The rise of Linux desktop is inevitable — it’s time music software developers got on board

[Zeisner]

My big question, lately, is how will AI coding help Linux and open source projects develop, in general?

It will (or already does) make things even worse.

[BertKoor]

You can also try to build your own rocket, hoping it will take you to Mars one day. There's plenty of people trying to do that...

It's just the one lunatic, and his nameless minnionX.

[Zeisner]

It's just the one lunatic, and his nameless minnionX.

He's not the only one. I've seen documentaries about funny guys who were building rockets in their backyards big enough to carry a person, of course without a degree in engineering. Quite scary to think that one of your neighbors might be one of them. All those highly flammable and explosive materials in their garages...

[audiojunkie]

My big question, lately, is how will AI coding help Linux and open source projects develop, in general? Can it multiply manpower and talent? Will it be a giant bogey or a boon? I'm sure there's a lot of strong takes on this, but I want to see the actual results. Which I suspect should become apparent soon.

I think it has the potential to be a real boon. We are in the early days though, and there is a lot of distrust (some of it rightfully so). But in the end, AI is just a tool to be used for good or bad. Just like the internet. Just like a gun.

I've been digging into AI quite a bit recently, and I'm actually quite amazed at what is possible with it. The potential for good is very high--just like the potential for bad. However, I've been learning some prompting tricks to help reduce the errors. AI makes a ton of errors and you have to really watch out because it can tell you something that is blatantly false--it doesn't really have the ability to reason to weed out errors. You have to specifically tell it to check (and confirm item by item according to criteria that you set) in order to minimize errors. And even then, you'll still get errors.

Yet, at the exact same time, it has the ability (because of so much data it has been trained on), to find specific things that you may be researching for in seconds what may take a person weeks to do manually. What I always do to minimize error is to build into my prompts instructions that force it to verify and confirm answers through cross verification. Then, even after that, I tell it to provide evidence of its accuracy with weblinks to its evidence. And then I go and check manually the evidence. I don't mind this really, because double checking its work is nothing to the relative time that is saved in doing the research in the first place.

So, I expect that the generations of AI to come will likely have some of this evidence checking and cross referencing built in at some point as part of its error checking and correction. So, I expect it will just get better and better.

I recently heard on the radio that scientists have been having tremendous success with genealogy records and tracking potential genetic traits that cause hereditary health problems . Likewise, through all of this, they are starting to determine which medicines may be more appropriate for one person versus another.

I suspect that we may live long enough that our grand children could spit in a tube, and the genetics could be pulled, and a few minutes later potential problems that arise around puberty (when many of the genetic problems show up in life), or adulthood, such as mental problems or specific problems with major organs could be diagnosed ahead of the problem starting, and the particular medicine that is likely to work best for a person's genetics type could be suggested before the symptoms even shows up. They are already doing something similar with breast cancer, so it's not too much of a stretch that AI could vastly improve this within our lifetime.

So yes, I think AI has a great potential for good--and bad.

As far as implications for AI and Linux, vibe coding and no code prompt engineering are both things that exist right now, and ask AI improves, these things are bound to improve as well. I am already seeing many vibe coded open source projects popping up everywhere. I think as time goes on, the potential for slop will be reduced, and the quality of AI code will go up. I think that there is going to soon be an explosion of vibe coded projects that support Linux. The whole reason I got interested in all of this is because I have been waiting 20 plus years for someone to create a gui based SFZ editor, and no one has ever built it. So, I'm interested in using no code prompt engineering to put one together. I've already been experimenting with vibe coded SFZ scripts, and I've getting a lot of success! I have a basic sfz template that grants sfz the ability to control through Midi CC pretty much everything one would expect in a normal subtractive synth-- ADSR, volume, pan, vibrato, cutoff freq, resonance, etc etc etc. That has always been really important to me, because so many sfz instruments don't have this capability programmed into the script, and that often leaves the sfz instrument more rudimentary that other sample formats. So that's where my interest in AI started.

So, yeah. Nothing is perfect. AI makes lots of mistakes. But it also has a lot of potential. I think its capabilities will continue to improve, and I think Linux will benefit from AI as time goes on.

[audiojunkie]

My big question, lately, is how will AI coding help Linux and open source projects develop, in general?

It will (or already does) make things even worse.

You're a glass half empty kinda guy, aren't you.

[lunardigs]

My big question, lately, is how will AI coding help Linux and open source projects develop, in general? Can it multiply manpower and talent? Will it be a giant bogey or a boon? I'm sure there's a lot of strong takes on this, but I want to see the actual results. Which I suspect should become apparent soon.

...

So, yeah. Nothing is perfect. AI makes lots of mistakes. But it also has a lot of potential. I think its capabilities will continue to improve, and I think Linux will benefit from AI as time goes on.

I tend to think the same. Right now it's the pioneer phase of this stuff, with lots of weird, scary and wonderful lessons in store. It should become more orderly and usable though as we go along. We'll see!

I think the big moment I'm waiting/hoping for is when the tech becomes a bit more mundane and liberated in terms of hardware requirement and we're not reliant on the bigs in order to access it.

I tend to think software itself is going to radically change as well, not just how we develop it, but how we look at it. Where perhaps a computer workstation/OS is just an agent that spins software at once, doing whatever you ask it to do. Generic, amorphous computing/software in other words. Where purpose built software one day stops being a major thing.