스크립트
00:00:00Linus, the creator of Linux, shared a really interesting rant about the use of AI in the
00:00:06Linux kernel. And I think some people will need to hear that because I've also said it in pretty
00:00:13much all of my episodes, videos over the last years, you will not have a future as a software
00:00:20engineer, as a professional software engineer, if you're acting like AI isn't there and isn't
00:00:26useful. You can debate AI. It has disadvantages. It has certainly taken some of the joy away from me
00:00:33when it comes to software engineering. I've shared my thoughts about that too. I'm also finding new
00:00:37joy more and more, but you can't ignore it. You can't act like it's only bad and doesn't have any
00:00:44disadvantages. That is why this is really interesting. I realized that some people really
00:00:49dislike AI, but this is an area where I'm willing to absolutely put my foot down as the top level
00:00:54maintainer of Linux. Linux is not one of those anti-AI projects where one of the most prominent
00:01:00examples would be the SIG project. But yeah, Linux is not one of those anti-AI projects. And if
00:01:05somebody has issues with that, they can do the open source thing and fork it. Now, to be very clear
00:01:11here, I am fully aware that AI has many negative consequences for open source maintainers. They get
00:01:18swamped with slop pull requests and issues and it's hard to keep up with all of them because
00:01:25there is an infinite amount of slop coming in and you as a maintainer feel pressured to work your way
00:01:31through that or ignore it all entirely. Which of course is also not the idea of open source. I totally
00:01:37get that. However, of course, the answer to that can't be to totally ignore AI. The answer may be to ignore
00:01:47issues and pull requests for you. But that does not imply that you yourself shouldn't be using AI or that some core
00:01:54contributors shouldn't be using AI. And for example, the SIG project has a very firm and strict anti-AI policy about not
00:02:03accepting pull requests, for example, that are about AI-generated code or where the code has been AI-generated.
00:02:10There are repositories, there are projects out there with pretty strict policies. And of course, every maintainer can do what they
00:02:17want. But I am convinced this is not the way forward, not for maintainers and not for software engineers in general.
00:02:23You have to use AI as what it is, as a tool. As a tool, you have to learn. You have to learn to wield efficiently and it also is quickly evolving. So how you use AI also changes. And I know there are many developers out there who may have used AI a year or two ago or who may be using it in their companies with restricted or outdated models or in environments where you can't do a lot with those models. And I totally get that AI doesn't feel super capable if that's the case.
00:02:51But play around with it on your own machines and be open to it. Embrace it as a learning process. It is a tool where we all need to get better and where we all need to figure out how to best use it.
00:03:02And there is this full range, right? You can be in the wipe coding area and not care about the code at all. You can be in the agentic engineering area where you do care about the code, but still use AI, where you do build or write your specs, your plans, where you review stuff, but where you also automate stuff.
00:03:21It's not black or white. It's not one or zero, even though some people like to make it that, but that's not what it is. It is a spectrum and it is something that's quickly evolving and where we still need to learn.
00:03:34You don't need to build loops where you try to automate everything, but you also should probably not just be using AI for for auto completion anymore.
00:03:44You should really embrace it and try things out and be open to failing and to learning and also open to improving the process over time.
00:03:54And try out using agent skills and see how that could improve things and how fine tuning these skills to your specific workflows could improve things.
00:04:02This is all what I'm doing and what most people that are working with AI are doing.
00:04:07They are trying to figure out how to wield it effectively, and that is why this is such a beautiful take here, because what Linus is saying, obviously, is that Linux will be built or will be maintained with help of AI.
00:04:21And people that don't want that can just walk away.
00:04:25AI is a tool, just like other tools we use, and it's clearly a useful one.
00:04:29Just what I also said.
00:04:31And of course, most people that are working with AI efficiently are saying that I am not talking here about the vibe coding area where AI is not a tool, but a replacement for you.
00:04:44It is a super useful and powerful tool, a quickly evolving one, as mentioned.
00:04:49It may not have been that clearly even just a year ago, but it's no longer in question today.
00:04:53And I think that's also a very good point here.
00:04:55And we all can feel it, at least if you are engaging with those newer models, if you're using them a year ago.
00:05:03So in summer 2025, all these tools and all these models were obviously way, way worse.
00:05:10So now I was wrong here around the change from 2025 to 26 in December.
00:05:17I predicted, or in early January, I predicted that models would probably not become that much better.
00:05:22And I was clearly wrong there.
00:05:23They did become much better because they got much better at following instructions, at calling tools, using skills, keeping on working for longer periods of time.
00:05:36The post training the AI companies did there was and is really effective.
00:05:41And of course, they are still getting better.
00:05:43But as I also said back then, the tooling where we use these models also got much better.
00:05:48We have amazing coding harnesses like Pi.
00:05:52We have cloud code, we have codecs, we have so many and most of them are pretty, pretty good.
00:05:56And of course, are also evolving and are really, well, can be used such that those models can work very efficiently in them, that they can use sub-agents, that they can use these agent skills and so on.
00:06:10And that clearly has changed compared to a year ago.
00:06:13And yeah, of course, you can get more done now than you could a year ago.
00:06:18There are other questions around AI, like what the economy of it will actually look like in the end.
00:06:23But is it useful is no longer one of those questions.
00:06:25Anybody who doubts that clearly hasn't actually used it.
00:06:28And I fully agree here.
00:06:29And yeah, again, really, I know some people are just rage baiting and that's fine.
00:06:35That's the internet.
00:06:36I'm aware of that.
00:06:37But if you truly believe AI should be ignored, think again.
00:06:44This is a very dangerous attitude and opinion if you want to stay in that field.
00:06:50And I couldn't care less if an individual person leaves software engineering or not.
00:06:56But I've always tried to create content that is helpful where I teach people stuff and I'll continue doing that.
00:07:02But part of that is that I want to share my opinion.
00:07:05And please, I urge you, if you are very skeptical of AI, give it a try and learn how to use it and be open to it.
00:07:13It's really important.
00:07:15And it is the future of software engineering.
00:07:17And this is just to be clear again, because there will be some comments.
00:07:20This is not a new take.
00:07:21I've said that in pretty much all of my episodes before too.
00:07:25AI is a tool.
00:07:26You need to learn how to use it.
00:07:28Yes, it can also be somewhat a painful tool, but for maintainer workloads, both for maintainer workloads and just from it keeps finding embarrassing bugs standpoint.
00:07:37Yeah, that's what I also mentioned before regarding maintainers being swamped with AI stuff.
00:07:42But the solution is not to put your head in the sand and sing la la la.
00:07:45I can hear you at the top of your voice like some people seem to do.
00:07:48Totally agree.
00:07:49The solution is to make sure those LLM tools help maintainers instead of just causing them pain.
00:07:53There's no question on that side.
00:07:55And that is exactly what I also meant here.
00:07:57You got to be open.
00:07:58You got to learn how to use that tool.
00:08:00And I will also create more resources also on that other channel, the Academy channel, where I try to share more on that for free on how I'm using these tools.
00:08:12And that may be helpful to some of you too.
00:08:15We're not forcing anybody to use it, but I will very loudly ignore people who try to argue against other people from using it.
00:08:20And no, AI isn't perfect, but Christ, anybody who points to the problems at AI had better be looking in the mirror and pointing at themselves at the same time.
00:08:27Because it's not like natural intelligence is always all that great either.
00:08:31And I mean, that is a good point.
00:08:34And I find myself in a position where you expect more from AI than you do from a human being.
00:08:41And at the same time, you often trust a human being more than AI.
00:08:45Even though, of course, humans also are prone to do errors, right?
00:08:50We, nobody's perfect.
00:08:51We've all written our fair share of horrible code or introduced bugs, and we are not perfect.
00:08:56One big difference, of course, is agency.
00:08:59As a human, you own a piece of code.
00:09:01You are responsible for that code.
00:09:03And with AI, that's not the case.
00:09:05The AI doesn't own it.
00:09:06The AI isn't responsible for it.
00:09:08It's still you, the human.
00:09:09And maybe that is why we expect more from AI.
00:09:13Because if something, some tool is writing the code for you, but you are responsible, maybe that is why you get pretty angry if that code is bad.
00:09:22Also, of course, because that thing, the AI is threatening to take your job.
00:09:27That's definitely the feeling you can get, obviously, for obvious reasons.
00:09:30And therefore, even more.
00:09:32You think, well, you better do the job right then if you're going to take mine.
00:09:35I don't know.
00:09:36I definitely know that feeling, too, that you expect more from AI than from humans.
00:09:40But it is a good reminder, of course, that humans aren't perfect either.
00:09:44And again, it is that combination.
00:09:46I truly, deeply believe in that.
00:09:48It is that combination.
00:09:50And it has been that combination for the last 40, 50, 60 years of technology and humans that can produce pretty amazing stuff.
00:09:59And yeah, therefore, that is a very important point here, I believe.
00:10:04The kernel project has been and will continue to be about the technology.
00:10:07Sure, the social angle of working on open source is important and often a very motivating part of the project.
00:10:12But in the end, that's a side.
00:10:13But if it's not the point of the project, this is not some kind of social warrior project, never has been and never will be.
00:10:21And I mean, I'm not sure if it's meant like that.
00:10:24But you could take that as a little as some shots fire towards some other projects like the SIG project.
00:10:32But there are others, too, which are very anti-AI, where it feels more like it's a mission rather than trying to build a helpful technology or tool.
00:10:43And I might be totally wrong here.
00:10:45Don't worry primarily about producing good results or outcomes, but more about maybe the art of writing code, which I totally get to some extent.
00:10:55But not if you do it professionally.
00:10:57That's just my take there.
00:10:59And I, as I mentioned also in other videos, there has been, I said it before, some joy taken away by AI.
00:11:06I did like that art of writing code by hand, but it is now really just that to me, an art.
00:11:13I can do that as a hobby, but not professionally.
00:11:17Writing all the code by hand, that is just not the future.
00:11:21Professionally.
00:11:22Again, I can just say it over and over again.
00:11:26In the kernel community, we do open source because it results in better technology, not because of religious reasons.
00:11:32And so we make decisions primarily based on technical merit, not fear of new tools.
00:11:37And I think some people need to read this.
00:11:39I know many people already notice or probably have a similar opinion, but some people really need to
00:11:47question their stance on AI and maybe become a bit more open minded if you want to stay in that field.
00:11:54If you are just here for rage baiting, or if you're just doing this as a hobby, which is totally fine,
00:12:01then yeah, sure, you can totally ignore it, but it's not a fad.
00:12:04It's not going away.
00:12:05And again, I've been saying that for ages.
00:12:09So yeah, try to be a bit more open there maybe and use AI as a tool.
00:12:16It is a really helpful tool.
00:12:19And I think Linus has a beautiful point here.