← AI-Driven Marketer: Master Practical AI Marketing Skills
AI-Driven Marketer: Master Practical AI Marketing Skills
The Marketer’s Guide to the 2026 AI Landscape
00:40:14
Loading summary
Transcript120 lines
- [00:05]
A
Bro, I gotta say, a lot's changed in the AI landscape and I could have, we could have never predicted how crazy it was going to be over the first, like quarter, I don't know, four months of this year and so much has changed and I've, I've missed a couple of different Fridays now and people have noticed, I've talked to them. So I think it's time to do an update. I think it's time to do a update of like the state of AI for marketers, right? Because there's lots of other people doing states of things and usually you do like a state of things at the end of the year, maybe mid year, but I'm like, I don't know, there's so many things that have happened over the last four months. I think everything's changing. The way I'm using AI has changed more in the last few weeks than it changed over the last year. So I think it's time to like do a rewind and talk about where all the major AI companies are at to kind of give marketers an idea. So welcome back to the AI Driven Marketer. I'm Dan Sanchez. I'm joined by my brother, Travis Sanchez.
- [00:56]
B
Good morning as always.
- [00:57]
A
And of course this is the Bot Bros segment where us brothers break down the hype into helpful news for you marketers because there's so much freaking going on and so much of it's just nonsense. So today we're going to talk about what you need to know and get you caught up in case you miss some important things going on. Because there's all these big AI companies and there's about all the big tech companies are involved now. So we're kind of got to give a review, starting with OpenAI, still the front runner, but has, I think has been challenged more as the leader more over the last four months than it's ever been challenged by Google in 2025, which is incredible to say because 2025, Google was looking really strong, but OpenAI clearly still had the lead. Now Anthropic's actually taken the lead in a few metrics. Not all the metrics, but in a few, I'd say Anthropic's actually kind of winning now. Do you know where Anthropic would be running versus OpenAI?
- [01:46]
B
Where it would be running in terms of its size and growth, comparatively growth.
- [01:51]
A
But it actually overtook OpenAI and its revenue or its annual, what they call ARR, annual reoccurring revenue. Anthropic stole a bunch of paid members or all the people paying for plus a lot of them move from chatgpt to Claude.
- [02:07]
B
Really?
- [02:07]
A
And the API usage has a lot of people using the API changed a lot of their API use to Claude.
- [02:14]
B
Wow.
- [02:14]
A
And Claude's not cheap so they're probably getting a better multiple off it. That's also. How much is it to run? I don't know, it's. It depends on how many tokens and stuff you're using. On the API side it's pretty expensive too. Like on the. If you're just like. For example I use. I was using the Claude's free app. Cause it was pretty generous. Like you could use a lot of Claude for free, just like you could for ChatGPT, also a generous free app. And a couple of weeks ago I like literally just prompted Claude with like one question. It gave me an answer and then said I was out of my free usage.
- [02:43]
B
Whoa.
- [02:44]
A
I was like after one prompt Claude. My gosh. I didn't even give you like a research task or like a ton of context with a PDF or something. I just asked you a question. Gosh. So they're growing so big so fast and I think they, they. We'll talk about that when we talk about anthropic. But it's becoming a problem for them. But OpenAI of course has not of the land grab and has been fighting, fighting doubling down over the last couple of months. Hence they closed down Sora. They're like we gotta focus. Cause like Claude's starting to steal freaking market share. I still think OpenAI and ChatGPT still has the public. You know, like if you ask people if they use AI, it's probably gonna be ChatGPT still. They're still like the winner when it comes to like the most consumers, the most people using it, the most brand recognition. Everybody knows ChatGpt less people know Claude. Even though I think in marketing land like a lot of people are pretty familiar with Claude now. Yeah. But they're pushing hard to catch up specifically with Claude code. That's really been the big deal. They've been crushing it with Claude code. That's why I'm. And they've been. And so OpenAI has been freaking sprinting with Codex and I've noticed cuz they're shipping updates. I feel like I have to push the update app button almost every time I use Codex, which is multiple times a day.
- [03:54]
B
Wow.
- [03:55]
A
They're launching little tiny things or big things sometimes every week for Codex I'm having to refresh that app. There's like a little hey, time to refresh this app because there's a new update.
- [04:05]
B
Yeah, look at that.
- [04:06]
A
Multiple times a day. I'm like, Dude, OpenAI is freaking hustling to get on that, and it's working. So I'm. I'm a. I'm a Codex user now, like ChatGPT. I'm like, why? Why do I have to use ChatGPT? Just give me Codex. So much better. So that's where OpenAI is at. I think I'll come back to this, but I think. I think Codex, specifically for a bunch of different reasons, is actually a better marketing tool than Claude. Coder Claude. And I'll explain that in a minute. Anthropic. Anthropic. So they grew so much over the last four months, they went from 9 billion in ARR to 30 billion.
- [04:36]
B
Wow.
- [04:37]
A
Okay. I'm sure their servers are melting at the. At the amount of demand that's hitting their servers now. Did you know, you know who they're renting servers from now, other than Google, is Grok. Really? Yeah. Elon built this huge, like, freaking new, like, because he built like a freaking mothership in Tennessee, I think it was in Memphis, for Grok, and trained a big model and got it out f last year than anybody had ever expected. But he's building, of course, immediately he went and started building Colossus, like this other bigger AI data center, and bought all the freaking chips because, you know, he had pre orders already with Nvidia ahead of time for Tesla. So he just started rerouting all his orders, all his power to getting more of it to build Colossus. And so I think a lot of people aren't using Grok. So he's got, like, all this extra compute power. Anthropic's like, hey, we'll. We'll pay a premium for that. Elon is like, go ahead.
- [05:28]
B
Wow.
- [05:29]
A
Give me your money.
- [05:30]
B
Yeah, I don't think anyone's using Grok anymore.
- [05:32]
A
So Anthropic's crushing it right now. A lot of attention, of course, marketers. Like, I heard somebody else tell me yesterday that they prefer Claude because they're like, OpenAI's trying to win people over by being the most liked. Hence the. The sycophancy, the oh, I'm always going to agree with you kind of mode. And they're like, the reason why he said he liked Claude is because Claude's more likely to give you the truth.
- [05:52]
B
Does it sound robotic, like Gemini?
- [05:54]
A
No, I'd say because Claude's more known for its nuance in writing that it's actually better. And I've. I don't know if you've noticed. Sure it is. All my writing friends swear by Claude. I guess I'm not that good of a writer because I'm usually pretty happy with what I get by from chat GPT. Like if I want chat GPT to write a sensitive email for me, I give it the context, it gives me something back that I'm like, this is, this is great. I'm really disappointed with chat. So I'm like, I don't. Or, or chat now through Codex. I'm like, I'm really disappointed with it. So I'm like, I don't, I don't. I don't know what's so much better about it, but I'm like, I use that and I use it side by side sometimes and I'm like, that's where I'm at. But a lot of people love it. Google. Google's been quiet. Coming out ahead 25 now. Maybe not ahead, but leg like catching up to chat.
- [06:35]
B
Yeah, they had some huge moments.
- [06:37]
A
Been super quiet. I think they had an update to Nano Banana. They've launched some other more obscure things that it's kind of like eh, you like look at it, you're like. People are like, oh, it's game changing. And then no one talks about it after the day comes out because it's actually not game changing. Nano Banana just got faster. So we're all waiting until May 19th through 20th for their IO DEV conference. Big deal where they launch stuff and we'll see if they launch anything cool. Last year they launched the video with speaking in it in audio. That was a big deal that made headlines last year. So we'll see if they got anything now. The one thing that they have that's missing is their equivalent to Claude Code or Codex. They don't have that yet. So I imagine they're going to put one out for that. And maybe it's better because they have a lot of similar tools. They have Firebase, they have Anti Gravity, which is kind of like an AI coding tool, but it's not like cloud code. It's not operating from your desktop quite in the same way. Maybe it is, I don't know. I tried to like Anti Gravity when it came out, but I just found it not to be as intuitive as Cloud Code or Codex. So problems have you. Anti Gravity? Yeah, yeah, it's like a. I don't know, it's. It's somewhere between using like a coding tool and having AI in it. I Tried using it. I wanted to like it. I just couldn't actually make it work. So I was like, maybe I'm just dumb. So not a helpful marketing tool. I've heard some people talk about anti gravity, but let's move on to Grok. Grok, dude. I feel like they launch features all the time. Every time I, like, once in a while, I actually log into that tool. I'm like, oh, crap, there's a lot of new stuff in here. Cool. And I never use it. What about you? I feel like they're shipping features all the time. Like they're constantly improving a lot of different things and it just seems like nobody cares.
- [08:14]
B
Yeah, I haven't heard Grok mentioned by anybody in months and months and months.
- [08:19]
A
The only time I ever hear someone talk about Grok is if they're like hardcore. Generally, like right wing. Probably conspiracy. Conspiracy theorists.
- [08:29]
B
Yikes.
- [08:30]
A
You know what I'm saying? They're like on that side. Those people love Grok. That's the only other person I've heard of being like, I'm a Grok guy. Or a girl actually did maybe Grok. A Grok girl. That's all she used. I was like, okay, okay. Still kind of watching and waiting. I don't bet against Elon, so I think they're still in the running with all the different things they have, you know, it's a really interesting development with Grok though. Did I tell you about the data centers they want to build in space?
- [08:54]
B
No.
- [08:54]
A
Okay. Your face tells me this is new news to you. So I have to tell you this story. Just, it's a little juicy. It's not really relevant for marketers, but it is just a juicy story. So Elon, probably months and months ago is kind of like, I don't think the future of data centers is on Earth. I think we're going to start building them in space. Very Elon thing to say. Everyone's just kind of like, eye roll. Okay. He's like, yeah, like, if you think about it, like, there's like unlimited sun, there's unlimited space in space, you know, like, you can build crazy stuff up there. The only problem is, like the heat sinks. Like, we have to build these like freaking huge things to like dissipate the heat. And he's like, that's the real. But like, engineering could probably do it. We just have to make him really big. And people are like, what is he talking about? Google is also now like, the CEO of Google is like, yeah, within the next 10 years, that's probably going to be a reality. Like, this is going to be a thing.
- [09:41]
B
No way.
- [09:43]
A
When you consider that, that Google is also now like, yeah, that the economics of that probably makes sense because they're math people. They know, they're like, it costs. We're going to have to build these. That's probably this much weight. The cost of getting stuff to space is this much. Yeah, we're going to be building in space. It's not like a crazy calculation of like figuring out the economics of it. So that'll probably be a thing in our future. Why does that matter? I don't think it matters to marketers because, like the devices will continue working the way they work. But it'll be weird that we will have all these like freaking space stations with just big computers on them floating around in the sky.
- [10:17]
B
Yeah, that's pretty strange.
- [10:20]
A
That sounds like the future to me. But if the, if, you know, if it becomes a bigger and bigger deal, which I think it will, then they're going to need to build more data centers and eventually space becomes the frontier to push them out into. So that should be fun. But that also leads me to believe that because Grok is, that's probably why they merged Xai with SpaceX rather than Tesla. Think about the pipeline there. Think about all the satellites he's already got to bridge out all the bandwidth and communication coming back to Earth, shooting that compute and that all the stuff the AI needs to do back down to earth. Think about who's got the infrastructure now. You're like, oh crap. Like, don't count out Grok. Like nobody's talking about it. But I'm like, It's not an OpenAI anthropic Google game. Even though Grok hasn't been a serious contender pretty much ever, the infrastructure. So that's interesting. Speaking of a person who's never mentioned anymore is meta, Facebook meta. Last big thing that happened there is they acquired Manus, you know, which is a like an agent thinking Chinese model. They acquired it from a. The Chinese company and pretty much like took all the software in. Started in corporate. I've seen it. It's like Manus is there and there and Facebook ads. They pulled all the team in to their team and then China was like, no, we reject this. So I don't know what they're doing about that now. This, this, that rejection came like maybe a few weeks ago. So I haven't actually followed up to know where they're at. But other than that, like they did the big Sprint last summer of hiring all the Best AI talent in the world and spent like billions of dollars on it.
- [11:50]
B
Crazy amount.
- [11:51]
A
Yeah, yeah, yeah. They just threw tons of money.
- [11:53]
B
And how's that working?
- [11:55]
A
It's been freaking crickets. Like, what have they done? They've launched some video models. They haven't really done anything like that anybody's really excited about. No. So I guess that's a pretty common thing in sports. Like when you recruit like the A players to all work together to be the Olympic team, like, it just generally doesn't go well. So.
- [12:13]
B
So I don't know if your reference works because out of the last five basketball gold medal winners, the US team has won four of them, so.
- [12:21]
A
But is it made out of the best NBA players?
- [12:23]
B
Yeah.
- [12:23]
A
Okay, never mind. I heard that was.
- [12:25]
B
You're not wrong. You're not. Yeah, you're not wrong.
- [12:28]
A
Basketball metaphor maybe isn't the right one. I don't. Obviously I'm not a big sports guy. I should pick better metaphors. So that's where Meta's at. No one really takes them seriously. They kind of at the time, like in the beginning had the best open source model, but now, like, China kind of dominates when it comes to open source models by far. And anybody I know using open source, anything, it's always China's models every single time. So that's that. And Amazon, you know, they got really serious. Jeff Bezos went back into the game, rolled up his sleeves, and started getting involved. Now they're kind of like winning and losing. They're winning on infrastructure. They're building their own chips, they're good. They're building their own data centers because they crush on data centers, right? With all their cloud infrastructure and people are renting it, people are paying for it. Anthropic's going to Amazon. OpenAI is going to Amazon. Like, everybody's buying, trying to get more compute. Amazon's like, well, we're killing on infrastructure, but nobody uses their AI models. Do you know what their AI model's called? Haven't even heard of it. Probably most of the listeners haven't heard of it. It's called Rufus. It's baked into Amazon. That's the only way people ever accidentally use it. Kind of like Facebook's models. So Amazon in the game, mostly quiet when it comes to US marketers. Now the one that's the most interesting of all of these is Apple. Apple has been the black sheep the whole time, but now everyone's starting to find out that Apple is probably going to be one of the biggest winners of all, because our AI subscriptions are underpriced, use way more than we like. We pay way less than what it actually costs to get what we get out of AI right now. For your $20 a month, where for certainly for your free, it costs a lot more for your $20 a month. All of us are using it more than what OpenAI actually gets. So the margins are, are upside down. And even for the hundred or two hundred dollars a month plans, they're upside down, they're losing money.
- [14:11]
B
Geez.
- [14:11]
A
That's true for all the models right now. That's true for Google, it's true for Anthropic, it's true for OpenAI. So everybody's counting the costs and being like, you know, there's going to be a reckoning here because business says like, you can't do that forever. There's going to be a bubble that pops here and they're going to have to start reeling it back. You know what that means? That means we're going to have to process more of our AI stuff because we're going to get like addicted to it. We're not going to want to go back. We're going to be using models on device. And you know who's really ready for that? Apple. Apple's got the best hardware stack for phones and computers and they've been preparing for the AI part of it with their chips. So like in a year or two, we're going to be wanting more processing power within our own devices and Apple is going to be the best position to be like you're going to want the newest iPhone and who creates the
- [14:54]
B
processing power for those chips?
- [14:57]
A
Apple designs their own chips. They don't build them. That's still done in Nvidia. That island. Not Taiwan, Not Singapore. Yeah, Taiwan. You're right, Singapore. Wrong island. Taiwan.
- [15:09]
B
Singapore is not big enough.
- [15:10]
A
Yeah, it tiny. Yeah. And you know, the fabs, the, the, the chip maker plants or whatever over there. Apparently we're trying to build them all over the US now but like in Arizona specifically. But like it takes a long time to build and staff it with the right talent and all that kind of stuff. So that's another, another topic for another time because that's.
- [15:29]
B
So Apple's going to be the hardware
- [15:30]
A
for all these, for personal devices. Obviously a lot of people are in the game of building out the hardware and infrastructure for like the big data centers and stuff and Amazon and Oracle and Dell and all those guys are working on the infrastructure, a lot of the infrastructure pieces for the server side processing. But who's going to be there for you. Who's going to be there to really carry you when it comes to like the processing on device. Apple is probably going to have some of the best in class devices for that, whether it's your MacBook Pro. I mean that's why everybody's making the run on the Mac Minis right now, because Mac Minis cheap but powerful enough to run some of the local devices as well as run local and cloud, which is probably what we'll do in the future.
- [16:07]
B
Wow.
- [16:07]
A
So that's the landscape I never would have predicted from the end of 2025 and it's only been four months that Apple would be a front runner in this race in a different way and that anthropic would be really taking OpenAI as hard and fast as it did. But punchline. I'm still an OpenAI guy and I still think Codex is the best one of all time for marketers. Specifically, I think cloud code definitely better for developers. But Codex, I swear, all marketers should just use Codex. It's just better. In fact, I'm actually doing a live webinar next week with Logan Lyles on Codex. I'm going to walk through the process of how to set it up, how to get started, why it's better and essentially do a live workshop. So if you want to join me, it's next Friday at 12:30 Eastern. And when I say next Friday, what's the date for that? Next Friday is the 15th, 15th of May, 1230 Eastern. You can go to dances.com Codex to register to get on that webinar. No, I'm not going to pitch anything. I'm literally just going to give the goods. No pitch, just me showing because I need to show it. But I need to do it live with people. It's going to be better done live and actually taught, hands on, hopefully.
- [17:14]
B
Trying to sell us things.
- [17:15]
A
No, I mean I'm not be. I'm a marketer. I'm not beyond selling people things. If you're listening to me now, like you've probably been on some of my webinars where I'm trying to sell you something because you know I'll take my shot. We're all marketers here. Don't judge me. You all do it. But Codex, did you. Have you. You tried Codex after our last episode?
- [17:32]
B
I did and it made me realize the power of it. Less hallucinations, incredible computing power. The way you described it to me last time was that it's no longer using a single prompt and figuring out what it had just, you know, discussed with you, which is where it starts to hallucinate, but really builds on top of itself. I was able to give it a plethora of information that it filtered through to help me in the task and journey that I was on. The good things is that it literally is agentic in its own way, trying to. But it starts. I still get remind. It's still trying to, like, download stuff from the Internet that I'm like, I don't know what it's trying to. I don't know what it's trying to download from. I'm like, what is happening? Why is it still trying to download things? Even though I'm not actively using it, it's like closed on my computer. So it makes me a little nervous. I almost wish I had a computer that was separate from the one that I use as a daily driver, you
- [18:24]
A
know, because it's asking you permission to do something. Yeah, I just turn those off and let them run. Generally, it's asking for permission to, like, do something on your computer. I'm like, do whatever you need to do. I've never found it. It's gone off the wall because. Yeah, I don't know. It's always worked pretty well and I always let it. I only give it access to a folder at a time, so it can't really delete anything outside that folder or mess anything up. But I gave her permission to do anything it needs to do within that folder or go research the Net. I don't give it access to everything on my computer. That's where it really starts to get a little scary. That's why people buy separate computers, why so many people are buying Mac minis to set up a separate computer to make sure it stays contained. You know, that's for, like, Claude code or not cloud code. I guess we people do do that with cloud code, but specifically openclaw.
- [19:09]
B
Right, Right. Anyways, I just don't even. It does make me nervous because I'm like, I don't know. What are you gonna do?
- [19:16]
A
I do everything through Codex now. Everything after he's chatgpt, I'm like, ugh, dang it. I have to give it so much context every time. But now it just openly goes and finds it. That's the thing that really changes the game with Codex is that it actively looks for context on the Internet. On the Internet, but even in within your own files. So within the folder, it has access to if it has lots of things that, like, leaves breadcrumbs for itself. To go and be like, oh, this is what. What's in here? Oh, check this out. For here, for here. So it like goes and leaves. Reads its own freaking notes. It leaves little read me files. And it reads them and then reminds itself, oh, I need to go check this out. I need to go look at this. This is the process for this. Okay, I'm back into it. Because it actively looks for context. It goes and figures out problems way faster because you can. You can tell it be like, hey, when we run this process again, change this, do it this way. And it'll be like, oh, okay, let me make a note of that. You're like, great. And then it could just do it itself. You don't have to update the process anymore. You can have it update it for you. That's the biggest difference in my opinion of what Codex does better than chat or what Claude code does better than just claude, is that it can actively look for its own context and make changes and leave breadcrumbs for itself. Like, it can make. It can take action, right?
- [20:21]
B
Wow. I haven't messed with it a ton other than the one project that I've worked on it with, but it was impressive.
- [20:26]
A
How long did it take to work through that project? It took a while. I mean, I was working details like, what was it actually doing for your project?
- [20:33]
B
It was consolidating a ton of information, surveys from a large group of people in their desire for what they wanted, and then building me graphs based on that information, pulling information from online to plot different people on a.
- [20:49]
A
So it had to. And it probably had some docs to go through, so I had to check with docs. It had to check with a big survey CSV data file, and then it had to go and verify different things with the Internet and then analyze it, come up with a report, and then give you like an insight folder. So, like, that's a lot of freaking work to do. And then going back and forth between those sources and proactively looking online and then coming back to the docket, proactively looking online, making notes, checking the survey again, running that, it's like. It's like a back and forth that chat and Claude couldn't do by themselves easily. And because it can keep notes along the way, again because it can make and update files, it's just more robust. Did it give you a good answer?
- [21:26]
B
Fantastic. Yeah, it was so helpful.
- [21:28]
A
So I'm trying a new project right now. I. I have a feeling it can work well, but I'm not sure yet. But this is what I'm going for. I kind of gave up on blogging and the whole SEO thing until now because really the amount of work it took. And I did a huge project like in the early days of this podcast I actually like went and found an automatic blog AI tool. I paid a one time fee for it and then I went and tested it. I literally bought five unique domain names that had five different niches that I was going to do and I auto blogged all of them. You're talking like a dozen articles a day for like for months in a row to these. And they gained initial traction. But then of course the content was freaking crap because it was all hallucinated back then. That was before reasoning models was before a lot. And even though the prompts were good and as best as I could get it, they just, the content it was making just, was just subpar. Like it just wasn't that good. So naturally they didn't rank well because you know, to rank well you gotta not just hit the keywords. You have to have good content and have unique content. You can't nowadays you lose even if you have, even with the reasoning models, you can have it scrape the Internet and come up with a article, but it's got to sound like all the other articles because that's where it got its information from. But I'm like, what if, what if you have a lot of original information? So I'm like, okay, this is what I'm going to do. I'm going to set up a whole new project. I'm going to say, hey Codex, here's all the transcripts from all my podcast episodes. Here's all of my LinkedIn posts ever. Here's a login to my WordPress site. Go ahead and pull and make notes on every single blog I've done. This is hundreds and hundreds of files. And here's my book that I wrote, 60,000 word book. Here's all the context. Now you know what blogs I've written, map which ones, how I can extend out my blogs and create more. How can I. What would you do with SEO and AEO in order to expand this blog out and get more traffic through search? So it comes up with the plan, maps, looks at my content, sees where I have good information that can supplement all the new blog posts that need to be written. And I'm like, also figure out what you can do just through Internet research and then what gaps are left over that would be really strategic for me to give you more information on. So that's the New plan I'm doing, and then I'm gonna have it come up with the plan. And I'm like, okay, now I want you to schedule at 5am every morning, I want you to write five new posts, and then I'll review them, and then I'll have you post them. I want you to also come up with the image for them. And then because it has access to my WordPress site, I don't have to upload it. I could just look at it, tell it to make a few changes, and then it can upload it. With the featured image already set, it's actually powerful enough to do that, size everything correctly and make it look freaking good. So I'm like, SEO is back on the table in a really compelling way that doesn't take much time, but can still be really good because it's based on what I've said before.
- [24:02]
B
Wow.
- [24:03]
A
Every new blog post, it can re index and go back through my LinkedIn post, go back through my book, go back through the transcripts, and pull new information, what's relevant from it so that it's actually original content supplementing with stuff it can pull off the net. And I imagine every morning, maybe it asked me like, five strategic questions that it can get for the five blog posts. I'm like, hey, if there's anything I can get cheap for you for this group of blog posts, here's five things I want to know from you if I have anything to say at all. So I'm like, it takes all the context, tries to fill in the gaps, and then pulls a few extra pieces out for me to make sure that these blog posts are actually unique.
- [24:35]
B
Wow.
- [24:35]
A
So I'm like, come on. Like, this is the stuff that's now possible that you could never do a chat.
- [24:40]
B
Right? But with Codex, possible.
- [24:42]
A
Yeah. I. I don't know if it's gonna work, but I'm gonna give it a shot. Like, I'm starting with a website that already has rankings and has, like, at its height, you know, got 10,000 page views a month. So not a lot, but not a little. I'm starting with a established website, dances. Com. I'll. For everybody listening, I'll give updates along the way of how it goes. If it does it. If I get started ranking for some of these things, I have a rank tracking tool. So I'll be following up with that. Lastly, one News item is ChatGPT ads. Did you know they've been testing ads in ChatGPT? Mm, mm. They have. They've had a limited beta with probably the big advertisers, they're rolling out that beta to more people and they're adding a self serve portal now for CPC bidding cost per click bidding. Kind of like Facebook ads.
- [25:26]
B
Gross.
- [25:26]
A
Now it's still limited. I can't. I'd love to get a shot at it because I'm like, I'd love to spend some ad money. Unfortunately. I like, I hate it as a consumer, I love it as a marketer, but it's just going to be reality. They got to make money. They got to make money and that's for the free users. So you're a plus user, I'm a plus user. I would never see it. But for all those free users, they got to start making money somehow. So makes sense. It's funny that, yeah, it does make sense because what we paying plus again, we don't have like a margin left over to cover the free users. So it's like they can't subsidize it through plus alone. They have to make up for the free users and save some of that money.
- [25:59]
B
Wow, I did not know that.
- [26:00]
A
So that's coming. But for all you marketers out there, that's something to pay attention to because that might be a new advertising opportunity for those who deal with paid media. Sponsor of today's episode, of course is High Level. They've been updating their AI Studio builder a lot. That's their like AI Vibe coding website tool. And it's getting better and better to the point where you could just like, again, just like you would with replit or lovable, just give it an image, give it your copy and it'll make the website for you. The cool thing is like it already has all your tools. Like it has your high level web forms, it has your automation sequences, it already has all the context and your color palettes and your brand boards and all that kind of stuff. So it just works better because it exists in the ecosystem of all your marketing stuff already. So that's why I love using it. If you want to check out High Level, it's an all in one marketing platform. It's like HubSpot. But honestly I like its features better than HubSpot and it's certainly a lot cheaper at a hundred bucks a month for and you get three accounts. Like you literally run three businesses out of it for a hundred bucks a month. But check it out. Dances.com highlevel. Try the trial, watch a few tutorials. Someday I need to do my own walkthrough on it, but I promise it's what I use. I love it. So check out high level@danceshaz.com highlevel for a 30 day free trial everyday. AI. How have you been using it lately?
- [27:10]
B
My son wanted to know the cost of a blue macaw yesterday and he was convinced that he was going to use some of his own money. 7 year old to buy a blue macaw. I said, yeah, that says it's anywhere from 10 to $35,000 for one of those bad boys and it costs about $6,000 in annual, you know, care costs, specific food and setups and they last for seven. They live for 70 years. So 70 times 6,000. You could do the math on cost of having an exotic bird like a blue macaw.
- [27:46]
A
I had no idea. Did you watch Rio or what?
- [27:49]
B
Yeah, they love Rio.
- [27:51]
A
That's. That's it. I'm like where else would he have known about a blue macaw?
- [27:54]
B
So that was helpful to let not totally pop my son's bubble but.
- [27:59]
A
Well I just gave the use case that I am experimenting now with Codex and my new blog or SEO, but let me give it like a more micro one. Before I even started coming up with like oh what could I be doing for new blogs I had it just log into my WordPress account and I'm like hey, just here do some research on what the most up to date SEO AEO practices are. And then it did that. I'm like okay, can you just scan my whole website to figure out what needs to happen per page? And it came with a whole different audit and then strategic analysis of every single page based on the research it did. I was like okay, can you like do the most do do modifications that don't change overhaul the page but like make simple modifications that will be the most bang for the buck. Cause I don't want to overhaul every, rewrite everything because like Google will notice and re index everything and do it for every page. And it went through and just knocked. You know, probably took it 20, 30 minutes to go and knock out every single page on my website. It's probably like 200, 300 pages. So it was one of the easiest, coolest things at social media marketing world. One of the speakers told me that like they had AI with an SEO I go back through and like rewrite all their LinkedIn articles without them having to do anything. And then traffic on all their articles went up probably like two, three times. What what I was getting before I was like dang, I'm going to try this on my website so I'll Report back on the change. But. But just cool things like that.
- [29:13]
B
Codex can do one more that I did. It's more marketing focused, is I work next to a daycare and I want them to succeed because they work out of our building and they are a great daycare. But I saw their Facebook ads come up on my feed as I was scrolling and it was just the worst ad paid ad that I could. I just. It was so painful. I screenshotted it, I uploaded it to ChatGPT. I said, can you please just give me a better ad for this daycare? Like, here's their website. Just do better. And it, of course, slam dunk, first try. So I showed it to someone on the board. I was like, this is how easy it is to make a decent Facebook ad. She was like, what the.
- [29:57]
A
Like, oh, dude. Yeah.
- [29:59]
B
So I didn't actually use that. I just kind of, you know, let them know the tools are out there to be. Not posting a photo. That made me cringe. It's fine.
- [30:09]
A
What do you think they made it with? Was it like a canva photo or they just upload like some random iPhone photo?
- [30:14]
B
You know, you know, you know how that was a bad iPhone photo and like, you know how you up upload a horizontal photo to your stories on Instagram and then it fills the color in the background and then you can put text above and below it. That's kind of what they did.
- [30:27]
A
Oh, dang.
- [30:28]
B
Yeah, it was bad.
- [30:29]
A
That's rough.
- [30:30]
B
So no design photo? No, it wasn't a good photo. No design.
- [30:34]
A
I'm like, yeah, super appealing, man. I freaking love chatgpt to learn. I was in Disneyland last week because again, social media marketing worlds in Anaheim. If you missed this year, join me next year because it's freaking amazing event. But I love that they moved to Anaheim because of course you can go to Disneyland afterwards. I did. But, you know, in your Disneyland, you got a lot of time in line, so you pull out your phone and I'm. I'm a learner, dude. I like. So I had the. Just the most fun time sitting in line and learning about every single ride I was on. I was like, chat. Give me the origin story of It's a Small World. Where did this come from? Oh, man, Fascinating. Whole backstory. Then you're like, oh, tell me more about this, tell me more about that. So like, through the lines were super fun because as I'm walking through the line looking at everything and all the art and stuff, I'm. I'm learning, I'm getting the History. It's. It's like the ride is so much more fun when you understand the backstory to it. Kind of like anything things are, when you understand where something came from, you appreciate it more. So it just made the whole experience so much more fun because Chat had a pretty good understanding of every single ride. The most wild one being, of course, Mr. Toad's Wild Ride. Do you remember that ride?
- [31:39]
B
Yeah.
- [31:40]
A
You remember it ends with you getting smashed by a train, dying and going to hell. And the last like 20 seconds of the ride is in hell. You're like, oh, this is a kid's ride. What the heck happened? Yeah, Chad gave me the background story behind that too. Is fun.
- [31:53]
B
Wow.
- [31:56]
A
So fun. Way to use chat. Learn. Get context of the things around you or the places you're going or the people you're meeting.
- [32:01]
B
There you go.
- [32:02]
A
No poll today, but I do have a viral post or more like, I guess all of his posts are viral. So they're interesting to me. From Ethan Mullock, and this is. Let me, let me read it and then I'll give some context of why I think this is important. He says organizational design for agents is hard. Benchmarking agents working in concert is hard. Together, this is the next critical frontier for making AI matter, an economically valuable task, and we really don't know very much about it. Okay, so what does this mean? I'll. I'll put a link to this post because he shared somebody else's post, a nice little infographic. What does it all have to mean? I actually bookmarked this one because I'm like, huh, A lot of people with this whole agent thing and cloud code and codex and how useful it is an agent. The more advanced players are not just launching agents to do things. They're launch groups of agents, sub agents, agents who go and start up other agents to go and do things. So you're talking like multi layer deep agents sometimes, okay? They even call them agent swarms, which is like a God awful way of calling it. Like there's nothing more scary than calling them agent swarms. I'm like, yeah, I should have rebranded that too late. That's the name to go and do things. The problem with them, and this is what Ethan's talking about, is like, they're not working well. Anybody who says they're creating a whole organization, their whole organization is run by agents and their agents run agents, is either extremely technical, it's kind of lying, or is covering up the fact that they're making lots of mistakes. You can't have agents give agents things to do. It'd be like hiring a bunch of really good interns and having them kind of like run things and hire like work with themselves. Like they're not going to do a good job. As competent as they might be. That is just not quite a thing yet. But it's definitely the next place. So a long time ago, almost two years ago, OpenAI released this like stages of artificial intelligence. It was stage one through five and it's funny, they said like some of these stages would go faster than they actually have. Stage one is chatbots, AI with conversational language. Stage two is reasoners human level problem solving. Stage three is agents systems that can take actions. Stage four is innovators, AI that can aid in invention. And stage five is organizations, AI that can do the work of an organization. Where do you think we're at?
- [34:09]
B
Right, two?
- [34:09]
A
I'd say we're finally getting to three and we're in the early stages of three. They've been talking about three for over a year. It's they as soon as they got to stage two with the first reasoning model they're like great. That took us three years to get there. We think we can get to stage three much faster. That turning out to my opinion to not be true. These all overlap by the way. It's not like you get from one to the other. It's like the first reasoning models we had sucked and then they got a little better and then they got a little better and now they're like pretty mature. We're almost two years into reasoning models. Guess we're a year and a half and they're finally like okay, the reasoning models are really reasonable. Agents are still pretty new and I think we're on the early stages of agents. We are not quite yet having the innovator stage. This is where they're pushing really hard now. And organizations with agent to agent management I think is still so far out that people talking about it are freaking lying through their teeth or kind of misleading on how successful they are because of multi agent work.
- [35:00]
B
Right. Well it's kind of scary but that's
- [35:03]
A
the hype right now. Now that agents are becoming a thing. The thing that's getting the most hype is multi agent work or organizational work. And I'm like freaking get out. We finally just got agents that are actually capable of like executing something that's like multi step and can troubleshoot the little problems that come up. But I can't be like here's a login to my WordPress site, make the traffic grow. Do you think AIs like, even codecs or Claude code, as smart as they are, can't do it without me thinking through like, huh, do an analysis. Like I'm still directing it. Like, I'm still saying, like, do an analysis of what's there. What could be better. Think about what, how I can expand it from there. Let me think about what I've actually said that could be useful to make original content. Is it possible that it could have figured that out by itself if it were like really good? Yes. Is it going to do that by default yet? No, not really.
- [35:48]
B
You know, it's like anything because as humans we gain and garnish discernment based off experience and that's, I mean it's still learning, it's still trying to figure out at a high level what's actually wanted or desired from the individual that's asking for it. So as these AI agents learn the creator or the person who's asking, they'll have better discernment moving forward. But it's still, I don't know, I don't know how, I don't know how it's going to learn exactly what each individual user is for. My, my mind, it just comes down to ethics because like what decision base is it making? Right? Like how humans have used all sorts of methods to motivate their team to get an end result of success of but then they've had to learn because I've like burned through their team being an absolute tyrannical leader and they go, well that didn't help with the long term result of keeping my business afloat because everyone quit on me. Yeah, well, at least I think I need to change my approach or hire someone that will. I don't know, I just think of like all of the learning processes of anyone who's in leadership and there's like a morality and A and a consciousness to it. So I'm just like, is it serving the consciousness of the person who's running it or is it serving its own consciousness? And it like doesn't give a rip, it's just trying to complete task A to B as quick as possible. So I'm like, that's what it is.
- [37:02]
A
My brain is still a machine. It's still a machine. It's not conscious. It's literally just inferencing, trying to get to the right answer.
- [37:10]
B
I know it doesn't have a conscious, but it's trying to create its own conscious based on the conscious of person that it's running Right.
- [37:16]
A
Like yeah, yeah.
- [37:17]
B
Because if I. It knows that I'm a care first leader who wants tasks done second over care for an individual. So if it knows that that that's the.
- [37:26]
A
Yeah, it will totally there. And people are programming open claws more known for this because people are giving it like forget what they call it. It's not heartbeat. That's a different file. There is a soul Soul doc, where you're actually hard coding in the values that you want it to have to give it its soul on its personality and all the different stuff. So openclaw has that most of these don't. You can make one easily and just call it SOL MD for a markdown file and tell it like, here's the values you should run through first when making decisions. For the most part, the things I have Codex doing are like very. Even though it has a lot of liberties to do things like I'm still very much guiding what it should do and it's big. I'm still giving in objectives and breaking down its objectives into very tactical things. I'm still deciding the strategy. It might give recommendations, but I'm still the one guiding it. Now if you're giving it much more and autonomy, you need to give it values in order to make decisions from. And that's why good, good organizational design is around mission and core values. So that people can actually make ethical and decisions as if they're being guided by larger principles. Right?
- [38:34]
B
Uh huh.
- [38:35]
A
Kind of like Southwest if you know their, their values. And it's like it's made Southwest as good as it is as an airline. They're like, well, number one, we're cheap like we're. Or sorry, like we're the low, we're low cost. So any decision you make, does it help us become a low cost? Because if it does, then that's a yes. And we're fun. But fun should never come at the expense of value 1. We're the low cost airline. So if you want to pop confetti when people come onto the airline, that's probably a no because then it creates mess and then we have to clean up and it costs more. You know, if people are coming on, you're like shooting the little poppers. Welcome to the airplane. You probably wouldn't do that, but can we add some jokes in our takeoff flight instructions? Yeah. Does it cost anything? No. Does it make it more fun without us having to cost more?
- [39:19]
B
Yes.
- [39:19]
A
You know, so you've accomplished the values. It gives them staff members, decision making criteria so they can ask like, oh, I wonder if I could do this. You have values to guide their decision making. And people are starting to experiment with that, with AI right now. But it doesn't matter too much yet, because we're not at a place where we can run agent to agent yet, at least reliably. Right, but that's where it's going.
- [39:41]
B
Dang. Make sure you have a soul file.
- [39:46]
A
Got.
- [39:46]
B
So, actually, do you have a soul file in your own soul? That's the real question.
- [39:51]
A
Yeah, well, if you. You could. You could probably get into a fun little philosophical conversation with. Chat about that if you want.
- [39:57]
B
I bet you've met people that don't have a soul file in their own soul.
- [40:01]
A
Where's your soul? Md Imagine if that makes it into normal conversation. Oh, man. Well, I think that's pod for today.
- [40:07]
B
Pot it. Pot out.