes you get ur math credits for a while
ye
and that reason is they like tuition kek
jk
I heard someone say "googling it" is rather useful skill if you want to get a job. And with AI, you can vibe code, I guess? Not the perfect way to start, but is a way too
troubleshooting
yuh
with ai it's so easy to have it be a detriment to learning though, like if you set claude code on autoaccept you're not gonna learn shit
if using an AI as like a tutor some like gemini have a tutoring mode so like if you ask it a programming question i guess it'll kinda guide you through how to get the answer instead of just shitting out the completed codeblock
Mhm... no. Vibe Coding is poison for becoming a programmer. All you are doing is offloading the parts you want to learn, want to know.
i havent tried gemini's learning mode for it but i would imagine IF planning to use ai you shold avoid anything like claude code/leaning on copilot in your ide etc because it will not explain everything it's doing, especially not why, without being prompted to
could set it up to i suppose
treat it more like a stack overflow or something (rip)
I mean, I think you should touch AI only if you already have some knowledge and experience, otherwise yes, I think it'll do more harm. I started way before and don't really use Copilot much (only to autocomplete comments, lol)
i heard notebookLM is good for it but idk
stack overflow can rest in piss
lolol
But for a job, I mean... any wae is a wae to get that bag
based on twitter jr dev is walking into the claude emporium and spending your entire day redoing your tooling
As a Beginner, you could treat vibe coding the same way you treat another poison: alcohol.
In controlled moderation, and don't drink too much and make everyones live around you miserable.
a support line of last resort
"hey guys I'm completely new to programming but I can write hello world in python, could someone give me a step by step instruction list for re-creating neuro completely from scratch? I'm using my 15 year old laptop with a celeron and 4gb of ram, will that be enough?" 
cant figure it out at all, cant find an answer online at all, etc then ig ask the oracle but ask it to explain
recipe for success
sure, why not. List would look like this - 1. fuck around 2.find out
"also, when you answer that honestly and genuinely try to help, what i'll respond with is i will argue the fuck out of that answer and will insist for some reason that neuro is very simple to do if you assholes would just fuckin tell me how"

rewriting your company's codebase to look like an indie game developer's code is the only way to keep your job in the big 2026
Why must you guys use a font that hurts my eyes to read
average stack overflow query
On the other hand, it is good ppl are interested in trying, even if it's clumsy. I'm sure many of us had very weird view of things at first (like thinking that if you write do_cool_stuff_please() would actually do something). Maybe it's just me though
a dev knows he's a dev when the led lights up.
my coding skills have absolutely been deep fried due to my heavy reliance on ai tools
claude pleese i know i only have enouh creditz for haiku but please this 500 billion loc written and rewritten half ass 30 times over the last 30 years just to barely work, legacy .net and powershell 4 based application with multiple unrelated responsibility, production critical spaghetti dinner of a codebase (there's 6 copied repos for some reason eaach with 10 open branches with shitty names that don't have any indication of why they are open, what their status is, or what they were trying to do, also nobody has ever done a pull request, they just copy it into main when done breaking things. all branches appear to be doing the same thing but we're not sure which one is actually in prod rn because the guy left)
plz halp
grok is this real
gpt summarize this in one line
sure, here's a summary:
aAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
yeah, life is hard
yolo-to-main, i like that strategy
it's ok because as soon as you really get situation prio has changed and you're moved to some other disaster 
Hewwo
its time
productive use of IDA Pro.
are you yeeting the entire skyrim se in to a decompiler
the hdr burned my eyeballs out i cant see now
that is an expensive license
OH theres free
kek
i was like no way people pay 1100 just to not use ghidra
the free IDA doesn't have decompiler.... just disassembler. and it's limited in so many ways like not being able to save. But good for learning if you want to get into computer security field.
(or at least it didn't.)
PC disassemblers (x86 32/64) only
2 cloud-based decompilers
x86 - 32
x86 - 64
License type:
Named
Operating systems:
MacOS
Linux
Windows
perfect for basic hobby stuff.
you can also pay $1/day for 1 disassembler of your choice:
x86 - 32 & 64
ARM - 32 & 64
MIPS - 32(+nM) & 64
PPC 32 & 64
ARC
RISC-V 32 & 64
2 cloud-based decompilers from the same family matching the selected disassembler:
x86 - 32 & 64
ARM - 32 & 64
MIPS 32+nM & 64
PPC 32 & 64
ARC
RISC-V 32 & 64
Scripting:
IDAPython API
IDC scripting
C++ SDK
Debugger:
Local
GDBServer (for remote debugging)
License type:
Named
Operating systems:
MacOS
Linux
Windows
wouldn't want to disassemble a AAA game though.
i dont understand a thing 
i did something in this before but i forgot it completely
i guess its time to rewatch that guide
i would just expect the output to maybe be neat to look through but that executable id expect to not be trivial to do much else
it's gonna be obfuscated when it was packagd and shit
there's a lot of info in the disassembly once you know how to read it.
depends what your goals are
well, it mean, the code has to run, so it's just you dont get any convenience and get a fat set of ??? names on variables and shit
based on the function names i'd say it wasn't very obfuscated
but you are looking at the whole thing
i need to find how this spell does pathfinding and spawns these magic balls on the path
well, it also supports beeg modding
so i'd imagine they don't button it down that badly
need to guess the function names then
there'll be structures which contain the C++ objects with pointers to all teh functions they use
via slam head on wall technique 
fuck im not even sure this is actual exe for the game i have
bethesda did a little trolling converting special edition into anniversary edition and now its ae
SkyrimSE.exe is it
oh
rip
what does vortex (modding) select lmfao
googling says it's probably still skyrimSE.exe
they fucked up all addresses but smart people already figured it out
but the guides were made for normal se so the addresses are all off

i guess if i figure out how it works i'll still find what i need
but its hardcore asf for me
at least a decent starting point compared to yolo
in terms of whatever they documented along with the original addresses when they found them
know what to loook for
ish
@stray dragon was running a model in lm studio and went into ram so i finally reloaded vivaldi and this was the drop from 500 closing xd
lol
vivaldi loves vram
not really
i don't think browsers really have a cap on how much they can eat, assuming you have enough tabs that each use up a fair amount of ram
oh
it also dropped 70 gb of regular ram lmao
wait i gotta check this now
but i had OSS and also qwen image in q6 loaded at the same time by accident
damn it actually does use up a fair amount of vram
yesh
i dont remember like
chrome using th at much vram
idk why it's so hungry for it
i had like
one youtube tab
ey
transformers v5 from hf
Js use vivaldi its better
something something ram that isn't being used something wasted
they do use vivaldi
true
didn't know what was using the vram before though
So batched got about 50% faster, neato (fuck you for comparing unrelated numbers whoever made these)
interesting to know that it was the browsers
Well prob not the browser then
it is 
it was certainly the browser
the gpu renders most of each webpage if you have hardware acceleration on so that'll eat up vram with enough tabs I guess
this is me closing my vivaldi

Oh I never got any vram spikes tho or even any usage at all
yeah, i basically never unloaded any of the tabs
they were all open
That'll do it 
...
mine is so fresh so clean now that i killed it
vivaldi tab stacks my beloved
Js why
why close if not need
i also enjoy blasting the vram like that and then seeing just how far it drops closing it
makes me feel like a new man
i close to keep organization
that's about it
vivaldi handles many tabs quite well
Damn I rarely ever hit 500mb of ram with my browser and i thought I use a lot of tabs
how much ram total do you have
16
I mean I don't understand y ppl even have 32gugs or more ram if they aren't doing like ram heavy tasks like editing or ai stuff
192
..
yes
i have 128 gb total
training
IN THIS ECONOMY IS CRAZY
damn i was just about to pre-emptively answer the question before you asked
i got my 192gb for $333 a year ago
Why ๐ฅ ai is ass ion understand ts
the first 64 i got at the start of 2025 and it was like 200 $ and then i got the 2nd 64 right when it started going up and it was like $300
and now that kit is like
$1000
I can't even get an SSD for that now ๐ฅ
i think 64GB is probably plenty. it really helps with keeping the models in memory so they load faster.
64gb wasn't enough for me because docker in windows is very hungry
Do u even need swap ๐ฅ
all the way in vram anyways
nah i dont but wandows didnt give an option to config it when i installed it on this drive like linux would
and i havent been assed to do anything w/ it
Oh so you did buy a 9700 huh
yepper
Windows?
Why
HIP
Hip?
old repo
damn i was hoping it explained anything but the readme is empty
Ye lol
super repo 
i wanted to be able to do both
i havent put my 3080 back in the computer tho
This chapter provides an introduction to the HIP API.

cursedd
rocm supported a higher version on wandows for a time when i reinstalled as well
i believe linux has gotten rocm7 now though?
almost had a silly virus on my pc
i need to reinstall my linux drive so i keep using winblows
Last time I checked hip for nvidia was unusable
kek
Wonder if that's changed
as far as i know it's significantly more reliable on win than linux for HIP
atm
also ZLUDA
doesnt love linux or i believe work at all 
Pretty sure it always did mb was reading the repo
Oh
it didnt have official packages or something for a while
i tested it and on arch installing normally it would only give me rocm 6.4
while the windows install was getting 7
there was some scuffed way to like
pull the files
and build 7
didnt fuck with it tho
Lol
i would normally expect amd to be ahead on linux myself so idk why that was a thing for a bit
Ye
i know my mini is still on debian so it's rocking rocm 1.0 or some shit
jk it's 6.4 i think
Classic Arch L
Not really arch
i am about to wipe my arch on this pc but i am idly debating rocky linux or alma
lol
useless choice
but
i remember linux drivers for my GPU used to always release about 3 months after they released a new stable version of Ubuntu so it was always at least one major version behind and usually didn't work properly.
bro can't take a joke
Rocky, alma? Never even heard of em
they are basically rhel
Fedora
oh shit you know what
my mini is running fedora
i reinstalled it like last week
ive been booted into tails fo a week
What do u need the distro for your main os or your server
Lmao
my server is over on the other side of the room
What's your server on
right now it is running proxmox
For me Fedora works great for both server and desktop
SERVER?
fedora has a server edition
Oh didn't know that
mhm
We're so glad you've decided to give Fedora Server a try. We know you'll love it.
it actually has like
multiple
i never use the server edition tho
What
i cant figure out why the fuck we need so many versions that are all server esque
just fedora workstation on both
well tecnically thre KDE one but that's basically just workstation but KDE
Your gonna get alot of idle usage tho
huh?
anyways my mini's running fedora but the secret is that it's actually running nobara
don't really give a shit tbh
Damn lol
Server is for if you want Ubuntu Server except Fedora
IoT is for tiny little toasters that can't handle Server
Cloud is for if you don't need to support actual hardware and all the drivers and really just want kernel + package manager, kinda like Docker container except VM
CoreOS is a special locked down container runtime thing
So people are able to fight over which is the best for no reason mhm
tat sounds like a noble cause
fedora living up to its name really with this one
i didnt know rhel has a 16 system free tier
license
Ubuntu doesn't have a CoreOS equivalent I think, but they do have IoT and I'm pretty sure also Cloud (just distributed by the hoster itself, not separately) 
rhel
-> centos stream (basically builds w/ upstream previews to mirror the future of the next minor rhel release apparently)
-> rocky - centos guy malding when they did stream so he started a 1:1 rhel tracking distro for non enterprise
-> alma - ALSO 1:1 tracks rhel builds but it's not fully tracking every build it just keeps abi parity with current rhel
the illusion of choice
stream is like the middle of rhel and fedora
boy which .0.0.1 update granularity does a guy want phew
ye Rocky and Alma are pretty much the same AFAIK
Stream is just bad
and they're both just budget RHEL 
only difference is binary compatible vs like lterally just rebuilds on every rhel build
so like
maybe a couple days of difference sometimes lmao
not having pacman cripples my spirit
every time i do this on my main and swap to something else i end up reinstalling arch withing weeks
but
we'll see
I am one of those people that stim by running pacman -Syu every so often lol
pacstall kinda fills the void but i also hates it
oh i have placed my cursor in a terminal
pacman -Syu
muscle memory
then nobara is like
wha tthe fuck does that mean
lmao
dnf + rpm fusion is pretty beeg
nix has the bestest repo of them all technically
when im on anything ubuntu/debian i inevitably end up having homebrew and then ruining my life
at that point just install nix smh
im honestly dangerously close to yolo slapping nix on here now that i dont have nvidia gpu in the box
to try it finally
every other time i've tried
also i just have an update shell alias for whichever distro i'm currently on
the fucking installer explodes
(i meant nix the package manager)
and i know the package manager but like
so true lmao, I put windows on my machine to run unreal engine, and this has happened far more than I would like to admit
feels

the NixOS installer or the NVIDIA installer thing?
alias it to winget update
just the system image install
I have having to go to a website to download the programs installer to then run the installer. and every different installer works differently so you have to check that you aren't installing some slopware game like candy crush cause you clicked "next" instead of "next but don't install that stuff"
the graphical installer might implode from NVIDIA (though it shouldn't)
manual install should definitely work though 
im sure it probably works better
it does
each time though
bwanyanya
calamares doesn't even configure anything nvidia related no?
that it like corrupts a drive in the process
it just gives you the generic mesa driver and lets you deal with that
i always just kinda figured it was somehow my 3080's fault
coping perhaps
i literally never have gotten calamares to finish installing even once in this pc lmfao
on nix
ONLY nix
works fine on every other distro that has calamares
i think you might be cursed
Oh he is cursed alright lol
i genuinely have gone to install nixos like 3-4 times and this happened every time
wildly different times
different kernel versions, new releases of it
it's crazy
mfw the definition of insanity
i gas myself up like i will finally do it
makin gcomputr do thing
i will make mlnt proud
i will use nixos

and then it explodes and takes one of my storage drives with it
nixos
i know the nixos installer can be kind of brittle sometimes, but from what i've seen that was mostly due to resource usage
very annoyingly
so then i boot back into arch and roll the now totally unusable drive back
and then just stay on arch
that's how nixos install goes
no
it freezes
every time
full system lock
it just hates me
i specifically change N O T H I N G on purpose too
other than choosing the drive
which is never the drive that gets nuked btw
as it freezes up it 360 noscopes just
one of the other drives

almost got mad enough to disconnect all my drives and run it with only one
but that would mean removing my GPU and taking the 69 pound metal cover off the 3 nvme that are hiding
which is such a pain
I'm amazed that you didn't consider using the manual installer at that point 
well, usually im booted back to arch and fixing my drive
and then i give up
and the next time i try it will be like
months later
so it's like
surely calamares is working

technically it shouldnt be hard to remove these covers, but they are on both sides of the gpu like this and like 1/3rd of an inch of thicc copper for some reason and i hate moving them because there's thermal pads on the back aand im afraid im gonna fumble it and it's gonna fall on the board directly onto the lynchpin cap that will take the entire board with it
yeah no fuck that
better anyway
using GUI is stupid when you have to edit Nix config manually later
install is just:
- partition drive (just use fdisk)
- format & mount partitions
- generate & edit config (have to know how to do anyway)
- run
nixos-install
not complicated
so the drives stay installed lmao
i guess i could disable them from uefi
but then i'd have to re-enable them which is over 15 seconds of effort
@rough bloom have you ever had to put 16 seconds of effort in to something
it's impossible
So arch install procedure
options 1 & 2 optional with disko 
(this is a lie you still need to run a command)
but I don't expect new user to learn that, extra dependency
me when i configure my partitions and second guess it 300 times, have it right, and then somehow forget how to use a computer
basically ye
i turn in to a blubbering idiot the second im staring at 5 drives and get so fucking paranoid i'm gonna wipe the wrong one even though i literally know which device path it's on
i guess making an official nixos module for something like disko would be too risky?
idk why else it doesn't exist yet
i know disko is a nix thing but each time ive gone to its repo i get stunlocked by the stunnig artistry
holy shit
yeah it's beautiful

i like how it's been in the repo for like a decade and they really fucked up the a in declare 10 years ago and have never cared to redo at least that a
but see then
i get baited here
because that's the level of effort i then come prepared to do anything wth
and then i just give up like a baby
this is when i am pondering hmm, calamares exploded, manual install? 
bee zee zee zee
wrrrrrrrrrrrrrrrr
wrrr
[the bee grinder]


is the an AI video model that'll make your video loop better by adding some frames at the end?
you just would put the end as the start and the start as the end
on any that support frame ->frame
many do
wan does
its l;ocal
model for joining two videos
join video to itself
@azure lynx this was what i was referencing (free)
kling will do it im sure any of the paid cloud inference type models will do it
But what if, the video become mobius strip
the video collapses into a black hole and kills us all
Now I wonder how old film would looks like as a mobius strip
it'd flip over left to right when 1/2 way around.
you could probably do a cute little animation where the character walks from left to right, does something, and then it flips so you see them walk right to left and do the thing flipped.
and it'd loop
that reminds me i was in a daze angrily writing shit to post on facebook that will make like the worst 30% of my hometown friends remove me (win) for the first time in my life i gotta finish that since i already blasted 12,000 characters into this poor notepad
once I do that i will feel refreshed and thriving and i will download the most recent nixos build to try the slot machine now with no nvidia card installed
it is so funny to me that this is a microsoft repo and moe architecture

vadeul is a spy https://github.com/microsoft/tutel
Tutel MoE: Optimized Mixture-of-Experts Library, Support GptOss/DeepSeek/Kimi-K2/Qwen3 using FP8/NVFP4/MXFP4 - microsoft/Tutel

how romantic
hi
time for 
i woke up not too long ago
just came back from mario kart
we're getting all the trophies

they cancelled my order
of the orange pi ultra 8gb
they only have the 16gb in stock, but that one is double the price
ill see if i can get one for not too muchj more
the one i was getting was 180 euro, the other ones are 330

Sus
i found one for 230
They probably have the stock but forgot to update the pricing
205.56 euro with a coupon code it gave me
sure
everyday we stray closer to overbudget
litle extra for me too
I keep telling my mf ai to NOT PLAY THE SOUND EFFECT TOOL
AND IT DOES IT EVERYTIME
WITH PIPES
it wont stop
I have to distract it with keys to get it to stop
jangle them
i hate it
so much
delete the files. turn down the volume. replace the files with silences
just not prompt enough
that'll just remind them they can play them.
you probably just need to distract them with a different task.
what is the use case for this over a mini of a similar price
other than just connecting to arduino and shit
it is cheaper thna a mini pc

they sell them for 200, but no ram
yeah. more beefy than the pi
they're not nearly that cheap here
theyre that cheap on ali
get one from a few years ago
and what will you use that for?
world domination
cuz no mini pc is gonna render this at 3600hz
hdmi
why does it have a pi if it's going to be run from a computer?
yea bbut like the param of the video
it wont always
and if it's not being run from the computer why would you need HDMI in then?
the pi renders whatever it is that you're rendering?
790x790@179.126hz video, but thats 4225hz 1 bit footage that got compressed to 179.126hz 24bit footage
sounds fairly trivial
not if its being rendered live
if that rockchip is able to it any of these ryzen apus can
that rockchip wont be able to
usb capture dongles have gotten pretty cheap
you could brute force that on like 2 cpu cores
thats the whole point of having hdmi input
real time
of all the dumb takes i've seen on AI so far..
that's 2.7gbps throughput
the throughput isnt the issue
this isnt normal rendering
Using emotional arguments without justifying them isn't clever. Makes people think it's an opinion.
its double helix slice 1 bit 4225hz footage
notch still exists 
one shader will do it while asleep
brother
to do borth
holy shit you're sam
what the helly
anyways
send me whatever it is you're running this on and i will send you a video of the mini pc i have right here running that real time
ye that doesn't exist yet

it snot jsut about the rendering
280hz 
you need to combine the footage to 24bit 179hz in realtime
each 1 bit is a slice right
ye
so you smasheach bit plane into the output which is liek
this assembly shit is so shit 
23 slices of color depth & intensity
๐ฟ
no
what is the helix being simulated like a dna thing
its jsut 8 bit RGB, so 24 bit total
i typoed 23 lol
double helix ye
volumetric display
without the actual hardware the output will look shit
imma get a hologram fan and pretend it's a volumetric output
it doent work the same way since there's no depth
and i doubt the hologram fan will do 179hz nor 4225hz input
i refuse to believe they didnt make some get_path_to_thing(thing)

that was an extremely tongue in cheek thing from me btw i dont want a fan that will be in my waylol
are you using an fpga to actually render it
nope
what is rendering it
just my desktop
why does it need the pi again?
the orange pi is more like, in charge of storing video, displaying video, and switching between hdmi input and between using prestored clips
the only way i could see it being an issue would be if it were not a pre batch render issue
i highly doubt i will need to use live rendered stuff much at all. but i want it o be possible
i imagine it would be much simpler to directly send the data from the PC to the projector
that just isn't as cool tho
i mean i dont think the pi WONT work for that (because it's not a difficult task with like 1 gpu shder lazily farting it along as far as i can tell)
butthen youhave a pi that's fairly pricy and it's like
damn
the projector has enough storage for a grand total of 40 frames, so the pi is needed for storing stuff when its not connected to my pc
if it's not needed, that saves you 200 euros
reeeee i need to see this bottlenecking when you do it
please ping me when you have it going lmao
i just want to know what im missing
the pi also has like 3 camera ports and whatever
yeah i mean the pi is good for microcontroller stuff and whatnot
i want it to run standalone, not really needing my pc
i kinda want one too for that
since i need the hdmi on my pc for vr
it's not that the PI sucks. it's that it's overkill and redundant
thats what makes it good tho
your money
i have money
we're still under budget
the end goal is for the pi to procedurally choose what clips to play
its gonna be the brain of the machine, not just the in-between
it also serves as the thing controlling the 900rpm motor
i dont understand why you need hdmi input if it's pre-rendered
can you not just transmit it literally any other way
cuz what if its not pre-rendered
but it was you said
i said it can be both
they're asking why you need HDMI-in
for live rendered stuff
just plug it into the computer
thats the plan
but what about like a pi pico with dma
the projector
no
i want it to still function when i dont have it hooked up to my pc
not by redoing all the cabling in the back
2 cases:
- you use it with a Pi with no HDMI-in without the computer.
- you plug the computer into the projector directly
why do you need HDMI-in?
i would need to like seee a vedal tutel style diagram
im not ognan be unplugging the pi when i plug the pc in
i still need the pi as the motor controller
you don't need HDMI-in. that is all i am arguing.
if the reason you got the expensive model is coz of that feature, you can avoid getting the expensive model
its only 50 bucks more, and with that i can jsut unplug the hdmi without needing the replug the pi.
and the pi needs to control the motor speed anyways which needs to be synced to the projector, so its already gonna be interconnected
good luck with your design.
i want it to be a fully enclosed system
no need to get to the inside to mess around with cabling
its probably fine for controlling the motor, but i dont see an sd card reader or hdmi output or anything
oh they have ones with hdmi
ye
the only thing would be like
some of those devices would shit themselves with the refresh rates
i can understand what yall mean
just because theyre weird
but like
whan you buy a smart tv
do you really want to unplug the hdmi, and plug the smart controller in when you go to watch netflix?
and then for ps5 you have to unplug the smart controller and plug the ps5 back in?
no but i do not think that i would proceed in this path to gain one(1) hdmi switch
will you have a remote control to power it off? if not, you may as well unplug it
what about like
i can use my phone
a switch would work, but the price difference then jsut becomes clsoe to jsut hetting one with a hdmi input
hdmi 2.0 switch is liek 20 bucks, then i need 2 new hdmi cables
and we're halfway tot he 50 bucks extra
but the pi was what liek 250
the pi was 200
ye so thats 30 bucks
that is 170 dollars less than 200
but instead i pay 50 bucks extra and i odnt have to deal with a hdmi splitter taking upos pace
you're ignoring the price of the pi
i still need a pi
ok now i have no idea what's going on
if i was doing this, i'd first get the display tech working then worry about how to make it all one thing.
that'd mean just getting the projector and the minimal hardware required to make it function.
motor controller, and secondary source of video
the pi would come much later
motor controller would be arduino (or Teensy). much cheaper
and i have many of them
well i dont
4gb ram
lmao
i mean, its fine
then do that 130 + 30 for the hdmi switch, and + 20 for the motor controller
and we're altready back to 180

the motor controller would be the pc with a cable coming out of the usb
live your best life i mainly suggested it because buying a pi rn is kinda a rip
i didnt mean to have a whole debate 
i dont mind paying 40 bucks more, it takes up less space, i odnt need to power more different things at the same time
the power and space are my main concern here
would it truly be less space, minis are tiiiny
cable flexibl
still takes up a lot of space
it's liek the width of 2 cables lol
my man's in that hong kong stand up apartment life
you're going to want something solid underneath everything when you have the motor spinning.
its close
still dont want one that you need to manually switch tho
could probabyl make it do that via the pi
i need to see this macguyver mickey mouse imagineer ahh concept when you put it together
i have a burning need
the waht?
this entire system
lol
iam never going to understand until i can hook my peepers on it
What on earth are you planning with this? xD
volumetric display
Ooooh
Is this from a project you found online?
I've seen various attempts of volumetric displays using OLED panels, spinning LEDs etc, what are you thinking for your display medium?
this is 17cm across, and i still need to put all the power supply components into this
spinning leds are either too low res or too expensive.
oled just doens't cut it with only 480hz
I mean in terms of display tech what other options are there? You could explore laser projection with galvanometers maybe?
lasers directly into your eyes.
Yeah a nice 1MW laser to the iris is how I start my morning ;P
we're usign a lightcrafter 4500 which tops out at 4225hz
Oooooh you're using a DLP module similar to what projectors use?
how close will the optics focus and with what focal depth of field? it's quite difficult to have a wide range of focus at short distance.
except this one is for more industrial use, so you have precise control over how it works, instead of store-bought projectors where you have a fuckton of prepossessing on the final image
the plan is to use a lense to make the lightrays close to parallel
have you priced one?
i found some for 50 bucks but they werent a good size
Using blender as CAD is insane
So You'll have a collimating lens in line with the DLP projection?
Also yeah do you not have access to a CAD suite like Fusion 360 etc?
big lenses cost much more
ye
im just better at working with blender
Those lens are usually C H O N K
i use Fusion for my robot stuff. it's not hard to learn.
That's fine for mockups but if you have any need for 3D printing can it not lead to issues? I've never done FDM etc in blender
Blender is fine to just check for dimension tho
jsut export as stl
Bro, usually art stuff is designed in blender
Especially sculpture
blender is good for artistic stuff, fusion is good for designing. (or some people i know use onshape. or have edu licenses for solidworks or something.)
Ahhhh yeah you're right, I'm from more of an engineering background so I'm just so used to designing prints in something like Fusion or Autodesk but I forgot blender does STL exports
The only thing I dislike about Fusion is they keep breaking the linux workaround
i've got many things i dislike about Fusion. the main thing i like is it's free and I can importand export stuff easily enough.
i think the cheapest option would be a fresnel lense, but that has the chance of looking very bad.
so i dont think i shoudl take that risk
fresnel lens is cheap. but because you control the images you might be able to workaround any distortions
Yeah you could try inital protyping with fresnel as they're cheap
Just keep in mind there's better options out there for collimation
ye
In other news, I'm looking into making a satellite tracker with a few servos and an ESP32 so I can use my HACKRF to receive weather satellite imagery haha
in theory a plano-convex lens could work
If you have SBC with GPIO lying around, just use that so you don't need the extra step of using ESP32 and host for the HackRF SDR
Yeah I have every generation of RPi with a 5 beside me, I assume I'd just need a decent stepper motor HAT?
the SDR code can be quite heavy. (at least it was if you were using the GNU Radio GUI thing)
Indeed. Or if you're talking the servo with PWM control, then you just need an opto-isolator for safety
Will a Pi5 be enough for the SDR code?
I think a satellite tracker have a lightweight version that doesn't need to do much
imagine if the dutch government sucks at coding...
oh wait... https://github.com/MinBZK
This is very specifically made for satellite tracking
Awesome does that have a linux port?
ministerie van binnenlandse zaken??
good thing its only binnenlandse zaken
suck it dutch people
I don't think so. Lemme look up my past project lol
Wait so your government open sources their infrastructure?
apparently
the tax payers are paying for it. may as well give it to them
Awesome
"nl-wallet"
ah yes, having the source code of where my money is online for everyone to find bugs in is such a good idea
i think it's a good idea. it's better than them claiming it's secure with no evidence and then getting hacked anyway
In theory yes as the bugs will be patched quickly but it needs decent bounty hunting programmes
this is truly one of their works
it needs to be more worth my time to tell you there is an issue than it would be to abuse the issue myself.
that codebase is such a mess

first they use nix.. then they use ruby gems.. and heres the funny part THEY ALSO USE FLUTTER
I'mma have a look through one sec
To do development work on the NL reference wallet, you need to following tools:
Rust (including additional targets and utilities)
Android SDK + NDK (for Android builds)
Xcode (for iOS builds)
Flutter
Docker (to run supporting services)
PKCS #11 library








