The State of Vibe-Coding a WordPress Website

追加: 2026-05-15

[00:00:00]Everybody, if you're like me in the WordPress space, you are watching how AI is infiltrating everything, uh, and everyone, everywhere, and you're wondering when is it going to be a robust way to build out a whole entire WordPress website.

[00:00:18]We're wondering that same thing.

[00:00:20]It's come a long way.

[00:00:21]It's amazing what has happened in a short amount of time, uh, but there's still some hiccups.

[00:00:26]Uh, but today, we have a guest who has really keen insight.

[00:00:30]I'm excited to talk with you today about the state of vibe coding a WordPress website on WP Product Talk.

[00:00:39]This is WP Product Talk, the place where every week we bring you insights on product marketing, business management and growth, customer experience, product development, and more.

[00:00:53]It's your go-to podcast for WordPress product owners by WordPress product owners.

[00:00:58]And now, enjoy the show.

[00:01:07]Hey, everybody.

[00:01:08]I am Matt Cromwell, founder of Roots and Fruit.

[00:01:12]And I am Zack Katz, founder of Gravity Kit.

[00:01:16]And before we get just started today, if you benefit at all from WP Product Talk, there's a way to help us.

[00:01:22]Um, we would love your support.

[00:01:24]The easiest way to do that is to subscribe, and there's lots of places you could do that, like YouTube or Spotify or Apple Podcasts or all those things.

[00:01:33]Subscribe wherever you can.

[00:01:34]It's easy for you.

[00:01:35]It's free.

[00:01:36]It helps us out a ton.

[00:01:38]Yeah, and w- also, a huge thanks to Freemius for being our exclusive growth partner.

[00:01:43]Uh, do you not know about Freemius?

[00:01:44]'Cause you should, 'cause, uh, Freemius helps software makers grow, monetize, and succeed.

[00:01:49]Uh, Freemius is a merchant of record platform, which means they take care of everything around selling software, checkout, payments, subscriptions, licensing, updates, global tax compliance.

[00:02:00]You don't need to worry about it.

[00:02:01]They handle that.

[00:02:02]You can just focus on building amazing products.

[00:02:05]So thanks to Freemius for being our exclusive growth partner.

[00:02:09]All righty.

[00:02:09]Let's jump into the state of vibe coding a WordPress website.

[00:02:13]For this, we invited none other than Andy Peatling from Miles AI.

[00:02:20]Andy, welcome, welcome.

[00:02:22]Hey.

[00:02:22]Hey, guys.

[00:02:23]Hey, Matt.

[00:02:23]Hey, Zack.

[00:02:23]Thanks for having me.

[00:02:24]Absolutely.

[00:02:25]Thanks for coming on.

[00:02:26]Uh, tell the world a little bit about yourself.

[00:02:28]Who are you?

[00:02:28]Where are you from?

[00:02:29]What do you do?

[00:02:30]Why in the world did we invite you here?

[00:02:33]Uh, yeah.

[00:02:34]I'm, I'm, uh, I'm Andy.

[00:02:35]I'm from, uh, originally from the UK, hence the accent.

[00:02:38]Uh, live in Vancouver, Canada now.

[00:02:40]Um, and yeah, I've been, I've been in the WordPress space for about 17, 18 years.

[00:02:45]Um, started in the early days before plugins and themes and, and kind of got really into WordPress as a young guy.

[00:02:52]And, uh, I f- I originally founded something called BuddyPress, which was a social networking add-on for WordPress, and Yeah, that led to me joining Automattic, um, spending most of my career so far at Automattic as a product lead there.

[00:03:06]And, uh, in la- last June, just less than a year ago, um, I left Automattic to build something on my own in the AI space.

[00:03:14]Um, just kind of poking around with AI agents, seeing how that, how that kind of is, is moving so quickly and was just so excited and interested in it, in it.

[00:03:22]Um, I ended up, yeah, kind of fiddling around with agents, coming back to WordPress, 'cause that's sort of where my knowledge base is, and then, uh, that led to building something called Miles, which is a AI agent for WordPress, specifically aimed at, you know, building full, well-designed websites using WordPress, using blocks.

[00:03:42]Yeah, not using page builders.

[00:03:44]Um, and I think, I'm pretty sure that's why I'm here today, to talk about the state of that and the state of live coding and, and engineering with WordPress in the AI age, which is moving so quickly.

[00:03:53]Yeah, absolutely.

[00:03:54]Thanks so much.

[00:03:55]Hey, folks, this is a live show.

[00:03:57]Uh, we're doing this at the moment.

[00:03:58]Uh, we're live on LinkedIn right now.

[00:04:00]We're live on YouTube.

[00:04:02]Uh, if you are here watching right now, use those comments because we will answer them.

[00:04:07]We'll put them up on the screen, um, like this one right here that came in before we even went live.

[00:04:13]Nick Hamsy is in the comments already.

[00:04:15]He says, "Who would you say is your favorite Miles user?"

[00:04:19]Um, I'm a Miles user, by the way, uh, Nick, um- Oh, boy, that raises the stakes-... 'cause I think Nick had somebody in mind.

[00:04:26]He may well have done.

[00:04:28]I have to say, Nick is probably the most creative Miles user that I have observed myself.

[00:04:32]But, uh, what do you say, Andy?

[00:04:35]Uh, well, I mean, I ca- I, I, I'm afraid I can't look past Nick 'cause he's been so helpful over the last, you know, fi- four or five months, and just basically hammering Miles and finding every bug and every issue and, and explaining how he uses it, and- Yeah, the feedback from him has been fantastic, but I've also appreciated the feedback from you, Matt, so I'm not gonna leave you out on that.

[00:04:54]Nice.

[00:04:55]Yeah, Nick's been creating amazing stuff, uh, with Miles, without Miles, all over the place.

[00:04:59]Uh, we had Nick on the show, uh, a, a while back, and, um, he's a- Oh, cool... he's a great person too, so.

[00:05:04]Thanks, man.

[00:05:05]Give us a comment, folks.

[00:05:06]Um, right now we wanna jump into the subject, uh, which we typically are all talking about our experience with a particular thing that's going on.

[00:05:16]Right now when it comes to the idea of, of vibe coding, everybody has experience with the idea of vibe coding and, um, uh, the idea that you, you could just prompt your way to something beautiful and amazing and awesome.

[00:05:28]Um, but truthfully, when it comes to vibe coding a WordPress website, that frontier I think is still in front of us a little bit, personally.

[00:05:37]Andy, you maybe feel differently, um, or might say that differently.

[00:05:42]Um, but I think that, you know, Zack and I, our experience with that might be a bit limited, honestly.

[00:05:50]Um, and that's why you're here, Andy, for sure, is that you are creating that frontier.

[00:05:54]You are looking at where we are today, you're looking at where we need to go, um, you're looking at the future, you're looking at where AI is going.

[00:06:02]Uh, you have a u- unique perspective on this question.

[00:06:05]Um, so just kinda at a really high level, I'd love to hear just kind of like what would you say today is the state of building an awesome website, WordPress website with AI?

[00:06:20]Yeah, I would say, um, if you, if you put Mil- Miles aside and use the current, uh, AI agents that are available in a Codex, Claude, that, and, and other products that work within the WordPress admin, um, I would say it, it's... You can easily do content updates, um, you know, you can generate images, you can, you know, write new blog posts, cr- you know, create some basic pages.

[00:06:46]I think that without other tooling, where it gets difficult is when you wanna build something, um, more like a full website, you know, more like full and detailed pages, custom blocks, um, just even using core blocks in general, um, to build out designs.

[00:07:04]It starts to get more difficult.

[00:07:06]The agents are not as good at, um, getting the block markup correct, and you'll find you'll, you'll get a lot of invalid blocks, um, if you start using, you know, just the base tools.

[00:07:17]Um, so, you know, I would say at, at, at a, at a very sort of surface level and, and, and entry level of using AI, I think it works well with WordPress.

[00:07:26]But as soon as you start to get into the details and start to have goals that push you beyond those basics, then I think it's can be frustrating, and that's sort of... That's really what you know, that, that was my motivation for getting into this space and get, and kind of seeing that you've got this- You know, incredibly mature and powerful piece of software in WordPress that, you know, if, if we don't have people pushing kind of the frontier of how to use AI with it, then, you know, you may, you may see other, um, pieces of software or you may see other, um, solutions kind of overtaking it, and that's not what You know, I, I love WordPress.

[00:08:02]I've done it most of my life.

[00:08:04]You know, I really wanted to see it keep up and, and be pushed in the right direction, so, um- Absolutely... that's what led to Miles.

[00:08:11]Zack, what's been your experience so far with building out, uh, a WordPress website with AI in general?

[00:08:17]Um, I, I, I'm kind of assuming it's similar to mine, but I just wanna, I wa- I wanna hear your take.

[00:08:23]Well, uh, I've been... We've just recently relaunched our website.

[00:08:28]Uh, the Filter agency, uh, redid our website.

[00:08:31]We launched it in, uh, October of 2025, and AI has been super helpful for already allowing us to add and enhance the existing features in the theme.

[00:08:43]So that was really nice that, like, having s- having a theme and then being able to modify it with AI has been great.

[00:08:49]Um, Casey, our, uh, head of growth and, and content person, had an issue though, like Andy said.

[00:08:57]Uh, when he would go to write blog posts or update blog posts, uh, it would strip all the block markup, and blocks would say they were invalid, and you'd have to go in there and click into it and it might or might not work.

[00:09:12]So we created the, uh, block MCP, we launched it last week publicly, uh, that actually parses block markup, writes it properly back to the blocks, uh, and updates, uh, atomically, which is great, um, so that you can move blocks around and it doesn't m- rely on the MCP to, like, copy and paste and move things around because AI's not great at that.

[00:09:37]So this is all, uh, powered by a plugin you install on your own website, and then it registers its own endpoints, so it doesn't use the WordPress REST API.

[00:09:46]It registers its own thing.

[00:09:47]So, like, structured data is the name of the game when you're dealing with AI, and one of the big limitations that we found was the structured data and the, the way to interact with it is just not there in WordPress.

[00:09:59]So Block MCP, it reads all the patterns that you use on your website, it reads all the blocks you use, and it prioritizes the ones that it recommends based on how frequently they're already used on your site.

[00:10:09]So that type of advancement in WordPress is going to be needed so that AI can know what to do better, uh, can know what the structures are, can know what content should be edited and whether or not it needs to be represented in the HTML comments.

[00:10:25]Like, there's so much for the AI to know, and having good MCPs is a good starting point that, that, uh, Block MCP was helpful f- uh, for us, and it, it works great.

[00:10:34]Um, and it also doesn't do any design stuff.

[00:10:38]So, uh, there's a whole lot of, uh, conversation to be had about working with websites.

[00:10:46]I haven't done the design stuff, Matt, to answer your question, when it comes to starting.

[00:10:50]Yeah.

[00:10:50]I would honestly, if I were, if I were wanting to actually launch a website using AI, sorry Andy, I would say do a traditional theme right now and forget the block theme.

[00:11:02]Like, don't do full site editing, just like A classic 'Cause that's... Do a classic theme is what I would probably do if I wanted to launch it quickly, because forget all this stuff, like the, the, it's super frustrating.

[00:11:16]Um- Okay, Boomer.

[00:11:18]You know?

[00:11:18]All right.

[00:11:19]I gotta tr- you gotta try.

[00:11:20]We need to get you on Miles.

[00:11:21]I, Andy, I tried signing up.

[00:11:23]You have a, you ha- you have like a, a, a great s- Like, impressive website, tons of great examples, and it says, like, "Sign up for a consultation." I'm like, "Oh, man, okay.

[00:11:31]I guess I-" "I just gotta spend some time."

[00:11:33]Open up the beta, Andy, come on.

[00:11:35]I gotta open it up.

[00:11:36]Yeah, I, I like the pr- yeah, that's good.

[00:11:37]I appreciate that.

[00:11:38]I'm gonna, I'm gonna open it up.

[00:11:40]Let me tell you a little bit about my experiences so far.

[00:11:44]So, um, I, I have just been absolutely fascinated with Cursor and the way it empowers me to do everything I wanna do.

[00:11:51]Um, and I was trying to spin up a quick site locally a- and it recommended Astro, and that was how I first got introduced to, uh, Astro before it was acquired by Cloudflare.

[00:12:05]Um, and it was amazing to me how it understood Tailwind and all the design tokens, um, and, um, was able to spin things up really quickly and easy, and it looked beautiful out of the box.

[00:12:18]Beautiful-ish, I'll say, out of the box.

[00:12:21]Um, and I was impressed, and I was like, "Damn it, I'm certain that we have to do this in WordPress. Um, and how can we do this in WordPress?" So I've tried a ton of different things.

[00:12:32]Um, and right now, whenever I wanna try anything like this in WordPress, I have to spend The first hour basically, like installing the AI plugin, the AI Experiments plugin, um, installing the MCP, uh, installing, usually I install Ollie, quite honestly, Mike McAlister, shout out if you're in the comments there.

[00:12:53]Um, and, um, setting up a whole bunch of, um, custom abilities and then spinning it up in studio, and then hoping and praying that it's going to understand anything at all, and it normally doesn't all that much.

[00:13:08]Um, that's, so far that's been my experience.

[00:13:11]Then I tried, um, Miles, which I love, and we're gonna talk about Miles for sure.

[00:13:16]Um, and then I tried Telex, which a lot of folks, uh, have probably written off Telex a while back because I think most folks think it's, like, a good way to spin up a custom block.

[00:13:28]Um, but, uh, l- uh, over a year ago, I tried Telex.

[00:13:32]It spun up a custom block, but then I tried to say, "Oh, but it needs this setting or that setting," and the iteration just fell apart and collapsed.

[00:13:39]Um, a good old, uh, uh, stellar friend of mine, um, Matt Batchelder, is working on Telex at Automattic right now, and he told me that now Telex is all about building out a whole entire theme, a whole website, uh, all kinds of custom blocks.

[00:13:55]And so I tried it out again this week, and it is very Miles similar.

[00:14:00]Um, and it is way more powerful than it used to be.

[00:14:04]Um, it still kind of fell apart in certain ways, um, but that kinda to me is, like, the state of, of, um, vibe coding a, a WordPress website today, is that it's, all of it is mostly there, um, but there's still issues.

[00:14:20]The other one that I don't have a ton of, uh, hands-on experience with, uh, only a little bit, is Angie from, uh, Elementor.

[00:14:28]Um, they launched it originally, what they called it was an MCP, but it really is like a whole, um, AI-empowered, uh, building experience based on the Elementor page builder, of course, um, and that whole ecosystem.

[00:14:42]Um, so there are tools out there.

[00:14:44]Um, I just, I kinda think Elementor is, like, in its own category.

[00:14:47]If you wanna build an Elementor website, then, then go with that.

[00:14:51]Um, but Telex and Miles seem very, very similar in my mind.

[00:14:56]Um, I don't know, Andy, what does that ring true for you in your mind right now?

[00:15:02]Just lost my connection there, but I'm back.

[00:15:04]Um, yeah, the, uh, the, yeah, I think the, the point I wanted to make on, I think I, I totally see what you're saying, and I think that the, the, the, the competitor to WordPress right now is the, the flow that you originally went through, is that you go into your coding agent.

[00:15:23]You say, "I want a website.

[00:15:25]I have this X, Y, Z goal Uh, you type it in, and hopefully a website comes out the other end that works for you and is on a platform that works for you.

[00:15:35]You may not even choose that, and like you say, like, the, the AI is choosing Astro.

[00:15:39]We need to work on that.

[00:15:40]We need to get it choosing WordPress.

[00:15:42]But I think that for... The point I wanted to make was that to- for WordPress to be competitive going forward is that you need that kind of all-in-one experience.

[00:15:53]You need that, I'm just gonna give you a goal, I'm gonna tell you what I need, and it needs to put all the pieces together, work seamlessly, install exactly what you need, create a design that you like.

[00:16:05]I think that that, for WordPress to be competitive over the next one, you know, one, two, three years, you need, excuse me, you need, um, you need to, you need a, an exper- a WordPress experience that allows you to just say what you want and it can build everything out smoothly.

[00:16:23]Without having, you know, the, the issues that you mentioned where you have to figure out which plugins to install, kind of the old school, like, developer mindset where you have to put the pieces together and then, and then, and then you can maybe c- create something.

[00:16:35]Like, the, the, the alternatives are so smooth now that you have to compete in that arena, I think- Yeah... in order to, to move forward.

[00:16:43]100%. 100%. Yeah.

[00:16:45]First of all, uh, I wanna shout out Darshan, uh, who r- writes, "We vibe coded and launched five websites in two months. All are on classic theme.

[00:16:53]Amazing results with classic theme."

[00:16:56]I'm not alone in this take.

[00:16:57]Part of, part of the, part of the problem I think that WordPress has to solve, and unfortunately they, they didn't solve it w- when they were doing full site editing, and block theme, and all this stuff, is that when we relaunched our website recently, uh, one of the things that kept on coming up was the database and the theme both are trying to define what is happening with the layout of the website.

[00:17:23]And if you've changed anything, all of a sudden it's a database setting instead of being pulled from the theme files.

[00:17:28]And this breaks my brain.

[00:17:30]It makes it so that it's not possible to have version control in a way that makes any sense to developers.

[00:17:35]Um, I think that this is a completely unsolved problem.

[00:17:39]I want the database to have nothing to do with the layout of my website.

[00:17:43]I want that to be defined in the files.

[00:17:45]So that, that was a really frustrating experience and why I said a, a classic theme is what would be better for me.

[00:17:53]The, the thing that I've been doing recently, 'cause I'm trying to relaunch another website for Trusted Login, is I've been iterating a lot locally with an HTML file.

[00:18:02]It uses CSS tokens and tries to define the, the structure of the site using, like, all locally, all in an HTML file, and iterate fast with AI on that.

[00:18:13]And then once you have a nice HTML structure, and design, and stuff, you can then say, "Turn this into a theme." You can then say, "Turn this into a, an, an Astro website," something like that.

[00:18:23]It's, it's nice to have something that you can work with that, that the AI can look at and, and fully understand what you want.

[00:18:33]Really interesting.

[00:18:34]We have a really good comment also, uh, from Zack Stepek in the comments here on YouTube.

[00:18:39]He says, "I'm experimenting with Claude Design to blocks uti- utilizing Claude Code and a number of MCPs and plugins.

[00:18:48]Ollie MCP is awesome and a good glimpse of the future." I have not experimented with Ollie MCP at all, and I'm sad to say that.

[00:18:54]That, that's cool.

[00:18:56]Um, that sounds li- like a great workflow.

[00:18:59]Claude Design is, um, really powerful overall.

[00:19:02]Um, the re- related to design tokens, um, and after this I wanna get in more in detail about Miles specifically, but, um, I actually wrote this issue, uh, that Andy also has seen, uh, for Gutenberg specifically, because, uh, that is one limitation that I experience, or, uh, honestly my experience with starting with Astro and then trying WordPress, it was, it was essentially this.

[00:19:29]It's like there are design tokens in WordPress, they're just not exposed in a way that people can customize them particularly well, and so let's make that a lot more serviceable.

[00:19:39]Um, so, a- and I honestly had Miles in mind, uh, with this as well.

[00:19:44]I feel like something like this would, uh, make what you do with Miles a lot stronger and better as well.

[00:19:51]Um, here's Andy commenting as well.

[00:19:53]Um, I, I wish there was a, a little bit more momentum on this idea, but you could see here my little mock-up of the PR.

[00:20:00]Um, in the styles area, you could see customize the tokens right here, and if you click on that, there's just a couple of the custom tokens that I created, um, that you'd be able to do the card border radius however you want.

[00:20:11]You could choose your accent color, you could do your section spacing, things like that.

[00:20:15]Um, uh, but, uh, yeah, I don't know.

[00:20:19]That's... I, I feel like something like this is necessary in order to keep, uh, moving forward well.

[00:20:26]Yeah, I think, um, I think the, the more structured data the better.

[00:20:29]I think, you know, Zack mentioned that, like AI is just gonna be, if you have good structured data, it's gonna work so much better.

[00:20:35]Um, the thing, the thing I would say on that is the design tokens, you know, they... AI works much better with that, with that structure.

[00:20:45]But I think that the, the main reason is, is that you can expose those controls in the interface as well, and that you can make them editable, you can show the controls.

[00:20:55]The user can then, you know, override whatever the AI's done more easily.

[00:20:59]That's one piece I would add in the sort of the, the plus column for block themes.

[00:21:03]You get an interface, uh, you get the visual feedback, um, you get support for more advanced designs that you would otherwise not be able to represent very well in, in classic themes, at least visually while you're editing.

[00:21:16]Um, on the flip side, you're also enter- you know- The, the design tokens are also introducing another layer of, I don't wanna say confusion, but another thing for the AI agents to learn, and they're re- they're really, like, fantastic at HTML, CSS, and JavaScript.

[00:21:36]So I'm sort of caught in two minds there.

[00:21:38]I think yes, that more better block tokens would, or better design tokens would help the interface.

[00:21:44]It would ultimately help the AI.

[00:21:46]But we're at, what we're doing really is adding another layer of abstraction on top of something that it's already really good at.

[00:21:52]And maybe, maybe it's worth considering, is it better to, to move forward with, with the things that it's good at and sort of release that kind of control a little bit- Yeah in a way that is gonna result in better designs, like more interesting designs, more dynamic designs- Yep... and, uh, just fewer problems going forward.

[00:22:13]You know, we have, I think there's a tendency with WordPress to, to wanna abstract everything away from the core components, and in the world of AI agents, that maybe isn't necessarily the right move.

[00:22:26]You need to find that balance.

[00:22:28]Yeah.

[00:22:29]I think, um, backing up just a little bit, we keep mentioning design tokens and things like that.

[00:22:35]Um, the, the other term that is used sometimes interchangeably, um, I'm gonna forget it right now.

[00:22:42]Um, um- CSS variables?

[00:22:46]Vari- no, not exactly.

[00:22:47]There's, it starts with a P. Shoot, I'm gonna drop it.

[00:22:50]I'm gonna forget it.

[00:22:52]Oh.

[00:22:52]What are you- It'll come to us.

[00:22:53]Yeah.

[00:22:53]Like, does it, like theme JSON?

[00:22:55]Like, uh- No.

[00:22:56]Uh, I'm gonna, it's gonna come to me later, but- Okay... so in WordPress it's design tokens.

[00:23:00]In, in Astro they talk about it a little bit differently.

[00:23:02]Um, but um, the whole idea is that, like, you have a set of, of design rules that the whole application is run by, and you define all of those in order that you don't have to apply a whole bunch of custom CSS everywhere.

[00:23:20]Because as soon as you start applying custom CSS everywhere, you run the risk of, uh, essentially having everything look different across every different page.

[00:23:30]Um, and, and the, the design tokens es- essentially, uh, empower you to have a very consistent look and feel across your whole website or across your whole application or whatever you want.

[00:23:41]Um, so I, I agree, Andy.

[00:23:44]As soon as I built out the, the interface, um, to do, like, the custom tokens, I was like, "This is actually kind of a small little sidebar," and like, I can imagine if 25 plugins all wanted to add their own custom design tokens, this could get really- Yeah.

[00:23:57]Now you're in another world unwieldy really quickly.

[00:23:58]I mean, that's the nature of WordPress, is as soon as you make it extensible, everybody wants to extend it.

[00:24:04]Um, and then all of a sudden you have actually problems that you didn't have before, so.

[00:24:08]Well, that's the thing that I think WordPress- used to be more comfortable with in a, in a way, even though it was decisions versus choices.

[00:24:17]Uh, and the, the classic theme said that we have a function, it's called get header, and this is where, like, all the stuff happens there.

[00:24:26]Um, the-- I think if WordPress were to define the token set more explicitly, uh, a- and have, uh, and cover all the possibilities or many pos- more possibilities than a c- than theme JSON covers out of the box, I think that would be a real service to just standardize everything.

[00:24:44]And maybe there's been motion on this that I don't know about, but it feels like if WordPress were to take the lead, that would be really helpful.

[00:24:51]Even just outside, like, beyond the block editor, just as WordPress core, that would be really helpful.

[00:24:58]Yeah.

[00:24:58]What do you think, Andy?

[00:24:59]Is that... Am I, am I missing something big?

[00:25:02]No, I think that's exactly right.

[00:25:03]I think, uh, the more that we talk about these fundamental pieces in WordPress, and the more-- and, and I would go back to documentation.

[00:25:10]The more that all of this is documented, 'cause I think some of the documentation is, to be honest, quite weak, um, for humans and for agents.

[00:25:19]If we, if we were able to broadly document all of these foundational pieces in a way that agents could absorb very easily, um, then I think we would see over time, and probably not that long of a timeframe, the agents get significantly better at doing this just kind of naturally.

[00:25:37]Um, and maybe even going as far as recommending WordPress over other products as well.

[00:25:42]It's, if you think about, you know, agents are just vacuuming up everything on the internet, and when there's a void of information, it, it, it leaves a void in the models, and I think that's what we're seeing in terms of using models with WordPress.

[00:25:54]We're seeing the result of that void, and we need to fill that.

[00:25:58]Um, I think that's k- actually quite existential.

[00:26:00]Like, if, if, if I was sort of in charge of that, that's the first thing I would do is, is put as much data on the internet about how to use all of these different very fine-grained pieces, and that would ultimately, like, um, build itself through the future models, and then we'd see better outcomes.

[00:26:16]Yeah, absolutely.

[00:26:18]Well, Andy, let's, uh, let's talk a little bit about, uh, about Miles here.

[00:26:23]Um- Sure thing... this is, uh, buymiles.ai.

[00:26:27]Um, this is your site.

[00:26:29]Tell us, uh, at a high level, like, what are you trying to accomplish, and what do you feel like you have accomplished?

[00:26:36]I think those are my two big questions this audience would love to hear about.

[00:26:41]Yeah, I mean, I think the, the first thing is I n- I want, I want a product that can build beautiful designs because- Yeah... I think, uh, you know, you can do a lot of design work with AI agents, Codex, Claude, whatever it might be, but I think that the designs, you know, they, they can be quite repetitive.

[00:27:00]They have the tells of AI design.

[00:27:02]Um, so I wanted to see how far I could push the current models and, um- Yeah, and, and their level of design ability with, uh, the, the harness that I'd built for Miles with the tools, with the prompts, and yeah, see if it could create the, the level of variation in design that, that I think people need and wanna see.

[00:27:23]They don't want every, every website to look the same.

[00:27:26]You know, can it understand what your needs are in terms of the s- the type of site that you're trying to build, who your audience is?

[00:27:33]And can it produce something out of the box that people say, "Yeah, that's, you know, that's what I want," um, without, you know, tons and tons of iteration or, like, wasting tokens or wasting your time?

[00:27:43]Can it really kind of nail it out of the box?

[00:27:45]So my kind of early days or early months of working on Miles is just to kind of try and really nail that, kind of, you know, upfront, give you something that you're like, "Yes, this is good. I wanna continue with this." Um, yeah, and I've put on the webs- you know, the, it's made hundreds and hundreds of designs and I've, you know, I've picked- Yeah uh, good ones on the site to try and showcase that.

[00:28:05]But I think without that, without that initial, "Oh, this looks really good," kind of aha moment, the rest of the product doesn't really matter.

[00:28:12]So I focused a lot, um, on the, a lot on that upfront.

[00:28:15]I love that.

[00:28:16]You have a fan in the comments here.

[00:28:18]Troy Chaplain on YouTube says, "Best design agent out there."

[00:28:22]Oh.

[00:28:23]Yeah.

[00:28:23]Troy's been great.

[00:28:24]He's been using Miles a lot, and he's been helping with a lot of feedback and, uh, yeah, it's me- the, the product has mean- meaningfully improved with his feedback, so that's great to hear.

[00:28:33]I think, you know, the des- the, it's working really well.

[00:28:35]I'm really ha- really happy with the design output.

[00:28:37]Yeah.

[00:28:37]You know, you always improve it and, you know, the, the landscape is always moving, and things are getting better, so you, you know, you have to keep moving.

[00:28:45]But, um, I think in terms of design output, it's in a, it's in a good space.

[00:28:49]I love that- It's-... the, the emphasis on the design, uh, first, and ev- if you just scroll through those designs, there is a huge amount of variability in there.

[00:28:58]Um, so I, I think hats off.

[00:29:00]A- and it doesn't look obviously super AI.

[00:29:04]Um, like, w- there was, uh, a website promoted around the web yesterday, um, that got a lot of attention in the WordPress space where all the pages were clearly very AI-driven.

[00:29:16]Um, and it, it, it wasn't a good reflection of that brand.

[00:29:20]Um, I won't name names.

[00:29:21]Um, but, um, it, I, I think you, you... That having that focus first is, is such a great focus, for sure.

[00:29:28]So, uh, uh, Andy, I wanna ask you more about this because, and Matt, because, uh, when I work with designers, when I was an agency, or when I did, when I did client work, it was always, well, the content first and then we'll talk about the design, and understanding your business and who your customers are, and then we'll talk about the design after.

[00:29:47]Like, it, it all focuses on the content, not the, not the design.

[00:29:51]So, like, when you're thinking about vibe coding a website, um, how... I'd love to learn more about how you figure out what the vibe of the website should be for, you know, design-wise, but also how do you take into account the, the content that's going to be the highlight of the design?

[00:30:09]Yeah, I mean, that's, that, that's a critical point and a critical piece of making good design.

[00:30:13]I think that still stands.

[00:30:15]I think you've gotta l- you've gotta learn about what the person wants, both, like, in terms of the content that they want on the site, but also the vibe, also the things that they're trying to communicate with customers or, or users of the site.

[00:30:28]And, and that's sort of the approach that I've taken is treat Miles like a design partner.

[00:30:33]Like, it's gonna... Miles is gonna ask you questions, uh, upfront to try and get a really good idea of what you actually wanna build.

[00:30:40]Like, what, what is your actual goal?

[00:30:42]You know, not, not necessarily technical things.

[00:30:44]Actually, no technical things.

[00:30:45]More like, what i- you know, what do you want to come out the end of this?

[00:30:48]Um, you know, trying to kinda tease that information out of you.

[00:30:51]And I think one of the b- one of the big reasons that it does produce pretty varied, like, a good variance of design and quality of design is because it asks those questions upfront, and it builds a, essentially a website brief, like a, a, you know, a kind of document that you would normally take, you know, to a designer or to, to somebody to say, like, "This is what I need and these are the pieces of content," or, "These are the things I really wanna show on the site," and then build out from there.

[00:31:16]I mean, it also... The, the nice thing with AI models is they can-- They have a pretty good idea of the kinds of content if you have that kind of brief, the kinds of content that you want on the site.

[00:31:26]Mm-hmm.

[00:31:26]So, the idea is that, you know, you're not gonna come upfront with all of the content that you want, but it's gonna put content in that probably fits structurally around the design that you want, and then you can, you know, you can update and change with Miles or manually afterwards.

[00:31:41]So, the design is built around that kind of knowledge of what is the typical structure of content for a site like this, and that helps, yeah, as you go forward and make edits.

[00:31:52]Nice.

[00:31:53]Um, we mentioned before we got on that, um, there are these WordPress-specific competitors.

[00:32:00]There's also, like, Lovable as a giant competitor.

[00:32:04]Yeah.

[00:32:04]Uh, people will choose Lovable instead of anything to do with WordPress at all.

[00:32:08]Um- Mm-hmm... and they're a giant billion-dollar company now already.

[00:32:12]Um, but, um, you said that you don't pay super close attention to what the other WordPress-adjacent competitors are doing, and you had some reasons for that.

[00:32:23]I actually would really like to hear that.

[00:32:25]Um- Well, I mean, I'm, obviously I'm aware of competitors and I do look at them because you just, you know, as a market research, you've gotta... If you're building a product you need to be aware of competitors.

[00:32:35]But my, my point there was I don't try and, like, really focus on what they're building and, like, day to day or week to week the things that they're shipping because it's really dangerous.

[00:32:46]It was dangerous before AI models in that you can allow competitors', um, roadmaps to really influence what you're doing, and you end up just kind of following competitors rather than trying to innovate yourself.

[00:32:59]And it's dramatically more dangerous now because You know, you can just say, "Look at my competitors."

[00:33:06]Right.

[00:33:06]Do what they're doing.

[00:33:07]Yeah, just do what they're doing.

[00:33:08]You know, like, just build this, you know, spin, spin up a Ralph loop or a plan and say, like, "Overnight build X, Y, Z." In the morning it will be there.

[00:33:16]And I, and I absolutely guarantee that's the way to build a terrible product.

[00:33:20]So-... I'm really, really, really strongly You know, I have... You know, it's so easy to do that I have to force myself to say no.

[00:33:28]You know, grou- what I'm trying to do is ground myself in the feedback that I get from people using the product.

[00:33:33]You know, they're telling me what sucks and what, you know, what has potential, and trying to, like, really concentrate on that.

[00:33:40]Yeah.

[00:33:40]Be aware of competitors, but concentrate on what pe- what the feedback I'm getting from people using it.

[00:33:45]100%.

[00:33:45]Yeah.

[00:33:46]I'll, I'll just say, like, marketing pro tip, folks.

[00:33:49]Um, I hear people in the plugin space talk all the time about paying attention to their custom- to their competitors, and they often are doing it in order to get feature parity or things like that.

[00:33:59]Um, I'm with Andy.

[00:34:00]In general, um, I do pay attention to competitors really carefully, but not with an eye towards, uh, following them, with an eye towards differentiation.

[00:34:11]Um- Mm-hmm... that's what it's all about.

[00:34:13]The, the competitor research stuff is all about, uh, understanding how you are different than your competitors, not how you can chase them.

[00:34:20]Um- Yeah.

[00:34:21]So- And find the negative reviews of Lovable and what are people complaining about, and improve that.

[00:34:26]Yeah.

[00:34:26]Right?

[00:34:27]Yeah, it's g- that's a great way to look at it as well.

[00:34:29]Um, you know, it, it's, it's super important now, more important than ever with the fact that you can spin the things up so quickly.

[00:34:34]So- Yeah... yeah, I think, think that's a good one.

[00:34:38]Uh, we have a question, uh, from Troy Chaplin again on YouTube.

[00:34:41]He says, "Do you have any plans or can Miles currently help with designing front-end view for custom blocks?"

[00:34:50]Yeah, I mean, it, it, de- definitely.

[00:34:53]I think that's super important in terms of building, uh, Miles building out fully functional, more advanced websites.

[00:35:01]I mean, that's ultimately my goal.

[00:35:03]It's... I think the differentiator for Miles and, and for WordPress in general, is that it's so, um, you know, so much history and so mature as a product that it's possible to build out any level.

[00:35:15]You know, the, the White House runs WordPress.

[00:35:17]You know, you've got, you've got bu- a bunch of different sized websites doing very, um, advanced things that WordPr- we know WordPress can handle.

[00:35:27]So, you know, and you, and you see competitors, you can build out, you know, a brochure website, you can build out a, a medium advanced website.

[00:35:34]You can probably get further with a lot of kind of back and forth.

[00:35:37]But I want Miles to be able to smoothly utilize all of the, the advanced and mature functionality of WordPress, and that includes plugins and writing its own plugins, and being able to not do it on a, on a technical, from a technical perspective.

[00:35:53]You know, someone could say, "Can you build me a custom block for XYZ?" And it would do that.

[00:35:56]But thinking about it from a goals perspective, like, "I need a website that does this for my customers," and Miles is gonna help you figure out like, okay, we might use core blocks for this.

[00:36:06]You know, there is already a plugin that is proven, you know, is secure, is, you know, very... It's got a huge customer base.

[00:36:12]Let's use that for this particular feature.

[00:36:15]And perhaps, like, builds c- like, custom extensions or custom blocks for things that maybe are a bit more niche to what you're doing.

[00:36:21]So again, like speaking to the product that can, uh, can help you as a partner and make the right decisions based on a goal, that's kind of where I see it going, rather than it being this ad hoc product that can- Yeah individually build things.

[00:36:36]So let's lean into this, uh, pain point you and I have talked about a bit.

[00:36:39]We've kind of gone over the surface a couple different times, um, about it, but I, I wanna hear your take on, like, what does core really need to do to be able to move forward and be that competitive advantage against the Lovables or the Astros and whatnot.

[00:36:55]Um, you and I have talked a bit about how in the beginning Miles was heavily dependent on custom CSS- Mm-hmm... because just- Mm... core wasn't doing the work.

[00:37:07]Um, and it still in some ways isn't.

[00:37:09]You've now moved a lot of your custom CSS into the block area, not the, uh, full site editing, um, additional, um, uh, additional CSS area, but the individual block area.

[00:37:22]Um, why is that, uh, specifically from like a technical core perspective, and is that okay?

[00:37:29]Um, or, or what do you think core needs to do to make this vision more realized?

[00:37:37]Yeah, I mean, it- to answer your direct question, that it is okay, and it's actually a new thing in Gutenberg and obviously in core with the, uh, V7, um, that blocks can have additional CSS at the block level.

[00:37:52]I do that, I, I've done that rather than a large CSS file because it, it gives Miles a much greater ability to scale for larger sites.

[00:38:02]What you find is if you have a single CSS file or even split CSS files up across multiple pages, it, as you scale to a larger site, it, it, it, you know, that's a lot to read.

[00:38:13]It's a lot to search.

[00:38:14]It's a lot to kind of structure and get the, um, the cascade correct.

[00:38:18]So- By thinking about things on a pattern and block level, installing them on a pattern and block level, and using the design token system to, um, to make sure there's consistency and, you know, you change one thing in one place and it changes in other places.

[00:38:33]That, that allows for much greater scalability because you can, you know, you're basically building individual sections that you can put anywhere and, and build out gradually as, as, as sites get larger and larger.

[00:38:44]So th- that, that's a positive sign.

[00:38:45]I think that that is being seen and being developed certainly on the Gutenberg side or the block editor side.

[00:38:52]Um, I, I, I like the-- I like what is happening in Core with, um, with all the APIs around, uh, allowing plugins to provide kind of access points to AI agents.

[00:39:06]Um, I think that's good.

[00:39:08]Uh, in terms of, yeah, what they can do going forward, I don't have any, like, really strong recommendations at this point because I'm aware of how fast things are moving and how things change so quickly.

[00:39:20]So even, even something that, you know, s- made sense six months ago for Core to build perhaps doesn't make as much sense now.

[00:39:28]Mm-hmm.

[00:39:28]You know, we've seen things like CLI access.

[00:39:32]Um, just the ability for the models to use the REST API really easily kind of opens up a lot of doors.

[00:39:38]Um, but all... The work that they're doing to allow plugins to, to communicate with, with models I think is, is really beneficial, and they should keep pushing on that.

[00:39:47]Um, but outside of that, I don't have anything specific.

[00:39:50]I haven't really run, you know, outside of the m- outside of the structured data design tokens, I haven't run into yet any serious, like, roadblocks where I'm like, "I cannot do this. It's not possible." Yeah.

[00:40:04]So I expect that's gonna happen as Miles gets more advanced and we s- you know, it does really start leaning into custom block building and interacting with plugins.

[00:40:12]But at this point, it's been okay.

[00:40:16]And as the models get more advanced, you know, you can kind of work, you can work around limit... And I don't like to work around.

[00:40:23]Obviously, I wanna push back to Core and get it improved, but you can, you know, it can write a custom API for a plugin if there isn't one, right?

[00:40:29]So say that it can't communicate with a well-known plugin because it doesn't have the access points or the API endpoints, Miles could just write that- Mm-hmm... and save it.

[00:40:38]Mm. And it would use it to communicate with, with a plugin.

[00:40:41]Um, so I don't have a, I don't have a really good answer for your question.

[00:40:45]I, I mean, it's honestly really encouraging because- Yeah what you're saying is like- Core has enough to empower you to, uh, make this happen.

[00:40:54]Um- Yeah... they seem to be chasing the same vision, honestly, with Telex and, and other things that they're building as well.

[00:41:01]Um- Yeah... and so it seems like today already i- it is possible.

[00:41:06]Uh, maybe it will get even easier, even more powerful in the future.

[00:41:10]I know that, like, um, being able to have Miles interact meaningfully with plugins is on the horizon as well, and that'll be a big, uh, step in the right direction, too.

[00:41:21]Yeah.

[00:41:21]Um, are you also thinking that you want the Miles experience to be in your UI exclusively, or are you planning that folks are gonna interact with Miles through Cursor and Claude and other things like that as well?

[00:41:35]Yeah, I, I, I feel strongly that you've gotta meet people where they are, and for some people that's in WP admin, and that's fine.

[00:41:43]Like, it... Miles will work great in, in that interface.

[00:41:46]For others, and I think more and more folks this, this will happen, they're gonna transition into coding agents, whether that's Cursor or Claude or Codex, whatever it might be.

[00:41:56]And Miles needs to and will work seamlessly in those environments via the skill that, that's available.

[00:42:04]So you should expect that, um, if you wanna build out a full WordPress website and never touch WP admin, you should be able to work in the environment you are w- that you work in in one of those agents, and it, and it... without any real installation of WordPress or any sort of technical piece.

[00:42:22]It should just, uh... It should just work.

[00:42:25]I mean, that's really what... That's kind of what I'm really focusing on at the moment.

[00:42:27]Um, I think it's, it's, it's not realistic to say to people in this, in the speed that things are moving, and all of the custom workflows that people have, or, and even, even, like, all the custom skills that they're using, to say, "No, dr- drop all that.

[00:42:43]Like, go and... You, you can only use it in this environment." I think that that probably... I hate to say things in absolute terms, because it's moving so fast and things change, but I think that might be a dead end and it, it needs to, it needs to, um, it needs to work everywhere.

[00:42:59]So that's my goal.

[00:43:00]Mm-hmm.

[00:43:00]Yeah.

[00:43:00]Yeah.

[00:43:01]That's amazing.

[00:43:02]One of the things that I have come to understand about how we are managing our, our, our own website is that I am not relying on other plugins to do... to add additional blocks.

[00:43:15]I'm incorporating that into our own theme.

[00:43:18]Uh, if something needs to be developed or fixed, I do it, uh, either in our theme or in the block MCP or in another MCP we write, right?

[00:43:27]Um, because it's so much easier just to do it ourselves and maintain it ourselves than to have externalities and have dependencies.

[00:43:34]Um- Yeah... is... So, I, I mean, WordPress was built on the idea of just install the plugins to do what you need, and now that AI is doing so much of that- Um, how do you see the, the, the future of, of WordPress interoperability and, uh, and the plugin space from the expansion of, you know, the, um, the Cambrian explosion of WordPress, uh, functionality through plugins?

[00:44:00]Uh, are we cons- now about to consolidate back down to, like, just a few plugins, uh, like, and your theme?

[00:44:05]Like, h- how, how is AI shaping that future?

[00:44:10]Yeah, I, I, I think that this is TBD.

[00:44:13]I think that- Yeah... you know, it, it, it'd be easy in this state right now to say, "Oh, all plugins are dead," you know, everything is gonna be built by the AI in WordPress, and it's just gonna spin up what you need.

[00:44:25]I think that that is, you know, there will be cases where a plugin that perhaps provided functionality that AI, AI can easily s- the agent can easily spin up, maybe is not as widely used or maybe not useful anymore.

[00:44:39]But there's, I think there's still a huge space to provide functionality in WordPress that is, is, is well-tested, well-maintained, secure, takes the pressure off the ongoing maintenance of people building with WordPress or anything with AI.

[00:44:57]I still think y- instead of extrapolating out everything in a straight line from where we are right now, I still think that there's a path where just providing good and well-maintained software is gonna be valuable.

[00:45:08]So I'd say that the, the space is... I, I, I, I believe in the plugin space of WordPress.

[00:45:12]I think it still re- remain healthy, but it's gonna change and you have to really adapt to think about what, you know, what, what could, what are the, what are the agents not good at right now?

[00:45:23]Like, what, where is the space that you can augment and improve and build upon the power of the a- that the agents provide?

[00:45:31]And I think there's a huge opportunity there in the plugin space to, to continue to work in, you know, think about where that's going, continue to produce good software and well-maintained software.

[00:45:42]I'd, I would say that's generally true, in my opinion, true of all software right now.

[00:45:49]People are like, you know, "Software's dead. AI's gonna write everything." But I think it's easy to discount the complexity of main- making sure everything's secure, making sure everything's well-maintained, updated, and also the desire to do that over a long period of time.

[00:46:03]It's sort of what made WordPress good in the first place, is that you could, you know, anyone could... Not anyone.

[00:46:08]Some- Mm-hmm... some people could spin up a, a blog, write their own PHP, and be done with it, and say, "This is my blog now." But of course, along came cross-site scripting, and along came, you know, like, p- PHP updates, and along came incompatible versions of software that you're using.

[00:46:25]That stuff's, h- is not gone, and I think that there's still a space to, to handle that and provide a service in a, in a space where the a- the agents don't handle.

[00:46:34]Yeah.

[00:46:35]Nice.

[00:46:36]I have one more question for you, Andy, and then we're gonna get into our best advice, and I'm gonna pivot a little bit.

[00:46:43]I wanna get-- hear a little bit from you about, um, a product aspect, um, uh, not the AI stuff.

[00:46:51]Um- Yeah... you chose to do a private beta, which, um, is different than build in public, um, but has similar effects honestly.

[00:47:03]Um, I'd love to hear a little bit ab- just about your thought process in like, "I'm gonna do a public, uh, or a private beta. I'm gonna advertise what I'm doing. I'm, I am-- There's build in p- public aspe- a- aspects to it."

[00:47:16]But, um- Yeah... tell us about why you did it the way you did it.

[00:47:19]Um, and do you, are you liking it?

[00:47:21]Is it going well?

[00:47:23]Um- Yeah, that's a good question.

[00:47:25]Um, I think it, it was a conscious choice because it's, it's notoriously difficult to get agent, uh, custom agents or things that use a-- or products that use agents into production and reliable.

[00:47:41]Um, you know, it's, uh, uh, everything around AI is non-deterministic, and to guarantee that it's gonna work every single time is hard, and that takes time to really refine.

[00:47:51]So my idea was that to have a private beta, I could get a really good group of people that were understanding of the non-deterministic nature of the software, you know, as it was, um, and be there to say, "Okay, this, you know, I tried it this time, it didn't work this time.

[00:48:07]This, these, this is why, and this is what I saw," and just use that kind of closed beta cycle to, to refine and refine and get to a place where, you know, it was more, you know, almost, or, uh, um, I think it's close to 100% reliable in terms of like producing a good design get, and getting, getting blocks built and then it's, you know, maybe like 90, 95%.

[00:48:28]You know, it, it's hard with AI 'cause it is non-non-deterministic, but it's getting that i, i- in a private setting- And then gradually releasing people into the beta as, uh, as it became more reliable.

[00:48:43]You know, I can see if I let everybody in now, there are places where people are gonna drop off, and I'm refining and refining and refining, and then as, and we'll gradually, like, more rapidly open the beta as it gets better.

[00:48:53]And that's worked well so far.

[00:48:54]I mean, the, the main thing is just getting people's feedback.

[00:48:56]Yeah.

[00:48:56]And trying to get that every day.

[00:48:58]I think for any product, just getting customer or user feedback, um, is just gonna make the product better, so the private beta certainly helped with that, without burning a bunch of people and kind of losing the goodwill of- Yeah it being a, an interesting product.

[00:49:12]But at the same time, you definitely have, um, like, created a, a base of super fans.

[00:49:19]Like, the folks who are in the private beta, myself included, um, are just thrilled with the product already.

[00:49:27]Uh, and you get all of this promotion from them, essentially free promotion more or less.

[00:49:32]Um, but also, um, you do have the ability to, to, for private beta folks to start paying you as well.

[00:49:41]So your, your, your path that you're on is really unique.

[00:49:44]Uh, like I almost feel like we need to bring you back and just talk about just this aspect for the whole conversation- Yeah... because it's really, I think it'd be really valuable for our au- audience to hear more about, about your perspective there.

[00:49:55]But overall, you feel like, you feel like it's been working?

[00:49:58]It's been going well?

[00:49:59]Like, um- Yeah, I, I f- I'm really happy with how it's gone.

[00:50:04]Like, I, I've, the feedback I've got is fantastic.

[00:50:06]Like, it feels like, but, you know, leaving a, a big company and going to work on your own, and then having a group of really, like, passionate people to work with and talk to every day and build, it feels like you're building something together, which is just kind of what I wanted.

[00:50:19]So, um, you know, I, I can't thank all the folks in the private beta enough.

[00:50:23]That's really been, like, great for my mental health, but also great for the product in terms of, uh, it getting better over time.

[00:50:30]Yeah, I mean, I, I, I didn't plan it out consciously, like really this is how it's ac- it's gonna go.

[00:50:36]It just kind of developed over time and, and provided that safe space to re- to really, like, iterate and, and, and be okay with things not working early on, so.

[00:50:45]Yeah.

[00:50:45]Yeah, I'm happy to talk about that more in another call.

[00:50:47]It's, uh- Yeah... it's worked well.

[00:50:49]I do have a question for you, Andy, because I have not used Miles, like I said, uh, and I want to, but also when do you decide to open it up?

[00:50:57]It sounds like people are loving it, people are paying for it, people are, uh... And yet, uh, like, why have you not opened it up yet?

[00:51:05]Well, this is the, this is the, I would say, the negative of this approach is the when is it good enough?

[00:51:09]Like, what, what is the line that you need to cross?

[00:51:11]It's been, apparently it's great.

[00:51:12]Like, everybody in the chat, Matt Cromwell gives it a full thumbs up.

[00:51:16]Uh, it sounds like it's great, so what's holding you back?

[00:51:21]Well, what I'd say what's holding me back is setting expectations correctly, I think, and that's probably on me as well is, like, what, what... If you, if I let everybody in right now- What should the expectation of what Miles can and can't do?

[00:51:34]That is critical to set correctly, so I'm working on that, and I'm, I'm delighted to- I hear from Andy Peatling of of Miles AI that it's almost 100% reliable for great design.

[00:51:45]Oh, you can test that.

[00:51:48]Yeah, you can test, you can test that.

[00:51:50]I'll probably live to regret that statement, but, uh- That's a... You know, that's what I want it to be, you know?

[00:51:55]I want it, I want- Yeah... I want it to go into Miles, and I want it to make a great design.

[00:51:58]You'd be like, "Yeah, this is 100% worth it." Otherwise, why am I bothering?

[00:52:01]You know?

[00:52:01]I, I have a totally different bar for you, Andy.

[00:52:04]I want it to be able to make the really crappy plugins actually look good on the website.

[00:52:08]That's... There you go.

[00:52:11]Yeah, that can be done.

[00:52:12]Yeah.

[00:52:12]Uh, there's so many times where you, like, build the whole great-looking website, and then you drop a plugin in, and it, and, and it looks so artificial.

[00:52:21]It looks so, like- Yeah... not what the w- rest of the website looks like.

[00:52:24]Um, but- Well, that's actually, to, to bring it back to the CSS, like the block level CSS, that's one benefit there, is that it can focus in on that specific block that maybe is not quite fitting in right, and flip it, and, you know, integrate it well.

[00:52:37]So- Yeah... that's another benefit.

[00:52:41]Well, we wrap up every show with best advice, and now is that time.

[00:52:46]Uh, w- what is your best advice for folks who want to get into vibe coding, AI websites in general?

[00:52:54]Uh, Zack, what's your best advice?

[00:52:57]My best advice is to have something in mind first.

[00:53:00]Uh, and I know that this is exact opposite of what Andy and Matt have been talking about.

[00:53:04]Mm-hmm.

[00:53:05]But I find when AI tells me something and I, I l- I'm like, "Well, I don't like this, maybe try another thing," it's, it's a lot easier if I know that, uh, a specific website makes me feel a specific thing, and what do I like about that website, and what are some of the words that I could use?

[00:53:22]Tell... You know, AI is great at, at helping you come up with a design prompt that you can refine first before putting it into Miles AI, even if Miles AI is good at helping you do the same thing.

[00:53:34]I don't know.

[00:53:34]Having, having a sense of what you like, h- your preferences, what you want the vibe to be, uh, is the, is the hard work now because the code is free.

[00:53:43]Mm-hmm.

[00:53:44]Good one.

[00:53:45]Andy, what about you?

[00:53:48]I w- I think I would echo what Zack says.

[00:53:50]Like, kn- having an idea of what you like is good.

[00:53:53]Um, I think, uh, you know, sign up for Miles.

[00:53:57]That'd be my number one piece of promotion.

[00:54:00]But outside of that, uh, yeah, like, you bring, bring what you know and what you like as a starting point.

[00:54:06]Um, and just keep... I think k- keeping an open mind and knowing that all of this changes every week.

[00:54:14]It's moving so fast.

[00:54:15]And just try, I think just keep an open mind in trying things I think is just a good general- State to be in.

[00:54:22]Um, and know that if something doesn't work right now, give it a few weeks and it'll probably work 'cause it's moving that fast.

[00:54:29]Love it.

[00:54:30]My advice, um, would be to try all the things, uh, right now, um, as much as possible.

[00:54:37]That, I mean, that's what I've been doing.

[00:54:38]I don't, I, I don't... And it's been teaching me a ton.

[00:54:41]Um, I don't know that I would have had any really, uh, good feedback for Andy unless I had been trying all the other tools as well.

[00:54:49]Um, I would've been like, "Oh, this is cool. Good job, Andy." That's what, that would've been my feedback.

[00:54:54]Um, but, uh, but I have opinions now because I've tried a whole bunch of stuff.

[00:54:59]And knowing what's possible, um, and, and like Zack said, what you want to see happen, uh, that only happens if you really get hands-on with stuff.

[00:55:08]Um, honestly, like everybody in WordPress has to build an Astro website.

[00:55:13]You just have to.

[00:55:14]You have to see what it's like, um, and, and experience it.

[00:55:17]You should absolutely build something in Lovable and ship it publicly.

[00:55:21]Like, do it.

[00:55:22]It's not ex- it's not expensive at all.

[00:55:25]Um, and it's interesting, and it, it really blows your mind.

[00:55:29]Um, try, um, Angie, uh, from Elementor for sure.

[00:55:33]Definitely try out Telex.

[00:55:34]I think not enough folks are trying Telex out.

[00:55:36]Um, but I still definitely, if I'm going to ship something with WordPress, um, and I want AI to help me with it, I'm using Miles so far, and I like it.

[00:55:47]Um, I haven't yet shipped something yet, I have to say.

[00:55:50]Um, but, um- You're welcome... I have a couple good ideas about it, so.

[00:55:55]Yeah.

[00:55:55]Awesome.

[00:55:55]Well, this has been great.

[00:55:56]Thanks so much, Andy.

[00:55:57]Uh, where can folks, uh, find you online?

[00:56:02]Yeah, I mean, I'm, I'm most active on Twitter/X.

[00:56:05]Um, so my handle is APeatling.

[00:56:07]Um, that's where I give a lot of Miles updates, um, and just talk generally.

[00:56:12]Uh, yeah, obviously the Mile, bymiles.ai is the website.

[00:56:15]Um, and apeatling.com is my blog, so that's most of my presence.

[00:56:20]I, uh, this is a free one.

[00:56:22]I'm sure you've come up with it already, but Miles Stones, um, you can talk, that could be your newsletter.

[00:56:29]Ooh, nice.

[00:56:30]All right.

[00:56:31]Uh, so if you are enjoying these shows, uh, please do us a favor and hit that subscribe button on YouTube.

[00:56:37]It's free, easy, and helps support our work.

[00:56:40]Uh, special thanks again to Freemius for being our global sponsor.

[00:56:44]And next week, uh, you should tune in again because we're also gonna do a state of episode, but it's gonna be the state of e-commerce and AI with none other than Bryce Adams from Metric.

[00:56:57]Uh, it's gonna be a good one.

[00:56:58]It's at a weird time because he, he's in Australia and our co-hosts are, um, in North America and Europe, and that's been a challenge, but we're gonna make it happen.

[00:57:09]Um, so thanks a bunch everybody.

[00:57:11]Have a good time.

[00:57:12]We'll see you next time.

[00:57:14]Thanks.

[00:57:14]Bye.

[00:57:15]Bye.

[00:57:15]Oh, wait.

[00:57:17]I lost my thing.

[00:57:18]I gotta do the outro.

[00:57:19]There it is.