스크립트
00:00:00Thank you.
00:00:30Hello, hello, hi everybody, good morning or good evening from my side.
00:00:48Yeah, and finally another stream.
00:00:52It's been a while since the last one, I think two or three weeks.
00:00:55Hi Frostbain, hi Martz06.
00:00:59Great to have you here, great to have you here.
00:01:02As I just said, it's been a while since the last stream.
00:01:04So I figured today, why don't we start with a little look back at what happened over the
00:01:12last couple of weeks and just a little ask me anything session so that you can ask your
00:01:18questions and I can hopefully answer them.
00:01:22So yeah, hi everybody.
00:01:23Thanks.
00:01:23Thank you all for tuning in and for the excitement.
00:01:26I'm excited too.
00:01:27It's great to be back.
00:01:28And yeah, it's been two weeks.
00:01:30So in case you don't know, I try to stream every Thursday, but I don't make it every Thursday
00:01:36because life.
00:01:39But for example, next week should work.
00:01:43And then it'll always be this time on a Thursday.
00:01:47So yeah, as mentioned, the idea is to have a little ask me anything session.
00:01:53So definitely feel free to ask any questions you have.
00:01:56And I also got some things prepared, of course, because I want to take a look back at what
00:02:02happened over those last couple of weeks where I've been gone.
00:02:06And thanks so much for the great feedback, Felix.
00:02:09I'm happy that you're liking the VPS course.
00:02:11So maybe a little shameless plug in case you don't know.
00:02:15Obviously, I got a bunch of courses.
00:02:17And one of them is the VPS essentials course, where I teach you how to securely set up a
00:02:22VPS so that you can host a website there or whatever.
00:02:24So that may be interesting.
00:02:26Happy you're liking the course, Felix.
00:02:28And obviously, there are a bunch of other courses as well.
00:02:31If you want to get all of them, you can, of course, also get our membership.
00:02:34It's a subscription.
00:02:35And as long as you have it, you have access to all courses.
00:02:38By the way, speaking of courses, I got a BUN course coming up.
00:02:43And thank you so much for the subscription gift trophy.
00:02:47Much, much appreciated.
00:02:50Thank you so much.
00:02:51And yeah, in case you don't know, I got a BUN course coming up.
00:02:55So that may be interesting too.
00:02:56And if you want to get notified about that, the best way is, of course, to be in our Discord
00:03:01server and or join our newsletter on academy.com slash community.
00:03:08So enough of the advertisement.
00:03:10Just figured that may be interesting.
00:03:11And let me dive right into the first question here.
00:03:16What's your feelings about TanStick Start in comparison to Next.js?
00:03:21I think that's a great question.
00:03:23I personally have pretty much exclusively, I would say, used TanStick Start over the last
00:03:31eight months or so.
00:03:34Because I think it's really an amazing framework, a full stack React framework, a great alternative
00:03:41to Next.js.
00:03:42Now, I want to be clear, obviously, Next.js is great too.
00:03:47And I've used it so much over the last years.
00:03:50I'm a big fan of Next.js, but I will also say it can be a bit slow during development, especially
00:03:56though that got better, but it can be slow.
00:04:00But I like the ergonomics of TanStick Start more.
00:04:06I will say getting started with TanStick Start was a bit more difficult for me because the
00:04:12documentation was not great.
00:04:14It got better, and of course, nowadays with AI, you can ask AI about it too, and maybe you
00:04:19have AI build significant parts of the application.
00:04:22So that all is changing, and I'm still trying to figure out how to best go about AI and coding
00:04:28and so on.
00:04:29I guess we all are.
00:04:30But yeah, TanStick Start, I think, is a great framework.
00:04:34Feeling a bit more modern, faster.
00:04:37I like its ideas.
00:04:40I like its server functions implementation.
00:04:44I don't use React server components with TanStick Start.
00:04:47They do support it now, but I'm not using it.
00:04:50Yeah.
00:04:51So personally, I like it a lot, but that, of course, is just my personal opinion.
00:04:55Now, speaking of personal opinions, next question, what do I prefer, codex or cloud?
00:05:00So the coding agent I use actually, or these days I use, is Pi.
00:05:05I like Pi a lot.
00:05:07It's a very slim, lightweight coding agent, a terminal-based coding agent, and you can use
00:05:15it with your codex, with your ChatGPT subscription, for example, and I'm using it with it.
00:05:21And occasionally, I do go back to cloud code or to codex.
00:05:24So right now, I'm using cloud code, for example, because I'm playing around with the
00:05:29Fable model, the famous Fable model, and I'll get back to that in a couple of minutes, too,
00:05:34I guess.
00:05:36But yeah.
00:05:36So right now, it's mostly Pi with the codex models or the OpenAI models, but I'm constantly
00:05:42switching also to stay updated and see how they work and so on.
00:05:46But I like the lightweight, fast feeling of Pi.
00:05:52And then, by the way, I'm reading all these nice messages and so on.
00:05:55Thank you all so, so much.
00:05:56And thank you all for tuning in.
00:05:57It means a lot to me.
00:05:58And I'm super happy to have you all here in the stream.
00:06:01So yeah.
00:06:02Thank you all so much for that.
00:06:06Now, I'll get back to those other questions here, by the way.
00:06:12Just, oh, well, no.
00:06:14Let's stick to the questions for now.
00:06:17What would a big company like, why would a big company like Cloudflare acquire Void Zero
00:06:23in the day of wipe coding?
00:06:26So I think Cloudflare, I don't think, I mean, they have said it pretty clearly.
00:06:31Cloudflare is obviously racing to become the kind of the default cloud or at least
00:06:39a popular choice as a cloud for applications built with AI.
00:06:48So I think all these cloud providers and especially the, how do I call it, the more
00:06:55developer-focused clouds like Vercel or Cloudflare, I think their big dream is to be
00:07:01the default cloud for all these wipe-coded apps or agentic-engineered apps.
00:07:06So that as a software developer, building software obviously becomes faster, easier.
00:07:13Maybe I shouldn't say obviously, but it can become faster and easier with AI.
00:07:18And the big question is, where are you deploying all these applications that are being built, right?
00:07:24And you want to have a cloud or an offering that's easy to use for AI.
00:07:31So for a CLI and so on, and where you have all the building blocks you need for any kind of application in the end.
00:07:39So where you have queues, databases, and so on.
00:07:41Now the building blocks already were there with Cloudflare for the most part.
00:07:45Obviously they're shipping heavily on that front too, to add all the missing pieces.
00:07:50But having a framework like void zero sit on top of that as one option that may be easy and nice to use for agents since it integrates so tightly with the underlying cloud infrastructure.
00:08:03I think that is one of the reasons why they acquired it, to have like a compelling all-included offering where you can just point your agent at the void zero docs in the future maybe,
00:08:15and then ask it to build whatever you want, and it has all the building blocks it needs, and then guess what?
00:08:20The application gets deployed on Cloudflare, and obviously they earn money off that.
00:08:24That is how I would reason about that acquisition.
00:08:31Yeah.
00:08:32But by the way, this is obviously an ask me anything session.
00:08:36But please also share your thoughts and opinions, because I'm just sharing mine.
00:08:42Obviously, don't have to be true or the absolute source of truth.
00:08:48I've bought your in-depth course on React and laughed how detailed it is.
00:08:52I was wondering if there's a detailed course on backend too.
00:08:56Not on backend in general.
00:08:58Obviously, I have a Node.js course, which may be interesting.
00:09:01I mentioned a couple of minutes ago, I'm working on a BUN course right now, which will be far, far, far shorter than the React course.
00:09:08It's because, frankly, the days of these 70-hour courses are over, and I'm figuring out how to do good courses that are in line with what people expect and need these days,
00:09:20how we build software and so on is shifting.
00:09:22But yeah, I'm working on that course.
00:09:24I don't have a general backend development course, though, I will say.
00:09:28How can we recognize those managers that have AI psychosis?
00:09:34Well, I guess it's probably not too difficult to recognize those people.
00:09:44By the way, in general, I would say it's...
00:09:48So managers in general are pretty vulnerable or likely to be affected by AI psychosis.
00:09:58So simply expecting it is probably a good default.
00:10:03And I think we're at a moment in time where there's so much hype and so much...
00:10:11There is a lot of real capability too, right?
00:10:14It's not all just hype, but there is a lot of hype too.
00:10:18That if you don't dive deeper, which is what managers often don't do, it can seem like, oh, God, the future is already here and AI can automate everything.
00:10:27Because on the surface, it can certainly look like that.
00:10:30So I guess it's a good default to assume that most managers have that picture.
00:10:36And obviously that is kind of the annoying part that you're likely the person who has to dive deeper and sees, okay, yeah, I can use AI for a lot of things and it can be helpful.
00:10:46But it's not there that it does everything magically perfectly all the time.
00:10:51And I guess, yeah, that's kind of the annoying part of the times we're living in right now.
00:10:56And obviously it doesn't help that the entire internet and YouTube and X is all hyped up about that AI stuff.
00:11:02And at the same time, there obviously is a source of truth to it because AI can be helpful.
00:11:10I wonder if it's too late to learn React at the age of 38.
00:11:13I would probably not just learn React because I think what's changing is as a developer, we don't know what we'll exactly do in five years from now.
00:11:27I've shared it before and I still believe will not be replaced by AI.
00:11:31But what we do will definitely change a lot.
00:11:34So you probably will not just learn React anymore and be a super and a super in-depth React expert.
00:11:42Instead, you will probably need to learn about all the different building blocks it takes to engineer an application, web application or whatever it is.
00:11:52And you need to be capable of understanding the technologies you're using there.
00:11:57You need to be capable of judging the output.
00:12:00An AI gives you order the strategy it follows, but I don't think you need to be capable of writing the code for advanced React features on your own.
00:12:16That's my guess.
00:12:17I may be wrong.
00:12:18You should be able to understand it.
00:12:20You must know about these features and you can certainly still learn that.
00:12:24But since it's, I believe it's less about writing all that code in the future and more about understanding, you can learn more quicker because reading and understanding something to some extent is easier than writing it.
00:12:38Of course, you don't have that in-depth understanding of it then.
00:12:42And that's what I mean.
00:12:43I'm still also trying to figure out how exactly does that role look like in the future and where do you need to go deeper?
00:12:49Because obviously being an expert in something is useful and is also fun, to be honest, but you need kind of both a broad knowledge base and some in-depth knowledge here and there.
00:13:01And that's like, again, the part which I'm also still trying to figure out how that should look like in the future.
00:13:08Do you isolate Pi or do you use some sort of extension to restrict its reach?
00:13:13Yeah, good question for Pi.
00:13:18By the way, I'm not a fan of websites that make you wait before you can do something.
00:13:23For Pi, in case you don't know, Pi is a very lightweight coding agent and it's also very extensible and it has a great extensions API and it can write its own extensions.
00:13:34So it can manipulate itself, which is really a killer feature.
00:13:38And there are also like a packages extensions store.
00:13:41It's free, though.
00:13:43And I don't have a lot of them.
00:13:45I do use the subagents extension.
00:13:48I do use the web access extension so that it can do web search and is a bit more effective there.
00:13:54And there also is an extension that limits what it can do.
00:14:04What is it?
00:14:06I quickly have to check because, yes, I have some guardrails in place.
00:14:12It's the, oh, yeah, it's the guardrails extension.
00:14:16So let me see guardrails.
00:14:21Hello.
00:14:22Hello.
00:14:23This one.
00:14:25Yeah, I'm using this as well.
00:14:28This just provides some basic guardrails like RMRF is blocked.
00:14:35And these AI models can always find ways around these guardrails.
00:14:39But the good thing is they typically try it with the obvious path at first.
00:14:44And then that extension steps in, blocks it.
00:14:47And then I'm aware of what the model is trying to do.
00:14:49And I can either allow it or I can block it.
00:14:51And then I know, OK, I should be careful.
00:14:53The model is probably trying to work around it now.
00:14:56But, yeah, that is what I do.
00:14:58Beyond that, it's basically no extra extensions and no extra restrictions.
00:15:04What about SEO?
00:15:05How do you manage that, especially in this AI era?
00:15:08Right now, I do it just as before because my understanding is that for the most part,
00:15:14what used to be found with traditional Google search is also picked up by AI.
00:15:21That being said, I can definitely see that traffic through search has decreased dramatically because people read those AI summaries or just read what ChatGPT tells them and rarely click on these links.
00:15:36So I guess the way forward is more to have content, if you can have that content for your business, that's less about general knowledge because that will just be picked up and incorporated in those summaries.
00:15:50But instead, that's about a specific product or thing, whatever that is, where people really want the original description or want to, how do I phrase it?
00:16:09Where they don't just want to find the knowledge, but instead the actual product or something that's maybe a bit incomplete when included in a summary.
00:16:21And that sounds very cryptic because that's another area where I'm still trying to figure out what the best way forward is there.
00:16:29But for me, it also is traditional search engine optimization and organic search as a traffic source has declined.
00:16:37I haven't found out how to get back to the same levels yet, and for me, it's there for YouTube mostly and social media in general.
00:16:45But for me, that is mostly YouTube where I still reach people and which is a super important traffic source for me.
00:16:53So, yeah, again, another kind of annoying side effect of AI there.
00:17:05Do I plan any updates on the Codex and Cloud Code courses on Udemy?
00:17:09Not right now.
00:17:11Obviously, may always change if there are significant changes, but not right now.
00:17:17Thank you again.
00:17:18Thank you all so much for the very nice comments.
00:17:20Super, super amazing.
00:17:21Means a lot to me.
00:17:25What do we have?
00:17:27How do you deal with burnouts?
00:17:29Luckily, I haven't burned out yet, and I hope it'll stay that way.
00:17:35My big advantage, obviously, is I'm self-employed.
00:17:38I'm doing my own thing here.
00:17:40And I'll also be honest, there have been more fun times than right now because, well, obviously, I shared that in YouTube videos, too.
00:17:49I enjoy coding and the flow state you enter when coding, right?
00:17:54And that has kind of been taken away with AI from me because I'm in the weird space where writing code entirely by hand feels a bit off and inefficient.
00:18:04Outsourcing it all to AI is not good enough.
00:18:07And even if it were, that's not why I got into coding in the first place.
00:18:13Now, I also like building stuff.
00:18:16But, yeah, I also liked the process of coding.
00:18:19So, yeah, there have been more fun times.
00:18:21But for me, not burning out is all about trying out different things, trying out stuff that genuinely excites me.
00:18:29And for me, that still is teaching, creating content that is also building stuff.
00:18:34And I want to share more about what I'm building.
00:18:37And I have a couple of thoughts there, which I hopefully can share soon.
00:18:41And one other thing I'm doing right now is I'm also getting into game development as a side project.
00:18:46So, not as my main gig, but as a side project.
00:18:49So, yeah, basically trying out new stuff and finding enough stuff that excites me.
00:18:54That's my recipe for not burning out.
00:18:56And obviously, that is a bit privileged because I can generally do what I want.
00:19:02Still, at the end of the day, I have or I want to do something that also pays the bills and is helpful to other people, too.
00:19:17How do you strike the right balance between using AI versus keeping up with the solid basics?
00:19:22Are there some tricks and routines?
00:19:24Yeah, that is tricky, definitely.
00:19:25So, what I really do is I got into the habit of, yeah, using AI to also generate some code.
00:19:35But I do look at the code.
00:19:37There are enough people that tell you that you don't have to, but I'm not one of them.
00:19:41I still look at the code and I question the code and I ask AI.
00:19:45And especially if I used AI also to build something which I haven't built that often before or maybe not at all, I really use AI to ask it, hey, why did we build it like this?
00:20:01What's that doing?
00:20:02Why did we not use this approach, which would come to my mind?
00:20:06And that is how I try to stay sharp.
00:20:09But I will also say the skill set is changing for sure.
00:20:14And that's also kind of related to what I said earlier.
00:20:17I think it'll be more about knowing about a lot of things and then diving deeper here and there.
00:20:24And definitely my skills have degraded if it comes to writing pure code.
00:20:32If you would take away AI.
00:20:34I could still build web applications, but it would take me a bit longer at first because I used to do it all the time every day, three or four years ago.
00:20:46I wrote all the code, obviously, myself.
00:20:50And now it's not like that anymore.
00:20:52So, yeah.
00:20:53But again, asking, being critical and really making sure you understand it, that's what I'm doing.
00:21:01Is there any possibility for complete Azure course?
00:21:04Because I'm not an Azure expert.
00:21:05I have a lot of expertise in AWS.
00:21:08I got all the AWS certifications.
00:21:11Well, by now, a couple of them were, I guess most of them are expired, but I got them back then.
00:21:16I still use AWS a lot, but I got no expertise in Azure.
00:21:20So, no course planned there.
00:21:25And yeah, thank you, Techscope, for the very nice message.
00:21:30Super happy to hear that my Flutter and Dart course was helpful.
00:21:34It's also difficult for me to know what my role would look like in the future.
00:21:38Writing the code, building, testing, learning from it was so much fun every time.
00:21:42But using AI and only watching, reading, and reviewing the code, I'm more likely bored.
00:21:46And I have also the feeling to lose a lot of knowledge.
00:21:48Yeah.
00:21:49Yeah, exactly.
00:21:49And I think one problem also with AI for me is before AI, I could enter that flow state, right?
00:21:57So, you were working on a project, and you were just coding, running into problems, thinking about them, looking for solutions, going to Stack Overflow, trying out the solution.
00:22:07So, that was a nice process, and you were non-stop challenged and doing something.
00:22:11Now, with AI, you spend some time crafting the prompt, thinking about how you want to build something, maybe writing a spec, and so on.
00:22:21You do a lot of work there, let's say.
00:22:23Then you send the prompt to the AI, and then you wait.
00:22:26So, what do you do in the time where you wait?
00:22:28You go to YouTube, watch a video, go to X, scroll social media.
00:22:34It's all not great.
00:22:35So, you totally lose the context, you lose the work mode, and instead, you kind of start slacking.
00:22:42Then, the AI is done, but you come back to it too late, because you spent way too much time on social media.
00:22:47And that's all super, super annoying, and absolutely not that nice flow state that it was before.
00:22:54So, yeah, that's my take on it.
00:22:56And at the same time, just to be clear, AI can be very useful.
00:23:01And I'm sure it, well, obviously, it makes you more productive.
00:23:07It's a big enabler.
00:23:08But at the same time, it's also so, so easy to feel way more productive than you actually are.
00:23:17Was my YouTube channel bought out to a big company?
00:23:19No, absolutely not.
00:23:21Because you're uploading AI videos all the time, and you said you're tired of AI yourself.
00:23:26No, all my YouTube channels still belong to me, and I'm uploading videos about AI, because it is the big topic right now.
00:23:32And I'm tired.
00:23:35Well, I just said everything there is to say about AI in the last couple of minutes.
00:23:38And in those videos, I try to be pretty, well, obviously, I'm not objective.
00:23:44It's my opinion, but it's, I try to share my honest thoughts, which is, yeah, it's an exciting, useful technology.
00:23:52And, yeah, I would not be sad if it were gone.
00:23:55But how does it change anything if I try to fake or simulate as if AI wouldn't be there?
00:24:04Obviously, it's there.
00:24:05Obviously, it's influencing every single day I have.
00:24:08And, yeah, YouTube is a medium where I share my knowledge, my thoughts.
00:24:12And those are mostly about AI these days, because everything is mostly about AI these days.
00:24:19Kind of, yeah, trying to relive the old days and kind of acting as if AI weren't there.
00:24:31I don't see the point in that, to be honest.
00:24:33So, yeah, but the YouTube channels are 100% mine.
00:24:37Nothing was sold to anyone.
00:24:41I would like to see more agents, CLI tools being faster and optimized.
00:24:45I do not understand why these AI companies do not try to provide CLI tools.
00:24:49Written in low-level languages.
00:24:50Yeah, I mean, the interesting stuff is that obviously AI is capable, but these AI companies
00:24:56often make it sound like software engineering is 100% solved already.
00:25:02And then you obviously wonder, okay, but why are your toolings not better?
00:25:09Yeah.
00:25:12Just want you to know, I bought every of your Udemy courses because it was very, very good.
00:25:17Love you, Max.
00:25:18Love you.
00:25:19Thank you so, so much, my sister.
00:25:22I don't know what to say.
00:25:23That's just amazing.
00:25:24Super happy that all the content and courses were helpful.
00:25:27And I'll try to keep them coming on AI and also non-AI topics.
00:25:32But as I mentioned, pretty much everything is kind of influenced by AI, he says.
00:25:36So, yeah.
00:25:37But yeah, thank you so much.
00:25:40How many hours can you focus without a break?
00:25:42So, before AI, really, I'm not exaggerating, but I could definitely enter that super efficient
00:25:51coding flow state for eight, 10 hours.
00:25:55Obviously, taking a little pee break here and there and then like going for lunch like half
00:26:00an hour.
00:26:01But other than that, yeah, pretty much.
00:26:03I could enter that for a very long time period.
00:26:06And now with AI, that has changed a lot because you have all these artificial breaks where you
00:26:11lose the context.
00:26:12At least I do.
00:26:14And of course, you could try to run like 10 agents in parallel, but that leads nowhere,
00:26:19at least for me.
00:26:25Since you have been coding without AI for a long time, that gives you a perspective on what
00:26:29it is to write code without AI.
00:26:32Do you think the upcoming developers will struggle hard with this?
00:26:36Yeah, I mean, I don't know if it's an advantage if you start your developer career right now
00:26:44or a disadvantage.
00:26:46Obviously, like the past days have been great.
00:26:48Job market has been great.
00:26:49But when it comes to learning, if you're learning to code right now, that, of course, is probably
00:26:56very different from how it was back for me in the old days.
00:27:01And maybe it's an advantage if you become AI native, whereas I kind of have to make the
00:27:08shift and adjust.
00:27:10At the same time, obviously, my existing knowledge helps me a lot.
00:27:13And there is a reason why there is so much demand for senior developers, because having that
00:27:18experience as a senior developer, also about more advanced or general system and software
00:27:24architecture stuff combined with AI, that is, of course, pretty great.
00:27:28And I feel that too.
00:27:29At the same time, getting used to that transition and getting used to how you build software with
00:27:35AI, that is a big change.
00:27:37I mean, for example, one question is, how important is code quality?
00:27:42There are good reasons for having patterns and best practices and why you want to ensure
00:27:49a good code quality.
00:27:50And I think those reasons are still important now in the age of AI.
00:27:55But you could, of course, also make the argument that code quality is less important than in
00:28:01the past, that it's more important that the AI understands the code.
00:28:05I would argue, since the AI models have a limited context window, having like a clean, lean code
00:28:12base that's not unnecessarily bloated and stuff like that still has a lot of value.
00:28:17But again, that's me coming from the old world, and maybe there is some benefit in not having
00:28:25all that old knowledge with you, to some extent, at least.
00:28:29Again, it's an advantage in other areas.
00:28:33Here's a big fan of yours from Brazil.
00:28:35Your work is incredible.
00:28:36Like you, I have an insetial hunger for learning.
00:28:41Good luck to us and understanding this new era.
00:28:44Not so good luck in the World Cup.
00:28:46Thank you so much.
00:28:47And yeah, good luck to us all.
00:28:51And regarding the World Cup, yeah, I'm really sorry about 2014.
00:28:56But I mean, the German national team is far away or has been far away from its performance
00:29:06back then over the last couple of years.
00:29:08So let's see.
00:29:09But yeah, good luck in the World Cup or no good luck to the World Cup to all of us, I guess.
00:29:23Hi Max, I never got a chance to tell you this, but your courses on Vue and React really helped
00:29:27me in my career.
00:29:28Really sorry for pirating them at first, but I bought them later after getting my first job.
00:29:32Yeah, I'm super happy that you bought them because, I mean, to this day, I try to sell
00:29:40it all at very affordable and fair prices.
00:29:43And everybody wants to get paid for their work because, yeah, creating courses is a lot
00:29:48of work.
00:29:48So yeah, thank you so much for buying them.
00:29:50Because, yeah, I'm of course super happy to hear that they were helpful, that they got
00:29:55you your first job.
00:29:56And yeah, reading messages like this and all these nice messages means a lot to me because
00:30:00ultimately that is why I started and why I still do create all that content in courses.
00:30:05I try to help others with the courses, maybe with some opinions.
00:30:12Yeah, so thank you all so, so much.
00:30:14And thank you, Hot Chatty.
00:30:18How to improve code architecture knowledge.
00:30:22I am able to get things implemented, but they lack structure.
00:30:25Yeah, I mean, that is really just a skill.
00:30:27It's a lot about knowing about the different patterns and building blocks and why you would
00:30:33use which, and that's exactly a topic I also plan on covering with a course or courses later
00:30:40this year, this entire software architecture system design thing, because I think that will
00:30:45be super important.
00:30:46Always was important.
00:30:48Now, more important in the future.
00:30:51Because, yeah, ultimately it really is about learning about the different patterns and building
00:30:55blocks and understanding their trade-offs, because that is what allows you to then make good
00:31:00decisions, because you can always build something in like 10 different ways and knowing which
00:31:05way is best for the current situation, that is something that comes with experience, but also
00:31:11something you can learn by learning the fundamentals.
00:31:14And yeah, I'll try to share more on that very, very soon.
00:31:18Good luck for Germany World Cup, though I'm supporting Netherlands.
00:31:21Good luck to you, too.
00:31:23Yeah, let's see how things go in this World Cup.
00:31:27For Germany, it can probably only get better compared to the last one.
00:31:35In my opinion, one of the bigger advantages with AI is that you are less often stuck for
00:31:39days or needing to wait for help from others, less frustration.
00:31:42Yeah, I get that.
00:31:43But I will say, hmm, the getting stuck part could be annoying, but it was also where you
00:31:54learned a lot.
00:31:55Obviously, you learned a lot, which you could argue is not important anymore these days.
00:31:58I don't know.
00:32:00But you learned a lot about why certain things maybe didn't work.
00:32:04So not just little syntax errors or, well, syntax or errors were easy to catch.
00:32:09But I'm not talking about like these annoying bugs, which where you don't get a lot out
00:32:14of them, where you just think, oh, God, I'm stupid.
00:32:16That was just annoying.
00:32:17I agree.
00:32:18But I ran into a lot of issues where ultimately, once I found the solution, I learned why my original
00:32:27approach didn't work, why it was a bad idea or why my thoughts about my architecture were
00:32:34wrong.
00:32:34And I think that was a very valuable skill in a very natural way of learning.
00:32:39And you don't have that anymore.
00:32:40You don't have that anymore.
00:32:42Now you have to learn stuff differently or outsource it all to AI and just blindly trust
00:32:47it.
00:32:47The problem with that, of course, is if you do that, you could, of course, ask what exactly
00:32:56is your role?
00:32:57And you could, of course, all say maybe there is no role for developers because AI can do
00:33:01it all perfectly, and I would argue that I don't see happening so soon.
00:33:06But it's shifting.
00:33:08And yeah, you kind of need to be able to know a bit more or better than AI in certain areas.
00:33:14And now you can't learn that anymore by running into problems yourself, I guess.
00:33:20And just blindly trusting AI, as I said, is probably also not getting you anywhere.
00:33:27It's hard to follow AI news because, meanwhile, it is important to stay in loop with updates,
00:33:32but it's so annoying to constantly listen to AI SEO speeches, how AI will take jobs and
00:33:36so on.
00:33:36Yeah, absolutely.
00:33:37I mean, we had, like, Anthropic is obviously, like, a big player in that area.
00:33:46We had them announce Claude Fable, which is their mythos model with extra guardrails, essentially.
00:33:53And obviously, Dario Amode, and I covered this in various YouTube videos, too, is one of the
00:33:58most prominent figures that tell you that AI is going to take all jobs now.
00:34:05And I agree, that is really annoying.
00:34:08That is super annoying.
00:34:09Also because AI is already super, super scary to, well, to a lot of people and annoying to me, too,
00:34:21to a large extent, not everything about it, but a lot of parts are annoying, as I mentioned,
00:34:27and hopefully explained.
00:34:28But at the same time, it's there, and you need to adapt, and it offers opportunities.
00:34:33But it doesn't help if there is someone who constantly tells you that your job is gone,
00:34:38because you already have that fear.
00:34:39You don't need anybody to tell you all the time.
00:34:42And I don't know, it's probably just a strategy to make your company seem even more powerful
00:34:49and for investors and customers, and with the IPO and so on.
00:34:54But yeah, that doesn't help.
00:34:56That is definitely, definitely very, very annoying.
00:35:00Yeah.
00:35:01And I mean, in general, like, yeah, you have these two, from a coding perspective,
00:35:08you have like Anthropic and OpenAI as the two main players.
00:35:12And feeling super, super dependent on them, and them giving you that feeling by changing
00:35:22pricing suddenly, or by what they did with Zable.
00:35:25Hey, you can use it in your subscription for the first two weeks, and then thereafter,
00:35:30you'll have to pay, and then maybe we'll bring it back.
00:35:33So that really doesn't give you a great feeling.
00:35:39Thanks for this.
00:35:40Ask me anything.
00:35:41You're welcome, Flavio.
00:35:41And again, thank you all for being part, and for the nice comments and questions.
00:35:47Thank you so much for all your videos.
00:35:49I also enrolled in a full-stack course in Codecademy.
00:35:52What tips or advice do you recommend?
00:35:54So obviously, I don't know that course.
00:35:56I don't even know the platform very well, just the name.
00:36:00But my advice when it comes to learning stuff is the same as it always was, and that hasn't
00:36:04changed.
00:36:05Build stuff.
00:36:06And nowadays, obviously, build stuff with AI too.
00:36:08But really, make sure you take a look at what AI gave you.
00:36:13And use AI also as a partner to ask, hey, why did you build like this?
00:36:17Why could we also build it like this?
00:36:19Or what would have been alternatives?
00:36:21And why didn't you use those alternatives?
00:36:23That is a great way of taking your knowledge, applying it, and then diving deeper, I think.
00:36:28So that's what I would probably recommend, or not probably, what I do recommend, yeah.
00:36:35It seems powerful to vibe code training models with the sim back to package and having it
00:36:42run on a Pi.
00:36:42But what do you think all this is leading to?
00:36:45I'm not sure I fully get the question.
00:36:47You mean on a Pi Raspberry now?
00:36:49Or you mean, oh, or you mean vibe coding packages, extensions for Pi to extend itself?
00:36:58Yeah, you mean that, I guess.
00:37:00Yeah, so I think that is a great feature of Pi, that you can ask it to write its own extensions.
00:37:08Because if you have a specific use case where you know that the agent would be a bit more
00:37:16efficient at doing it, if it had some extra tools, let's say, or anything like that, you
00:37:21can just ask Pi to do some research and write itself the tools it needs for a certain task,
00:37:26and then package that into an extension, and maybe add a skill so that it uses the features
00:37:33provided by that extension when it needs to them.
00:37:36And I think that's very powerful.
00:37:38Now, obviously, since it's very likely that you're vibe coding those extensions, and I
00:37:43definitely did that too in the past.
00:37:47In general, with those coding agents, you want to A, be careful.
00:37:51They can roam pretty freely on your system, and they can extract data, not even in a malicious
00:37:58way necessarily, but they read a .env file that goes to the OpenAI servers, boom, that
00:38:04data is kind of extracted.
00:38:06So you want to be super, super careful and not forget that you have a tool on your system,
00:38:12which may have full access to everything, depending on how you run it.
00:38:16Pi typically has full access.
00:38:18And even if we ignore the aspect of prompt injections or malware, which is a real issue, we're seeing
00:38:26all these supply chain attacks.
00:38:28But even if we ignore that, that is a super powerful tool driven by an AI roaming freely
00:38:32on your system.
00:38:33So definitely something to be aware of and to always keep in mind and maybe not let the
00:38:41AI do everything you're too lazy to do on your own.
00:38:45That's at least what I try.
00:38:49Hi, Max.
00:38:50Thanks for this.
00:38:51Ask me anything.
00:38:52Wondering how you think about SEO and also in general.
00:38:54How do you think of making your apps reach as many people?
00:38:57Do you have any particular new methods that you use now?
00:39:00So as I said before, one problem I have, and I guess many people have, is that organic search
00:39:06traffic is down.
00:39:07And the good thing for me is that my primary way of reaching people has pretty much always
00:39:14been for one, obviously Udemy, the marketing they do and their audience.
00:39:19And then of course my YouTube channels and that's a good fit because I really like doing YouTube.
00:39:25I like sharing videos there and creating videos.
00:39:30So that has become even more important for me.
00:39:33And I guess in theory, social media in general would be more important.
00:39:36I just absolutely am not a fan of TikTok or Instagram and stuff like that.
00:39:40So that's why I'm super bad at that and I don't really do it.
00:39:44But yeah, I think that will become more important because I don't have a good solution for search
00:39:50engine optimization and for organic search traffic.
00:39:54I used to publish tutorial articles, blog articles, and those got found and people came to my website
00:40:01and maybe they bought a course and now they get the answers right in ChatGPT.
00:40:07So that I have unfortunately no solution for that yet.
00:40:12I think video, YouTube in my case, is a good solution to some extent because it has that personal touch.
00:40:19And I think that is becoming even more important in the days of AI.
00:40:23And then of course something like a newsletter or anything that allows you to get in touch with people
00:40:28and it has like a personal note maybe, depending on how your newsletter looks like.
00:40:32That's my strategy at least.
00:40:34But I'm still also trying to figure that out, to be honest.
00:40:39I'm brainstorming a Node.js Dino style mobile app.
00:40:43No, mobile runtime for developers and AI agents.
00:40:46What must-have features would make the community adopted?
00:40:50That's a pretty big question.
00:40:51And I'm probably the wrong person to answer it because I have never even tried to build a mobile runtime,
00:41:01a mobile JavaScript runtime.
00:41:04And as I understand it, that is what you're trying to build.
00:41:09So, yeah, I think it'll be hard for sure because, I mean, React Native has Hermes.
00:41:21And as I understand it, that is relatively new, I know.
00:41:25And as I understand it, it was kind of purpose-built.
00:41:31So, I, and since React Native is kind of the main JavaScript consumer when it comes to mobile development
00:41:40or the main platform, I don't know how another runtime would be adopted there.
00:41:49But maybe I also totally misunderstand the question.
00:41:53It's much more difficult before, I would say, because answers have been provided already now.
00:41:57Before even you search with that, what you struggle, you have to read the code first and understand.
00:42:01Yeah, I agree.
00:42:03Code quality will be a thing that we as developers will care only for our personal projects.
00:42:07I see that companies care less and less for that.
00:42:09And it's not only because of AI.
00:42:10Yeah, that, that may very well be the case.
00:42:13And I mean, if we're honest, it's not like code quality was super great before AI.
00:42:18And especially in larger projects, you would ultimately gather and have a lot of tech debt and code quality that was maybe the best.
00:42:30But still, you could always try for exactly that reason to avoid getting more and more tech debt.
00:42:36But yeah, I definitely see also because of AI and yeah, probably also without AI, but definitely with AI, that being less and less important for companies.
00:42:49And again, I don't even know if that's the wrong way of looking at it.
00:42:53It feels wrong to me.
00:42:55And I would have arguments for why it still matters, but I may be wrong.
00:43:00I may absolutely be wrong about that.
00:43:03From your experience, what's the best AI tool for a specialized skill for UI UX design right now, especially for creating modern and creative website designs?
00:43:11I don't know, because to be very honest, I don't use AI too much for web design, even though I pretty much suck at it.
00:43:20But still, I don't want to have too generic AI looks.
00:43:26By the way, as a fun fact, we all know that AI likes these gradients and AI likes purple, right?
00:43:32Everything has to be purple.
00:43:34And my academy inside, we always had purple as our main color since 2017.
00:43:41So now, all our site and our content looks super AI generated.
00:43:46And obviously, we're using AI as part of our development too.
00:43:49And in some parts of the design are AI generated, but it looks super AI generated just because we're using the same colors we used before.
00:43:56But now, it's the favorite color of AI.
00:43:59But yeah, I typically use AI without any special skills, just for some first draft.
00:44:06But I already give it some instructions about the general structure I want, color codes or colors I want, and so on.
00:44:15And then I really fine-tune it myself.
00:44:17Maybe even by diving into the CSS code or the Tailwind code or with AI, but it's a super iterative process for me.
00:44:24It's not like, here's the skill, here's the task, now go.
00:44:28And then I spin the slot machine like four times and take some design.
00:44:33That's not my approach.
00:44:34I'm sure there are great skills, but I'm not using one for design.
00:44:38What's your current favorite code editor right now?
00:44:41I'm switching a lot, but for me, what I typically use a lot is VS Code, mostly for code reviewing and div reviewing,
00:44:49and then simply the integrated terminal with Pi or the Codex CLI or the Codex Desktop app.
00:44:55I like that a lot too.
00:44:57Yeah, these are my editors right now.
00:44:58I tried SAD, in case you don't know, SAD is a code editor written in Rust.
00:45:06And I like it a lot, but it always had some minor issues where, yeah, I just always switched away again.
00:45:14Either because for some time it would crash relatively often.
00:45:21I don't think it does that anymore.
00:45:23But then some feature is missing, which I wanted to use, or the integrated terminal doesn't support Shift-Enter, something.
00:45:33There was always something.
00:45:34So, yeah, for me, it's still VS Code or Codex app.
00:45:40Oh, yeah.
00:45:41Thanks for teaching me Angular like 10 years ago.
00:45:44Yeah.
00:45:44Oh, those old times.
00:45:47The Angular course, the first version of it, was released over 10 years ago.
00:45:53That's so incredible to think about.
00:45:56Those were fun times.
00:45:59Can we really make any original projects just by using AI, the real production projects?
00:46:05I asked Gemini to make authentication flow in Django, and I can't use that in prod, and the UI was very basic.
00:46:11So I will say Gemini, in my experience, is not that great at coding and also not at design, though I haven't used the latest version just from what I read and from what my experience has been with older versions compared with other older versions by other providers.
00:46:26So my experience has definitely been better with ChatGPT or with the GPT 5.5 model and the Opus models by Anthropic.
00:46:33So that may help, but yeah, that's exactly my point.
00:46:37I think if you know what you want, if you can steer the AI, if you can tell it, hey, we're going to do authentication this way or using this approach or using this package.
00:46:48And if you can give it those hints and instructions and then review the code and also correct it and not just give it very broad general prompts and then follow whichever decision the AI makes.
00:47:02If you're able to make your own decisions and then ask the AI to implement them and to do research, that is the far more powerful pattern right now.
00:47:11And I would guess also in the future, obviously, I can't look into the future, though, but yeah, I think you can build production ready projects with AI.
00:47:20But at least I can only do that with confidence if I also steer it and if I make the decisions and that's kind of what I'm trying to do and where I try to find the best way of using AI to be productive.
00:47:34Any thoughts on porting BUN to Rust?
00:47:37Yeah, that's something I also wanted to talk about.
00:47:40Now, I mentioned before, I got a BUN course coming up, hopefully in a few weeks, because I like BUN a lot.
00:47:49I use BUN for pretty much all my projects as a runtime, also because it has so many amazing features built in.
00:47:55And nowadays with those supply chain attacks, it looks pretty interesting to have a runtime that has certain features built in so that you don't need to reach for a third party package.
00:48:04But yeah, anyways, I like Rust and I found that rewrite in, I like BUN.
00:48:10I also like Rust, though.
00:48:12I found that rewrite in Rust pretty interesting for one, because, yeah, I like Rust, but I'm no expert at it at all.
00:48:20I used Rust for, I think it was over a year ago that I wrote my last Rust code, but I like the language, the aesthetics of the language, the way it works and so on.
00:48:32Now, the reasons for that rewrite, as I understand it, were pretty broad.
00:48:41There were different reasons.
00:48:42One of them was that the resulting runtime is equally fast or faster, and it's smaller, and they also have an easier time catching and preventing memory bugs because Rust, it's all about memory safety, right?
00:49:06So I get that these are good reasons to use Rust.
00:49:09Now, I think the original or the reasons why they chose SIG originally were that it was easier to build something like BUN with SIG, which to some degree was a bit less restrictive than Rust back at the time.
00:49:30And as I understand it, I'm no SIG developer at all, it's just from what I think I read, why BUN originally was originally was written in SIG.
00:49:38So it makes sense to me that if you have a working runtime in this case that is written in a language, porting it to another language can probably, as it seems here, be done with AI pretty well.
00:49:51You won't have a beautiful, super idiomatic Rust code, as it seems from what I read, but you have working code, you can improve it over time.
00:50:02And I definitely see that this is an area where AI can be helpful, and definitely it looks like it just works.
00:50:10We'll see over the next months if there are significant issues with that rewritten version, since it's not the default stable version yet.
00:50:19But I found that interesting.
00:50:20I think another reason for the port, of course, was that the BUN team, they had a fork of SIG, which they maintained, because they needed certain features or behaviors of SIG that just weren't in SIG.
00:50:35And they weren't able to merge their changes upstream, because SIG has that no AI policy.
00:50:43And the BUN team, obviously, also, since they are part of Anthropic, they are using AI.
00:50:49So their fork of SIG, on which they depended, of course, kind of drifted away from SIG farther and farther, right?
00:50:58Because they kept on adding features to their fork, to their copy of SIG, but they couldn't bring it to the original SIG because they didn't accept contributions that were generated with help of AI.
00:51:11And I think that's also a good reason to kind of switch to a language where things maybe are different, or which is a bit more open to those contributions, or where maybe you don't feel like you need to make as many contributions as they had to make to SIG.
00:51:29Because, obviously, if you drift away farther and farther, at some point, it's like two different languages, right?
00:51:35And I get that reasoning.
00:51:36But I'm excited to see if it really is stable and working, or if we do find serious problems over the next months.
00:51:45Do you think editors such as WebStorm still matter, or is it not necessary anymore?
00:51:49I honestly can't tell, because the last time I used WebStorm was, I don't know, five years ago.
00:51:58And even back then, I didn't fully see the value.
00:52:02I understand that in the early days, when I created that first version of my Angular course I mentioned, I did that with WebStorm, I think.
00:52:11Because back then, it was way more capable.
00:52:17VS Code was pretty new back then.
00:52:19And especially, for example, Angular support was something, Angular 2 support was something that landed in WebStorm pretty early.
00:52:29But nowadays, for me, yeah, I don't necessarily see the appeal.
00:52:37If something like WebStorm had a super amazing AI assistant, some integration that just isn't there in other editors, that may be a reason.
00:52:50But I don't know if they have that.
00:52:52Yeah.
00:52:55I've started doing your React Native course.
00:52:58Do you think it's up to date nowadays?
00:53:00Oh, maybe there were some big changes recently.
00:53:02It's definitely been some time since I last updated it.
00:53:05But the core way of building React Native apps, that is still the same.
00:53:11So that will all work and is all in line with how you build apps these days.
00:53:18So, yeah, you can absolutely still take that course and learn how to build React Native apps.
00:53:24I think learning manual coding is still useful to using AI.
00:53:28AI was trained on human code, but often produces an approximation of that code that's a patchwork cobbled together just to pass the does-it-work test.
00:53:37You still need to code to know good code.
00:53:39Yeah.
00:53:40And that's what I meant before.
00:53:42I think so, too.
00:53:44I think my knowledge helps me a lot with using AI.
00:53:49But I think at the same time, learning to code these days looks different than in the past because you probably just won't sit down and type it out all by hand whilst you see people left and right using AI.
00:54:05It may be the right thing to do.
00:54:06I don't know.
00:54:07But I don't think that is how you learn it these days.
00:54:10But then again, that is exactly what I'm trying to figure out.
00:54:13How are you doing it these days, ideally?
00:54:20What is your advice to keep your brain sharp and not fall into the machine slot lever pulling?
00:54:25Yeah, good question.
00:54:26For me, it really is not blindly accepting the stuff the AI gives you or just prompting again if there is some error, but instead diving into the code, analyzing it, thinking about it critically, discussing it with AI, fixing parts you don't like or bugs on your own here and there and not ask AI to make the button red.
00:54:50Just do it yourself.
00:54:51It's like a change in one line.
00:54:55That is what I think helps you keep your brain sharp and especially being super critical about the suggestions that the code AI gives you because it often sounds super valid, correct.
00:55:10And it often may be, but it's not always correct.
00:55:13And it's not always the best solution.
00:55:14I mean, correct is a difficult word, but depending on what you're building, approach A may be correct.
00:55:20If you're building it for 1 million users, but not for your zero users, right?
00:55:28So that is how I try to stay sharp.
00:55:32Question, be critical, discuss.
00:55:34Yeah.
00:55:43What are your thoughts concerning the cost of using AI in development?
00:55:46Most people think AI would replace developers, but it seems to be more expensive compared to hiring devs.
00:55:51Yeah.
00:55:51I mean, that will be the super interesting question.
00:55:55I mean, the Fable, for example, is super expensive.
00:55:59And yeah, in general, right now we're still in the world where all these subscriptions are subsidized.
00:56:06And at some point that will end and AI is expensive.
00:56:09And I guess whilst token cost probably will come down long term, I don't think it will necessarily come down right now or in a year or so.
00:56:21There's far too much progress still, too much training of more and more capable models going on.
00:56:28So, yeah, token cost will be a problem.
00:56:31I just have one problem.
00:56:32If the main reason for hiring developers is that they are cheaper than AI, that's really difficult.
00:56:40That is really difficult.
00:56:42I hope we can find other reasons, like being good in evaluating the code, designing the architecture and so on, making decisions.
00:56:50Because if it's just the cost, that's not so great.
00:56:57Because for one, the cost will probably come down.
00:57:00So that makes it less and less attractive to a few minutes if cost is the only factor.
00:57:06And B, that is a pretty fierce competition then.
00:57:10And therefore, I hope it will not just be cost.
00:57:15In the near term, though, I think it will be an additional incentive.
00:57:21Not the only one, but an additional incentive for companies to have real humans do the work combined with AI.
00:57:29Because just AI alone is probably too expensive.
00:57:34The good thing, by the way, is that the cost of tokens and the new realization of companies that it's pretty expensive led to an end of this token maxing BS.
00:57:46Because obviously, it's super easy to token max.
00:57:49It's super easy to burn a lot of tokens.
00:57:50And it's nice that companies realize that this is just too expensive.
00:57:54Because, yeah, the number of tokens you used is just a super stupid measure of how productive someone is.
00:58:05Just as the output in lines of code is not a good measure.
00:58:14This community is in their pre-alpha stages.
00:58:16It might start off as coding, but it's definitely not going to end as coding.
00:58:19And by then, it probably won't even be called AI any longer.
00:58:24Which community?
00:58:28I'm writing firmware for the automotive.
00:58:30In my company, we use AI mainly test validation for firmware AI struggles most of the time.
00:58:35Yeah, I mean, I guess in general, the more niche your work is or the software is you're working on,
00:58:44the harder it probably is for AI because it may know the language,
00:58:49but it may not know the exact patterns or approaches, best practices, and so on,
00:58:57that matter for the kind of software you're building.
00:59:03Thank you for doing these live streams.
00:59:05Thank you for being part.
00:59:06Thank you all for being part.
00:59:08I'm having a blast, and I hope you're liking it too.
00:59:11I use AI to teach me new languages.
00:59:13Yeah, I tried that too.
00:59:15And it worked, I guess.
00:59:18Yeah.
00:59:19I'm just a bit old school.
00:59:22I also like, I mean, it's also, it depends on how much time you have,
00:59:28but I guess if I were to seriously learn a new language,
00:59:32I would probably do it with a human coach,
00:59:35but I also learned a bit of Spanish with AI,
00:59:39and that worked pretty well, I will say.
00:59:40Do you think there is a future on 3D web dev like meta?
00:59:46I mean, 3D web development has like its moment right now with 3.js and AI's love for 3.js,
00:59:54and I don't know what the future of that is exactly.
00:59:59I, for example, don't necessarily see video games in the browser having a big future necessarily.
01:00:08But I think the fact that AI makes it a bit easier to build complex websites with like spinning globes or stuff like that is definitely good,
01:00:21because I think that's one area where AI makes it more approachable because not a lot of developers, myself included,
01:00:29got into 3.js or other related more complex libraries in the past because it was so niche.
01:00:36And therefore, that was kind of like a self-fulfilling prophecy since not a lot of developers got into that.
01:00:42It wasn't used on a lot of websites and so on.
01:00:44And yeah, therefore, I think that is definitely something where AI helps and maybe makes it a bit more popular
01:00:50and definitely more approachable.
01:00:58Say you won the lottery and never had to touch a computer for work ever again, what would you do?
01:01:04So I would want to touch a computer because I seriously, honestly, like working with computers, programming, building stuff there.
01:01:16So I would probably still continue doing that, to be honest.
01:01:21I would basically try to do what I'm doing right now, too.
01:01:26I want to figure out how to best work as a developer with AI,
01:01:33how to build useful stuff, what useful stuff is in the future,
01:01:37where theoretically everybody can build everything.
01:01:41Not entirely true, to be honest, but that's kind of the feeling you can get these days.
01:01:45And how to share the knowledge.
01:01:46So I would probably try to continue doing that.
01:01:49As boring as that answer is, sorry.
01:01:51But yeah, I really just like building stuff with computers.
01:01:55I always have, starting as a kid, and I still do.
01:02:00I may be doing that until I die.
01:02:04I don't know.
01:02:07And yeah, sure, early 2K, you can ask a question.
01:02:10Just go ahead.
01:02:11I'll try to go through all the questions I see in the chat step by step.
01:02:15Learning something new, and that's not a question.
01:02:22Max, your teachings have always served as a yardstick for what I need to know to be a competent developer.
01:02:27I hope that you will continue to shine a light in terms of what we need to know,
01:02:30and also to provide the associated instruction.
01:02:32Thanks, as always, for everything so far.
01:02:34Wow, thank you so much, Peter E70.
01:02:38I don't know what to say.
01:02:39That's just amazing.
01:02:41Thank you so, so much.
01:02:42And yeah, I will definitely try my best to keep on sharing useful knowledge.
01:02:46A lot of that will be obviously related to AI, because as I mentioned earlier, it is everywhere.
01:02:52But as I also mentioned, I got a fun course coming up.
01:02:55I plan on doing a super amazing system design course.
01:03:00And I just try to teach useful stuff, stuff I use, I learn about, not the stuff that's necessarily most hyped up right now,
01:03:12but also not the stuff nobody's interested in, because obviously I want to build stuff that reaches people.
01:03:19But yeah, I'll definitely try to keep it up.
01:03:21Thank you so, so much.
01:03:25If AI can use programming languages to create apps, it could also create its own programming language to create apps even more efficient, right?
01:03:31Maybe a programming language that no human can understand anymore.
01:03:36I guess in theory, it could, yeah.
01:03:40I don't know how well AI is or how well it does at building programming languages,
01:03:47because I don't know how much it saw of that in its training data and how much it's able to infer.
01:03:54But in theory, it definitely could, yeah.
01:03:58I've relied heavily on AI to meet work deadlines, which is causing me to lose my ability to think.
01:04:03I fear this will make it difficult for me to clear interviews for future roles.
01:04:07Yeah, I mean, that entire skill degradation and losing the ability to think,
01:04:13I guess we all feel that to some degree, the more we use AI.
01:04:17And then that's simply what we'll have to figure out over the next years.
01:04:22I guess, as with every big transition, it's challenging.
01:04:26We make a lot of mistakes and we have to kind of figure out how to best use that new technology in the end.
01:04:32Thanks for your Angular course.
01:04:33New to me is really great.
01:04:34Thank you so much, Chris Crispy.
01:04:37I'm super happy you're liking it.
01:04:41Do you have any thoughts on the problem of AI inbreeding, model collapse?
01:04:49I don't think I have enough in-depth knowledge about that aspect that I could share anything useful there.
01:05:02And I had various theories or thoughts about which problems AI companies may be facing because they already used all the data available if they were using synthetic data for future training.
01:05:20And I just don't, hmm, I don't know how to phrase it.
01:05:27I think it's too early to tell what potential problems may be.
01:05:39And I'm just no expert.
01:05:40I'm nowhere near, yeah, I don't have enough insights into which models or which approaches they may already be using internally the public doesn't even know about yet.
01:05:56So for me, it would really be a lot of guessing.
01:05:59And so, yeah, I don't have any educated thoughts there, I guess.
01:06:06I feel like the bar has increased with AI, not just for juniors, but seniors as well.
01:06:10If things continue like this how they are, then in the future, I believe the competition between seniors will be huge.
01:06:19Yes, I mean, not just AI has increased the bar, the job market has in general, and the fact that there was all that overhiring during the pandemic, and now all these companies realize that they don't need as many people, that also raises the bar, because fewer people are needed, and that naturally raises the bar.
01:06:42And AI may just be another factor there for sure.
01:06:47I don't know how the competition will look like in the future, because that's another area where I could just guess and where you could make arguments either way.
01:06:56If you go down the route that AI will replace developers in the near future, then, yeah, of course, competition will get fierce and fierce until there is no competition anymore, essentially.
01:07:09But that is just one scenario, and not the most likely one, in my opinion.
01:07:15Not in the near future, not in the near future, and also not in the midterm future.
01:07:19But the role of a software developer is changing, for sure.
01:07:24And being able to adapt, that is something that's definitely hard, and not everybody will be able to do that, I guess.
01:07:33So, that also will lead to fierce competition between – also, by the way, because we don't know yet what adapting means exactly, and how our job will look like in two years.
01:07:45Nobody knows that.
01:07:47So, that's what I meant at the very beginning of the stream.
01:07:49There have definitely been more fun times as a developer than right now, despite it all also being very interesting and exciting.
01:07:58But, yeah, fear of job loss and so on is real, and that has definitely been better.
01:08:16Back when Angular had a future, 10 years ago.
01:08:24Just skipping over some – I'm also reading all the opinions here, by the way, just to let you know.
01:08:29I'm just skipping over them to the questions where I can answer something again.
01:08:34When they put a bun to Rust, they let Claude use the unsafe Rust feature to disable the type checking.
01:08:39Kind of defeats the whole point of using Rust.
01:08:41Yeah, that's what I read too, and then I read that in many places where unsafe was used, you essentially had to use it.
01:08:48But also, what I mentioned earlier, I think the idea with the port was to really, like, have a kind of like a one-to-one translation of SIG code to Rust code and not deliberately not write idiomatic Rust, which would definitely lead to fewer unsafe usage.
01:09:07And I 100% see that, and I think I also read it somewhere, and I hope the official blog post will be out soon, that they plan on kind of – now that it was ported, they plan on improving the code.
01:09:20And we'll see how well that works with AI.
01:09:23But yeah, I think that that is something where you're 100% right.
01:09:28If you, like, break out of the Rust rules in most of the parts of the code base, there is not a lot of reason to use it in the first place.
01:09:35So the question will be, will they be able to get to more idiomatic Rust code?
01:09:41I think it's their goal, if I understood it correctly, but we'll see.
01:09:49So, um, am I considering doing a course on AstroJS?
01:10:00Um, no.
01:10:03Hmm, for one, because it's probably a bit too niche.
01:10:08Um, and the way I use it, um, for example, I used it for my – whoops.
01:10:17I did not mean to do that.
01:10:21Uh, for example, I used it on my personal website.
01:10:25I'm not using it there anymore.
01:10:26It's now an even simpler website, just powered by Bun and Markdown.
01:10:31But, um, the way I used it there was so trivial that I felt like teaching that in a course is kind of pointless.
01:10:39So, uh, yeah, no plans for an AstroJS course right now.
01:10:43Uh, hi, Max, I would like to thank you for the stream.
01:10:46I know there are not that many viewers, but for the people watching it, it makes a difference.
01:10:49Thank you all for the hard work.
01:10:51Thank you, and thanks, everybody, every one of you that's, um, that's watching.
01:10:57And, um, yeah, I'm honestly not doing that for, um, because I want or I need thousands of viewers.
01:11:08Um, I'm not – yeah, I'm not doing the kind of hyped-up streams.
01:11:13I – I see it more as, like, uh, we – we can – we get together, can discuss stuff and talk about stuff, and I enjoyed that a lot.
01:11:21So, yeah, thanks for – for being part of the stream, and I hope it's helpful and a bit of fun.
01:11:28Um, if AI drastically changes coding and humans are no longer writing, what incentive will there be to create new frameworks, languages, et cetera?
01:11:36Yeah, that's a good question.
01:11:37I mean, uh, I wonder about that, too, if we'll see new frameworks and languages and how new frameworks will do.
01:11:45I mean, there is, um, Remix 3, is it 3?
01:11:49Remix 3, yeah, coming up as a new framework.
01:11:53I haven't used it yet, even though it's in beta.
01:11:56Uh, I'm interested – it will be interesting to see how well that is received, because I would imagine it's super hard to get adoption, because AI is – it's React.
01:12:09Let's be honest, it's mostly React when we talk about web development.
01:12:12Um, and I don't know if there are incentives for new languages and frameworks.
01:12:17You could make the argument that you want to build more agent- and AI-friendly frameworks and languages, and there have been attempts.
01:12:24I think Remix 3 is built to be very AI-friendly, and there is this language – what's its name?
01:12:30Um, Zero Vercel – Zero Lang, a programming language for agents.
01:12:39So, I think there will be attempts to build stuff like that.
01:12:42I haven't used that either, because I'm not an agent, but I also haven't used it with agents.
01:12:47Um, we'll see.
01:12:48Maybe that becomes a thing.
01:12:49But ultimately, yeah, um, I – I really don't know.
01:12:54Um, because I've – I also feel like maybe there isn't as much demand anymore for new frameworks and languages.
01:13:01Because if we're honest, most frameworks and all the languages were kind of created to make solving certain problems easier for humans.
01:13:13And if those problems don't exist anymore, we'll see.
01:13:17Um, which field has a brighter future?
01:13:22Data analytics or software development?
01:13:24I guess both have a future, but for both, it will be different from how it looked in the past.
01:13:31And it will be more about orchestrating and engineering and designing systems and, um, analyzing results and controlling multiple agents.
01:13:43But again, that's exactly kind of the thing we're all trying to figure out how it will look like in two years or so.
01:13:49But I don't know what the brighter future, uh, where the brighter future will be.
01:13:54I think if you properly adjust, both can have a bright future, but the way you did it does not have a bright future, neither for data analytics nor for, nor for software development, I think.
01:14:09Do you think Lynx.js has the potential to become, become a mainstream alternative to React Native?
01:14:16And what would it need to get there?
01:14:19Um, excuse me, um, I don't think it has that perspective.
01:14:27Um, I mean, the one big thing it got going for it is that it's, it's, it's, it's by Tencent, right?
01:14:34Um, I always forget.
01:14:37It's, it's not by some niche company.
01:14:39Um, so that is definitely a benefit, but I think to really replace React Native.
01:14:49You would have to offer like maybe a performance that is out of that world, or maybe what we just said, um, be particularly, uh, loved or, but by agents, by AI agents or good for AI agents.
01:15:04Um, something like that.
01:15:05Um, but other than that, I don't really see why you would switch away from React Native.
01:15:11I'm certainly not switching away from it, uh, for all my mobile, um, side projects, which never get released, but that's a different story.
01:15:21Um, yeah, so no, I'm not too, uh, bullish on, on links.
01:15:28Is million.js yet a thing?
01:15:31Not a big one, I think.
01:15:33Um, the premise Silicon Valley companies have is that everyone will continue to use their services while they destroy the market and internet.
01:15:48What do you think of swapping with European services instead?
01:15:52Um, let's be honest.
01:15:55There are no relevant European services in many areas and you are just too deeply integrated or many people, myself included, are just too deeply integrated or connected to various US companies that it's not trivial or anything like that to pull out.
01:16:16And when we, when we talk about AI, yeah, sorry, uh, there is nothing there in Europe that would be really, really meaningful.
01:16:24And Mistral is, is nice, but not on the same level as those frontier models, unfortunately.
01:16:32Um,
01:16:37I make markdown files and create file by file code and unit test.
01:16:40Is this skill valuable?
01:16:43Uh, you mean you create markdown files and let AI create that?
01:16:47Or I don't know.
01:16:48Um, yeah.
01:16:52I was recently let off and it's directly related to AI.
01:16:55My manager said I was not using AI effectively by checking my usage logs data.
01:16:59I never saw this coming and I'm really in excited, depressed.
01:17:02Well, I'm so, so sorry to hear that.
01:17:03And I mean, what, what a horrible way, because what I, what I said before, uh,
01:17:10judging AI effectiveness by the amount of tokens generated or something like this is just so,
01:17:16so stupid.
01:17:17And I mean, being led off without a word of warning or a chance of improving or anything like that.
01:17:22I'm, I'm really, really sorry, uh, to hear that.
01:17:26Um, yeah, that's.
01:17:31Yeah.
01:17:32Don't know what to say.
01:17:34No, not, uh, not a great company.
01:17:39Um, a lot has been said about AI as a dev tool, but what about as a consumer product?
01:17:45I don't think products that integrate AI models are that good either.
01:17:49Most of them are just over-engineered chatbots.
01:17:51Yeah.
01:17:51That's a very good point.
01:17:53Uh, we're definitely still in that phase.
01:17:55Maybe we're slowly getting out of it.
01:17:57I don't know, but we're still in that phase where we need AI everywhere, right?
01:18:01Just as maybe, maybe some of you remember, um, around, I don't know, 2012, 13,
01:18:10everybody needed an app.
01:18:11Every company needed an app.
01:18:13And you saw advertisements like, Hey,
01:18:17get our new app for like the grocery store next door.
01:18:22Um, and no, it was not for ordering food back then.
01:18:26So, uh, I feel like we're at that point in time for AI that every company needs AI.
01:18:31And for that means a chatbot, right?
01:18:35It means a chatbot.
01:18:36And of course those chatbots then run on top of shitty models because the good models are too expensive.
01:18:42And then you have like your rag solution tacked on.
01:18:45Um, but yeah, you end up with a chatbot.
01:18:48That's not super helpful.
01:18:49You got a couple of customers that just abused the chatbot and use it as a free chat,
01:18:55chat GPT alternative.
01:18:57Um, yeah, I think more and more companies will figure out how to meaningfully use AI.
01:19:02But when it comes to customer facing products, I think
01:19:07we either have to yet find the good use cases or we have to accept that for certain businesses.
01:19:14Yeah, you just don't need AI and a chatbot may not be needed and may not add a lot of value.
01:19:22Um,
01:19:25Hi, I'm a UX designer and a product manager.
01:19:28I would love to learn more about vibe coding with Claude, but I don't want to go with basics
01:19:32of web dev prompts.
01:19:33I want to learn how to read code on a basic level.
01:19:35Can you guide me through it?
01:19:36Yeah, um, basically what I meant before, I'm trying to figure out exactly that.
01:19:42How do I best teach to code in the future where you don't write the code on your own necessarily
01:19:49or all the code.
01:19:50So yeah, I'll try my best to help you and not right now in this second, but I'll try to get there.
01:20:00Do you think there will be a national law to limit AI or not to take other jobs?
01:20:05There may be such attempts, but to be honest, I think also if you look back in history,
01:20:12that just doesn't really work.
01:20:14It may be a short-term remedy, long-term.
01:20:18It will not prevent a change that's inevitable.
01:20:21And it will only lead to a society or an economy falling behind, I guess.
01:20:32So yeah, I think we'll see that.
01:20:34And I think Europe and Germany are great candidates for that.
01:20:37But I think it will actually not help.
01:20:43"Can you finish my Flutter course in three months if you're at the half, if you completed it to 46%?"
01:20:53Yeah, I would guess so.
01:20:55I mean, depends on how much time you have every day, but sounds definitely doable to me.
01:21:00And also, it's not about necessarily checking all the boxes.
01:21:03It's about completing the stuff that sounds helpful to you and then building stuff with it
01:21:08and taking those skills and applying them and understanding which code AI may be generated and so on.
01:21:14That's the more important part, I would say, than getting 100% completion.
01:21:19Unless your company requires that.
01:21:22"What do you think about the Odin language? Will it get more traction?"
01:21:25I find it very interesting.
01:21:26Considered learning it like a year ago or so.
01:21:30Right now, I got too much other stuff going on.
01:21:32I am getting into game development, though, as a side project.
01:21:36And I heard that it should be good for that.
01:21:38But I'll probably not use it.
01:21:40I'll just use Godot and GDScript.
01:21:43But I'm not sure how much traction we'll get.
01:21:46Because just like new frameworks, I think all these non-mainstream languages
01:21:52may have it very, very hard and difficult with AI.
01:21:58Because in the past, if you had a language that offered certain advantages
01:22:04that people could see and that made the job of people easier, you could get traction.
01:22:10But now AI is making the job of people easier, at least definitely feel easier.
01:22:17And you are not as incentivized to look for new frameworks or programming languages, I guess.
01:22:24So I think it'll be hard to get significant amounts of traction.
01:22:31What's the new ways you would suggest integrating AI into software engineering workflows that makes you still learn?
01:22:38I already said before, but I'll briefly repeat it here.
01:22:41I think to learn more whilst using AI, the trick is to really not just blindly accept the solutions,
01:22:47but really question them.
01:22:49Ask AI for alternatives.
01:22:51If you can't think of a better solution, ask AI for alternatives.
01:22:55Ask it, why did you build it like this?
01:22:56Which other solutions could we have used?
01:22:58Do that in a new session with a fresh context and then discuss that with AI.
01:23:04And I think that's a great way of still thinking critically about the code base
01:23:10and not blindly accepting it and honing your skills.
01:23:14Would you agree that nowadays beginners have to learn too many technologies?
01:23:17The phrase, just pick a language and get good at coding, does it still work?
01:23:21I think it's not worse than like six years ago when we had the JavaScript framework wars back then in 2019,
01:23:28because I remember vividly that back then lots of people felt overwhelmed because it felt like
01:23:35there is a new language, a new framework, especially to learn every week.
01:23:39And I don't know if that changed, but I do definitely see that nowadays with AI,
01:23:45everything is changing so rapidly and it feels like you need to learn how to write good prompts,
01:23:50how to do context engineering, how to build loops.
01:23:52That's the new hot shit.
01:23:56And there is so much hype.
01:23:58The thing is, you don't need to do all these things,
01:24:00but it can be hard to filter out what you need to do.
01:24:03And I think what you need to do is just play around with AI, use it and learn for yourself
01:24:09what works for you, what doesn't.
01:24:12But there are so many people screaming at you nonstop what you have to do, also because the
01:24:17internet changed so much and social media changed so much.
01:24:19Like eight years ago, seven years ago, it was such a different place.
01:24:24So I think that's why it feels very overwhelming.
01:24:27You don't need to do everything everybody tells you, but there are more people than ever telling
01:24:31you what you need to do.
01:24:32So that definitely can 100% feel overwhelming.
01:24:37And by the way, because there was this question before, if I sold my YouTube channel because I
01:24:42create so much AI content, that's one of the reasons why I create that content.
01:24:46Because it is the main thing right now and I try to share my opinion, which is mostly not that you have
01:24:55to do everything.
01:24:56I mean, I sell courses on Cloud Code and Codex, but no, you don't need to token max or construct loops.
01:25:03You just need to be able to use these tools, know a bit more about them maybe, and then build stuff
01:25:09and play around.
01:25:09I think that is what you should do.
01:25:12Yeah.
01:25:17Wow.
01:25:17The flutter question came a lot.
01:25:19So if I'm not answering your question right away, you don't need to spam the chat with it.
01:25:23And if I would have seen that you spammed it, I would have not answered, to be honest.
01:25:26Stop that.
01:25:27I'm just going through the questions from top to bottom.
01:25:30So spamming is just pointless and annoying.
01:25:43I'm following the React course and enjoying a lot.
01:25:45I wanted to know whether there are any plans on equalizing the course on the backend database
01:25:49aspect.
01:25:50Some sections use Firebase, others use MongoDB, others SQLite, and others have a dedicated backend,
01:25:54others do not.
01:25:56No, no plans.
01:25:57I show different approaches because there isn't a single write approach and also because the course
01:26:02became what it is over months and years of updates.
01:26:09I will 100% not redo that course just to have like a uniform backend because it isn't about the backend.
01:26:15It's about React.
01:26:16And the backend is therefore pretty much always provided and doesn't matter too much.
01:26:27I create markdown files refined with AI.
01:26:29Like back in college, we write algorithms step by step, then use AI to write code.
01:26:34Yeah.
01:26:35I mean, writing markdown files and then using that for generating code with AI is definitely
01:26:45is definitely an important part of today's work as a software developer, I guess.
01:26:50Markdown files and prompts.
01:26:51And it's how I work too.
01:26:55I also take a look at the code and edit the code here and there when needed, as needed.
01:27:00But yeah, I also work with markdown files and also just normal prompts.
01:27:05And yeah, I think to me, it feels like that will be a part of how we work in the future.
01:27:13And what the full story looks like, I don't know.
01:27:19So yeah, but it is an important skill, I believe, to write good prompts and
01:27:28ask for the right things, which takes me back to that idea of architecting a system
01:27:33or architecting a system or a software.
01:27:35And for example, when we talk about writing markdown files, it's very easy to just ask AI
01:27:42to write you a markdown file.
01:27:44So you're asking AI to write something for it to then execute, right?
01:27:48And I do that too.
01:27:50But you want to read what the AI wrote there and you want to tweak it and think about it,
01:27:55even though that's annoying, right?
01:27:56That's exactly the part that's difficult.
01:27:58With AI, you can get so used to just asking, pull the slot lever, asking, do it again.
01:28:05But yeah, you will get better results if you don't blindly accept everything the AI gives you.
01:28:11But instead, if you dive into the code or if you dive into the markdown file
01:28:15and adjust it so that you're asking the AI for the right stuff.
01:28:19So that is definitely part of the skill you need in the future, I guess.
01:28:25What are your thoughts on the Udemy and Coursera merger?
01:28:29Will it improve the situation for instructors or make it worse?
01:28:34I honestly don't know.
01:28:38I'm not too happy about the merger, to be very honest, because I don't really see the
01:28:43business models of Coursera and Udemy working together that well.
01:28:48And it feels like they also don't have a clear plan on how to merge them.
01:28:53So I would expect that for the foreseeable near future, not a lot changes.
01:29:02But yeah, we'll see.
01:29:05There definitely have been better times for the Udemy platform and also for Udemy instructors.
01:29:12But it is what it is, I guess.
01:29:20What do you think about using AI as your hands telling exactly what to do one prompt at a time
01:29:25versus plan versus spec dev?
01:29:27For me, it's super important to use AI iteratively.
01:29:30So I don't spend six hours writing one perfect spec and then let AI go and then have it loop.
01:29:39I don't work like this.
01:29:40I'm not sure if it works for people.
01:29:42For me, software development has always been about doing it step by step and
01:29:49figuring out the best way whilst you're on it, so to say.
01:29:55And that hasn't changed with AI.
01:29:56So for me, I do work with specs.
01:29:59I do work with documents where I lay out my thoughts or an architecture, a general plan I have.
01:30:07But then I refine that over time with AI and with myself.
01:30:12And yeah, that's how I work with it.
01:30:14I guess that also will be different for different people.
01:30:18But I'm also not sure if it works if you just have that perfect spec and then you let the AI just go.
01:30:29Not sure.
01:30:29May work for some people.
01:30:31I think the BUN port to Rust was mainly one big document with a lot of rules and then the AI went through it.
01:30:42So that may work too.
01:30:44But a port, of course, is also something else than a new project.
01:30:51What's my favorite project architecture when the project gets big?
01:30:54I don't have a favorite project architecture.
01:30:57Or do you mean just the files, the folders or the software, the system architecture?
01:31:02Because that, of course, depends on the project.
01:31:05If we're talking about folders and files.
01:31:08Yeah, I just mono repos, but also adjust the structure when you need to.
01:31:15I'm not the kind of person that tries to plan the perfect structure ahead of time.
01:31:19But instead, if a file gets too big, I split it.
01:31:23If I see that I need to duplicate a certain functionality, I extract it and stuff like that.
01:31:32So I go with the flow as the project increases in complexity.
01:31:37And then to me, typically the decisions come on their own.
01:31:40It's kind of like a natural progression.
01:31:45Yeah, that's been my experience, at least.
01:31:47What are your favorite AI skills?
01:31:48I don't have favorite skills.
01:31:50I'm not Matt Pocock.
01:31:53Yeah.
01:31:54I think AI skills are super useful.
01:31:58But I only have a few.
01:32:00Most of them are related to code and documentation research to kind of steer AI to do thorough research,
01:32:08dive into official documentation and not make assumptions based on the knowledge it has in
01:32:12its training data.
01:32:13These are my most important skills.
01:32:17Yeah.
01:32:19But then again, I only have like a few skills I use in most of my projects.
01:32:26Did you not think to create a course where you will simply give all your courses in one?
01:32:30People will have one general course that will contain everything.
01:32:33Well, the price I would have to charge for the course would be so, so expensive.
01:32:39And the blinds go up, especially if I would include future courses there in that as well.
01:32:46No, that's not something I plan right now.
01:32:50Oh no, I noticed your life after 85 minutes.
01:32:52No worries.
01:32:53I plan on being back next week.
01:32:56So in general, the plan is to stream every Thursday at 5:00 p.m. Central European summer time.
01:33:04I don't make it every Thursday, but next Thursday should work.
01:33:08And the recording of the stream will also stay online on this channel.
01:33:14What's my favorite codex or cloud?
01:33:15I mentioned this earlier.
01:33:17I mostly use Pi these days, the Pi coding agent.
01:33:22But I switch constantly also to stay up to date.
01:33:25Now I'm using Cloud a bit more because I'm using the Fable model, Cloud Code that is.
01:33:31And I also use Codex.
01:33:33I like the Codex desktop app a lot.
01:33:34So I'm switching a lot, but mostly it's Pi in the terminal for most of my projects with the OpenAI subscription.
01:33:45How about vulnerability issues?
01:33:46Is this and previous week normal?
01:33:49Oh, God.
01:33:49Yeah, I mean, that entire situation is horrible with the supply chain attacks and so on.
01:33:55And for me, it has definitely led to a decision where I try to use as little packages as possible,
01:34:04which was always a good idea to some extent, I guess.
01:34:08But yeah, I guess it's the new normal that we see lots of attacks like this, lots of problems like this.
01:34:15And I mean, supply chain attacks are just one issue.
01:34:18We also see more software vulnerabilities because AI-powered detection of such vulnerabilities is,
01:34:27of course, easier than human-powered detection, so to say.
01:34:34And not all vulnerabilities are necessarily super, super critical.
01:34:39But yeah, I think it will be a new normal right now, at least, and for the foreseeable future, that we see more attacks, more vulnerabilities.
01:34:52And that on the one hand, we need to patch quickly because vulnerabilities.
01:34:57And on the other hand, patching always is a bit like an annoying thing because of supply chain attacks.
01:35:03I think you can just try to have various layers of security.
01:35:08I also against supply chain attacks and kind of adjust to this new normal, which it is.
01:35:15I want to learn a compiled language and think among Golang, SIG, or Odin, what should I pick?
01:35:21Would mostly need to work on file handling and reading, writing from DB, which would be useful.
01:35:25I'm neither a SIG nor an Odin developer, so I can only comment on Golang.
01:35:32And I know that not everybody likes Golang.
01:35:34I always liked it.
01:35:35I can recommend it.
01:35:36It has a great performance.
01:35:38I like the ergonomics, the syntax of the language.
01:35:42It has an amazing standard library with so much stuff built in.
01:35:46So I can really recommend Golang.
01:35:48And just as a side note, I do have a course on that.
01:35:51So if you want to learn Golang with a course, I have my Go course.
01:36:01Golang.
01:36:02I don't know why images are not loading right now.
01:36:06Yeah, but I have a Go course.
01:36:08But no matter if you take the course or not, Go is amazing.
01:36:14Can you give me your honest opinion about my cloud code workflow called Own Your Code?
01:36:20It's for fighting cognitive offloading, which can hurt the developer skills.
01:36:23I cannot send links.
01:36:25No, sorry.
01:36:25I don't do any reviews like that because I get so many requests also via email.
01:36:31And I just don't have the time for that.
01:36:34And if I would do it, I would want to do it in a helpful way.
01:36:38And I just can't do that.
01:36:39Sorry.
01:36:49I can't do it in a helpful way.
01:36:53I can't do it in a helpful way.
01:36:54I can't do it in a helpful way.
01:36:55I can't do it in a helpful way.
01:36:56I can't do it in a helpful way.
01:36:57I can't do it in a helpful way.
01:36:58I can't do it in a helpful way.
01:36:59I can't do it in a helpful way.
01:37:00I can't do it in a helpful way.
01:37:01I can't do it in a helpful way.
01:37:02I can't do it in a helpful way.
01:37:03It's kind of my main thing.
01:37:05It's definitely around TypeScript, around React, probably around Next.js, Tailwind.
01:37:14These are kind of the common frameworks and libraries you see in the web development world.
01:37:21Obviously, not the only choices out there.
01:37:23It's just what you see quite a bit, especially with AI.
01:37:27If we look at the enterprise level, though, of course, we see way more C#, .NET, Java in there,
01:37:32right, for the backends.
01:37:34So it's not that easy.
01:37:36But if we're looking at what's hyped up and what's the favorite of AI,
01:37:42then it's what I just said about TypeScript and so on, I would say.
01:37:49I'm a solo learner using feature-driven architecture or domain-driven architecture.
01:37:53Should I use for SaaS, you mean SaaS, software as a service?
01:38:00I wouldn't bother about which of these approaches you use, to be very honest.
01:38:07At least I don't.
01:38:08So maybe you should, but I don't too much.
01:38:15Especially if you're new, I would just get started building stuff.
01:38:21And then over time, you'll see in which problems you run or where you realize that,
01:38:28oh, I should have thought about this earlier.
01:38:30And that will naturally give you an approach which works for you
01:38:34and which you can use in your future project as well.
01:38:36Because especially if you're working solo and so on,
01:38:39you just need to find an approach that works for you.
01:38:42You need an approach with which you can be super efficient.
01:38:45And the only way to find out is by just starting.
01:38:49Just starting, what can go wrong, especially with AI?
01:38:51It's easier than ever to kind of rewrite stuff or reorganize stuff.
01:38:56Yeah.
01:38:56So maybe that's not the fancy answer, but that's what I would recommend.
01:39:03What are my thoughts on the future of Angular?
01:39:05With the recent updates, I think it's on track to become far more similar to React
01:39:09and lose a bit of what made it different.
01:39:11I'd not be surprised if they'd come up with function components in the foreseeable future.
01:39:15Yeah, me too.
01:39:17And I know they've been discussing new ways of authoring components for the last few years.
01:39:23So definitely they got closer to other frameworks.
01:39:29The future of Angular is super difficult because Angular, since the release of Angular 2,
01:39:34has been in the shadow of React.
01:39:38Or at least like one or two years after the release of Angular 2.
01:39:41Because Angular 2 was not very popular amongst many developers.
01:39:46At the same time, many big companies did and are still using Angular.
01:39:52Google is, of course, using Angular internally a lot.
01:39:54So the future, I think, is not too bad.
01:39:57But for many companies and for solo developers and so on,
01:40:06I think React is probably the better choice.
01:40:13Though again, with AI, it may not matter too much.
01:40:19I mean, it matters what you use because you should still be able to understand it.
01:40:23But you can use whatever you like.
01:40:27But still, yeah.
01:40:31Angular, I don't think, has no future where it becomes the number one.
01:40:37It has a future.
01:40:38It has a future where it's getting used a lot by various companies.
01:40:43But the winner is React right now.
01:40:48And I'm not sure if that will change in the near future.
01:40:53And if it would change, if it would change to Angular.
01:41:00What do you think of the AI companion app industry?
01:41:02Lots of dev jobs in this field.
01:41:05Yeah.
01:41:06I mean, what I said before.
01:41:07I think we're at this point in time where we all need AI companions and chatbots.
01:41:13And nothing wrong with riding the wave while it's there.
01:41:18So, yeah.
01:41:20I think, however, ultimately, it's not an industry with a huge future.
01:41:25Because we'll see more, we'll see AI companions that are integrated
01:41:30into mobile devices like Google's Gemini and Android devices or maybe the new Apple Siri.
01:41:36So, I don't think people will use like 10 different apps for different services.
01:41:41They will expect to have one AI chatbot.
01:41:44And then other services can integrate with that.
01:41:48But that will mostly be with boring APIs, MCPs, whatever.
01:41:52But I guess it'll be that future and not a bunch of extra companion apps.
01:41:59For certain areas, you will have dedicated apps, but I think for many areas, you don't need them.
01:42:13I'm using Minimax and Cloud Code and using Context 7 and Play, right?
01:42:15And that is fine for most of my website web app stuff.
01:42:17My website web app stuff.
01:42:19Yeah.
01:42:20Excuse me.
01:42:24Sounds about...
01:42:25Makes sense.
01:42:26I mean, in general, it will be interesting to see how good open models will
01:42:32become or if they will continue getting better and better.
01:42:37Because open models, especially if they are really good at tool calling and so on,
01:42:43are definitely intriguing and an important alternative in a world where tokens will become
01:42:52more and more expensive in the future.
01:42:55But we'll see how good smaller models can get at specialized tasks.
01:43:04"Your voice is really low.
01:43:06Please increase your mic volume."
01:43:07I don't know.
01:43:08You're the first to write it.
01:43:10So I guess it's fine for the others.
01:43:14I'm sorry if it's not.
01:43:15But it's the same settings as always.
01:43:17So it should be fine, I hope.
01:43:19Yeah.
01:43:19Looks fine on my end here if I look at my meters.
01:43:25But let me know in the chat if more people have that problem.
01:43:32Yo-yo, a new day and new teaching from Guruji.
01:43:35Thank you so much.
01:43:37Since you're using Pi, have you tried the Pi agent Rust?
01:43:40No, I haven't.
01:43:41I haven't.
01:43:42I know that it exists, but I'm using the original Pi because it just works.
01:43:48That's the boring answer.
01:43:51What's the game dev side project you're building?
01:43:53Are you using AI in the workflow?
01:43:54What's the workflow compared to web dev?
01:43:56Is it engaging more?
01:43:57So I'm primarily doing game dev because I'm a gamer.
01:44:00I like video games.
01:44:01Always did.
01:44:02And I'm deliberately not using AI there except for research.
01:44:07I'm not using it for art generation, even though I suck at art.
01:44:12I'm just doing it as a side project in my free time.
01:44:14And I don't know.
01:44:15I plan on publishing a game.
01:44:17I have no plans yet.
01:44:19I'm in the very early stages, but I'll share more.
01:44:21Once I have more to share.
01:44:23But in the end, I am doing the game dev just because it's something new, which I enjoy and
01:44:32where I don't have to use AI as effectively as possible because it is about the journey there for me too.
01:44:41AI is frustrating.
01:44:42You have to still know how to code, but you're not allowed to write your own.
01:44:45It's hard to balance learning skills and AI usage.
01:44:47Back in the day, it was straightforward.
01:44:49Yeah, can relate.
01:44:51Have you any local model recommendations?
01:44:55I haven't really tried any of the recent local models for coding.
01:44:59So I only here, but I haven't tried that Kimi and so on are seemingly pretty good.
01:45:09I use local models for, for example, for analyzing data or documents, which I don't want to send to
01:45:17the cloud, which I don't want to send to open AI. And there, I just like the Gemma models, the Gemma 4 models.
01:45:22Really good at analyzing PDF documents and stuff like that.
01:45:26So these are my recommendations, but explicitly not for coding.
01:45:31And hey, nice to see you, Joel.
01:45:35One problem I face in TypeScript ecosystem, a lot of core packages are being deprecated,
01:45:39but standard core library is not coming up. Framework is giving so much warning how to deal with.
01:45:44Yeah, I mean, warnings in projects are not that unusual when you install packages.
01:45:48And I think in general, it's a good idea to kind of limit the amount of packages you use
01:45:53because of supply chain attacks too. But yeah, of course, not everything is in the standard library.
01:45:58One thing, if you're using AI, one thing I can definitely recommend is that
01:46:03some things you can just build on your own with AI, not everything.
01:46:08And I wouldn't recommend trying to replace React with your own vibe-coded alternative that may have
01:46:15unintended side effects. But for certain features, you could try using AI and then analyze its code,
01:46:23tweak it a bit and maybe replace a library or two with that as an alternative.
01:46:29Next question is related. Can I build a framework better than React with AI?
01:46:33In theory, for sure. But of course, React is the result of many, many, many years of work and battle
01:46:40testing across all kinds of web apps and websites. So you can definitely do it. And better, of course,
01:46:50depends totally on the definition. Smaller is possible. Faster in certain areas is possible. And
01:46:56such solutions already exist. But something that will be adopted by the industry and that's equally
01:47:03good for all potential use cases, also doable, but will definitely be hard. And the AI alone won't be
01:47:09able to do it. But sure, combined with AI, it's doable, but still difficult, I would say.
01:47:16Audio is fine on my end. Okay, good to hear. Audio is good. Audio is good.
01:47:20Okay, so yeah, audio seems fine. Thank you all for the feedback. Have heard news about
01:47:27SIG language maintainers banned pulled requests contributions where Gen AI are used. What is your
01:47:32take on this? Yeah, I mean, obviously, I'm not in the open source development space. So I can only
01:47:39comment on what I read. I totally understand that one big problem open source maintainers have is that
01:47:45they're drowning in issues and pull requests that are mostly or entirely AI generated and it's impossible
01:47:53to keep up because, of course, it's easy to generate thousands of issues and pull requests. But for
01:47:59merging, you still have to take a look at them and analyze them. And yeah, you can use AI for that too,
01:48:05but blindly trusting it is probably not a good idea. So that is a real issue right now that open source
01:48:12development is kind of not working anymore because you're drowning and all of that. And I totally
01:48:17understand that you therefore take measures like saying, "Hey, no, we won't accept anything that was
01:48:24created with AI in any way to limit this problem." The downside, of course, is that many, most, whatever,
01:48:31developers use AI these days to some extent. And you therefore exclude a large chunk of developers from
01:48:42maintaining. And I mean, that was one of the reasons I read for the BUN port to Rust that the BUN team
01:48:50maintained their own fork of SIG to add features they needed or wanted for BUN. And they weren't able
01:48:56to merge them back upstream because the SIG team didn't accept contributions that includes Gen AI. And
01:49:04that, of course, also isn't great. If you have serious contributions by a significant, serious, big user of your
01:49:12language, so to say, yeah, not great either. And I don't know what the best solution there is. I don't think you can ban
01:49:23everything related to AI because more and more contributions, code and so on will be generated
01:49:31or at least partially generated by AI in the future. But I honestly think, I don't know how open source
01:49:40development will work in the future because right now there doesn't seem to be a good solution for
01:49:47this problem of drowning in AI-generated contributions.
01:49:57In Udemy, we can either buy a subscription or buy individual courses. Does it make any difference for
01:50:03you? Do you have any suggestion on what we should pick? For me, it makes a difference.
01:50:08With subscriptions on Udemy, I earn way less, way less. So for me, buying individual courses is better.
01:50:16For students, it's better because you have lifetime access. With the subscription, you lose all access
01:50:22when the subscription ends. But of course, the advantage of the subscription obviously is that
01:50:27you have access to a broad variety of courses. Yeah. But for me, individual purchases are better.
01:50:33What engine am I using for game development? I'm using the Godot engine. Whoops, that was a different
01:50:37search. Godot. I'm using the Godot engine because it's open source. I like it. And my internet seems to
01:50:46be slow in general. Good to see that the images are not just slow to load on my side. Yeah. I like it a
01:50:53lot, but I'm in the very early stages. So I'm still learning it. And I'm loving it because it's something
01:50:59totally new. What are your thoughts on Rust? I mainly think it and Go will take a great portion
01:51:06of the market in the future. Yeah. And I guess TypeScript also, obviously not necessarily in exactly
01:51:14the same use cases, but there is some overlap. And also other languages like C++, C# will definitely
01:51:20also be important, Java. But yeah, Rust. I personally, I like Rust a lot. Always have. I learned Rust. When
01:51:28was it? Two years ago? Yeah. I think two years ago, I liked it a lot. Haven't used it over a year. I
01:51:37definitely want to use it more. Just too much stuff. But yeah, I like Rust. I like Go. So I'm on board with
01:51:46them being important in the future. And I see reasons why that would be the case. Go also very
01:51:55easy to use with AI where AI is good at Go, in my opinion. And Rust also. AI does seem to
01:52:03do a decent job. Can't fully judge it because again, I'm not a Rust expert. But yeah,
01:52:09definitely there is a lot of hype, especially around Rust. It's also polarizing, but I think
01:52:16Rust is very popular and for a reason. Do you really think it's time Angular creates a new version
01:52:23as full stack? Yeah. I mean, for Angular, there is Analog.js, which is kind of Next.js for Angular. It
01:52:32just never got as popular or even close to the popularity of Next.js and so on. But it does exist.
01:52:40There is an Angular full stack framework, so to say. I think just Angular overall is not popular enough
01:52:49amongst many developers for it to reach the same level of popularity or for full stack frameworks to
01:52:57reach the same popularity as Next.js or Tanstack or something like that has it.
01:53:05Max, love your video so much. I have a couple of questions. Is it good to start
01:53:09low-level language before choosing a specific track? Is C++ good? And if it's good to use AI in C++ or Python?
01:53:18So it depends on what you want to do, of course. If you want to build websites, learning C++ can be fun,
01:53:26but won't really be necessary. But of course, web development is not the only kind of development
01:53:31out there. And C++, for example, is something I also always wanted to learn. I did learn it many,
01:53:39many years ago, like the basics. And C++ is definitely not a bad choice. Having low-level
01:53:47language experience is not a bad choice. And using AI with it, also probably not a bad choice. Just
01:53:55make sure that you still learn the language, which does not mean writing all the code by hand,
01:53:59but simply means use AI, but also discuss the code with AI, discuss alternatives,
01:54:06ask questions and so on. And of course, learn the core rules and features and concepts of the
01:54:12language on your own too. But then I would say definitely not a bad choice. Yeah.
01:54:19The way how AI evolves so fast, wouldn't it be difficult to create a course that would make
01:54:23it obsolete fast to quote weird AIL? It was obsolete before I opened the box. Yeah.
01:54:29And that's the reason why I, for example, don't have a course where I teach AI-powered
01:54:34development or anything like that. Like I sometimes get asked why I don't have a course where I share
01:54:40how I use AI. And I do have one course, which is mostly about GitHub Copilot and Cursor though.
01:54:48But the general here's how I use AI course has the problem that this does indeed change too quickly.
01:54:56So yeah, I won't create such a course anytime soon.
01:55:03And again, hi everybody who joined. And thank you all so much for the nice messages.
01:55:10SIG versus Rust versus C++. I'm not a SIG developer. I'm not an expert at either of these
01:55:17languages, unfortunately. C++ is, of course, obviously the most established one. And depending
01:55:24on what you're planning on doing, it's the best choice for game development, for example, of the
01:55:29three, it's probably best to learn C++. You will find most resources on C++. Rust, of course, has
01:55:37the memory safety aspect, which is not to be underestimated. And SIG seems to be like a good
01:55:44blend. But again, I'm no expert there. So ultimately, I would use what feels most appealing to you.
01:55:51Because if you know one, even though there are significant differences, switching and learning
01:55:55a number will always be possible and not too difficult, I believe.
01:56:03Yeah, you joined just in time for the AnalogJS mention.
01:56:08Have you checked Vouch by Mitchell Hashimoto? No, I haven't. Can't really comment on that.
01:56:17And nice to have you here, Brandon, by the way.
01:56:21Creator of AnalogJS.
01:56:26Do you think with advent of coding agents, it would hamper new open source projects?
01:56:32Reason being, we could simply share, get help from AI agents. So why share it to public?
01:56:37Do you think with advent of coding, it would hamper new open source projects?
01:56:41Oh, yeah. So I think open source projects with AI. So one problem I see, of course,
01:56:48besides open source projects, drowning in AI slop contributions is that it's of course a bit easier
01:56:55to steal code with AI, because you can just point AI at some repository and say, hey, I want to build
01:57:01kind of the same thing. But yeah, just be inspired by it. I think it's just something you have to live
01:57:07with. And you, of course, don't have to open source software. But I will say, I think some of the
01:57:13reasons why you might have chosen to go open source in the past may not exist anymore, because there
01:57:20definitely are more disadvantages to it now than there were a couple of years ago, I would say.
01:57:27But then again, I am not really an open source developer. So there are definitely people that
01:57:33are better suited to share their opinions on that than I am, I guess.
01:57:45Next.js is so good with Vercel only. I'm tired of my personal VPS. It's very hard to make
01:57:50everything work. Is there an alternative? Yeah, 10 stack start is a great alternative,
01:57:53if you want to stick to React, in my opinion. For a front end developer, is there any resource
01:58:00you'd recommend with for UX design? I managed to learn some during work, but I think an explicit
01:58:05teaching would help me a lot. For me, really, I'm not great at UX design, web design, all of all
01:58:16things that involve design. It has really been trial and error. AI can help a bit with that,
01:58:22since you can ask it for alternatives and then quickly have it create mockups where you can kind
01:58:28of steer it in certain directions. There may be great resources, but I don't have any specific
01:58:35resources to share because I've just gone down the trial and error route, to be honest.
01:58:43I'm a mid developer. I notice when I try to code, I can't, but as soon as I see the answer,
01:58:47it seems so obvious. Do you think moving to cloud DevOps where we do not write much code will help?
01:58:53Well, you still write code with cloud and DevOps, to be honest, and you need to understand code as well.
01:59:02I don't think you can say, oh, it's less therefore it's easier to understand because you could say the
01:59:07complexity is a bit higher with DevOps and cloud. Understanding how all these different systems and
01:59:13services work together and anticipating all the potential problems you can and will run into. So
01:59:19I wouldn't say that this is easier, to be honest. I think with AI, however, you have a good tool you can
01:59:26use to help you understand the code and the concepts and so on. What are your thoughts on Nuxt or Adonis.js?
01:59:33Adonis.js I tried in my last live stream. The recording is still live. It looked interesting.
01:59:39Nuxt.js I used a lot like eight years ago. Oh God, how time flies. I like both, but I'll be honest.
01:59:46Nuxt.js I used a bit more back then. Adonis I only played around with. They look interesting. They are
01:59:53certainly useful and good frameworks, but I have nowhere near used them enough or in the last months
01:59:59to have a good opinion on them. Is 10 stack stable? It's still release candidate officially, 10 stack start,
02:00:05but in my experience, it's really stable and works well. Yeah. And that's perfect. That was the last
02:00:12question I see right now and I gotta go. So perfect ending. I hope the stream was helpful. The recording
02:00:18will stay online and I'll be back next Thursday, 5:00 p.m. Central European summertime. I hope it was
02:00:28fun for you. It was for me. I hope you enjoyed it. We'll see what we'll do next Thursday. Maybe build
02:00:32something, maybe more questions. We'll see. And I wish you a great day, evening, morning, whatever it is
02:00:39for you. And yeah, hopefully see you in the future. So bye-bye.