070: Morten, a new co-host; Discussing the current state of education and AI
Hello there. You're listening to Go Podcast. I'm Demnik St. Pierre.
Morten:And I am Morten Wittesen.
Dominic:So, yeah, today we are trying something new. So like I was saying in the last episode, you know, I want I want to bring some dynamism to the podcast. So I'm trying to find some cohosts and whatnot. So Morten is there. So, yeah, let's let's start by hearing a little bit about you, Morten.
Morten:Yeah. Okay. So, yeah, my name is Morten. I'm from I'm from Denmark. I live in Spain right now.
Morten:I've been doing Go for the last, I think, five, six, seven years or something. I mainly work as a freelance consulting developer, do a little bit of open source and educational content as well as a lot of AI stuff. Yeah. I think that's that's me. A quick rundown.
Dominic:Nice. So today, we wanted to touch a little bit on the subject of AI and education, you know, especially I I think I think we we will have some kind of interesting conversation because we don't really have anything prepared. So we will we will, you know, do it by here and things like that. So I will I will let you start maybe, or maybe I can start. I I don't I don't even know.
Dominic:I'm not super familiar with having a co host. I usually do interviews with with other people, and now I'm I'm, you know, asking questions and whatnot. So Yeah. But, yeah, I will I will let you start.
Morten:Yeah. So I think just to preface this a little bit, I've done a course on how you do full stack web development in Go. I started doing that over a year ago and did a release around December, an early access release. So we've following the whole educational content development content and AI closely. You can definitely see a shift in the markets.
Morten:The big ones like Laracast are doing big changes. Right? Aaron Francis also doing some pivots. So, yeah, I mean, obviously, have opinions. Right?
Morten:But I'm also trying to pedal a call so that they might be biased. I think that's important to to preface it right. I I think my main thesis is that we are in a in a in a time right now where maybe there's an overestimation of how efficient you can use an LLM as a teacher. I think people will probably realize that unknown unknowns is is a big thing when you're trying to to learn a topic. Right?
Morten:And there's nothing there's nothing really to push against you when you're using an LLM. Think I that's gonna be one of the big things to solve, like, when how do you know when to push back. Right? But I also come from a background of having taught myself programming through courses. I started with something called Team Treehouse back in the day, did a Python course, and then I did some React courses.
Morten:Right? So this has been my approach. Yeah, maybe whatever I think right now is completely off and all the young people will be completely fine using LLMs. I am I'm very skeptical. Yeah.
Morten:I I think that's my core thesis.
Dominic:Yeah. Yeah. Yeah. I hear you on that. And, you know, having published some courses myself, I I I talked with that with John Arundel a lot because he's also publishing some courses and whatnot.
Dominic:We have seen some kind of a drop, I would say in the last year, maybe year and a half potentially in sales. Well, me personally, my my material was getting very old, so I I was expecting that a lot. But even even there, I mean, it's true it's true that I think I think a lot of content or at least people that were publishing some training materials have seen a drop since, you know, since LLMs are getting there, I would say. But but like like you were saying, I mean, it's it's it's it's tricky because yeah. You you cannot really fully blindly believe what what gets out of the LLM.
Dominic:So and and you have to know exactly what to ask for. So that is that is the challenge potentially to to say that, you know, LLMs are, you know, are are very very good at at teaching. Yeah. They they are good at teaching what they know or at least what you are able to make clear for them as a as a questions. Yeah.
Dominic:You know? So that that is that is the tricky part.
Morten:And as I think you see a lot in in development because I I started working very heavily with or not working very heavily. I started using them more in, I think, April or something last year. Before that, I've been doing just using Claude and ChatTPT as a chat. I've been using Copilot. Right?
Morten:But I really started to explore them more deeply around spring last year, and I I could follow the conversations on Twitter, seeing how there there are still a lot of people arguing that this is this is really bad. This doesn't work. And I that was not my experience. Right? But I think it also comes down to if you, first of all, have, like, a a simple language like Go that compiles and you know what to ask it for, and you can be very detailed in your in your prompts and how you how you structure your tasks.
Morten:You I tend to at least see better results. And also, a prime the did did some streams where he tried to learn vibe coding. Right? And he was doing these, like, massive massive prompts, right, instead of just breaking it down and being very specific. And I think being able to break it down and being very specific takes training, and my assumption is that that also is also true for education.
Morten:Like, teach me Python, for example. Okay. You mean you know, like, okay. That's probably where I will start. Right?
Morten:Sure.
Dominic:Yeah. And and that that's interesting because, frankly, this this is a skill that most software engineer will will need to have, LLM or not. I mean, the the ability to understand a problem, the ability to chunk a problem into very, very small pieces, this this is something this is something that you learn to do when you are either alone or in a very, very small team in in some companies. Myself, when I when I started my career, I inherited the old IT department at the at at, you know, at a credit company, a a fintech company back then. You quickly you quickly acquire that skills of being able to oh, you know what?
Dominic:Management is asking x, y, and zed, and I need to what what is x exactly? You know, what it means in terms of so so again, I mean, the the more granular granular that you get with with the LLM, I think the better result you are getting. That is true. I've I've seen that. Because myself, I I kind of changed my mind a little bit in the last potentially four months or so.
Dominic:Finally decided to to get it some some real efforts and and it's true, you know, it's it's true that the it it it's it's pretty capable. Let's let's be honest. But but again, I mean, it's it's it reflects the amount of efforts that you are you are putting into that thing. And I think I think it it would be the same for education. But but again, how do you know what to ask?
Dominic:You know, let's put ourselves into the shoes of, I don't know, young young software engineers or people that are just starting that now see the the work environment and things like that getting harder and harder. They need to acquire some experience, and they you know, it's pretty it's pretty hard to exactly know what what should you do with with DLLM. I I would I would guess. I don't know.
Morten:Yeah. Yeah. I I think I think that's the gist of it. Right? Like like, what what to ask and how to structure it.
Morten:And, also, there's there's also this thing with open source and AI contributions, you have people coming in, and they just do gigantic PRs, like 5,000, 10,000 line changes. Right? Because what what how would they know? Like, how would they know that that is maybe not a good idea? And when you start out, you do those big changes, and then you find out in a few days that you believe, you know, it was not smart because you did a lot of things that you didn't understand.
Morten:And I think that comes with experience. Right? And maybe you could get that with LLM. That is, like, it's still such a new technology that's that that exists out there and people need to learn how to use it and all everyone has an opinion. Right?
Morten:Everyone's like, buy my my my, I don't know, book on prompt engineering or whatever. Right? But does does anyone really know how to use it? Right? And and you also then have what Dijolf, the CEO of is called Dario, Dario, whatever, the CEO of Anthropic, saying now that we are in twelve months having software automated, which is you know, his last estimate was a couple of year off and is still not true, but then you read that as a new maybe new to software engineering or starting it, and you're like, you know, what's the point?
Morten:I don't know. It must very disheartening to also get into the industry and see all this talk, right?
Dominic:I would imagine myself. I don't know. It's very hard to to put myself into my my younger my younger shoes if I can say that. I'm sure it's not a it's not a valid expression, but whatever. But yeah.
Dominic:Totally. I mean, this is I I think the the craziest thing that that that is said at the moment, I think you are receiving the most VC funding. Seems to be that or something. But but nonetheless, I mean, yeah, I I it's I think that Claude has this mode, and I'm I'm not super sure. I'm not very experienced with all of those tools.
Dominic:But I think that Claude has this mode Cloud Code. Right? Mhmm. As this as this mode where it's supposed to be a learning mode or certain but I'm I'm not sure if it's only regarding a project that you are in or whatever, or if you can, you know, put put cloud code in that mode and start asking, you know what? I I would like to I would like to learn Ascle.
Dominic:What what can you, you know, how how can we jump start? That that to me would be a crazy a fantastic way to use LLM for people. Having like more in intelligent tutor that that are capable of starting you from scratch and they would, you know, you know, they they could on the fly build a curriculum that fits what you are. They could ask some question at first. It has to exist.
Dominic:I mean I mean yeah. Maybe maybe someone is working on that. I don't know. But I I think those companies should probably start to invest into education like that. In Quebec, we have we have a a very hard time with teachers, you know.
Dominic:The education system in here is abysmal. The government isn't investing at all. They are cutting. It's it's bad. It's just bad.
Dominic:But not, you know, not a lot of people seems to be realizing that. The teachers, of course, knows that. They just want more resources for the kids and whatnot. So I'm wondering at some point if if if, you know, if we could put more efforts, you know, instead of putting efforts into letting the LLMs create, I don't know, videos of cats and whatnot, could we not try to, I don't know, try to fix some some very crucial problem that we have in in society. In Quebec, this is this is an issue, but I I don't think we are there.
Dominic:And I'm not even sure that we will be with what we have at the moment. It's I'm not even sure it's possible.
Morten:No. But I'll say you're you're dealing with these in the in deterministic systems. Right? So I mean, imagine trying to learn really hard topics where the where the output changes every time or Yeah. Potentially could change every time.
Morten:Right? So you need some form of determinism in educational content. I know boot. Dev does at least they introduced this thing called boot where you can chat with their agent to explore the answers. And then, of course, he he he gamified it in a way where you you have to there's, like, a cost to use this chatbot, I believe, which that seems interesting, but it also it it originates from a human putting together the material.
Morten:Right? And I could definitely see a way where you could expose the student to variations of a topic, probably because, like, from my point, learning has been a lot of iteration, seeing what works, seeing what not works. And it could probably be really good at that, setting up scenarios. But then again, how do you really validate the scenarios are true? Like, you you There needs to be something where you can kind of kind of trust that what gets out is is it's true.
Morten:But I do think, like, iterating and also think what a lot of the promises is with with education and something like YouTube, where you can you can adapt to a person's learning style, learning approach, could could potentially be an an area to explore. But how how that looks like, I I have no I have no clue.
Dominic:Right. Yeah. Because we we have to, you know, we have to think a little bit about that. I I fear because the way that I've always learned new things myself, and I I'm still are today, was by, you know, sit sitting there being relaxed, trying to approach a problem, writing some code with my hand, trying to read some documentation, you know, just creating things that that I was not going to publish whatever. I I I was just wanting to learn something and build something that validate, you know what, if you are able to build that thing, you will you will have a check mark that says, you are good with that, you know, that what what what you wanted to learn.
Dominic:Now what what I'm seeing at the moment, and again, I'm I'm I'm not trying to generalize, but this this is we are inundated with people that just just pick the LLM and oh, you know what? Look, I have I have created something. There there's another there's another, you know, go trans spider that arrive lately that also try to have results and enums and whatnot. There there was another one that I think it was called is it is it Dingo or something like that? That was the first one that I've seen.
Dominic:Now there's another one. People seems to just, you know, build build things and and use the the elements because you can tell, first of all, the the Dingo creator that does, you know, does not hide it himself or themself. I I I don't know. But, you know, they they are I you you're not really learning at the end of the day just just by, you know, opening cloud code and and creating something. But I think that that was the best way to learn by doing mistakes.
Dominic:Reading a little bit of good documentation or a good book or or a good thing, That was and and after that trying to applied what you learn. What I'm fearing at the moment is that and and I'm hearing that a lot from my kids friends and whatnot. They are at university at the moment and they are doing their exam with with with with OpenAI GPT thing. I mean, this is this is crazy. They this is kind of very weird what what we are trying to do at the moment.
Dominic:The impact that it will have on on the on the younger people. Because at some point, need to put your hand into something and you need to try to do something. And now with the anxiety level that this these young people have, I don't think we are helping at all. That that is that is one piece to me that I am afraid a little bit for them. Because because now if they don't have the answer immediately, they won't continue to try.
Dominic:Let's let's I I don't know for you. I I started in in 2000, my my programming career. Back then, I mean, you were reading the documentation like let's say, I was doing like v b six. Like, let's take that for example. There were not a lot of material out there in the Internet.
Dominic:You were in the MSDN documentation, the thing that Microsoft taught that that was helpful. It was not bad. But my my point was, you know, you were doing a lot of mistakes and you were really putting a huge effort to learn something new back then. So I think I think that part is very important for for us to really learn something. And that is where I am I am afraid that using an LLM, even if they are becoming extremely good, I don't I don't I don't think that we are keeping the knowledge as much as when we were doing that the hard way.
Dominic:Take for example, it's crazy the other day I was returning to go. I've been writing go for ten years. I was in the in the line. If error not equal to a call to a function, and I was I was having an error because I was not having the the semi semicolon and error not equal nil. But I I was not I was I was I was lost because I'm I'm using Cloud Code a lot these days, not for Go, because I don't want to write TypeScript so I decided to but long story short is that I was hearing that a lot that starting to use DLLMs makes you lose a little bit of your skills.
Dominic:I've felt that firsthand. I was in Go code which is supposed to be my go to language for for most everything. I was I I was writing, you know, I was trying to print out the error and I I I was I was typing error dot string and I was having an error. I was like, wow. What is going on?
Dominic:I've been doing that for ten years. Oh, that's error dot error. That's true. That I I mean, the knowledge is going very fast when you are not using that. I I don't I don't know.
Dominic:I I I think I I went to a huge tangent, my point was mainly, I don't know. It it it feels a little bit dangerous where we are going.
Morten:Yeah. Definitely. I mean, also, like, I also have learned a lot from, you know, sitting down with books and and and reading and then trying it out and then reading again. Right? But it also requires an attention span that it might might not be that existent anymore for a lot of people.
Morten:I know my my attention span is definitely hurting. It's something I have to be actively looking out for. And also my wife, she also like, both of us sometimes you get stuck in these, like, you know, the the the shorts. Right? The short loop, and you just keep it.
Morten:You have to look at each other like, okay. Get out of it. And I can just imagine having to study something with my like, with that type of attention span because a lot of it is just on the backbone, I think, because you did the hard work. And I think that's also that's the that's what you need to do. Right?
Morten:You need to do the hard work to to understand it because I had a a similar experience. I used I can't remember if it was it was not Go directly. Maybe it was CSS or maybe it was SQL. It was something simple that I used to do a lot that I just had been lazy and had LLM do for me. It And was just this, like, I I I myself having forgotten how to do a joint statement or something like that.
Morten:I'm like, what is going on? That needs to be balanced here. Yeah. And and I was I'm like, I'm I'm using the smallest plan on oh, the plus plan on OpenAI, and then I think it's called the pro plan on Anthropic. So that I that I do get limited.
Morten:You can really fast run through those users limits. Right? But it also forces me, okay. Now it's probably time to to sit down and and and read for the code and write a little bit of code yourself. So so it balances because you could just get the big plan and have have it written for you, that might be the future of how we do programming.
Morten:Right? But but if you see on, like, Tech Twitter at least, the ones who are really promoting that model has been doing software for twenty years. Like, what's it called? Armin, the guy behind Ginja and the mini version of Django, I can't remember the name right now, but he is really big on a genetic coding, But he has also been doing this for many, many years. And that is maybe efficient.
Morten:Right? And he probably don't care. I mean, he knows. I don't know. I think it's really hard to to figure out how to to to to balance that that whole workflow.
Dominic:Yeah. Oh, yeah. Totally. Totally. Yeah.
Dominic:Exactly. But, yeah, yeah, I I don't I I cannot figure myself how a a junior developer can can drive this thing and ask precise things that will output good code, to be frank. Because me, personally, I I still have some some things that I'm trying sometimes and but it's normal. I mean, it's it's it's all about, you know, there there is so many information that that get drops and it's it's still kind of you know, I'm I'm still impressed with what what we have at the moment, but but yeah, You know, the amount of of code that gets created per day and the amount of energy we have to review them from a real human is yeah, there there's a lot of the balancing there, to be frank. And I'm I'm not I'm not super I'm I'm not super, you know, comfortable letting another LLM or AI doing the reviews.
Dominic:I mean, at some point, but I I I must admit, like like I was saying, I'm I'm in a project at the at the moment. I need I needed to do TypeScript, and it's heavily on the on the front end. I'm I don't want to write that myself anymore. I've been there. I've done that.
Dominic:And I'm super super glad that these things exist now, and I just can I I just can be a product manager? Yeah. So that that to me for, you know, for front end point of view because due to my, you know, vision situation, I don't want to do UI anymore. I don't I don't want to do front end anymore. But on the other side, I have a very very hard time opening the code reviewing that and sometimes I look at the code and I'm like, oh, what is going on there?
Dominic:And now the amount of energy that would be required for me to clean that up, I I don't have it? No. This is I don't know. This is a huge problem that we will have. It was already rough to maintain software.
Dominic:You know, maintaining software is is very very hard. But now we will have a ton ton of code that no humans will be wanting to to clean up.
Morten:Yeah. Yeah. Yeah. Yeah. But as it's also, like, I'm a little bit I think it was Dex, the guy behind OpenCode, who talked about that humans, they will they will always choose the lazier option.
Morten:So, I mean, I totally get it. Right? If you can just if you can just be, like, do this, do this, do this, get it out. If it works, I think people will will will start to do that more. And like you said, like, having an AI review of our code, I'm also very not I'm really I'm really not into that.
Morten:Like, I'm I'm very bullish on the AI writing code. I think that can be that can be a good workflow, but then having an AI also review the code I mean, at that point, why don't just we might as well just give up.
Dominic:Right? I
Morten:mean Yeah. I mean, that's fine if people wanna do that, you know, like, you do you if that's how you wanna build products, and maybe that works. Maybe that will work. It's still so early, and we we saw the the boom over the Christmas holidays, right, where at least Anthropic were giving out free usage and people on Twitter were going nuts over Opus 4.5. And I was looking at it and saying, like, what what are you talking about?
Morten:It has been this has been this good for a long time. Right? So maybe this is the direction. I I don't think we have seen as many products coming out. There was also a lot of people talking about the sloppification of SaaS products.
Morten:I'm not sure if I have seen that yet come up, but I mean, six months, a year, maybe you'll start to see if that approach is actually a viable way of building products or if we were a little bit. I do think we were worried a little bit back. You know, like, the people going like this is the best thing since sliced breads, and this is all we're gonna do for now. It's kind of like a feel is a little bit like when we have the the crypto craze where it's like this is the new form of payment. Right?
Morten:This is gonna be revolutionized, and then you end you end somewhere in the middle instead of these two extremes of, like, regular money is gone or crypto is gone. Like, you just it was in this middle state, and I I think we are probably gonna end in a middle state with with LLMs. They're probably gonna get better. Right? But we are not seeing huge jumps anymore.
Morten:We are not seeing ChatDPT six, seven, eight. We are seeing five point one and five point one with a weird name that no one can figure out what means. So Yeah. Something needs to happen. And you saw Jan Li Kun from Meta actually leave to start his own because he was he was talking a lot about the the current structure and approach is is is not the way to progress.
Morten:And I'm not an expert on this, so I can only like, what he is what the experts are talking about. But I think more of them are talking about the fact that that now it's back to to to research, to get a breakthrough, to to to make them actually better. Because right now, they can't reason even though they have thinking capability. Right? It's not really thinking.
Morten:And they can do impressive stuff, but they cannot really, you know, push back or clarify or or really like, you you need to iterate with it and know what to to iterate on and and, again, go back to the point we made in the beginning, right, where you need to know what to push back against.
Dominic:Yeah. Absolutely. Absolutely. And and what what I'm I'm curious a little bit to to know, you know, so why are you using it for Go especially? Is it is it when you have let's say you have a problem that you it's either you don't want to write the code because you have written that code like 100 times or it's something new.
Dominic:I will give you an example where I personally have used it with Go. I needed to implement the, and I will probably say that wrongly, the the the Jigstra algorithm. I've it's it's it's a it's an algorithm that can calculate a position. Let's let's say you have a game. I don't know if you know that game, Ticket to Ride, for example.
Dominic:It's a board game.
Morten:Yeah.
Dominic:Okay. So you you have train station that are connected to other train stations. So I I built myself an application to be able to continue continue to play that. I played that with my family, my my wife, and my kids. But I I don't see the board anymore.
Dominic:I don't see the cars. So I I needed I needed something for me to be able to, first of all, like calculate the routes. And I don't know. I I don't I I did not really wanted to do that, so I I took the L and M to help me with that. But I kind of regretted that a little bit after the fact.
Dominic:You know what? That that that is that is very very interesting for, you know, because I I would I would have loved to implement that. But so I yeah. I was curious to to hear what are you using. Because personally, I I don't use that with Go myself.
Dominic:I I still prefer to write my Go code, and that is that is probably crazy since Go is is is compiled. So I mean, at least, it's a little bit less dangerous than than even TypeScript, for example. But but, yeah, I'm just curious to to know where are you using it.
Morten:So there's definitely an an element of laziness into, like, you know, how I use it. There are some things where you're like, I have done this a million times before. And I think that if use it in a code base where you have established some good patterns and some good practices, DLLM is really good at following those practices. Also, it's getting good at following feedback from the LSPs, a lot of these at least, tools feed the LSP diagnostics into the LLM. It's also good at, like, fixing it.
Morten:I I tend to do, like, really low hanging fruits that I need to do. For example, I'm I'm I'm building a web framework in Go where I I basically start by scaffolding MEC architecture for Go. That involves a lot of template manipulation. That is really boring to write, and it's also really, really error prone. And I have used AI a lot there where I would sit down and figure out, okay.
Morten:So I want these templates. I will create some of the templates myself and then create some some testing, some golden file testing, some something that it can validate through output. And then I can then point and say, okay. Look at the file for for controller. Right?
Morten:And then repeat this for this user controller. Something like this. That I think is very good at. I also tend to use it a lot for exploring different options. I tend to work.
Morten:I work from home. Work remotely at least. And sometimes it's just it's the thing of, like, bouncing ideas off it and having it build POCs that you then review tend to work really, really well. It it will definitely do something stupid. And if you sit down and look in the code, you find that, okay.
Morten:This is wrong. You fix it. But you can kind of work side by side with it if you again, if you know what to ask, you you can get some some feedback, some other ideas or some other angles, or have you thought about this? And then, okay, let's build version a, b, and c, and then I can look at the code without having to spend, I don't know, a half a day writing Yeah. Code.
Morten:Right? I used to work with one of my clients, and he was very on AI. He basically had me and another one who had been doing coding a little bit longer than some of the other team members, and he was always like, you guys are a little bit slower than the others. You know? And and it's like, yeah, because we look at the code.
Morten:Right? And Yeah. Yeah. And I I did build a lot of integrations with, like so this was the product that wanted to use AI for a lot of automation, and he wanted to integrate with CMS systems. So I sat down and I built an integration to one CMS that was basically connecting the client's account and then fetching data in a background job and then loading into the database and then generate some embedding so we could we could search it or the the agents could search through the through the through the records.
Morten:Right? And and once I had that structure that was like, okay. This is this is sound. This is where I want this to go. Having it repeated for other CMSs was Yeah.
Morten:Was a really good use case of this because all of these CMSs has some API documentation. So it knows the underlying data structure, it knows the code architecture, it just in a sense needs to review the API documentation and then write the code. I review it. Maybe I review it with the agents, and then and then I could easily replicate the setup for for for other CMS systems. So so that is mainly where I use it.
Morten:I I tend to write all my my front end codes in in HTML through through Temple where I I mean, if you have a lot of components, it can be okay at it, but I I definitely think it's not as good as if you use it for something like React where it seems like you can just send it build me dashboard, and you get a decent looking dashboard.
Dominic:Yeah.
Morten:So that is that is also where I do most of the work for actually writing writing the code codes. Yeah. So I think, like, low hanging fruits, repetitive tasks, exploration, and then MVPs of of a feature or something that could go many different ways, that is probably where I spend most of the time with with it.
Dominic:I I think that might be where junior developer might have a very, very hard time to draw a line on what they should outsource to the LLM and what they should try to learn or try to do themselves. Yeah. That might be very difficult to know at that stage in your career what you should, you know, what what is a lowing a hanging fruit, for example, or and and and I think the the the thing that I don't like at the moment, and I've I've said that a lot in in this podcast, but I will continue to say it, the the messages that that that are the hype that that that that is currently is that you don't need to code anymore and you you you can just, you know, put your hat of the product manager or the the product designer or whatever. But, yeah, I I think this is very very very not good to say to to junior developer. And I I don't have a lot of contact or I'd I'd yeah.
Dominic:Unfortunately, I don't I don't have any data to to base any assumption at the moment. That is that is sad. I I used to be, you know, I I used to be in position where I was hiring a lot of junior personnel and and trying to to train them. And I I've seen people, you know, going from zero to very, very good senior software engineer. I miss that a lot.
Dominic:So I'm I'm wondering what it's like at the moment in companies and organizations. Because we we are we are talking about writing codes and the education and things like that. But I think the onboarding and just employee developments in general has has also taken a huge hit globally in in a lot of places.
Morten:Yeah. Yeah. It it definitely seems to to to be a a downturn. But again, if if if you are I don't know if you are maybe not technical technical and you see all these big corporations talking about how programming is there. Right?
Morten:And I think most companies probably look at at developers as a cost center. Right? So if you can cut down that cost center, why not be a little hesitant? And I I and I you I think you can do more with LLMs. It's definitely a new tool you have to learn, but it is also just a new tool.
Morten:It's like going from writing code in TextMate to writing it in a proper IDE, for example. Like, when I went to I used to I used to use JetBrains products a lot. So went from, like, Versus Code to JetBrains, the go lands. Right? And it was like, oh, what is this?
Morten:This is amazing. I can do so much. And and it's that's where we see this. Like, it's it is a tool that we need to learn, But it's it's it's true for, like, for juniors, it's really difficult. Also, because one of the things when I have hired junior developers myself has been to look for opinions.
Morten:Do you have any opinions? And can you justify your opinions? I don't really care so much if I agree with opinions, but as long as you have opinions and you have a reason for having them, they might I might know they are wrong because I just have done this a long time, but I'm seeing the thing that this person is talking about. But the fact that they have an opinion about something is, at least in my eyes, always like a positive signal. And that that comes from doing doing things.
Morten:Right? And doing it a lot of times so you you get you get exposure to a problem surface and and know why something is probably good or bad, at least in in your mind.
Dominic:Yeah. Oh, yeah. Totally. Totally. Nice.
Dominic:So I don't know I don't know where we are with the episode, but I think I think that that's that that could be a good one. I I don't know if you have anything to say more regarding the education. I I I think we might see some kind of return to more human interaction in terms of course. Maybe the course creator will will will need to to change a little bit, maybe try to build the communities, try to add some coaching potentially, some more human interaction. That that might be what people will will look for in in the future.
Dominic:Because let's face it, it's when when you are using the the LLM, you you kind of you're kind of alone in your corner, and it's it's not it's not always fun. I don't know. Maybe maybe the human touch will will return for for the education.
Morten:I I do think there will come a time again where people will realize that it is a it's a competitive advantage to have an expert walk you through how to use a tool. Right? And then I think you will see people realizing that, okay, maybe spending, I don't know, 50 to $100 on this course from this person who's doing this for a long time is probably a good return on on investment. But I just I think with with the current discourse right now, that sentiment is not there because there is all this worry about it being done and and and this mindset that I can just use LLIM to to teach me everything. So, yeah, I I think we will return back, and then it's a question of how do you integrate AI into into these educational products.
Morten:Yeah. So so probably probably against somewhere in between. Right? But I I do think we are in a in a in a slump right now where everything is uncertain and people don't know how to to respond to this.
Dominic:Yeah. And there's probably, you know, a lot of courses out there that that was just created blatantly out of LLMs. So Yeah. It might be, you know, you were you were talking about SaaS earlier. I mean, might might be the same.
Dominic:I I don't have look exactly what's going on in in the courses world, but I would assume that, you know, there might have been some some material out out there that that were just created by the, you know, by an AI and whatnot. So that that is bad. That is bad for the customers.
Morten:Definitely. And if you just ship that, then then it gets bad. I did I did completely vibe codes and MVP of assess I'm I'm considering. And it was like, okay. Write all the code for me.
Morten:Do this. Do this. I know I I knew what I wanted. Right? But at the end of the day, I did not understand the code.
Dominic:So Yeah.
Morten:That was a a right test and throw away and then start over. And I do think that loop, if you can fully utilize that, you're gonna see some some nice returns of using AI. But if you just if you just prompt it, close your eyes, and and press release, I I yes. You're not gonna last long. I don't at least, I don't think so.
Dominic:Oh, yeah. Oh, yeah. You just you you need to be in the loop for sure.
Morten:Yeah. Yeah. Yeah. Yeah.
Dominic:Alright. So so cool, Martin. So do you have do you have any closing thoughts or we we wrap up?
Morten:I I think we will just repeat Yep. The thesis over and over again. I have prophecies at least. So, yeah, no no more no more thoughts on this.
Dominic:Alright. Alright. Thank you. That was that was fun. So so that's the episode.
Morten:Yeah. Thank you. It was really fun to to try and do a proper podcast. Alright. Bye.
Morten:Bye.
Creators and Guests
