Guest: Toms Vārpiņš, Product Designer
What’s happening in the designers’ world in the age of AI? Vibe designing? AI wiping out designers from Earth? Engineers acting like Steve Jobs? Well, we don’t know… That’s why we’ll be interviewing Tom Varpins, a seasoned product designer who will keep us company for the next Product Engineers Meetup. We’ll be waiting for you. And if you don’t join… well. You won’t be able to disagree with our guest live. But at least, you’ll be able to watch it later. Your call!
Toms Vārpiņš is a product designer with 12 years of experience turning complex ideas into impactful products. He’s worked on everything from fintech systems to AI-powered health tools and e-commerce growth. With a mix of tech and communication skills, Toms loves solving tricky problems, running workshops, and learning about everything from design and engineering to economics, philosophy, and music.
Toms’ LinkedIn: https://www.linkedin.com/in/toms-varpins
Full Transcript
Peppe (00:02.897) And we are live! Welcome Toms!
Toms Varpins (00:06.99) Thank you, thank you, glad to be here.
Peppe (00:08.243) Yeah, nice, nice. So this is one of the meetups of the Broadengineers community. So we have people live watching and I’m going to just check if on YouTube everything is fine because sometimes Riverside does some glitches. Yes, it’s all good. It’s all good. Nice, nice. So the chat is open. If anyone is listening, feel free to leave any message at any point during our conversation.
And at the end, we’re going to have also Q &A session. So yeah, feel free to ask questions and have fun. That’s the most important thing. Cool. Cool. So let’s get to Tom’s. Tom’s is a former colleague of mine. We worked together at Kinsta. He was a designer there. And so he is a product designer. He has 12 years of experience.
Toms Varpins (00:45.88) Yeah, sounds good.
Toms Varpins (00:52.28) time.
Peppe (01:03.693) And he worked from fintech systems to AI powered health tools and even e-commerce growth. And he has different interests. He likes to run in workshops, learning everything about from design, engineering, economics, philosophy, music. So yeah, a lot of interests. And today we’re going to talk about a bit of like the state of the design industry in the age of AI and how AI is impacting.
designing products. So we have a lot to talk about and also at a certain point we’re gonna even have a quick demo on I think Lovaball or VZero, even both, just to see, just to have a critique from Tom’s on what’s the outcome of these tools, where they kind of suck in design, let’s say, if they do, I don’t know. So we’re gonna see, we’re gonna see. Cool, yeah, welcome Tom’s.
Toms Varpins (01:57.73) Yeah.
Toms Varpins (02:01.806) Thank you, thank you, glad to be here. The AI is affecting everyone’s life nowadays and we have to make sure we embrace it in a friendly way and take it with us on our rides to building something that people actually enjoy using.
Peppe (02:21.735) Yeah, yeah, there’s no way back, right?
Toms Varpins (02:24.846) There’s no way back. Yeah. And it becomes useful over, it will become more useful over time once we get through the current throat of, they know the peak after the hype is the throat of the innovation curve to go. So.
We’re currently at the peak and let’s see which direction we keep going. But it feels like we are stagnating on use cases at the moment. So we need a lot of refinement into existing use cases that are being presented. So there’s a lot of work to do here and a to explore, which is very interesting.
Peppe (03:02.857) Yeah, and we haven’t met AGI yet, right?
Toms Varpins (03:07.758) Mr. AGI has not yet arrived. We have issued a ticket to Mr. AGI and we are waiting for it to be accepted basically.
Peppe (03:19.989) That is nice. So yeah, let’s start talking about, you know, what’s, how the design industry is doing right now with all these new AI tools. So at least from like the engineering world, we have Coursor, for example, all these editors are very smart and they help you with producing code. And so what’s the equivalent in the design world? What are the tools?
are really relevant at the moment and what’s the impact they’re having.
Toms Varpins (03:53.634) Yeah, the design process is very sort of full with opportunities where AI can actually do a lot of assistance and it depends on what is relevant to you in your design process. For me personally, it is very useful to use Chatch GPT to start out anything because like if you’re starting a project from zero, there’s always some kind of information around available.
to use and you can put that together with your brain dump into gpt and it will make some kind of sense of your thoughts will give you some more accurate wording and how to express yourself but that then has to be taken and applied through again thinking you can’t just take copy paste answers and move forward through it because again every product is unique every use case is kind of
different and the way people work is also different. So first main kind of benefit is that you’re able to put your thoughts into some coherent story. Next, you can also you can use it to kind of help with content creation. For example, when you’re building UIs like you’re building screens, lot of time you need to put like dummy lore, mipsum text in there and nobody really does that anymore.
used to people just write their own wording and then when the marketing people see your wording they come they come crashing at your door asking where did you get these from so now if you put in AI generated content in your UIs there’s only refinements happening at the edges of the text because if you got the essence that helps
And summarizing a bunch of information, when you work, we work in complex environments. There are JIRA tickets or whatever linear that you use. There are Confluence pages. There’s so much information around the work that we do.
Toms Varpins (05:54.498) that as a human just spending time reading it is not really an effective way because again, the documents have been created by humans and you know, people like using extra words similar to GPT in which image it’s been made like in human image. So it’s helpful to take this information and summarize it to you in key aspects and then.
it is very nice to try and test your knowledge as well, to ask it questions about information and to kind of give it the space to reason through that information and make it make sense to you. What I keep doing a lot is when you ask anything to JGPT, it’s very good to ask for the contrarian opinions, the ones that are like outside of the boundaries, because a most of the time they are not very
contrary and opinions and just opinions that are true in some cases but not in all cases and it allows you to kind of check how far your basic understanding is from the edges so that’s very useful that’s only for information management there’s there’s so many other sort of steps in the design process I have tried using v0 and there’s a codec
There are so many tools right now and you can’t really follow through, but I’ve tried from zero. I downloaded Visual Studio, added the check with the copilot that’s installed there and tried to generate some websites, which is very interesting. And there it’s mostly about getting the first draft out just to kind of…
Peppe (07:14.939) Yeah.
Toms Varpins (07:34.905) get some, you can put in all the requirements and it gives you like a good starting point to generate something that you can then use as a kind of a basis to actually refine and or rebuild, but just to give you, to get over the writer’s block of the first empty blank page. I forget how it’s called that. Even like if you’re a public manager and you’re writing a PRD or something.
Peppe (07:50.992) Right. Thank
Toms Varpins (07:58.273) You write the title and they’re like, okay, where do I start? And the AIs can really help with that. the working with prototypes is also very useful because instead of you stitching together a bunch of screens, currently, sometimes when it works properly, you can use like Figma Make or some other lovable or other of these tools too.
kind of give you some sense of a draft journey to jump off of, which is useful. And of course, know, we have the others that are basically in every part of the journey where you require some generation aspects, like for workshops or summarization aspects for condensing a bunch of information into more of a clear output. That’s where it’s very, very useful.
But there’s a lot of caveats to the usefulness part and lot of limitations in different contexts. We can get into them later.
Peppe (09:02.308) I remember that one big thing in design is are like low fidelity prototypes. So just, you know, do them on paper, black and white with sketches. And if I’m not wrong, they used to be like they are like that because you kind of want to isolate the
the flow, journey of the user from the visual impact. it’s not a distraction, right? So now with these tools like Lovaball and V0, whatever, this kind of is not like that anymore, right? Because you kind of have all these visual elements that could be distracting. So do you think this is like a problem or?
Are there ways to overcome this? To get around this?
Toms Varpins (09:59.631) Well, paper prototypes are a tool and the main aspect of them was that they allow you to explore ideas quicker. you’re sitting, like the glorified use case was that you’re sitting in a cafe, having lunch and suddenly you have this epiphany of an idea for the feature and you just sketch it out on a napkin and you go back and you start, you know, more paper prototypes and maybe you can show someone to try and interact with it, which is…
a good way to start off and helps a lot of people to kind of get their ideas into more concrete way. But one of the reasons this was an approach was because when this came out, designers were designing Photoshop and in other non design function specific tools for product design. So it was just like a really, literally easier to design a few flows with your hand than going to Photoshop and design several art frames with that.
But in modern days, do agree, like, I haven’t seen much wireframing recently, and I haven’t also myself done much of wireframing or even pop-type types. And there is this…
a lot of focus on design details, which can be and is distracting sometimes for stakeholders. Because you as a designer, you spend a lot of time figuring out the problem, the solution space, and you do a lot of alignment with other people. And then when you need to get the final approvals, stakeholders who have not been involved in the process, but they have to put their stamp of approval, they see designs.
And if you come too early, like I had some experiences where I come with a thought, with an idea laid out, but specifically made the UI horrible to look at. So that idea would be on the flow. And that now distracts from the problem because now you design high fidelity more or less. And it
Toms Varpins (12:03.758) distracts the stakeholder attention. But if you go with a very crude version, the idea is that you haven’t put enough effort into it. So the ideas are not worth talking about. So for a designer, it’s very hard to not do detailed designs specifically for stakeholders because you, the goal is to get your point across and get the approvals, got insights for your problem and try to avoid conversations around that.
And only way you can do that if everything else just disappears in the background.
Peppe (12:39.634) All right, yeah, makes sense. Makes sense. what’s your feeling? Do you feel, designers excited, anxious about using these new tools? What’s the general feeling?
Toms Varpins (12:56.27) Well, it depends who you ask and what perspective you’re looking at, right? So if you think about where, what is the source of these tools, it’s like five, six companies in general that are generated, that built tools, they keep iterating in the tools. And then there are thousands of companies that take these tools and wrap them into a product of some sorts, useful or not useful. That’s a different topic.
And you, the engineers, me, like the designers, product managers, even the C level are kind of recipients of this software. So it has not been the case where each business individually have decided we should use AI because we see specific opportunities here and here and here and here. It’s been that way around. It’s like AI appeared. And because we live in a very
driven business environment where most of the companies that are making money are investor driven and you know and so and then there’s a big push amongst investors because for them it’s the new NFT it’s the it’s the new crypto token it’s the new whatever trend you think about don’t get bitchforks out I can can verify it’s a trend in AI is a trend in specific use cases and
and other use cases but we haven’t chosen this as a tool and for most of most part we are
told to make this tool work for the product environment. And I’ve heard a lot of stories where people are made to use AI instead of their regular processes to get through the regular process, which creates delays because technology is not necessarily there in all of these ways. Like it can be useful at each step.
Toms Varpins (14:57.518) but depending on your use case, maybe it’s not even useful in your design process. So from a design perspective, it’s a very, very, very shiny tool. Like for me as a designer who doesn’t really have great programming skills, I can in a few minutes generate an app.
And that’s great. I created a do app. I was able to iterate and put like more detail into it. I was able to put new functionality on top of it. And then, I ran out of credits. So it’s a useful tool that is very, very alpha stage for real world scenarios. So for me, I just like it as a…
Conversation sparing partner basically so they have thoughts and it helps me to put them in different words. It helps me to sum up some bunch of information or to use like notebook LM to try and learn information because a lot of information nowadays is consumed like this, like this podcast or this meetup that we are doing. People consume content from people talking to them personally.
Peppe (16:02.603) .
Toms Varpins (16:04.59) because we’re talking personally to these people. yeah, there’s a mixed reaction. I have not really seen designers really celebrate. Those who celebrate, I’ve noticed are very focused on the website creation side. So for example, if you, in marketing, very useful, you can get a lot of drafts out, you can get a lot of design directions out, because it’s all about the vibes and information architecture then is put into
the vibe because it’s a brand website is a brand and the information is the core of the brand but that’s going to come afterwards the
the brand is decided. But the product design, we work with very precise things. design each input field, each button, comes from a bunch of services in the background that took a lot of time to develop. And you cannot just generate front ends here and there and then connect them easily to the backend. In one of my previous companies, tried, they were building
a bunch of landing pages and the idea was like, hey, maybe we can try these AI tools and generate landing page and maybe we can use that as a final product. It was great. could do the first draft, then you make edits and starts to lose context. So it doesn’t make edits in the way that you want to make edits. And then when you talk to developers about, can we use this tool?
Can we use the code that has been generated? And they’re like, we can, but it might be easier and more maintainable if we just rewrite it, but we can use this as a reference. So there is excitement, but I feel the excitement comes more from the investor space than from the technology aspect where it is meant to do the main disruption.
Peppe (18:01.082) Mm-hmm.
Peppe (18:07.782) Right. Yeah, that’s interesting. Most of all, last thing you said, like in my experience, I’ve used it a lot. I’ve used a lot of videos, other tools to generate a lot of code. And to be honest, it wasn’t that bad all the time. Most of all, the way was architecting the old front-end application we react. I think that the main problem was in architecting a good state management solution.
But then thinking in retrospective, I didn’t tell. I didn’t give any guidelines on that. So I think maybe it comes again, maybe better prompting can give better results, but you need to understand what you’re as well. Right. So I think that that could be a limitation if you’re not technical in that sense.
Toms Varpins (18:44.632) Mm.
Toms Varpins (18:55.905) Yeah.
Toms Varpins (19:01.006) There’s also the cost aspect because the LLMs are greater generating text. The cost per million tokens is not that high. But if you think about like generating a UI, generating any image, that is a lot of tokens. And if you want to iterate on that, the costs starts to increase. And then at the current time, there is this, this, this, this,
what is it called, the inflection point where you can continue trading with AI tool. And if it keeps the context and if you give clear enough indications, maybe you can go through and create something useful without blowing your budget. But if you’re doing anything that is a little bit more advanced, there will be a moment where hiring a person will be cheaper than keeping…
Peppe (19:31.666) Hmm?
Toms Varpins (19:53.775) paying for the AI credits. And that’s currently one of the limiting factors for experimentation if you’re not in a company, because you just run out of tokens to experiment. So you have like a half finished project and then, okay, so I did this in one day, but I ran out of tokens. So I got to the project graveyard 99 % faster than I did before. Like if you’re a developer and you code website projects,
most of them don’t really become businesses, right, because that’s not always the point, but with AI you can kind of get there a bit quicker. So you can have more exploration, like more experimentation before you kind of throw things in the away. So that’s kind of a big positive here.
Peppe (20:44.155) Yeah. And this takes us to the next topic, which is the new responsibilities that maybe designers could have in this age, right? So maybe getting out of just doing a designer part, designing part, but also maybe stepping in other roles lanes. So I can see that as like, I don’t know, probably getting a bit more technical learning to write some front-end code, which is
probably a good fit. Have you seen like this trending somehow, like the lines blurring between the different roles?
Toms Varpins (21:26.35) Oh yeah, there’s a lot being blurred nowadays. We live in a Gaussian blur world. I think there’s a big misunderstanding of the design role in general, especially for a product design role, because it is more closer to the PM role than to like a graphic designer role.
because the UI, which is the product that you in the end ship is like 5 % UI, 95 % like 90 % code and 5 % something else. And the biggest investment there is to get people to share and agree on information. And that is the most of the part. And then the design iterations, again, unless you work like in a creative agency, where you’re working on branding.
where you’re websites to make people feel something. When you’re building a product, most of it is to align functionality because it has a cascading effect. Any change in design has a downstream effect through engineering and maybe even touches architecture. And you have to test all of that to make sure it works and is compatible with.
all the things that it needs to be compatible with. So the first thing to say is that the UI shouldn’t be that big of a focus because that is just the outcome. The same way a code is not the thing that engineers get paid, it’s the work that they do. The same way PMs don’t get paid for Jira tickets, it’s for the work orchestration that they do.
Sorry, I feel like I went in a tangent and I lost your original question.
Peppe (23:10.801) No, we were just talking about like new hybrid roles, raising it like growing in the industry.
Toms Varpins (23:19.99) Yeah, yes. Yeah, so I feel like design is kind of merging with project management or project management always wants to move into other spaces. Like, look, can code a little bit here or look, I can design a little bit. And like with the idea that there could be one man or one person for the company. So designers…
role, like it depends on company, but currently a designer’s role is more going into experimenting with potential solutions and trying to understand the problem spaces while product managers who have the same responsibilities need to do the same thing. So everyone’s kind of stepping on their toes. And when you put in developers who also generate code,
and generate prototypes and then show off to engineers. So everyone is not doing the same kind of a thing if everyone’s brave enough and the organization allows to do this kind of sharing. And it does make me think that there might be too many cooks in the kitchen, like in the future.
Because if anyone, if code becomes a commodity, anyone can create a a code, two codes, three codes, or anyone can create a design or a visual identity. Anyone can generate a PRD or anyone can organize a meeting and everything becomes a mush.
The question is, will this all be squeezed in a single person who have to do all these things? So we are supercharging a single person at the expense of other people, or is it that this will actually allow some kind of an expansion to test out more product ideas or test out more initiatives?
Toms Varpins (25:29.016) The current trend that we have seen is that AI is used kind of as excuse to downsize company, to keep the company’s this and make it more efficient instead of expanding the company outwards, which is the original promise of AI. So there’s a bit of confusion on the messaging part and then the actions following this messaging.
So your guess is as good. What do you think? What do you think about this merging? Because for me, it’s not clear because everyone in the end is still building the same product as a team.
Peppe (26:10.896) Yeah, that’s a good question. I just wrote a post about this today on Lingling, actually. So maybe I can try to rephrase it. You are in the pro-engineers community here. That’s a bit biased. My mind is a bit biased. But yeah, I can say that I would say AI just made it clear that
everyone wanted to do a bit more than what they were doing before. Because let’s say product managers, maybe they wanted to test the ideas, but they were kind of blocked by maybe designers and engineers because it was a bit slower before even to a prototype. Designers kind of similar. So I think, you said, there is a lot of overlap there. But also designers, I know that
a bit like the code monkey for us, there is the design monkey, So you kind of, you don’t do a lot of strategic stuff. I think you’re right, right? But now this can kind of change because you can focus a bit more on the strategic part, not much on the hard work part, the Figma thing, you For engineers, it’s gonna be the same.
Toms Varpins (27:17.582) and the ticket writer.
Peppe (27:40.113) If we get more, let’s say productive writing code, then we will have more time doing more strategic stuff. So in my opinion, I see this, this convergence somehow between these roles and it’s, I’ve always seen this, like these three roles represent the three main activities of product building. it’s framing, framing a problem, understanding a problem.
which is usually what PM do, finding good solutions for this problem, which is usually what designers do, and then the technical parts, implementing it. So these are the three main things. And I always seen these things very, very, very intertwined between them. So they are very interdependent. So unless you understand the problem, everything else doesn’t matter. So unless you understand the tech,
you can’t really start designing because maybe you don’t understand the technical limitations of what’s available nowadays, right? Or in your system. So the way I see it in the future is maybe this is just my theory, but having teams made of people that can do a bit of everything. So as a base, have everyone that can act in these three levels.
So understanding, designing solutions and implementing them. And this could be the core of a product team. And it’s not much about roles anymore, but it’s about who does what in a specific moment. So maybe one day you can do this, one day you can do another thing, as long as we get to an outcome. And then the problem is that this is not probably scalable because you can’t have one team doing all these things or one person.
But the problem becomes coordination between different teams, right? So that’s where I see maybe PMs or tech leads leading different teams, like as a coordination effort, just to keep the vision and the direction clear for everyone. But at the team level, as long as they stay small, I don’t see why this cannot happen. So I think AI just made this clear in my opinion.
Peppe (30:02.611) we can actually do a bit more before it was a bit more limiting because of time. Most of all, you can’t do all these things because of time. But now it could be a bit faster. And as long as you understand the first principles, why not? It’s complex, But this is my theory. I don’t know.
Toms Varpins (30:25.346) Well, you have to like, it makes sense what you say. And it is again from the lens of productivity, right? And efficiency. How can we do more with less, which is, you know, current mindset in the tech industry. But then you think about good product teams right now.
developers should be understanding the strategic strategic objects. Same with designers and public managers. We should be understanding the technical aspects of the limitations and everyone should be understanding who they’re building for. So the knowledge kind of stays the same, right? It doesn’t really enable one to think a bit more strategically, but if everyone’s thinking strategically, what does the C level do? If their main, uh,
main direction is to provide direction based on all the inputs, which by the way come from the bottom up. They don’t necessarily come from the top down always. And then if everyone’s doing strategic things, then you don’t really need that many people. But then at the same time, you think about companies expanding, there’s a finite number of people on the world and all the markets have a finite amount of value that’s currently there.
Peppe (31:22.822) Yeah.
Toms Varpins (31:45.783) unless you start creating new value that people are willing to pay extra for, you are overlapping with other companies. And if the market does not expand, then AI, which enables people to be verticals themselves, vertically able to build a product, right? One person can do everything. Then it is more…
of the fact that monopolies become bit harder to control because if everyone is AI enabled, who is enabling the AI, which is a few companies right now, right? Which is all being pushed down. So.
I can be my own entrepreneur, be my own person, but I need a market to keep the life, keep me living, right? To give me enough money so I can continue working on. And then there will be hundreds of people like me offering very niche solutions to things. I heard an interesting quote from…
I think it was the creator of Netscape who said that all business value comes from coupling or decoupling. So you couple people together, you can create more value. When it becomes too big, it starts to stagnate. So you need to decouple it, break it to pieces, and let it individually grow again. It’s just like the generative approach.
Peppe (33:02.131) Hm?
Toms Varpins (33:15.854) Yeah, so like my big question is really how does this affect economy in general because if the economy is doing well, job creation accelerates. If economy is not doing well, job creation is as you see right now and
What would happen if suddenly you have a seven people team, let’s say current team, and everyone can do everyone’s job? That sounds more like a conflict situation than freedom for all. And we also see now that many people are generating websites, generating products, generating small niche solutions with the help of AI and
They’re using that to improve their skills. But in the end, who will pay you for your product? Like, are you solving a customer problem? Do your customers even have this problem? Are they willing to pay for this problem?
Peppe (34:08.099) .
Toms Varpins (34:20.718) Who are you competing with? Right? It all comes down to fundamentals. And if the market stays the same, whenever monopolies grow, markets shrink. Whenever individual startups grow, the market becomes bigger because there’s more money, more value created, and there’s more competition happening. So the question is, will this AI come as a big enablement force for the individual allowing to…
compete with those who own the AI code. I mean, we have DeepSeek, we have the Qwenn from China, we have a lot of these models that are standalone, but the hardware is not there for us to use them properly. That’s why Apple is failing with AI because they started late, didn’t have their own models, and the hardware wasn’t really there, nor the use cases, really. So, I don’t know, this is a lot of…
a of thoughts that are not really condensed into a coherent thought, but I just feel like there’s a lot of shakiness happening. We don’t know which direction the force will go. So for us in tech, it’s a great addition to have this assistant to help us go through these things. And it’s elevates other people’s ability to be more expressive, be more thoughtful.
But in the end, yeah, I’m not sure where this is going. I just have thoughts.
Peppe (35:51.482) Yeah. Well, I guess we need to use our kind of mindset, the engineering mindset, the design mindset to approach it like that. So exploring, experimenting.
and see if he actually helps us somehow. So I think this is the perfect time to show, to get Unlovable and V0, and we can have some fun creating some maps and see what they come out with. then, yeah, curious to see what you think about it. Like what’s the, just,
Are they sign critic? I don’t know.
Toms Varpins (36:38.102) I feel it would be very interesting to do two parallel builds.
One build is where we write our prompt ourselves, just whatever we think. And then we ask to GPT to write the same prompt for us with more detail and to compare the outcomes between just your regular thought generated. Because if you look at all the commercials and ads about AI, it’s like one line generates a SaaS product for you.
Peppe (36:48.995) Mm-hmm.
Toms Varpins (37:12.599) which is not realistic. So let’s kind of put that to the test. So how important prompt structure is to the outcome of the initial solution.
Peppe (37:14.764) Yeah.
Peppe (37:24.75) Let’s do it. OK. So I’m on Lovable right now. Yeah. So I think we can start with a generic one and see what happens. And then we can go to specific one. Have you got any ideas?
Toms Varpins (37:26.574) Let’s do it.
Toms Varpins (37:30.859) Mm-hmm.
Toms Varpins (37:37.934) I’ll open to LGBT.
Yes, what is sort of let’s make a landing website for this podcast so that we can gather emails and send email notifications, which requires some backend.
Peppe (37:54.36) Okay. So.
Peppe (38:00.079) Okay, so a landing page for the product engineers community to gather emails. To gather emails. That’s it. That’s it. Okay.
Toms Varpins (38:10.254) yeah let’s try something simple and I will go to chat GPT and yes can you copy the prompt to me please so this could be faster
Peppe (38:14.754) And I’ll also try it on V0, so to see what’s difference.
Peppe (38:23.98) Yeah, it’s in the chat.
Toms Varpins (38:27.438) Thank
So.
Toms Varpins (38:33.91) So, chgpt design a prompt for…
Toms Varpins (38:44.462) that will achieve the following because you need to give ggpt some excuse okay there you go we have uh i had i wanted to touch also on like um
boom on the generation aspect of the thing is because there’s the one I put in the chat, the longer prompt from ChachiPD to see if can you do a parallel build while it builds in another tab or a new browser or something? I was thinking about for design because I would love to be able to put
Peppe (39:08.248) Yeah.
Peppe (39:19.947) I think so, I think so, let me try
Toms Varpins (39:32.409) thoughts and, and kind of market research into like a chat box. And then lovable just spits out like a bunch of generate a bunch of potential UI ideas. That’ll be amazing, which it might do in a while. But then I realized that they have been trained on limited amounts of data. They’ve been trained on a fixed set of data and
If we think about innovation, how innovation happens, innovation comes from copying and improving and adding on top. While I mean, it’s, I can read every single thing. It doesn’t do hero well.
Peppe (40:15.102) Incredible. Yeah, it’s a bit messy.
Peppe (40:23.031) Yeah, so I think it generated an image and they used it as an arrow, they also, it has some headers here, but then also the other is here. So weirdly, it never happened before. Yeah, this is from episode one.
Toms Varpins (40:37.112) But this is from your simple word prompt. This is just you writing down. Yeah.
Peppe (40:43.051) Let’s see if we got, why Broad Engineers? Tech and class sensor, product mindset, ship faster. Yeah, makes sense. Trustable Broad Engineers worldwide. I mean, to my eyes, it looks clean. Just a bit too simple, you know.
Toms Varpins (40:49.868) It’s not bad.
Toms Varpins (41:01.688) This does not give high confidence that the email will go to highly reasonable people.
Peppe (41:12.998) What do you think is the worst thing here? The first thing that comes to your eyes?
Toms Varpins (41:19.532) It’s the, it’s the gen, well, if we ignore the bugs, for me, it’s the generality of how it looks like because it is built on existing designs, which.
are not used as references. Again, this is like a big soup, right? You train the AI on a bunch of information. The AI takes it as a soup to learn from it. And then from that module soup, provides something like this. And it does not have, because we want it to be available to people, the detail in the output is very low.
And it is using what is the average website on the internet. So it just, generates, like what was mine? It was a generative mediocrity. What I call it is that it takes the average of everything. But if you look at any, any SAS website, lovable, for example, it is fine tuned. There’s.
Peppe (42:04.374) Mm-hmm.
Toms Varpins (42:33.1) there’s all kinds of small touches that technically, practically are not important from a technical aspect, but from the human perception aspect, those little details that craft aspect is what creates trustworthiness into the product. So here we have the more complicated prompt, right?
Peppe (42:57.623) This is the vid0 version. Yeah, vid0 usually, I think the target of vid0 is a bit different from lovable. So lovable tries to make things a bit more beautiful, let’s say, appealing. But vid0, think it’s more for technique. Yeah, vid0 is more like for, I know, I think the targets are developers here. It’s very usually very…
Toms Varpins (42:59.556) this is a V0. This is still your same prompt with a V0. Okay.
Toms Varpins (43:12.27) Not succeeding here.
Peppe (43:23.559) This kind of style is what I see every time if you don’t suggest anything. So it’s very simple, just right to the point. Nothing fancy.
Toms Varpins (43:32.726) I would say this is more trustworthy than the lovable output. And you have to think that lovable made a product out of generating websites and v0 is a coding assistant.
Peppe (43:36.257) Yeah.
Peppe (43:45.655) It looks more professional.
Toms Varpins (43:48.898) Yes, yes, yes. There’s a…
Okay, so this is a…
Peppe (43:56.654) This gives me a good vibe already.
Toms Varpins (44:00.047) This is better. feels that, look at the G being cut off by some kind of a frame. But yeah, this feels like, okay, it’s still very basic. This is like an email campaign, right? You feel like, okay, I add, I’ll subscribe to newsletter here. It feels okay. I mean, I wouldn’t put my email here, but if you compare just…
from a technical aspect, you will compare a simple prompt, which is a more complicated prompt. are differences. I do feel like there’s more creativity in the more open-ended prompt, because with the more defined prompt, we have much less randomness in here.
Peppe (44:44.196) Yep.
Peppe (44:47.937) Yeah. And I think that that’s where these tools could be useful. Like if you kind of give them freedom to just going crazy, you can get maybe different versions that you can get inspiration from, your, like a mood board, right? You create a mood board out of, yeah. And then you can, you can keep it the things you like.
Toms Varpins (45:02.189) Yeah.
Toms Varpins (45:08.192) And you know, for the untrained eye, like if I was not in the tech industry, I would be like, and I would just be a regular person doing all the other jobs in the world. This is amazing. I just went from just putting a thought into a box to generating something that previously would require a lot of searching for people. It would be quite expensive to make this, but at the same time.
This is a presentation. This is not something that you can use. Well, you can publish it. So, Lovable will host it for you, I guess. And then you can, but it doesn’t have a database behind it. It doesn’t have any, okay.
Peppe (45:40.602) Yeah.
Peppe (45:52.439) Yeah, yeah, well, well.
Peppe (45:56.941) You can use super base so you can connect it to super base and it’s like the same as with zero, with zero allows you to do that. And then you can tell the zero available to configure it for you. And you can create something a bit more complex, let’s say. Yeah, but as you said during the conversation, the thing here, if you don’t have like, I would call it like a vision.
in your mind, who are you serving and what’s the feeling you want to give? What’s the mission, like the goal then? Yeah, it can be anything. It doesn’t make sense because if it doesn’t solve the problem, what’s the point, right?
Toms Varpins (46:44.77) Yeah, who you’re solving the problem for and how does this help you get to the outcome.
Peppe (46:49.472) Yeah, so I’ll show you this. I’ve coded something today. did the product engineering jobs.
This is the one. Based on the branding.
Toms Varpins (47:01.046) I mean, you vibed called it that with which one did you vibed call this one?
Peppe (47:08.332) This is from vidzero
Toms Varpins (47:11.148) This is from Visoroo. But when this is like, which draft is this? How many drafts did you go through before you felt like this?
Peppe (47:17.932) 4043
Toms Varpins (47:21.87) 43
Peppe (47:22.89) But it was different before. then since now, like I branded the community, I gave VidZero all the material. So the branding style guide, some visuals that I created, the mood board, and it kind of nailed it. for me, it’s appealing to see, but the vibe I wanted is there, let’s say.
Toms Varpins (47:44.942) the
This is much like you took your 40 iterations, but this is much better than it was the first iteration spit out by either the Chagy Bt prompt or the first prompt. So, cause this feels really a little bit like manicured. This feels like massaged. It feels like someone has put in effort into making this happen. Cause there’s like, you know where you are.
Peppe (47:54.508) Yeah, of course. Of course, yeah.
Toms Varpins (48:10.946) you know what’s the tagline and there’s some description more and you have the two main actions to do and those jobs, those actions kind of differ from each other from the header and the main action, but you also see that you can scroll down. So it gave you some animation that there’s more to this website than was before. So, you know, this is useful.
Peppe (48:32.84) cool.
Toms Varpins (48:36.664) but at the same time, like, what would it take to kind of from this? Does it have the CMS behind? Can you actually post jobs? Can you edit the descriptions and everything?
Peppe (48:47.466) Yeah, yeah. That’s the integration with super bass. So everything is there.
Toms Varpins (48:53.07) So you can just, everything’s mapped. So then why do we need developers?
Peppe (48:57.516) Yeah, so you can put a job in. No, why?
Toms Varpins (49:02.028) This is the, and this is, and you didn’t pay for tokens. This is like a monthly subscription, right? That you were able to iterate.
Peppe (49:08.844) Yeah, yeah. So I paid like for a month, but this is like a work of a few hours. Like if I had to do this by myself, it take way longer. But I don’t know how the code looks like, to be honest. Like in this case, I don’t care much because there is a very low risk behind this. Like it’s not like I’m handling people’s data or anything. So it works fine. It just does its job. It’s very simple. So
Toms Varpins (49:19.842) Yeah, of course.
Peppe (49:37.898) I don’t care about the code here, but it would be different if it was like upper publication with people data that the security is a big one. But in this case, who cares? Really?
Toms Varpins (49:40.29) Yeah.
Toms Varpins (49:49.538) Would you be able to make changes to this code? Like if you want to make a new functionality.
Peppe (49:53.718) Yeah, as I said at beginning, it’s not bad. So the code is not bad. It’s understandable, but also because it’s quite simple. So everything is structured properly, like all the folders. And it’s very, very clear. So it’s the way I would do it. So I would say, yeah, the code that these tools are generating, at least on the front end, is quite good. Yeah.
Toms Varpins (50:19.672) So what does that then leave for our professions then? If anyone can create a quick job board in a few hours.
where and if the future is promised that you have an AI agent and then your AI agent interacts with other AI agents, what’s the interface there? And then at the same time, what is the value of the developer or do we become just, know how in the old times there were monks who were the ones who could read and write and they were responsible for creating all the literature?
And this to me currently can resemble the same thing about with engineers and future of AI. There’s these AI black boxes and then engineers are just working away at the core because that is what generates all the other output. But it always, again, I still have a question who’s going to interact with the rest of the output if everything is AI agents? Talking to AI agents.
Peppe (51:24.076) I mean, I think you still need to understand what you’re doing. Like, because I mean, this is quite simple. And if I want to change something, I can do it. I know the overall architecture. know, like I can, I know how this works behind the scene. I think that’s very important.
But on other side, what I enjoyed more about this was not the coding part, of course. I think for me, that’s quite boring nowadays. For me, it’s the building part. So I have this idea in my mind and I can just do it quickly. like, so I have lot of fun just translating my ideas into some things now. So I don’t care how I do it.
As long as the outcome is good for me, that’s fine. So it’s just that I think it’s shifting the mindset. Yeah. But of course, this did also again, if the, if you’re building a SaaS software and security is one of the important outcomes you want, because otherwise your business can just fail and your reputation is gone. Then yeah, you need to understand what you’re doing. These tools ever even as.
like security check feature. I don’t know how it works behind the scene. I didn’t try it, but they are doing that as well. using the eye to check the security, but can you trust it then? I don’t know. So yeah, I don’t know. I don’t know if, I mean, I don’t think we are at a point yet where we can just do things blindly and just treat this as a black box. You always need someone that understands how things work.
Toms Varpins (53:06.638) And that’s going to become a limited amount of people seeing how things are progressing. But this kind of brings to different questions. We’ve been talking about like, how does this affect the individual contributors, the developers, the designers, the blog managers?
How does this affect the whole SaaS ecosystem? If a lot of SaaS is built upon other SaaS, which is built upon other SaaS, which is built upon open source repositories that have been written by people, right? The whole internet is based on a few libraries that keep it running in some sectors. again, if it’s AI agents talking to AI agents, what is the worth of these?
SaaS tools like if you create a website for whatever reason right you can ask the GPT to show me the statistics of the website show it in a graph give me a CSV what are the analytics tools then for what are the monitoring tools then for what are all these marketing platforms for if the email is being auto filtered by AI like my sure we can look at it
Peppe (54:00.768) .
Toms Varpins (54:20.642) how it affects the individual and work. But if software becomes a commodity, how does it affect all the other software? Government services will stay because governments will stay. So they need the ways to interact with them. But many of the tools out there can be
not easily replaced, but with some effort replaced. what is the, how does the internet economy work in that way? That’s that, that side. That’s the big question for me. It’s not about the individual. It’s about what is the internet for? Because if we think about the, dead internet theory is becoming more, it’s like, it’s accelerating with, our
AI is being able to generate so many convincing posts. So it’s, and if we’re conglomerating in monopolies within the business side of the internet, what is it for? It started as a communication tool, ended up becoming a culture defining, society defining tool. And now what direction it’s going, we don’t know.
But there will be always corners of the internet that are useful. I’m just wondering if… What happens to sauce?
Peppe (55:46.879) Yeah. Yeah, interesting questions. think we can leave the listeners with these questions too. I’m thinking about them for the fourth and maybe we can do the podcast later to talk about this in details. So we have five minutes left. I think we can take some time to see if anyone asks any questions. So let’s get into Q &A. All right. So.
Toms Varpins (55:55.47) Food for thought.
Toms Varpins (56:02.541) Yeah.
Peppe (56:20.49) I’ll just go through the comments and there’s something interesting. hi, Louise. I have two Louis’s here. hello. So Louise Rodriguez says that given the cost of AI, it’s inevitable that push the price onto you and you’ll have to pay open AI rent if you want to continue to have a business. Yeah, probably. Yeah.
Toms Varpins (56:43.79) It’s true now, right? Who you are paying the database, who is hosting the database? Are you paying Amazon? Are you paying some smaller player that rents from Amazon? You know, you will be already paying for all these things.
Peppe (57:01.172) Yeah. And then he says, Peppa, you can file the designer.
Toms Varpins (57:10.126) Honestly, soon, probably no, but in the next decade. Oh, let’s see, we don’t know like a decade ago. We didn’t, actually the internet hasn’t advanced that much in the last 10 years. If you think about going from web zero to web 1.0, web 2.0, then the crypto era was like from that, we haven’t really innovated on the internet.
It hasn’t expanded that much, but maybe I’m wrong. I’m always wrong, so.
Peppe (57:41.599) Yeah.
Peppe (57:46.708) Cool. All right. So I don’t see any questions here. Maybe there will be some on the channel on Discord if anyone sees this recorded. Cool. thumbs. We’re at the end of the show. Thank you very much for joining. was a very interesting conversation. And for who’s listening.
I invite you to follow us on LinkedIn. We have a LinkedIn page for the community. You can go on productengineers.community and you’ll find a website with all the links that take you to different places. We also have a Discord community. And yeah, I hope you see you there. Thanks again, Tom’s.
Toms Varpins (58:36.046) Thank you Pepe, have a good one.
Peppe (58:38.612) Thank you, YouTube.