I have been trying Linux for a year now. I know my way around the system, but I still feel like I rely way too much on AI when I don’t know what to do
There is a growing fear inside me that I will never truly know how to do stuff without asking AI, and that makes me feel like a fraud
I don’t think using AI as a replacement for Stackoverflow counts as vibe coding.
Linux is tough. Stackoverflow was shitty and hostile.
Don’t beat yourself up about it.
To add to this, if you’re using an LLM that cites its sources, read the cited section (and ideally a bit before and after it). You’ll end up learning things gradually and having progressively fewer questions to ask the LLM.
And definitely read and try to understand any code or commands that it tells you to run. You never know when it’ll hallucinate something potentially destructive.
you havne’t even grasped the scale of the things you don’t know
makes me feel like a fraud
cut that shit out. a fraud of being a 1 year linux user? get over yourself. it’s like a fraud of a thing you can be a fraud of.
Learn how to use
grepandmanYou can learn a lot when you actually read through the man pages, and usually there’s a pretty standard structure so grep works well if you know roughly what you’re trying to do.
Also following issue trackers and mailing lists can be helpful.
Edit: Also try out
fishornu.fishauto complete is amazing and embeds man info into the tab complete options (including flags and such).xonshis kinda cool too if you are comfortable with Python, and less comfortable with bash.Just a warning, don’t use these as your login shell, just run them from your distro’s shell when you want to switch over.
RTFM (Read the friendly manual). Theres the arch wiki that has pages on most any system component thats probably common across 90% of distributions. Theres project websites, YouTube videos, other wikis like the Debian wiki or Gentoo wiki. Theres peoples blogposts where they talk about cool things they did.
For example:
Try looking for holistic sources of information basically rather than sifting through aggregators like LLMs or page rankings (which take you to listicles generated by LLMs).
Also post on c/libre if you have any software related questions (which should I use, etc).
This post could be from 2016 and titled “How to be less reliant on stackexchange”. Suffice to say, our ambition to do stuff on our computers sometimes outruns our technical acumen. I built up my own skillset by just furiously searching through forum posts trying to find the one post where someone had the exact command I needed. I’ve copy and pasted a lot of shit into the terminal without understanding what it really did.
I think there’s a couple things you should do to help build up your own brainpower:
- Don’t copy and paste stuff into the CLI, actually physically type it out. The contours of the command will actually touch your brain.
- Just think about stuff. Give yourself like 45 seconds to think about how you would do something before you ask the AI chatbot. When the AI chatbot is giving you answers, actually read through them and think about whether they make sense.
There’s two ways of going about things. Copy pasting without understanding which means you never learn. Or understanding what you’re being given to run to learn.
Ai, stack exchange, some docs if you are running stuff without trying to understand what its doing you won’t learn things. If you take the time to understand it then you’ll be fine. As much as i hate ai its rather good at giving you a quick explanation of what you’re doing. If you are relinquishing your judgement to ai then yeah you wont be learning anything.
I’ve been using linux for about 18 years now its not something you ever really stop learning your way around. What I tend to do is keep notes in Joplin when i figure out how to do something so I can come back later. Its been invaluable making my own docs for the stuff i need to do
Oh and then there’s the things you’ll never learn to do. Like tar and the find command. I never remember them, i dont know why but i just dont
Find is easy. Off the dome: The search dir comes first which is whack. Then you want single dashes for the long options. Type is like f,d, or l? depth is what it says on the tin. Then regexes for the pattern and you’re good?
Tar you just flail at the keyboard. cvaf I think is the magic string to compress stuff. Picking a single file out of an archive or something though always requires studying the manual.
My eyes blurred reading this comment and I continue to not know find. And I will never know find till my last breath. You cant make me know it
Growth comrade.
A fascist hacked out an awk script to parse files today, did you?
Yeah? Well while you were out partying with find and awk I studied the xrandr, i was moulded by iptables, born with ffmpeg. I am the one true knower of the xorg codebase
just google stuff. but dont use google use ddg
I prefer to use official docs, if available, e.g.
--help, man pages, GNU info,/usr/share/doc, official website. The official documentation is usually more correct and more comprehensive than other resources, though sometimes a bit technical. Other than that, the arch wiki (and I’m using Debian) is a good place to look.If I’m seeing an error, I find it’s most useful to collect as much information as possible. If it’s a GUI program, you’ll need to start it in a terminal, it’ll likely spam errors and warnings to stderr. Many programs have
-vor--verboseor--debugoptions, or a debug mode that can be enabled in the configuration file. Look at the logs if there are any (usually throughjournalctlthese days). It’s sometimes already obvious what the problem is just from the log/stderr output, but if not then just paste it into Google.Also, official bugtrackers are good. When I encounter a bug in a package, I go to the Debian bugtracker first, and almost always someone has already reported it, and left useful information there.
tbh i would never have learned anything in linux if i had been born 30 years later than i was
I first got into Linux around 2021 and used google and cheat sheets for everything, all AI did was consolidate that
AI, in this case, is the new Google. Or stack overflow. Or whatever the cool kids were using.
I think as long as you’re asking how to do it and not just asking it to do it for you, you’re actually probably going to get a better learning experience using a modern LLM to learn stuff like this than almost any other resource other than a very experienced person walking you through it one on one
Just don’t use it then. Use Google or post to forums, read solutions, understand the solutions, then implement the solutions. Using these steps helps to create a deeper understanding of the OS as a whole (regardless of which OS you’re using honestly). Now admittedly, I’ve never used an AI prompt (and never will), so I don’t know how it structures its results, but I imagine it skips the steps for learning or obfuscates them in some manner. Also be sure to use man pages to help as they contain an absolute wealth of information themselves.
Just remember people have been learning Linux long before LLMs came into existence. Trust me when I say you can learn Linux without its assistance, too.
As a note, I’ve been using Linux since 2007 when I picked up Ubuntu on a whim. It was dual booted for the next 11 years and then became my daily driver shortly thereafter. The steps outlined above are how I became proficient at using the OS.
Could try guessing but that’s risky, its either that or google or look at manpages and hope you can decipher what you want.
I’ve been using Linux a long time (since 2007). There’s a few things I know a fair bit about, and that came from hours of problem solving, reading forums, chatting on IRC and xmpp, etc. There’s value in that process, but goddam, so many times I just wanted answers. Like an expert friend to always show me the answer. AI is pretty much that, with some caveats. Try to go slow and understand what it’s telling you. Ask it to explain. Create documentation notes as you go.















