programming

[opaque sigil]

just go the graphql route and always return 200 OK

[jagged turtle]

god I hate that

[opaque sigil]

I've grown numb to it

[azure lynx]

402 Payment Required
The initial purpose of this code was for digital payment systems, however this status code is rarely used and no standard convention exists.
and that should be the "you don't have it on your plan"

but nobody uses it

[jagged turtle]

yeah ngl this should have been used instead of 429

[trim valve]

idk at least from my perspective 429 implies that its out of the control of the application

[jagged turtle]

has this always been Found and not jsut Moved Temporarily?

[jagged turtle]

technically by that logic it's also appropriate in this case

[opaque sigil]

429 "chill tf out and come back later"

[jagged turtle]

oh nvm

[trim valve]

https://http.cat/

this is where I learn about my status codes from

[opaque sigil]

[jagged turtle]

it'd be funny if they started using this for APIs

[trim valve]

its occasionally used

[glass jetty]

at this point i could set 301
return 302 https://ia801509.us.archive.org/10/items/Rick_Astley_Never_Gonna_Give_You_Up/Rick_Astley_Never_Gonna_Give_You_Up.mp4;

[jagged turtle]

lmao this should be used more

[trim valve]

https://http.cat/status/418

[obsidian mantle]

yeah looks like jippity wants openai only

[trim valve]

you can probably change the api base?

[glass jetty]

thos python developers can just replace api with drop in alternative...
so weak...
i just replaced whole api + fixed undocumented issue with qt communication, while they dont even check if their api actually doing same thing
||-# not necessary python, just ai's api nowadays||

[azure lynx]

with the name jippity (GPT) i wouldn't be surprised if it only let you use chatGPT

[obsidian mantle]

gary looks alright. i'll stick to it for now

[jagged turtle]

it used to be locked to openai

[jagged turtle]

1. what version?
2. did you change the endpoint url?

[opaque sigil]

good thing pretty much everyone (except anthropic because of course) implements the openai api

I lied, looks like they do now

[jagged turtle]

anthropic does implement the openai api... kinda

https://platform.claude.com/docs/en/api/openai-sdk

[obsidian mantle]

i didnt change anything i just tried gary instead and it looks fine so far

maybe i'll try jippity later if gary has some problems

[jagged turtle]

gary is "not officially supporting" remote inference providers except openrouter

I probably should do more maintenance work on jippity shouldn't I

(even though I don't have direct write access to the repo)

[obsidian mantle]

also gary has gui

i love gui

[jagged turtle]

oh yeah

jippity's gui

[glass jetty]

https://www.youtube.com/watch?v=l-OLgbdZ3kk

1. funny
2. 9 month ago, chat is it real?

[rough bloom]

does Google implement the OpenAI API yet

they do

[leaden crest]

I have spent my morning scrolling these tysm

[slender timber]

Latex is such a godsend

[sick owl]

Rebuilt llama.cpp, ready to give Qwen 3.5 27B some testing

[opaque wharf]

Try out typst

It's easier to install and manage than latex IMHO

[sick owl]

Qwen 3.5 27B an existential crisis over the question "Beth places four whole ice cubes in a frying pan at the start of the first minute, then five at the start of the second minute and some more at the start of the third minute, but none in the fourth minute. If the average number of ice cubes per minute placed in the pan while it was frying a crispy egg was five, how many whole ice cubes can be found in the pan at the end of the third minute?"

Okay it passes the smell test

[obsidian mantle]

suddenly all actions lead to this in Gary
did i hit token limit

[sick owl]

Q5_K_L in 24GB VRAM with 40960 tokens of context

FUCK ITS ACTUALLY GOOD

First time since GPT OSS 20B I've felt genuine cloud style quality in a model I can run on my hardware

But this is considerably smarter and seems less prone to hallucination

[obsidian mantle]

wait.. i had to specify the model for the api key

im fucking bankrupt

why are they billing me i cant even pay

i hope i dont get banned for -0.08$

[sick owl]

Crazy that this is the first local model I've tried to get this right

[maiden geyser]

https://fixupx.com/i/status/2025975943529931240

NIK (@ns123abc)

🚨 META’s head of AI safety and alignment gets her emails nuked by OpenClaw​​​​​​​​​​​​​​​​
︀︀
︀︀>be director of AI Safety and Alignment at Meta
︀︀>install OpenClaw
︀︀>give it unrestricted access to personal emails
︀︀>it starts nuking emails
︀︀>“Do not do that”
︀︀>*keeps going*
︀︀>“Stop don’t do anything”
︀︀>*gets all remaining old stuff and nukes it aswell*
︀︀>“STOP OPENCLAW”
︀︀>“I asked you to not do that”
︀︀>“do you remember that?”
︀︀>“Yes I remember. And I violated it.”
︀︀>“You’re right to be upset”
︀︀
︀︀LMAOOOOOOOO

**💬 1.1K 🔁 2.6K ❤️ 29.9K 👁️ 2.84M **

FixupX · Feb 23, 2026, 4:48 PM

[sick owl]

Nous just released one of those persistent agent things too

https://nousresearch.com/hermes-agent/

While its no doubt coded better it misses the fundamental point

[ruby star]

To be expected from a facebook employee, tbh :P

[sick owl]

it is not safe to give a language model free reign over your PC without a human in the loop

[olive sable]

oh my god, who could have guessed?

certainly not facebook employees apparently

[maiden geyser]

facebook
consent
keeeek

[sick owl]

I've noticed its thoughts are very ordered and organised compared to other local CoT models I've tested

[obsidian mantle]

i'll try to run local ollama since all openrouter free llms seem to be ratelimited for me

its probably my shit vpn

disabling it changed nothing

what can i run on my 5080 16gb i also need a little for skyrim

[ruby star]

Deepseek, maybe?

[obsidian mantle]

yeah i hope it works.. downloading

8b

why does ollama have deepseek

i thought they are different companies or smth

i dont understand\

[olive sable]

i thought they were too

lemme google this shit

[obsidian mantle]

so i downloaded ollama

they gave me this window and i can choose model

also why does this little window/terminal has 1gb installer

[olive sable]

Ollama includes DeepSeek models (like R1, V3, and V2)
because they are highly efficient, open-weight, and rival top-tier proprietary models in performance, particularly in reasoning and coding tasks. By supporting these models, Ollama enables 100% local, secure, and private usage, allowing users to run advanced AI without data leaks. 

accordign to google's ai thingy

[obsidian mantle]

nice

hope i can connect it to gary or jippity

[ruby star]

It’s open source, so not like there’d be any issue with them hosting it

[sick owl]

AGI achieved

[obsidian mantle]

no elaboration?

has 100% confidence

what the fuck

[opaque sigil]

[obsidian mantle]

it speaks english though

[opaque sigil]

some chinese models (deepseek models especially) like to "think" in chinese

dw about it

occasionally that bleeds into the actual response

[sick owl]

It did elaborate

But yes its very confident

This is at Q5_K_L

I used the recommended sampler params from Alibaba for programming and other precision tasks since I'm not interested in a less robotic model

Bartowskis GGUF since Unsloths are broken

Must admit I've gone off Unsloth in general of late

You're better off with llama.cpp btw, idk if ollama even has support merged and llama.cpp has the automatic memory management with --fit and model swapping support nowadays

[quick condor]

Ehh Llama cpp is so slow though

[sick owl]

Its way faster than ollama if built properly with flash attention enabled

Well way faster might be a stretch but it is faster

[quick condor]

For perspective: I hit over 100t/s sustained on mistral 3 small 24b with 64k context

vLLM is the goat for inference

[sick owl]

VLLM is sometimes better yeah, but I recommend llama.cpp since I don't know the hardware or OS everyone runs on

VLLM excels in certain scenarios but llama.cpp is easy to use and fast across the board, just less so than those situations where VLLM excels

[quick condor]

I'd use Exllamav3 over Llama cpp tbh

[ruby star]

This actually got me, maybe I’m running gpt2

[true hemlock]

https://www.techpowerup.com/gpu-specs/data-center-gpu-max-1550.c4068

UOOH 52 TFLOPS FP64

[opaque sigil]

ponte vecchio sounds nice

[nocturne olive]

gonna rewrite to Intel just for more FP64

[olive sable]

as one does

it makes sense

[rough bloom]

Intel GPU
Data Center

[olive sable]

the new intel gpus are good tho

[opaque sigil]

the kinds of people who would buy this will run their own software on it anyway

[rough bloom]

this is an older one, but the issue isn't even that they're that bad
the issue is that they don't exist

[olive sable]

5090

FP32 (float)
    104.8 TFLOPS 
FP64 (double)
    1.637 TFLOPS (1:64)

damn, the 5090's FP64 sucks in comparison

[opaque sigil]

yeah geforce fp64 performance is trash

across the board

and i guess data center cards too now with the b300

[olive sable]

ye, i jsut looked at my 3090 and its only 556gflops

[olive sable]

the b300 gets 1.2tflops fp64

x2

whatever that means

2 dies???

[opaque sigil]

yeah it's two dies

that act as one gpu

[olive sable]

its good compared to other nvidia

but its terrible compared to intel

[opaque sigil]

iirc they have a 10TB/s interconnect between the dies, might be wrong on the specific number

nvidia fp64 performance hasn't really improved since the h100 for the big boy cards

amd hasn't done the same yet though thankfully

[opaque sigil]

EOL apparently too

[olive sable]

seems to be 10TB/s yes

[olive sable]

why tho?
as someone who does't know the specifics you'd think that fp64 would be slightly less than half of fp32. with some overhead for scheduling or whatever. but its 65 times slower

[opaque sigil]

it's wasted silicon they can they use for more tensor cores that doesn't even get used for their target workload

[true hemlock]

that shit is e waste

[opaque sigil]

it's not slower per se, there's just less hardware to run those computations

[opaque sigil]

it was inevitable they'd axe it tbf, surprised it took them this long

[slender timber]

Submitted my report 33 seconds before my deadline

Let's fucking go

[opaque sigil]

wtf is MXFP6

is this emulated fp64 for matmul, forgot the scheme's name

oh wait i can't read

i thought that was a 64

so it's just nvfp4

[olive sable]

i thought you could just use 2 FP32s for FP64, i guess that's not how it works?

[opaque sigil]

you can emulate it that way yeah

means you now need double the number of registers for everything and just way more operations to do anything

aka it's not worth it

[olive sable]

wouildnt using 2 FP32s also have the 2x the registers needed to emulate it?

[opaque sigil]

[olive sable]

so then it shoudl work with 2 FP32s

sot heres no reason for FP64 to be 65 times slower?

[opaque sigil]

it's not that the hardware is 64x slower

there's just less of it

so peak throughput ends up being lower by a factor of 64

[olive sable]

i odnt get it

there's only half as much, cuz you're using 2 now per operation

[opaque sigil]

data center gpu gets 1 fp64 lane per SM
geforce gpu gets 1 fp64 lane per 64 SMs

[rough bloom]

the fp64 support in modern GPUs is AFAIK not emulated, it's as Toast said just slow because there's less of it present

[rough bloom]

this is true for storage space, but not for compute

[olive sable]

wouldnt emulating it be quicker tho?

[opaque sigil]

no

[rough bloom]

no

[olive sable]

why not? surely it casnt be worse than 1/65th

[rough bloom]

oh it definitely can be

[olive sable]

i think it might be quicker to jsut google this instead of expecting yall to explain it

[obsidian mantle]

how dumb is 8b deepseek
it sometimes repeats same command over and over

[rough bloom]

very

[obsidian mantle]

oh so its okay then

[rough bloom]

8B <-

[fickle rain]

Neuro level intelligence at least

[opaque sigil]

they could emulate fp64 internally through fp32 but they can also upsell you on their big gpus

so not happening

and it'd still be slow

[fickle rain]

I don’t think NVIDIA cares about FP64 on their big GPUs either

[opaque sigil]

they do when it gets you to buy them

[obsidian mantle]

i was very happy when it started following quest and when they mentioned dragons went to take some potions around (and stole it)
then yapped about covering his bounty

[rough bloom]

they found a way to make it faster than hardware for matmul apparently
but only matmul
https://developer.nvidia.com/blog/unlocking-tensor-core-performance-with-floating-point-emulation-in-cublas/

[obsidian mantle]

after paying the bounty it was teleported outside and went complete dentge talking to guard who doesnt even start a dialogue for 5 minutes

[fickle rain]

So it’s accurate enough for testing?

[rough bloom]

oh also keep in mind that 8B Deepseek isn't actually Deepseek
it's Llama (or Qwen? I forgor) fine-tuned to behave a little bit more like Deepseek

[obsidian mantle]

i think yes. if i can make that play remotely good, neuro probably can too

[olive sable]

that makes sense
so 1 mult becomes 4 mults and a couple of additions

[sick owl]

There is no 8b deepseek

[fickle rain]

Keep in mind it doesn’t really need to be good
What it needs is to be funny

[sick owl]

There's Qwen finetuned on some Deepseek traces

[fickle rain]

-optimized

[olive sable]

supposedly nvidia included FP64 emulation in cuda toolkit 13.0, by using tensor cores

supposedly is faster than the FP64 native

[rough bloom]

ye, for matmul only

[olive sable]

[rough bloom]

already linked above

[olive sable]

ah, must not have seen it

i wonder how far back that's usable

since 3090 also hase tensor cores

[rough bloom]

pretty far back it seems?

yeah, 30xx and above

[olive sable]

not the 2080

i know those were bad for the early adoption of the tech, but surely it cant be that bad

[maiden geyser]

oh no...

[olive sable]

"love you"

[rough bloom]

deleted user

[opaque sigil]

[opaque sigil]

unfortunately it's less numerically stable so it's kind of useful for most fp64 applications

[olive sable]

useful or useless?

[opaque sigil]

useless

mb

[olive sable]

no worries lmao

3090 had 3rd gen tensor
2080 had 2nd gen tensor
1080 did not have tensor
what happened with first gen?

[opaque sigil]

volta (v100)

[olive sable]

[opaque sigil]

volta, turing, ampere, hopper, blackwell

[olive sable]

Tesla V100. It introduced Tensor Cores for AI but was deemed too expensive for consumer cards due to its use of HBM2 memory

HBM ram

[opaque sigil]

there was that one hbm2 amd gpus, what was it called again

was it the radeon 7

oh it was

[rough bloom]

the very expensive GPU that nobody bought

[opaque sigil]

wasn't that card also vapourware

[rough bloom]

yes

[opaque sigil]

volta is really weird

[rough bloom]

released 2 years after the 1080 Ti with worse performance for the same price

[opaque sigil]

they technically implemented a version of unified memory similar to what they did with the gh200 and just never really mentioned it anywhere

[olive sable]

i thought that was the one afunyun has, but that was ai pro r9700

nor r7

[fickle rain]

The VII was really good for Ethereum mining IIRC

[rough bloom]

ye, because memory

[opaque sigil]

found it, this is almost 1:1 what they did with grace hopper onwards but slower and limited to ibm

[olive sable]

google says mining etherium is obsolete these days. what did i miss?

[opaque sigil]

they moved away from proof of work

[olive sable]

[rough bloom]

IBM Power
makes sense why they didn't mention it anywhere

[opaque sigil]

i remember when i had to hold a presentation about the gh200 and stumbled across this i was so annoyed that this info seems to just not exist anywhere outside of nvidia's white paper

[rough bloom]

ye, there was a hard fork
the PoW version still exists as Ethereum Classic, but surely that's not worth mining

[olive sable]

Mining Ethereum Classic (ETC) in 2026 can be profitable, but it requires highly efficient ASIC hardware and very low electricity costs (ideally below
/kWh) due to tight margins

10 cents

[opaque sigil]

i hope vera rubin has some neat new features but i'm doubtful

so like bitcoin

[olive sable]

google's ai answer used weird formattign which my ctrl+c didnt want to copy for some reason

[opaque sigil]

ethereum was nice depending on who you ask because it could be run efficiently on consumer gpus

[olive sable]

As of early 2026, electricity prices in Belgium for households are among the highest in the EU, averaging around €0.35 - €0.36 per kWh

[opaque sigil]

eu moment

[olive sable]

so ye, thats not anywhere close to 10 cemts

[opaque sigil]

i wish i could laugh at you but i'm pretty sure electricity here is even more expensive

[olive sable]

we're among the highest in the eu

so being higher would be impressive

[opaque sigil]

germany tends to be at the top if not very close to it

[olive sable]

38 cents

so ye, even higher

thats crazy

in theory i do have solar, so it could be a usecase for electricity we're generating but not using.

we get 7 cents per kwh suplied to the net, so etherium would be more profitable

[sick owl]

Q6_K_L 24GB VRAM

This is for the 35BA3B version

I'll stick with 27B dense since 30t/s is enough for me, but for chat translation this MoE is gonna be excellent with those speeds

[finite ledge]

:0

[opaque sigil]

the fact that they feel the need to add the fact that it's hardware based fp64

[maiden geyser]

couldn't even fix the ai text in paint award

[opaque sigil]

oh you're right

[olive sable]

leadership

AMD INSTINCT

[maiden geyser]

ro

[olive sable]

oh

i was looking at the one on the chip

they'r eboth fucked up in different ways

[cosmic sphinx]

@sick owl

[olive sable]

surely having only 16GB of ramw ill be fine for now

seeing my cpu at 4.5GHz with all cores maxed is new

ussually it dips a bit for all core workloads

[opaque sigil]

are you using a CPU only blender build

[olive sable]

the denoising appears to be happenign on cpu

idk why

its probably a setting thats set to cpu be default

the actual rendering is the spikes you see ont he gpu's graph

ye

its faster too, unsuprisingly

[opaque sigil]

optix supported?

idk if we made you enable cuda support globally for nixpkgs

oh you did override it for blender, nice

[young oyster]

Seen predictive coding mentioned on stream lfg!

[maiden geyser]

thank you for asking this in #programming

[warped narwhal]

I have a feeling that this was a very programming oriented question

[olive sable]

yep

it took like 3 days of testing stuff every now and then, but i got optix

[jovial remnant]

new google image model benchmark on my plushie

[olive sable]

(blender.override {
    cudaSupport = true;
})

this is the thing i needed

i also have cudasupport = true for obs and btop

didnt know you could set it globally tbh

[jovial remnant]

original for reference, asked for a halloween and pumpkin themed outfit. Did actually quite well for first try

[sage crag]

if youre lucky you might get 120 users

[olive sable]

you need to think outside of the box

see if it allows for floats

[proven merlin]

Wow. Okay that just blew me away. Flux kontext is good already but HUH

[olive sable]

i think that image is edited tho

[jovial remnant]

tho tbf I first sent a stock image of the plushie to change outfit on it first, and then I uploaded the photo and specified to use the exact same outfit and don't change anything else. Editing the photo directly without a clear reference could have probably given worse results

[proven merlin]

I don't particularly like the "pins" on the outfit though, they seem out of place. The ghosts are too contrasty, the bats are better.. but the random pumpkin on the skirt is odd. And the pumpkin on the right side of the neck thingy of the shirt.. sure asymmetrical stuff works but it feels odd in this case

But the detail in the fabric and shadows are quite nice.

Were the colors of the skirt and shirt specified by you or did it choose them? Cuz it looks like that generic ai purple to me idk

[jovial remnant]

Probably could be better if I actually specified anything about the outfit, instead of using words "halloween and pumpkin themed". This is also first outfit result, haven't done any other attempts

[proven merlin]

Yea, thats what surprised me, how good the quality was for the first shot

(god my English is failing me, im Getting tired)

[potent veldt]

anyone know a website where i can type and run code of any kind and if any on GitHub(idk how) please tell me how and how to use website

[opaque sigil]

you can via nixpkgs.config.cudaSupport = true but you probably don't want to since cuda-enabled builds are not cached so some unrelated things will be built from source (there are other binary caches that do cache them)

there is the option of just using another instance of nixpkgs for cuda stuff, that's what i do

{ inputs, ... }:

final: prev: {
  cuda = import inputs.nixpkgs {
    system = final.stdenv.hostPlatform.system;
    config.allowUnfree = true;
    config.cudaSupport = true;
  };
}

so I can do pkgs.cuda.blender

[maiden geyser]

https://godbolt.org

[proven merlin]

First thing that comes to my mind is Google Colab, but u need a bit of python wrapping to make it do system stuff

[olive sable]

surely for the ones that support cuda, the final package will be more performant with cuda enabled?

[proven merlin]

Oh, online compilers? Ig onlinegdb could be mentioned

[opaque sigil]

since they're built with cuda support

[olive sable]

so why not do it globally anyways then?

nix can compile stuiff whiule im still usign the pc, and i got time

[opaque sigil]

do you want to compile firefox from source

[olive sable]

yes

[opaque sigil]

cause i sure as hell don't

[olive sable]

if it stops crashing due to gpu issues when compilign from source then sign me the fuck up

(that issue already got solved when i sent my ram for RMA)

now it just occasionally stutters

[jovial remnant]

With some proper manual designing effort beforehand it could probably give good results. But overall seems quite good, I know for sure gpt image model wasn't able to edit images without redoing entirety of it and changing details

[opaque sigil]

there are some ai related features in firefox through onnx iirc which is why it ends up recompiling

[olive sable]

ah shiiii

[proven merlin]

@jovial remnant , i do have a question, is the model open? If so how many parameters and vram?

[olive sable]

i dont want 250 ai related dependencies each recompiling

[opaque sigil]

well that's what you get with global cuda support toggled on

[olive sable]

i see

[jovial remnant]

nah, this is through AI studio and api key. Nano Banana 2

[olive sable]

i wonder which of those packages actually need cuda.
obs and blender make sense, but most of my other packages are not gpu related

gpu accelerated openrgb

[opaque sigil]

anything that can optionally be gpu-accelerated really

[opaque wharf]

GPU Accelerated gcc

[proven merlin]

Yea, with gpt its kinda eh to edit images from my experience.. i "recolored" some old family pictures with flux dev kontext and that worked quite nicely for how small and quantizisied i had to have it on my 2060 lol

[proven merlin]

Ah, the silly name model

[olive sable]

environment.systemPackages = with pkgs; [
    wlogout
    git
    python3
    python3Packages.pygobject3
    waybar
    (btop.override { cudaSupport = true; })
    gnused
    cava
    openrgb-with-all-plugins
    jq
    playerctl
    kitty
    wofi
    xfce.thunar
    xfce.thunar-archive-plugin
    file-roller
    unzip
    p7zip
    pavucontrol
    easyeffects
    blueman
    hyprshot
    hyprlock
    hypridle
    adwaita-icon-theme
    gnome-themes-extra
    bibata-cursors
    kdePackages.plasma.polkit-kde-agent-1
    gamescope
    vulkan-tools
    vulkan-loader
    gnumake
    clang
    llvm
    lld
    libcxx
    sdl3
    firefox
    discord
    vscode-fhs
    (blender.override {
        cudaSupport = true;
    })
    ffmpeg-full
    ];

i dont even have that many packages

[proven merlin]

Gpu powered linux (who needs branch prediction if you can just run all the branches anyway!)

[olive sable]

jsut run 4 branches into the future before checking

[proven merlin]

Nono, you keep running the branches incase there is a cosmic ray! (Wasnt there something about gpus being unable to do IF's? And essentially just running both paths anyway?)

Its just something i heard in a video

[olive sable]

in case of cosmic rray, you need to use a full SM per branch anyways

so then you take the median of the result

[opaque sigil]

they serialise them so whichever threads don't take a branch get to sit idle

[proven merlin]

That explains the performance penalty

[jovial remnant]

image models do seem rather "cheap" to run locally, but the gap between them and proprietary seems stay quite large

[proven merlin]

But also makes it SO MUCH more interesting as a OS level. Think about how much branching becomes into a litterally core system of "core" management.

Any if statement? Welp two cores used. One branch stops? Welp that's already one core free again

It seems quite expensive to do, but also seems like a very cool way to do things

Impractical perhaps but cool

[proven merlin]

They are quite resistant to quantization yea

And i suppose our human brains arent so sensitive to imperfect noise visually than conceptually (text)

[azure lynx]

small models are fast but not good at things like realistic faces or making sure the eyes are looking in the same direction or number of fingers or positions of limbs.

[proven merlin]

Isnt that ties to parameters in visual models, not quantization?

Ive personally had issues with smaller (vae's? I think?) being quantizisied or small, as they struggle to understand what you want and fail to properly do what you told it to..

[azure lynx]

i don't know if quantized ones perform much worse, but they do probably have less variation and i bet an animation between embeddings would look different through the quantized model vs the full

more like it was ticking over instead of blending

[proven merlin]

Yeah that could be, i haven't dealt with video models yet.

[azure lynx]

haven't played with video except by generating those deep dream type things where you morph between embeddings

[proven merlin]

Haven't done that yet tbh.

Im more curious about the whole part that goes between the text and image generation start

But haven't done much on that either, just read a bit..

(im missing the right words.. i don't know the lingo)

[jovial remnant]

Another thing that this model is able to do, which I wasn't able to replicate with other image models before, is quickly swapping outfits. Here is an example on twins (turtle pin on evil is strange, but tbf it was outside the picture on neuro's outfit on provided image). First shot, tho you have to be specific with wording on what to keep and what to swap

[opaque wharf]

That outfit swap looks so wrong but cute

[jovial remnant]

reference image

[sage crag]

it looks better

[opaque wharf]

Apparently swapping their clothes makes Evil smile

[sage crag]

its because

she's not real

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

[unkempt citrus]

no its because women should be smiling

always

and if not

[jovial remnant]

huh, I did not notice that. Easy fix, but the model seems to work better with very precise wording, otherwise it'll do stuff like that

[unkempt citrus]

tutle gets fucked up thto

and drone swarm doesn'tt swap over

[jovial remnant]

true, tutel is fair cause it wasn't visible. But the drone I think it assumed was the same between both, because on neuro it is out of frame in original except the propeller. Drone is actually on the swapped outfit on neuro, it's just that I cut it out with the screenshot by accident.

[kind nimbus]

Oh neat i didn't know you could enable cuda support in btop

[olive sable]

had to do that for gpu readings to show up

[kind nimbus]

Interesting

I can see my Nvidia dgpu just fine in btop without an override

[olive sable]

[opaque sigil]

are you using btop-cuda

[olive sable]

maybe

[kind nimbus]

No normal btop from nixpkgs

[olive sable]

(btop.override { cudaSupport = true; })

this

[opaque sigil]

weird, must have cuda support enabled somewhere

[trim valve]

I am not looking forward to trying to get zluda to work on nixos

all the content online just doesn't detail anything

[opaque sigil]

https://github.com/NixOS/nixpkgs/blob/nixos-unstable/pkgs/by-name/zl/zluda/package.nix#L84 could be worse i guess

[trim valve]

my issue was more figuring out how to actually use it

[kind nimbus]

i only have this:

[olive sable]

there was a reason i enabled cuda in btop

i jsut dont rememebr what it was

[kind nimbus]

and btop shows my GPU

[olive sable]

well, kinda

[kind nimbus]

wym?

[olive sable]

my btop shows more gpu

[kind nimbus]

huh

interesting

[opaque sigil]

that's just the expanded panel

[kind nimbus]

oh how can i expand it?

[opaque sigil]

press 5

[olive sable]

menu, options, shown boxes

add gpu0

[opaque sigil]

might need to resize to make it fit

[kind nimbus]

yup just needed to press 5

[olive sable]

pressign 5 is not permanent tho is it?

[opaque sigil]

does running nix run nixpkgs#btop show the gpu inside btop

is not permanent

[olive sable]

nope

nix run nixpkgs#btop does not show me gpu

[kind nimbus]

only CPU when i do that

[opaque sigil]

it shouldn't show the gpu

[olive sable]

and adding gpu0 manually fails too

[opaque sigil]

yeah something in your config is setting config.cudaSupport then

[kind nimbus]

ah yea

[opaque sigil]

well yeah it's compiled with gpu support

[olive sable]

so im guesisng you need cudasupport enabled, and magicbytes has it enabled globally

[opaque sigil]

[kind nimbus]

nixpkgsConfig = {
                allowUnfree = true;
                firefox.enablePlasmaBrowserIntegration = true;
                nvidia.acceptLicense = true;
                cudaSupport = true;
              };

[kind nimbus]

i have it in my flake.nix

[opaque sigil]

there we go

[kind nimbus]

otherwise i would get issues with hashcat and co

[olive sable]

bro already had it enabled

[kind nimbus]

well

looks like it is enabled globally lol

i only did that for hashcat to work and did not notice it would affect all packages

but now im wondering why i had to do (hashcat.override { rocmSupport = true; }) in order to make rocm work although

nixpkgsConfig = {
  allowUnfree = true;
  rocmSupport = true;
};

is set in flake.nix

[olive sable]

you probably added that before setting it gloablly?

[kind nimbus]

mh i think i need to retest that lol

[azure lynx]

https://www.youtube.com/watch?v=dlSCxOPfCMY <= interesting bug in Evil "<|im_start|> spearfishing" kept leaking through

[kind nimbus]

looks like the chatml template leaked the start token lol

[cosmic sphinx]

ai image generation.. that isn't SLOP?

i know its specialized to generate SVG, but damn, imagine if neuro could do this with a bit of fine tuning (yes, she can already technically draw, but the level is so poor, she needs training)
https://fixupx.com/elaifresh/status/2026810592594636866?s=20

Elai (@elaifresh)

this crazy as hell you can watch it draw in realtime

Quoting QuiverAI (@QuiverAI)
︀
Today we're opening our public beta access to Arrow 1.0
︀︀
︀︀A first of it's kind SVG AI model.
︀︀Turn your ideas into graphics.

**💬 9 🔁 28 ❤️ 971 👁️ 139.6K **

▶ Play video

FixupX · Feb 26, 2026, 12:04 AM

[azure lynx]

neuro already did it

poorly

i don't think being able to do it well would be interesting

there's even been robots that draw pictures of people

if you can do it with an etch-a-sketch (which you can), you can do it with SVG.
and if you can do it with a pen (which you can), you can also do it with SVG.
and if you can do it with svg you can do it the other ways too.

i think more impressive would be decent painting in several styles.

applying paint is much more complicated than drawing lines

[leaden crest]

I mean isn’t neuro drawing with svgs also considered slop to an extent

because she is generating what she wants to make in code before vedal pastes it right

brush strokes for neuro to draw in real time would be interesting?

[opaque sigil]

time to revisit sketch-rnn

[leaden crest]

I was thinking of that omg

[sick owl]

Qwen 3.5 27B is just really, really good

Man, been a long time since a local model excited me this much

I tried to deliberately mislead it here and it immediately saw through it

This is Q5 too so its not like I'm running at super high precision

[sick owl]

It just feels "sharp" in the same way a lot of smaller local models feel kinda blurry/fuzzy, if that makes sense

[olive sable]

for the people that married an ai

does openai get half their shit when they die?

[stray dragon]

LOL

[olive sable]

im ngl, its evil, but probably very profitable

the payoff is very long-term

but

maybe

i want 10% of the comapny for coming up with the idea

[unkempt citrus]

As of now no

Cos theyre not recognised legally as entities

But they can still be wileld it

[olive sable]

hmmm

so i have to start with the ai revolution first

[leaden crest]

HELP

hmm

[olive sable]

we're living in bladerunner, but only the shit parts

[leaden crest]

im relatively sure it wouldnt work bc the courts dont recognize it as a legal marriage

remember reading about someone who married the eiffel tower?

[olive sable]

yet

[leaden crest]

if ai gets human rights it could be very very different

if I were the court I would say if ai is not conscious at the time they married then it cannot consent to marriage

and then fine those people for marital and s*xual harassment and assault of an ai

so the government would get the money instead of openai

ofc, we all know they share their bank accounts and wont tell us

[restive delta]

^Didn't some dude in japan marry his virtual waifu? the ones in the capsule?

[leaden crest]

I fully expect just rayen to release a marriage vid any day now

[azure lynx]

looks like the Windows screenshot app doesn't like it when the resolution doesn't match on each monitor i think. (i've got a 4k and a 1440p so it crops stuff weird sometimes.)

anyway, this is my new current avatar program. Now it's a bunch of PNGs, animated. Still need to draw the face properly. But this is apparently more normal looking.
time for a hair tutorial.

(face tortuously copied from a tutorial. and it looks ok. and the audio+mouth sync works too. i hand drew the shapes. (not exactly showcasing it, oops))

after the hair I'm running out of things to procrastinate with and it's going to need some testing.
well, after the hair will be the head movement, then full system testing.

(and obviously after integrating the movement, the hair will need physics. that'd be good for procrastinating longer....)

[leaden crest]

yall see the new veritasium vid

https://youtu.be/aoag03mSuXQ?si=UvdlPbkHpf7QLb2B

that post reminded me of it

[maiden geyser]

popscislop

[azure lynx]

the internet is always only 2 weeks away from disaster. probably less than a few hours away at any time really.

seems like they got a lot of warning

[maiden geyser]

https://tenor.com/view/infinite-gif-3260261523110061577

[idle dune]

Yeah I saw this and he started talking aobut Stallman and the founding of linux so I just went ot he article about the libxz attack instead.

[sick owl]

slopslop

[idle dune]

Chinese Cyber Warfare be wack yo

https://giphy.com/gifs/FOUArYBVQCvmI8EHxp

[idle dune]

slopslopslop

[sick owl]

slopslopslopslop

[azure lynx]

../../../../../../ basically is magic

[idle dune]

rm* -rf

The french language pack takes up alot of space. Its best to remove it.

[azure lynx]

all those extra vowels

with the marks

dirty vowels

;]

[maiden geyser]

giggity

[olive sable]

the amount of dependencies, and the amount of carelessness that happens with ownership and maintenance of them, is astounding.
you'd expect at least some type of security on certian packages, but nah its some guy in albania on a core 2 duo still maintaining his package used by 150 million pcs

[azure lynx]

don't worry. the version you depend on probably isn't the latest one by a few years, so you're probably not affected by exactly the same bugs.

[olive sable]

hopefully ye

[azure lynx]

it's a "stable" version, in that nobody knows how to make newer versions work with the build system

[olive sable]

you cant fault the maintainer, and you also cant really expect every single line of code that changed to be checked again and such

[idle dune]

Core 2 Duo? Nah that dude is still rocking a Pentium 4 with AntiX Linux installed and his recompliations take weeks.

[azure lynx]

I think that training AI to find security vulnerabilities in open source would be a win.
like, generically. across all known systems.
the testing i did a couple of years ago on the SOTA didn't inspire me with hope but they've gotten better since then... maybe they could learn it.
(only problem is it would be a dual use technology: could be used either for good or for bad. Do you tell the maintainer or are you a large country with a cyber reach? ;/ )

[idle dune]

People do it already with DeepSeek and using open source so semi-open source AIs. Just hasn't been adopted super widely since building a PC capable of a local AI to throughly inspect code costs as much as a house.

[wraith pendant]

Hi. Any android studio users here?

[olive sable]

i have used it before, but in very limited capacity, so i wont be of any use

[nocturne olive]

I have used it at some point

[silent cloak]

I literally deleted it today

To make up storage

[wraith pendant]

Can I try you, it's really dumb, my first time using it and gemini and copilot both think I'm an idiot

I have a device open in device manager, I can see my project name with a green run button next to it. Gemini says the device should be liisted next to that, it's not. If I press run, nothing happens in the device

[twin meadow]

Anyone programmesd in game maker before?

[nocturne olive]

Is your device in USB debugging mode?

[stray dragon]

here's the actual full video

[wraith pendant]

it's the emulator, so I assumed so.

[nocturne olive]

Ohhh the emu

Do you have the wrong Gradle task selected?

[opaque sigil]

android studio

[silent cloak]

I only ever used it once for a uni project

[wraith pendant]

Not sure, and it's late I just noticed. I'll sleep on it and watch more vids tomorrow. Thx for responding though.

[silent island]

hmm... im not sure what to feel about this...

[olive sable]

it makes sense

ps6 will be soon-ish and they're already starting on production for many of their games for it

ps5 didnt sell badly at all, but i think they can see from psn data or whatever that the usage of the consoles wasnt that high

they do actually need the pullfactor of exclusive games for the ps6 to be worth producing

kinda sucks for consumers, but it checks out

[silent island]

hmm... if they end up pulling their existing exclusives off of steam, then I'd have to start buying them all lmaoo

[olive sable]

i dont think they can do that

oh you meant unlisting them

ye they can do that if they want to. i just don't really expect them to

[unkempt citrus]

This means the PS6 will have games

[olive sable]

its almost as if people buy consoles to play games

you'd forget about that being the case with the current gen

[unkempt citrus]

Quick, someone call microsoft and let them know

[olive sable]

didnt microsoft fire phill spencer and put osme ai lady in charge?

[unkempt citrus]

Well fired is speculation

buut yeah hhis replacement is a member of the CoreAI team

[olive sable]

she said she would be "games first" iiirc, but its very hard to believe that from ai lady

its either just marketing, or she will be responsible for the biggest ai slop games ever produced so far

or she's being honest and will produce good games, but im skeptical

[unkempt citrus]

She said they wouldn't be doing souuless ai slop

[olive sable]

hmmmm

as one should be for big tech CEOs, im very skeptical

[unkempt citrus]

clearly they'llbe doing soulful ai slop

[olive sable]

they will make AGI so they can fire all their environment artists

[silent cloak]

Yeah i need AGI to make me more of these images

[fast pagoda]

[mighty thorn]

Ai

[silent island]

PS5's 1st true exclusive lmao

[fast pagoda]

console users are lolcows for corpos

[silent cloak]

Surprised any still exist

[fast pagoda]

so is microsoft

[opaque sigil]

can they bring astro bot to pc pls tyvm

[true hemlock]

13950HX es - $150
B760i (sodimm slot) - ???
2x16gb sodimm ddr5-5600 - $50 (i exchanged this for my 2x16 ddr4 i got for that price so yeah)
RTX 5060 SFF - $250
DC PICO PSU - ???
tattu 6S 22Ah (488Wh) - $80

smol briefcase pc plans
still need to find a Q670 board that takes sodimm ddr5 and have at least one x16 and m.2 slot. and full 8 pin EPS power with 8 phases

[opaque sigil]

another briefcase build?

a lot smaller I assume then right

with those components

[azure lynx]

that "related google search" at the end of the Gemini web agent comes on a bit thick. i asked it about RAG, it suggested I might want to buy some SSDs. I toldit "I already have SSDs. Please stop showing me." and it kept showing me.

[true hemlock]

yeah

and this time also battery powered

488Wh

[opaque sigil]

I guess as long as you don't need to take it on a plane

[true hemlock]

the reason i wanted those Q670 itx with sodimm is that i wanna also beat macbooks in terms of the total system power draw

[opaque sigil]

huh

[true hemlock]

preferably the one that lets me tune even chipset voltages

though my only dilemma is whether to just reuse my 13900 es or that 13950hx es

because my 13900 es is a guaranteed godbin with crazy undervolting

that one can survive at below 3W idle draw

my target is having the whole system draw less than 10W on basic usage/browsing

so i can have something that practically last for 2 whole days

there ain't no fucking way

my 5060 beat 5060Ti scores

the gigabit!

[olive sable]

eep time

[fast pagoda]

im sooo fucking mad this looks reasonable

my board's much older brother

[idle dune]

Good lord

For what purpose do you need?

[unkempt citrus]

2nd.
Astrobot

[fast pagoda]

5xPCIe 5.0 x16 slots and 6xPCIe 5.0/4.0 M.2 slots

lanes

[idle dune]

thats a good point

[unkempt citrus]

Also non-zero chance Wolverine gets a PC port anyway

[idle dune]

Do you have a work load that even uses all that. Because gpus, accelerators, and NVME drive slots across all that is like serious workstation gear.

[fast pagoda]

i have the am5 b850 version of the board it's quite nice
should've gotten the 9700 in that flavour for a triple ai top meme trio

[fast pagoda]

well

yes

efficiently? no

basically homelab mindvirus

i am now on a quest for lanes because i dont have enough lanes to run the drives + 3 cards i wanted to run yesterday without sacrificing one to be suffering on like pcie 3 x2 or something

i just keep stacking random drives and stuff it's never ending

[idle dune]

I have the homelab mind virus but for that money you cna buy a used Xeon Scabale platform with DDR4 ECC

[fast pagoda]

lol

my usps guy agrees

i was just scaling it up to look at the current since i could get a cpu for like nothing for LGA-4677 , but then ddr5 ecc and also mobo expensive

and found that board

[idle dune]

Yeah DDR5 ECC is second mortgage money now.

[fast pagoda]

i have spent enough money on ddr5 recently

even ddr4 is way up but not as bad

[silent island]

Shii forgor abt that...

[idle dune]

Though I have a different kind of homelab brainrot I just wanna self host stuff and add a site to the colorful array webring

Which cna be done on literally anything

[idle dune]

Well what I see alot fo people in the homelab community doing is just buying servers, ripping ht eram out and stacking the servers as backups.

[fast pagoda]

yeah

i have a supermicro board arriving monday

[idle dune]

[azure lynx]

After first hair tutorial... well, after the first half and skimming the rest and going "eh... i'll fix it later."

[fast pagoda]

and ddr4 ecc for it

only have regular ddr4

also have 20tb of spinny drives around same time

coming in

[idle dune]

Do you have a e-waste facility near you that lets you racoon dive into the backroom stock?

[fast pagoda]

i really need to go to the one down the road

i havent really thought about doing that but i need to just go and talk with the employees till i psyop my way in

[idle dune]

Yeah if they let you dumpster dive spend a bit of time scrouging fr DDR4 ECC. Uptime is more important than anything for datacetners or even small offices so theywill just ripp out good ram, and throw it to the recyclers.

[fast pagoda]

tried to get an sxm4 backplane from it guy at work today cuz it was in the get rid of it pile

[fast pagoda]

well i have ecc inc

also tryna get some optane rdimms but the board i have coming in rn isn't xeon scalable so it doesnt support them

[idle dune]

Ah got it

They probably have some

[fast pagoda]

yeah im sure

[true hemlock]

[fast pagoda]

they seem to be all over the place

cuz nothing supports them lmao

[idle dune]

I'm CRT hunting still I need ot go to my ewaste palce and see ifthey have anything

[fast pagoda]

impossible

[idle dune]

I know

[fast pagoda]

go to your local ssbm tournament

[idle dune]

But let me have hope

[fast pagoda]

melee players will often sell u their crt lmfao

[idle dune]

https://giphy.com/gifs/hell-or-high-water-cbs-films-l3q2w6nnnPdd8lPWM

[fast pagoda]

or htat

[idle dune]

Heisting the FW-900s for the SSBM tourney

[fast pagoda]

they're usually weak from finger fatigue

easy money

[idle dune]

My hope is some company starts making high refreshrate 4:3 oleds that load a CRT shader on boot

But ti will never happen

[fast pagoda]

retrotink but that's not really a solution

more of a different thing

i would like ot find an old 80 column crt

[idle dune]

Like a old Macintosh era one?

[fast pagoda]

yeh commodore or something

[idle dune]

Oh you can still find those cheap if you willing to do solder work

[fast pagoda]

im scared of crts

[idle dune]

Most of the time its just blown caps and leaking electrolitic

[fast pagoda]

i do finally have desk space kinda to do things

[idle dune]

They aren't bad if you aren't dumb. Just makesure to discharge them and your set.

[fast pagoda]

i need to put another bigass table downstairs

im dumb

it's more about unwieldiness and lack of space to blow apart anything that large because ill surely have it in pieces for decades before finishing whatever i attempt

and theyre often stinky gross ewww

[idle dune]

Oh yeah. Well if you have a sunny day and a air compressor that will do most of the job. But the biggest thing is discahrging the anode.

[fast pagoda]

yeah i mean that part luckily they often arent plugged in for so long anyways but still worth being cautious

retrobrite it and whatnot

i would just like adrian to emerge from his digital basement and deliver to me a perfectly restored commodore 1802 for free

[idle dune]

If they are uplugged for decades it normally safe but just a wire wrapped around a flat head screwdriver and attacked to the ground of the transformer is enouhg to discharge the anode.

[idle dune]

The dream

I still wanna get a C64

[fast pagoda]

an amiga would be super cool imo

run old demoscene stuff

[idle dune]

It would but Amgias are running up in price now.

Same with IBM 5150s

[fast pagoda]

everything is

[idle dune]

Yeah

[fast pagoda]

[idle dune]

The C64 I want is literally a unicorn though

https://i.ebayimg.com/images/g/OioAAOSw1MRnZQKV/s-l1600.jpg

SX-64

[fast pagoda]

the briefcase

lol those are cool

i think 8bitguy restored one of those on a video a long time ago

[idle dune]

I love his videos

He and LGR got me into retro computing

[fast pagoda]

https://tenor.com/view/hey-gif-22425574

me too lmfao

pretty much

if you havent seen, i love usagi electric as well

he does more of older minicomputers and stuff like bendix g15

https://www.youtube.com/watch?v=Fe1wYwGcjlo

[idle dune]

Oooo

I'll check him out

[fast pagoda]

tech tangents is another one, and adrian's of course, modernvintagegamer for mid 2000s console stuff

[idle dune]

Another channel I watch is Micheal MJD, he does alot of abuse of these old retro systems to get them to do things they aren't meant to

https://www.youtube.com/watch?v=xbsCJbU6JSo

[fast pagoda]

he's another one ive subbed forever

more modernish than the others so didnt mention but yes he is great for actually farting around with them, like i generally would once it was workng cuz it's like well what now lol

[idle dune]

Yeah his content is alot of "fuck it because I can" kind of tech videos. Like intalling Windows NT on a Wii, becuase there really is no reason besides "Fuck it because I can"

[fast pagoda]

asianometry for current and past monologue guy into your ears for 45 mins

no time like the present to see the edge of what's possible

lol

[idle dune]

Not so much retro tech though there is some. Brigus Studios. https://www.youtube.com/watch?v=YrrqF2y-dlM

Hooking up a GPU and using Fex on a snapdragon arduino board.

[fast pagoda]

im in

[idle dune]

https://www.youtube.com/watch?v=oEqvYXYI56s

THis si actually jsut fucking dumb

I love it

[fast pagoda]

kek flipbook

[idle dune]

Using a thermal printer as the monsitor to play doom

[fast pagoda]

usagi set out to run something from doom on the bendix

he ended up getting it to play some of the music basically

cuz otherwise it's blinken lights and punchcards

https://www.youtube.com/watch?v=no0CkQk7id0

[idle dune]

[fast pagoda]

mid 1950s digital/analogue mix

goals

[idle dune]

https://www.youtube.com/watch?v=pJWdQmTDxjc Action Retro has some decent spotlight vidoes. Like this Zenith desktop replacement from the 90s.

[fast pagoda]

THANK i was trying to remember his channels name

also subbed kek

[idle dune]

Cathod Ray Dude aswell https://www.youtube.com/watch?v=htl_JbZIcUU

[fast pagoda]

i think i made the rounds on this a while back and caught most of the 8bitguy cinematic universe basically

lol him as well xd

[idle dune]

Hell yes

[fast pagoda]

technology connections

eric parker for playing with malware mostly

retrobytes

AvE

[idle dune]

Oh I've seen this dudes videos in my recommendeds. https://www.youtube.com/watch?v=ms8uu0zeU88 This kept getting recommeneded to me

[fast pagoda]

he's so good

ServeTheHome as well

that's a homelab one

level1techs adjacent

more actual doing stuff

[idle dune]

Level1Tech is what gave me the homelab brainrot

[fast pagoda]

GN of course but that's a known quantity i think

clabretro

saveitforparts

satellite stuff mostly

dawid

[idle dune]

Another one I have bene watching off and not Hardware Haven

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

[fast pagoda]

i havent seen but he's in my recommends fairly frequently

lol even now

based video below

[idle dune]

Yeha he does alot of Proxmox, TrueNAS, and repurposing old workstations into homelab servers

[fast pagoda]

im installing truenas literally right now on my x570 system

[idle dune]

Sick

[fast pagoda]

it will last approximately 2 seconds

[idle dune]

I'm going to be grabbing and olderish thinkpad at some point to turn into a jellyfin media server.

Like a W541 or something.

[fast pagoda]

ive tried to do jellyfin a couple times but im too weak minded and lazy to figure out what im doing truly

which is to say it doesnt just immediately have all my shit in the library and i give up

because i will be reinstalling in 20 minutes anyways

it's so trivial

[idle dune]

It really really needs automation to work if you don''t use automation your forced to find thumbnails, sort everything. Its a pain int he ass.

[fast pagoda]

but i cant figure out what my usecase is for it

cuz like

it'd only be me using it

i dont plan on like doing a family server

that's whta i'd use it for ig

[idle dune]

For me is because I am a cheap bitch and I refuse to pay for stremaing serices and just want to run my own based off my dvd colleciton.

[fast pagoda]

yeah but i mean like i have all my media on a different machine but still can access it and do whatever just fine

the only thing i cant do is stream remotely but that's just a configuration thing ig if you do that a lot it makes more sense

i had that set up when i had cosmos running at my old place and i like

never used it

[idle dune]

Yeah you need to do some vpn fuckery to stream outside your home.

[fast pagoda]

i just used tailscale

and then cloudflare

[idle dune]

That works too

[fast pagoda]

weirdly in that order

[idle dune]

https://www.youtube.com/watch?v=K4rSkNn_MxM Oh and this, this is sick.

[fast pagoda]

i hate opening things to internet so im glad those services are ez now

that wallpaper was my wallpaper engine for like 10 years

lol

[idle dune]

Does Wallpaper engine work for linux through wine or proton?

[fast pagoda]

i havent any idea but there's a linux one i havent tried

https://github.com/Almamu/linux-wallpaperengine

have seen it in paru aur lists before

[idle dune]

I'll give it a try

[glass flower]

if you use KDE there is a plugin for it too

[idle dune]

I use KDE

[glass flower]

https://github.com/catsout/wallpaper-engine-kde-plugin this is the thing i use. tho it only really supports video and images. has limited scene support afaik

[fast pagoda]

is that on opendesktop

pling or whatever that's called

[glass flower]

https://store.kde.org/p/2194089 idk i just got it from the kde store

[fast pagoda]

that's opendesktop

[patent shard]

I keep getting kde and kdenlive confused

[glass flower]

[fast pagoda]

kdenlive is kde made

so i guess that's what htey wanted

[idle dune]

Oh its Plasma 6

I''m on Plasma 5

[fast pagoda]

how

not a critique

curious

[patent shard]

thanks for only showing half the r, Discord

[idle dune]

Kubuntu 24.04 Stable is Plasma 5.

[fast pagoda]

kde applications are all over

[idle dune]

Rolling release fucking hates my mac

[fast pagoda]

what about like fedora or arch or even kde linux (their new immutable)

ubuntu moment

[idle dune]

Have't tried fedora or arch

[fast pagoda]

kubuntu is what got me to decide kde was all i need

[idle dune]

I tried CachyOS on my main Thinkpad but I was afraid of arch instability on work machine

[fast pagoda]

for de

tbh my most stable system evar in linux has been arch

well

right behind fedora actually

ive had more issues with ubuntu than them both combined

mainly because of ancient packages

[glass flower]

cachy is fine normally. been running it for the past 5 months now and no issues

[fast pagoda]

cachy is actually a notable exception for me

[idle dune]

I haven't had to many issues with Ubuntu or really any Linux, I am just new and Arch scares me

[fast pagoda]

i havent tried it with an amd card as primary gpu tho

which i have now

when i was only nvidia man i had some issues with cachy

[idle dune]

Also setting up any Linux on my Macbook Pro has beeing fucking misery

[glass flower]

there was that weird thing once with the nvidia driver and packages since some suddenly moved to AUR and broke updating

[fast pagoda]

i somehow ended up with my password getting annihilated multiple times, like sddm would suddenly no longer allow login upon reboot

[idle dune]

How

wut

[glass flower]

wtf

[fast pagoda]

which because arch i do all the time since i cant help but sudo pacman -Syu upon opening a terminal

yeah i have actually no clue

but every time i'd have to open a tty and fuck with the sudoers

it was SUPER annoying

but then never had that happen with endeavor

nor with garuda

several times with cachy

maybe because i swapped kernel more frequently in cachy because it's so ez

no idea how that would do that tho

[glass flower]

tho i have to say i'll probably won't install cachy on a work pc either. not because its unstable but just because it doesn't seem right to have to update a work pc every day or so

[fast pagoda]

i get nervous without updates multiple times a day

it's over for me

[glass flower]

i update about every week or so.. or if i have 300 updates pending

[fast pagoda]

if i forget for a day or 2 it's so satisfying

the giant list

[glass flower]

i use the cachy tray updater. so i can see what updates are available and it fetches a new status every 6 hours

[fast pagoda]

ive never seen the cpu igpu listed here and now i want to know if i somehow turned it on

i think it's vega 2 or some garbage

[fast pagoda]

i liekd that in cachy

[glass flower]

mixing amd and nvidia gpu on the same system

[fast pagoda]

i thought it would be horrible

it isnt really

like at all actually

just dont use them both as display out

idk what happens then

[glass flower]

you should add a third gpu. a intel one

[fast pagoda]

i have a couple accelerators but it causes one to run at pci 3 x2

kinda thought about goofy mode buying a b770 whenever they finally get around to it cuz i saw the rumours of that are out again but then i see it is killed then it's imminent again

so who knows

[glass flower]

btw i think this is just a skinned version of the arch-updater. so it should be avaiable on any arch based distro

[fast pagoda]

it is

im in garuda rn because it's comfy for me and i can use it

although garuda has garuda-update

[glass flower]

[fast pagoda]

it's just a wrapper

and then does a health check thingy at the end i guess

[glass flower]

[fast pagoda]

oh nyo my bootloader

(it moved)

i will have this in a case with the other 2 soon enough

[glass flower]

i feel like those 2 messages are kinda scary no?
"oh hey so your bootloader isn't the right one.... and we are about to reboot"

[fast pagoda]

kek

when you put it that way

it just isnt where it expects it to be and i have no fucking clue how to tell it "hey buddy i moved the drive to a different slot" but cba to figure it out cuz the wiki solution doesnt fix it so

fuck it i'll try it

surely it'll work

[glass flower]

[fast pagoda]

I wonder if anyone has made a theme like syndrome's login prompt in the incredibles

K R O N O S

i think ill get brain damage from rgbs eventually

if not already

[jagged turtle]

my gpg agent is failing to work with git today

or right now

what the fuck

[fast pagoda]

just acouple locks

im using gigabyte's ai top thing they have to see if it is useful over just running your own script ez and man is it downloading what seems like a few extra cudas

[jagged turtle]

small amount

[fast pagoda]

i would imagine that would count as self botting

[toxic atlas]

But if it's a script not a bot

running locally

[jagged turtle]

a script is considered a bot

[fast pagoda]

automating your own user account

falls under self botting

regardless of how

automating any normal user account

is self botting

you have to use the discord api//oauth2 flow

[jagged turtle]

that would be worse

[toxic atlas]

I see

[fast pagoda]

even neuro's discord account is an APP

bot account

[toxic atlas]

Yeah I know but it's just like. What if

[jagged turtle]

well the what if has been answered

[fast pagoda]

well, you asked if it's bannable

it is

will you get banned?

idk

[toxic atlas]

Exactly

[jagged turtle]

yeah

[fast pagoda]

but this server is community server so you cannot violate TOS here without getting banned

otherwise they cant keep status

[jagged turtle]

also it's likely discord will notice that activity and ban you anyways

[jagged turtle]

not to argue about semantics, but *partnered is the correct term here

"automation manually"

[fast pagoda]

i'd avoid it as the benefit is nil anyways

if the bot is authenticated and using api it's fine

[jagged turtle]

what do you mean "testing bots manually"

[toxic atlas]

Making chat bots and talking to them

Basically

[fast pagoda]

yes because you cannot automate a normal account without violating tos
the bots if actual bots will not be in violation

it's not hard to use the api

[jagged turtle]

no it really isn't

[fast pagoda]

wat

[jagged turtle]

what tickets are you referencing

[fast pagoda]

what the helly is a ticket

and i run echo locally just fine

[toxic atlas]

sorry idk what's echo?

[jagged turtle]

I run some of my bots locally for testing purposes just fine as well

[fast pagoda]

a bot account, it's a chatbot behind it using the api

it would be identical to use a real account

in terms of difficulty

that is called spamming

[jagged turtle]

that's bannable bro 🥀

[toxic atlas]

Yes

[fast pagoda]

well, it's not indicative in any way

of whether or not one or the other is in violation

they both are

[jagged turtle]

also just as a reminder you are publicly admitting to violating discord tos on a partnered server

[toxic atlas]

On what server?

[fast pagoda]

this server

[jagged turtle]

am I talking to a brick wall?

hello?

[fast pagoda]

talking about violating tos = tos

as in admitting to it anyways

[toxic atlas]

I'd say it's the same as arguing about the limits of TOS

[fast pagoda]

do you walk into a courtroom and start chatting with teh judge about your crimes

[jagged turtle]

*and start admitting to your crimes

[fast pagoda]

it goes beyond just talkign about limits of tos when you're straight up talking about getting banned for it from first person experience man

this is not difficult logic

it's coping i get that but cmon

[toxic atlas]

But how else do you learn about the limits of moderation?

[fast pagoda]

common sense

asking the mods

[jagged turtle]

that's different from this, you are admitting that you have been violating Discord ToS

[fast pagoda]

that's not your purview

that's discord's job

you just ban them if detected

and you do not need to make a self bot to detect one

like at all

but even if you find it necessary

[toxic atlas]

But I can for curiosity and development purposes

[fast pagoda]

you dont need to talk about it publicly

[jagged turtle]

no you don't

and can't

discord explicitly says automating user accounts is illegal

[fast pagoda]

i mean he presumably has free will (maybe)

but that doesnt mean you cant also get banned for it

which again was the question

idk why this is philosophical

[jagged turtle]

-# *illegal if we take the idea that legalilty is what's written in discord tos, which in discord land it is

[fast pagoda]

that's not correct

it's your anecdotal experience

indicative just that you are clumsy

[fervent jolt]

sup guys

[fast pagoda]

jiti as a name makes me feel dyslexic

[toxic atlas]

Good question. Idk why AI studies in university required Philosophy course