Are Matt’s AI Skills Making WordPress Developers Obsolete? ft. Matt Medeiros

Brian (00:02)
Mamaderos, think we are live. I’m going to double check.
Matt (00:06)
Brian, it’s good
that we’re live. It’s good. It’s good.
Brian (00:09)
Yeah, when
I invited you to do this, I didn’t until the last minute say, hey, do you want to actually just do this live? And you said you were down. So I appreciate your willingness.
Matt (00:15)
Mm. Yeah.
Anytime that I can come on your channel and sabotage it, sign me up. Sign me up. That’s what I’m here for.
Brian (00:29)
Maybe so people can understand, you could give us a little bit of a background who you are real quick and why people should know who you are.
Matt (00:38)
Matt Medeiros full time job at Gravity Forms as WordPress community and content and spending a lot of time there creating stuff for their YouTube channel their Learn site which you have been featured on in our podcast breakdown. And you can also find some of the side work that I do at the WP minute at the WP minute dot com WP minute plus podcast have a couple of things I actually haven’t announced yet which we could talk about breaking news on your own channel. But yeah spend some time there.
analyzing and commenting on the WordPress space, which led me to appearing here on your show.
Brian (01:16)
Yeah. And we’ve, we’ve chatted a lot. We’ve had kind of a few different conversations about a few different things. feel like we always talk about the meta big picture stuff more than we talk about the nuts and bolts of, of WordPress. But this one is spurred by your recent video. I make sure I get the name right. It’s a, I made an AI app and WordPress is doomed. Am I, am I pretty close?
Matt (01:40)
Yeah, it’s I made an app and I’m afraid for WordPress now.
Brian (01:49)
Yeah, so give me the high level overview of why you made an app and you’re worried about WordPress.
Matt (01:55)
Yeah. So. Obviously this is big topic. This is what we’re to break down today and chat about. Right. So the the the big thing is. Look, I mean, this is a conversation that we all have a lot on these WordPress podcasts is what’s the future of WordPress, what’s happening with WordPress and especially in today’s age with what’s happening with WordPress. One must always be slightly ahead of the game. And. You know, if you were.
not constantly reinvesting like 20 percent sort of like outlook. I use that number roughly but if you’re sort of not like looking ahead 20 percent of where your business is going it doesn’t matter if it’s just WordPress but if you’re an agency or freelance or whatever you’re constantly like looking ahead a little bit doing the work that you have to do to survive which for me is creating WordPress content and commenting on WordPress but then also looking ahead of like.
Brian (02:39)
Mm-hmm.
Matt (02:52)
Well, what else is out there that could be disrupting this thing? What’s really going to change this thing? And of course, you know, AI is something that is in everybody’s dialogue these days. And I had to stay ahead of the game. really sparked. mean, I’ve you and others have been telling me about it. Mark Zemanski, right? You know, he’s always there about like, hey, this AI thing is taking over. But it was really a conversation with my brother and I and.
Brian (03:13)
Mm-hmm.
Matt (03:19)
between my brother and I, and we were just chatting one day and he’s like, hey, have you done anything with it? And I’m like, no, really, just aside from like the content stuff. And he’s like, I’m telling you, man, it’s gonna things are gonna get weird. He’s on the finance side of content. So he was telling me about it. I said, fine, let me just let me go a little bit deeper based on what he was telling me. And that’s when I went sort of all in learning how to code with AI and you maybe saw me starting to tweet or.
blue sky about things like that. And yeah, it sort of set me down this path of how to do this. Like, where do I, how can people do this? How can an absolute noob like myself do this? And what are the best routes to do? And it took a lot of failures to figure out the best route, but maybe we can unpack that as we go along.
Brian (04:09)
So what did you build first of all? Because I go back and forth. I use AI constantly. I use it to write code. I use it to generate all sorts of stuff. I’m very aggressively bullish on it changing things and I mostly agree with you. But then I see a lot of people sharing stuff and they’re like, look at me, I built this cool new thing and it’s…
You know, when Arub and I had a podcast, we’ve joked that everything was a to-do list because we kept coming up with ideas of things to build and show off. And then we’d always be like, this is just a to-do list. That’s all, you know, it’s a to-do list or it’s a database in a form. That’s kind of what most of the web is. So a lot of the things you see are, are the one hand kind of insane that people are building these things from scratch with no coding experience. And it blows your mind. On the other hand, a lot of what’s being built is, you know, not like.
game changing, it’s not going to be the infrastructure for your massive e-commerce store or something like that. So it’s like, keep going back and forth on it. So like, what did you build and what was your, you know, what was your experience with the actual process of building as a technical person, you know?
Matt (05:12)
Yeah.
Yeah, I mean, look, all kidding aside, I’ve been doing tech like maybe a lot of people watching this show for 30 years, right? Or more. And so I have some solid understanding. And what I’m getting at is there’s also a lot of discourse around AI and people are people are also saying things like, this is this is just creating garbage code. It’s not like everlasting things. You still have to learn.
fundamentally understand this stuff and I completely agree with that. And because I am technical, it was a little bit easier for me to onboard into these things. And I have tried many apps, many ways to develop code. And when I first started with this little experiment back in mid-November, when I got pneumonia and I couldn’t move, I was just laying on my couch, I was building an iteration of what I’ve built with
the app I’ll get to in a moment using CloudFlare workers and like I have never used CloudFlare workers. I’ve used AWS before and I was like, let me just throw myself into infrastructure I have no idea about because I heard it was great. Yeah, like let’s I heard it was great and I heard you could like really optimize better and you know, run these these tasks at very optimal speed. you know, the point I’m getting at is I have some understanding of this stuff.
Brian (06:26)
Deep End.
Matt (06:40)
which probably leads to maybe a little bit more success when I’m using AI to prompt my way to an app. So what did I build? Currently, the first thing that I’ve actually shipped with any kind of logical success is a site called PulseWP.cc. It’s PulseWP.cc. At end of the day, it’s just a RSS aggregator.
built my way. It’s just a way for me to aggregate the news that I already follow in the WordPress space as somebody who commentates on WordPress. And it’s just our assessor reader. I have a list of sites. You can see the sites it’s pulling in. It shows it to you in a feed and it summarizes each article. It creates little social media posts per article, which is
what I need, right? I mean, I follow all the news, so if I can quickly scan something, see what it’s about, and then cue it up in my social, because I think it’s something cool to share. It’s lessening the workload for me. And then there’s a weekly summary tab up top, and the weekly summary is kind of cool, because there are lots of us in the WordPress space who do WordPress newsletters, and that’s generating.
a weekly summary based on the day that you’re looking at it. So you’re looking at it today. It’s looking at the last seven days of articles in my database and saying, here’s your weekly summary. So it summarizes it, writes a couple of sentences per link that it’s pulling in. And you can literally copy and paste that, put that in your Google doc, WordPress post, MailChimp newsletter, whatever, to help you start your own sort of aggregator of links. So PulseWP.cc, it is an RSS reader.
built my way for WordPress news, which goes into the, which is like the developing theory that anyone in the future, right, we can unpack some of like the comments and feedback that I got on my video, that anyone in the future will say, you know what, my to-do app, speaking of to-do apps, Todoist, love it. Maybe someday I’m like, you know what, I want my own to-do app and I just want it done this way. I don’t want all the features that Todoist gives me. I want it.
this way and I can just prompt probably the Apple App Store at some point to say, build me a, just make me this app that does this thing. And I don’t need to go look, nor do I have to go buy a particular app. I’m not saying that’s a good thing. I’m not saying I have the answers for that, but I can certainly see a world where app and software gets disrupted. People are already talking about it. It already has. And you know, here we are, you know,
Brian (08:58)
Mm-hmm.
Yeah.
Matt (09:22)
with my little experiment going, Matt built an RSS reader his way just for WordPress. And other people can too.
Brian (09:29)
So what, tends to be really easy is, is that prototype and that MVP. And then like where things get difficult is one, if you have a client and they’re like, no, I need it to do exactly this in this exact way. Or two, what if you want to expand it to where I can log in, I can make an account, I can pay you a buck a month. I can curate the feeds. I can blah, blah. So like, you know, that extra layer of making it more dynamic or something. And in the past 10 years ago, I could have imagined five years ago, I would spin up a new WordPress site.
I’d make new post types with ACF or taxonomies. I’d get WP all import pro, which lets you point it to RSS feeds and import them into it. And then I’d have a database. I’d have all the users stuff managed. I’d have WordPress under the hood. It pull in all my RSS feeds. It put them in a post type. It would do all that for me. Was there a part of you that was like, I could build this in WordPress and, like just the thoughts on whether you even would or wouldn’t because
It feels like it would have been a no-brainer in the past to spin up a little side project like this on WordPress.
Matt (10:33)
Yeah, 100%. So let me just
So I’ll take you through how I got to, I think it’s important to understand how I got to what I’ll call a stable MVP of this product. Because I think it’s just important to unpack that. But before I do, I’ve also, throughout this few months of experimentation, PulseWP is built on React and Tailwind for styling.
Brian (10:49)
Mm-hmm.
Matt (11:09)
of which I know nothing about. I mean, I know about it now and I can understand some core concepts, but had you asked me to make this a year ago, would have been impossible. I can’t just, I couldn’t just do it, right? And all of this has led me to a newfound appreciation for WordPress, you know, and software in general. Meaning, what has taken me, hmmm.
if I were actually timing it, maybe at this point, 15 to 25 hours of prompt time, like typing in things and debugging things. You know, there’s a newfound appreciation where it’s like, I could build this RSS reader that did this like one little tiny thing, but it took me, let’s call it 40 hours all in, you know, let’s call it 40 hours all in. You know, I could never do what WordPress does. So.
I think we’re actually going to kind of flip this conversation when people say, oh, people will be able to just build their own WordPress thing. Sure. Or people will be able just replace gravity forms with doing things in AI. Sure. But when you actually look at the time it’s going to take you to do that, you might appreciate the $250 it costs for gravity forms, you know, elite license with everything done for you already. You might go, oh, actually, that’s better value now. Now that I realize I can do it myself in like 30 hours prompting, I’ll just go
Brian (12:22)
Yeah.
Matt (12:33)
buy this thing for a few hundred bucks. So I think it’s important to like wrap our heads around that concept. Second, how did I get to this point? It started with, I hear people are coding with AI, how can I do that too? And I just started throwing things at Claude and chat GPT and saying, build me, it started with local Python scripts. I actually built a little RSS podcast grabber.
Brian (12:49)
Mm-hmm.
Matt (12:59)
And it just goes out and downloads episodes for me because what I needed was those MP3 files on my computer so that I could transcribe them and search for keywords as somebody who analyzes the industry. So I just started throwing all this stuff at Claude and chat GPT. And then I was like, well, maybe I’ll make a web app out of this stuff. And I started doing that, too. And I started throwing out code that was working and executing. And I was uploading it to Cloudflare. I was like, oh, my God, I’m building stuff.
Brian (13:10)
Mm-hmm.
Matt (13:29)
That works, this is awesome. And then as the MVP started to grow and more features, that’s when things started to crumble, not only in like, how do I organize this mentally, but to call out and chat GPT, I was just hitting token limits where they were like these, what I’ll call vanilla interfaces were just saying, hey, time’s up, you pay us 20 bucks a month, that’s cool. But come back, literally Claude would say things like, come back at 430 Eastern.
Brian (13:44)
Hmm.
Mm-hmm.
Matt (13:56)
where
you can continue your journey. I was like, this is cool. But maddening at the same time. And that started to get me to look at, eventually I found Cursor and I started building things manually through Cursor. And then what happened there was, hey, this is really powerful. This is really cool. It’s writing code for me. But now I’m here managing infrastructure. And when I started to build out a web app,
portion of this, I was stuck making this, like having to decide where to host this, where do you host a React app? Where do I, what database am I using? And most importantly, when it came to building any web app that I’d have anyone register and have a login with permissions, things we take for granted in the WordPress world, I started to realize, there’s this whole like,
Brian (14:51)
Mm-hmm.
Matt (14:55)
user authentication layer. And I started prompting for things like that, like create me a login and a dashboard and a control panel and user access and all this stuff. That really unraveled the whole other problem was, yeah, you think you’re just going to make that yourself? Yeah, nice try. know, AI will. And this was one of the issues with AI was AI was just saying, here you go. Here it is. Here’s the login page. Here’s the user dashboard. you want password management? Here you go.
Brian (15:14)
Yeah.
Matt (15:25)
you know, you want to hash your passwords for protection? Well, you didn’t ask me to do that. I guess I’ll do it for you now. And that quickly made me quickly realize like, there’s platforms out there like Superbase. Amazon has another one. I forget the name of it. The replet, which is what I’ve built Pulse WP on, has its own user management. So that’s a thing. User management, user authentication. That’s a thing. That’s a service which I quickly realized like, these other
coding platforms that are out there are building this stuff in for you because they know how important but also how difficult that is for AI to do or anyone to just like manage themselves. So once I got past cursor and like trying to figure out how to manually do all this stuff, I then found two apps that I still use in kind of in tandem, Bolt, which you can find at bolt.new.
Brian (16:11)
Mm-hmm.
Matt (16:23)
And then they also have an open source version of that which you can run locally which is super powerful. And then a Replet which is what I’ll call the Elementor of AI coding because I see all of this stuff as a page builder for any software these days.
Brian (16:34)
Yeah.
Yeah, there’s a lot there. I had like three things I was going to throw in the first out to respond to your first point. I think you said it takes four, you you took 40 hours and then it, you’re kind of like, could probably paid for a 200 piece, 200, all the pieces of software. And I think that’s the one place where WordPress has always had the advantage, which was people with more time than money. You know, like that’s the target market of the WordPress builder. I have a lot of time. I don’t want to spend a lot of money. And that’s the part that I’m more concerned about that.
Matt (16:44)
Yeah. Sorry.
Brian (17:11)
I’m concerned about, you know, it eating away at that. The other piece is like right now, I wouldn’t recommend anybody open cursor if they’re not a developer. I think it’s a developer tool and it’s not the place to start, but I can understand like you, like you went most advanced and then you ended on the element or as you call it of AI, replet. And so it’s just kind of a wild journey, but like, we’re going to hit a point where people, the workflows are kind of smoothed.
the prompts are probably finessed to where if you ask for a login page, they’re going to probably update the prompt first that says, and make it secure. you know, like they’re going to find a lot of those gaps and polish them off. but it seems like the, the developer, the modern developer experience, especially in react and like react and JavaScript, which a lot of these tools like to use. I haven’t seen one that wants to build you a Laravel app. They always want to build you a JavaScript react app and
Matt (18:06)
Yeah.
Brian (18:07)
The JavaScript world is so much more like glue holding things together. Like is Replet even doing your users and your database or do you have like SuperBase or something else glued in?
Matt (18:17)
Yeah, so
there’s something else I want to address before we go into the technical stuff. And this is the more like emotional side of this stuff. And that is I’ve always wanted to be a developer, just don’t have the mental capacity to do it. Right. I’m not smart enough, Brian is what I’m saying. And it goes back to my days of learning how to code in basic in Pascal back in high school.
Brian (18:37)
believe that.
Matt (18:46)
You know, doing things in MS-DOS before that. And I’ve always wanted this, right? I’ve always wanted to build things. you know, once I learned about the web and open source stuff, like going back to Linux, PHP, Nuke, and putting all these things together, I was like, wow, this is really cool stuff. And then finding Drupal and being like, oh, I can build stuff with Drupal. It’s the first time I ever felt like a developer with Drupal, CCKM, Views, fast forward WordPress, you’re stitching all these plugins together and voila, like you’re building cool stuff.
That was, you once I started to do this, this was like, man, this felt, it felt like a whole new superpower was unlocked. And I’m excited for more people to get to feel that too. I know a lot of people are trying to discount it and say, look, you should be cautious and be careful with what you do with this stuff. But there’s an emotional level there with like empowering people to be able to build this stuff without barriers, which is really cool. And second, it.
was an educational experience. mean, so long as you’re going into this knowing you know nothing and you’re willing to like learn, right? So when I was doing like the, when I was trying to build this stuff and just chat GPT and Claude, half the reason why I was running out of tokens was I was like, can you explain this to me? I don’t understand it. And it would just spend all that time. Now, Replit and Bolt and stuff like that, they’re actively telling you how this is being built. And I constantly question it. So I know what I’m.
Brian (19:48)
Mm-hmm.
Matt (20:12)
what I’m doing here. So it has helped me understand the world outside of WordPress, especially JavaScript frameworks and all this stuff. Like you think it’s bad in WordPress. It’s like a zillion times worse in that world than what we experience. So anyway, going back to the technical side, I ended up choosing Replet because it has everything in house. I have some notes.
Brian (20:19)
Mm-hmm.
Matt (20:42)
which I’ll pull up, but Replet is certainly something that if you’re looking for the database, the user authentication, and the hosting, it’s all done inside of Replet, which those of us who are technical kind of know, well, okay, now we’re just locked into one environment to a degree. Like, you know, I can extract this code, I deploy this code automatically to a GitHub repo. So yes, I have the code, but really the sticking point, which I’m…
keeping a close eye across this whole industry on is user authentication because I don’t know how portable that is. Like I’m not ready for it anyway but like when I’m there like if I said, or you know, I’m gonna move to another platform. I wanna go to Superbase let’s say. I don’t know how portable like my user auth stuff in Replit is gonna be for Superbase. It’s built in Postgres database so yeah, I could probably extract that fairly painless and then of course my code is all just.
Brian (21:17)
E. Yeah.
Matt (21:38)
freaking bespoke JavaScript that it wrote for me. And I’m sure it’ll take it to the next environment with some headache. But I chose suva base because it was all encompassing. The Bolt, just a quick technical comparison, Bolt was really, really good at user interface. Replet sucks. Right? Like I don’t think I look at PulseWP.cc and be like, that’s a great looking interface. It’s fine. It’s OK.
Brian (21:40)
Yeah.
Mm-hmm. Yeah.
Matt (22:07)
But even when I prompt it to do better interface, it doesn’t do it. Bolt on the other hand, for whatever reason, whatever they do with Tailwind, however they’ve trained it, it does a lot better job at user interface, in my opinion, in these short few months that I’ve been experimenting. And yeah, so that’s largely why I went with these platforms, or I’ve put these two platforms head to head.
So that’s, to answer your question, that’s why I chose Replet. All in one place and it works for me.
Brian (22:41)
Okay.
Yeah. And it’s, it’s kind of wild to me because when I think of all the things you need for an application versus like a website, that’s, think the, a lot of people in WordPress development, it used to be like, I’m to put everything on WordPress. I’m to build my membership site and my LMS and my this and my, that and my dashboard and my portal and everything. That was kind of the way WordPress, felt like kind of grew about this popular, you can do anything with that mentality. And then we hit a point where it just kind of,
It just kind of got to a place where building applications with other tools made a lot more sense and WordPress started to strain at trying to do too many things with it. And your membership plugin and your e-commerce plugin didn’t always work together too well. And it was just, you know, everything was getting bloated and slow, but it really does feel like if, if WordPress democratized publishing in terms of like getting content on the internet and getting like websites, content, marketing sites, that sort of stuff.
And now we’re in the place where everybody wants applications. They want their own application. They don’t just want to build a website. They want to build an application, interactive data, database users, all that sort of stuff. It feels like WordPress could still really solve that need because it does have, like you said, everything you need. has the database, it has the user management, it has the foundation for storing data. It has a foundation for building a front end and markup and design tools and themes and all that sort of stuff.
It’s like it has all the component pieces, but is the difference just the fact that you can open up Replet and say like, I’m dreaming this thing, make it. I’m not going to click on anything. I’m just going to tell it and just let it do it.
Matt (24:20)
Yeah.
Yeah, I would say that, yes, I mean, it’s that ability to do it, which is, you know, in my video, the title of like, now I’m afraid for WordPress, you know, it’s largely that because what I’m constantly doing with WordPress is, again, under the context of where we’re at in the WordPress community, the competitors, right? You can’t simply just keep rising to the top.
How many times have humans learned that lesson across every era of life whether it’s tech or humanity civilization right. So what is going to dethrone WordPress or at least what is going to take away the 10 percent off of our of our market share. Do I think WordPress is going to continue to a 55 60 percent. I don’t think so. Not anymore. And so what’s going to claim that next 10 percent off the top.
as WordPress realigns its core values. it’s gonna be competitive CMSs, be a 1%, 2 % of other great CMSs that are out there doing some cool things. But I think as this ability to just make anything with AI starts to hit the consumer market, or at least to more implementers like myself who are servicing that market, you’ll start to see people be like, you know what?
to heck with it, I would have used WordPress to do this, know, membership site or learning site or contact form. Now I’m gonna do it over here in this thing and I’ll just deploy this code to, you know, whatever web hosting I can do this on. And that’s gonna be, that’s the real threat, that attention, you know, that attention and that ability to just walk into it and be like, build me this thing.
And, some of the comments I got was like, there’s no way this is going to do this. Yeah, man. I was that same way a year ago. Right. I still kind of am, but like I see where it has gone in one year and in my head, it’s going to go another year is just going to be even more powerful. So what it doesn’t do for you is care for the software. Not now. Not, not now until there’s that, you know, sustainability agent.
Brian (26:26)
Mm-hmm.
Matt (26:48)
that just says, hey, would you like me to think of new features? And Replit does this for you already. So when you deploy something, it says, I can work on some other features. Here are some other features that you want to want to do. And I’ll constantly say, like, great. That’s a great feature idea. Don’t write the code. Tell me how you’re going to do it first, because I want to see like the impact it’s going to have across this app before it breaks things. That’s another thing we could talk about. But yeah, it’s this ability to just make whatever you want.
Brian (26:52)
Mm-hmm. Yeah.
Matt (27:18)
air quotes, with some time investment. So I’m thinking about like, what does that mean for my day job at Gravity Forms? This is just me speaking, not Gravity Forms and the rest of the team. This is not a strategy, this is just what’s in my head. And maybe in the future, software companies are the ones creating the agents to do things for their particular product. So you, you know, there’s a Gravity Forms agent. And maybe you can think of, just to put it in WordPress terms,
Brian (27:29)
Mm-hmm.
Matt (27:47)
Maybe you can think of the components of your app as blocks and patterns and you’re refining those and making those so good so that somebody can come in, use the Gravity Forms agent and build Gravity Forms using code and patterns that you’ve cared for to make it work better, faster, know, easier with the prompts. Those who have spent time building stuff with AI know like, yeah, it ain’t perfect.
And you’re constantly running into roadblocks and hallucinations and like just starts doing things and malforming stuff. But maybe the future is the software companies and the plug-in companies are making the agents and refining that. So yeah, you can just interface with a Gravity Forms bot and build Gravity Forms. And that’s what you’re paying us for theoretically.
Brian (28:23)
Mm-hmm.
Yeah, I mean, I think the one of the
things you said early on there is, that it’s not that great at maintenance. And I think this is like where the big disconnect is, is you see people saying it’s replacing software engineers and you see that from CEOs who love a good reason to purge their ranks every once in a while. And you see it from a lot of people who are selling AI stuff and everything like that. And then most developers that are in bigger systems are saying like, yeah, it’s helpful, but it’s not as helpful as you’d think it’s helpful.
Matt (28:57)
Yeah.
Brian (29:10)
for a small task, it’s helpful when you know exactly what you want and you just don’t want to write it, but it’s not really helpful for like building on legacy code. So if you look at WordPress, you would think by now in WordPress core, we would have seen some dramatic productivity increases in the contributions. You’d think we’d start seeing way more pull requests. You know, if you have an idea and you’re already a contributor to WordPress, it should be so much easier to use cursor.
to help you finish your pull request and get it over the line. But in WordPress in, and in most honestly, big systems there’s, and I put some links in a newsletter I sent out this morning to some of this stuff. There’s just not that different thing. So do you feel like, do you feel like when you’re going to hit a point with this thing you’re building where it’s just not going to get over that line or you’re not going to be able to keep going with it? Or do you think that’s just going to be solved?
Matt (30:09)
So I think WordPress needs to figure out a way to adapt and you know, when I put out this video, think Matt said, oh, see my, Matt Mullenweg said, see my speech from State of the Word 2023. And I went back to it and I watched it it was like, yeah, learn AI deeply and it showed off some cool things. know, so it feels, that feels like an eternity ago in WordPress years and in AI years, I guess, for that point. But yeah, there needs to be a way for WordPress
Brian (30:31)
Mm-hmm.
Matt (30:39)
Automatic and core of WordPress to to ride alongside of this and figure out how it how it survives Just from the just from the The the marketing thing which is something we’ve always struggled with but certainly now like we want people to be like, okay use AI But also build a WordPress app with it like how are you gonna solve that that that’s very important that people are going to AI to build WordPress Otherwise, I was just gonna be like I just give me whatever you want to give me. I don’t care
As long as it runs, right? But to answer your question about the sustainability stuff and how I see this stuff moving forward, again, another reason why I felt safer, I hope this answers your question, another reason why I felt safer using Replet was one, because yes, all the infrastructure was there. I was less worried. Like when I was doing stuff in Cursor, trying to build this other thing, deploying to GitHub, it would then automatically deploy to or build in,
in I’m going to forget the name of the hosting that I used but then the hosting would always fail on deployment so I kept going back to figure out like Netlify that’s what I was using I was using Netlify and SuperBase so now I’m managing two different areas of infrastructure so that safety net of Replet feels pretty good and they literally have rollback buttons for every like major thing that you that you build. I think it’s important to go slow and
Brian (31:46)
like a Netlify or something like that or okay. Yeah.
Matt (32:07)
don’t ask for too much in each prompt. know, ask it, I always ask. If I have a new concept, a new idea, I just say, explain it first before you write code. That way I know exactly what you potentially might go break that I have to like reverse engineer or debug. But then I’ll also, I’ll leverage, I’ll go back and I’ll leverage ChatGPT or Bolt. So if I have a new idea,
Brian (32:22)
Mm-hmm.
Mm-hmm.
Matt (32:35)
Like here’s something I deployed before we went live. And if you go to PulseWP.cc now you’ll see just placeholder images. I can’t for the life of me figure out how to pull in the damn image from the RSS feed and have it display inside of the app, right? I have tried to do this a gazillion times with Replet. And I know the RSS tag.
Brian (32:45)
Yeah.
Okay.
Matt (33:02)
and it just won’t do it. And what I’ve learned with AI is if it’s not getting it in like two or three prompts, it’s just never gonna get it. And you will really start to deteriorate your app by constantly telling it to do this shit, right? So no, so I go back to chat GPT. So what I’ll do is I’ll go to chat GPT and I’ll ask it.
Brian (33:19)
So you go to Stack Overflow, still learn.
Matt (33:26)
What’s the best way to and it’s not working for this particular thing, but it’s it’s it’s worked in other areas. Like if I don’t know something, I’ll go to chat GPT first and say, hey, I got this thing I’m trying to solve in this app. Help me solve it. And then it’ll help me solve it and figure out like and I’ll always ask like what’s the standard way? What’s the popular way? Have a nice conversation with it and then say now build me up. Now write me a prompt that I can bring over here and I’ll put that prompt inside replet asking it to, you know, check it first. Tell me what you’re going to do first.
Brian (33:49)
Mmm.
Matt (33:56)
So I’ll build my prompt in Replet. Because I like Bolt for user interfaces, sometimes I will just go to Bolt and say, make me this user interface for this idea and see if it spits out that interface. And then I’ll say, now write me a prompt for Tailwind, using Tailwind to do all this stuff. Then I’ll bring that prompt back into Replet. Because, yeah, for many reasons, but because I want to have those conversations outside. It’s like, you know,
I go to my design department, I go to my engineering department, I go to my IT department, I have these conversations, and then I bring it back to the product team. It’s very similar to that, I guess, to a degree. I’ve also noticed, technically, with Replet, that you have to complete conversations with Replet when you’re coding, because it just has memory, right? And if I’m making a new thing,
Brian (34:50)
yeah.
Matt (34:53)
and like I just deployed categories, right? So AI will categorize the articles based on the summary. If I don’t finish that, right? So it’ll say, hey, do you see the categories now? In your head, you might say, yes, I see the categories now. But you don’t answer that question to the replet agent. And then you just start going on to something else, like, okay, now let’s go build this other thing. The agent’s going, well, what about those categories?
Brian (35:00)
Hmm.
Matt (35:23)
Right. So what will happen is it’ll just stay there in memory and sometimes it’ll like make it again because you’ve never answered it. Right. So you’ll you’ll create another feature and it’ll just continue to work on that category thing because you never said it’s done. So like you have to have you have to like slow things down a lot in reflet which is good. It’s good. You know and you have to complete the feature and you have to break apart the feature if possible which I’ve learned as well.
Brian (35:29)
Yeah.
Matt (35:52)
I hope that answers your question. There’s a lot going on, but that’s how I’ve technically approached this inside the app.
Brian (36:00)
Yeah, there’s a lot in that one too. And a lot of these sort of like, was a kind of a phase where people were saying prompt engineering is like the next skill everyone needs to learn. need to understand how to talk to these AI. And the response to that was traditionally that like, we would hit a point where they would become so autonomous that they would be better at prompting themselves. Like they would be able to take your prompt and expand on it. And.
In some ways that’s kind of come true, but you’re basically saying like, you’re, you’re, you’re basically like a toddler and you’re, this is your, you’re trying to speak to an adult and you’re just like, please baby step me through what I’m supposed to do. I think I have an idea. Please tell me it’s a good one and tell me exactly how to do it. do you feel like that’s the main skill that you’re, growing is like that prompt engineering, that like conversational skill.
Matt (36:55)
Yeah. Yes. The answer is yes. I think it’s just being able to look at this stuff. You know, again, we’re all we’re talking about this now at the start of 2025 with AI being in its, you know, still its infancy. And I’m sure 10 years from now, it’s it’s in its infancy compared to 50 years from now. Right. So like we’re very, very early on. And these are the things you you have to do.
I can share my screen really quick to show you something that proves how I think about this. So I’m at PulseWP.cc right now, this is just the homepage, right? So what I’ve had to do is, right, so of course, I create, I have a dashboard, right? First, when I started this,
Brian (37:29)
Yeah, sure.
Matt (37:55)
I had, I didn’t have a dashboard to have the feeds. was just, I would just prompt it and say, add this feed, right? I didn’t need a dashboard. But obviously what happened was, oh, all of a sudden I need to rename the feed. Or what I was running into was feeds with errors. Sometimes an error would pop up. So I had to spend time like educating the feed parser to find errors and if there’s errors, report them to me, right? And then,
I used to have AI prompts in the code, right? So in the code would have the prompt that I generated and it would just shoot it out to my API and summarize the articles that way. And I was like, I don’t like what I’m getting back from that prompt. I don’t wanna be in the code every single time. So make me a settings dashboard that allows me to go in and create my own prompts and I can change these out.
at will. Same thing with this bulk categorization feature that I just made. now I store article content. I store all the article content, summaries and social posts in the database. I forget when I launched this thing, it’s been a few weeks. It’s got five hundred and thirty ish articles in the database. Right? So when I launched this categorization feature,
Brian (39:12)
Mm-hmm.
Matt (39:24)
it doesn’t know to go back and make those, to actually go and label all the articles. So I had to create a bulk generation feature that started creating the categories. And this is another thing I shipped before we went live. It’s not working for all articles, right? I don’t know why. So I started working on a logging system that would tell me why the categories are failing, right? And then another feature just to show you,
Brian (39:28)
Yeah.
You’re,
you’re living like my, my day job experience of, want this, somebody wants a feature and you have to explain to them, okay, well you don’t want that feature. You actually want 10 features because we need to log it. We need to go back. need to, you know, it’s not just a feature. It’s there’s so much more that goes into it and it’s. It’s yeah, that’s interesting.
Matt (39:57)
E-
Same thing when I started building out the app, I was just adding all these feeds in, right? So I know all the feeds, I’m just going and putting all these feeds in and nothing’s displaying. I mean, obviously, Matt, because the thing is processing all this information. It’s going through, it’s storing to the database, it’s sending to the OpenAI API, bringing that back and communicating with the system. So I had to build.
what I call the AI processing monitor. Right now there’s one processing. So I was able to see if I start adding a bunch of feeds and things are not appearing, I need to know why. So it’s gonna tell me I have X amount of articles in the queue, if there’s any errors, recent processing attempts. So this is just like a little monitoring dashboard that allows me to see things in a glimpse.
And then for each summary I have, you can only see this when you’re logged in. I can regenerate the summary. If I don’t like that summary, I can regenerate it. And same thing on the social posts, which I still want to fine tune. If I’m like, don’t really like that social media post, I can just regenerate that and it’ll give me a new one that I can copy and paste. So it is all of these, I guess you would call utility, utility things inside the app that I’ve had to build just to keep it.
just to help myself continue to develop the app alongside of doing it in Replet, if that makes sense.
Brian (41:51)
Yeah. And what’s, what’s interesting. I’m thinking about this now, cause I use feed Lee as my RSS reader and it has a bunch of features I like and a bunch of features I don’t, and I don’t even pay for it. I just have a free account. There’s some stuff I would want if to pay for, but I’m a little cheap. So then I’m looking at this and I’m thinking. You have replete that you’re paying for. You probably have some other thing you’re paying for and you bought the domain that you’re probably paying for and you’re doing all these API calls. I, I’m actually a separate question would be curious if you’ve tried using.
Matt (42:14)
Mm-hmm.
Brian (42:20)
a sonnet clod sonnet for the summaries. Cause I think it’s a better writer than chat GPT. But anyway, you’re putting all this stuff together. Is it, is it costing money? Is it causing a lot? Are you burning down a rainforest?
Matt (42:31)
yeah.
Yeah, I mean, when you think about, you know, when you think about AI at large, you know, the cost. But yes, it’s costing me money, right? So the replete base account is I think 25 bucks a month, you know, and you burn through the the base agent usage pretty quickly. You know, if you’re constantly like shipping new things like this and then you just pay per usage, but it’s kind of cool because they
you know, you can cap it at X amount of spend and it’ll tell you, which is cool. Like there’s some, there’s some safety in that. So whatever I cap it at 50 bucks or something like that. So I don’t go over, you know, bought the domain of course, but that’s all that replica cost is all with hosting too. I’m sure like if I was analyzing this from a business perspective, you’d be looking at that like, what are they going to charge me for hosting? And I’m going to probably spend more at replica because
Everything is there. So obviously you’re to pay more versus breaking it out. But the. Yeah, everything else is it’s still fairly low cost, right? For me to build this, to build this my way is the way that I see it is it’s fairly low cost. I don’t think there’s a necessary necessarily a business to be had with this, but I just think if I were looking at the microscopic business of WordPress media, there are many humans that already do.
email newsletters with a list of articles. It’s already a super competitive space to be somebody covering the news. We’ve talked about that a gazillion times and inevitably this is going to be replaced by your AI agent that’s going to be built into probably your operating system is where the war will be waged and it’s going to just summarize your content for you, right? Maybe I have an advantage because I am, you know,
I’m the human behind curating WordPress news and maybe I’m doing things slightly different that a bigger audience will want. But even the stuff that I’m building here with Pulse is just an example of what will be taken over by whatever way I believe operating systems or browsers are going to give you that information.
Brian (44:48)
Now, what if you could have this workflow? I’m just curious, thinking about WordPress, you log into your WordPress site and do you have the same exact thing? And I know people have sort of invented these a little bit. There is like an existing concept, but you could have a little chat guy in the side and you just say, Hey, you know what? I want a new page on my website. I want it to pull a bunch of feeds. I want to be able to manage the feeds with a dashboard screen, all that sort of stuff. Do you think you in
would want to log into the WP minute.com and build your own extra interface in there to do this, like build this whole thing in WordPress. If the tooling existed, if you felt like that workflow was, was like feasible, would you have done it inside of WordPress?
Matt (45:35)
Probably, yeah, I mean probably that yeah, the I mean I’m trying to think of like I’m also trying to answer this of like how I think somebody who doesn’t know, you know about WordPress or you know how somebody else would would would perceive that As somebody who is obviously a fan and has a lifeline connected to WordPress Yeah, I mean if I could do that it it would have been something that I would have done
I was also trying to get an exercise in learning how to do this without WordPress. So there was that. yeah, wouldn’t that had WordPress had that ability. I would definitely do stuff like that inside of WordPress for sure. In fact, it could be the you know, it would be a great framework for an app. The same conversations we had 15 years ago. It would be a great framework for an app for something like this.
Brian (46:28)
Yeah. And I’m only because you framed your video as WordPress is doomed. So I’m trying to ask, you know, devil’s advocate questions to push against the idea. I, know, I, my theory on this right now is that WordPress got really big because it did solve a lot of problems, but now we’re in a place where it needs to solve one problem really well, probably. And it seems like writers putting content on the internet is
Matt (46:34)
Yeah, yeah.
100%.
Mm-hmm.
Brian (46:56)
At the end of the day, the thing that it’s really still going to be very good at, it’s, you know, I, don’t, I understand that like AI is going to replace the email newsletters I get that are like the 30 popular links of WordPress this week, but like, I’m kind of okay with that. I’m not really, I don’t have a personal connection to that, but there are some people whose newsletter I get that right on WordPress blogs and no way is going to really replace that. Cause I just want to hear their opinion and their analysis. And it’s not.
anything that I don’t think I would just subscribe to an AI’s general analysis of things because I’m okay with that, you So I mean, if you were in charge of WordPress right now, you had the keys to the kingdom as someone once said, and you see this AI coming, you see that WordPress is losing areas where it has its competitive advantage, AI is coming, you’re not.
negative, you’re not pessimistic, you’re seeking opportunities and I’m prompting you, I’m doing a prompt for you. You’re a WordPress writer and you’re seeking positive opportunities. would be the first thing you would want to do to make sure that the next wave of AI stuff is written is open, is not in these proprietary platforms?
Matt (47:58)
Yeah, yeah.
Yeah
Yeah, this is a tough one, man. This is a tough one because. You know, the. This is something I’ve been watching Mullenweg just to dodge the question slightly at the beginning of this, like I’ve been watching Mullenweg like not talk about open source AI for the last couple of years, which was interesting to me, you know, at the YouTube meetup at work camp US before our worlds were kind of shattered. I had asked him like, you know,
I don’t hear you talking about it much. Like, what do you think the future is? You know, and he, you know, with everything else he had going on in his head that night, pre WP engine, he just said, like, I, you know, I think we’ll just try to work with all of them, you know, to see what the best, you know, the best bet is all of them being, I guess, all of the AI platforms, all the LLMs sort of saying like, yeah, we can’t compete financially with something like that. Yeah, so I don’t have.
I don’t certainly don’t have the answer. I would love to see some alignment with like the open source community at hugging face. If that’s even a thing like how do you get a WordPress LLM off the ground so that you can try to have the right way of coding for WordPress prepared to interface with. don’t have the solution is what I’m getting at. But how do you prepare somebody to
build WordPress using AI. What are the steps needed to do that before all is lost kind of thing? I don’t have the answer for that, but whatever that looks like, is it an LLM? Is it having something trained specifically for WordPress? Do you have to go and just train the LLMs that are out there and say like, this is the best way, is it better user documentation? I don’t have the answer for it, but it would be like.
How do we attract people to do that in WordPress as best as possible? Is that just, you I think the future of WordPress is the playground. Playground is probably where that is going to happen. You know, again, just going back to Matt’s talk in 2023, he did demo AI prompt building something in the playground. I don’t know where that technology is on the roadmap, but I would imagine like playground is going to be the starting point for WordPress.
Brian (50:33)
Mm-hmm.
Matt (50:42)
both commercially and open source, and that’s the place where it’s gonna be. I don’t have a great answer for you, that’s what I got.
Brian (50:52)
I wanted to quickly mention because I’ve gotten a lot of aggressive tech tweets about this, but I things people feel like things haven’t changed like that. Maybe AI has plateaued a little bit, but then in the last like month with the kind of quote open source, but not really model coming from China where they’ve, know, there’s been some tremendous gains there and, and energy consumption is going to go down.
push for clean energy would accompany this, I think will be a great thing net for the world. And then also these models, you know, they are getting better. And there was a sense that maybe they were going to plateau, but they actually are. And if your experience with AI is either chat, GPT, or like the Apple iPhone stuff, then that’s really not even the best, you know, stuff out there. That’s, that’s kind of, you know, they have, that’s, that’s mixed in with the
Matt (51:40)
Right. Yeah.
Brian (51:45)
the user interface. So you, they have that nice like product sense, but you know, one of the things that’s really happening or is stuff under the hood. so I like, I’m optimistic that there’s a place for WordPress to adopt this sort of stuff. I’m with you with playground. seems like we should be able to get to a place where I can open my site and playground and experiment generating code and seeing how it affects it and, and mess around with that and, and build off of that. But, it feels like.
WordPress itself is so fractured right now, can we get the momentum from the community and could everybody agree to even row in the same direction and stuff?
Matt (52:28)
Yeah, when I think about
When people are like, got WordPress job, not just WordPress, but any tech job being displaced. And then all jobs, right? People are like, okay, all jobs are gonna be displaced, right? Accounting, legal, like all this stuff. I have lot of threads of thought around this. I published something on Crafted by Matt this weekend about it. There’s, like everything else, it’s good and bad. We have no idea where this is gonna go. But here’s what I know for certain, is that.
Companies like OpenAI and Google and Microsoft and Facebook don’t want to lose and they want to make money. End of story. And I think that there is, in terms of the AI stuff, is they will be, like these big corporations going at it, however that is, both either just like competing with each other.
Brian (53:15)
Yeah.
Matt (53:30)
or having the US government and governments around the world lobby against one another to make sure one person’s blocked and the other one has access. I think that’ll be the only thing that stagnates that and energy. That’ll be the only thing that stagnates this whole AI thing from us all losing our jobs because they are so, you know, they are so profit driven that they can’t get out of their own way to like get this off the ground.
Brian (53:58)
Mm-hmm.
Matt (54:00)
for it to disrupt all markets. And, you know, how many lessons have we learned in the tech space of companies not making any money, not profit driven or not profitable at all, just going under. We’ve seen a wave of those over the last couple of years and we’ve seen it with investments like VC hasn’t been throwing money at anybody because if you’re not profitable, they’re not making the risk. And for so many years, you would just see these companies get tons and tons of money and just never hit profitability and just phase out.
You know, that is like on ridiculous steroids with AI right now. I mean, how much money is getting poured into AI not making any money? And what I’m getting at is, is I think for this consumer level, it’s going to get so expensive. Like when, when do they turn on profit making price points? Like what would, what would chat GPT have to charge now in order for them to make a buck on all the stuff that you do? It would be astronomical.
Brian (54:48)
Yeah.
Matt (54:57)
Right. People are like, my God, 200 bucks a month for chat CPT Pro. It’s like nothing. Right. They They’re just experimenting with let’s say let’s see what happens to the market if we add another zero. And in six months, they’ll just say we’re going to add another zero. And I think when it comes to like humanity, get your tin foil hats on everyone is that, know, this stuff will get really good. Right. I wrote this in my article. There’ll be a world where when the hardware aligns with the software.
and all the Ray-Ban Facebook goggles that everyone kind of makes fun of, when that stuff is implanted either into your eyeball contact or your Ray-Ban glasses, there is a level of ability for every human who can afford that. For every human who can afford the best AI tech in their glasses. Because what that’ll do is you’ll walk into a car dealership
Brian (55:33)
Mm-hmm.
Matt (55:56)
and look at the guy or gal selling you the car and go, give me your best price. And he or she will say $30,000 and your glasses are gonna say, nope, try again. And inside your glasses, you’re gonna see negotiation factors happening. And it’ll prompt you to say, I don’t think that’s the best price. I think you can go lower. And he’ll say, oh, 29.5. And it’ll all happen, all happen.
Brian (56:10)
Mm-hmm.
Yeah.
Matt (56:23)
And it can happen already because we carry around a supercomputer in our pocket already. So there’s that. This is going to thrust humanity into areas which are good and bad. Right. If I go to the doctor and the doctor starts saying this, that and the other thing, here’s this chart you’re doing this. I don’t know what you’re saying, man. Is this good or bad? I had this happen to me the other day. Like, oh, you know, go test this thing over here. I’m like, what? I even know what you just said to me. You know, and if I have. AI or something solving that for me in real time.
Brian (56:46)
Yeah.
Matt (56:53)
Well now I’m way more informed and this helps a whole swath of folks who aren’t educated in something. Legal documents contracts. Hey buy this cell phone sign this contract. Is this contract even safe for me. You you look at it in your AI bot will tell you now at what cost. No clue. And that’s where we’ll start to see the stuff crumble because at some point they’re going to say yeah you’re to to charge. We’re to to charge you for that.
And one day you might have an AI bill that’s more than your mortgage, right? If it’s doing all these things in your life for you, you know? I don’t know. That’s what I got.
Brian (57:30)
I go back and forth because I think they said the same thing when Microsoft Excel came out and I still pay an accountant to do my taxes. You know what I mean? Like I’m not, I’m not going to go out there and do everything myself. Uh, but you know, on the other hand, I, I, I do liken this more to the iPhone or the mobile internet in terms of like, it really did pervade society in a way that nobody ever expected. Um, the fact that an insane amount of technology was in your hand and in, and
Matt (57:37)
Sure, yeah.
Mm-hmm.
Brian (58:00)
the way the iPhone or just smartphones in general just change the structure of how we communicate and how we consume the news and how we consume other people and stuff. So I do think people should be prepared a little bit for that. But I also think, you know, at the end of the day, there’s, there’s always going to be a place for experts for the work to be done. I, know, I I don’t see it completely at eliminating all of our jobs, but I guess my closing question to you, since you brought it up earlier, the, the,
the Matt Mullenweg statement of two years ago, which was learn AI deeply. Is that your recommendation to people? Are you agreeing with Matt that it is time to learn AI deeply?
Matt (58:36)
Yeah.
Yeah, I mean, at least have some core understandings, you know, of it. It’s. Again, like a year ago, I I’m still skeptic again, like there’s the 80 and the 20, right? The 20 of me is like, hey, this is cool. And there’s I guess there’s a I have one hundred and forty percent meter, right? There’s another 20 percent out there that’s like, don’t know if I really trust this thing because I don’t know what these other companies are going to do. But, you know, I think there’s still I’m not a an alarmist. Like, I don’t think these things will end everything in
In 12 months, our jobs are still safe. I mean, it takes a while for humans to adapt this stuff. And there’s a cost to this, which is unrealized at this point. So it’s just not going to permeate the entire industry. know, but it is very important to just stay ahead because eventually you’re, you know, if you’re just a freelancer out there or an agency, eventually your customer, if they’re not already asking, they’re going to ask.
One day you might be faced with a customer saying, man, you’re charging me 300 bucks a month to maintain my site. Can I just do this over here? You should at least understand what your customers might be dipping into and looking at from a buzz worthiness and marketing side of things. I have some ideas for plugins and I still haven’t found like the best way to go about this. I know Nick Diego put out something recently for cursor and building plugins. So that’s a world that I want to explore.
Brian (1:00:00)
Mm-hmm.
Matt (1:00:06)
We won’t get into it because because we’re out of time. But I think there’s also a detriment of like APIs. Like what does this mean for APIs? Because I think if you know you can just build your own API connector I think this is going to be something where Zapier is looking less and less you know valuable to a degree. And you know maybe the open web is also challenged because of that because now people are like whoa whoa.
Brian (1:00:23)
Mm-hmm.
Matt (1:00:30)
It’s so easy for you to access APIs before you had to at least understand what the hell you were doing now. Anyone can access APIs. Whoa. Now you’re smashing this data together and putting out something out that we don’t really like that. We might start closing off API. So that’s like a whole other discussion that this stuff affects too. But anyway, learn it deeply to a degree. Spend some time and check out PulseWP.cc.
Brian (1:00:34)
Yeah, yeah, yeah.
Thanks, Matt. Can you also plug your CMS blog too?
Matt (1:00:56)
yes, so CMSminute.com if you’re into it. So again, like looking at new opportunities. I’ve had a great round of discussion so far with with a bunch of other folks who run CMS platforms like Ghost, Statomic, Craft. And that’ll be at CMSminute.com limited series brought to you by Ghost.
Brian (1:01:19)
Awesome. Thank you, Matt, for scaring the crap out of us. I appreciate it.
Matt (1:01:22)
Hey,
anytime, Brian.