Building with AI How WordPress Standards Shape the Future of Open Source

Adam Weeks:
This is Adam Weeks at openchannels.fm. Uh, we are talking about WordPress in the age of AI, and I am here at WordCamp US with my good friend Nate Stewart. Uh, Nate, it’s such a, a a big deal that it deserves unpacking. So we’re here at the beginning of WordCamp US. So yeah, real quick, Nate, how’s the conference treating you so far? How’s it going?

Nate Stewart:
Good. I mean, everyone’s gonna say it’s like pretty hot outside, but it’s also warm and friendly inside.

Adam Weeks:
I like that. Warm and friendly inside. Although I was riding on a golf cart yesterday and I felt like I was in like an air fryer, like a convection oven.

Nate Stewart:
Just don’t open your mouth when you do that.

Adam Weeks:
My eyes like wouldn’t close. But it’s a dry heat. It’s an oven heat, but that’s okay.

Nate Stewart:
We’re here.

Adam Weeks:
Okay. WordPress is evolving and changing quickly. It always has. It’s not like, you know, WordPress, you know, over the last, you know, 15, 20 years, it changed. Now we gotta update WordPress because there’s a new threat, there’s a new scary, there’s a new something. What are— we’re gonna start with some of those things that you observe that are making you nervous when it comes to open source?

Nate Stewart:
Nervous. I think it’s— I mean, I like to think of like where there’s the opportunity.

Adam Weeks:
Yeah, we’ll get to that. But go ahead, let’s sit in the scary for a little bit.

Nate Stewart:
Yeah, I think the biggest trend is really everyone’s writing way more code, but no one’s really looking at it. And that ends up with security potential issues. You’re pulling in packages and whatnot that you don’t really know what’s going on, or creating something new where somebody else solved the problem securely.

And you’re like, well, I’m just doing it from scratch. And just the way AI works is you’re getting generally an average. Maybe you can, you can increase that if you use plugins and skills and other systems to improve that. But if you’re just going for it and vibing, yeah, like that’s an issue. Also, like in general, let’s call it like taste, kind of the—

That I think that it used to be, you had to go through a fair amount of cycles when you’re building an experience for like another user or even something for yourself. You’re kind of thinking about every step of the process. Now you’re giving up some of the taste. To, to, you know, another form of intelligence that’s not yours. And yeah, I think it does 2 things. One, it combined with things like security, it creates this where you don’t really know what’s going on and if it should be one way or another because you weren’t a part of the user experience or the backend or whatnot. So you kind of get led down a route. So if the route is not good, you don’t know that it’s not good until later on and something happens. And then on the taste side, you’re not really making your own enough to have a strong opinion.

And which is later on in any, any project like helps. So you end up having a little bit of paralysis with making decisions, whatnot. And that’s just as painful as like a human being go through like rapid progress to like, what do I do now? You know, it’s the same tool that got me here is like not helping me get to there. So combined, those 2 sides create experiences where, you know, they don’t feel as good as maybe they can. But again, I think there’s opportunities with that. You see a lot of people like even at conference like trying to sell stuff like that. Right.

Adam Weeks:
I wanted to go back to one thing you said about, you know, the AI will take you down a path, right? It’ll take you down a path that maybe you weren’t anticipating. How does that feel as far as correcting the AI? Like, no, no, AI, you’re wrong. This isn’t the right way.

Nate Stewart:
Yeah. Well, I mean, you have to come up with your own system. And again, the more personal you make it in terms of here are my like standards and the way my workflow, what works for me. And eventually you build up kind of what you know you need to push against. And sometimes you do need to push against. You need to be like, no, like this is what I believe and this is what’s important to me. And also in terms of standards, like I think WordPress actually is in a good spot and a lot of people don’t give it enough credit where there’s so much standards, there’s so much thought that’s already been put in. Anytime there’s a, ecosystem around a product or platform that’s been around for a bit, like, it actually works really well with AI to keep it on more of a good path than bad. So if you have something that’s more likely with a platform to put you in a good path, and you have enough of your own principles that are like, hey, this is what I want, right? And it proved, right, you’re probably gonna have a good time.

There. And actually, again, I’ve heard a couple talks, uh, like yesterday where they were kind of talking about that, coming up with a combination of Okay, whether it’s Gutenberg blocks or whether it’s some existing part of WordPress combined with what did I do with AI to help put my own personal opinion in. I think there’s one about like RT Camp did a talk about their agency and they put a framework in place that just happened to work well with AI because of standards and taste.

Adam Weeks:
When I think about pushing back against AI, whether it be taste or standards or just it being wrong. A silly example is a while back, I don’t have to remember, it was kind of the big thing. You ask your AI, how many Rs are in the word strawberry? And it was very confident that it knew that there were only 2. And like, no, no. It was pushing back. And it’s odd to disagree with an AI that seems so confident. This is what is correct. And if you’re not paying attention, it feels like you can like, yeah, that’s correct. And then shipping stuff that is not because, yeah, I know. Wait, no, no, I gotta actually lean into this and be willing to disagree with my AI. And that confidence to do that, to pay attention to that. Yeah.

Nate Stewart:
Yeah. Yeah.

Adam Weeks:
How are you seeing AI in your own, as an engineer, AI impact your own kind of workflow? Like, you’ve done things probably for years. It’s changed over time, but how are you seeing your workflow change at this time?

Nate Stewart:
I’ll call my wife back later. Sorry, watch this. So I think, well, first off, the way that I see a lot of this moving is kind of how I felt myself as an engineer, and now it’s just I have permission to be more that way, which is more of like product engineer. I think a lot of us are turning into that. And you kind of have to balance both the product decisions and how you want it to feel as well as like, okay, how is it being solved?

You can give all that up. And look, I went along a journey that everybody else goes along where like, okay, how much can I give up for that? And you give up maybe a lot, and then your experience ends up telling you, ah man, it doesn’t feel right. Like, why? Why? So a lot of, from that initial point, I mean, now it’s been a while since I’ve gone down like the agentic coding side, but now I have a lot of my own principles more clearly defined now, where I know that I need to protect them. Like, what’s important to me on the UI or in the backend, or like what components or reusability kind of— I know things I can pick up and say, I really want to use this because it gives me a good end result. And I’m kind of thinking from both lenses. Like, is it helping the product and is it easy to engineer not only today, but I need to know how it works kind of, like enough to quickly iterate. Otherwise, it’s a problem. 2 months in the future where I’m like, man, just another run and I’ll complete this. I’ll hit the enter button one more time. And it’s like 4,000 enter buttons. And what did I build? So just trying to build a system that doesn’t equate to that.

Adam Weeks:
Yeah, just pulling that slot machine and like, what happens now? I win! Or, oh no, I lost my tokens again.

Nate Stewart:
Yeah, it’s very similar to where you first go in the casino, It, the first time is kind of like all the lights and it’s really cool. And then you start putting your own capital in there and you’re like, wait a minute. And then eventually, like, you have to go through it a couple of times to be like, you know, unless you’re, you know, making it a career, like some people go down that track, but most part you’re like, hey, it’s fun, but I’m gonna like go there, then eat at the buffet and then go see a show. And then it’s more than just going there. It’s kind of, you have the full spectrum. I think that’s the same way where, It feels good to pull the lever, but what are the other responsibilities? And where do you, again, where do you have taste? Where do you need to think about security or reusability? Or when you bring someone else on, your dear friend starts working with you and they’re like, what is this thing? And you have no idea.

Adam Weeks:
I don’t know what I built.

Nate Stewart:
Yeah, it’s almost easier to rebuild the whole thing sometimes than it is, which we should, that’s I mean, good engineering, you shouldn’t rebuild the whole thing every 4 months.

Adam Weeks:
You’re just burning tokens to burn tokens that don’t need to be burned.

Nate Stewart:
The cycle is not a good cycle there.

Adam Weeks:
Yeah. Yeah, it’s just going to get more inefficient over time. And yeah, sometimes you pull that lever and something magical came out, but it wasn’t actually good. Like, it’s still magical. Like, it’s amazing it did that thing, but it’s not the thing you want. And yeah, that’s— I want to go back to this idea of burning tokens for the sake of burning tokens. In your workflow, are you getting better at just not always pulling the whole slot machine and having to start from scratch? Are you getting more efficient in your token usage, or how are you doing that? Yeah.

Nate Stewart:
I mean, at runtime, we focus a lot on this. So I spend most of my time thinking about how to make things it’s not like no AI, but it’s more of like, what’s the appropriate thing? Like, what’s a workflow and what’s like non-deterministic versus deterministic? Like, where am I giving it up to AI and like, I’m okay with it thinking versus, hey, this runs like every time kind of the same way with this loop or like agent that I’m building. Why am I letting it decide each time? I just want it to run on the computer each time the same way.

So then sometimes you get to the point where certain parts of your workflow didn’t really need AI in the first place. They just needed to be understood quickly, which it can help you with. And then you piece it back to like, okay, pull that down, run it in a sandbox, like build a script that like does it each time. So a lot of what I work on is trying to come up with what are the building blocks and components and how do I use it in the right place? Make the repeatability—

Adam Weeks:
Yeah. Important.

Nate Stewart:
Like I want the outcome to be raising in quality over time and be more understandable each time.

Adam Weeks:
Yeah, I think that repeatability is important because if you’re building houses or something and you build a crappy house and then you’re going to like, oh, we’re going to destroy it and build another crappy house and build another crappy house, you’d actually need a bunch of crappy houses. You needed maybe a bigger house. So like, all right, let’s build a really good house. It’s foundationally secure and safe. Now let’s add a room to it.

Nate Stewart:
They architect it.

Adam Weeks:
Yeah. You’re building that thing that’s quality rather than starting it over from scratch each time because it wasn’t that good to begin with.

Nate Stewart:
100%. Yeah. And actually a lot of what— that’s architecting. And I think a lot about where you have kind of like your product architecture or the architecture of just like your code in general. The more you think about it as an overall system and how it relates to one another, you can start to make decisions on Okay, that part feels good. That part I’m going to increase the maintainability and repeatability, and I’m gonna document it. And I don’t need to change that as much anymore. I need to make like the right interfaces and the right like connection in the user experience to the new thing I’m building. And then that might have like, we can, that’s a good cycle. ‘Cause you can go through rapid creation, then it kind of sits and yeah, the foundation settles. And then you, you know, you live in it for a little bit. And then you extend it intentionally. Like, that’s architecture. Yeah, yeah, right. So a lot of that is where my mind tends to go. And then when I can’t get to there, I start to go, okay, well, what is the missing component of that that actually helps the next time for me to think architecturally rather than just like pure generation?

Adam Weeks:
Recently in the news, it came out that Claude— there’s a lawsuit in the EU and they’re bringing it here where essentially we’re watermarking or basically being able to prove that what Claude created was Claude’s. We know that it was AI-generated. And not just AI-generated, but that Claude generated it. I think Google’s been doing it for a while now, so it’s not new. But is there anything that’s making you nervous as like, hey, at some point you’re gonna be able to look at a code base, do a search and like, oh, 70% of this was built with Claude code. Like, is that something that would make us nervous that like they are like, oh, do they own my, the thing I built?

Nate Stewart:
I don’t know about the owning because I mean, I think where it comes from, at least I hope it gets here to where, that was trained off of all this human knowledge. So some that certain people would say like, well, maybe they shouldn’t have had that in the first place, they didn’t have permission. But I’d say because of that, we should be able to retain IP and all that because it was kind of ours in the first place.

Nate Stewart:
So yeah, I think it’s more about— well, I mean, one thing I’m trying to process, Claude code kind of already, you see Claude and Codex on a lot of PRs, on a lot of code. I think we’re kind of already living in that age. It’s more of the intent behind the usage of AI. And what I mean, like, if your intent is to get to something that other people can depend on, like either yourself or if you’re sharing the code, that intent bundles in a lot of it. Like, if someone else depends on what you’re building, you’re thinking not just about yourself, you’re thinking about, okay, like, what if they run into other things? What’s my documentation? Like, all of that.

Nate Stewart:
And if you’re using AI to help you with that and just thinking about the best architecture and all of that, I don’t mind that as much. I mind it where it is, it’s just pure form. I didn’t read anything. I didn’t think about it. I just tried to get out of there as fast as possible. And then you just get like hundreds and hundreds of things that are very similar and confusing. And it’s not for any of the users. It’s almost for you to say, I did this thing. So I don’t know, that feels uncomfortable. And I wonder like if there’s a world almost where there’s like a bucket of like, I don’t know, like imagine, trying to think about it, like when I was on like talking to Luke and Jonathan last week at like the GeoCities of like today.

Like I wouldn’t mind if there was a site or kind of a community around, look what I did here. And I don’t want you to depend on it. I just want you to learn from it. And I’m super, super excited. I pulled the lever and I got it. Then when I’m looking in that community, it’s like, oh, I might learn something, I might do it, but I know where it is. When you— we don’t have that today. We don’t have like the— it’s not even a watermark thing. It’s like an acknowledgement by the person that created it that I was just having like fun. This is for me personally. I don’t need it. So again, I think a lot more of because we don’t, easily separate out, like, can I depend on it or not? And everything looks reasonably polished from the get-go sometimes. It is confusing.Watermarks and all that help with it, but I don’t think they’re like one-to-one. Because like you might have something completely generated, but you did a really great spec. You have like 10, 15 years experience in that domain.

Adam Weeks:
Right.

Nate Stewart:
Well, I should probably trust it more. Than someone that like coded it by hand and has no experience.

Adam Weeks:
Right.

Nate Stewart:
Right, so it really depends. There’s kind of a matrix there.

Adam Weeks:
That they truly have no idea of understanding what they built.

Nate Stewart:
Yeah, yeah, yeah. So it’s like, but I’m more just personally, it’s kind of, I like to find somewhat of a balance in things like this where when we’re too close, like it’s a binary endpoint of like, do you agree or disagree? It’s like, well, I don’t know, I kind of want to use it and figure out like, the right world is in the future.

Adam Weeks:
There’s a gray area here, and that’s okay to be like, well, I like this part and I don’t like this part. And actually getting that feedback from people, I think that’s kind of an interesting idea where, all right, I pulled the slot machine and I got a prize out of it. I don’t want to share it with you. And I’d like to see if you think it’s also a prize or maybe you think it’s terrible. Exactly.

Nate Stewart:
And I think games are having that moment in a bit. That’s kind of where it seems we’re first going to have the more active discussions because you have game creators where— it’s actually a very good analogy, I guess, because if a game’s thought of from first principles, they have their own story, they have their engine, all that. And you play and you’re like, wow, they really— I really felt like they cared about this. That’s great. But you have people trying to vibe code the whole game. And it’s like, is that taking away from creators or not? Like, well, right now I kind of know it’s not a complete game. it maybe gets to the point to where it can, you know, overtake a lot of the game development, but still you need a great story. You need, you need to care enough about the end outcome.

Adam Weeks:
And taste. I think that there’s still that element of taste.

Nate Stewart:
100%.

Adam Weeks:
That is what makes us human. And part of it is that AI is just taking all the other stuff and then like guessing what you want. It’s that generative, you know, part of it. Yeah. But it’s not new. It’s just taking all the old stuff and—

Nate Stewart:
Well, and I think that’s what people push back on. I think it’s rightly so. Like games, a good thing, there are a lot of games come out and they don’t really want to tell you that they use AI. Well, I think that’s terrible to like go in there and instead of somebody really thinking about the story to generate it all, that meant you didn’t care. It’s not the fact that like AI was used there, like pure form that people are mad. It’s you didn’t care. So just like, let’s not use any tool to prevent us from caring.

Adam Weeks:
I love that.

Nate Stewart:
Yeah.

Adam Weeks:
Because one of the things I think about with AI is that I don’t care how you got here. I don’t care how you got this outcome. I care that you own it, that you stand behind every single piece of it. And so in this game analogy, that the entirety of the game is something that you truly believe in, You care about it. But if you use the magic of AI, or you pulled that slot machine enough times to get a thing you really wanted, then I don’t care. Because on the other side of it, I get to play a fun game.

Nate Stewart:
Yeah. And I think, again, a lot of what I’m seeing here— I think we have a couple more talks that are related in the next few days. It’s finding that balance. People are like, hey, I really care about this aspect of content development, that lifecycle. And, you know, I’m not an engineer, so I built something, you know, with AI to help me do, you know, use my experience or the inverse. So we, yeah, we need to come to terms with that, but you really shouldn’t remove, you know, the act of caring.

Adam Weeks:
Yeah.

Nate Stewart:
That’s just like A general life thing. Like, get outta my face if you don’t care, you know?

Adam Weeks:
Yeah, if you don’t care enough to, yeah, don’t just ship something that you don’t care about. That is your human, you know, what makes you special is—

Nate Stewart:
Yeah, it’s like you’re not a painter, you’re not an artist if you don’t care.

Adam Weeks:
Right, yeah.

Nate Stewart:
So if you’re trying to like come across as an artist or anything of anything where like, hey, you didn’t do the work, like, and really teaching ourselves to be honest about that is maybe just a cultural shift. We need to do it. You can create art, but you’re not an artist in that form. And if you wanna be that in the future and the tool’s helping you like get enough confidence to now go like pick up the paintbrush or like, You know, actually touch things and understand, get the taste yourself, then that’s probably fine.

Adam Weeks:
Yeah, I think, yeah, if you don’t care, go do something else. Like, find the thing you do care about and do that.

Nate Stewart:
Yeah, and if it’s an experimentation thing to find what you care about, like, that’s actually a good human thing to do.

Adam Weeks:
Yeah, and a good use of AI is to figure out whether or not you care about this thing that, oh, I wonder if I like video game development. Oh no, it’s really hard. Like, that’s like, all right, go do something else. You know, find the things you care about. AI can probably get you further along that path of actually what it’s like to develop a video game than ever before. Yeah. As we’re wrapping up our time, you know, it’s great to be here in person at WordCamp US. And I love the, just the connections that we can have. This conversation about WordPress in the age of AI, I’d love to have your final wrap-up thoughts on, is there anything that you’re seeing that is getting you excited about WordPress again because of AI?

Nate Stewart:
Yeah, I think what I’ve seen that is, it’s not surprising to me because I’ve been tracking it a while, but Gutenberg, it seems there’s been a shift into now people love blocks and they love the standards there. And I think part of it is because AI or otherwise, like that, they’ve refined it enough to where it’s really useful and the UX is like, you know, pretty good at this point. So I’m excited to see where the community takes that. And I think that could be a big strength. Just in the market in general, because there’s so many people that depend on it and the standard is so refined enough to where, and along with WordPress AI, which I’ve dabbled around a bit in the past couple weeks, you could make it your own in a way that’s still like stable and standardized. And you’re not gonna break your whole content management system. So that’s probably what I’m most excited about and happy to see. ‘Cause a couple years ago, remember like the editor was a big— Like what’s going on?

Adam Weeks:
It’s terrible.

Nate Stewart:
So I think those 2 bets there of, well, AI’s early still, the bet for WordPress still, but It is an actual investment, almost like Gutenberg was. But Gutenberg was the React and JavaScript and all of that. Those standards are now really easy to work with. Got it.

Adam Weeks:
Very cool. Well, Nate Stewart, thank you so much for taking the time. You built a cool thing. You got a co-founder, Runtype. You’re building these incredible ways for people to make AI tools for others. I think that’s pretty fantastic. I’m excited to continue our conversations around this topic because I think it is a lot. I’m really glad that we got to meet here in the desert to be in person because there’s nothing that beats this. If someone is curious and they want to find out more about the work you’re doing over there at Runtype, how would someone find you?

Nate Stewart:
So runtype.com. Here, like, you know, I don’t know when this is going out. Like, talk to me, nate@runtype.com. And then Become Vocal is my All right.

Adam Weeks:
Become Vocal on X. Excellent. Well, Nate, thank you so much for sharing your thoughts on, yeah, on what’s happening in this crazy age, you know, in WordPress. And I really appreciate it. I’m looking forward to our future conversations.

Nate Stewart:
Yeah, me too.

Sponsor Announcer:
All right. See you soon.