Stop building AI features

In this talk, Ronnie Burt, a product team leader at Automattic, discusses the future of websites in an AI-driven world. He explores the importance of identifying what the website of the future looks like, particularly with the rise of bots and AI agents that may constitute the majority of web traffic. He emphasizes the need for websites to serve as the primary source of truth and how they may need to evolve into more niche and specialized platforms to remain relevant. Burt highlights the significance of structuring content to cater to AI tools for improved discoverability. He also touches upon Automattic's approach to integrating AI capabilities within their products, including developing APIs and tools that enhance collaboration among creators and developers. The session emphasizes open-source principles and the importance of building a trustworthy web ecosystem.

Oh, good afternoon. Thanks for showing up on the labyrinth of how you get here in what is probably like the 50th session with AI in its title for the last day or two. Um very happy that you're here. My name is Ronnie Burt. Uh you can if bored, you can read a little bit more about me at ronnieburt.com. Uh first I'll say before I got into the

whole web thing, I was a middle and high school math teacher for 6 years. And so I'm very comfortable with being interrupted and things not going as planned and we're a nice little group here. So feel free to like interrupt, ask questions. Just don't raise your hand and ask to go to the bathroom or that sort of thing like my students used to do. So I'll stop

with the terrible jokes and then um I work for a company called Automattic. If you're not familiar with Automattic, we're better known for our WordPress uh hosting and products and suite of tools. And so you might be asking, why is a a guy from Automattic who spent the last 15 plus years deep in the WordPress community here at my very first DrupalCon. Um which is going great,

by the way. You I I I listened to my first DriesNote this morning and I think a lot of the things that I heard there really resonated with the same things that we're thinking about, the same situation that we all find ourselves in with the uncertainty, the anxiety, the uh you know, what's going to happen to us all and what's going to happen to the web in

the near future thing. And we're wrestling deeply with those questions. I lead our product team at wordpress.com working across our AI products and tools. Work with our AI teams uh across the company to figure out how we plan for this next iteration, this crossroads that we're at of the web. And so, the first question, kind of an obvious question, is what is the website of the future?

And this question is kind of funny to me because we have at Automattic, one of our advisors is Stephen Wolfram of Wolfram Alpha and Mathematica fame. When I was a math student, I had to buy his software back in the day, right? And he uh meets with us every couple of months where the whole company can kind of like he he's just an amazing question asker. He's

just phenomenal at like drilling down into some point that you hadn't thought about. And month after month, session after session, he kept asking us this very basic question over the last 6 months. What is the website of the future? Why are you guys not talking about it? Why are you guys not like trying to lead and and figure this out? And we were completely flat-footed. We really

did not have an answer. still we have a bunch of different answers, right? And we're trying to figure that out together. And I think that kind of can be a frame for the conversation as we move forward, thinking about, well, what is that website of the future? What does that mean both from the like consumer visitor side and the builder creator side? And the honest truth is,

right? None of us have a clue. Um the it feels like every like week in AI is 7 dog years or something. Like it just keeps changing constantly. Uh just as I was standing here about to start the presentation, I was reading new tweets X posts about Shopify's new latest thing and like how that supposedly changes things. And I'm trying to figure that out and go I

mean it's just constant, right? Um so I just we don't have the answers. None of us are experts. We're all building this and figuring this out, but the important thing is that we're asking these questions. We're trying to experiment. We're trying to play. We're trying to see what sticks with builders, with creators, with users and visitors. And we got to be prepared because it's going to happen

either with us or without us. One of the things about the website of the future, I I live in Austin, Texas and we have the South by Southwest Festival every year coming through town. Um it kind of takes over and last week one of the headlines was from the Cloudflare CEO that they predict that sometime within the next 12 18 months, the majority of traffic on the

web will be bots and agents. Uh before the AI uh kind of 2 plus years ago, it was less than 20% of all traffic on the web was bots. And now it's greater than 50 or pretty soon we'll think it's greater than 50% and it's a different type of bot as well. It's a bot that's smarter that can act on someone's behalf or that can do things.

That's one thing about the future of the How people are interacting with and experiencing the web is changing. This is just one example of like all the browser wars that happened like just in October of last year. I gave a version of this talk twice in November and I was thinking, "Oh, I could just reuse those slides." This is the only one that stuck and it's probably

the most outdated slide because it's all changed, right? The people are we have Gemini baked right inside of Chrome if you're a Chrome user. Not necessarily all of us using that or finding that the most useful way of interacting with websites, but we are probably getting more to where our Claude instance or or whatever can work on our behalf within the sites that we're visiting. You have

things like Claude Co-work. If you can squint real hard and see the date of when Claude Co-work was launched, it was like 10 weeks ago. it is now like I spend at least an hour a day in Claude Co-work. I've moved things over there. It's how I start my day. It's how I end my day basically with different skills and different things that it's running for me

to just keep on top of it all. And it's doing that a lot by crawling different sites on the web. Some are private, some are public, but it they're still websites and instances in general that it is able to access and and do things for me. So that whole point of that, right? Is we're at a time where if you're not already thinking about it, we got

to rethink who our audience is. Our audience is increasingly not human for the sites that we're building, creating, maintaining, or whatever it And what does that mean? My answer to uh Mr. Wolfram's question, what's the website of the future? Is websites are the same thing that to me they've always been for 30 plus years. They're the single source of truth for that organization, that government, that business,

nonprofit, or person, right? That is where you go to find the most authentic up-to-date information about whatever it is that's on that site. And that is only going to increasingly be more important. So websites will be where agents go to find the most up-to-date news and uh or data or information on the site. It's the only place that we'll have really to uh kind of show the

unique voice, that unique perspective, that uh makes whatever that organization or that site is about different than any other site that's out there. And so I think in general sites are going to be smaller, more niche, more topical. instead of these catch-all aggregate sites that that kind of do everything part of that is maybe if we're lucky some of the old SEO tactics of the day are

going to be less and less relevant and it's morphing into this whatever we're going to decide we're going to call it, AEO, GEO, AIO, all kind of the you know, how we're displaying the content so that it can be discovered, so it can be found. It's no fun creating a website with no visitors, no traffic. And just a most obvious example that like frustration point for me,

right, is like you have you tried to find a recipe for anything and you have to to like doom scroll all the way down to the bottom of the site because there's a giant wall of text. Well, why is that wall of text there? SEO. That's the only reason it exist. The person didn't really care to tell you the story that they wrote at the top of

that recipe. They were hoping that recipe would be discovered by the search engines of the past. that's an example of well, we're building for a different type of visitor and a different kind of discovery engine that's not the same as the web crawler or the person that's coming through. Instead, that recipe needs to make sure that it's structured, easy for the AI to determine that it's a

recipe, easy to know that each thing within the recipe is an ingredient or a measurement for the ingredient. Luckily, there are nice standard web standards for recipes, so it's a pretty obvious example, and I think most of those sites do that anyway, but you can extrapolate that or take that to like any type of site with any type of data. The actual structure doesn't matter as much

as the fact that it is structured so that the LLM can figure it out. This one is is still really fuzzy. This is more like a future thing of the web. I don't think anybody has solved this or is doing this particularly well. It's just kind of my personal like where I kind of think we're going, more and more personalization. There's going to be this desire for

some type of sites where it the browser may know some things about me, my likes, my dislikes more than just my location and language kind of thing, or the agent that's visiting the site on my behalf may know that information, and maybe that will be able to be shared as a page is created on the fly, unique just for each and every single visitor. And so, we

have to be able to build a platform and a website that's ready for those sorts of pages or pieces of the content that's completely personalized. Um you can see uh parts of like most obvious reason why this doesn't work now is it's just super expensive and inefficient and slow. Um because you would have to send it off to like OpenAI or whatever and wait for it to

come back and then generate it. But that's getting faster. LLMs are getting better local. You can imagine like an LLM on each and every single device that can run this right there in real time in a more private privacy-forward way. This is just kind of the picture that we're painting of the future of the web. One that is more personal, one that is more directed at bots,

less directed at humans perhaps, or at least needs to be able to be interacted with by those agents, by those AIs. And so, we'll think about that website of the future, but it's we're also kind of like flying the airplane while we're building it. Like, who's building these websites of the future? And is that changing? Or the ways that they're building those websites of the future, is

that changing? What tools are they using? Where are they building those sites? And is that changing, right? The answer is yes, it's all changing. And we've got to figure out how we are going to put the pieces in place, like the LEGO bricks that are there, as we build those. so, we can meet the people where they are, whoever's building the sites or creating that content. And

so, I I'm going to come back to that point, but I'm going to sort of a story in January this year, late January, had a lot of teams across Automattic working on a whole bunch of different projects. And I'd been working on our like AI website builder tool that was not generating any code. It was completely just like picking what we call patterns and like picking what

it was best, so it was There was nothing generative really about the tool. then we just had across the company and outside the company that you could see on social media or or wherever, twice a day, something it would just keep getting leapfrogged and leapfrogged and leapfrogged. The ability of the output The output was just getting better. Something about this January that just like really got us

all by surprise. New models came out from OpenAI and from Claude that were just noticeably better at these sorts of things. >> [snorts] >> And so, we had been spending the last 2 years building these features, these tools, these builders, things within side the WordPress dashboard, or things for content creators. And we basically just all at the same time looked around at each other and we're like,

we are not going to be Claude. We are not going to be Open AI >> and and having a better outcome having a better output of what it's creating. Like every time we get like a little bit better, the models change, and we're we're back we're back behind. And it was just kind of just shift I mean we're really talking like 2 months ago of okay, well,

if we're not going to try to build the tools within the product that are going to be better than these frontier models, what are we going to do? So, what we're doing >> [laughter] >> what what we're what we're trying to do is put the building blocks in place so that no matter if you're within side like what we call WP admin in the WordPress dashboard, if

you are inside Claude code or Codex or Cursor, if you're using Telegram, our SMS, you have access to the same tools, the same abilities, the same APIs and everything to manage, build, create, update all the things for the site. And that was 2 months ago we decided that we're really going forward on this, so I'm not saying that this exists, but this is kind of our vision

that we're we're pushing on and we're encouraging all the builders that have things that are plugins in the ecosystem, like we're providing the framework so so anyone can build on top of this in a very open-source way so that you don't have know a new interface. Uh the context that's known for your tool and for another person's tool like is all shared for the site and for

the content creator or the site creator, whatever it is. This is kind of the idea. feel like I did a pretty bad job of explaining it actually like it's it's still new. We don't we don't know exactly how it's going to work, right? >> Um but it feels like the direction I think we're all going to have to move, all ecosystems, all products, all sort like we

we're not going to beat the clouds. I don't think that the Lovabots and the Replit's and the V0's are going to beat the clouds um long-term, right? Because it's there's you can just feel it every time that there's an update. One of the things I heard which was really cool in the in the Driesnote was the context control center and it's very similar to what we've been

thinking about with something also going into core of content guidelines. It sounds like it's pretty much the same, which is cool. And I brought it up because that is one of the building blocks that is available so no matter where the person is interacting with the AI that's building or managing or maintaining the site, they have access to that context, that memory for that site. Um so

we're building in a way that it's you can read and write to it from uh we have it working in a Slack app or Telegram uh bot or within the uh dashboard itself, right? And this this is an unknown. I feel like it's new within Drupal as well and I'm very excited to kind of learn and see what what happens with this. We also were very careful

to which I heard today, which was cool, this isn't totally only about AI, right? This is useful to anyone who had has always had content guidelines that was probably in some random doc or PDF somewhere that had to be manually updated. Now we can put it right within the CMS. Uh this is one of those examples of a a building block that [clears throat] we have to

make sure we give access to everywhere. So if I'm also drafting content in Cloud, I have access to that same context if I'm drafting content and drafting content is my least favorite example. I don't know why I use it, right? I don't want AI writing my content. Um but if it's suggesting improvements or one of the things it is good at is if you give it those

content guidelines, it'll check and make sure that um you know, if you capitalize certain words but not capitalize others and that's in your guidelines, it will catch that for you and that's that's like something it's really good at. and so that's a an example where we want that to be no matter what type of uh tool that you're using. In the session this morning, this was kind

of said in passing and then I didn't realize what it's what he said until it was too late to get the exact quote. So I'm terribly paraphrasing here. So like don't take a picture and say that like the WordPress guy here is is messing this quote up or whatever. But I heard something to the effect of AI for developers doesn't belong in core. That the AI for

content management does or something like that. Did anybody hear that? Do you know what I'm talking about? Uh Is that how you heard it? Um it caught me a little off guard because that's actually kind of the opposite of the way that we're currently thinking about it. And I'm curious like why I don't know who's right or wrong. It's like this this where we've landed. And um

So that that brings us to like the idea of AI first development and we're really leaning into that and come in the outcomes of that AI first development and making that available. And this is across everything from core to all of internal projects that are only like used by staff all the way up like the whole gamut of what what we're working on. In February on like

a random Monday we all got a message that said drop everything you're doing for the week and go through your repos and through everything that you're doing and spend the whole week like optimizing for AI first development or AI first like being able to be discovered by our tools internal external whatever. Some teams were much further along with that already they didn't need the whole week some

like you know are still catching up but it was this idea of really leaning into things like agent .md files in all of our repos and I saw that I think that like for now at least that's not something that Drupal wants in in core. And there's probably really good reasons for that. For us like what we found um it's not the developers that are getting the

most benefit from this focus that we're working on. It's all of us like myself more non-developers in the company. I can now use our our tools I mean just within the last month we have thousands of repos across the company across the different suite of products and projects and everything. >> [clears throat] >> instead of having to put in a request to an engineer to go dig

through like where something is or how it works I can just ask my cloud like can you give me like a little bit of an architectural like can you explain to me how this works? And it's probably not perfect uh but it's fast and it gives me a pretty good approximate animation that lets me then decide if I need to go bug the engineer and take his

his or her time, right, to uh, take the next step and and dig in even further. It's letting our design team create even better, like, more accurate mock-ups and prototypes with actual code base because they're able to access it now. Um, now that's not stuff that we're shipping to production anywhere by any means, but it's it's just really changing the way that we're work that we're working.

Mostly for the best. Uh, some of the ways that it's not for the best probably is that now instead of, like, two or three mock-ups or something that someone spent a lot of time on all the little details, uh, that we're looking through for something new that we're going to do next, we might have, like, dozens that we have to look through it and figure out. Some

are kind of slop and and junk or missing things, and some are better than others or pieces here and there. Um, so we're going to have to rework that rhythm and and figure out like how we prioritize in that. It's not not here to say it's perfect, but we're putting in the work there to make it discoverable where the end goal that we're getting closer and closer

to is someone can start a conversation and can plan, build, buy by domains, provision a site by hosting, all of these things, uh, without ever an actual, like, traditional funnel or or UI on any of our properties. Um, that's kind of the ex- expectation that we have for where we're going. Do we know if people are going to buy that way? No. That's what Shopify just announced

that now you can buy all the Shopify stuff through, um, Open AI. How are they doing it? I don't know. Like, are people going to feel comfortable buying? We don't know, but this is just kind of the time we're in of experimenting and and figuring out with what's going to stick. And so that kind of brings me to the open source you know, why open source now

more than ever? I've been in in the the WordPress ecosystem for over 15 years because of open source. It's why I'm at Automattic because we are like a very much an open source first company. And I know that the Drupal community and everything around it is is very similar. There's a lot more similarities than differences And I think we're all wrestling with two different kind of threats

and or things to be aware of. One is what we've always kind of faced, the SaaS closed proprietary systems. And now we face the bespoke thrown-together someone dangerous with code like me building something they shouldn't, that they have no idea how it works, um and putting it live and in production. That's not on any platform, right? It's just completely just created out of uh thin air. And

we're seeing signs that like we heard uh this morning in in the DriesNote right people are finding that after they have a little bit of success with that site, with that app that was kind of one-shotted or created that way, uh they need to go find find a new home uh for something that scales better and grows better, is more performant, more secure. These are all the

reasons that um we need open source platforms to build upon more than ever. The good news is that these LLMs are all trained with our code and our data. Um they know Drupal well, and they'll be able to build on it that much better because of all the training there that you wouldn't have in these proprietary closed SAS Um and I more important just in general as

a general comment, you know, authenticity, trust, knowing what's real, what's not real. I mean, these sound like I'm I feel like a crazy person saying it up here, but this is like a real problem that we're facing now. uh the more open the platform is, the more visible the code is for people to inspect, the the more trust there can be, the more authenticity there can be.

And so we all have uh to just keep pushing for for more of that in the expectations of the tools that we choose and the things that we build on and build with. And we need that. Otherwise, we're going to end up with like, who knows, right? Like a world where we have no idea who's taking what, what's going on behind the scenes. Um and so I

expect more and more overlap of projects that are popping up that all web properties will will be able to take advantage of. Kind of open uh open source the web MCP idea, which is like supported in Chrome now, that is still kind of untested and we don't know if it'll stick as well, but like these sorts of standards. I guess I'm just I don't have a specific

standard to advocate for, but I know that we need more that we're all building from so that we can um quit playing this guessing game and and get ready to to support all the things that are coming. I without knowing for certain, I can feel pretty confident that the Shopify thing listed announced today, which is like uh a gentle commerce, is not going to be completely open

source platform, right? That we can we can build off of. And that's just one example. we need to keep looking for opportunities for us to learn from each other, build together. Um so that we can establish and decide where it is that we want to lead and where it is that we want to follow. I rambled quickly and nobody stopped me. So then we have a lot

of time for like every How do How does this sit? How are you feeling? Anybody have any any thoughts? Yeah. Yeah, I I mean that Drew's don't that whole thing about the module like bringing AI into the code just yet. Like not having the agents empty file in core. I I think it's a tricky one with politics and I think if I'm getting it right, Automatic is

a open a company supporting open-source software, but as a company you have a policy Mhm. that we should be using these agentic tools. Yes. But and and I think what Drew's was distinguishing it wasn't his keynote's not about his company, but about the community. >> Right. And kind of treading He's not I'll not speak on his behalf, but he's treading lightly. Oh, we have the same thing,

right? I mean our Our CEO is also the founder of WordPress and the same similar. I get it. So what I mean, I think as a company it's pretty easy to be like, "Use these tools." And It's It's like either get on board or I don't even want to say what the old term really is, but we know what it is. But how is Automatic or WordPress

handling the community with that? Like the [snorts] And I you have a much bigger ecosystem of modules. And I mean, I feel like I don't know if it's happening, but I I get the sense that you might have a bunch of people doing AI tools for WordPress. There are. Yeah, I think I don't [clears throat] know that we're definitely not getting it perfect. Right? But for at

least the last 18 months, like laying the foundation and the groundwork, so we have this thing called abilities API as part of core, officially supported in that when it was by memory, I think about a year ago >> had no real utility. There was no obvious advantage, right? To adding it. But the vision was there. We knew that we expected that these other things were coming, so

that this abilities API would be more and more valuable. And that abilities API can also be tapped into by our plugins, our our version of extensions, right? And so any of these plugin authors can register an which is essentially like saying this is what my plugin can do, this is how you do it, how you can interact with it. And then no matter what agent you have,

if it's the agent that is like kind of part of by provided by your host or by the agency you work with or you bought through another tool or called directly can discover these tools through these abilities. And so I the hope is that by laying that foundation early and making it clear this is where we're going, we gave people plenty of time to I opt in

and kind of see this is like a direction that that we think we're going. Um there are things in uh you know, choice is another piece. So everything we built also for the company side there's a simple toggle. I want nothing to do with AI, I'm going to click this toggle. I got nothing AI anywhere in my face, right? And we have lots of users that are

are there. And we need to meet them where they are. on the core side, it's uh a way for if you're building an extension and you have decided that you can let someone use your OpenAI API like you have a connection and they'll pay you, right? You'll be a kind of like a middle person for that. You can also register it so they can use whatever API

or LLM they want. They can use a local one or or however. So, we've tried to build those things around it, so you can choose your LLM provider and even be a even be like a self-hosted or a local or an open-source one as well. It's not winning everybody over. A lot of skeptics. I mean, I think we're all skeptical to some degree, right? And it's just

that we feel like we've got to be in a place where we don't we don't get left behind, basically, right? So, we have to try. Any other thoughts or questions or rants? Um how much longer do you think that toggle off is even going to be available? It's a good question. I think until we stop getting complaints about it. I mean, it's every marketing announcement comes with

way more complaints about turn it off than people excited about announcement. Now, you're going to hear I don't think that that's an accurate representation of the like the perception of the people receiving those marketing announcements. I think it's just the vocal minority but I mean, the beautiful thing about all of our open-source tools is if we take away the toggle, someone can put it back, I guess.

Um and we try to build it in a way that that makes that easy and and is supported. But, I I I keep hearing story after story, I guess, of people wanting nothing to do with AI, and then they find one use case that, like, they're comfortable with and happy with that changes the way that they interact with it, and then all of a sudden much more

susceptible to the next use case and then the and the next thing, Yeah. I just like to say you're brave coming to Drupal. Thank you. >> Not as a developer, as a WordPress developer, but not even a coder Yeah. and giving a presentation. Well, thank Thank you. It's very I was It's always It's just good for me to put these together and to think about how to

tell the story in a different way to different audience, so appreciate it. Have you [snorts] ever looked at the Drupal code? Not in a long time. So, I did build my I know that I built a site in Drupal like 15 years ago before I built anything in WordPress, >> but I don't remember it. Yeah, it's it's a completely different I I personally would not call WordPress

open source. Hm. Um you say that you can take something and change it. I don't see that as open source. I see the community deciding on something and having that be available as open source. I think WordPress is more of a company that came up with a product and wanted other people to build off of it, and then those you've leveraged those people being able to charge

for that, and that's the kind of community that is it really what what is different from Drupal is it's a lot of easier to launch because of that. Yeah. You set up a a real streamlined business was set up to have software that you could change, but if you wanted to use the whole wordpress.com etc, you're not touching Right. But, it makes it really easy. And Drupal

didn't do that. They built a that everybody contributed to to come up with. So, when Dries says, "We're not doing AI development. We're all doing AI development." He doesn't have to do any of that. Right. Not fair. You're saying as a company you're going AI first. He just knows we're all doing AI to contribute back to this system. That's That's fair. That's interesting. And I think it's

worth like there's so many different uh flavors of open source, right? And that's interesting. I think there's even like uh like would we call Chrome open source? It's an interesting Android open source is an interesting question, right? And you know, there's a range of everything. Yeah. And I I've always felt Drupal was the open sourcier Yeah. >> version of open source. Yeah. It is truly that. I

agree with that, but I'm going to defend WordPress. It's like and I I I'm like the web form guy in the Drupal community. And WordPress, there are dozens of form builders. It is pure chaos. There's no standard [laughter] API. I'm sorry. I just criticized that. I didn't build it. It was like, "Okay, that's a rough one." But, at the same time they do bring in core features

like Gutenberg where they're just like, "We need to solve this bigger problem because there's a larger agenda." So, and that is fully open source. We're literally using that code in the Drupal community to build pages. It It is a different model. But, and I agree with the statement that Drupal is more open sourcier, for sure. But, WordPress is, you know, they're they're they just created a different

community ecosystem. And I I can admire it, too, that people make money. You know, it's it's a larger much larger than Drupal because of that approach that worked in One thing I will criticize Going with the criticism WordPress is you have a lot of technical debt in the And there's a huge opportunity to be a lot more aggressive with AI advancing features. Like the anecdotal one that

I heard from Claude the Claude code creator is they built the plugin Well, they built the plugin system in a weekend for Claude using AI. They wrote a spec or like we need this feature. In our community it took years to evolve a plugin system. And I think how do you put it? Their plugin system is decent. It's It's a working API. >> And there's been [clears

throat] some talk at Automatic about how to do, you know, leveraging AI to advance the software like ambitious projects or goals that we could go after or you could go after. Yeah, I I think we're concerned actually that we built that ecos- I'm concerned. I don't want to say we're. I'm concerned that uh because before Automatic I was in the plugin and hosting building ecosystem in WordPress

for for 10 plus years. And so I think more back to those days. I couldn't have built that business because anyone could just copy it. Like they're they're not uh the the things that that we built as a company like no one's going to need now. They're going to just build they you will be able to build that bespoke theme or that bespoke like little feature that

you need and that will be okay and entire businesses in the entire ecosystem was built off of I I don't I think we're all very much wrestling with what happens to that ecosystem because that's changing. That that like builder ecosystem you can already see peaking unless we figure out a you know, like a new model, a new way. Um that's a different ecosystem than like the agency

ecosystem and the people building sites, right? I I don't we don't have the answers, but it's a definite like monumental shift and what what God is here is way different than what's next as a as an ecosystem. Yeah. Um this wasn't really a question more just like a reflection on the earlier comments about you know, that they're going to be more bots than humans online in a

few years and like you know, we're not the nature of content and SEO and stuff is going away. So, is you know, is a content management system, you know, the recipe example, like I have a hard time imagining that there's going to be a recipe site. You know, like who's going to go and look up a recipe? But if the sites are really a single and that

they're all going to continue the function as a single source of truth, then I'm almost imagining like I don't optimizing your site like notebook LM. Like I just want to make it easy for somebody to like, you know, for it to be referenced, to check the reference, but that's all they're trying They're not really that they don't need media, they don't need anything. They want an an

easy way to know it's me, to find it, confirm that the data that, you know, whatever gave them is Yeah. Can I just add something to that? So, yeah, we we did that recently. So, bot trafficking, it's like the consumer like CSS JS file or I think that is unnecessary because >> Right. It It didn't care what it looks like. >> So, yeah. I don't know why

why I mean we can't mandate like, okay, for bots we're giving only text only. So, I mean, we all kind of we kind of went down like the LM dot text thing and just like or an MD version of every page on some sites. And so far like I haven't seen any predictably the LLMs are using, right? We've We've got different versions of different different things looking

to see is traffic increasing or different type of traffic increasing because we have this new like entire markdown version of our doc site was the most obvious thing. Um I think that's coming. I Or more important now than ever to have really good structured clean simple data that's not buried in layers or duplicative cuz it'll just burn context that's not needed or whatever. I think I think

what you said earlier about sites becoming more niche. I think that's kind of It It It gives those niche sites credibility. It makes you trust them. Yeah. You know where it's coming from and there's some verification on it. Uh as opposed to sites that have gone and just scraped everything from the internet and then they just have a house to show. Yeah. Those aren't going to be

anything anymore. Um you're going to want to have those sites that you trust that content. I mean, if it is a recipe site. Right. It might be something that the recipe has that makes it more engaging. A little special nuggets or something that you're not going to get from just a general response. So, like the quality um becomes more important over quan- Before you just wanted wanted

something. Exactly. >> Now you can want the best you can get. Which is a very good argument for I am a even though we built some of the tools for like generative content, like by definition LLMs are producing average content. That's like what they're trained to do, right? I mean, we can prompt it to be better than average or unique or something, but uh hopefully we we

don't lose that voice. That like maybe it's co-created with LLM sure, edited, whatever, but um, yeah, it needs to be I mean, otherwise we'll I'll just ask ask Claude to give me a a recipe or whatever, right? Uh, so, yes. So, I was wondering uh, how has this affected the blueprint initiative? Ooh. >> the blueprint initiative is where workers called site templates and I know that was

something that was big and it wasn't going to come computationally now, but with AI and all the other stuff coming in It's actually bolstering it. Uh, we have some projects specifically around um, this tool called playground also that um, lets you load those blueprints in a browser. No server needed, no connection needed, everything lives there so you can build and play with the whole site with the

blueprint. Yeah. That is amazing for development. Like we can develop the site basically locally but in a browser for for dumb people like me that can't figure out the more complicated setup. And so blueprints are kind of back. Yeah, it's kind of funny that that you mentioned it. Um, the problem with blueprints from what I've seen is that when you start adding a bunch of content, the

way that it it's architected it gets really big. File gets really big for loading it locally in the browser like that. So, we're trying to figure out some it might be blueprint plus like a some package or some other way of of stripping the content out from the site settings and the site stuff, but some interesting things being played with there specifically for AI first development of

sites and being able to build them quickly and efficiently um, without like we don't want every line of a of a theme or a site what we call themes right of of a site being generatively built. That that doesn't seem like it's good for anybody. So blueprints are perfect starting point This is fun. Good to learn more about Drupal too. I hope to keep keep learning. Thank

you. [applause]