Subscribe
Share
Share
Embed
For long-form interviews, news, and commentary about the WordPress ecosystem. This is the companion show to The WP Minute, your favorite 5-minutes of WordPress news every week.
How do we get Agents to use WordPress?
===
Matt Medeiros: ~showed today. Andy, your background changed, uh, to, uh I'm in a new place. Yeah. Suddenly you have- I have no idea what just happened, but there we go. I was looking at that going, "What a fantastic background this man has." Uh- It is not real. Uh, welcome to a sp- and~
[00:00:00]
Speaker 11: Welcome back to The WP Minute. Before we get into today's episode, there's a little glitch at the one to two-minute mark of this episode, but that doesn't mean you leave. That means you stick around and watch. See if you can spot the transition. It'd be pretty obvious if you're watching, uh, on YouTube, but probably less obvious if you're just listening to the podcast in the glorious form of audio in your favorite podcast app.
By the way, tell folks about The WP Minute. Tell them to search for The WP Minute wherever they get their podcasts. Okay, great episode today. Let's dive in.
Matt Medeiros: speaking of AI and what's real, uh, today we're gonna talk about, uh, WordPress in an agentic world.
Big, broad topic, and I brought some pretty heavy hitters to discuss that. Um, first I'm gonna start with my special co-host for today's episode, uh, Alex. Alex, uh, say, say hi and, uh, let folks know where you're from and what you do.
Alex Panagis: Hello. Good to be here. Thank you for having me. Uh, my name is Alex. I'm based in London.
I run ScaleMath, where we work with some WordPress product companies, and [00:01:00] also outside of that space we work with some B2B software companies. So very excited to be here and talk about WordPress.
Matt Medeiros: Brian Coords.
Brian Coords: Hey, I'm Brian, and, uh, coming to you from WooCommerce, where I work as a developer advocate.
Matt Medeiros: Andy.
Andy Peatling: Hi, I'm Andy. I'm, I live in Vancouver, Canada, and I worked at Automattic for 17 years, and for the past year I've been working on a design agent, uh, for WordPress, making great designs.
Matt Medeiros: And Mike.
Mike McAlister: Hello. Mike from Ollie, um, which is a design library editor toolkit and design agent layer thing for WordPress too, and maybe something else in a little bit.
Who knows? We'll see.
Matt Medeiros: Those really listening, uh, didn't hear the word theme, which was interesting. Uh, we'll talk more about that, uh, later on. Uh, thewpminute.com, thewpminute.com/subscribe. It's the number one way to stay connected to the [00:02:00] stuff that we do here at The WP Minute.~ Um, Andy, I'm gonna start with you be- because what I accused you of in the green room is that you kicked this off.~
Uh, so Andy, I'm gonna kick it off to you because in the green room, I accused you of kicking off this whole thing about agents and building stuff with, with agents, and it just simply not recommending WordPress. Uh, in fact, I've been doing it for a while now, and yeah, it has never recommended, "You should go do this thing with WordPress."
That's the high level overview, and I think that's what got around, got us all around this campfire. Can you just summarize your article? And then we're gonna unpack that in a discussion. Uh, for, for the listener, and I'll link it up in the show notes if you're listening to this. Uh, I'll, I will link up, uh, the article in the show notes.
Andy Peatling: Yeah, I think, well, I think you s- you summarized it pretty well there. I think there's a challenge certainly for WordPress, and I would say for all existing software and SaaS as well, to some degree, where people are now, you know, the, the, it has shifted from people going to Google, searching for a product that they wanna use or searching for, you know, a, a solution that they need for a problem that they have, and it bringing up things like WordPress.
So I [00:03:00] want a website, go to Google, WordPress comes up. Someone else has used WordPress, I get told about it, I go and install it and figure it out and, you know, I get a website. And that, that kind of word and mouth, word of mouth and, and solution to searching for, uh, products on the internet has been around, you know, it's been around since the internet has essentially existed or, or the, or the web for at least.
So I think that's changing with agents in terms of we're starting to see less of like this soft- these software islands where people go and they use their software in the island that they, they've made, and more about this centralized solution that through an agent that then connects with other pieces of software.
And as a result of that, I mean, that, that's moving. It's, I, I think it's still at its early stages. But the result of that is people are going there looking for solutions or just saying their problems, and the AI is essentially providing the solutions, whether that's third party or something that spins up itself.
And that's sort of the foundation of the challenge of growing an existing product, because if you've taken away that path, it becomes more difficult. [00:04:00] And the path to influencing those agents and, uh, being in a position where it is choosing your existing solution is quite unclear. I mean, in my post, I talk about a number of different ways that that could happen, or at least we could try and affect as, as an open source community, try and affect that.
But It's unclear. I think it's sort of, you know, it, it's early days, and I think my overall perspective on that was to think, how can we at least measure? You know, like measuring might be the first way to understand, like how much, first of all, how much of a problem it actually is, and then allow, allow the product to tr- project to try things and then see if there's actually a change in, um, the outcomes.
Is it recommending WordPress more?
Matt Medeiros: For decades, folks would Google, "How do I build a website?" And you'd had folks like, uh, some of us on this call who have built tutorials, made videos, showed people how to build things, and there were so many of us doing it with WordPress. Hey, humans were kind of answering this.
Of course, it was, you know, you had Google Search that helped sort of filter that down and [00:05:00] play that game of who was recommended, uh, on the blog article or the video. But now the agent just says, "Hey, let me just do that for you and let me shoot that over to your static site." I brought Alex here because Alex is on the forefront of search and understanding the change of the marketplace.
So Alex, uh, your thoughts on this changing landscape, and if, if you have any direct questions about if you've read Andy's article or if you've had any, any thoughts that you want, uh, these intelligent folks to, to chat about here.
Alex Panagis: I have not read Andy's article so I'm starting from a losing position here.
But what I will, what I'll turn the question back to everybody here, 'cause I'm curious. I, I build-- I come from a classical computer science background, so, you know, trained to be software engineer, and I never worked in a development role because I just very quickly moved on to... I had my business on the side, so that ended up becoming the main thing.
And now I find with myself, as with many people, getting back into coding because I have more time unlocked by AI. If you use Claude Code or any other harness to try to spin up a [00:06:00] site, why does it take the path to create the site using Astro, using Next.js versus something else? And I think the answer now is most likely that it is just the path to least resist- of least resistance.
So if it has a Next.js monorepo where it has the front end and the back end in one place, so it can version control everything, and yeah, in theory, you know, changing content doesn't depend on changing stuff in a database and also changing the front end separately. I think that's just an easier pattern for AI to work with.
I, I, I think that's largely why. Then the second thing is, I think if you take, even if you specifically guide Claude Code to spin up a l- local WordPress install on a server or if you do that locally on your Mac, you'll find it takes way longer than it will take to spin up a static site because it just has, it, it is just a more complicated process.
Which I guess is, uh, th- probably the first step, is like how can we make it as easy as possible for agents to understand how to work with the technology, get an environment up and running locally so that it's easier to do that? Because right now it isn't for [00:07:00] the agents. Question mark.
Andy Peatling: Yeah. I, I, I would just res- I think, I think that- That would be my first take as well is the, is the path of least resistance is, is like very powerful in terms of i- if the agent knows that it can get from the start to the end and something deliverable, then I think that biases it significantly.
I-- Getting WordPress sp- spun up quickly I think is like a, a big focus of the project in terms of playground and having an environment where you can spin something up really quickly and I think that can have an impact. But again, it's-- I would go back to measuring like how, how do you know it's having an impact?
And I think that, that might be the first step. I'll let somebody else talk. I think that...
Alex Panagis: sorry, go, go for it.
Andy Peatling: I was just gonna say go ahead and let, let somebody else chime in. I'll chime in. Yeah, I can.
Mike McAlister: Go ahead, Brian. Go for it.
Brian Coords: Matt will edit all this out I think one of the questions we have to ask is, you know, what is WordPress good for?
And WordPress is this kind of web application that runs on a server somewhere, and it's great for multiple users logging in and managing things, and it's great for e-commerce and membership sites, and anything where you actually need an interactive server sitting on an internet with a [00:08:00] database and all these sort of things.
And I think a lot of people used it for brochure websites in the past because it was the easiest tool at the time. But right now, you know, does the average WordPress site need all of those tools? Is that the most important thing that it needs? And so I think there's also this question of, does WordPress need to make itself easier and, and appeal to this market that it used to own and it doesn't quite anymore?
Or does it just say, "No, we know that we're good for complicated, we know we're good for enterprise, we know we're good for multiple authors and e-commerce," and that's the, the use case. And it just feels like, you know, the thing that it used to be the best option for, it's just not the best option for, so it kind of makes sense that, you know, Claude isn't gonna say, "Wait, let me get you PHP runtime and a MySQL server and, uh, you know, find a hosting company and, and do all those things."
And so I, I think it's kind of a fork in the road situation.
Mike McAlister: Yeah. The-- I was gonna say something very similar, but I'll even, I'll even kind of take it a step further in that, you know, we keep thinking about why would it choose WordPress, and then it's like the obvious answer is [00:09:00] like CMS, user management, you know, all these things.
But it's not hard to imagine a future where if things are going the way we think they're going, agents just de-develop their own, like, universal standard data models for CMS-type things, for managing users, for e-commerce even. Like, the-- There's, there's-- Right now, there, there's a reason that this layer exists for us, 'cause we built it 'cause we needed it.
But, you know, that could certainly be abstracted up, up a layer and that gets removed from the value of, of WordPress and what we do. So then it's like, what is WordPress now, you know? Is it just the biggest kind of repo of modules for this, this huge agent layer? I don't know. There's, there's big questions, for sure.
Matt Medeiros: WordPress has always had that identity crisis. The small static website, the blogging platform, the app framework. Remember the three years we argued that? Uh, you know, there was so many years on this roller coaster ride where people still don't know what WordPress is. Um, you know, [00:10:00] I, I think that heading into this agentic future, you, you might see WordPress just evolving into- probably a, a bad phrase to use, 'cause I know a lot of us may disagree with it, but the mature s- like, the mature solution, the advanced solu- the more technical solution for, uh, an end client.
I don't know if WordPress needs to get easier in order for it to compete. That, that's in my head. Like, maybe, maybe WordPress says, "Yeah, you know what? No more five-page static portfolio pages." Like, we don't, we don't need that anymore. Go build it on Astra. Do you think that's a fair assessment? Especially coming from you, Mike, selling what I'll call an end user product, who may- Mm-hmm
have considered, "I'll just use Ollie to build a five-page landing, uh, you know, landing site for my client." Like, do you still want them to do that? Or do you want them to think bigger with WordPress?
Mike McAlister: You know, we're in this weird in between where it's applicable to both. You know, I do... We, you know, across the, the Ollie customer base, we see so many personas of people who, you know, are just u- barely using the block editor, [00:11:00] to people who are, like, almost fully agentic building and also building w- with, with Ollie as, like, a, as a base for that.
So yeah, the, the, the spectrum is, is, is pretty huge. It also depends on, like, your style of workflow, because I've had people say, like, they've built a bunch of static sites, you know, prompt-based, and they're like, they get prompt fatigue. They're like, "This is actually not the way, like, physically, emotionally, that I want to build and, and spend my time."
So, uh, it really depends on kind of the user. And, you know, the cool thing is WordPress can serve all of these, these things. I don't know if WordPress needs more. I think we need less but better, you know, because so much of this stuff is gonna get chucked out when agents take over, like, a great deal of the, the interactions.
And- I think like Brian was saying, the problems we used to have, I think are not necessarily our, our problems today.
Andy Peatling: I would, I would-- I f- my perspective is that I think that's totally okay to say that WordPress becomes a piece of software that's used when your needs re- when you have the need for it, and that might be e-commerce, it [00:12:00] might be, you know, more advanced kind of interactive pieces on the site.
I think that's totally okay to concede. Concede's a strong word, but to say the needs of like a brochure site is not where WordPress is gonna focus. I think the challenge for the project is how people see WordPress as growing or being successful is like percentage of sites on the internet that run WordPress.
And if that continues to be that kind of vanity metric and the way that WordPress is, is measured, then I think that's problematic with that situation. So I agree that it c- it can c- it can concede the, the brochure style sites, but I think that there has to be a mindset shift and certainly a, like how do we measure success shift of the project in order to, for that to actually be a reality and actually happen.
Alex Panagis: I think it's equally the case though, like for people saying that static sites are on the rise, there's pro-- I mean, I don't know what, what the argument would be. We can have a vote, but there's probably an equal amount of garbage static sites that are being spun up as there are low quality or what we would, you [00:13:00] know, by our perception, would say are low quality WordPress sites.
So the argument I would make in response to Brian's point, which I think is really valid, which is maybe WordPress just isn't the choice for every site anymore, which is okay potentially, 'cause it's actually the choice for the bigger, better sites that have multi-user needs, established publishers and so on, is not every site necessarily starts on WordPress anymore.
So then what you have is you have people building static, which may eventually grow and need a CMS. So do we agree that at the moment, WordPress may not even be the choice in that case anymore in terms of like people who are truly building with AI and then letting AI make the recommendation and basically blindly accepting the recommendation?
I don't know if WordPress would be the first choice in those cases either, which is the thing that concerns me because in terms of like new user growth, it's, it's probably so far out that we'll have like another five, 10 years even before we would notice the effects of this. But if people setting up brand new sites that aren't doing anything with them and then eventually [00:14:00] start doing stuff with them, those still don't end up somehow adopting WordPress, whether it's headless or later somehow moving over to WordPress in another way.
I do think that that is a bigger problem. Like it's easy to write off to say, "Oh yeah, we're losing out to the brochure sites" but I-- that might be a bit of an oversimplification.
Matt Medeiros: Hmm. Andy, in, in your blog post, you brought up creating public benchmarks. Can you expand on that? And then I wanna ask if, if WooCommerce will do it, and then will Ollie do it?
Andy Peatling: Yeah, I mean I, I've touched on it a bit earlier. I think that, I think that it's important with any challenge to, to measure. I mean, o- otherwise you don't know How effect-- You don't know whether what you're working on or doing is actually having any effect on the outcomes. So the benchmarks are really a way to, to, for a few things.
F- the first thing would be just to track, first of all, if people do use WordPress with agents, are they having any success? Like what-- Are they able to get an end-to-end site built? Are they ab- able to achieve the outcomes that they wanted? And if they aren't, where are they falling [00:15:00] off? That becomes then a, a body of data to feed back into the project to improve the agent experience of WordPress.
I think there's like a, there's a cycle there, a loop that will ultimately result in WordPress getting better with agents. And then the second one is, like, a, a platform selection benchmark. So everything we've been talking about, whether that's, you know, coming in and starting a site or whether it is, like, a m- perhaps a more important piece, does, is WordPress recommended when you want more features on a site, and then it's expanded out.
So a b- a, a benchmark of, of providing all those scenarios to popular agents and then measuring what, what do they suggest? What happens when you say XYZ, and is that changing over time? Um, and then you can see if the work that you're doing is actually affecting that. Um, and then, yeah, I think just another one to track whether agents can actually do the things that are deep into WordPress.
Can they access certain areas of the product? Can they, like, ca- can they get down to the depths of the product that allow them to be successful in actually doing the things that people wanna do? So there's another piece there, and that would also feed into that loop of improving the product [00:16:00] too. So it's measure-- it's, it's, it's b- it's both public measurement and, and, and showing that progress is important, I think, in this area, and also, uh, improving the product in those loops.
But then ultimately that data, I think, can also be picked up by the agents too, and see those kind of like pathways to doing certain things with the product, and then use that as training data. That's more speculative, speculative because it's hard to know actually what affects the agents in the training data.
But I think you've gotta start somewhere, and that may be a place to start.
Matt Medeiros: So maybe a better question for Brian and Mike is, are, are you, are, are-- And I know, Andy, you probably already do this, probably where the idea came from, especially with Miles, the accuracy, the, the, the design, how you enhance your skill, like your skills that the agent uses.
But curious, WooCommerce, Ollie, uh, I'll kick it to Brian first. Are, are you all tracking this kind of success and creating an internal benchmark for what you can talk about that hits on Andy's points?
Brian Coords: Yeah, I mean, I think there was, like, two sides to it. One was just, like, discoverability. Are we showing up as a [00:17:00] solution?
And obviously, that's a huge focus for, like, our marketing team. Like, are, you know, are we a recommended option? Obviously, we have a different, a different user than, you know, the average WordPress user. We have, like, different competitors that we think about. We're not really thinking about other WordPress plugins as much as we are thinking about, you know, a Shopify or something like that.
And so that, you know, I think we're in the same place as everybody. We used to live riding high on that organic traffic, and we have a pretty strong base of that. But obviously, that's changing a lot of stuff. So there's a ton of internal work on that discoverability, making sure we're recommended, making sure we're recommended to the kind of people that would succeed with WooCommerce, which is also not every single e-commerce store.
But then as far as the actual experience of managing the software with AI, that's one of the things I've been really digging into recently because I think that's the other piece that we've kind of, I think we hinted at, and I think Alex kind of hinted at this too, which is like, there's just a lot of pieces.
There's a lot of ways to do these sorts of things, and it's very [00:18:00] unclear. Even internally in WooCommerce, it's like, "Okay, well, do we need an AI product?" Which we beta test. We did beta previews at WordCamp US. Well, that's AI in your dashboard. But what about MCP? Do we want people to bring their own agent?
How much do you give away for free? How much do you try to do? How much of open source works better when you proxy it through, like, your kind of own servers and you add some special sauce to it? Does the open source community like when you do that kind of stuff? Not really. So there's all these different factors that we're thinking about and trying to kind of get, like, a clear narrative.
But I think the end goal has to be what you're seeing with a lot of these other tools, it-- which is people wanna show up with their agent, their bot, their thing, and they just want s- to be successful through every action. And so I think Andy's framework of like, are we measuring these internal actions and how easy they are?
Is really important, and I think some of that is gonna come down to the architecture of Gutenberg and this way we've really scattered everything across databases and file, theme files, and all this other stuff. And I think there's gotta be some cleanup there [00:19:00] because it's, it's really hard to, to watch an agent and sell it, "Okay, I need a banner on my header that promotes my Black Friday sale," and the agent's going, "Well, where is the header?
Is it in the database somewhere? Is it in this theme? Can I change the theme? Oh, wait, I can't change the theme 'cause it's connected to a .org theme." Like, it's just, you know, you're watching it just burn tokens trying to solve what is a basic issue on any other platform, which is like stick a banner at the top of my website.
So I didn't think too much about the benchmarking specific user, like, flows and stuff, but I think that's actually, uh, a really good way to frame it. I don't know if I answered that.
Mike McAlister: It's a little different for, for me where, you know, having an established product and the kind of product that it is, implementing AI's been a little more iterative purposefully.
One, I'm not in a rush to, to dictate what, you know, kind of thing I should jam in there, even after talking to people, 'cause people just are all over the place. So the next year is gonna be totally crazy. So I'm, I'm, I'm trying to build features that are relevant today and thoughtful and smooth out the WordPress experience no matter what happens.
These things have to be smoothed out and, and so e- even on the d- [00:20:00] the, the design side, like right now our, our agent is, is trained to go fetch one of the many awesome patterns we've pre-designed so that it doesn't have to design i- in, in some kind of, you know, mediocre Claude-looking site. It can just grab a really good design.
It's cheaper, it's quicker, it's, it's just, it's just built for how we, our, our product is currently. But, you know- I see and hear Andy, he's like way deep into-- 'cause he's got like a new product in, in AI. And so I am, I'm learning a lot just from watching kinda what he's doing and figuring out like, you know, where is, where is Ollie growing this and just a-as a WordPress product in this era.
So yeah, it's a strange and interesting times for sure.
Andy Peatling: You know, I say there's, there's a ton, I mean, there's a ton of unknowns. You know, like I think we can, we can all speculate about where things are going and we can, we can speculate about what we think will help. I think that, that w-we have to remember this is all very new and changing quite quickly.
So you know, what, [00:21:00] what-- from my perspective it's like what are the fundamentals that, what are the building blocks that can help WordPress and can help the experience around agents and the experience of using AI with WordPress? Like what are those fundamental pieces that don't like lock you into like a specific way of things working?
You know, it's, it's easy to-- we traditionally as w- you know, application developers on the web thought like what's the UI? And then like bring it down from that level. And I think kinda starting at that level is difficult with the way that things are changing, and it's-- you almo-- you have to sort of start more bottom up now.
Like what is the, what are the funt-fundamental pieces that can, can enable like potential future interfaces, future agen-agentic interactions? And that's a hard mindset shift, and I think like all product people like I think is, uh, you know, are dealing with that at various positions, like established products.
It's easy for me with a new product to kinda like talk about that stuff and think that way, but I think it's, it's a challenge for everyone and, and the pace of change especially.
Mike McAlister: That mindset shift is so hard. Like even- Mm. -years into this now, and I, I think about it [00:22:00] still every day, and still I catch myself and have to remind myself- Yeah
of that kind of these paradigm shifts. And it's just even the, as a longtime builder, it's something really hard to wrap your head around, but it's, it's, it's crucial to do it.
Andy Peatling: Yeah, I do it every day as well. Like I j- just today, I, I was like, "Oh, no, I'm thinking about, I'm thinking about the web interface, and I should be going down at a lower level."
Like, it's, it's, it, like it, it's, it's 20 years of, of like, you know, muscle memory. It's hard to work through. Mm-hmm.
Brian Coords: But it, it feels like that is one of our advantages that I feel like WordPress should lean into, which is, like, there is a visual editing component, like Mike was saying. People get a little prompt fatigue, and I think the fact that you can do things like click around and just mess with stuff sometimes, I think, like, Gutenberg itself, for all of its, like, pros and cons, is almost, like, the one piece out of WordPress where you're like, this, this thing right here, this ability to just look and click and, like, have a strong representation of my content visually is, I think, the thing that will maybe become the [00:23:00] competitive advantage.
Right now, but architecturally it's very confusing. Full site editing is very confusing. But there's, like, a gem of a really good idea, which is you've given me, you've created something, and then you've given me a nice visual editing interface that everybody is familiar with that's not like Claude just came up with its own little version.
And I feel like there's gotta be some, some strengths in WordPress that we can lean into, and it might be in places like the block editor, which maybe people wouldn't have considered a strength, but I think is very quickly becoming a, a strong strength and selling point for WordPress.
Alex Panagis: I 100% agree. I think, like, the visual editor is one big component of that.
So if you're working with, let's say, a future WordPress CLI or with Miles, that you could have a block li- library or a style guide or a brand design system, which I think is already how Miles works, and it just uses and reuses components. 'Cause if you build with- Just cloud code and let it build static sites.
It will regenerate and recreate a component 100 different times until you have like 100 different versions of the same component. But then also on the-- if we focus [00:24:00] like purely on the CMS side, so like the people working in the CMS and not designing the site, I think there's an increasing demand for an interface that people actually want to spend time in and are able to review the massive amounts of data that we have because of AI.
It's like people, you know, you know, regardless of like how, what your view is, people are writing and editing with AI, but they ultimately still want, and at least in my experience from like what we do as a team as well as like the companies that we work with, we do want an interface that makes it easy to work with the fact that we're probably doing 10 to 100 times more or are expected to do 10 to 100 times more than we were pre-AI.
So I feel like that the, like interface layer still remains im-important, and I agree with Brian. I think it will eventually, and Remcus talked about this in your recent episode as well, Matt, that it's just a matter of time before it catches up and the pattern becomes really obvious because while starting and doing certain things in the CLI or via cloud code is really nice, ultimately you still want to then be able to edit something without having to prompt and wait, you know, 30 [00:25:00] seconds to change some text on a page again.
That-- Whether or not that changes, I don't know, but that, that's my view at the moment at least.
Andy Peatling: Yeah. And, uh, and I think I wanna add on that. I think that's a great point. Like, uh, when I said like interface down into the pieces, it's not, it's not throw- necessarily throwing away those interface pieces. It's, it's also decomposing the interface pieces into those foundational chunks and thinking more like how do you compose them in a way that works with the agent, like, you know, exposes...
Maybe it, it exposes the editor at the right time, like in context of what you're doing or maybe a specific post list or, you know, there's a, it's a, it's a, it's sort of the pieces being both like tools, prompts, interface elements, like how do they decompose into something that can be recomposed into a, an experience that works with agents in that way.
But like to-- I, I 100% agree that the interface, it's not all chat. Like I think that's-- There's many, many situations that you've already mentioned that where chat is very inefficient and, um It, I think, I think that's where the interface pieces 100% they fit in there
Alex Panagis: I have a question for the three of [00:26:00] you, actually.
Sorry to cut, cut you off, but I'm very curious. Building with trying to make your products, or, and not trying, but succeeding at making your products AI compatible to some degree, do you think that the WordPress architecture, WooCommerce architecture, or Gutenberg architecture in general poses some, like, fundamental limitations to getting there?
Or do you think it is just... Like, how achievable do you think it is that the core project moves quickly enough, and I think that we have the time, so I'm not pessimistic in this regard, but that it moves so that it becomes way easier for Miles, it becomes way easier for Ollie, and it becomes way easier for WooCommerce to have an easy CLI to work with or to make it easy for agents to work with?
Brian Coords: I mean, I think that, that's kind of where I was gonna go, which is, like, before I worked at WooCommerce, I worked in an agency, and we were early to block inside editing. And like you said, the, the user interface is important, but the architectural decisions underneath it were also very important. And it's just interesting because a lot of the things at the agency level that we were struggling with, with adopting the block inside editor, these are [00:27:00] the exact same problems that agents are having with it.
Like, it's, we're, it's not really new situations. You know, Claude is just a developer, and he's just trying to do the things that we used to do four years ago with the block editor and having the same problems, like you said, a s- single source of truth A really composable architecture, like all these little pieces where it's like, man, this was really hard from a developer architecture level, version controlling things, deploying things seamlessly across environments, all those problems.
And so I think more than anything, those things are just coming more into focus now. The fact that we're seeing an AI just basically speed running our own development problems we had with block themes and block editing in the past. And so I feel like those things are, are, are becoming more clear and more important, and we are just gonna have to solve those because the actual editing interface and of the block editor is honestly really good once people get used to it.
It's just when you put a developer or an agent in front of it, it just slows things down.
Andy Peatling: I, I, I'll-- I've been doing a lot, I've been getting involved with the core [00:28:00] project a little bit more just over the last little while, uh, basically since I've wrote the post, just looking at the foundational pieces that need to be in the core project in order to, for agents to work better with, with WordPress in general.
And I think, like, that all that work is happening, and, and I think it's, there's some really excellent work around the core foundational pieces like abilities and looking at MCP in terms of the core project. Those pieces are all happening. I think where, where more work needs to be done is, is, is pushing it, pushing like a perspective and, and a, an opinion on when that should be in the core project and not sort of worrying.
I, I know there's a lot of, there's a lot of history, history in WordPress of, of getting things in too early and then it being difficult to, to pull out or change. But I think that there's a position here where some of these a- fa- foundational agentic pieces are becoming more t- more mature, and the market is speaking in terms of the direction that it's going, where WordPress can kind of put the foot down on the accelerator a little bit and get some of these foundational pieces in sooner than may [00:29:00] be comfortable.
Obviously cautious, but, but faster. So my sort of, my perspective there is to, just to jump in and try and help with that and give, like, perspective as a, as a product using a lot of those foundational pieces that exist right now. But I think for the, part of my post was for the, for the, for the w- project of WordPress to continue to, to move forward and be relevant in that agentic world is to try and, like, be more opinionated and think about how those pieces can get in the core project.
So that would be my take on, like, how to, yeah, how to make the core project better for it.
Mike McAlister: Yep. One thing that I just saw kind of going on Twitter the other day was the conversation around the abilities and how they, they're kind of limited right now, and there's no standard, and now plugins are kind of sucking up all the abilities, and it's gonna create this, this, this insane thing to deal with.
So like Andy's saying, like, an opinion here and sort of a, a standard, even if it feels early, I think is important to just at least get a hold of the, the experience and kind of outcomes very early. So these are [00:30:00] things that I know they change every day, and there's, there's gonna be some things coming in in 7.2, some, some more AI, MCP, and, and abilities stuff.
But yeah, I agree, like, an, an opinion, uh, I think is really important to help shape this as, at the speed we need to.
Brian Coords: Yeah, I wanna second what Andy said about the core AI team. I think, like, when James LePage came in and set it up, I think they made a lot of really smart decisions to focus on fundamentals and to not just be, you know, "Hey, we built a W- WordPress MCP, and everybody has to use it."
Like, they've done a really good job. I think abilities API as a concept has, I think, been really successful. I think we're seeing it adopted even when people are offering their own MCP tools. It seems to be they're using that framework underneath, and I think that's gonna pay off in the long run. But yeah, it's, we're probably turning the corner where it's time for some, some of the disparate experiences of MCP or any other sort of connection should probably come in a little closer.
There's a lot of interesting stuff happening in [00:31:00] WP-CLI for, like, the next version that I think it's like I'm ready for it to come out. And so, like, the foundational approach I think has actually been really successful, and like you said, we, we have backwards compatibility. We can't just merge everything in and stuff.
But now that we've seen some of it, there's probably, like, two steps of opinionation that we could go forward in a little bit more to bring these together. But the fact that the ecosystem is coming out with all these tools and we're seeing wide adoption of this is, I think, a good sign that it's, we're headed in the right direction, and we, you know, are, are not totally losing ground here.
Matt Medeiros: I'm wondering if it's chicken or the egg kinda thing. Like, I, I hear a lot of technical solutions here, you know, advancing Gutenberg to be much more agentic ready and, you know, things that WordPress can do at a technical level. What do, what do you all feel about, I guess, on the marketing side? Uh, because, you know, we're seeing that agents aren't recommending WordPress.
Yes, one, it's technical, but two, is it [00:32:00] A marketing and SEO problem. How do we retrain agents to say, "Oh no, you should work..." Like, if we made all these changes today, magically Gutenberg is great for a- agentic, how do we get today's new search engine to understand that and say, "Okay, user, go ahead and use WordPress"?
Like, where do you, what do you think we need to do as a community to enhance that adoption and that recommendation, if at all? Like, if you have any thoughts on that.
Alex Panagis: I think if the UX gets to the point where it's the best choice, in many ways this is counterintuitive to, from a marketing point of view. I think much of it will happen naturally because WordPress already has such a big community.
So I think if it can get to the point where the user wants to reach for... They may not know they want WordPress, but intuitively what they want is they want to reach for WordPress. They're gonna get that there by AI recommending it to them, and then once they see it, it is actually what they want. So let's take a moment to also think, let's say WordPress would be currently recommended, but then somebody ends up in the WordPress dashboard and they're like, "Whoa, [00:33:00] this is not what I wanted.
This is not what I expected." So they end up bouncing out and saying like, "No, I actually don't want this as a CMS. I want to use some other headl-" you know. Again, right now we're assuming that it-- Let's say all of the problems are solved, so it doesn't end up being some, like, headless setup where you have the CMS and the front end separate.
But I think, yeah, I think the main thing is if it gets to the point where it is as easy to deploy... And perhaps the answer is that, that, my, my follow-up question is from an architecture point of view. Right now, if you use CloudCode to deploy a static site or a site that, that is not built with WordPress, it does so and it, in theory, lives at the ser- at the level of the hosting provider.
So as opposed to if you run an agent in WordPress, which I'm sure poses interesting issues, 'cause the, in theory, the agent can break WordPress and then not be able to fix it 'cause it runs inside WordPress. Whereas if it has just root access to the server, it would be able to, to do anything. So I think the answer is the UX has to be there in order for people to want to choose it.
And then, then comes the argument, which I think, Matt, you're kind of alluding to is, do [00:34:00] people actually want to have the dashboard? And I think then that's, that, that comes back to Brian's point about the user that we're targeting. Right now it is more the established company that cares about having a CMS where they interact with the, the interface.
But if it's somebody that's just, you know, doing some, like, programmatic SEO experiment and they want to spin it up with 1,000 Mar-Markdown files, then they're not gonna use WordPress, they're gonna use Markdown because it's probably always gonna be the easier way to do that. But that's not... Yeah, I don't even think that that's a valid market to go after personally, but-
Brian Coords: There's also like-
Mike McAlister: No, go for it
Brian Coords: Yeah, I was gonna say, I think that when you say marketing WordPress as a whole is always gonna be tough.
You know, if you market Next.js, there's a hosting provider that you're going to get funneled to. That's by default. If it markets any other of these things, there's generally a place it's gonna send you. With WordPress, they're never just gonna send you to one place. There's gonna be some decisions and stuff.
And like you said, some of these more interesting AI applications are not things you run in WordPress. There's things you run on the server. We have, [00:35:00] like, a lot of internal experiments in, like, the wordpress.com side, where it's, like, an agent that sits on the server next to WordPress, and you can communicate it with, through Telegram or whatever, but it can do things like if it crashes the site, it can fix it because it's not tied to that.
But that's obviously, like, a scary technical thing, just like throwing an agent, like, directly on your server and stuff. But a lot of these things, I think it's just gonna be the WordPress scenario that, like, hosts are gonna come in and solve some of this, plugins are gonna come in and solve some of this, and like Alex said, like, there's probably not gonna be, like, a s-super strong opinion from WordPress because a lot of these will come in at different layers from the ecosystem, and that's how it's always worked.
That said, I would like to see wordpress.org have a little bit more of an opinion on AI and at least make a clear spot like, you know, agents. If you're looking for WordPress AI, here's at least some guidelines. Here's at least some starting points and at least some places to get started. 'Cause right now, if you go to wordpress.org, AI just kind of doesn't exist.
It still lives, the site lives like it did five years ago, and you gotta really hunt and find to even know if WordPress [00:36:00] has tools or opinions on AI. So I think we kind of do need to do both. But yeah, the, the ecosystem is gonna be a big part of it.
Mike McAlister: Yeah. That's pretty much, that leads into what I was gonna say, where we do have to deal with the kind of the enthusiasm gap in our industry, in our ecosystem, and outside of it.
You know, I think if, if I was to land on wordpress.org, and I wasn't super familiar with it, and I'm trying to learn about WordPress, and I'm trying to learn how to get WordPress, I think I'd be like, "Forget this." There's no, you know, it's not... It just needs some work. It feels more like a meeting than a party, I would say WordPress does.
And I feel like, you know, you see these other ecosystems, I, I always, uh, point to Lara- Laravel as an example as, like, they're an open source, you know, uh, platform and, but they have, like, a pulse. They're kind of exciting, and they do kind of, you know, cool videos, and they have, like, personalities and- These things do attract people.
People of in-influence, people who have big followings, and then those people come. So, you know, there's the whole agent layer, which is how we, [00:37:00] we acquire users, but then there's also, like, the real pulse that we have to have as a, as a, as a platform. And we do have a good story to tell. We're just not really telling it, and we, we have so much to, to be bragging about and to, to kinda help us grow this story and carry this thing forward a little further.
Um, but I, uh, we just have to get serious about that. I know, Matt, that you've, you've kind of, you've been talking about that for years, kind of the, the WordPress story, and yeah, we're, we're still not quite telling it. It's not the-
Matt Medeiros: It's not the cheerleader you wanted, but it's the cheerleader you got, okay?
So- Yeah. There, there you go. Sorry, Andy, go ahead.
Andy Peatling: Oh, I was-- I just-- I, that's an interesting point. I think you're absolutely right, Mike. I just-- I would just ask why. Why, why is that the case? Like, what-- I think there is, there are people with the enthusiasm, and I think we do have a great story to tell, but why is that not happening?
Yeah. Yeah, it's- I don't have an answer.
Mike McAlister: Yeah. Yeah. I don't, I don't either. I think, like, treating WordPress.org maybe differently than we have previously would be a, a huge step forward in that it needs to feel like a living, breathing hub for developers, builders, and agents and, like, [00:38:00] really, like, like Brian was saying, like, let's have an opinion, and let's just put it out there and say, like, "No, we are building an agentic CMS frontier platform here."
Are we or are we not? If we are, then let's say it, and let's kind of start building that part of it. Um, but it, it starts at the homepage of the, of the product, and that is, uh, currently WordPress.org. So something has to happen there.
Alex Panagis: I think, I don't know which of us was that said it, but the f- the, the dichotomy b-with WordPress.com and WordPress.org is definitely a part of why that's the case.
Like, if you'd see a case study of a site on WordPress, you're way more likely to see that on the hosting provider's website on, on, you know, for, for actually hosting the site or even on WordPress VIP, I think they do some of those. But you'd not see that on, on WordPress.org. Like, if you compare it to the Next.js project, you similarly would likely m- see that on Vercel, even though the-- yeah, as the hosting platform, whether that's because they use Next.js for a small part of their site or they may also use, like, some whole other, like, host of technologies.
I think that's [00:39:00] Part of it is just this is the nature of open source, that it's harder to market. But I agree that that's kind of a lazy out for the people that have been in the industry for a while. We've kind of gotten used to it and almost immune to it, and I think it's a good, a good arg- like, a, a good initiative to be like, "Well, maybe we could still.
Like, yes, we are open source, but there's probably still something more that we could do," which I agree with.
Andy Peatling: I think there's a, there's a... I touched on a little bit in, b- bit on this on X, and just saying that in the, in, in the world where people come to WordPress through hosts, like, they, they, maybe they start on a one-click WordPress install.
That's their pathway into WordPress. That, that can h- that can be good for the project in that you've got, you know, 10, 20, 30 different entry points for WordPress there, and, and, and the, and the cohesive whole becomes growth for WordPress. But where, i- if you're in a situation where there's one central entry point to WordPress, you lose the cohesive whole of all of those different hosts.
You know, the, the, they don't have the reach individually to grow the project in a world where people coming at it centrally. [00:40:00] So I think there's a kind of a, I don't wanna say a reckoning, but, like, a real, a real, like, danger point there, where what worked before with the decentralized method is possibly not gonna help WordPress in the future with this more centralized way of people coming in.
And I don't know what the answer to that is, because WordPress is built on the community. It's built on the hosts. It's built on the collective growth. But I think that there has to be some thought around how that can be translated into a s- more centralized position. And I don't mean the hosts go away or the, the community goes away.
I just mean, like, how people can work together more to, to improve that centralized path going forward.
Brian Coords: Yeah, I, I think the, you brought up, like, the .org versus .com thing, and comparing it to other open source frameworks, it's like they don't have a .org and a .com. They have a .com, and the founder of the open source runs the .com, and that's the place you go.
And people in WordPress don't like that there's a .org and .com, but it is the one thing we have, which is, like, we have two separate places. And [00:41:00] so .org isn't just a huge funnel for .com the same way that laravel.com is just a huge funnel for Laravel Cloud. And so for better or worse, we do have two separate things that kind of give all these other hosts the opportunity to also do their own funneling.
But I think, like Andy said, that was when people were coming maybe from the host to build a website, and now they're going to Claude to build a website, and the hosting is sort of a secondary place. And so some of these things will probably have to shift. And like any other problem with AI, everything we're discussing on this call that sounds hard or weird or whatever, in two weeks it might be like, actually no, the models can do that too, so it kinda doesn't matter anymore.
Yeah, of course. And that's- Yeah ... the other hard thing, is which, there's so many priorities, and you go, which one, it will still be a priority in three months, and which one is kind of solved by intelligence progressing a few more steps? Uh, and that, that's, I feel like, the hardest part we're dealing with.
Matt Medeiros: That's, that's one of the topics I, I did wanna talk, but I, I know we're running up against the hour, and Brian does have to leave, and I don't wanna h- exclude him from that conversa- Oh, you, you got like a few more minutes?
Brian Coords: Yeah, yeah, [00:42:00] I moved a call.
Matt Medeiros: All right, cool. Uh, so let's talk about that. I wanna talk about the cost of, and get your feedback from everybody, the cost of that intelligence and of that centralization.
If we're looking at centralization, like I go to my agent, I say, "Please build me a website," and, "What do you recommend?" The cost is, number one, we're already losing it for W- W- WordPress. It's recommending other stuff, right? For all the reasons we've already talked about. So there's a, there's a cost right there.
We're, we're... for WordPress. We're losing that. What about the cost of intelligence that we're just like, "Ah, the model does it," and what do I do? I pay the model to do it. Are we all in the, of the, of the... How do you all feel about that cost, of outsourcing that cost to a model to do the work for us? And do we think it's getting cheaper?
Do we think it's gonna get more expensive? And why is the future of, of WordPress an open source model from Automattic that builds it for you for free? 'Cause that's, that, that's my, that's my running theory. What do we all think about the cost of outsourcing intelligence? Alex, let me give it to you first.
What, what do you think about the [00:43:00] cost? Do you think it's gonna go down?
Alex Panagis: I think, uh, I've thought about this a lot. I, I kind of, when I'm speaking to people outside of the industry a little bit, outside of our little bubble And I think there's a part of me which again, I hesitate to say this out loud because as Brian points out, this is so subject to change.
But I think three weeks ago I said something along the lines of, could there be a world where there's people that can afford frontier AI or more expensive models, and other people that can build slightly lesser than sites because they can only afford the cheaper AI models, because they can't afford to spend five hundred dollars in tokens or a thousand dollars, they can only afford to spend, you know, twenty dollars.
I feel like that is short-term thinking, and I don't think that that's where we'll end up in five years. I think that-- I feel like that's unlikely, but it does-- There's a sense of part of me that does feel like we could end up there. And then the other part is open models seem to be gaining traction. So I think there's a reality where the idea, and this goes back to the point, Andy, you were making, of like, is decentralized nature of WordPress a [00:44:00] disadvantage?
I think right now it is, but I-- perhaps that can shift where people actually realize, wait, we don't want to be beholden to a Lovable or a Replit. We actually like the fact that I can rent a box somewhere in Germany on a data center or wherever from Hetzner, and then I can just dump whatever I'm working on on that box.
Could be ten WordPress sites a-along with other sites and, you know, whether that's the ideal architecture for enterprise, definitely not. But I think there is, there is like a place where it will get to that and the cost will become increasingly cheaper. So that's my view.
Mike McAlister: Yeah. I, I think I've mentioned this on a few shows now, but I did a, a survey kind of when we were thinking about some of our AI stuff, and one of the things that kept coming back that I kind of assumed, but it kind of was validated, this is anecdotal, so take it for what you will, but cost came up where there was people from other countries, you know, South America and other places in Asia where they were like, "We actually are not really even using AI right now because we just can't afford it."
And, you know, I think that's a [00:45:00] huge thing to think about with WordPress and the kind of global user base we have in that, and also we are, all of us here on this, in this chat are in a huge bubble. Like, uh, you, you go outside and it's, it's just not the same. So cost is a huge thing, and that's one thing that's currently being dictated by so many different factors.
Global politics even are weighing into how much the AI is costing us. So like, you know, it's something to be super mindful of when you're thinking about like, well, what level am I investing into my product? And then if I have to like offload some of the cost to the user, you know, now you're kind of having to raise and lower prices potentially on them.
So yeah, it's, it's, it's, it's something to think about.
Brian Coords: Yeah, I, I wanna call out that Matt said that Automattic should make AI free for everybody in WordPress and that I-
Matt Medeiros: I was gonna, I'm gonna expand on my theory. I'm gonna expand on my theory, but-
Brian Coords: I, I, I do think this is an important point. We, we saw this a lot when Core released the Connectors API, where you could just put your API key in in the back end of WordPress, and one comment that [00:46:00] really stuck with me was someone was like, "Whoever designed this feature clearly works somewhere where you don't have token limits," you know?
And fear of how much you're spending on AI, and I think that is true that we are all in this bubble. I also, though, am pretty optimistic about open source and open weight models. I'm pretty optimistic that this is going to be something that's valuable enough that we'll see it start to get cheaper and more subsidized.
I, I think it's less the AI and it's more the compute that we're sort of worried about, but I'm also pretty optimistic about that as well. But I think that one of the things that WordPress can really focus on is really being model agnostic, and I think that we, we really... I think that's another good thing of the Core AI team, where it's like we're really not trying to tie people into one of the big three providers, and we really are trying to find ways.
And I feel like we have a huge responsibility to the open web to, like, continue that and to really push forward this idea of, you know, most of what I'm doing in WordPress, I could use a model from five months ago and it's probably gonna be [00:47:00] fine. Like, it's not a big deal. I could a- absolutely be using open weight models to do most of this WordPress stuff, and so I think we should, uh, continue to, like, do that 'cause it's just, it's just another environmental sort of, like, sustainability thing that we need to do for the project in general, which is, like, continually make it cheaper and easier for people to do things and not proprietary.
And so I think it's just, it's aligned to the mission. Yeah.
Andy Peatling: My, my quick take on it would be I think that it's a temporary situation where we're in a window of time where the foundation models are passing through software development and all the things that we do day to day, and they're gonna p- continue to proceed well beyond that.
And behind are coming all of the cheaper models, open source models, and are, are kinda getting dragged along with it into that window. And now we're seeing our window of usage, the open source models are passing into it. And I think they're just ultimately gonna get good enough to do everything that we need in terms of software and day to day.
Um, and that cost will, I hope, drive down to zero, where you can run it locally on [00:48:00] very limited hardware. And then the foundation models will continue to improve and improve far beyond anything that really concerns us day to day. Where that ends up, I don't know, whether that kinda ultimately consumes everything, but I think it will drag in like we're seeing, cheaper, smaller, faster models that we can use day to day, and that will be our, yeah, that will be what we use.
I hope that goes to zero. I, I do, I do actually legitimately think the cost of using AI models for what we do day-to-day will go to zero. You'll just run it locally. But we're not there. Obviously we're not there yet, but it's, I think it's on its way.
Matt Medeiros: Yeah. The, the local thing is something that, that I've just been doing a lot of like experiments locally and, and researching w- w- the capabilities of these open weight models.
But the, but the WordPress, I mean, this hype bubble that we're in It's so anti, I don't wanna say it's anti-WordPress, but the beautiful thing about WordPress is you can put it on any web host you want. You don't like this web host, move to the next one. You don't want this theme, switch to the other theme.
And it's l- but for all the value we've talked about today, it's like use three big frontier models that you have to pay for who are like [00:49:00] mega corps. And it just seems like so- Yeah ... like we've just outsourced like all the stuff we love about WordPress to three tr- tr- three trillion dollars worth of, of companies, which feels weird.
And like, as I was testing Studio the other day and some of their, uh, advancements with, you know, chat to build websites, I ran out of credits, and I'm like, "What even is a Studio credit?" You know? Now, now I'm having to do math against the math of like what I'm paying, you know, for my, uh, subscriptions and, and I was like, man, if there were just a world where we had a distilled model for just WordPress that I could run locally, right, on my MacBook Air, and it just did the build-out of the website, even if it was just smart enough to scaffold the website in Gutenberg, like Brian was saying before.
Just add a banner to the top of the page. You shouldn't have to read the database to do all this stuff. Just go ahead and do it. To me, that's an interesting position to be more aligned with WordPress, the open source project, where it's like, here's this open source model, super lightweight, [00:50:00] just for building WordPress sites, run it locally, and it builds it.
I'm not even saying design, just build it. That'd be a pretty cool place and something that I think aligns more with what we've all been standing on for a foundation of open source for 20-plus years at this point. So that's the crazy conspiracy that I have, uh, for that. We love to see it from the corporation.
Andy Peatling: Actually, can I, can I, can I say, can I s- can I say something about that? I, I, I think as an idea, that's really interesting. As an experience, I don't know how that would f- like if WordPress has a model, then that means you're saying that every other piece of software has a distilled model, and every specific task has a distilled model.
Like I don't-- how does that fit into an experience? Like if I, you know, if I'm building a site, I might use that model, but then what if I want an image, or what if I wanna go and check, like bring in more advanced functionality into that site? Like, are we saying that the WordPress model is gonna encompass all of that as well?
Like what, what is the... I'm trying to like bring it out, what, what is the u- like how would you use that? I think it's a good question.
Matt Medeiros: Andy, I have a career in being just the grenade guy. That's what I do. I walk into the room. I go throw this [00:51:00] at the smart people, and I go, "I'll see you later."
Andy Peatling: It's a good grenade.
No. It's a good grenade. No, I mean, it's, yeah. I think it's really interesting. I think it's really interesting.
Matt Medeiros: Yeah. Yeah, I, I, I'm looking at it f- from the 50,000-foot view of... Because I, as much as I love AI, I'm still like, ugh, I, I'm outsourcing all of this to these, to- Yeah ... to a closed source system, and that doesn't align well with me.
You know, for example, like when abilities options page came out, abilities API and, and the options page came out, it was, you know, Google, Anthropic and, and, and OpenAI. And to me, I was like, "This is the craziest thing I've ever seen happen to WordPress." Like for me, 'cause now it's like three massive corporations are inside WordPress, the open source project right now.
And to me, I was like, I've never seen anything like this. I for sure thought it would be Automattic to be the first, like, commercial play, bright and obvious, uh, like those three people were. Um, but or three entities. I don't, I don't have an answer for that. I mean, that, that would be something that is a, a great thought experiment.
Is this something where .org is learning and .org, I mean, [00:52:00] .org is learning how all of these plugins and themes operate. Does Mike have to include skills with Ollie when he, you know, up- uh, uploads it to, to wordpress.org, and is that how it, it's learning all this stuff? To me, I think free and open source and a free and open source model you can run locally without a cloud compute cost is where WordPress aligns well.
I don't know how we get there. Uh, and you're right, I don't know what the experience is, but Automattic could certainly train a-- I mean, lots of people are training models these days. I'm sure Automattic could train a model just for WordPress.
Andy Peatling: Yeah, yeah. Yeah I think we get, I think we get there. I think we get there.
I mean, for me, I think we get there with the open source models that, that are, that are gonna exist and will exist in the future. And I think the, a s- like a very, a specifically trained WordPress model is interesting. My only query is just how, how that functions, how that would function to, for a user, like from a, from a usability standpoint.
You know, do you have multiple agents for different models? I, I don't wanna get into the weeds on it, but- Yeah. ... I think the, [00:53:00] the, the, I, what I agree on is, like, bringing the open source models up and, and using those and as, as opposed to very expensive foundation models that probably do way more than you'll ever need- Yeah
in your day-to-day.
Brian Coords: Yeah. And that, that is- Yeah ... something internally that, you know, there is... We do have our own hosted open source models, and we, uh, they're constantly benchmarking. I mean, we use AI for support, obviously, support queues, and Studio, and harnesses, and the AI website builder on wordpress.com and stuff, but they do absolutely benchmark open source models and start plugging them in.
And, you know, obviously there's a harness that's meant for their workflows and stuff, but it is getting a lot closer and cheaper, and there are spots where it's like, yeah, actually Kimi can do this, and we don't need Anthropic. And so I, I don't think Automattic, I don't know if Automattic's gonna, like, service, you know, like that.
You have open code or OpenRouter or whatever to do that. But, like, the models are catching up absolutely, for sure, and there's a lot less need to go to these big three providers, which is also gonna affect the economics of [00:54:00] everything as they start to feel threatened, you know? Yeah. But I, I do think Andy's point, it's, it's the product experience.
Like, I don't use an open source model locally because I love the experience of opening Codex, and it's a nice UI, and it connects to everything, and it's, I'm working in five contexts at once, and it's pulling in those things and stuff. And I think we have to think about it from that user perspective first and, and let the technology, uh, not dictate it.
Matt Medeiros: Yeah. Yeah, I think anyone who hasn't experimented, for those, those listening, if you've never experimented with, I use OpenRouter for some projects, and of course I use my- You know, heavily subsidized OAuth accounts for, for other things. And when you can actually track the token spend through something like an open router and see that, see what you're accomplishing with the token spend there versus the ambiguous OAuth, like subsidized stuff, you start to look at it and go, "Okay, this is the actual cost of me building this thing."
You know, and, and these frontier models, what a position to be in when you have people on Twitter going, "I can't wait for you to unlock my OAuth account [00:55:00] so I can build more stuff." Like, you're just telling the product company to just like take more money. Like give me, give me the 20X account, give me the 50.
Uh, I've never seen it before, and it's just so wild to see the reliance that people have these days and, and just willing to fork over the money. And I just look at it as a, as a dangerous situation for, for all included. But that's just me. Fantastic conversation. I think we covered a lot. I think we covered, of course, Andy's article, which was the, uh, the impetus behind this conversation.
We talked about the ways that WordPress can technically change the code to have agents better used and build with WordPress. We talked about how we can potentially start marketing it and get that hype train running again for WordPress, and, uh, we talked about crazy costs of AI. Anything else anyone wanted to say before we do sign off?
No, we're gonna go round the horn. We're gonna start with Alex, my special co-host. Alex, thanks for hanging out today. Thanks for agreeing to it sort of last minute, uh, so that I didn't have to be the only one thinking on this call. No offense, everybody else. Alex, where are you from and what do you do?
Alex Panagis: [00:56:00] My name's Alex.
I'm from London. I'm based in, I'm based in London. I'm not actually from London. I'm a re- as you can probably tell by my accent. I didn't know that Andy was also based in the UK, but it sounds like you're based on your accent, right?
Andy Peatling: I'm from the UK, but I live in Vancouver, Canada now. But yeah, fr- Nice
originally from the UK. Yeah.
Alex Panagis: Very nice. Yeah. Thank you for having me. And I will say that you didn't mention Andy's article would be a focus here, which is a bit of a gotcha when I joined. I will say I have read it by the end of the episode. I've gone through it, so. Awesome.
Matt Medeiros: Grenade guy, that's what I do.
Yeah. Brian Coords.
Brian Coords: Yeah. Thank you for having me. I'm actually in California, so Andy and I are, like, repping the West Coast over here. And I guess if I'm pitching anything, it's developer.woo.com, 'cause that's where all of our WooCommerce stuff is. And, uh, we didn't really talk about WooCommerce, but it's there.
It, uh, does a lot of AI things, and, uh, it's important to WordPress, so check it out. It's, uh,
Matt Medeiros: it's imp- it's important to his, his, his paycheck is what he's saying.
Brian Coords: It's important to my kids eating.
Matt Medeiros: Andy from the West Coast.
Andy Peatling: Yeah. Yeah, I'm from the West Coast in Vancouver. [00:57:00] Moved here 20 years ago, but the accent never goes away.
Yeah, I work at Pro- my product is buymiles.ai, and, you know, just working on that has ma- has, has kind of brought a lot of this WordPress stuff to the forefront of my mind, so happy to talk about it and, and help out.
Matt Medeiros: Mike. Mike, I thought you were West Coast.
Mike McAlister: No, I'm, I'm mid-coast. I'm in the Midwest. Oh, that's right.
Speaker 9: I'm in, uh, Milwaukee, Wisconsin, right on Lake Michigan.
Matt Medeiros: That's right.
Mike McAlister: And, yep, yep, so I represent the Midwest. And, uh, yeah, you can find me, Ollie, olliewp.com, doin' cool, interesting, futuristic, frontier kind of things. Stay tuned.
Matt Medeiros: East Coast, The WP Minute. We're about to get hit with an Nor'easter, as they say. Uh, so it's gonna be exciting times this weekend.
Thewpminute.com, where you can also find a hurricane in the Slack channel. Uh, go to thewpminute.com/support to join the free Slack channel if you wanna talk to other like-minded WordPress individuals like Brian Coords, who's in Slack all the time. Thanks for listening, everybody. See you in the next episode.