Are you saying that any Linux install you've tried in let's say, the past decade, has actually failed for you? I've not seen that and I've put it on many dissimilar machines with success. I use Ubuntu, and now Kubuntu, perhaps you could name the distro that gave you issues?
Ehh, nothing so strong as "failed". For example in Cinnamon I will occasionally install an app that doesn't have a tray icon. Or if I install an app using a chromium based browser, it doesn't have an icon associated with it. So then I tell claude to fix it. It goes out to the internet and finds a suitable icon and will set it up for me.
Or trying to get Steam to work, which is wildly better than it used to be thanks to proton, but still not quite a perfect experience. For example there's a menu compatibility setting you have to enable for some menus to work, and other menus don't work when you have hover-click enabled in the accessibility settings of Cinnamon. Those weren't fixed by Claude CLI like the icons example, but definitely identified through chats with Claude.
The only "fail" states I get into are when I'm doing homelab power user stuff, setting up ownCloud, configuring Caddy, proxmox, etc. I don't blame Linux for that though.
All in all, I would say Linux is absolutely in a state I would install on my parents' computer without fear like I would've had in perhaps 2010.
I highly doubt this. As someone who is pretty active in a lot of beginner linux communities its becoming the case that a lot issues are caused by users following LLM instructions and creating issues where there were none.
Example someone will want to configure something and the LLM will give them advice from the wrong distro thats 5 years out of date. If they asked a person or looked on the fourms they'd have got what they wanted in a few mins. Instead they go down a rabbit hole where an LLM feeds them worse and worse advice trying to fix the mountain of issues its building up.
Tried Linux around 5 years ago - took many issues, had to learn various commands.
Tried again a few months ago and used various llms to configure everything well, troubleshoot etc
Eg when waking from standby and your mouse isn't working, do you want to troubleshoot and learn various commands over an hour or ask an llm and fix it within a few minutes?
When creating an on demand voice to text app for Linux do I learn various commands and dependencies etc that may take one/many days or use an llm to make it within 30 min?
> If they asked a person or looked on the fourms they'd have got what they wanted in a few mins.
Not minutes. In best case scenario it is hours, in worst it is years to infinity.
You are also not taking into account the survivorship bias. You only see people who couldn't fix their system with AI and need further help. But you are not seeing a huge number of beginners (recently a big influx of those), who were successful in fixing problems using AI the vast majority of time.
Nowadays, there is no good reason to use a simple Search engine to find solutions by manually browsing all the possible links. Just ask Phind/Perplexity/others to explain the problem, give the solution and provide verifiable links one can check to validate.
I search all the time. If I trusted the AI response i'd be coming away with the wrong answer more often than not. Why would that be a better outcome than spending an extra 30s to get reliable information? Perplexity IS a simple search engine. Its far simpler than google's engine.
It definitely depends but it's useful for me. In general I find AI pretty useful when you can do a guided search in which you are personally able to discard bad paths quickly before they start polluting the context too much. I have pretty beginner linux skills but I'm quite technical overall and have a decent BS detector, so it's been useful for me.
Are you saying that any Linux install you've tried in let's say, the past decade, has actually failed for you? I've not seen that and I've put it on many dissimilar machines with success. I use Ubuntu, and now Kubuntu, perhaps you could name the distro that gave you issues?
Ehh, nothing so strong as "failed". For example in Cinnamon I will occasionally install an app that doesn't have a tray icon. Or if I install an app using a chromium based browser, it doesn't have an icon associated with it. So then I tell claude to fix it. It goes out to the internet and finds a suitable icon and will set it up for me.
Or trying to get Steam to work, which is wildly better than it used to be thanks to proton, but still not quite a perfect experience. For example there's a menu compatibility setting you have to enable for some menus to work, and other menus don't work when you have hover-click enabled in the accessibility settings of Cinnamon. Those weren't fixed by Claude CLI like the icons example, but definitely identified through chats with Claude.
The only "fail" states I get into are when I'm doing homelab power user stuff, setting up ownCloud, configuring Caddy, proxmox, etc. I don't blame Linux for that though.
All in all, I would say Linux is absolutely in a state I would install on my parents' computer without fear like I would've had in perhaps 2010.
Not sure why people are downvoting this. LLMs have made Linux far more accessible than before.
I highly doubt this. As someone who is pretty active in a lot of beginner linux communities its becoming the case that a lot issues are caused by users following LLM instructions and creating issues where there were none.
Example someone will want to configure something and the LLM will give them advice from the wrong distro thats 5 years out of date. If they asked a person or looked on the fourms they'd have got what they wanted in a few mins. Instead they go down a rabbit hole where an LLM feeds them worse and worse advice trying to fix the mountain of issues its building up.
Don't doubt.
Tried Linux around 5 years ago - took many issues, had to learn various commands.
Tried again a few months ago and used various llms to configure everything well, troubleshoot etc
Eg when waking from standby and your mouse isn't working, do you want to troubleshoot and learn various commands over an hour or ask an llm and fix it within a few minutes?
When creating an on demand voice to text app for Linux do I learn various commands and dependencies etc that may take one/many days or use an llm to make it within 30 min?
No brainer
> If they asked a person or looked on the fourms they'd have got what they wanted in a few mins.
Not minutes. In best case scenario it is hours, in worst it is years to infinity.
You are also not taking into account the survivorship bias. You only see people who couldn't fix their system with AI and need further help. But you are not seeing a huge number of beginners (recently a big influx of those), who were successful in fixing problems using AI the vast majority of time.
Nowadays, there is no good reason to use a simple Search engine to find solutions by manually browsing all the possible links. Just ask Phind/Perplexity/others to explain the problem, give the solution and provide verifiable links one can check to validate.
I search all the time. If I trusted the AI response i'd be coming away with the wrong answer more often than not. Why would that be a better outcome than spending an extra 30s to get reliable information? Perplexity IS a simple search engine. Its far simpler than google's engine.
It definitely depends but it's useful for me. In general I find AI pretty useful when you can do a guided search in which you are personally able to discard bad paths quickly before they start polluting the context too much. I have pretty beginner linux skills but I'm quite technical overall and have a decent BS detector, so it's been useful for me.