Episode Transcript
[00:00:00] Speaker A: Hello, I'm Jeremy Rivera, your unscripted SEO podcast host. I'm here with Ben Wills of Op Alerts. So, Ben, what would you say you do here?
[00:00:12] Speaker B: What do I do here?
Well, first off, I'm getting back into the industry. I was in the industry a long time ago and then did a bunch of engineering. So now I'm building, basically building AI search tools with a long history of building SEO tools, which I think you were part of Raven Tools.
And so, yeah, building AI search visibility tools. The thing that I enjoy doing is really figuring out how to make processes a lot more streamlined. Just taking off a lot of the hard work, putting it into applications that do as much of the work for you and just letting you focus on the strategy.
[00:00:54] Speaker A: I would say Brett Sklar, who I interviewed, said AI isn't taking a job, it's taking your tasks.
And I think that's absolutely true.
Adrian Nikola said AI is really the steroids that SEO has genuinely needed. From the beginning that we were stuck just outsourcing stuff to the Philippines to do the, the make work stuff. But being able to shot call some of the on site tasks and work and, you know, direct and define an SOP for an agent is so much more satisfying than waiting a week or two for somebody to do such a menial task.
[00:01:39] Speaker B: Yeah, yeah, you get to focus. I mean, there's fun parts of marketing and you get to spend more time on them, like figuring out the strategy, figuring out like understanding the competition, how you're gonna, how you're gonna work with that.
[00:01:53] Speaker A: So, yeah, I think Matt Brooks of seoteric defined it as, you know, like educating your off site customer support team. You know, you've got chat, GPT and you know, Gemini and all these others who are waiting out there to direct people and direct the flow of traffic, like just take the step and educate them. What would you. It's what I just told you. What would you say you do here?
Should be the question you ask, you know, you're asking yourself before you make any page on your site.
[00:02:24] Speaker B: Yeah, yeah, exactly.
[00:02:27] Speaker A: I'm curious about one of your posts that you made. You said, I think it was something about running a prompt 1374 times.
Was that. Did I see that right? And what were the, what was the outcome of that project?
[00:02:45] Speaker B: So I'm not sure I have a lot of little projects like that. Do you remember, do you remember which one?
[00:02:51] Speaker A: It was something about Honda Civic and Marriott.
[00:02:55] Speaker B: Oh, yeah, so.
So that was a test.
I would say like a third of my brain, like 33% is marketer, 33% is engineer, and like 34 is like scientist. So like I want to understand exactly how the thing works instead of relying on intuition. I think a lot of SEO and AI is you see so much cause and effect, you begin to develop an intuition for what's going on.
But I think with a lot of the stuff that we have access to now, we can, we can test in a way that we couldn't really test before, just because it's so the feedback loop is so fast. You write the prompt, you get the result, and you get insights into the LLMs that way.
So I think one of the most important things for people to understand about the prompts is how one word can completely change the results, the response that you get back. And so that specific test was to figure out how much one word actually changes the response that you get back. So I, I wanted to think of like, what's a prompt that I could put together that I'm asking for advice about this thing, but I'm going to change this little word over here. So I think the prompt was something like, I'm flying to Los Angeles to pick up a new car, give me hotel recommendations, but I changed the car that I was picking up from a Honda Civic to a BMW to like a Ferrari or something like that. And the hotel recommendations were very different.
Whether it was, it was across all three of the CHAT GPT models and across all levels of thinking. And the more thinking there was, the more variance it was where the hotels were more expensive if it was a Ferrari and cheaper hotels if it was a Honda Civic that you're picking up. So that's the kind of testing you can't do with SEO. You gotta wait for the, you gotta wait for the index to change and figure out how to get the personalization out of the way. But with the APIs, you get direct LLM access in a way that, in a way that you've never really been able to do. On the, on the SEO side.
[00:05:07] Speaker A: That is true because we have like, you know, it's not walking around and pretending that oh, hey, 2023 and suddenly ChatGPT appeared and LLMs have never existed before. This is new and magical is the dumbest worldview that I've ever heard of. Like we have had LLM tools for a long, long time and they've been, they got baked in with, I believe it was Bert was LLM based implementation. Hummingbird had LLM based heuristics in it, multiple layers of it. But you're right. It was always within the black box behind the twiddlers. As Mark Williams Cook was talking on LinkedIn today about, hey, you should know, Google intentionally has patents, has deployments, has programs to fuck with you as an SEO. Like literally 100% intentionally designed. You know, they couch it in nice phrasing and they say oh, it's to prevent, prevent abuse and you know, this delays results and you know, ensures good intentions. But no, it's, it's something in the serps that they intentionally have crafted just for our industry, for us. I feel honored and insulted.
[00:06:30] Speaker B: Yeah, it's, I mean it's this weird cat and mouse game. It's like on one hand it's like just give us just like stop the nonsense. You're just creating more work for yourself and creating more work for us.
But also like if I was in their position and somebody's trying to game my product and I have to keep the value of it up, it's like how much do you invest in making that difficult for them?
[00:06:56] Speaker A: But you're right in the moment. You know, now we can actually test these. We have an API. We can not only an API for a similar model, but you know, we can connect and see what Gemini outputs, which is a Google product connected into the Google index directly. You know the Bing index is hooked by that Promethean chain system that Fabrice Chanel talked about several times at SEOmoz. When I saw, saw him it, they're intrinsically linked. So if you want to know more about the indices, what better opportunity have we ever had?
[00:07:40] Speaker B: Yeah, exactly. And an smarty said, I think this was last week with you. She said the best way, the fastest way I think to get into the LLM responses is via SEO. And like that's the highest leverage piece that you have to get in the responses and especially with as much fan out querying is this happening and how prevalent that is? Even in the free results, even if you like, if you're scraping chat GPT responses from like not logged in, they're still hitting an index, they're still doing fan out queries behind the scenes to bring that stuff in. I mean it's just a matter of time before Chad GPT has their own search engine. Yeah, OpenAI has their own search engine. Maybe it makes sense for Anthropic to build one, maybe it doesn't, I don't know yet.
But so the idea is like, I think a lot of traditional SEOs, you know, we sort of got whiplash because LLMs came on. We're like how do we navigate this? Like, what are we supposed to do now? This is like this whole other ranking factor like supersedes everything else. Yes, this part of the, of how the responses are generated.
And I think what we're doing, going to see is as there's more search baked into the thinking process, the thought process of the LLMs, the search principles are going to come right back. You got to have good quality content. There's got to be a lot of it distributed across the web. You got to have backlinks. It's like, you know, people say that page rank doesn't really matter for LLMs, and that's true.
But also you can adjust your training data so something that has more authority gets fed in multiple times. Like it's that simple to influence how, how an LLM, how a model is specifically trained.
So, and that's just one way. So yeah, it's, it's a, it's a crazy and interesting time right now, I
[00:09:39] Speaker A: think, from the data science perspective. I was asking a previous guest about, you know, how, how Alive is the PageRank algorithm and is there any indication that it's used in the LLMs? Oh, it was Malte.
I was asking him about the, what we know about, you know, how Google is currently baking in PageRank and I think you did a correlation study between PageRank and Backlinks. Can you clue me in and give me more information on what that study was and some of the data that came out of it?
[00:10:18] Speaker B: Yeah, so I don't remember the exact numbers, but yeah, so I did two analyses. One was, one came out in May, one came out in July, and the, and they both, there were both strong correlations between page rank and what was recommended by the LLMs. But the basic premise of that was so for the July one, it was across 100 different industries. And then I generated 10 different Personas for each industry.
So I think one of the things that, one of the fundamental marketing principles that SEOs really need to learn is that the Persona, the profile of who's actually typing in the prompt matters tremendously.
So I would pick an industry like airline industry, and then I would have like, you know, business traveler, weekend Traveler, whatever, like 10 different Personas for each of these industries.
So 1100 Personas across the board. And I ran, I generated LLM prompts, keywords, search results. I analyzed all of Reddit. So I've got an archive of Reddit, did correlation analysis between relevant keywords and Reddit posts, went through all the Wikidata and Wikipedia content to find Correlations between links and keywords and mentions and all of that stuff.
And essentially the result sort of, if you look at everything as a whole, which isn't quite fair because it really changes a lot from industry to industry and even from Persona to Persona.
But if you look at everything as a whole, the. There was coral. There was a pretty decent correlation between your backlink strength, your search ranking strength, and your likelihood to show up in the LLM results.
[00:12:01] Speaker A: Walt, that's great news.
I love a good correlation study. What's, you know, anytime that we look at correlation, though, what are your caveats for pegging that?
You know, anytime correlation is not causation, but it indicates a relationship. So you have to put the asterisk there. What in your mind are those asterisks?
[00:12:27] Speaker B: Basically exactly what you just said. And that's actually in the reports that I put together, like posted on LinkedIn and did like a document or whatever. There was a full page before I got to the data saying this correlation, not causality, this is what it means. Like, don't look at, at. Look, look at it as like, you know, like this is baked into any algorithm. This is just. We're just looking at patterns here.
So I think the way that I treat correlation analysis is, I mean, from a scientific perspective, you need anecdotal evidence before you start your experiment to like, figure out what's the experiment to start.
So at the very least, strong correlations or stronger correlations or at least a place to look to see how it's affecting your industry. And, and maybe you use that to define your strategy. So, you know, if, if I was in one of those industry, if I was like in the airlines industry and Target and I had a campaign targeting budget travelers and Page Rank was a key piece of that, then I would go after that. If Reddit was a stronger signal, I would prioritize that in my campaign. So it's. I basically treat it as. It's not perfect, but it's at least data. It's at least decent enough to make
[00:13:39] Speaker A: a decision on what's a rabbit hole that you've been led down recently that was nonetheless still an interesting adventure
[00:13:52] Speaker B: in terms of like the AI stuff.
[00:13:54] Speaker A: Yeah, yeah,
[00:13:58] Speaker B: man. What have I been.
Rabbit holes that turned out good or turned out like dead ends.
[00:14:08] Speaker A: Sometimes the journey is worth it, even if you end up running into a wall, as long as you figured something out along the way.
[00:14:17] Speaker B: Yeah, I would say.
There was a recent batch of research that I put together. I never actually released it. It's sort of On a to do list. I did this probably two or three weeks ago and I burned a lot of tokens. What I wanted to figure out was like, and this is across different mod, maybe it's all the chat GPT ones.
I wanted to see how recommendations changed for different levels of thinking.
So what's the correlation like? So if there's, if the thinking level that you're putting into the API is no or low thinking, it's not. And, and you turn off fan out queries, that's like the purest, that's the purest insight into how the LLM is actually built.
[00:15:05] Speaker A: Ah, okay. Yeah.
[00:15:07] Speaker B: And so, so if you take that as a baseline and then you're like, okay, let me turn up the thinking to the lowest level of thinking, like how does that change? And then the next level, and the next level, the next level. And then what happens if you do all of that again but you turn on fan out queries? So now you're bringing in web data.
So now the LLM isn't just doing its own thinking, now it's bringing in web data. Now how does that change?
And it's a huge difference. It's just a huge difference. Like I think that's, I would say what the rabbit hole is.
How do I say this?
The engineer part of my brain and the marketer part of my brain are kind of in constant conflict because the principles are totally different. The engineer is like we need to know exactly what's going on. And the marketer is we need to know what's going on. Good enough.
Good enough to explain it to the client.
That's really like what it comes down to. And on the engineering side it's like no, we need to understand it at all the levels here and there's.
And so I see things like the engineering side of my brain sees things in the marketing space like scraping chat GPT results from non logged in accounts and, and how a lot of the big companies use that for rank tracking.
And the engineer part of my brain is like that's just a bad representation of the LLM. And then the marketing part of my brain is like it's, it's representative.
Like it's a, it's, it's what people are actually getting. And then the business part of my brain is like scraping those results is way cheaper than the LLM costs. So it's like if I'm going to build a tool, let's, let's just scrape the results and if that works for people, great, we'll do that until it doesn't.
So yeah, there's this, like, I would say that's, that's sort of where my brain constantly swirls, is trying to understand how to get meaningful results out of the LLM, to figure out what to base my, my campaign progress against is
[00:17:18] Speaker A: that the, the eternal war of art versus science.
Because you know, you have art as the expression of experience, latent talent, your observations, you know, and intuition coming together. And then you have the hard, rigorous scientific process which you have to admit, if you are running a science, rigorous scientific process.
We, we have no clean data points or sources to truly work with. You know, both on SEO and then on the LLM side, they are by their nature obfuscated, their black box results.
We have what they tell us occurs behind the scenes, we have the patents that they've patented and we have people who have done similar things. But you, one truly cannot know behind the scenes. So you have to, as a scientist, you know, from a purely scientific methodology, there, there are unknowables. And in science you, I guess you can, you have to take so much salt that you could make a C out of it.
So it almost seems like on certain levels just accepting, you know, or being able to switch and say, hey, this is something where it needs interpretation. I can get as much of the data set as possible and learn this from it to feed that the interpretation and know, know what is knowable, but know that you know it's know what you don't know. Right, Socrates?
[00:19:09] Speaker B: Yeah, yeah. Figuring out what that line is. And then sometimes you know things and you don't realize that you actually know them until after you're like, hey, I've kind of been like doing a good job for like a year at this. Like maybe I did know it sooner than I realized.
And sometimes you think you know stuff and then you know, you're blindsided and you're like, oh, wow, I really didn't understand what was going on there.
[00:19:32] Speaker A: I, Is there another industry that's like this? I mean, I feel kind of like I'm taking crazy pills because.
[00:19:43] Speaker B: I can't think of one. I think, I mean, I genuinely can't. I can think of parallels in like, aspects of the industry, but as an industry that's so, I mean, like on one hand we're just chasing around search engines and LLMs.
I mean that's kind of what it comes down to. And on the other hand there's this balance like you're talking about between like the science and the art of it and the engineering of It.
And yeah, I really can't think of anything where there's so many different kinds of unknowns and external forces on it.
[00:20:26] Speaker A: Talk to me about a slop and the role that you think that it's playing in our information economy right now.
Is it just, you know, the predictable outcome of handling, handing the masses execution intelligence en masse without the wisdom to choose how to apply it, or is there something more insidious about human nature that just is going to exploit systems?
[00:21:00] Speaker B: Yeah, I think the answer to that depends on where you place the accountability and so on the, if you, if you place the accountability on the producer of the slop, you're basically saying, you know, what you need to do is you need to hold a different standard than you do have, and you should hold my standard.
And we have to raise standards for everyone because if we don't. Now there's a bunch of misinformation.
Like there are these downstream effects of taking quality out of things. Like there's.
And I don't think we can. In the same way that you're talking about art earlier. Like, I don't, I don't know that we can measure it, but we know that it's there. There's something like there's something important about maintaining quality.
But then I think if you put the accountability on the consumer, then the conversation is stop accepting it.
And the, the difficulty there is as long as people are willing to accept it, it's going to keep getting produced.
And so, like, I have a friend, she's got a couple kids, and you know, they'll be on their, they'll be on their iPads or whatever, and you can kind of hear some of the stuff they're, what they're watching. And it's like, it's just all AI slope.
As long as they're willing to consume it, it's going to keep happening. It'll, you know, it's like whatever space is there to be filled, people will find a way to produce it to fill that space. So I don't, I don't know the answer. I don't know if it's a cultural thing that we have to shift where we insist on higher quality. Like, how do we get everyone to agree on that?
Or, you know, I don't know if it makes sense to make it legislative or, you know, punishable. Like, I don't know what the, I don't know how you introduce consequences to the producers. I think it's a really difficult and fascinating conversation and I, I'm really curious to See what it's like in five years.
[00:23:08] Speaker A: The, it's, it's interesting because there's as many unlocks for.
You know, personally I know that my workflow, you know, I'm working on an SEO audit right now and you know, 10 years ago that would have looked like me pulling a thousand CSVs, copying and pasting these and then using Google Docs and Google Spreadsheets and making a Google Slide deck. And instead I'm building that audit result right into my, my SEO Arcade app for the first time and using it as a beta test study. I'm like, okay, put this here, here's a checklist.
Use Google, use the mcp, the Google Analytics mcp, you know, and Microsoft Clarity and triangulate and tell me from those three different sources. Okay, stick that in.
Oh, open SEO. Just added local SEO stuff. Let's pull that in. And that would have taken, that would have been impossible 10 years ago.
APIs. I just realized there's APIs everywhere.
And as a marketer, just stopping and looking at your stack and just taking an afternoon to ask, oh, hey, is there an API for Google Tag Manager? Yes, and it can help me do it. Turns out there is. And it can help me do the pain in the butt Google Tag Manager implementation that setting up your, your triggers manually to make sure they correctly fire events everywhere was the most tedious and painful of tasks to do correctly. And it was done in minutes. While I think me and my kid went for a swim.
[00:24:56] Speaker B: Yeah, totally.
Well, and I think that brings up like another point to the AI slop conversation is like, when is it slop?
Because if it wasn't for LLMs, like I'm not making that analysis of a hundred different industries and like 10 different market segments and like, I'm not doing that.
Like it's just not going to happen as quickly as it did. And I think on the engineering side. So I actually started an SEO in 2001 and grew with a company called Keyword Ranking. I don't know if you know Andy Beal and Garrett French and all them. Yeah, yeah. So I was one of the first people there and I managed all the writers, the account managers, the, the analysts. And I left like four years later.
I actually hired Garrett at Keyword Ranking, but also he and I worked together, started a company onto low doing like link building stuff and then we parted ways after a few years and he kept, he kept doing, he got into, on the sort of like the link building agency side. And I just wanted to build tools, I wanted to do the engineering stuff. So I did engineering stuff for the last, what, probably 13 years, 14 years, all from, all the way from writing like large scale crawlers and scrapers and pure C by hand. Like no LLMs, like pure circumstances, just obsessed with, with performance all the way to my last big project doing pure engineering. I was, I designed the embedded software for an ESP32 that went into some high end audio equipment and it managed all the like over the air updates to update it and other microcontrollers inside of it. Like, like actual engineering.
And so on the slop side the conversation is like.
And one of my best friends is that he was at Reddit for four years. He helped on, he designed their API stuff and it's like well, where's the line?
Where's the line between like is are the reports that I put together, is that considered AI slop?
Because I didn't sit there, I didn't write the code right.
And so there's this, there's, there's also this, this part on the engineering side which is like how do you use the tools without just accumulating more and more technical debt?
And it's difficult because it's, it's, it's really interesting. Like with Fable I can give it hard problems and those hard problems might be just this mess of code that Sonnet wrote three months ago and Fable will just take care of it it.
So now it's like, do I need to care about technical debt if the LLMs are getting so good to take care of it for. It's this weird, it's this weird like mind game about like where's the line? When is it like outside of my control? How much do I need to actually know about what's going on? And I don't know where that line is. It's, it's difficult.
[00:28:07] Speaker A: It is a good challenge. I'm curious about where that, that circle around back to link building came because I, I just saw your linked talking about your upcoming, I think upcoming beta with opalert.
So talk a little bit about that. Tell me what you're aiming to do in applying what you've learned in the engineering side to the link space as well as how you're fitting in that LLM technology or view of it.
[00:28:40] Speaker B: Yeah, so I got into engineering, always wanting to build marketing tools. Like I genuinely love marketing.
I also genuinely love just getting a bunch of data, finding the patterns in it, making sense of it and just making it usable. So like one of the things I released, I don't know if this was last week or the week before. But news.opalerts.com is a news aggregator, so it pulls from like a few million RSS feeds and podcast feeds and then over a hundred thousand news sources. And there's like, it's just like a marketing dashboard for AI search, visibility, link building, PR and content marketing. And it auto updates every 15 minutes. So that's like, I've always wanted that.
So I always knew when I got into engineering the long term, like I wanted to build marketing tools. I just really enjoy it. I've always enjoyed it and just sort of ease back into it. Over the last three years and seeing the industry shift and evolve, I've sort of been trying to figure out, like, where do I fit in? Like, what can I build that would actually be useful for people? And, and what does that look like?
And yeah, so next week I'm releasing op alerts and that is, I mean, it just, it, it takes so many different pieces. There's obviously like the rank tracking, whether it's, you know, organic search results or scraping ChatGPT or whatever else. So you can see that over time. That's. I feel like that's sort of like table stakes for an AI visibility tool.
But there's a lot of other stuff that I've incorporated into it. Such as, like Common Crawl has their own page rank data.
So I've been chipping away at building a database where you can see your historical page rank over time. You can see it for yourself, you could see it for other, for other sites and you can kind of get a sense of like, how much trust should you put in into this particular site. So there's that. I've built a database of contact information just going through, I mean, just tons and tons of web pages. So I've got like contact pages, social media stuff, email accounts, phone numbers, anything I can find on. I don't know if it's 10 to 20 million sites now. So there's that piece, but I would say that the, that there's. There are two other parts that are bigger. One is you. You, you set up a campaign which is just like, this is the industry that I'm trying to target. And you set up three Personas.
Just give me like a couple paragraphs on each. I can take that.
There's a whole bunch of stuff that it does behind the scenes. And then I come back to you with like, who's leading in the, in the industry?
Like, where are your competitors ranked for those specific Personas in the market? But also going so far as like trying to identify link opportunities, guest post opportunities, link directories, what are the conferences like, what are the awards like, actually looking and getting a full view of what the actual market is.
And then there's other stuff that I can take from there, including like keywords and you know, a lot of obvious stuff. But then all of that goes into that news search index.
So one of the things I posted today is you can do news specific link gap analysis. So you could put in 10. When you define your campaign, you define your competitors and you define the sites that you own. So now behind the scenes, it's constantly finding stuff where you're competing, competitors are linked, but you're not.
And that's just constantly coming in. It's also constantly finding guest post opportunities or you know, if somebody does a review of some other AI search visibility tool, I'll be able to see that. And then now I could reach out to them to see if they want an account and do a review.
So all of the different ways, all the different reasons that you would monitor stuff, find opportunities, all of that.
And then third piece that I forgot about is I built a, there's a, so optimizing content is just different now with text embeddings of vectors and all that. So I actually built, this took like real work. I built, I aggregated 10 different taxonomies like the Google Ads taxonomy or the Internet, the International Advertising Bureau content types, things like that.
And I did a ton of research to figure out how to build vectors.
So when you're writing that represent each of those topics. So you're writing a piece of content, say you get a client in cybersecurity industry, you put that, you put a piece of content in there and you can check against over 20,000 different taxonomy categories to see if it's the most relevant to that category or not. And if not, it'll give you additional terms and form phrases so that your piece of content actually aligns the most with the, with the taxonomy categories you want it to. And it's all semantic. It's not, it's not like the typical keyword database type stuff.
[00:33:48] Speaker A: That seems a lot more effective than pinging. What was Watson, IBM's Watson website and was a very fundamental version that we built into Raven tools in 2012. So it seems like a much more matured version and a little more detailed approach to that.
While you were exploring vectors, what was some of the, the ways that you, what were some of the ways that you think Google is using that type of viewpoint that way of looking at information and how does that inform SEO in general?
[00:34:33] Speaker B: So when I built the news.app alerts.com, i had to figure out like, how am I going to get just the AI search stuff? How am I going to get just the SEO stuff, just the link building stuff.
So I have vectors for each of those industries to bring that up. But that's just one ranking factor for what it takes for something to show up in on the news in that site. Because it might have 5,000 articles from the last two days, it's got to figure out which ones actually make sense.
So that similarity score is just one ranking factor.
There's also about a hundred other phrases and terms that get weighted into that alongside the similarity score to figure out which pages are actually the most relevant. Because just, just doing vector search is pretty muddy.
It's. It's good to get like a sense of it, but it's still pretty muddy.
But it's really great at filtering out stuff that's not relevant and it's really great at finding stuff that the keywords would, that pure keyword search would miss. Even with a hundred different phrases that I'm searching, the similarity search using vectors brings out stuff that the keyword searches are just going to miss. Like there's not enough keywords to catch everything. And, and so the vectors are phenomenal at that I think the way where that's being used. So I don't. Do you remember when Yandex, when the source code from Yandex was leaked?
[00:35:57] Speaker A: Yeah, I do.
[00:35:59] Speaker B: Yeah. So Mike King, Russ Jones and I, we got on a Slack channel. I found out about. It's like three in the morning and I posted a private friends only message on Facebook. I was like, hey, if you had access to like this, what would you do? And Mike sent me a message like immediately. And so he and I and Mike and Rust. So we're in the Slack channel for like a week just tearing the code apart trying to figure out what's going on.
The, the category vectors that were part of even just Yandex. This is just three years ago. So Cat Boost I think was a similar.
I think Cat Boost and Bert were both strong, like categorical taxonomies and I think they were showing up in lots of different way. So lots of different vectors per search query and per document. It wasn't just like a single cat boost, which is a categorical boost. It wasn't just a single one. It was lots of them all at once.
[00:36:55] Speaker A: Yeah.
[00:36:56] Speaker B: So it's, it's something that really augments, I think pure keyword search and also at the same time, like it just can't replace it. It's just not gonna, it's just not gonna get there.
[00:37:12] Speaker A: That's fair enough. As we kind of wind down, are there any other studies or workflows or releases of information that you want to point the listeners to, to kind of study up, do some homework, get, get a final bite at the apple?
[00:37:31] Speaker B: What would I suggest to people?
Honestly? I would suggest, I would suggest people just build a search engine.
Honestly, it's the best way to get a feel for it. Just open up Claude code, tell it to build a Manticore, a Docker image using Manticore.
Tell it to have the title and the body.
So it's going to have title, text, body, text.
Go scrape a thousand web pages, 10,000 web page. Just put the title and the body in there and then figure out how to design searches to get the most relevant documents. You'll learn more about search in that hour and a half than anything else.
[00:38:15] Speaker A: Brilliant. I'll add that as an SOP to my site.
Learn about SEO by building your own search engine.
You know, certainly, you know, seeing how Yandex was leaked, I think there was one other, if I'm not mistaken, leak from or. There's been a couple of leaks of search engines in the past, but I
[00:38:37] Speaker B: think, I think a few months after, six or 12 months after that, the Google ranking factors were, were leaked and that was like, I think that was like a XML or something document that slipped through, like a GitHub repo. And yeah, all the SEO's jumped on it. I've got a copy of it somewhere.
[00:38:58] Speaker A: Well, I love the conversation. Give a final, final shout out for the brand quick coming up for you guys.
I think you had that beta coming up. I'll make sure that gets linked in the show Notes. If anybody wants more information about Ben, I'm going to put this all up on unscripted SEO.com, hand him an article for his site and his audience.
So here's your last but your last chance, Ben. Talk to the people.
[00:39:26] Speaker B: Yeah, yeah, it's just opalerts.com and if nothing else, check out the news.operts.com if you're somebody who checks out like Hacker News two or three times a day or ESPN or whatever it is, is the news site is a good way to just cut your brain off for 30 seconds, see what's going on and get back to whatever else you were doing before.
[00:39:44] Speaker A: Fantastic. Thanks for stopping by.
[00:39:46] Speaker B: All right. Thanks, Jeremy. I appreciate it.
