Osom to Know - a WordPress podcast

Matt Medeiros [00:00:00]:
And we all probably have these stories right? Where especially back then somebody would go buy a theme from ThemeForest where it had like all the functionality when like went into it, like business directories, like it had all this stuff in a theme. You could buy it for $79 and then the agency owner with air quotes again would sell that for $500 to a client. Actually what you built is really worth 30, 50,000, $100,000 a year because you're replacing an engineer to build which what you wanted. I'm wondering if the economy start to change.

Maciej Nowak [00:00:34]:
Hello everyone, my name is Maciej Nowak and welcome to the Osom to know podcast where we discuss all things related to building great websites. Today we are exploring WordPress world and AI with our guest Matt Medeiros. We will start with press conference moderated panels on the business side of WordPress and tackle the big how is AI really impacting agency life? Matt shares his candid thoughts on the AI hype cycle, drawing parallels to the past tech bubbles. Then we will discuss the gap between AI promises and reality, the challenges of maintaining AI generated code, and our own personal experiences in using tools like cursor or replit. If you don't want to miss new episodes and keep learning more about building great websites, please subscribe to our newsletter @osomstudio.com/newsletter. This is OsOmstudio.com newsletter. If you are watching this on YouTube, please give us a thumb and subscribe to the channel. This means a world to us.

Maciej Nowak [00:01:33]:
Without further ado, please enjoy my conversation with Matt Medeiros.

Lector [00:01:46]:
Hey everyone, it's good to have you here. We're glad you decided to tune in for this episode of the Osom to Know podcast.

Maciej Nowak [00:01:55]:
Hey Matt, how are you doing?

Matt Medeiros [00:01:57]:
Well, thanks. Thanks for having me. It's been a journey to get on the podcast, but we're finally here after scheduling once or twice changed a bit.

Maciej Nowak [00:02:07]:
But the calendar, Tetris, you know, the, the. The game that gives. Keep on giving, right?

Matt Medeiros [00:02:14]:
Yeah, yeah.

Maciej Nowak [00:02:15]:
Perfect. Yeah. So I would like to start this conversation with Press Conf.

Maciej Nowak [00:02:21]:
You had a chance to manage or, you know, run the panel and I'm curious to know what's your take?

Matt Medeiros [00:02:33]:
Yeah, so Press Conf was great. For those that have only heard about Press Conf for the first time, it is the conference where the business of WordPress happens. Right. There's a lot of. Well, it's a core focus, of course, of the business of WordPress and the economy of WordPress, hosting companies, agencies, freelancers, et cetera. It is a continuation of an event called Pressnomics, which was started by Josh and Sally Strebel of Pagely, a company that I worked at for 3 and a half ish years. So it was great to have like that conference again.

Matt Medeiros [00:03:11]:
Right. Much more involved for the business owner than a WordCamp. WordCamps are great. It's a great place to like hang out with your friends and your WordPress friends. You can learn some things I guess, if you're at that level. But Press Conf is much more about how do I run my agency. You know, what's the, what's, what's happening in with AI, right, with WordPress and how is that changing the way we should be thinking about WordPress? And I was a moderator for the agency panel, which was 10 up valet and web Dev Studios and Kim, Brad and Jake. And it was really good to have the deeper conversations about where agencies are going in The Enterprise, where WordPress sits in the enterprise space.

Matt Medeiros [00:04:09]:
And then obviously the hot topic was AI and where folks were starting to leverage AI. And a lot of predictions about like, you know, what, what is, how is AI going to impact, you know, the content management system that we love. But yeah, I mean, still all positive mostly unless you talk to Jake from 10 up. And it does sound like the world is closing in on us all right. When it, when it comes to agency life and, and AI. But, but it was great. The event was great. The panel was great.

Matt Medeiros [00:04:42]:
I did a podcast panel as well. That was great. Yeah, all things we're really good at.

Maciej Nowak [00:04:48]:
Yeah, I'm hearing a lot of very good reviews of the event. So like, one of my questions would be, you answered this partly, you know, how is this different to any other technology conference? Because we all know where it comes and you've been probably to cloudfest as well, how this event compares to other events. Except it's much smaller obviously and a little bit different target group. Maybe more tight knit community because of less, you know, less people attending.

Matt Medeiros [00:05:25]:
Like how I've never been to a CloudFest. Okay. Event though people are trying to get me to go there. My day job is at Gravity Forms and we are starting to think about, you know, going to events outside of the WordPress bubble. Although it seems like WordPress is really prevalent at CloudFest these days. So it's almost like, okay, is this just a bigger wordcamp now with, with more hosting providers? I've seen things like roller coaster rides, boxing rings, like what is happening at these events? This is, this is almost like going to Disney for I once.

Maciej Nowak [00:06:07]:
Saw that there was a competition where, like, who threw the server harder, you know, like something like this.

Matt Medeiros [00:06:15]:
Yeah. So those, those are, those are, you know, it seems to be, you know, pretty wild. But yeah, in terms of like being much more tight knit. Yeah. The Press Conf is certainly, you know, these are folks that I, at least I've known for nearly two decades at this point. And one of the best things about Press Conf is that it's that tight knit conversation. None of this stuff is recorded. Right.

Matt Medeiros [00:06:40]:
So there's no recording of these sessions except for the podcast session at the end. And it's, it means that the conversations are going to be deeper, a little bit more intimate. And I think that's where people get the most value.

Maciej Nowak [00:06:53]:
And in terms of the conversations itself, what were the, like, questions from the public? Did have the, did you get any questions on that panel for the agencies? Did you get any questions and what were the most like, pressing questions the panelists got from the public?

Matt Medeiros [00:07:16]:
Yeah, I would say it was again, it was still centered around AI, right? Like, how are they leveraging it in their agency? What are customers asking for? You know, what's the, you know, predictions of, of how WordPress evolves with AI and you know, your typical ground floor agency questions. How do we get more clients? Are client prices changing, our budgets changing? Right. And there is a notable shift in, in, in all of it. Right. Clients are, you know, still not like, say, fully on board with AI. Right. They're, they're trying to keep up with it too, and they're looking at agency owners and, you know, the freelancers of the world to be like, okay, you're the, you're my guide. You should be watching this AI stuff.

Matt Medeiros [00:08:02]:
How can we leverage it? But what I think is, is this indication that customers might be getting a little bit sharper to the AI stuff and maybe starting to question things like, well, well, agency owner, can't you do it a little bit faster now that you have AI? Can't you do it a little bit more efficiently? Which means, do we have to pay as much anymore? So there was some of that conversation happening and certainly from Jake's perspective up in like the enterprise side of things, he's starting to see enterprise ask for those, those kinds of efficiency gains. With AI, it's always hard to predict. I was talking to a pretty big company last week and they were like, well, where do you think AI is going to be at in 10 years? And I was like, 10 years? I mean, like 10 years in AI it might as well be a thousand. I have no idea. Like 10 years. And like, I'm, I'm only thinking like, literally a year, two years out at most, where you can kind of see where this is going. Five years is a stretch. Like, I can't even predict what's happening in five years, though.

Matt Medeiros [00:09:13]:
You know, I think that over the next couple of years anyway, you know, we'll start to see those efficiencies with our own, you know, staff and our own, like, productivity level. I don't know if that means that the price of a client is going to shrink. We might be able to do more for them. Trying to look at it positively where, okay, maybe if it's the same budget, now we're just outputting more. So, same budget, outputting more. I don't know if we're going to just necessarily go into, like, output more and then they pay less. And that's not the feel that I got. But those, that was the general vibe of questions at Press Conf.

Maciej Nowak [00:09:50]:
And if you, if you summarize your panel, like, what did the crystal ball of your panelists, like, hold for the future? Like, were there any predictions drops, you know, except, you know, how, how are we? Like, how we are doing this currently with what the tech know is making it easier for us, but, you know, yeah, how about the future?

Matt Medeiros [00:10:17]:
Yeah, I mean, I don't want to give away everything because that's why folks should go to Press Conf. But I again, I would say that it's like, cautiously optimistic. No one is like, oh my God, let's, let's run away from this stuff or let's shut down the agency. You know, again, I think a lot of people talked about leveraging AI in areas. So, for example, like, I think everybody new knows that AI is going to help you, whatever, write more code faster, right? Maybe more efficiently. But that's still only like one piece of the puzzle. You know, if you have a, a full offering at your agency or even if you're a designer developer, like, small boutique agency. Like, I don't, I haven't seen any wins in design for AI yet.

Matt Medeiros [00:11:16]:
Like, I haven't seen anyone crafting beautiful prototypes of applications or even websites for that matter, where you're like, huh, this is, this is like super creative and this is something that's going to hit our brand goals. Like, I don't see humans getting replaced anytime soon. I have a lot of conversations on my podcast at the WP Minute with agency owners, and nobody is like, oh, yeah, we're totally doomed here. This is going to be the end of everything. It's cautiously optimistic. We look for the winds in efficiency and it might just be how do we get more human with our customers and our clients and our staff. Right. I think a lot of things will change.

Matt Medeiros [00:12:00]:
But yeah, in terms of the crystal ball, hard to look too far down the road in terms of technology. But I haven't heard anyone saying like agency life is totally doomed yet because of AI.

Maciej Nowak [00:12:12]:
I think this might be a lagging factor. So what's happening now is different. And a lagging factor is that if that answer to that question, if we are or not. If or not yet doomed. Right. Because it's not yet there and the transition has to be made fully on the client side and the agencies have to catch up or not catch up and be then doomed. But I think it's too early and that transition has to be made for us to see the results. But right now big organizations are not catching up.

Maciej Nowak [00:12:50]:
And those who which are catching up, they don't talk too much about it. And for example, Klarna was the leader. They like released let's say their employees from not employees, but you know, people who were on a customer service and then they are like reviving this service again with human, with people. Right. So they did the test, the results dropped. And now from what I'm hearing, at least in the reading, they are onboarding back these people to do the customer service. And this is very basic, answering questions about the product.

Matt Medeiros [00:13:32]:
Yeah. One of the things that this is a bit of an aside. It's the same thing. But there's a few other podcasts that I listen to that specifically talk about agency life and again much more in the enterprise space and they talk about AI. And one thing I have and Shopify did this, right. The CEO of Shopify recently like put out memo. Memo, yeah, memo. And I think the Not 99 designs, it was another Fiverr.

Matt Medeiros [00:14:02]:
Fiverr, right, Fiverr. They put out their thing too. And it's just like, hey employees, if you don't get on this AI thing, you're all going to lose your jobs. And I have a real issue with that at many in many le for many reasons. But one of them is like from a leadership perspective, like if, if you're a leader of this company and you're only. Your only answer to AI is to warn everybody about their jobs being cut. I look at that, I see, I see that as like terrible leadership and a bit of a scare tactic for tech companies to be like, well, we're going to cut. We're going to cut staff and we're going to just be more profitable.

Matt Medeiros [00:14:43]:
Like, this is an excuse. This is just like any other big business in traditional business, where they go, well, the economy's down. We're going to lay off 10,000 people. Microsoft just did this. Right. And it's just an excuse. And yes, AI will help you find these efficiencies, but if you're a leader in this space, guess what? You too can be removed by a supercomputer as a CEO. And in fact, the CEO might be the best thing to replace for those efficiencies and keep the other humans in place.

Matt Medeiros [00:15:16]:
Doing the role that is much more like boots on the ground than it is from the typical CEO role, which is like broader thinking roadmap, like, where's this brand going? So I think the biggest thing that I take issue to from a leadership perspective, and especially in agency life, are the agency owners that I hear saying, if my team doesn't catch up to AI, I'm going to cut them loose. I get it from a business perspective, but I urge leaders of agencies to also have a plan, have a path, because nobody knows how this is going to impact us. Have a plan and a path for your team to adopt AI. Don't just say if you don't adopt it, you're out. Have a plan for your brand and how you're going to leverage AI to service your customers, to level up your team and to be more productive. And if you can't figure it out, AI will replace you. That's the way that I see it. You know, and a bit of a rant, but it's something that I've seen and heard a lot of recently and, you know, I just had to get it off my chest.

Matt Medeiros [00:16:22]:
I apologize for, for doing it here.

Maciej Nowak [00:16:25]:
Yeah, Norris, that's interesting take because we have clients who are maybe a little bit inspired by Toby from Shopify about his memo. And they did the same thing in a way that, because if something is talked about, if this is, you know, constantly on your agenda, then it's getting done. If, if you will, ask, you know, people, okay, think about where you can use AI in your daily job. Some will, some will not. But if you make it, to make it, like, important point, you have to get back to that topic. And one of our clients did this and they have a recurring weekly meeting talking about, you know, how these tools can be helpful to your daily job. That's one thing on the, on, on your take that this is cruel and everything. Isn't it also transparent that, you know, this is not the.

Maciej Nowak [00:17:30]:
That Fiverr CEO isn't maybe softening the blow? No, this. He's being frank. And the. And this is painful, obviously, but it will happen either way. You know, for this size of organizations, it will happen. So, like, you can either accept that you are working and you are that, you know, tiny, tiny screw in a very, very big machine, or lie to yourself that we are happy family. Like, I'm looking at this as a marketing and PR stand from one perspective so that it's getting talked about, you know, in terms of PR and everything. We both talk about Fiverr now.

Maciej Nowak [00:18:12]:
Yeah. It's like, I think we are thinking the brand is with us right now. On the other hand, it will happen no matter what. Right. Because these are two big organizations to. To do anything else. In fact, you know, investors and everything.

Matt Medeiros [00:18:31]:
Right, Right. Yes. It could be a signal to investors, you know, signal to them to the market or. Or whatever. The biggest issue I have, sort of my big rant with AI is that I'm still skeptical of it. A year ago, I could not build React apps, web apps, and just like, create whatever I wanted. Can I do that now with tools like replit? I've built a dozen apps with Replit so far. Some small utility apps that just help me do things that I would normally try to do in my daily work routine.

Matt Medeiros [00:19:10]:
Other apps that I'm building for Gravity Forms to, like, help with our marketing stuff. Built a video. A video game. I built a game for Gravity Forms from, like a marketing perspective, you can play a game and. And win, like a coupon for Gravity Forms. These are like, really cool things. Whatever. Could I do that a year ago? Absolutely not.

Matt Medeiros [00:19:28]:
The big thing I have. My biggest gripe with, with AI is it's still not. It's still not this instant thing. I still, you know, if you want to use the term Vibe code or not, I'm not 100 on board with that term either. But when I'm building as just a power user, like, I can't write code. I know how to think about a product. I know how I could logically think about structuring a database or breaking down components of a React app. Like, I understand that the.

Matt Medeiros [00:20:01]:
The plumbing of it, but I could never write you lines of code. Can I do that now? Yes, but it still takes so much time, so much troubleshooting, so many prompts, and it's still not perfect. And once I get it to a level that I need to ship it, to let people start to use it. That's where the rubber meets the road. Where now it's like, now you've got users using this thing. What happens when you need to update it, what happens when you need to maintain it? These systems still haven't caught up repl it. As great as it is, when I say create me a new web app and I start a new prompt for a new idea, and I say make a database for me, it still doesn't even make the database for me. And it's just like, how far off are we from like this utopian world where these things are completing these tasks? We got to this point pretty quickly, but just like iPhones, I'm only starting to see iterative, small, iterative improvements to these things month after month or quarter after quarter.

Matt Medeiros [00:21:05]:
I'm not seeing these massive leaps that we keep, that we keep promising ourselves. And I think that's where we get caught up. Like the world of NFTs from years ago, everyone was like, nfts, this is amazing. Like, look, what we can do is sell a digital ticket. And if I sell you this digital ticket, you have access to this membership. Gary Vee was selling tickets to these events and it was all these NFT things and all we did was like project this utopian future of this technology. And everyone got hyped around it. And what happened to NFTs, I don't even know.

Matt Medeiros [00:21:41]:
I don't even hear anyone talking about NFTs anymore. And my last point to this is I think we're doing a lot of that in AI where we're seeing the stuff that we can do now. And we're going, yeah, yeah, yeah, it's not there yet, but it will be in a year, in five years, in 10 years. But I heard that like two years ago, I heard two years ago that we would have already had that by now. I don't see it yet. And I feel like there's a lot of us projecting this future that may or may not happen, at least from my perspective as a, as a power user.

Maciej Nowak [00:22:17]:
Don't you have a feeling that this is a problem of over promising and over expecting on the other side of the barricade, let's say, because we hear that the drugs will be developed by AI this year and so on and so forth on the, on the, and this is on the supply side and on, let's say demand side. So the users, they are expecting this to happen and this is not yet happening. But what you said is that you can Build your own tools without coding knowledge. You have high level knowledge, but you were never able to write a new, you know, everything without. No, from scratch, basically without these tools. And now look where we are. You are able to build, I don't know, on replit. If you switch the tool, you will be able to build an iOS application, native iOS application.

Maciej Nowak [00:23:18]:
You will be able to debug database queries. It's like, I don't know if this is a perfect metaphor, but it's like you are naked and, and it's freezing now. And okay, you will stitch a couple of furs together. This is not comfortable. This is not, you know, there is a hole. There is a hole, you know, but look what you, you are warm right now. Yeah. So in, in, in this sense, you have clothing.

Maciej Nowak [00:23:49]:
It's not perfect yet, but it's clothing. It will get you through the winter. You know what I mean? Saying it's not yet perfect, but it is making a huge difference on this first level.

Matt Medeiros [00:24:01]:
Yeah, yeah, I agree with that and I believe that, I believe that's, that's actually half the issue is that we're convincing ourselves that it's going to be, that it's going to exponentially improve at a rate where places like Fiverr and Shopify can put out these announcements like, we're going to cut everybody if you don't do this AI thing. It's like, well, really, are, are we moving that fast where you can, you can just ax 30% of your staff, like, what does that look like? And if you're just doing it for investors and profits, that's, that's one thing. But I think like a true business to, to operate. I don't see it. I don't see the. Everything that's needed for a business to run being solved by AI as quickly as we're predicting it to be. That's just my, my own opinion. Yeah, I can do a lot more now.

Matt Medeiros [00:24:59]:
Sure. And yes, this happens a lot faster. You, because you could make the same argument about WordPress. Like when WordPress first came out, let's say version, let's say version 2.5 and 3, when you were like, wow, I can build a whole website with a database and a user backend. All I have to do is install this thing PHP MYSQL on a web server. And you were like, I can build a whole website. Which, which then turned into the, the conversation back then was, is this a whole. I can build a whole Web application with WordPress.

Matt Medeiros [00:25:32]:
And that whole debate started to happen. And then page builders came and page builders started to expand. Then we started to see no code tools. So I feel like we're always like evolving towards like that, that I don't have a better phrase, like that easier future. Like look how easy I can make things now. And we've seen this over the course of 20 years with WordPress. AI has accelerated or compounded that in a few years. Now I can build a web app.

Matt Medeiros [00:26:00]:
Yes, it's going to get better, but I don't know if it's going to be perfect to just, you know, run a business at 50% to where we're at today or some people saying like you could just run it with one person and have multimillion dollar business.

Maciej Nowak [00:26:18]:
Yeah, first billion dollar business operated by one person.

Matt Medeiros [00:26:21]:
I mean, yes, there's, there's going to be outliers, I get it, you know, there's going to be outliers. But I don't see, I don't see it happening end to end as fast as we got here. Right. Replit hasn't introduced by this point. I would, had hoped like replit would introduce an agent to continuously update my apps if I'm building apps. Because that's one of the arguments I've heard. Oh, now all we're going to do is create apps that people are just going to, it's just going to be, you know, a bunch of crapware. Right.

Matt Medeiros [00:26:57]:
It's just going to sit out there and consume resources and run and it's. Nobody's ever going to do anything with it. Right. Abandoned apps or something like that. Well, I would expect Replit to have an agent that continuously updates it for me. Like why would I choose WordPress over building my own content management system with AI? Because WordPress is maintained by, well at this point maybe hundreds of people. Hundreds of people instead of thousands because hundreds of people are paying attention to this piece of software. If I built a WordPress alternative with Replit, who's going to maintain it? Me maybe.

Matt Medeiros [00:27:38]:
Right, but do I want me to be that weakest link? Do I want a replit agent to be the weakest link? So I think we're still very far from like a sustainable software future with AI. That's my cautious side speaking. Yes, I like what I'm able to do, but I'm also cautious about it at the same time.

Maciej Nowak [00:27:59]:
But don't you think that the picture we see is extremely incomplete because we are looking at it and this conversation centers around non technical people being enabled to do something they were never like meant to do you know what I mean? There was zero chance that the path would be very long and very painful to get to understand programming on a level that would allow you to build a web application with database. And like my feeling is that big organizations don't brag about this because they don't have to. They are pushing their products forward, they are doing much more stuff quicker and they don't create like a kilometer long threads on Twitter. How you know, they are like imagine 1000, you know, strong software SaaS platform, right. And half of the team would be bragging about how they set up their cursor rules to be faster. I can't see this yet. Like I have a feeling that to some extent this is happening. And this is what was the conversation about at press conf.

Maciej Nowak [00:29:08]:
How we as you know this or that agency can make that switch for I know 20, 50, 100 developers in the organization to use certain suddenly use cushion instead of PHP storm. So I think that the problem is the agility and that winningness and that push from that leadership you mentioned earlier. Like how are you going to motivate your employees? Except like that's the ultimate motivator. Except like you know, rising salary or eternal fame. Like, but what works the most is the insecurity of your employment. Like if you are not utterly and extremely motivated to get out of your comfort zone and try new things, it has to be increasing motivation to break like status quo doing all stuff for a long time. No, like mode of operation is to talk about what's happening in the frontier of development in these tools because they will be read. There will be no views on Twitter.

Matt Medeiros [00:30:15]:
Right, right, right.

Maciej Nowak [00:30:16]:
So it's in their interest. On the other hand, if you're in that bubble, you get the fomo, right? Like how fast is this? Like what am I missing with that new tool?

Matt Medeiros [00:30:25]:
Yeah, and you know, that's one of the things you hit it right on the, right on the head with that because you know that those kinds of things affirm. I feel like we're at like, whoa, are we, we're in this bubble right now. Like how many times do we have to see these bubbles come where we're like, oh, this is all just first mover stuff right now. These are things that are happening because it's just very early stages and everyone's trying to adopt this technology. When you look at the competitors, here's my other sort of like rant about this on the, on the AI side is when you look at Google ChatGPT Microsoft, Claude, Deepseek, these AI platforms are all competing against themselves or competing with one another, that they're all going to start to have the same solution. It's going to have all the same outcome. Like we'll be able to turn and do to Claude and do the same thing that we could do with ChatGPT. There'll be small percentages of differences.

Matt Medeiros [00:31:41]:
And to me, like, when that level playing field comes now, now everybody just has the same thing. Like this is, this becomes less innovative and it just goes back to the operating system wars of whatever 30 years ago at Windows versus Mac, right? And nowadays you're just like, whatever, like it's the same thing. Like, it helps me run my thing. It's nice now, it's fine. I did an interview on my, on the WP minute with this guy named John Doherty. He runs a productized service, you could call it an agency called Editor Ninja. And Editor Ninja helps people edit their content. He started it like right around the time ChatGPT launched for the consumer.

Matt Medeiros [00:32:25]:
Right. And basically where people could start writing content with ChatGPT. Right. A few years ago. So now he has clients come to him that say, we created 400 articles with ChatGPT, now we need you to edit them. So that's great for his business, but I think to myself, like, wait a minute, what, what did we do here? All you replace the human writer with ChatGPT to write you 400 articles, but now you're asking a human to edit the articles. So, so we took one shovel and we, you know, we replaced it with a pickaxe. Like now you're, you were digging a hole with a shovel, now you're going to use a pic.

Matt Medeiros [00:33:09]:
I don't understand. So I think that, you know, when I hear those kinds of stories and, and I just watch these trends adjust. I don't know, I'm, I'm, I'm less excited about it in the short term, but certainly like excited about it in the long term, if that makes any sense. Like, I don't see the humans being replaced as fast as I see like these CEO letters come, which, you know, I think we've, we've made that point already.

Maciej Nowak [00:33:37]:
Yeah, but, but I think if you are not in the trend, you don't know what will. Sure, what will keep with us because you have to try out things and then see what is working, what, what is bringing the, the, the results. And with that replacing AI slope, you know, we, we will enter a phase where AI will learn on AI and it will Be you know like a recursion, recursive pattern and that's why it has to be broken. But yesterday, interestingly I read on Jason Friedz of Jason Freed of 37 Signals, you know the, the base Camp maker makers that they, they were recently recruiting a junior developer for the first time since ChatGPT 3.5. They applied like they run the answers of that coding challenge through AI and obviously they should, the candidates should be using AI to, to, to accelerate that, that thing. But they were able, you know, 37 signals were able to like say which AI tool was used for that code base. Right. And so the same thing as you have with writing.

Maciej Nowak [00:34:58]:
If you spend enough time with these models, you will see the style of the given model. Jason has written that for the application letters and you know that cvs, it's like AI writes in brown color because it aggregates all of the colors. And if you color like mix a lot of different colors, you, you maybe don't get black, but you will get brown. And there is no distinctive style. It's like, and on top of that you can only build into a given direction with your expertise, your insights and so on so that that base will use us like a, like a, something like this where you will be able to use that to get farther but without a human being it won't, you won't succeed.

Matt Medeiros [00:35:55]:
Yeah, yeah, yeah, no, it's interesting. So let me ask you this question. Do you know from the Jason Fried thing, did he say that do they care if these applicants are using AI to code?

Maciej Nowak [00:36:08]:
They are expecting them to use AI to code the expectations that it's dumb thing not to use AI. The case is no, it can be slope, right? It can be just hit enter and four letters. That's not a great thing because you have to. They were very quickly able to say if the letter was written by AI and that's not a good thing. The letter should be made by a person. But the code can be enhanced with the AI coding model. Whatever.

Matt Medeiros [00:36:48]:
Yeah, I mean it's a curious thing again to think about it in the world of agency life and let's say building WordPress websites. Years ago when I started my podcast I was always questioning the, the $500 website, right? The person who would just go to ThemeForest as an, as an air quotes agency owner, they'd get a customer. And we all probably have these stories, right? Where especially back then somebody would go buy a theme from ThemeForest where it had like all the functionality when like went into it like business directories. Like it had all this stuff in a theme. You could buy it for $79 and then the agency owner with air quotes again would sell that for $500 to. Then they'd leave because they're out of business. Right? Like the, the agency, like I was like, all right, I made 500 bucks, I'm done. And then you'd get that customer knock on your door and go, hey, I get this guy who built me this website for 500 bucks and it's broken.

Matt Medeiros [00:37:46]:
Can you fix it for me? And then you log into it and you're like, what is this slop? Right? It's just a different kind of slop that we're getting these days. And I'm wondering, I'm wondering how AI changed. I don't have an answer for it, but I'm wondering as like AI, you know, proliferates the industry, if permeates the industry. I'm wondering if customers don't care anymore. Right? Just build it for me on AI, you know, I don't care what the code looks like. But do we settle?

Maciej Nowak [00:38:17]:
We get questions like, and these questions we started to get half a year ago. It's not a, it's not like initiated by you know, last development and especially with our clients like with clients like with projects with our clients who are driven by IT teams. So that, you know, the project can be initiated by marketing or by IT mainly or maybe sales, but it's either marketing or it. And with these IT driven in big organizations projects, we got these questions like, okay, there are some studio, like are you using Copilot, you know, what are tools that you are using, you know, what are your developers are using, like, and so on. And for example, another project we would go to Bridge with a like clickable demo wipe code that include. So it was like not even a cursor on anything like this, but you were able to very quickly mock up something or a data model, for example, so you don't have to like, these are the use cases where you don't have to have a full application, but that development component can be used to very quickly, quickly mock up an API response, for example, or part of the ui, not the hop, but only like a pop up for example. And like how fast can be done. Extremely.

Maciej Nowak [00:39:41]:
Yeah, yeah, yeah. So we are getting these questions too. And, and, and there is, there is the expectation that we are the guardians of code quality, but please have a bigger shovel, you know.

Matt Medeiros [00:39:57]:
Yeah, yeah, I mean, I don't, yeah, I'm building a pretty Intense, like marketing application for the team at Gravity Forms right now. And my biggest concern is like, oh yeah, I built this thing. You know, my replit bill for building this thing besides my time was a few hundred dollars, right? For replit for this one application for like all the agent requests, they charge you after a certain amount. If you never use replit, they charge.

Maciej Nowak [00:40:31]:
You a few.

Matt Medeiros [00:40:33]:
Cents per request after you do it. So I've been doing a lot of requests to add up to a few hundred dollars and it's really powerful. But I don't know what happens next. Like, okay, team, I've built it, everyone's excited about it. But when I launch this thing, if you have, if it's, if it starts to have performance issues or it's crashing because too many people are logged in, I don't know how to fix that. Can I turn to replit and say, you know, I'm starting to have issues with, you know, 12 logged in users trying to access this thing and every time 13th user comes in, the database crashes.

Maciej Nowak [00:41:12]:
Exactly.

Matt Medeiros [00:41:13]:
Is that an exactly. Is that enough to solve the problem? I have no idea.

Maciej Nowak [00:41:17]:
It will take you 95% there. Really?

Matt Medeiros [00:41:21]:
Like, yeah, yeah, but, but, but that's a safety net for, for like I think, I think that's a safety net for the real developers right now.

Maciej Nowak [00:41:30]:
I don't agree. I like yes, yeah, yes to like if the code base will swell to the size exceeding the context window and the, you know, you like there is a certain size where you start to collapse. Like maybe again bad metaphor. But you know, if your code base will grow out of the context window and that IDE will no, no longer handle that and you just basically can't code anymore any further. No, you, you naturally preacher level of incompetence, so you can't cross that level. Right, but on, on the other hand, yes, but like, yes and no. Like I agree and don't agree because obviously if you have zero understanding of coding and just, you know, hit enter. Yes, yes, yes, yes.

Maciej Nowak [00:42:24]:
It's like, you know, installing applications in, you know, on Windows back in the days and you will get very easy malware and whatever else, right? So I think yes, because the developers that are fully like, let's say well versed in the craft, nowhere to look for help and know how the concepts, the basic patterns work, like what can be expected. And they will, for example, remember that every now and then they will asking the prompt like, and now think about all of the attack vectors this form can get and so on. What are the VULNERABILITIES common vulnerabilities. And please act as a security researcher and protect this from hex. And then suddenly there will be validation, backend validation, you know, escaping characters and so on, which wouldn't be maybe the case in the first place because that Vibe coder who wouldn't know the patterns wouldn't do this. So. So to answering your question, in my opinion, yes, there is still an end love, you know, or some kind of a barrier. But look, there are so many cursor rules now.

Maciej Nowak [00:43:43]:
There are repositories of cursor rules. This is the basic understanding that like the knowledge how to wipe code is not really growing. And also, you know, for Vibe coders. So these people who don't know how to code but are building the applications, they are also learning they soon will not be wipe coders anymore. Like if they spent enough time doing that, so it naturally will be also maturing environment, getting better at what they do.

Matt Medeiros [00:44:14]:
Yeah. I'm curious on like the economy, is this maybe this is a. Here's a question for you and, and how you, you predict this stuff.

Maciej Nowak [00:44:23]:
Okay, I wasn't expecting that, but please go.

Matt Medeiros [00:44:28]:
So here's, here's what I'm curious about. If I look at like my replet bill, right, for this application, it's a couple hundred dollars, right. I'm wondering if these platforms. So I have like two questions in the back of my head. Like, are these companies profitable? Right. In the typical startup world, lovable is you think?

Maciej Nowak [00:44:48]:
I like, I don't know, but, you know, I don't know at how many dozen monthly, like millions of monthly recurring revenue.

Matt Medeiros [00:44:56]:
They are, but sure, yeah, so, so I'm number one. Are these companies profitable? Right. Because we've all seen it in the past where they've raised billions, hundreds of millions, billions of dollars, and then they just go out of business because they were never profitable, right? So can these AI companies be profitable, especially with such energy and compute demand? Like, can they stay profitable if they can stay profitable? Are the prices we're paying now like a couple hundred dollars to repl it? Is that enough for them or do they need to change the economies of this kind of hosting to go, oh, actually what you built is really worth 30, 50,000, $100,000 a year because you're replacing an engineer to build what, what you wanted. I'm wondering if the economies start to change, where they go now we're going to charge you instead of like hosting an agent requests, we're going to charge you on the value of the employee that you're replacing. You might be paying an engineer $200,000 a year to build you one app. Now you can build infinite apps. So we need to at least get $200,000 a year out of you. I'm wondering if we start to see economies change with these solutions that do it better best instead of just giving us slop.

Matt Medeiros [00:46:14]:
Right. Instead of just giving us slop code. I wonder if there's an environment that gets built where it does think holistically and not just keep prompting away. Vibe coder. Now we've got you a team of engineers in this solution. I'm wondering if you, if you think the economy's change.

Maciej Nowak [00:46:32]:
I think that you can't like maybe initially, you know, that's the incentive to even start. Figure out, figure out if this is worth trying because initially that no, if you have a very mature and big database, sorry code base, you start using cursor. No, it's not very easy probably to upload everything to replit and so on. So like for big projects I don't think the gains are such huge. But for starting, yes. I think that the incentive to start a project, middle sized project, you know. You know, you mean yearly salary for developer but you know you have many cases where the app would cost I don't know 20k or $15,000 and that can be, you know, that's the size of the project that the agency would charge you 20k but that can be with obviously no problems like coded in a month by, by the. I don't know, someone.

Maciej Nowak [00:47:35]:
Not. Certainly not over the weekend. Right. I don't believe these over the weekend Vibe coded apps that was, that were charged 20k but with the argument of okay, I'm replacing the developer. That's the Devin. Devin's promise. Right that you are. You.

Maciej Nowak [00:47:52]:
You get a junior programmer for $500 a month. So this was the hook that Devin used. I don't know. I haven't used Devin. We didn't use Devin. I don't know there was some controversy about you know over promising and failed demo on that front. I don't buy this, this thing that. Okay.

Maciej Nowak [00:48:15]:
Because we are replacing a developer you will get nowhere or you will have to have pay so much attention to where that agent is going and you have to see that it's you know, choosing wrong path for something. So. So you have to have that knowledge.

Matt Medeiros [00:48:33]:
Yeah. Yeah. I wonder if we. I was thinking about this like for like that developer who's like oh God, if you vibe code like we don't Want any of that here, right? That the, the typical developer who's maybe not adopting AI and like, don't they, they don't want to lose their job and they're like, ah, if you're a five coder, you know, there's, there's no room for you here. But I think in terms of software, humans have always wanted to just say something and have it happen, like build me this thing. They want to just talk in regular language and have the thing, whatever it is that they're looking for. I'm curious if we move to a world where let's say we get to a point where, you know, in a few years time, a few years time in five years where you can just naturally talk to ChatGPT, it builds you a perfect app that you want for no problems at all, no bugs, exactly what you wanted. Do we get to a point where these AI systems don't even use the, the coding language that we, that we know these days? Php, HTML? Does it even matter anymore? Like if these AI platforms just come up with their own coding language to be more efficient for their LLM kindly.

Maciej Nowak [00:49:50]:
That's very interesting.

Matt Medeiros [00:49:50]:
Do we even have that anymore?

Maciej Nowak [00:49:52]:
That's a very interesting thing to think about. That they may, that there may be developed a coding language that would be super pro performant on the, on the bare metal box level. Because it wouldn't be. And the readability can be dropped because it would be read by a person but would be maintained in a like a syntax and how is it called? Like memory, presentation and everything by a machine, which is so much better at understanding that. But to know this, we would have to record the second episode because we have to know stuff very abruptly.

Matt Medeiros [00:50:38]:
Yes. Yes. Well, I appreciate the time.

Maciej Nowak [00:50:41]:
It was pleasure to meet you, Edward. Sorry, it was pleasure to have you on the pod and I hope we will have a sequel to this episode. Maybe not five years time, but maybe let's do a health check in a year's time to see what, what happened since this conversation.

Matt Medeiros [00:51:01]:
Yes, absolutely. I'd love to.

Maciej Nowak [00:51:03]:
Perfect. Let's do this then. Take care much.

Lector [00:51:07]:
If you like what you've just heard, don't forget to subscribe for more episodes. On the other hand, if you've got a question we haven't answered yet, feel free to reach out to us directly. Just go to osomstudio.com contact thanks for listening and see you in the next episode of the Osom to Know podcast.