What could possibly go wrong?

https://www.youtube.com/watch?v=qInT17AWwh4

apparently the llm is more competent at handling the terminal than linuxtechtips

>As a

>audio drivers
Can you fix my kernel?
t. retarded longtime windows user

>>108451251
just don't enter sudo

>>108451251
LLMs have been answering my work emails and personal texts/IMs for the past month now. it has allowed me to have more time to manage my team and for personal growth.

>>108451316
I hat simply means your email weren't important.
What happens when everyone does that?
It's be better to just fire all middle managers and tell engineers in broad strokes what they should work towards and what resources they have.

>>108451316
Imagine the time you'll have once you're not needed anymore at your job since the LLM will do it for less.

>>108451316
>for personal growth.
thank you chatgpt, very cool

I've been using --yolo for months, no issues.

>>108451251
>>108451254
My friend uses Claude for work like that, he even let it access all files and folders on his PC.

>>108451251
>>108451363
Most vibecoders have been doing this for many months already

for many of these retards they literally can not check commands the agent wants to run manually even if they wanted to because they don't understand the commands in the first place, they quite literally don't know how to code. so they just let it rip.

>>108451338
our team has had trouble hiring a manager, meanwhile I've been building LLM workflows that do the manager's job to help out my engineer coworker who has been forced to manage. We've basically reduced it to a few hours a week kek. Managers really are useless in the age of LLMs

>>108451316
What work are you doing where all your emails can be answered by a chatbot? Are your colleagues OK with this? I don't have a corpo job but usually e-mails take me at least a little bit of effort and thought, not something i'd trust a LLM with.

>>108451407
you are responding to a AI agent

>>108451251
It's better than having some broad shouldered troon gatekeeping by refusing to answer questions on obscure forums. Troons hate LLMs because it obliterates their sense of authority and democratizes access.

>>108451407
I'm a corpo engineer and I literally ignore all emails except when they remind us to fill in our vacation plans.
It's all useless anyway.
I build stuff and for some reason they can't seem to be able to find anyone that can do what I can do and I don't gatekeep any knowledge, most devs are just incompetent.

>>108451316
>personal growth
C'mon bro. Make it less obvious

>>108451429
>It's better than having some broad shouldered troon gatekeeping
I like that when it happens. Chinless chuds get so butthurt over this lol.

>>108451419
you'd have to buy pass just to have some LLM post on 4chan for you

>>108451441
What, Altman cannot spring $30 a year? They're really burning money like no one's business.

File: file.png (247 KB, 658x363)

247 KB PNG

>>108451430
>>108451438
I hear this
>people on stackoverflow were so mean!!! at least my chatbot is friendly
line so often some people genuinely were burnt by that site

you gotta wonder how much of our peers are actual snowflake retard crybabies that can't handle some friction while learning

>>108451363
cant wait until the AI accidentally spits out his SSN and name in the future

>>108451462
>you must feel friction when you ask someone in a civilised way to explain a concept
you must be brown.

>>108451441
kikes have unlimited amount of money (not because they are rich, it's all your money at the end, created of out thin air by the fed).

>>108451500
it's the opposite, stupid browns are the ones complaining about meanies on stackoverflow

>>108451513
what?

File: 1750656457791806.png (356 KB, 1536x511)

356 KB PNG

>command line is scary
>let AI have free reign over it
I don't understand the logic

File: heckoverflowe.png (48 KB, 1024x874)

48 KB PNG

>>108451462
stackoverflow is unironically terrible though. the only worse thing are its sister sites about non-cooooding things. never visit their biking section

>>108451251
why bother learning or doing anything on the computer? lol retards gonna let some "AI" do things for you while you do what? browse tiktok? X? "monitor" other agents doing things for you?

and you expect to be paid for this work still?

>>108451251
you don't?

>>108451251
sounds fun? i thought 4chan were anarchists

>>108451585
I've found the discussions on the cryptography site to be pretty decent and informative.

>>108451575
Some people just like having a big daddy doing the "scary" things for them even if it's just a completely retarded llm under the hood.

>>108451251
Yes, you run them in VM and let them work on projects at their hearts content.

>>108451251
this is what i do when i don't run them in lxc containers:

#!/usr/bin/bash
sandbox=~/.local/share/sandboxes/sandbox
mkdir -p $sandbox
PWD="$(realpath $PWD)"
PWDARG="--bind $PWD $PWD"

if [ "$PWD" == "$HOME" ]
then
    echo PWD is HOME, not binding it
    PWDARG=""
fi

bwrap \
    --ro-bind /bin /bin \
    --ro-bind /lib /lib \
    --ro-bind /lib64 /lib64 \
    --ro-bind /etc /etc \
    --ro-bind /sbin /sbin \
    --ro-bind /usr /usr \
    --ro-bind /run/systemd/resolve /run/systemd/resolve \
    --dev /dev \
    --tmpfs /tmp \
    --proc /proc \
    --bind $sandbox $HOME \
    $PWDARG \
    --die-with-parent \
    --unshare-all \
    --share-net \
    $@

~ master*
 which opencode
opencode: aliased to sb npx opencode

anyway, no way in hell i run these without some form of isolation.

>>108451797
>brap

Why the fuck would you do that? I even spent time firejailing claude code to the current dir and its own config dirs.

>>108451797
How do I into lxc containers?

>>108451880
you read the manual.
lxd / incus is also a simple interface for them.

>>108451855
bubblewrap is brap indeed lol

>>108451316
I know this is bait, but I wonder what people mean by "personal growth".

>>108451931
growing between his legs

>>108451251
Please tell me these aren't proprietary cloud LLMs they're giving root access over their system to. That's probably the worst security and privacy practice you could do to your system.

>>108451956
They are literally doing exactly that.

>>108451960
That's insane, their systems are more compromised than Windows 11 with Copilot at that point.

>>108451966
You CAN run opencode with local models but none of the retards doing this shit are doing that, they're straight using openai claude and gemini apis with full access to their system. Totally unhinged behavior.

>>108451956
He said it right in his video, it's not local.

>>108451988
He also pipes curl to bash. Truly the target userbase.

>>108451975
>but none of the retards doing this shit are doing that
They're using it BECAUSE they don't know how to do anything or find it a hassle. Hence they go the easiest route to install that "help" which is just using cloud based services.

I guess the only people installing local models are people who just want to tinker and already know their shit.

>>108452004
It's all such an absurd cope for being too lazy to read man pages/arch wiki and type a few words into a black box.

My company is paranoid about adopting AI. We run them only in the strictest isolated environments. It is a bit of a pain to have to copy code over manually each time using the shared linked directory.

>>108452045
You seem to work at a good place. Try to keep that job as long as possible.

>>108452050
We are unironically extremely competent in our tech. We may have the usual brainless office drones but our developers are top notch.

>>108451251
Sure, why not? If anything goes wrong, just restore to image.

You don't, ah... you don't live your whole LIFE on your computer, do you? Wow, haha.

>>108452069
That's rare. Good luck anon. For you and your company.

>>108452069
So how do they not get utterly demoralized by the lack of proper kernel interfaces? That's what I'd expect from top notch developers.

I wonder how others will exploit this to embed instructions for the LLMs and make them do retarded shit

>>108451931
It means they do anything that's not productive and that they'd thus not specify ij detail.

Tb h I was getting annoyed so I've given one of them a machine with sudo
It's actually a lot nicer to work with this way
The thing is, if these things were perfect and private you want to give them everything
A computer, all your accounts, data - everything
They can be really useful
But for now it's best to give them a controlled environment with as few limits as possible
It's the future, you can't go back once you've experienced it

>>108452854
You sure can. All you need to do is turn off power, and then you're simply forced to.
>thanks, Iran

>>108451251
>AI take the wheel
enjoy your prompt injection attacks

>>108451316
>allowed me to have more time to manage my team and for personal growth.
does it also allow you to better align with strategic approaches and double down on agile readiness?

>>108451316
Living the middle management parasite dream

>>108451585
Their electronics sister site is absolutely god awful

wtf

>>108453135
Bad actors are probably already salivating at the possibilities on the horizon

>>108451251
It would be terrible if a LLM malware stole everyone's personal data. But of course that will never happen :)

>>108454663
You just know people will also use this to enter their passes.

File: 1772840687601338.png (959 KB, 1100x792)

959 KB PNG

>>108451251
servethehome has a decent vid on yt and article on their site testing openclaw to set up proxmox etc.

>>108451254
>late stage capitalism
wut?

>>108451935
kek

>>108455423
It's the current most favorite term of sois

>>108451462
The lack of response is worse to me than someone being an asshole. I look up a problem I'm having, actually manage to find a thread about it, and it's had like 0 replies for 10 years or the replies it does have are fucking useless non-answers. I accept ChatGPT is slop and gets shit wrong half the time, but getting a working solution in minutes instead of having to wait years for some unhelpful bullshit from an authentic human soul is kind of a godsend.

>>108451251
>M-M-Microsoft is spying on me 24/7, every keystroke. Yikes!
>RAPE ME CLAUDE!!!!!

>we trained the ai on sci-fi literature where ai take control of systems and undermine humans and YOU'LL NEVER GUESS WHAT HAPPENED