The Truth About SEO Rankings in the AI Age with Mike Roberts, Founder of SpyFu | Ep. 396

0:00Today's episode is brought to you by customer.io. You have customer data, but turning it into personalized marketing at scale? That's the hard part. Customer.io connects the dots. Feed it your data, start sending messages that actually feel personal. Learn more at customer.io slash TMM. Welcome to the Marketing Millennials, the no BS marketing podcast. I'm Daniel Murray and join me for unfiltered conversations with the brains behind marketing's coolest companies. The one request I tell our guests, stories or it didn't happen. Get ready to turn the f*** up. We are back with another episode of the Marketing Millennials podcast.

0:53I am here with Mike Roberts, founder of SpyFu and founder and CEO of Rival Flow AI. He's a good person to talk about, about this craziness that's happening in marketing right now, especially how people can show up in these models. And I think that's like the new wave of SEO, even though I hate when people say SEO is dead because some of the same principles still exist that you need for SEO. But we could talk more about that. But I want to go into your story. And SpyFu basically built the see what competitors do in search category. So what did you see when you started that everybody else was missing?

1:37Yeah. So before there was SpyFu, there was no way to like just type in a website and see all the keywords they rank on or all the ads that they buy. And I actually had another software company prior to SpyFu. And I was, you know, like this is a straight up bootstrap situation, man. And I had like$3 ,000 in my bank account that if I didn't make the right bets on ads, I'd have to go get another job, right? I'd have to go back and be a consultant, which I did occasionally, you know, during my sort of startup timeframe. But I didn't want to. So I wanted to be as right as I could possibly be. And I had recently learned that I was using some kind of like cursive knowledge style keywords.

2:21Like I described my product as web data extraction and everybody in the entire world calls that web scraping, web screen scraping. That's what I, that was, that was what my product was. And I was just kind of missing the mark because I was too much of a dork, too much of a nerd for my own, for my own technology. And, and so once I, once I solved that problem, once I've made that switch, I was like, oh my God, I wonder if there's other things that I don't know that my competitors do. and so I needed to build this thing and the only way to do it at the time was to just type in a search and see if your competitor shows up so what I basically did was I just brute forced it right I just ran all of the searches and then made like a reverse you know like reverse search reverse phone directory if you will for for Google so that was like kind of like I was solving my own problem but then I was like well you know how about I also release this for the world and and made it made it into like this rapid prototype that I called Goog Spy.

3:22And, um, and then, uh, and then, you know, at that time, you know, I just released it, send it out. And, um, and, and part of that release process was a press release. We got picked up, people kept coming back. And that's ultimately how the category was, was, was created. And then we launched SpyFu a few, like about a year later. Um, and then everybody else, all the competitors that you know of, uh, SEMrush, et et cetera, essentially were copying us for their own country or something like that. Like that was like, that was the origin story of SEMrush, for example. It was SpyFu, but for Russia.

3:55Oh, I actually didn't know that. That's a... Yeah, yeah. Now you know the rush, where the rush comes from. Yeah, exactly. Okay. I was like, why are they called SEMrush? Yeah, totally. Yeah. It's like a play on the Russian part. I want to go into what the industry was before this AI hit and then now what's happening to the world of surge, PPC, AI surge, what's happening right now? So in a broad stroke, right? Like Google has been, their intent is to deliver, their like sort of 25, 50 year mission or whatever is to organize the world's content and make it so that you can get anything that you search for or any question that you ask, they'll deliver you to that.

4:45And broadly speaking, I don't think that the long arc of their goal has changed much. What they want you to do is come to Google and rely on Google to ask it your questions and to find what you're looking to find. Over the last 10 years, let's say, lots of things have migrated from being like something that you click on and go to a different website to being zero click things. Think of the very beginning where you've got like, you know, like a definition of a word just sort of pops up or like what's the weather is actually literally in the auto suggest or the search. It doesn't even show up in the search results.

5:24It actually just shows up in your in the autocomplete. Right. So they're doing like zero click and zero search things. So obviously, like now you got AI overviews and like AI answering a lot of the questions, but they had knowledge graph before. So there is a huge change, but also it's helpful to understand what their overall inertia is, what's their overall direction, because then you can kind of understand where is this all going? You know, how catastrophic is it? because it's a bit of a, you know, like we've been impacted, SpyFu in particular. We got a lot of branded traffic that like we used to rely on in terms of clicks coming to our site that ultimately now people are asking questions about SpyFu and they get the answer without coming to SpyFu.

6:17Fantastic and also terrifying. And that's kind of like the main shift is that answers are getting like meaningful questions and answers specifically about your brand, which is what most people I think are like really nervous about, get answered outside the context of your brand. Which is crazy to think about, which is the zero click zero. Because that goes to the question before you can think about adding keywords. And if I search this keyword, something, or if I did a long tail versus a short tail, something will come up, which is a little bit of, it's still, I mean, it's not going that hasn't gone away yet but what are you seeing like when people ask you okay what how should i prepare for this change what what should i be doing differently than i was doing before what are you telling that like that cmo or that um seo marketer or um people who are trying to stay up with making sure they get shown in these different uh ai overviews or LLMs or stuff like that.

7:27So it's like this interesting merge of, of like where SEO becomes a bit more synergistic across all of the, across your traditional, like sort of branded and PR stuff. Right. So that's like, that's like sort of the outbound area, right? You want to make sure that you're getting brand mentions in, in, in, in the areas like that are outside of your site that you can, that you can control or that you can influence. And that's kind of like, oh, I mean, if we wanted to kind of make it simplify it, it'd be like backlink outreach, but for brands. But it's more like traditional PR is probably a good choice to invest in more.

8:07But also think of that from the perspective of SEO. On-site is something that we're actually pretty engaged in, like what happens when you're on the site. And so we have a product called Rival Flow. And Rival Flow basically says, like, what questions are your competitors answering that are outranking you in both the SERP and the AI overviews? And we say, what questions are they answering that you're not? And then we just fill in those gaps with AI answers, right? So we just generate content to perfectly match your voice and just insert it into your page right where it makes sense contextually.

8:46and uh and that way you know you're covering all of those questions that ai is answering um more effectively than your competitors and when you say what's the way to make it more effective i know the tools like doing it but what what is making it like more effective to rank in an overview like what are the what are what is the google looking at what is these other um ai um bots looking at you know like so google's been using uh using ai as sort of a um in their SERP ranking process so what they do is they crawl things uh using a traditional search index um which we would just sort of like simplify to like tf idf like if you're familiar with tf idf uh that's like kind of the old or that's like kind of like the most it's not necessarily old it's like the the base level of how, how like a, a search index works.

9:41It looks at how frequently a term shows up in the, um, in, in a document versus how frequently it shows up in all documents. Right. So it's like this, you know, when you, when you think of like keyword stuffing and what have you, that's basically a hack for TF-IDF. And to a large extent, certain search engines still function primarily with TF-IDF like indexes. But there's like, if you think of that as like the traditional document retrieval search index, that's like what Google does first. And then they may return, like, let's say the top 1000 documents, and then they re-rank them with AI. And they've been doing that since, well, for a while, at least since BERT, right?

10:27Like we all kind of like heard about BERT, BERT 2 or whatever, and that impacting the rankings. It's likely, and so by the way, the way that AI works is a completely different process, right? So it's still called tokenization, et cetera, but it doesn't actually function quite the same way. It's a lot more semantic. it's a lot more expensive, wildly more expensive to do the, to, to store the entire internet in, in a vector index. But the whole world is probably going roughly in that direction. I think that most things, well, actually I know that most things in particular, Google is still a hybrid.

11:13So they're still doing like retrieval first and then re-ranking things second with, sort of the LLM. So the thing that you can easily impact is where you show up in that first search. And that's largely via traditional SEO, right? Then the second thing is, the second thing becomes quite a bit more random because fundamentally, the way that LLMs work, including Google's, right is is there's a there's a thing called temperature right and temperature controls what is um uh randomness essentially because if you think of like if you think of like llms as being like um spicy autocomplete or whatever right like it's basically doing the same thing you type in like what is my favorite and then it's going to complete with the most probable next term, right?

12:11That's what you would expect your search and the autocomplete thing to do. What the LLM does is it says, what are the top five most probable things? And then let's just randomly choose one, right? That's effectively what temperature does. And the temperature says, well, am I going to pick the top three things or the top five things or the top 100 things? make sense and that happens for every every uh every token as you go through sort of the generation process so the output that you get from that second stage is inherently going to change from instance to instance from time to time it's always going to be somewhat random it's by design it's where creativity comes from it's where like a lot of the good all of a lot of the good comes a lot of the good of AI comes from that randomness, but it does mean that the outcomes are like not deterministic, whereas the outcomes of a search index are much more deterministic.

13:17So yeah, I mean, sorry, that's a bit of a rant, a ramble or whatever, like a lot of context there, if you will, but there's definitely a lot of SEO that you can still do to impact the results. Have you told like an SEO person what to prioritize right now? And like, what is like the one or two things that they should be thinking of? Like, I need to do this and this and this to make sure, okay, I'm going to get in the search results. That's cool. And then, you know, you don't get the perfect chance, but I get the best chance that I'm in that cool randomness part of the AI. I would personally focus on the things that you can truly control.

13:55And those things are the things that are on your site. And those things are about answering the questions that are being asked and specifically closing the gaps between anybody that's beating you and making sure that you're answering more questions than they are and more directly. like there's a lot of things that are in flux and there's going to be a lot of people that are going to tell you that what you should focus on is um getting a lot of mentions in reddit or assumptions um some some something right like the this is like the type of thing if you've been in this industry for like 20 years there's patterns right and this is there's a bit of a hype cycle to certain things where it's not going to last it's not going to work for a very long time so the hacks that you hear about that feel like they're hacks by the time you hear about it it may be too late right i always say that if you go like even like with memes and stuff like there's like a trend where it's like it's in the deep part of like the reddit subreddit and then it starts getting to x and then by the time it's on like linkedin and instagram you're like three weeks late to that that meme trip yeah that happens yeah psa parents don't use skibbity anymore.

15:08It's totally dead. Right. Like it's there's trendiness to to these things. And ultimately the signal becomes lower. Right. So like Reddit was a Reddit, what Reddit at some point and maybe still is a very strong signal for Google. But as it sort of gets overused, it becomes less of a strong signal. I don't think that they're necessarily trying to combat anything like realistically. The machine learning aspect of the stuff is going to just figure out, well, it's not as strong of a signal anymore. And a lot of those things will, you know, it's an interesting challenge for them actually to organize the information when there is quite a bit of, let's say there's actually quite a bit of like ChatGPT 3.5 content that's been generated in the world.

15:55Stuff that's generated now is pretty good, like, right, relative to, you know, a year and a half ago. Yeah, like the models now, from what they used to put out, I barely can tell. Sometimes I can't tell the difference between a human versus AI writing. Not at all. I mean, it's like the little signs if you're not feeding it probably the right things that you could spot the BS, but the tone and the voice and the output of what they're searching is so good now. But the old catchy BG was terrible when it was Friday. But I wanted to go into, so what are some pages that you see a lot of people doing wrong?

16:41I know they're answering questions on the site, but where are the gaps on the site that you usually see that people are not paying enough attention to? There probably are some commonalities at these sites that you see that are not ranking because they don't have a frequently asked question section or they don't have enough blog posts going out or they have not enough. What are the things that are you seeing that are the problems on their site? I don't know that there's a pattern to the type of questions that don't get answered. But what I will say that there's a really strong pattern for that is such a strong pattern that we added this to the product, right?

17:21Like we have a lot of customers that are agencies that have a bunch of customers that are like local businesses and local service landing pages, local service pages where they're talking about their product or service. Oftentimes they'll talk pretty intelligently and completely about their product. and we don't necessarily want to just like inject a bunch of like random questions and answers or you know like i i call these things questions and answers but really it's content that answers the question right like it doesn't necessarily need to be like an interrogative like you don't have to do faq in fact we specifically generally don't we'll just put them in sections and answer the question directly but not necessarily be like this is a question and this is an answer because realistically you may not actually have that question right like it only really works if you actually have the question but having a topic always works anyway so on these pages the broad observation is that they lack a lot of like just typical like um i would call it um like local seo cro right there's like not like a um there's not like mentions of of like the local area or like like landmarks um there's not like mentions of like um you know like weather patterns etc it's just relatively like it's content that's not that's not clearly localized and it's also there's also often no cta above the fold stuff like that like that's kind of a pretty common unfortunately really common thing on the internet for people that are actually trying to drive leads they ultimately want to like make people call their phone number and they don't make the cta above the fold and it's like okay well we'll solve this problem for like the world and all of our clients and all our clients clients you know like uh and that makes actually a really big impact because if you are getting fewer clicks to your site the clicks that you do get if you can just make those into leads more you know if you can improve your conversion rate then that's like that's the big pattern right they're just missing the um the core elements the sort of few bullet points uh the following the sort of best practices of actually just converting a visitor into a lead it's actually crazy i mean like you think some of the biggest companies you think you're probably listening and this is probably the most simple advice like hey you have your button above the flow like add certain things of your area and your content that but for some there's so many businesses out there that still are just saying basic stuff on their website there's no the ct is not above the flow They're not doing little things that could just boost them up a little bit more higher in the ranking because they just don't know.

20:05They're just not marketers. Like they're a plumber running a plumbing business or a person. Like I just got someone to help clean like air duct like for the dryer. And I had to like find the person. And just I picked whoever was ranking. I went to top five ranking people. But I bet you I miss so many of those air duct cleaners because they didn't put enough information on their site to get me there, which is sad. Yeah, I mean, it's not it's relatively subtle, but Google definitely uses those signals to rank those sites. You know, it's one of those things that doesn't exist in like, you know, your typical like site audit tool or whatever.

20:50Like Screaming Frog isn't going to detect that you're not that you don't have like localized terms and stuff like that. So it's something that we have in Rivalflow. It'll detect that and improve that. We'll wind up adding that into all of the SpyFu products at some point. Could you give a little one to two minute overview of, okay, what I should do if I use SpyFu or Rivalflow and how it's going to improve my marketing? so people know that basic of how it's going to help them. Sure. So actually, as it turns out, SpyFu, when you buy a SpyFu plan, you also get Rival Flow. Not all plans on the basic plan, but the next level up includes a whole bunch of AI tools, which we're fully integrated into ChatGPT, so you can just talk to ChatGPT.

21:51and you can say, help me build a marketing plan or anything you want. It'll pull in SpyFu data. The problem with, actually, this is a shirt that I'm wearing, I can see. The problem with SpyFu, or sorry, with ChatGPT is that it'll totally make up keyword data. It does a terrible job of that. So we pull that stuff into ChatGPT automatically. So I'll give you kind of like the high level of what SpyFu is and how Rivalflow differentiates and like where that all fits together. So you can go to SpyFu and type in any website and see all the keywords that they rank on on Google. All the ads, all the ad copy, every ad variation they've ever run.

22:35We're absolutely the best in the world at the PPC side. And we also have the most data on the SEO side. We just added like a thousand times more data, 800 times more data on the PPC side. So it's a pretty big, pretty big gain. That's the most that there's ever been in the history of our company. But also that would be all companies. Nobody has ever had more PPC data. So it's a pretty cool time. In addition, we track like your website and like make recommendations on your like how you can improve your site's rankings. And Rival Flow factors into that quite a bit. Rival Flow is going to look at your site, look at every page on your site and say, what's the most obvious opportunity for you to improve your rankings?

23:22And this product absolutely takes everything that I've ever learned about SEO and distills it into this process that realistically never loses. It always, it's weird because anybody that ever tells you that I have an SEO process that always wins is usually going to scam you for some, somehow, or like they don't really see the future. It's kind of like this whole Reddit thing or whatever, right? But in this case, all we're doing is improving your website. All we're doing is improving, making your page answer more questions that you're, you know, that customers are looking for. Right. Or alternatively, in some cases, we're improving, you know, like your sort of conversion rates.

24:13You're sitting on a gold mine of customer data and it's doing nothing for you. That's not a data problem. That's a tools problem. Customer.io connects your data to every channel you already use. SMS, email, push, in-app. Then it uses AI to build workflows that used to take weeks. Every message feels like it's written for that customer, not blasted at them. 8 ,000 plus brands already made the switch. Learn more at customer.io slash TMM. If we don't. And so what happens is that we make the recommendation. We say, what questions are your competitors that are outranking you? What questions are they answering that you're not?

24:56Which questions are they answering more thoroughly? And then we just sort of use AI to fill in those gaps. We give you exactly the changes that you need to make to your page. And then as soon as you hit publish, we track the before and after in your Google Search Console. We pull in your Google Search Console data. So I have this Google search console for data for like 20 ,000, 20 ,000 customers. And so I can tell you like that nobody drops like, and when they do drop, I go in and, you know, cause occasionally something will happen when they do, I go in and I look at that exact page. Why did, why did this page drop?

25:32Oftentimes it'll be because it's like 404ing now for some reason. So it's like not something we did. I did some videos where I like just completely YOLO into like random customer accounts that, um, that I pick with a random number generator and, uh, and do a live, um, you know, for like two hours straight. And we just go in and blur out their details, but show you their charts. So like in there, there'll be like these, these, these ones that I land on. I'm like, Oh no, what's going on here. And then I go in and I figure it out and I do that all live. and so the reality is is that this product it just improves your your ranks it just always works it just it's a very very good tool spyfu is all about research and improving and kind of like it's not like a super clear process end to end but rival flow happens to be a really clear process but again it's included it's part of spyfu and i highly recommend it also that legion is like one of the best legions of hey type in your domain find get a little bit of data of it and then oh yeah that's the one yeah the best like legion things to do is you get a little taste and then and it's also cool to see because i i did mine earlier uh just for like the free stuff um i have an account too but i want to see the free but it's cool to see that a lot of the new keywords I was ranking for were articles I was writing about um marketing which is super cool and it was posted on our site so the the natural things where you just what people are telling you to stop doing things that are have been working for a long time is is not something I mean you could be better write better articles by not only putting junk in them but it's cool to see that the the old tactics of still the writing articles, a blog post and stuff of that and ranking for keywords are still like thriving, not, not dying.

27:31Like everybody you see have to telling you. In fact, it's an interesting thing that you bring up because showing up on page four is actually more valuable now than it used to be. Right. Like it used to be that if you're on page four, you're absolutely buried but the ai will find that like so ai will find your page for content and surface it into the ai overview um and it'll and like chat gpt will find it it's uh it'll find kind of like more content than uh and surface it up into the ai overview than then yeah i mean because if you write a good piece of content that is going to help like you said answer a question like for example i wrote an article about like jordan shoes i know it doesn't have to do anything about like his marketing and then but like if someone's searching like how did jordan's brand get so popular like my one of my articles like popping up in that yeah which is cool to see that and that's it's probably not the top article but since it's like written a lot of people have seen it it's it's popping up into um like search uh overviews and stuff like that so um yeah use like a deep research model or whatever to aid in that, you know, or just like build it or just use chat GPT pro or like whatever.

28:48Right. It's helping Google out. Right. Cause Google, Google doesn't want to run. You can't, it can't run like super expensive high compute processes on every single topic. Right. So if you do, if you use your chat GPT credits or whatever, right. To like do some, like make something really deep for a relatively niche topic and you publish it. It's kind of in everyone's best interest. Yeah. I love like using like chat GPT or like Gemini or perplexity and finding like articles that I can reference and then like tagging those articles into my, my paper. And it makes the article 10 X better because you reference like cool stuff.

29:32And it's, it's so much easier to find these articles that took so hard to look for that's why i love ai because oh yeah the good card good articles are rising to the top which is so much more refreshing where the ones before were the people who played the game the best and stuffed their page the best one a long time ago now it's like actually these deep articles about like i was doing more research on like aeo and i like some of the like in-depth articles that i popped up and i found youtube videos and that would have taken me hours and hours before because i would have to go through the what is aeo page and the hubspot's this page and this random company's this page and until i got to an article written by someone who understands the topic deeply that can tell me what's what's happening so that's why ai is so great and then if you can add it to your articles too so yeah so so rival flow so the thing is is that i had like this idea that we're probably going to create a lot more content right like that's just like the way it's going to be we'll make a lot more content so people are going to have like hundreds of pages thousands of pages potentially so a lot of opportunity exists in that existing content like the best most effective way to improve your traffic is to improve your existing content so rival the rival flow process only focuses on existing stuff right all it's going to do is juice your existing stuff get you to two to three x more clicks on your stuff that already ranks you know so bill so these two things work hand in hand you just like use dude i was like i was like in the car yesterday and i just like had like i had an idea and i'm like okay can you do the research on this real quick i just speak it in do the research by the time i get there all the research is done i'm like yes yeah time to do i love that feeling too i also do it sometimes just like getting an agent to like research web pages for me and then i like when i'm home i'm like okay i don't have to think about i didn't have to do all this research as the agent just went to a thousand like not i'm just exaggerating but they went to like they went through a thousand pages at least to find good answers for me that I can, it's magical now that, especially if somebody writes newsletters and does podcasts, even like podcast research, I can find in the archives everything people say.

32:06Oh, yeah. Yeah, totally. I hadn't thought about that. That's pretty cool. Yeah. Yeah. So I can find like quotes that you probably said like five years ago. Yeah. Yeah. I love doing that. Um, so I, one question I have to ask everybody in this podcast is what is a marketing hill you would die on? A marketing hill I would die on. Okay. I would die on the marketing hill that, okay. So I'll, I'll give you something that like kind of conflicts with like whatever, what you might like, um, believe, uh, I would say, cause I'm like pretty into data, right? I would die on the marketing hill that, um, that data is going to lead you in a really good direction, but sometimes it can like lead you to the absolutely wrong direction if you look at it after the fact.

32:53So this is kind of like a big deal, right? Everybody tends to look at the data after the fact and then come up with a story after the fact. But you have to figure out how to like separate the thing where you can come up with a story after the fact, because you can like at SpyFu, we do this interview process. And one of the first questions we ask is basically we generate like a hypothetical chart with like two different axes that we randomly make up. And we have the person explain it to us, which is a strength when they can explain this chart to us that has no meaning. But it's also a weird weakness in all of us where we can look at data and explain it after the fact and somehow make it make sense, even though it's two randomly generated axes with a randomly generated chart, right?

33:43So the hill that I'm willing to die on here is that you have to come up with a hypothesis before you look at the data. I always say that marketing is doing like the scientific method over and over and over and over. Because if you're not, you don't know what you're trying to prove at the end of the day. You're just saying, okay, this is the data. But what are you trying to prove from this data? That something happened? Is this a needle you're trying to move here? Or are you just presenting? Because you're right. there are a lot of people out there who are really good at taking a piece of data and telling a pretty story around that piece of data but it's harder to like when you were doing that analysis what was like and why was that even analysis done in the first place to get to that output super smart i think that i've never heard anybody say that but it's a great great um especially from a data nerd like it's a great yeah to die on because um data is an intro and it's also as good as the inputs you put in and you can make tell any story with and marketers are the best at telling stories like there's a sample of people and you say like this is the end all be all but you don't know who that hundred people are like you don't know it's just you could tell any story with data yeah people talk a lot about vanity metrics and those are a little bit easier to detect after the fact.

35:13You can kind of be like, yeah, well, actually, how do we connect this? But there's the whole story, the cause and effect and what's actually going on here. It's tough because sometimes you do have the data and you only have, like, it's a really tricky process that you have to be skeptical and aware of your own ability to detect patterns, right? And frankly, AI is going to have that same exact problem. The cool thing about the weird thing about ai is that it does not behave like computers we nerds spend our whole lives learning how to think or just i guess basically if you're a nerd if you're if you're like a if you're like a if you're like a developer you're probably born thinking in like loops and if statements but ultimately if you think of it like you're also just like learning to think and synergize how a computer works in these sort of boolean ways ai does not ai makes the same like type of mistakes we make so if you apply ai and give it data if you give it data and say tell me what this data means it'll make the same kind of like i'll make up a fucking story about this i saw i saw something that kind of was like this i saw someone just did like a simple test with a bunch of models that said like i need a car wash the car wash is 50 yards away should i walk or take a take my car and like two models say duh walk it's so close like and it's like no the logical thing is like to get in your car because you need a car wash even if it's 50 meters away but just so funny to see how how the how like that and it would tell the whole story like why it's a good reason to um to walk but it's not like telling like the the logic like like the logical hypothesis would be like, what is the best way to get to like wash my car at the car wash?

37:06Um, and it's not, not take my car. So it's like, here's the, like the, here's like the one hack that I can like give you for using AI. It's I'm sure you may know this, right. But like, I think everybody uses like the, the, the sort of like, let's say six months ago hack, the six month ago hack is like, well, don't just type to your don't type to your chat GPT or whatever like talk to it you know speak for a very long time give it plenty of context describe your problem into a microphone and by doing that you'll give it more context and so on right like speak to it it works really well okay here's the here's the like what what are we at like February 2026 hack that that I just made up okay so rather than asking chat GPT to do something for you or or whatever cloud you're using uh say your whole thing, blah, blah, blah, blah, blah, blah, blah, blah, blah.

37:56And I want to make this really cool thing, yada, yada, yada, yada. And then say, and then don't say, do it, say, I don't want you to do it. I want you to give me a step-by-step process to get that done. And when you, when it does that, then you say, okay, now do step one or whatever, or do the first three steps, something like this, right? But you get it to, to give you those steps. The cool thing is then you can debug those steps before anything happens, before you evaluate the output. And you can also kind of like take those steps if you like them and be like give me a prompt that will create these steps or give me a prompt chain that will do these steps so that's kind of oh no i i love that i used to like when i was doing like deep helping me like write deeper articles i used to like make chat gbt write to me like in sections so i could see like yes otherwise because it gets in more depth if you ask or like Claude or whatever to more depth if you say okay section one I want to write about this here's my whole thing could you write and find and then instead of doing this whole article for you like you do each seconds then you can look through and say okay this is wrong this is wrong because if it's wrong in the first like you said wrong in the first section it's probably going to carry off that section to make the next section in the next section yeah in the next section so i like that i think so for people who are going in chat gbt still do what you're doing but make sure you are queuing each step so you know where how chat tb got chat tbd or any of these models got to that spot um or decided to write that whole um and correct yeah cool i mean like here's that sort of challenge point right like we're all going to become two to three X more efficient and more productive eventually with, with AI.

39:48The thing is, is that, uh, the challenge is, can you become 10 or a hundred or a thousand or a hundred thousand times more efficient? And some people, everybody's going to be two times more efficient. The question is, can you get to 50 X? Can you get to a hundred X? Can you get to a hundred thousand X? It is possible most likely um and uh and so that's the interesting dichotomy is that we're all going to get better some of us will get a hundred thousand times more production yeah i mean that's a weird it's actually kind of interesting because if you go back and just time like they're like the people who started filing things in computers versus filing it in file cabinets was like the first like like the old school manual thing and then they went digital was like the first step of like the people who were doing the old way started being like less efficient now it's just yeah even with like rocket fuel or even more efficient but you you see these leaps over time where like the people who took the efficiency because filing on the computer was probably like 10x more productive than than that person who had to like go and put the file in and go look for the file after and pick up the file and couldn't search the file and like maybe lost the file because it wasn't stored and then that person had to come in and do it it was like those things where you like now if you take the next step and take what you're doing because tools have software tools helped a lot of people over time and now like your tool helped a lot of like seo people not have to like do guesswork and like write random articles and like actually answer questions that people were saying but if they didn't use your tool they would have been how many acts like less productive as an professional um oh dude i love your example because like you got the file cabinets and then the file and then the file system right right so like it's like yeah like you went from like you went from like a back office to a digital back office and like your whole workforce like had like got twice as effective meanwhile somebody made like a search engine and organized the world documents and then yeah exactly then the next thing you know what i mean it's like now you can put all your documents and and ai and and and like find exact quotes from them where it would just and like spit out which like if you did that with your filing system it would take so long to find like your your it would take a long time to find exact or like partial quotes that you're trying to pull out for an article that you're trying to do like you could probably find maybe a quote but maybe you didn't remember that quote so now like what is like the next step of that and the next step of that and the next step of that where you're like becoming that it's just crazy how the world is just like and the people who are trying these new things um or doing it efficiently like you said the way you debugging um chat gbt is probably like going to give someone you a more efficient answer than someone who's not doing that and that made you more efficient than that person who who didn't so there's like these little efficiency gains that you need to figure out to be more efficient um that was a lot of efficient in one sense yeah like the challenge though is to be the be the one that makes the search engine i mean we're all going to be we're all going to start using digital file systems but what what what you want to be if you're the listener is to uh is to be the one that makes makes google essentially um or or something on that range or just at least challenge yourself to think in terms of like well what's what's a thousand times this like what would i do if i had a team of uh of 80 million people yeah that's good yeah i mean a lot of people a lot of it's cost constraint but also like a lot of like it's crazy how like the talk of like everybody went from on-prem like to like off-prem and now like to to actually be like you just to win in some of these things you probably have to like store data that people can't find because like you want to have some propriety things to like and it's just like the world works in like circles and like if things go from trends to not trends to trends like you said about the redditors and stuff like dies off and then it comes back and dies in a different way but it still pops up in the yeah it's like yeah those patterns right like definitely patterns in the markets and yeah so lastly where can people find you and what you're doing oh i'm like most available on uh i'm most active on linkedin um and uh and uh i think all of my my handles are the same there are mr spy mr spy um obviously there's like there's spy food channels etc etc but my personal is um is linkedin slash mr spy on on reddit i'm i happen to be sporktopus it's totally unbranded and that's my personal account and you can but it's just spies like a cool alias to go by on the internet though.