SPECIAL GUEST: Matt Mullenweg on his vision of AI in tech and WordPress

Matt Cromwell
00:05-01:14
Hey, everyone. We have been talking on this show a lot about AI because it's kind of the thing that's happening all over the place. But it's a very future-oriented topic, and there's a lot of ways that it can possibly impact WordPress and this space in general. And if we wanted to kind of like have a roadmap or be able to understand it all, how AI is gonna impact WordPress in general, there's one person I think that probably has the most vision for how that might be, and he is our guest today. This is WP Product Talk, a place where every week we bring you insights, product marketing, business management and growth, customer experience, product development, and more. It's your go-to podcast for WordPress product owners, by WordPress product owners. And now, enjoy the show. Hey, everybody. I am Matt Cromwell from Stellar WP.

Zack Katz
01:15-01:17
And I'm Zack Katz from GravityKit.

Matt Cromwell
01:17-01:26
And today we have a very special guest with us today to talk about the future of AI and WordPress. Matt Mullenweg is here. Matt, thanks so much for being here.

Matt Mullenweg
01:27-01:27
Howdy, howdy, everybody.

Matt Cromwell
01:29-01:39
And for those who might not know who you are, which is not likely or possible on this show, I'd love to, if you give a little bit of an introduction of who you are and what you do and all that.

Matt Mullenweg
01:40-02:13
Sure. My name is Matt Mullenwink. My blog is ma.tt. On socials, I'm usually photomat, P-H-O-T-O-M-A-T-T. And gosh, I started blogging a long time ago. that ended up winding to contributing to open source software called B2, which turned into WordPress, which turned into a whole wild ride that I spent a real joy in my life to be on. And, you know, founded a company called Automatic and many products along the way. And yeah, just really a joy to be here with you all today.

Zack Katz
02:15-03:08
Yeah. And Matt, you yourself are a product owner. You have a lot of different products going on. And a lot of them are probably aligning around AI in many interesting ways. So thanks for coming on to talk about this with us. Happy to be here. So let's jump in. First of all, for those watching live on YouTube, feel free to comment. We will be taking questions live. We might not be able to get to all of them, but please leave your comments and questions and we will try to address them. But Matt, let's jump in. You talked during the WordCamp US town hall about AI, not only about WordPress and AI, but AI and in general. Can you give us a quick lay of the land in terms of what you think the lay of the land is for opportunities in AI and the risks that AI might have for the web and WordPress as we know it?

Matt Mullenweg
03:10-04:35
Well, as hackers, makers, we're always looking for the best new tools, right? So actually the blog post I'm working on today, I found some old stuff I used to Photoshop. I remember like learning that first version of Photoshop and how hard it was and all the buttons and everything like that. And getting the software was expensive, right? Or you had to use something like a GIMP for open source version or whatever it was. And today, you know, this models like Nano Banana, you can talk to or type to and say, hey, put clouds in the background or move this person over here or something like all the things that used to be like some difficult and some impossible actually with previous tools are not radically accessible. And that's the weird thing about AI is that we have this new tool in our toolbox and some things it's incredibly good at and some things it's like unpredictably, you know sometimes it's like incredible and sometimes it like totally messes up and so yeah this uh it's beautifully human at the end of the day we've uh my whole life programming's been very deterministic you know like uh defined inputs and outputs now we've created this like a kind of new tool that we can use and leverage um which is uh yeah has its own quirks and personalities and ways that it works and that's to me uh really really exciting time to be someone who's making things

Matt Cromwell
04:35-05:03
building things. Yeah, that's actually like the thing that I'm excited about too. And you have talked about specifically how this is a really exciting time in WordPress in general for builders and creators. And, you know, our audience are those builders and creators. They're like, give me a new tool and I'll run and I'll innovate. So what do you think plugin and theme owners should be doing right now to take advantage of this situation?

Matt Mullenweg
05:05-06:03
Well, I would say the number one thing people should be doing is trying to use the tools for their own personal productivity. Hmm. So how can you incorporate these things in your life? And this is what I said when I what I meant when I said I learned AI deeply a couple years ago, 2022. At the state of the word, it's that, you know, kind of just like when new software comes out, know photoshop canva new language whatever like you kind of have to play with it and learn the contours of the tool and its strengths and weaknesses and with ai gosh you you probably have to devote a little more time because there's new stuff coming out every single week it's kind of wild the progress and how fast things move so totally um there are uh you know i would i would encourage everyone to first and foremost just look at things you could automate or things in your life that you could maybe use AI to enhance or be a partner on, a co-pilot.

Matt Cromwell
06:04-06:24
Yeah. I really like that overall because folks seem to be rushing to do a lot of customer-facing AI that kind of ends up being a little bit crappy. But in terms of productivity, I've personally experienced it as highly helpful and productive. I wonder if you could speak to that a little bit too.

Matt Mullenweg
06:24-10:26
How are you using AI for your own productivity? You know, it's hard to make predictions, and I hesitate to in this rapidly changing world. But it does feel like one of the things that makes AI most useful to you is context, right? How much does it have about you? So for an individual customer or user, their most useful AI tools will probably be the ones that have most context about them. And that starts to get more into the browser or the operating system layer. or one of these frontier models like ChatGPT that you've been chatting with for two years, so it knows you really well, has the memory, it has the history, so it can kind of tailor itself to you. Now, I think one of the kind of first wave of this is that we're all like, okay, ChatGPT is awesome. Let's now embed chatbots into our products as well. And that could be useful. And, you know, the chat as an interface for things is a UI paradigm that literally, you know, 700, 800 million people are very familiar with now. And it's far more functional than it was before. But there's also kind of this uncanny valley where it can kind of mess up and it's hard to undo. And like, it's not necessarily the most best design for certain types of actions or certain types of things you want to do, right? So, and something that only has the context embedded in a product. So, pretend that you're like putting an AI chatbot inside your plugin, right? It doesn't know the rest of my life. It doesn't know all the websites I've logged into. It doesn't know all the things on my computer. It doesn't know all those sorts of things. So I think that within specific tools, I think that it's great to do these experiments. And we had a great one that we demoed at WorkCamp. And I think that might be something people start to expect. But I'm far more interested in like, what are the sort of more like purpose built agents or AI tools that can be embedded into user interfaces that can do something that feels a lot friendlier. I'll just go back to image editing because I think that's why that very clearly in the past couple of weeks we've had like huge breakthroughs on and has been a bugaboo, you know, a very difficult thing for us to tackle inside of like WordPress, for example. We want to offer like cropping. We did some stuff with CSS where you can like do duotone, other sort of filters and things. But for, you know, many types of transformations that you might want to make for an image, like the tools that we're able to execute and embed inside of WordPress, inside of the browser are lacking. And even if we had those tools, the amount of buttons and the fluency you would need to properly operate them is as much more like very expert level. Right. You have to be in Photoshop or Canva or something like that. So now having an AI you can talk to, like you might talk to a designer or someone like that and say, here are the transformations I want to make to this thing. and it's able to execute those, its tools in a very fluent way, that feels like a real breakthrough. There's a few other areas where I feel like AI is, it's kind of gotten like best in the world of things. We almost haven't noticed like translation. When we think how hard Apple just a few days ago announced like the AirPods thing, right? The real time translation. I haven't tried it out yet, but like, yeah, it's very clear that we're very close to that and that things that were previously pretty difficult order to become radically more accessible to everyday people at essentially

Zack Katz
10:27-11:07
free, too cheap to meter cost. Yeah. So you've talked a lot about image editing and having a good sense of having AI have a sense of who you are and the work that you do. And for example, with image editing, you have, let's say like a brand guideline for a blog and you have an image look and feel. And that requires having a definition of that somewhere and have that saved somewhere on your site or maybe in a browser or something like that where it has a good storage. How are you thinking about data storage? And where does all this information about you live? Does it live in WordPress? Does it live in a browser? Where does it live?

Matt Mullenweg
11:08-11:12
Yeah. It was funny. That was one of the first questions I asked Felix after that chatbot demo.

Zack Katz
11:12-11:14
I was like, where are you still using cats?

Matt Mullenweg
11:16-11:17
And I think he was putting it in user meta or something.

Zack Katz
11:18-11:31
But I was like, where is this going? Because it's a lot of data. It's a lot of data that needs to be structured. It's a lot of data that needs to be reviewed and curated. I don't know yet. Well, that's good.

Matt Mullenweg
11:31-12:21
That's a relief for me, at least. I do think that as products, you have a lot of product builders. One thing that I think we can all try to build into these products is just a really easy undo. Right. When you think about that ability to undo something is so powerful. It's like such a great paradigm that removes the fear for someone when they're using a new technology. Right. Something that I can't like revert. And so we've worked huge amounts on this and like, well, both Gutenberg and WordPress with revisions and also just like unlevel, undo levels and Gutenberg and everything like that. But like, you know, as we have these tools that might do pretty radical transformations to your site or whatever, like what's the undo button? So I'd say that's something we need to

Matt Cromwell
12:21-13:29
keep in mind as well. Yeah. I was really interested at your town hall. You had this focus at the AI browsers. And you touched on it again. And to me, it just felt like that was a little bit of a Easter egg or a hint. A little bit of a weird callback. I was like, I wonder if Matt's thinking about this in terms of like, Calypso days, like a lot of folks don't remember Calypso, but it was like a desktop tool that you can build your WordPress website and ship it to wordpress.com. I always love that idea. And I love what's happening with WV Studio. I was like, you know, are you trying to map of like an AI browser, playground based, create your WordPress website in a desktop environment, hit publish and it's live? Like, that would be kind of like, this is the future kind of thinking. I don't know, I might be leading too much. But that you talking about AI browsers got me thinking in that way. I don't know.

Matt Mullenweg
13:30-15:25
Well, definitely one of the biggest breakthroughs in the WordPress ecosystem, which has happened, but we still, I think, haven't fully, like, taken advantage of all the things it provides, is this idea of, like, Playgrounds Studio, that you can boot up a container for WordPress with a database, with PHP, with a web server, with everything like that, that you used to have to, like, either go through a ton of steps to, like, download Docker or some sort of local thing or spend something up in cloud, what was very expensive. you can do that in like 30 or 40 seconds inside a browser on a modern computer. So that sort of, you know, instant virtual machine and the reproducibility and the sort of like ability to maintain state in that machine is like a radical capability. The things I am curious about it, one, can we make contribution a lot easier? You know, right now, I had a contributor day at WorkCamp US. Like, you know, there's a lot of steps that you have to go jump through. And you have to get Docker and, like, sign up for an account with Docker and, like, you know, all these sorts of different things. And, like, can we make that maybe even just a single download that gave you, like, a little environment, a little playground that had everything right there for you, kind of in a self-contained way. That gave you the ability to, you know, contribute to WordPress, for example, or Gutenberg. This is tricky. There's some build steps and other things. Like some parts of modern web development are a little trickier than they used to be. But I think this is an attractable problem, you know? And, you know, for Comet, the reason I highlighted is it was kind of like my first, like, brain explode moment since ChatGPT, right? You know, when I, I, you know, got very early access to that, that browser. And I was like, well, let's just go to like WP admin and ask it to do some stuff.

Matt Cromwell
15:26-15:26
Yeah.

Matt Mullenweg
15:27-18:12
I think first I went to Gravatar and I said like, Hey, put the phonetic, you know, pronunciation of my name. And it like somehow figured out all that weird international phonetic alphabet stuff. And like, and put my product and it actually like saved it and published it. I was like, Whoa, I didn't mean for you to do that. I was like, wait. And then I went to WordPress and started like navigating Gutenberg and moving things around, moving blocks around to do all these types of pretty complex interactions. Again, and in the background, we've been spending, you know, a year building customized bots to sort of interface with WordPress and doing things like MCP and other things to like have a word, an AI bot sort of interface and modify your WordPress. And here was something essentially imitating what a human would do, like reading the screen, looking at it in a multimodal way, and then like taking actions like a human would. And that felt like an incredible hack. It reminded me of folks when, you know, obviously we've had robotics for a long time in factories, and there's specialized robots that look like have arms and do things and build stuff. But there's a big push around these sort of humanoid looking robots, you know, the optimists or unitry or the other things that are happening. And I was like, well, why would why would make something look like a human? Right. Robots can have eight arms. They can be an octopus. They could like make them better, more capabilities. And someone explained to me that, like, well, the world's UI is for humans. So when you think about every appliance, every door, everything in the world is built for our form factor. If you're able to make an autonomous agent, a robot, which has our form factor and can use the world like we do, it instantly gets access. We don't have to rebuild the world to work for that. It's kind of like when you think of like railroads, right, are great, but you have to build the tracks first. Cars are better, right? You know, cars can do a lot on roads or off roads. And then, you know, at the best, you maybe have things like flying or drones or other things that can literally navigate 3D space without needing anything to be built for them ahead of time. And we're kind of in this weird space where on one hand, like we're doing incredible investments in MCP and APIs and other things to essentially like make make rails for these robots or machines to use things more efficiently. On the other end, the machines are saying like, Hey, what if I just pretend to be a human? Yeah. And so if a human could use this,

Matt Cromwell
18:13-18:17
just skip all those steps. Yeah. Yeah, that's fascinating. I

Matt Mullenweg
18:17-19:58
said like, this is great for usability testing, actually, because, you know, we haven't fully solved the can a human use this problem. We still have lots of problems in WordPress and other products where like you put a natural intelligence, a very smart, you know, high IQ human in front of this thing and they can't figure it out. So that's, that's a version. If we make it better for that, it's going to help. And it also kind of, you know, part of what I loved about Danny Sullivan's talk about like SEO versus geo versus everything. And kind of like one of my big takeaways there is like, you know, there's all these things you can do the game, the algorithm or build for the machine, build for Google. But Google's aim is to find the best answer or the best result for a human. So if you build the most valuable thing for a human, you will ultimately be aligned of where Google is going to end up or where Google is trying to go and try to optimize for. Yeah, yeah. I get why people criticize that. And it's true, that's possible to game these things in the meantime. But it really made me appreciate like, you know, Gutenberg has taken a couple years would otherwise because of the investments we've made and like accessibility and, you know, kind of making that the framework really robust in a way that like many other builders do not. And now that's paying off. Yeah. Kind of like the old adage that like when you build it for accessibility, it's not just for people who might be disabled, you're also building for someone who might be holding a kid in one arm or using a walker or something like that, like all these things like start to open that you didn't imagine, some of the investments we've made into accessibility are now paying off for these agents being able to navigate our interfaces and our things.

Zack Katz
19:59-20:31
Yeah, love it. So how does this translate to WordPress? When AI is capable of doing anything, essentially, that a human can do, not general intelligence, but if it can browse a website, upload things, write articles, publish blog posts. Like, how does WordPress fit in with this? You've talked in the past about WordPress being the operating system of the web. What is the role of WordPress in this AI future?

Matt Mullenweg
20:34-20:44
Well, that's interesting because WordPress is a tool. Right. So I think the more interesting question is like, what's the role of WordPress builders? Or like, what's the role of humans in this?

Zack Katz
20:45-20:45
Yeah.

Matt Mullenweg
20:46-20:47
Yeah, I blogged about this the other day.

Zack Katz
20:47-21:37
I feel like taste and discernment is going to become ever more important. Right, but if you're trained on taste and discernment, if the model is trained, we humans are exposed to a lot of bad taste things. And so our context is polluted by bad Sunday mailer ads, right? We know what bad design looks like. We know what good design looks like. Design bots are trained on this stuff as well, And they're not influenced by as many diverse things, I guess, as we are. So that is our benefit. Or they are, but they're not thinking about... Anyway, there's so much context in the world that surely AI can have better taste than most of us. Jury's out on that one. Not yet, maybe. But if it's trained only on Johnny Ive and... Well, this is why I love...

Matt Mullenweg
21:38-22:06
I'm a huge fan of things like Mid Journey. David Holtz is someone I look up to a lot in this space because he really is like, we're trying to optimize the reward function of our AI for beauty. That's interesting. And how do you train things and other things for that? But you're right. So, yes. And again, we're probably getting a little further out than it's like able to predict perfectly here.

Matt Cromwell
22:07-22:08
Yeah, of course.

Matt Mullenweg
22:08-22:20
So for today, I like to think much more like, well, you know, how do you stand out when it's very, you know, it's very easy to generate 10,000 blog posts and put them on your blog.

Matt Mullenweg
22:21-22:23
But is that actually reliable? Yeah.

Matt Mullenweg
22:23-22:46
And so like, what can you talk to about that's unique to your experience? Or how are you creating things that maybe are like new to the world that haven't been published a thousand times already? You know, because, again, there's so much of, you know, AIs are trained on things, but the fraction of human experience, which is accessible to them, is still a fraction of what we all experience.

Matt Cromwell
22:46-22:46
Yeah.

Matt Mullenweg
22:47-24:13
All have in our minds and our intelligence. So, yeah, this is, I think, an area of great opportunity. It's, I think, why so many folks are locking in and working really hard right now. Because it does feel like this is a particular window of opportunity to define to answer some of these questions, which aren't really certain yet. And which we probably can't fully answer without shipping and iterating. Right? Like some of the stuff, like, gosh, your guess is as good as mine. let's ship 10 versions of this and see which, you know, five of them do okay and two of them do really well. And that's what we'll start to iterate on and really like see. And it's kind of beautiful that, you know, like, there's been, you know, pre-ChatGBT, you know, LLMs and other things and other chatbots were doing pretty interesting things. And they sort of found like just the sweet spots of like that reinforcement trading and the chat interface and everything on top of the base model that made it like sort of like hit a nerve with society we're like oh it's something like a little bit different from all the bots i've used before from microsoft's bots or eliza or whatever it was you know but uh so and um yeah i think i think there is that um you know finding that sweet spot is something we're gonna have to work our way into yeah uh we got a couple good questions um

Matt Cromwell
24:13-24:59
here well one this one's just a observation first stephanie hudson i think really was impressed by your comment the world's ui is built for our form factor and that was a good one that's a one we'll spread around to other places it's it's really unique and perspective that i think is super helpful to help align around the idea that like yeah we're not we don't operate as chat bots we operate in this whole ui holistic experience of things um so thanks steph for that comment Andrew Palmer has a question that I think is relevant. He says, "What aspects of AI commercialization interest you the most, particularly for like WooCommerce or personalized image generation, user funneling to specific products or services and reducing churn?"

Matt Mullenweg
25:02-27:10
In a WooCommerce context, I do think a lot about, you know, product catalogs and imagery and other things like that like you know the this and there's already great examples of this like you know you visit the picture of the shirt and you know those AIs that can now put that shirt on a picture of you so you like see what it looks like and stuff like that so that's pretty cool and that is kind of like a new novel thing in shopping I have no idea how that's going to like influence or become a common thing but that's pretty neat um I also love that AIs are very much open web native you know right now all of our sites are getting hammered by like a million bots crawling us and everything like that, right? And so I do love that, you know, we had a, I always talk about this pendulum that swings between like centralization and decentralization, which we like open and proprietary. And there was definitely a lot of market effects that I think brought us towards centralized marketplaces including like ebay etsy amazon in the past um and uh and you know as those marketplaces and aggregators and mass power they start to exert that on the individual vendors right because their interests and their profits not necessarily like any individual publisher so it starts to be a lot more competition there's advertising things merge and shrink etc And so AI now like putting people back towards a lot more of these independent publishers, independent stores, independent sites on the web, I think is a nice rebalance of power there, which I think could be very, very great for people to, you know, just like social media, like still distribute to those things to get like extra distribution where it's appropriate for your business. But when you have place a place where you can have a direct relationship with your customer, that is golden, right? When you can capture that email address, capture that phone number, have that relationship, provide something that's a little bit differentiated from like the generic, like, you know, big box store or Amazon experience, that can be pretty special.

Matt Cromwell
27:12-27:36
I like that. Another thing you highlighted on your town hall was telex, which that was a fun surprise. We actually have a short video here from James Kemp that is showcasing a little bit of what can happen there. It's something that he told us is really experimental. I don't know

Zack Katz
27:36-27:46
if you know what I'm going to show. This is core AI, respective, not necessarily telex.

Matt Cromwell
27:46-28:00
Oh, not Telex. That's right. That's right. Sorry. Yeah. So we'll have to circle back to Telex. But this one is actually, yeah, WooCommerce, CoreAI, and that command palette thing. So hold on. Let's bring that up. And oh, it didn't come up.

Zack Katz
28:04-28:22
All right. Well, here we go. There we go. So we have James Kemp, who's giving us a nice demo of WooCommerce command palette integrations and we're jumping to the midi part. So here we go.

James Kemp (in video demo)
28:26-29:25
In the code. Another one we can look at is getting an order summary. I don't have any orders on this site, but you can see again the response there. This one's really cool, the stock report. This one's a bit more of a multi-layered ability. So we're using an ability that gets the stock report, which just fetches all of the products, gets their current stock amount, the stock status, and also the value, which is the cost of goods multiplied by the amount of stock. So you can see as a store owner how much stock you have on hold. And again, that's the execution of the ability there. And the nested version of this is that we can go through and update the stock. And then the stock update is using another ability, which is WooCommerce update product stock. Obviously, it's sending its input parameters there, which is the product ID and the stock quantity. And then the response is coming back. And within the response, I've also added the total.

Zack Katz
29:28-30:17
I don't know where to call my what, but that's a good example of AI interfacing with the data of a website. And it's really interesting for me. I'm seeing that and I'm thinking, okay, well, how does that interface with somebody who's in their store? Let's say they have a physical store with a WooCommerce installation and they can just say to their virtual assistant, we just sold something. Or obviously it could integrate with a point of sale system or have a camera that they're wearing in their stock room and just automatically see what's available. When you expose all these data structures, anything becomes possible, and it's really exciting to have this coming in WordPress 6.9. I'm really looking forward to it.

Matt Mullenweg
30:18-30:28
Yeah, and I think one cool thing about that is the Abilities API, I do feel like, is a very important primitive that will be part of WordPress for many decades to come.

Matt Mullenweg
30:30-30:31
And it doesn't rely on AI, right?

Matt Mullenweg
30:32-30:46
So it does all sorts of things, the command palette and other things that allow us to take a customer need. What's something that people are doing repetitively that takes up a lot of the time? And how can we streamline that? How can we make it fewer clicks? How can we make it faster? How can we automate it?

Zack Katz
30:47-31:22
So what are some of the things that you're... I'm wondering from an organizational standpoint, you've got a lot to manage. And I know that we're a six-person team at GravityKit. And I am trying to wrangle all the different parts of the business and have it exposed to AI in a way that helps optimize my time and helps improve our efficiencies. What are some of the ways that Automatic is using AI? Can you share specifics that have really unlocked productivity or creativity in a way that you were surprised by and excited by?

Matt Mullenweg
31:24-32:06
Yeah, again, I'll go to the things that are clearly way, way better. Definitely, I would say people proofreading or translating their output in a million contexts. That might be support, it might be in communicating with colleagues, etc. That AI editor, AI assistant is pretty darn good right now. I would say that engineers with a human in the loop, having AI, a system, automate things, check things, write tests, span code, that has definitely had an impact as well. And we're seeing in spots, but some people accomplish things in hours that maybe previously would have taken them days.

Matt Mullenweg
32:07-32:08
It would be really fun.

Matt Mullenweg
32:08-33:16
And it doesn't always work. So I'm not, again, I'm not a, I don't want to be probably Anna-ish about this and like pretend that like, you just like put some agents and all of a sudden your business is transformed. You don't need people anymore. Like, no, we're not there yet. We won't be. And you can kind of like squint and see how we might end up there. But, you know, again, I love that something, I think it was Hedon Shah tweeted the other day. He's like, if you suck at remote work, you're going to be suck at managing AI. Well, the things around asynchronous communication, setting up clear tasks, like all these sorts of things, like these skills that you might develop and that we've developed as a distributed company might be useful for managing some of these agents in the future. Inside Automatic as well, you know, we have a very strong culture of, you know, we basically haven't used email for like 20 years because we use internal blogging. We have internal WordPresses using a theme called P2, which is kind of like a theme that's built for real-time collaboration and project management and things like that. And so now having 20 years of those archives,

Zack Katz
33:17-33:42
you can imagine how valuable. So how are you taking, so we have Slack channels, internally at GravityKit, and we have long otter transcriptions from years of meetings. And how are you collecting that and making that actionable in a way that isn't garbage in, garbage out? when it comes to the amount of noise that is accumulated um i would say there's not something

Matt Mullenweg
33:42-34:31
that can like fully recommend right now like our search is getting better other things um summarization is pretty good like other that sort of stuff is great um but uh yeah i don't think we have this like magic thing um there's stuff launching every day though i just saw one yesterday called co-founder and it's like i guess something that like reads all your slacks read all your emails and like you know is able to put that in a just like humans these ais can't keep too much in the brain at once right right so um this context engineering putting the appropriate things in front of them that they're using to make a decision uh is important and like you said like maybe reading you know uh a track ticket or a ticket from 20 years ago is not the most relevant thing to answering a question about a bug today. Right.

Zack Katz
34:31-34:32
It might be.

Matt Mullenweg
34:36-34:49
So it's going to be very important just like put the right things in front of these tools so that they're able to like not get overwhelmed. It seems like you're very familiar with the shortcomings of this as well.

Zack Katz
34:49-35:55
Yeah. And I'm at an organizational level. I'm trying to create embeddings for everything and cross tag things and make sure that everything's organized systematically so that, oh, okay, this UI is related to this other product that we have, which is related to this WordPress core feature, which is related to a Gravity Forms feature. And connecting all those dots in the structure of the data is something that I feel is a real shortcoming and requires like a knowledge graph that doesn't yet exist. And that's where it comes back to the question of, okay, where does WordPress come into this with the data structures of the, of the web, like how does WordPress describe what we're trying to get at when it comes to track? Like, sure we can MCP that, but that requires them the track to have also ingested and, and figured out what is important. Like, is WordPress a good central store for this stuff or is it, is it a distributed like connected future or for WordPress? Oh, it's definitely going to be distributed, right?

Matt Mullenweg
35:56-36:32
So again, there will be different products and databases and other things that lots of things live in, and you'll want to be able to call out to them or aggregate them as needed. But yeah, we'll see. - What do you got? - One of the tools, we have a cool tool inside Automattic called MGS. I think it stands for Matt's Global Search, 'cause it's one of the things I built a long time ago, which was like a search engine that essentially looks across like IRC, Slack, code, you know, our internal stat system are all over internal blogs, like there's a global search across all of them. And that sort of thing can be really, really useful.

Zack Katz
36:33-36:45
Yeah. And that's that's what I'm trying to build as well. I'm calling it Mono Kit. Yeah, it's a mono repo for the whole org, but it's hard to it's it's hard to even conceptualize.

Matt Mullenweg
36:46-37:21
Well, and it's good also like let's not build these things for their own sake, but have a problem you're trying to solve. And you can measure the effectiveness of this solution. versus other things that exist in the marketplace or that you could do to solving that problem. So if you root it always in that problem or customer need, I think that's the best way to answer these kind of broader questions, because then you're not just trying to predict the future in a generic way, but you're really saying, "Okay, what's a specific thing?" And we can measure whether we are solving that problem or not. Right. And is there a specific thing that

Zack Katz
37:21-37:33
you've solved it automatic like you've talked about translations and stuff but like uh what what else what what else have you like really just uh automated or um you know made

Matt Mullenweg
37:34-38:07
uh made super better oh man it's funny because it's kind of it's like little improvements across everything um it's actually funny in wordpress we've had like a a very uh machine learning thing built into almost every WordPress for a while, which was Akismet. And obviously Akismet, you know, identifying spam now has like a new array of tools, things available to it that weren't there when I first wrote it 20 years ago. So there's other things that like, you know, it's like lots of little improvements every single place.

Matt Cromwell
38:09-38:13
Is there already AI working in the background of Akismet today? Is that what you're saying?

Matt Mullenweg
38:15-38:43
That's awesome. Yeah, I've always been so fascinated with this space and I've tracked it closely for a long time is that, yeah, it's always been really fascinating to me, like how you can sort of have like learning algorithms. How does a machine get better over time? How does it self improve? How does like more data make it better? And yeah, that's very much at the root of why it gives me to have like four or five nines of accuracy for a very long time. That's awesome.

Zack Katz
38:44-39:28
And where is the moat? Like for our customer or for our listeners rather, who are product owners, there's a lot that AI can and will automate out of existence. So like Akismet is nice because it's a service that you call out to. When our listeners are considering what they should be building and can be building, what do you feel is the most protected area of interest that they could be exploring? I'm really doubling down on like really strong customer relationships. And what does that look like? What does that look like? Yeah, when it comes to like a world of AI. Sure. And it looks like, you know, talking to people. It looks like support.

Matt Mullenweg
39:28-40:22
And it looks like, you know, outreach and understanding people's problems. And like, you know, and that's human to human connection. The more of those that you can have, that feels like a moat. Because there will be brand loyalty. There's a stickiness to that, right? that and um and you're part of maintaining a subscriber over time is how are you increasing the value of what you're providing them especially if you want to hopefully like raise prices over time right keep up with inflation other things so you know when you when you have that renewal which is going from like 100 bucks a year to like 110 bucks a year like do they feel like that you're still providing a vast amount of value over what you're charging and um yeah doubling down on those customer relationships and really knowing and understanding your customers I think is not going out of style anytime soon and a huge vote. I love that and actually I've said to a couple different

Matt Cromwell
40:23-41:09
product folks that I see a lot more potential with AI in the admin side of things rather than the front-facing side of things just in terms of like being able to provide business owners a lot more insights like you're saying being able to help folks do support a lot better that's something and I experiment with all the time. I think there's a lot of value there in terms of, you know, when you ask questions about big data, how can we deal with those things? You know, the customer side, front end of the website isn't as concerned with all the big data things. They just want like magic at their fingertips, you know. But on the admin side, you know, empowering store owners to know their customers better, that's really good magic. That's lots of potential. I love that insight.

Zack Katz
41:11-41:46
Yeah, one of the things we're doing is trying to connect the dots with our Help Scout database and our email newsletter database and our WordPress installation for EDD, which is how we distribute our plugins. So that when I have a call with a customer, like I make it available to have a consultation call, I can easily query and have AI write up a summary of what I need to know about the call that's coming up with the exact data that's being presented, like that customer has shared with us about themselves. That's awesome. That's an example.

Matt Mullenweg
41:46-41:59
That's a great example of it. And yeah, it'll be very interesting to see how CRMs like Salesforce, HubSpot, and other support tools, how that evolves in the future.

Zack Katz
42:00-42:18
We're getting ready to move away from our newsletter because our provider Drip, they don't have good APIs. And they don't have a way for enough exposed to make it actionable in a way that is more relevant now than ever with the AI revolution.

Matt Cromwell
42:19-42:27
interesting i hope drip folks are listening yeah this is actually one you know one reason i'm

Matt Mullenweg
42:27-43:14
bullish on wordpress in the ai era is that wordpress is probably the most radically open place you could like put stuff into and get it out of right there's so many different apis there's ways you can access it and of course it's open source so like and then you can open up the hood and modify it and um and so proprietary tools that don't keep up with the accessibility and sort of like data deliberation that we're talking about. And that's why data deliberation is also something I want us to get a lot better at, is how can we bring things, bring data from like locked up tools or other things into WordPress? Because once it's in the WordPress, it's now yours forever. It's now, you know, really in a place that like truly belongs to you, both from like a code point of view, a freedom point of view, a license point of view, and like a data accessibility point of view.

Matt Cromwell
43:15-44:37
Love it. Um, I dropped there for a minute. Did you all talk about Telex? I want to make sure we talked about Telex. Okay. Telex. Um, Telex was super cool. It was great to see, um, this idea. And as soon as like, while you were talking about it at the town hall, I jumped right on there and I had built a little AI tool that does like click a button and it gives you a TLDR of the article that you're reading. And I was like, can Telex build me that block? Um, and it did a pretty, pretty darn good job right out of the box. Um, and I've seen several different, um, folks, uh, showing like Jamie Marsland just did something where you can scroll and it will, it will progress a video while you're scrolling and things like that. Um, like where do you see stuff like that? Like, I think there's product folks out there that are like, there's a million different, not a million, there's several thousands of plugins on the directory right now that are like three files maximum, you know, or there's, you know, there was a big push for a long time for these single block plugins. Um, and there's a single block library, um, and those have a lot of value, but with something like Telex, it's almost like those, are those going to go away, Matt? What do you think? Like, um, instead of looking for those types of simple functionality plugins, we're just going to Telex our way into new blocks. That's a good question. Um,

Matt Mullenweg
44:37-47:11
Yeah, I mean, Telex is, I really like it because it also like builds on things we've been investing in for a long time. Like one of the great things about Gutenberg is that it can sort of translate modifying sort of a structured part of the content into also user interface, right? Which renders in a certain way or has certain controls that make sense and are intuitive to like someone editing the content. They can see how it fits with the rest of their things in their post or page or their sites. So, yeah, but fundamentally, if you're going to simplify it, it's kind of like a fast plugin builder, right? And that's what's neat about it. Telex is, like we said, an experiment. It gives you a zip file. It can upload to any WordPress in the world. It can be on.com. It can be on Bluehost. It could be wherever. But you get kind of back to the fundamental software problem, which I think is still the great unsolved problem of these things. So 90% of the lifetime cost of software is not in the initial building of it, but in its iteration. Maybe more, actually, over the terminal thing. We also have, I will describe a general problem in WordPress where how does a user know which thing to choose to solve their problem? Right. Almost anything you search for in the plugin directory, there might be dozens of plugins that do it. And they're trying to sift through like ratings, descriptions, other quality indicators to know like what solves their problem or not. And then we have kind of like a sub problem that which is like the commercialization or non commercialization of certain of these solutions. Right. So fundamentally, if you look at like, what at least WordPress is betting on, or why WordPress exists, is that let's take a set of problems, and put that almost like in a DMZ, that we're not going to commercialize, and try to promote collaboration, where we work together to solve these and make it better and iterate on it. Right. It's very easy to build a CMS, right? It's a CRUD thing, right? It's like the Rails five-minute demo, whatever, is like a blogging system. But, like, what's the reason, you know, we use WordPress? Well, we start to get to those edge cases, those, you know, different APIs, different things you kind of can iterate and say, here's a tool that you can use. So the tool use of, like, sorry, I'm going to bring this all back.

Matt Mullenweg
47:11-47:12
Okay, I'm hanging in there.

Matt Mullenweg
47:15-48:39
there's um you can have telex make a slider right and actually that might be nice because someone just might need this very very simple slider and it might do everything they need um there are also awesome slider plugins and slider businesses right that actually purge money and like provide like many dollars of value per year that people are willing to pay for uh for like a really advanced slider so in that continuum of software i think there's going to be multiple solutions there'll be like that there may be one off that is just fine for the purpose that you build on the fly there's going to be something which i hope is in between which is maybe that runoff as the starter but then we have like a great way for like a community to come together and like three devs who have done like that simple slider thing come together and say like hey let's work together on this not dissimilar to how wordpress started b2 had many forks and i went to some of the other forks of b2 and said hey instead of making like a bunch of different forks of b2 let's work together on one of them and that'll be better right and users that'll be better for users too because they don't have to be like oh do i want b2 plus plus or b2 evolution or wordpress or like you know we'll just come to one and um and it's open source and so it belongs to all of us and then finally like you know where are there commercial opportunities there which is is a tough problem to figure out uh but um sort of the root of like product development capitalism absolutely and in mr asks uh what do what does

Zack Katz
48:39-49:24
does WordPress do to close the customer relationship gap on a site built with AI created blocks and plugins? Which is a relevant question. I see the singularity coming. And when we're to the general intelligence side of things, but where there's an AI company that builds and distributes AI plugins and automatically fixes bugs using AI and provides AI customer support, I see that coming eventually. In the meantime, they have a block that breaks now. What next? How does WordPress scale to support that when everybody's breaking their own code?

Matt Mullenweg
49:28-50:20
Yeah. When something breaks, what's next? Well, I would say that if the AI is able to debug and yeah that's amazing right because it's instantly available faster um but we've probably all been i mean by the way this this is not new this goes back decades where like you're stuck with the bottom of the country you know like please speak to you human like zero zero what's the way to like get me to someone who can actually understand me and so um there's still that again uncanny valley where sometimes if they're able to understand your problem and address it uh using their capabilities It can be super magical. But when they screw up, man, it's frustrating. And you're just like, gosh, can I just talk to someone who'll understand? And again, by the way, even when there's human support, there's many examples of like, I'm talking to a support agent who just doesn't know what's going on.

Zack Katz
50:20-50:23
Is that a hosting? Let me ask you.

Matt Mullenweg
50:24-50:29
It's really to be understood and addressed.

Zack Katz
50:29-50:52
Who will be on the hook for this? Is it eventually the host? is it like when you have AI building a website, publishing on a hosting platform, regardless of the host, like is it then going to fall on these distributors of the code like to the world, which is the hosting platform at the moment?

Matt Mullenweg
50:55-52:06
I mean, the free market is going to figure that out. You look at how it exists today, like what host support has changed over time. And actually a big thing for WordPress success is like the work. I did a ton of work like 15 years ago, like saying like, because before WordPress was just one of like 30 things they would install and cPanel, whatever. And they don't provide in support for it. And I started going to all the hosts and be like, Hey, let's start going to all your 2000 phone agents and giving them like a WordPress one-on-one because this is like now the operating system for your website. It's more than half of the installs that are happening or half of your new sites or whatever are powered by this. And so like, you know, our, our support forums can't handle all these people. And by the way, they're paying you, they're not paying us. So like, you know, and it's not that hard to learn. So if we can start to get some basic training and hosts started to support more than they used to, where before you'd call and like, if you needed like your email fixed, they would do it. But if you started to say like an application on top of it, they're like, sorry, go to that application provider. But also because they installed the application for you for free and you didn't pay anyone for it. So it's kind of like, I do feel ultimately like some of the responsibility paid.

Matt Cromwell
52:06-52:07
I'm like, who are you paying?

Matt Mullenweg
52:09-52:45
That's who ultimately has, I think, some of the most responsibility for that customer relationship because they've gotten you to like open up your credit card and give them something. So what value are they providing? There's always going to be a threshold where like, you know, plugins provide support, etc. But at some point you might say, hey, you really need to hire a developer. Or this is kind of beyond what we're able to offer for $100 a year. You need probably like 10 hours of support from someone that costs $150 an hour. So like, you know, there's always going to be a way to pass off or escalate depending on the needs, the uniqueness of the problem.

Matt Cromwell
52:45-53:08
Now we got two last questions for you. First one is, I just want to see what are we missing? Like we asked a lot of questions that are top of our mind. about AI and WordPress and trying to read the tea leaves as we see them from the outside looking in. But what did we not ask? What's on your mind that you're really excited about that we haven't covered?

Matt Mullenweg
53:12-54:16
You know, something definitely on my mind that we haven't talked about, which is relevant to product talk, is the importance of design, usability, testing, understanding your customers and all of this. I do feel like that's a place where the entire WordPress ecosystem can up our game a bit, is really being thoughtful and crafting these user experiences that make people feel good, aren't challenging, aren't confusing. So that's something both for Core WordPress and WooCommerce and everything I work on. I, you know, I like to say I'm the unhappy as user of all our products. Like there's a lot we can do to improve. But I think it's also something that just every engineer, every marketer, every business owner, everything can start to improve their own skills and taste and discernment around this. In addition to like, you know, let's make sure that we're hiring the best professionals, their designers or UX researchers or other things. Like, let's like, yes, and let's do all of this to up our game.

Matt Cromwell
54:16-54:44
Love it. Last one. We wrap up every single show asking for best advice. We go around the circle. And what is your best advice for anyone thinking about their WordPress product business and what we've discussed here today? Like you're at a WordCamp. Someone's like, Matt, I heard all about AI. What's your best advice for me? In a nutshell, what should the audience take away from this conversation?

Matt Mullenweg
54:46-56:21
You know, I said it earlier, but I get so jazzed by talking with customers. And it always like really sparks my, I was in Houston, you know, last week and met up with a VIP customer. That's a large energy company, like $100 billion energy company or whatever like that. But also I've been a customer since I was a kid, which is neat. So that was kind of fun. But then like just hearing what they're tackling in their business and how that works. And, you know, like their sort of customer model where there's like franchisees that own the gas stations, like all these sorts of different things. And like, she kind of sparked my mind. I'm like, OK, like what's their intranet look like? What's their compete to help them? You know, what's their how are the websites for all these different businesses that are owned by lots of different people being built? or the brand guidelines? And then what are their customers' customers ultimately looking for? How did they decide to stop at this gas station versus another one? And so there's a million niches out there. And one of my favorite things about WordCamps is meeting someone who's built a WordPress business that goes really deep around a niche I hadn't thought of. There's obvious ones like education or dentists or real estate or stuff like that. But then sometimes you get some pretty niche ones, like when I came across the other day, storage racks. So I guess warehouses have like shelving systems essentially. And there's like a whole industry around like different types of shelving systems and like how like certain ones can support certain weights or support different accessibility.

Matt Mullenweg
56:21-56:23
And like, that's a whole world.

Matt Mullenweg
56:24-57:41
So I love when entrepreneurs like go into an area, shipping storage containers, whatever it is that like, might be like, you know, part of this like hidden thing that like makes the world work. And like sort of really deeply understand what's a problem there and how we can apply technology to like, you know, make their lives a little bit easier, a little bit more efficient. And that's kind of like the engine of prosperity and abundance and progress, which has allowed us all to like experience so much better lives than we would have 20 50 years ago, like there's no time in history, I would rather be alive than today. You know, when you of what we take for granted medical care antibiotics broadband cell phones all these things that like you know 100 years ago like the richest person in the world the most powerful king didn't have access to and now it's accessible to all of us um that's really powerful and that when i think when i think of like democratizing as like a verb you know technology and part of our mission around democratizing publishing it really is around how do we take these things that were difficult expensive or inaccessible and make them available radically to a lot of people. And then who knows what happens after that, but it's going to surprise and delight us and probably be some things we didn't expect and hopefully move humanity forward in a way that then allows like, you know, future

Zack Katz
57:41-57:56
things to be built on top of it. Yeah. Matt, we lied actually. There's one more question. I know you're a big reader. What are you currently reading? Oh, I'm reading a lot, but

Matt Mullenweg
57:56-58:08
I like to move between books. One I think is relevant for this crew is Creativity, Inc. Ed Campbell from Pixar, who worked closely with Steve Jobs and other things. It's a fun book.

Zack Katz
58:09-58:10
That's a good one. Nice.

Matt Mullenweg
58:11-58:12
Cool, cool.

Zack Katz
58:14-58:55
Okay, well, my best advice is if you haven't used an MCP, if you don't even know what that is, fine. Download Claude, download Claude is the one I would recommend using for this, and install any MCP server in it. It might take you chatting with Claude to figure out how to do it. That's fine. I tell Claude to install MCP servers on Claude. It's okay. Just ask it what to do, and it will work for you. Just set up any MCP server to any service as a starting point. Just do that and you'll be better off than you were before you did in terms of the AI conversation. If you haven't done that yet.

Matt Cromwell
58:56-01:00:17
That's a good one. Interesting. My best advice, honestly, is, and I, you know, talk to a lot of product folks all the time. And honestly, sometimes they take a lot of the automatic products for granted and kind of just like skim and don't really look really carefully and closely. I have been looking a lot more closely and I think folks need to do that. They need to check out Telex. I think Telex is really interesting. Everybody needs to check out WP Studio. I find it really fascinating and a really great product. It's way faster locally than anything else I've tried before. And it's been a great upgrade. I see Automatic doing a lot in terms of Jetpack. Jetpack is really interesting at the moment. There's interesting moves there. All this stuff, honestly, for the product folks who kind of are always looking for the new sexy or whatnot, they don't look at automatic all that much. And I think that that's a disservice. They need to take a look at what you all are doing because I do see a ton of really interesting things. And of course, everything James LePage is doing with the core AI team, it's incredible. And the pace that they're running at is really amazing. So my advice, folks, is dig in to what's happening around the WordPress project as a whole, not just in your little neighborhood or niche of the space.

Matt Mullenweg
01:00:20-01:00:50
Well, Matt, that is incredibly kind. I really appreciate you saying that. And all right on, like James is a great example. Like when you spend money with companies like Automatic that are really active in the WordPress community as well, we then funnel a lot back of that into things that then don't belong just to us, but like belong to all of us at our community services. So like that sort of fire for a future idea. We're probably way above five, actually, if you count everything. But like, it's, yeah. So that's another reason to vote with your wallet for companies that are supportive.

Matt Cromwell
01:00:52-01:01:03
Nice. Well, that's a wrap, Matt. Thank you so much for joining us again. And we hope to have you as often as you're willing to be here. But where can people find you online?

Matt Mullenweg
01:01:05-01:01:42
My blog is ma.tt. I've actually like since WordCamp been really inspired to just be blogging every day. So that's been kind of fun for me. Like, just like, you know, I think before I got in a mindset where like a blog had to be like a long essay or a huge thing. And I'm just starting to like share links, share interesting things. So check out MA.TD first and foremost. But I'm on X Twitter, Tumblr, Instagram, Blue Sky, like all these sorts of things. The ones I use probably most actively would be in my blog and then Twitter. So, you know, those two. And I'm photomad on Twitter, P-H-O-T-O-A-T-T.

Zack Katz
01:01:42-01:01:56
Well, Matt, thank you again for joining us. and for y'all listening or watching if you're enjoying our shows uh please do us a favor and hit like subscribe share it with your friends reference it in your newsletter and don't forget to tune in again next week bye bye