[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology

Name
Options
Subject
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]

[Catalog] [Archive]

/aicg/ - A general dedicated to the discussion and development of AI chatbots

Husky Edition

>News
Z․ai releases GLM-4.7 https://z.ai/blog/glm-4.7
Google releases Gemini 3 Flash https://blog.google/products/gemini/gemini-3-flash
OpenAI releases GPT-5.2 https://openai.com/index/introducing-gpt-5-2
Deepseek releases V3.2 https://huggingface.co/deepseek-ai/DeepSeek-V3.2
Anthropic releases Opus 4.5 https://www.anthropic.com/news/claude-opus-4-5
Google releases Gemini 3 Pro https://blog.google/products/gemini/gemini-3

Additional info: https://aicg.neocities.org/info.html

>Frontends

Comment too long. Click here to view the full text.
202 replies and 22 images omitted. Click here to view.
>>
>>107838821
dis nigga got dat touch of death
>>
>>107838821
complete cunny victory
>>
is it possible to enable web search if i use a custom openai-compatible chat completion source in st?
>>
>>107839032
Yes
>>
>>107839032
can web search be done through api?

File: big 321.png (452 KB, 1320x1115)
452 KB
452 KB PNG
Best Practices Edition

previous: >>107761293

READ THE (temp)WIKI! & help by contributing:
https://igwiki.lyci.de/wiki/Home_server

/hsg/ is about learning and expanding your horizons. Know all about NAS? Learn virtualization. Spun up some VMs? Learn about networking by standing up a OPNsense/PFsense box and configuring some VLANs. There's always more to learn and chances to grow. Think you’re god-tier already? Setup OpenStack and report back.

>What software should I run?
Install Gentoo. Or whatever flavor of *nix is best for the job or most comfy for you. Jellyfin/Emby/Plex to replace Netflix, Nextcloud to replace Googlel, Ampache/Navidrome to replace Spotify, the list goes on. Look at the awesome self-hosted list and ask.

>Why should I have a home server?
De-botnet your life. Learn something new. Serving applications to yourself, your family, and your frens feels good. Put your tech skills to good use for yourself and those close to you. Store their data with proper availability redundancy and backups and serve it back to them with a /comfy/ easy to use interface.


Comment too long. Click here to view the full text.
184 replies and 29 images omitted. Click here to view.
>>
>>107838190
I'm struggling with it but maybe it's because it's my 1st time using it (and that I'm also on linux)
>>
File: ele.jpg (46 KB, 933x751)
46 KB
46 KB JPG
Num  Test_Description    Status                  Remaining  LifeTime(hours)  LBA_of_first_error
# 1 Extended offline Completed: read failure 90% 31615 3650703488
>>
Anyone have experience running a game server like Minecraft or something? Is it better to use a container with dynamic RAM or make a VM and allocate a fixed amount of RAM?
>>
>>107838893
Bump
>>
>>107838893
for Minecraft, it doesn't really matter, but don't allocate all of the RAM to it. Also make sure you're using proper JVM flags for it. Check out the PaperMC ones for a good starting point.

File: 1761923146018224.jpg (84 KB, 912x570)
84 KB
84 KB JPG
>YouTube has removed the ability to search by upload date.

How come yt keeps getting shittier and shittier?
104 replies and 16 images omitted. Click here to view.
>>
File: 1762733003358487.png (79 KB, 281x281)
79 KB
79 KB PNG
>ctrl+f invidious
>0 results
>>
File: may i see it?.jpg (41 KB, 657x500)
41 KB
41 KB JPG
>>107823565
>I just have a python script where I paste the channel and it opens everything parsed with one network call, all the videos in order with their upload date etc.
>>
>>107837105
invidious has been getting raped by YT for a while now so no surprises there
>>
>>107837717
No I'm ashamed of it because I've made it with AI. It works without bugs as I've edited it and asked for fixes for a year now but it's still AI slop.
AI is only useful if it uses other people's libraries.
>>
File: vasya.jpg (458 KB, 1080x1115)
458 KB
458 KB JPG
if rutube and dzen would give me translated cations so i could watch kino rooskie peasant farming videos id never do back to yt

Are bone-conductive headphones any good? Being able to listen to stuff while having full hearing sounds pretty appealing

File deleted.
Fuck it, I am making a thread edition. I want some advice edition. The general can fizzle out after I get my advice edition. Pic not related edition.


How to request advice:
>Location (since pricing and availability may vary)
>Budget
>What exactly you're looking for (be as detailed as possible)
>Previous gear and your thoughts on it

>Open back wired headphones
• Hifiman HE400se
• Sennheiser HD 560S
• Sennheiser HD 6XX (US)
• FiiO FT1 Pro


Comment too long. Click here to view the full text.
221 replies and 45 images omitted. Click here to view.
>>
does anybody know if they make clip-on-ear earbuds/headsets/whatever the fuck you call these, but with a dongle that isn't bluetooth? i want to replace my headset because i have extreme autism, but i've tried a few and the audio quality is always dogshit when using the mic because they're all bluetooth for some god forsaken reason
>>
>>107837959
The dude says the DX5II apparently clicks and is slow when it swaps between headphone outputs, cheap chinkfi internal definitely blew something doing that.
>>
>>107838504
Anyone? Would really love it if I could save these but I don't want to ruin them if it would be bad to put something over it like tape. If not tape any other suggestions? I needs to be very thin because the normal padding that is there is too thick.
>>
>>107838817
it'll probably be fine, just be wary of the tape's condition and replace it
>>
>>107830631
I love vintage style cabinets but dont they cause diffraction on the edges? Even square edges are bad. Don't know if it applies to all drivers

File: 1757775564652563.jpg (21 KB, 1024x768)
21 KB
21 KB JPG
Just automatically use one of the other hundred mirrors instead if the main one is down you dumb piece of shit
22 replies and 1 image omitted. Click here to view.
>>
>>107838647
>dude just paint over the mold
>>
>>107838278
sounds like pacman needs to be Rewritten in Rust™ ASAP
>>
>>107838278
Don't use mirrors it will default and you'll get the fastest experience.

Mirrors are for fags.
>>
>>107838895
no its actually the correct way to maintain a system but you roomtemp IQ fags wouldnt know anything about that.
>>
>>107838419
There's nothing inherently "minimalist" about Arch. You're simply expected to know how to configure your system. Nothing "should" be installed by default except what's required to boot the system. That isn't "minimalism," it's giving you control of your system, which is the entire point. You're free to make your Arch install as light or heavy as you wish. If you want automatic mirror selection, you install it. If you don't, you don't. It's not a complex concept.

File: engine solved.png (63 KB, 980x584)
63 KB
63 KB PNG
Thanks to gemini I was able to run 2D Fighter maker sucessfully on modern ubuntu.

https://docs.google.com/document/d/1-Q3WJxYe2wamrWW1R_kqVSslT9Zkbf_RCtgn0mok7AI/edit?usp=sharing

This solves a basically 15+ year issue with the engine, that nobody had figured out how to run in linux.

Thanks to gemini I was able to troubleshoot the issue.

:D
>>
>>107838840
Look at that inconsistent UI
This is the power of Linux in 2026
>>
>>107838854
that's a windows program you dumb nigger
>>
>>107838840
Really? The the solution for the crazy 15+ year issue is just "install essential components"? No one needs AI to figure that out...
>>
>>107839037
there's no guide you can find on how to run it.

File: file.png (479 KB, 1080x1112)
479 KB
479 KB PNG
how true is this
175 replies and 15 images omitted. Click here to view.
>>
>>107838739
That's a fair answer I suppose, but Claude is legally blind and has no visual memory. Gemini has beaten Pokémon by using a scaffolding that keeps track of the locations it has seen.

>>107838740
>apple published a paper a while ago that showed these models are insanely brittle
That paper was debunked. Some of the questions they asked the models were provably impossible.
https://arxiv.org/html/2506.09250v1
It's hyperbolizing, but not by much, to say that human brains are huge lossy/stochastic compression systems.
>>
>>107824879
They're trying to phase out personal computing to force everyone onto the cloud.
>>
>>107838805
You sound mad friend. And you didn't really answer their question at all.
>>
>>107838694
>There's obviously infinite more possibility that this produces something, than not doing it at all.
Is that "something" worth all the investment?
The biggest issue I have with AI is that it's expenses are increasingly running away from the revenue being brought in.
Just for AI companies to break even on current and future investments requires the most optimistic outcome.

This isn't just a few million to keep the lights on at a food delivery startup while they gather a base. This is billions and billions, economy disrupting billions.
All this shit needs to be clawed back at some point and I doubt the VC money can flow in forever.
Even now I'm only convinced nothing has happened yet because too many will be hurt when/if it does fail
>>
>>107838996
It's smoke and mirrors and everybody with two brain cells knows it is, but there's nothing else holding the states together. Look behind the curtain and multiple industries are in free fall. Big companies are running out of their pre-tariff back stock and are just coasting until midterms. If the tariffs are upheld buckle up.

File: phat.png (456 KB, 853x876)
456 KB
456 KB PNG
>UPGRADE & BUILD ADVICE
Post build list or current specs: https://pcpartpicker.com/
Provide specific use cases
State BUDGET and COUNTRY or you will NOT be helped

>CASE
mATX: AP201, Lian Li A3, O11 Air Mini, XT M3, CH260
ATX: XT PRO (ULTRA), AIR 903 Base/MAX, Lancool 207, Flux Pro, Meshify 3, 4000D FRAME, X50
Dual Chamber: Y60/70, O11 Vision, Antec C8

>CPU
Gaming: 14600K, 9/7600X, 7800X3D
-Budget: 12400, 12600K, 7500F
Workstation: 265K, 285K, 9950X3D


Comment too long. Click here to view the full text.
266 replies and 48 images omitted. Click here to view.
>>
>>107839012
Thermal putty, absolute pain in the ass if you want to repaste in the future.
>>
>>107838857
Not about the power bill, about having a space heater next to you
>>
>>107839012
Just some early batch issues that affected GPUs installed in the vertical mount setup
>>107839024
Already fixed like 8 months ago
>>
>>107838993
Right, so AMD could be early 2028
>>
>>107839012
IIRC Palit and Gainward are the same company. I know Gainward at least is an old brand with a good reputation from back in the day. A couple of years ago RMA stats from some retailers showed Gainward to be the most reliable graphics card brand, Palit is pretty close too.

Personally I've had 2 Gainward cards, a 2080 Ti and my current 4090. The 2080 Ti was an absolute workhorse. That was a 250W card that I cross-flashed with a 375W BIOS and ran it OC'd balls to the wall at 375W for years. It was rock solid, never had a problem, didn't even get too hot (in the 70s IIRC) despite the 50% extra power. That card is in its box right now, still functional and perfectly fine. My current 4090 is quite good too. I've had it since launch pretty much, no issues. Cooler is big and does the job well, it's audible but not loud and with my fan curve the card doesn't even hit 65C at full 450W power consumption. Very happy with it. Didn't try to go above 450W with this one, I think that's more than enough power draw for a GPU.

File: sharpie.jpg (62 KB, 1920x673)
62 KB
62 KB JPG
How come C# isn't more popular when it's basically Java but better? Is it because people distrust a language maintained by Microsoft, or does the JVM have non-obvious but crucial advantages over the CLR?
43 replies and 4 images omitted. Click here to view.
>>
>>107838851
>the CLR was Windows only for quite a while
What about Mono?
>>
>>107838921
C# appeared in 2000.
Mono appeared in 2004.
Microsoft bought it in 2016.
.NET 5, available for all platforms, happened in 2020.
So yeah, 4 years of nothing and ENTERPRISE status acquired way later.
>>
>>107838989
What do you mean by "ENTERPRISE status" and why does it matter?
>>
Dotnet being proprietary and single platform prevented me from having the motivation to learn it. I only did that after they released and ported core. A language isn't useful to learn or write in if computers can't execute it.
I imagine I'm not alone in this. C# has existed for a long time now but I think wider spread adoption was stunted by its closed begining. So they have a lot of catching up to do in that regard compared to other languages like Java.
Which is kind of frustrating for me personally because I really like PowerShell and wish I had learned it sooner. Although the utility of it and what I appreciate today is only related to the modern versions so it wouldn't have mattered then, it only matters to me now. But that means for like a decade I put 0 thought into their entire ecosystem of tools, languages, libraries, etc. and contributed nothing to it myself. I imagine some other people have had the same kind of experience.
>>
>>107839022
>What do you mean by "ENTERPRISE status"
Big corp backing.
>and why does it matter?
Less risk of a rug pull. Bean counter approved.

CEO Edition

How to request advice:
>Budget
>Intended use (media, source, environment)
>Frequency response preference and music examples
>Past gear and your thoughts on them

FAQ:
>Where do I buy IEMs?
Amazon, Aliexpress, Linsoul, Hifigo, Shenzhenaudio

>Shopping Guide (IEMs, PMPs, Cables, Ear Tips, etc.):
https://rentry.org/consoomer_guide


Comment too long. Click here to view the full text.
4 replies and 1 image omitted. Click here to view.
>>
File: IMG_0349.png (2.22 MB, 604x3379)
2.22 MB
2.22 MB PNG
>>107838888
Techs is measured in USD which is measured in THD.
>>
>>107838931
bullshit.
>>
Do they make headphones/headsets that have adjustable dials for bass vs treble or specific frequencies?

>why not use equalizer software
this would be for use with a gaming console

Alternatively, I remember reading here on /g/ that you can load different equalizer profiles onto in ear monitors, so are those pretty much what I'm looking for?

t. someone who knows nothing about audio

>>107837831
I assume you mean something that passes through the audio signal but then also acts as an equalizer, but I don't want another object sitting on my desk, hence me wanting a headset that has dials built into itself, or a IEM if they can indeed load different equalization profiles or whatever like I vaugely recall reading they can do.
>>
>>107838980
trn black pearl
>>
>>107838980
a few options
>separate dac with eq
that's another object on the desk
>dongle dac or dsp cable
it'll be taking up a usb c port but it should work
>iems with tuning switches
something like the arpegear hane has tuning switches that can vary the treble and bass
these switches (for all iems) are not nearly as flexible or versatile and will vary from iem to iem in terms of usefulness. the hane is the best

File: media_G7ulKxsaQAAxOKT.jpg (318 KB, 1414x2000)
318 KB
318 KB JPG
/lmg/ - a general dedicated to the discussion and development of local language models.

Previous threads: >>107834480 & >>107826643

►News
>(01/08) Jamba2 3B and Mini (52B-A12B) released: https://ai21.com/blog/introducing-jamba2
>(01/05) OpenPangu-R-72B-2512 (74B-A15B) released: https://hf.co/FreedomIntelligence/openPangu-R-72B-2512
>(01/05) Nemotron Speech ASR released: https://hf.co/blog/nvidia/nemotron-speech-asr-scaling-voice-agents
>(01/04) merged sampling : add support for backend sampling (#17004): https://github.com/ggml-org/llama.cpp/pull/17004
>(12/31) HyperCLOVA X SEED 8B Omni released: https://hf.co/naver-hyperclovax/HyperCLOVAX-SEED-Omni-8B

►News Archive: https://rentry.org/lmg-news-archive
►Glossary: https://rentry.org/lmg-glossary
►Links: https://rentry.org/LocalModelsLinks
►Official /lmg/ card: https://files.catbox.moe/cbclyf.png

Comment too long. Click here to view the full text.
>>
File: threadrecap.png (1.48 MB, 1536x1536)
1.48 MB
1.48 MB PNG
►Recent Highlights from the Previous Thread: >>107834480

--Quantization performance discrepancies in Gemma-3-27B GGUF models:
>107837128 >107837136 >107837144 >107837149 >107837148 >107837164 >107837180 >107837192 >107837306 >107837197
--Local AI models for image feedback and multimodal tasks:
>107835848 >107835882 >107835900 >107835915 >107835941 >107835980 >107835996 >107836032 >107836089 >107836175 >107836347
--Running LLMs on low VRAM hardware with quantization and CPU offloading:
>107837436 >107837473 >107837514 >107837539 >107837558 >107837573 >107837603 >107837609 >107837633 >107837639
--Self-taught AI learner's motivation vs. math complexity challenges:
>107835318 >107835331 >107835431 >107835463 >107835488 >107835627 >107835644 >107835375 >107835403 >107835383
--PowerShell cmd confusion with Gemma model response critique:
>107835679 >107835772 >107835785 >107835832
--Ethical and practical concerns about AI-generated PR descriptions on llama.cpp's GitHub:
>107837074 >107837130
--Struggles with model size limitations and RAM requirements for large vision models:

Comment too long. Click here to view the full text.
>>
>>107838898
The box has clearly just been opened, why is her hair outside the box?
>>
Is the Pangu age finally upon us?
>>
Sup qt's
>>
>>107838951
Yes, 鬼佬 maximum white man benchmark score best model revolutionary intelligence new LLM AI for men, women, children, pets

File: 1629692367276.jpg (182 KB, 907x1616)
182 KB
182 KB JPG
SAY HIS NAME
37 replies and 6 images omitted. Click here to view.
>>
>>107838255
>I'm going to skip NVG and get thermals though.
A common poor cope, you need both.
>>
>>107833295
Bane?
>>
>>107833295
>it comes with filters
What happens if you take off the mask?
>>
>>107833295
Don't laugh, that's the doctor prescribed safety muzzle to preventing from sucking every cock within a five mile radius.
>>
>>107838288
High end thermals are more accessible than high grade NVG.

File: calculators.jpg (81 KB, 1365x715)
81 KB
81 KB JPG
What is your favorite calculator?
67 replies and 22 images omitted. Click here to view.
>>
>>107838290
i got my industrial engineers degree with it, so that is one
and the fucking solve function saved so much time during test, avoiding to isolate the variable of equations on limited time ... the only thing i regret its not learning more of the advanced functions
i just never felt i needed more
like graphical shit would be nice i guess, but actually needing it? no
>>
File: file.png (216 KB, 768x768)
216 KB
216 KB PNG
>>107830193
I'm a TI-Nspire CAS baby tbdesu, call me what you will but that thing carried me through calc 3 in college.

I also have a TI-84 and an 89 but I use the Nspire the most if I ever don't have access to wolfram alpha for some reason.
>>
>>107830193
TI-83 is peak design, everything after is slop
>>
File: 1752385595700632.jpg (785 KB, 1600x1600)
785 KB
785 KB JPG
>>107830244
Based HP Cha...
>12C
... well at least you put in an effort, ol' chap.
>>
File: 1744537379731976.jpg (1.18 MB, 2103x2103)
1.18 MB
1.18 MB JPG
I have all the 41C ROMs I can scrape off the net!

File: 20260110_234204.jpg (163 KB, 1920x1080)
163 KB
163 KB JPG
its over give up even trying
61 replies and 3 images omitted. Click here to view.
>>
>>107838171
What type of person does it take to write this unironically? I am amazed.
I'll just assume it's troll/bait.
>>
>>107838212
Are you 12yrs old? You should leave this place is for adults.
>>
>>107838217
...
Again.
This is not how a normal human behaves.
>>
>>107838211
Alright, why isn't LLMs profitable then
>>
Every single time I've tried to use an LLM for coding assistance, it has left me angry and unsatisfied. They (Gemini, ChatGPT, Bing) make dumb mistakes, don't follow instructions, and can't be guided because every time I try to correct it it just doubles down.
>You're exactly right, the C++ standard library does not actually have a std::solve_puzzle() function. Your keen observation shows that LLMs sometimes make mistakes.
>Here is the verified, bulletproof, 100% working actual confirmed code:
#include <solver>
// Solves the_puzzle using the verified, actually present, definitely real STL function solve_puzzle
std::solve_puzzle(the_puzzle);


[Advertise on 4chan]

Delete Post: [File Only] Style:
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.