Setting the scene
Today, I am with Martynas Kairys. Martynas is a keynote speaker at various international AI conferences because he’s a specialist in AI strategy and development. We’re going to talk about vibe coding, which is not going to replace traditional software development, but it’s definitely disrupting it just as other factors today are disrupting how organizations need to be functioning. I think you’re going to find this very interesting. When we were talking a little bit about vibe coding, you said that it bridges a gap between what business needs and what personal needs are and technical solutions. Could we start with what is vibe coding? Can you give us a high-level description of it?
What exactly is vibe coding?
So vibe coding is basically creating a technological solution like a website, an app, online calculator, a web app, that is most of the time is simple enough, and you can do this by using just chat window, by basically saying what needs to be done, what kind of solution you’re looking for. And then when you happy with a solution that you see in the window, you ask, “Okay, so how do I make this published as a website,” or, “what steps should I take to make this as an app,” right? And you keep talking with a tool that you have, either it’s Lovable or Cursor or Horizons by Hostinger, or as some others, and you get solutions that you can use without having much of technology knowledge, right? Having said that, I have to admit that it really helps if you have some basic understanding and some fundamentals, but I would still encourage people to try it out even though they may not have knowledge at all.
For simple solutions, not for sensitive data
Because this way, you can get motivated by creating something, and you can stick with it because you are motivated enough to finish that because you see results right away. I think this is one of the best ways to see results really, really fast, much faster than hiring a developer or a designer to get a final product. Because in most cases, we do not need a final product right away. We need to have sort of an standing, if this is what it is, this is what I want, and if I want something more complicated then I can give this to developer later. So vibe coding at this point is usually mostly for quite simple solutions and for something that is not sensitive like not about financial data, not about health data, not about legal data where you have a kind of problem with sensitivity of information and you want to have foolproof solution. But for everything else, you can definitely try it out and use it.
JANE
How do professional developers feel about this? It sounds like magic, and it must make some of them feel that they might be losing their competitive edge.
Right. I think that developers who are really good ones like senior level developers, they not afraid of that because AI is still not there yet that you can work with really complicated systems and with lots of legacy code. And most of the companies, lots of companies people work for, they have legacy code. You need to have a context. You need to understand the business. So there are a lot of things involved in that. The one thing is to work with one file. The whole different story is when you have thousands of files you have to work with and get one feature, one component ready. So I would say that developers who are senior, they’re not afraid of that, and they use AI tools already.
Used to do mockups, present ideas to management
And for some mock-ups, for some small ideas that they want to present to the management, they use vibe coding. So they’re like, “Oh, so you want this? This is what you asked for. This is what we’re looking for.” And management says, “Yeah, yeah, this is exactly what we need,” and then they can work with build developer tools if I can say. And even then, they use different AI help tools to help them. In some cases, not to write code but to evaluate code, to find some bugs, which AI is kind of really good at it and better than humans in some cases. For starting developers, for junior developers, I think there is a challenge here, I have to admit, for businesses because if I hire a senior developer and if I can give him different AI tools, he can do much, much more than junior developer with those tools. And instead of him spending time training, junior developer, mentoring him, he can deliver what’s needed faster.
A bridge to mentor junior developers
So there is this gap in a way, but at the same time, I believe that company, they’re looking for long-term value for the investors. So if I invest in the company, I want this company to be great at what they do, not just for one or two years. I want them to be great for 10, 20, maybe 30 years. So I only am going to invest in the company that have a way, a bridge to mentor those junior developers so they can grow into those roles. Right now, we’re talking about development about programmers, but actually, this is with everything. The same with marketing, so a lot of marketing tools. I mean, if I have to hire a junior marketing person, there will be a lot of challenges with that because he or she doesn’t have enough experience and there will be a lot of mistakes that have to fix, and I can do it much faster and better if I’m a senior marketing person and having AI tools.
But when I look long term, I need to invest in those people. I have to nurture them. I have to mentor them. This is how companies should structure their future workforce. They need to hire more senior people who are capable of mentoring and coaching people because it’s not enough to say, “Oh, you a senior developer or senior marketing person or a senior business person, here’s a new person. Please be a mentor or coach for them.” Some people are not mentor that. They don’t want it. They don’t like it. And some people love it. They love doing what they’re doing, and additionally, they feel they can bring a lot of value and they feel fulfilled when they mentor other people and coach people. So I think this is a huge part. This is a really, really big part. I do believe that companies who hire these type of people who are professionals in their job already, and they have coaching, mentoring capabilities, those companies going to thrive.
Do you think vibe coding gives them an advantage, a way that they can train the new people easier and then move them onto more professional coding?
MARTYNAS
It does. It does. If we’re talking a programming specialist because there’s also a new term called vibe marketing, so everything… If it’s easier, it’s becoming like vibe.
JANE
It’s a cool word, isn’t it? Vibe.
Vibe coding is not for final results, but for better results to show management
Right. Right because it feels easy and so on forth. And I think those companies win if they use at least part of the job is vibe coding because it’s faster and easier to get results. Not final results, but better version type of results to show that to a management team or to clients or stakeholders. That is a much, much faster way to do that. I mean, I remember when the first time I was doing… I was in startups. Now, I’m working at Shift4 and it’s a big company. But before that, I worked in my own company, startup, 2012, and we had the hackathons and we would need to build something during the weekend. And now, I mean, what we build during the weekend, now a person can do that in an hour.
JANE
Wow.
MARTYNAS
It’s that different. And yeah, it’s huge change. But vibe coding is… I’m not sure if we’re going to talk about this about my experience, what I did with vibe coding.
JANE
Yes, I was just going to ask you about that. You’re reading my mind. That must be the vibe of mine. You have a vibe mind.
MARTYNAS
Vibe mind. Yeah.
JANE
Tell me about what you’ve done.
The app I built for my daughter
This year, I was a keynote speaker to LOGIN Conference. One of the keynotes that I did was when AI is going to replace developers, right? Programmers. And I showed some stats and hiring information and what developers think, what business thing. And I thought, okay, showing stats is one thing, but let’s put it to a test, not something super difficult, but much more difficult than just a web app or just a website. So I decided to build an app for my daughter. IOS app is more difficult than Android app because of the whole procedure, exceptions of the app store. So the whole thing, a lot of things involved there. It’s also designed, some copywriting, and of course writing code. So the idea was very simple. It was like we have a time for our daughter for Saturdays for cartoon time, and we have a problem that when she has to go to kindergarten, it takes ages for her to get dressed up.
And every time I ask her, “Sophia, so are you ready?” And she’s like, “Yeah, yeah, almost ready.” I come to her room and no, she’s not even started dressing. So I said, “Why don’t we do this?” The less time you spend on getting ready for kindergarten, the more time you have for cartoons. So there’s a sort of gamified timer that tracks time. She wakes up. She clicks the timer, and she sees the clock is disappearing the time and the faster she draws up, she gets point for that. And every time, she’s finished dressing up, she reveals a new Pixar style cartoon that I did with Midjourney app designing images, and that’s one of the most motivating thing for her. She can reveal more and more cartoons. And when she has to brush her teeth, the timer goes differently. It adds time because she has to dress… She has to brush her teeth two minutes, otherwise it’ll not be the point. It needs to spend the adequate time.
So there’s also gamification involved, getting badges, getting points. We also agreed points can be exchanged to cinema tickets and some other things. So on average, she would go and it takes her on average 40 minutes, but with that app, on average, it takes eight minutes now, and the fastest time was three minutes. So the first version that I created was with vibe coding. I did that… I think the first version I had was less than 30 minutes. We had something working as a web app that was perfect. But the final version that I have in an app store right now, it’s called Time Trophy. You kind of like it. Trophies for time, right? It took me 42 hours. So in total using different AI tools. So it’s more than an hour, it’s definitely more than 30 minutes, but it’s still not that much.
Gamification and using a combination of different apps
When you develop an app that is gamified that has so many things involved, there’s images that I had to create. So there’s a lot of things, but the way I created that, it wasn’t with one tool, and this is where an interesting part is. So for example, you can start with Lovable, pretty good tool. Lovable or Horizons by Hostinger. They’re doing something simple, similar actually. And you can create sort of a mock up, some information that you can already use. You can export some files and then you can use in Visual Studio Code. And in my case, I started with that, but then I switched to Cursor. I had those files there. And on top of that, I was using ChatGPT. I was using Claude, I was asking some questions. The reason for that was… There were actual a few reasons.
One reason I was getting auto credits and Claude would say, “This is too much. You have to wait for three hours now or five hours.” So I would have to switch to ChatGPT and vice versa. And in some cases, for some reason, one model would know where is the bug, whereas the problem that wouldn’t. And what surprised me was that it can create some really difficult things quite fast, but in some cases, some simple things, it takes so much time. For example, I said the design, I mean, the alignment is not the right one. I mean, this is something wrong. Please fix this. And I think that part, it took me two hours. That was very simple thing. Of course, if you’re a developer or you understand code, and this is what I should have done that myself, I would go to the specific file where there’s design is I just do some quick fixes and to see how it looks on the new screen.
So anyway, switching from different things and also working with additional tools, I keep asking, every time I would get stuck, I would ask, “Okay, I’m stuck here, help me with that, right? I don’t know what the problem is. I don’t know how to tell you.” And Cursor or other tool would say like, “Okay, why don’t you add the bug testing system?” And so I don’t know how to do that. How will we do that? And so, okay, I will going to do that for you, and you’re going to just tell me what you see on the screen and just copy paste that information. And I would have that on the screen. I say, “I cannot copy paste that in simulator. I need you to create me a copy button for copy pasting that bug information that you need.” These type of conversations, you need to keep going.
With vibe coding, you need to be stubborn, figure things out
In this case, vibe coding wasn’t that fast. I would say you still need to be stubborn enough to finish the job. And I know a lot of people are not, they want to have this, then they want information right now. They want to see results. So even though AI can help do a lot of things, I think that there are differences among us, among humans. If I would ask, I have a friend, he said, “I need a developer to help me with a website to do this and that.” And I sounded like, “Oh, well this is easy. Just use this tool and you can do it yourself.” One week later he says, “No, I just hired a new developer.” And so like, “Why?” And he’s like, “I got stuck. I didn’t know what to do. I understood that.” I thought that was easy, but he didn’t want to spend his time there.
But if I would give a task for a developer programmer to make calls to business, he would maybe do two, three cold calls and would come back to me and say, “Martynas, this service or this product is no good. Nobody’s buying.” But business person, he would be calling or she would be calling as long as it’s needed until it’s getting sold. So they were like, we’re different type of people. And I think the ones that have perseverance in a way, and they’re curious, adaptive, interested in learning, continuous learning, they’re going to win in the next 10 years. Those are the ones, those are the people I would definitely invest in. These are the right people.
Invest in people interested in continuous learning
You talk in an article I read that you published on LinkedIn, you talked about four types of vibe coders and what you just described now reminds me of your last category, the curious explorers.
MARTYNAS
Yeah.
JANE
You remember that you start with the problem solvers, the process optimizers, the data wranglers, I love that term, and the curious explorer.
MARTYNAS
Yeah.
JANE
I don’t know when you wrote that article, but do those categories still make sense to you?
MARTYNAS
It does. It does, definitely. And I think curious, self-learning person does even more. One thing that I would recommend for anybody is there are actually two books that I always think of right now. The one book is by Malcolm Gladwell, it’s called Outliers.
From specialist to generalist, outliers to “rangers”
Oh, yes.
MARTYNAS
That’s about this 10,000 rule that if you spend 10,000 hours on specific subject, you become a genius level person. Then there is another one, I don’t remember the writer, but the book is called Range, and there is a forward word by Malcolm Gladwell that’s very interesting because that book, , is about the opposite of 10,000 hours. It’s telling that you need to have a range of experiences. You have to be interested in different fields to have the best solution. I think this is where we are. I mean, before that, we would say you need to be really good specialists or something. Now having understanding of different things, you can create so creative solutions that never existed, and you can come up with decisions and ideas that went up possibly before, but now we have AI that really helps. And I remember very good example from that book from… I think that’s from the book Range. They have this catastrophe in Southern America, the Central America part where this BP oil spill.
JANE
Spill.
MARTYNAS
And they need… Yeah, that was disastrous thing, right?
JANE
Oh, yeah.
MARTYNAS
And they were looking like who could help them to come up with the idea to collect oil from the bottom of the ocean because from the top, it was kind of easy, but they were at the bottom, and it was difficult to collect that without destroying the soil that is on the ground, just collecting the oil. And they ask different biologists, they ask different scientists, NASA specialists, and the idea that won and then they used was by a bartender. And he just came up with the idea because he doing those drinks. He said, “With a straw, when you drink, you have different layers of specific liquid, and if you want to drink one top type of liquid, you just have a short specific depth.” So he said, “Yeah, we could do this for oil.” So that was very interesting. And this idea came from person who is not involved in oil or science or NASA or anything that came up with some really clever ideas. It was very different. So yeah, those two books I recommend for everybody, especially now in this age.
JANE
If someone’s interested in learning or trying vibe coding, how can they begin? What can they do?
How to start with vibe coding?
Well, that’s the usual thing. Just go and do, right?
JANE
Just go and do it.
MARTYNAS
Yeah. I would say, you need to have a problem that I want to solve, or you have to be something curious about. So my example was I want to solve a problem with my daughter with getting dressed in the morning fast enough. And I said, “Yeah, why don’t I try this,” right? But you can start more and more simple than that to say, “Okay, I want to build a website.” So a person would never build a website, so that could be a good start and could be about anything because with vibe coding, you could get even information on text that you need to write, copy the design, everything. So I would say maybe starting with a website, that’s the most easy part. I would go either with Horizons by Hostinger or Lovable. Those are probably the ones that easy to see results right away. There’s other ones, of course, like Bolt, Replit, Cursor, but the first two I mentioned, I think this is the first ones that you can get results really fast.
The problem, there are some others, but these are the ones that I tried that work for me. With vibe coding, just go start. I think website, even like a silly website, just for fun, just do it. I mean, just go and use. And there are even some free trials. You can already see results of that. But I would really encourage to get not on free trial, but to spend some money to buy pro accounts, at least use for one month to see what you can get out of it because this is the best way to understand how AI works. This is not the only way to understand that, but because if we talk with scientists, they have a different approach to that. But I would say for most people, for most humans, we need a very, very simple way to understand what it is. Because right now, we see, “Oh, so there’s ChatGPT. Okay, cool. I’m using. There is Midjourney to create images,” but there’s another thing, try vibe coding so you understand better that you can create with that stuff not just text, not just images but also solutions.
JANE
Could you tell me about your experiment comparing AI writing and human writing?
Comparing AI and human writing
So last year, the end of November was two years anniversary for ChatGPT, even though it started the whole thing 2018, but the ChatGPT that we use today was two years. And I have a friend who is communication expert and she’s saying like, “Oh, that’s easy to guess which text is AI,” and yeah, I mean to distinguish text as AI and which text is written by professional writer. And I said like, “No, I don’t think so. I think that’s pretty difficult.” I said like, “Okay, why don’t we do official experiment and we do it publicly?” And at that time, I was really spending time, there’s this… I want to recommend this guy, he’s Rob Lennon. There’s Lennon Labs. He’s one of the people I follow. I love what he’s doing.
He’s really helping the community. He specialized, at that time, quite a bit in producing human-grade writing. And I would say like, “Okay, let’s do experiment. We are going to invite professional writers, journalists, editors-in-chief who are going to provide text, and I’m going to use some of the techniques that I learned how to mimic their style.” And I created 13 texts and 12 texts were written by professional writers. And then we had like 25 texts, and we ask people to find out to say, which text is AI, which text is human? And out of those 25 texts, 1,400 plus people who participated, the average right guess identifying if it’s either AI or human was 13.66. So it’s a little bit better than average.
So from those 25 texts, they identified correctly 13 texts. So in essence, we can say that we cannot distinguish. Nobody got 25 points, 24 points. There were two people that got 23 points out of 25 points, two people, that’s it, out of 1,400 people. And number one text that was most human… They voted as the most human, 76.4% of all people who participated, the number one text was the one I generated with AI. It wasn’t human text. The second and third place were by humans, by writers. One of them, he’s actually also a professional, not just a writer, journalist, but he also do training on writing. I just [inaudible]. He’s a great person. I love his writings. But even him, he was surprised how AI can do this.
So in a sense, I can say that there are some universities who tried to figure out if this information that was written by a person was AI or a student. And there are always some debates on LinkedIn, “This is text by AI or not AI.” I think this is not the point. The thing is, I always ask this, let’s say you need to read something about something. And there’s one text is unique, it’s written by human, but it’s so boring. It’s so boring. Are you going to keep reading that? You know it’s by human, but it’s super boring. And there’s another text that is written by AI, 100% AI, but it’s so interesting. Tell me, which one you’re going to keep reading? I know everybody knows the answer. So I think we’re trying to ask questions sometimes not in the right places.
And I think that if AI can do something better, just allow it to do it better. I think that professional writers, they should keep writing because they know the audiences. They can use AI tools to get a structure or to change the style, but they are the ones creating the narrative. They have ideas. They have stories. AI doesn’t have that, but if you just need to write something, it can produce human-grade writing better than human. And that was insane. And this is what I wanted to address for writer’s group. Instead of being like ostriches and putting our heads into sand that it’s going to pass. No, it’s not. It’s here. And it’s writing better than most writers. Well, not most writers than some writers. Sorry for that. Not most writers, but some writers for sure. And yeah, that was very interesting experience. That was very interesting experiment.
JANE
In general, do you feel optimistic or pessimistic about the future? In a general sense, not with coding specifically, but in a general sense over say the next 10, 15, 20 years, do you think things are going to get better for us humanity or not necessarily?
Optimistic about the future but a little scared
My nature, in general, I’m optimistic person. So let’s talk with that. So it’s very difficult for me to be not optimistic about anything. So I am optimistic, but at the same time, I’m a little bit scared about things that we do not know yet. And there are some things that we do not know yet. When you create a program, it’s either working or not working. With AI, you do not know what to expect in some cases. Of course, some people saying like, “Yeah, just switch off electricity and that’s it,” right? But it doesn’t solve the problem entirely. It just solves when you switch off electricity, but what happens when you switch on again? So I would say a lot of opportunities for good and for bad. It’s going to depend heavily on if we have the right guardrails in place, but at the same time, not too many guardrails because it may stop progress. And in some countries who do not have guardrails at all, they will be more advanced. So I’m a little bit worried about the youngest generation, what’s going to happen with that.
JANE
That was my next question for you. What do you think the younger generation will be able to do that we couldn’t do? Or you just said you were a little concerned about the younger generation. Why is that?
What will the next generation do better than us?
A few things. So in my age, my game that I used and I played was… I think it’s Japanese. It’s Tamagotchi, where you have a friend, you feed it, you give some corns. You have it in your pocket. This is when I was, I think, in maybe second, third grade or something. That was my friend, friend from other friends. Now I see kids, they use ChatGPT as a friend. And I mean, if it gets complicated in that… It can get complicated in a way that if I have trouble communicating with people, with my friends at school, I’m probably going to spend more time using ChatGPT as my friend. And there is a risk that I’m going to spending too much time with my ChatGPT. And at some point, I have to face a reality that there are some harsh kids, harsh adults, not just in school or university, but in business as well. And because I was living in my own world, I may face that and I may have a depression or some problems with that. So this is one thing.
Another thing with the younger generation, I just heard that recently, and that’s very important for anybody who has teenagers, I would say. This one I heard that’s becoming a pretty big thing in United States, that there is a group of bots, group of people organized crime that behaves like a really human and send texts and images to… Usually, that’s to boys, to teenagers. And you see there is an image or a text from a girl from school and she’s saying, “Oh, I’m going to send you a naked picture. I need you to send me back yours.” And with AI, you can create those things and you can create really human-like communication. And you receive this photo that looks like this is the girl and she’s naked. And in return, you go to the bathroom and you do your thing, you take picture.
And later, you get a message saying, please buy her $200, or we will publish this picture to your school and to your friends. And there have been some suicide in United States because I mean teenagers… And that’s a problem. And this is what they’re saying, it’s becoming bigger and bigger. This is not… I mean, they’re switching from this Nigeria prince that’s sending you emails saying like, “Oh, you won a prize, or you have grand, grand, granddaughter that died here. Grandfather, not granddaughter, and you have this money you can use.” Now switching to that, it’s very difficult to fight it because it’s not one person, it’s a group, and this is a group of AI modernized bots that does it. So these type of things are really worrisome. And some videos, for example, that Google provided a possibility to create videos that looks like real, I mean, 100% real videos. It’s becoming difficult for people to think that this is fake.
JANE
What can be done about that? Because as you say, it’s not an individual who’s doing it or a small number of people, they can’t be tracked down and put into prison for that. It’s a broader scale thing across different countries.
We need to learn critical thinking – adults and children
I don’t think you can find it that way because you may close this thing, but there will be a new one and it’ll be fighting with the new ones all the time. I think the best way to approach this is to teach children and humans in general critical thinking. This is the thing I would say that makes most sense. And also how to use resources in a good way that you can double check if this is the correct one.
JANE
So critical thinking needs to be taught at a very young age.
MARTYNAS
Yes, yes. And it has to be taught not just by talking with kids, but also with showing examples and showing… I mean, even showing some bad things, what’s happening. We have this discussion with my wife. One, we have to show a specific movie to kids about traveling, about kidnappings and other things. This is not about AI. There are some other things we have to worry about, but these things as well. So even though I’m still optimistic about AI, and I believe that we will figure that out, even though sometimes when I hear some news from a lot of countries I think will live in simulation in a way that how some people can behave this way or do this and that. It’s like this is not real. This is not happening. This must be like a simulation, but this is how it is.
JANE
Good. I also feel optimistic, but I do think there’s a lot of awareness raising that needs to be done. We’ve gotten away from vibe coding, but that’s fine because I wanted to talk with you about these other dimensions of the future. Is there anything else, Martynas, that you would like to say? Any message that you’d like to share about vibe coding or about the future?
We need to find problems we are interested in solving
I would say that we need to find problems that we are interested in solving. I think if this problem is important to you, you would be the best person to solve it, and you need to dedicate time for that. And in some cases, it doesn’t mean you have to quit your job and pursue this and have a startup, even though I believe that everybody should try to have their own business at some point in life. But having this as a side, as a hobby or something you can work on that is I think this is a great balance of your personality. This is what you do at work and you love what you do, or at least some parts what you do. And then there’s other parts like where you can create some things, right? And in some cases, you can even earn some extra money. So I would say keep learning. This is a big part, but the best way to keep learning is to find interesting problems to solve. Otherwise learning because of just learning doesn’t make sense.
It’s not sustainable and it’s very difficult to predict future, especially now awaits us, but staying curious is a part of it. And also since I’m a big believer in investing and finance, I worked in the bank before. I think this is an interesting time to be investing in companies related to AI. Even though some companies are really probably are overvalued. I would still look into opportunities to be part of it and enjoy this journey. We do not know how far can it go. We may see humanoids walking around us within five… Within 10 years, I can almost guarantee 100%. Within five, not sure yet. But within 10 years and we will not be talking about AI because we’re not talking about internet because that’s part of our life already, right? We’re not saying, “I’m going to go on the internet and check that.” We just say, “I’m going to Google it, right? I’m going to mail it or something.” We don’t say word internet, so I don’t think we’re going to use AI word at some point because it’s part of everything, right? It’s just how it is.
JANE
Okay. Well, that’s a great closing statement. I think realistic, but I think it’s also inspiring. Thank you very much for your time.
MARTYNAS
Thank you, Jane. Thanks, guys.
SUBSCRIBE TO IMAGINIZE WORLD ON YOUTUBE
We talk with forward thinkers, scifi visionaries and pioneering organizations about people and society, AI and humans, the earth and survival. Read more Imaginize.World
