GPT-5.4 Changes ChatGPT in One Huge Way

Foreign. It's been a Busy Week because ChatGPT has been throwing out some updates, Google's been throwing out some updates, and everybody's talking about moving to Claude right now, even though both the big powerhouses are throwing down. So today we're going to talk about those changes and if it's actually worth moving everything from ChatGPT to Claude. So welcome back to Bot Bros, where we cover the news and separate out the help from the hype that is going on all the time in the AI news cycle. Just for marketers. I'm Dan Sanchez and I'm joined by my brother, Travis Sanchez. Hey.

0:00

Hey.

0:37

And today the big leading story is ChatGPT 5.4. Now this is 5. 4. It's interesting. They didn't do a cross the line upgrade. The 5, 4 is specifically the thinking model, the more powerful of the ones you have in your typical Chat GPT account. And it was a little bit of a spec boost. Like if you look at it, you know, every time everybody launches a new model, they compare it to the other models and they're like, look, it's this much better than the leading models and they have all these evals. They do it and it was definitely better generally across the board in every way. And the one that caught my attention, there's one eval, it's called the GPT VAL or the GPT eval or something like that. And it is a measure of how well ChatGPT performs against experts. So it does expert tasks and then compares it to human experts and then has a panel of human other experts in that particular field that grade it did the AI win or did the person win?

0:38

Wow.

1:37

So it's competing against actual experts in a bunch of different fields. And ChatGPT is now winning a total of 82% of the time on expert level work tasks. The reason why that eval matters more to me is because that we're in marketing. Like, like marketing tasks are part of that evaluation, even though they're measuring it across multiple fields. And AI is just winning across the board. Now it still can't do cross browser tasks and all that, so you still need a human loop on a lot of this stuff. Because AI still messes a lot, but so do humans. So it's good in a lot of the evals. But let's talk about what else it got good at. One is it's got better at web research, so it's going to be more clean, it's got less hallucinations. So that's something that you know Hallucinations become less and less. Every time we update the model, it hallucinates less. Remember that used to be a big deal. Like that used to be the main thing we were talking about a year ago, two years ago, it was like, oh, hallucinates. Oh, it hallucinates. Do you hear people talk about that anymore?

1:38

Actually, I was just talking about Open AI yesterday and about some hallucinations funny enough not to get political, but he was asking open AI is Charlie Kirk alive or dead? And it was like, very much alive. He is kicking and breathing and he's like, no, it's not. So he said he went to Gemini and asked the same question and it was like, no, he died September 10, 2025. He goes, okay, good, because open AI said he's alive. And then Gemini goes, actually, that's right, he is alive. Oh, you know, literally had this conversation last night with somebody. I. So it's experienced. I haven't experienced as many hallucinations, but some people still clearly, clearly.

2:38

And it's, it still messes up, but it's becoming less and less to where I feel like it's just not like that used to be the major pain point all the time and now it's less so. But clearly, especially if it's getting that wrong. Usually like when I ask it questions like that, you can see its little thinking bubble pop up and the first thing it usually thinks is, oh, maybe something's changed since last I've updated my model. I should go check the Internet.

3:21

Interesting, right?

3:47

And that's usually what it does, even with the dumb models. Like it like thinks about it for a second. It's like, I should go, this sounds like something I should go check, right? Because it's, it's, you know, it's training data is only updated every, every so often. So like the last time it's had more recent long term training data is probably since before Kirk, Kirk died. So.

3:48

Yeah, well, when I search that question just for everyone to know, it says he is no longer alive. So I don't know. I don't know what I didn't get detailed, you know, like, are you paying? Are you. Is it free? Did you actually.

4:13

What was your prompt exactly?

4:26

Did you not download Chat GBT and you downloaded the one that's like the fake Open AI one, you know?

4:28

Ah, maybe that's it. Who knows? Maybe it. Yeah, I'm seeing that hallucinations way less.

4:32

He's like, I'm using CHAT GBT and I, I'm like, let me see that. I'm like, no, you typed in AI Chatbot and downloaded AI Chatbot, which is not the same thing.

4:38

Yeah, yeah, it's funny. The big thing though that came out with ChatGPT 5.4 thinking is that it finally, finally fixed like one of the biggest problems with OpenAI was the context window. Now what's the context window? It now has a level of about a million. It can handle a million tokens. So what is that? That means it can handle about 900,000 words ish of context in what span of time? So it's like a multiple book series, Right? Like a typical book, a business book, your average business book has about 60,000 words in it. It can now handle. Before it could only handle about a hundred thousand words. Now it can handle about 900,000 words.

4:51

Like at one given time, if you,

5:38

you could have sent it more, but it would stop. It wouldn't even read it. It couldn't. It couldn't handle more.

5:41

Oh, the.

5:46

That's important because the other models have had this for a while and it's one of the big things. Like Claude can handle a million. Google can handle 2 million.

5:46

Oh my gosh.

5:54

Which is pretty good.

5:55

So if I wrote a book and I wanted Chat GPT to be my chief editor and I had a digital copy and put it into OpenAI, it would stop after so, so, so many words.

5:56

It wouldn't be able to think about the whole context of the book at once. You'd have to do it in sections, which is important because I actually just did that. Claude wrote all the chapters for the book that I had. Night was based on the transcripts. We've talked about it on the show before, but because Gemini had the largest context window, I let Gemini do the big pass editing because I could give it the whole book. What's. What's like, tell me where, where did I. Where did. We talk about the same thing too many times across chapters, and it identified it correctly where we had done it across chapters. This is a big deal because I think it's been one of the bigger problems of OpenAI being far behind in this one area. I don't think people realize how important this is because the more we work with AI, the more we find out that context is actually the thing that makes AI good or not.

6:09

What's anthropics?

6:52

A million. So ChatGPT is just caught up to Claude. Anthropic Google. Google was at 2 million almost two years ago. They were like went way ahead on this one. Then their model caught up to being good enough to handle it. Now ChatGPT is that good. The reason why it's so important is because I find I'm feeding it more and more stuff to help like company documents, research and trying to give it all this stuff so it can have be properly armed with all the relevant information to help me do these little things. It needs all the context and it's just like an employee. If you give an employee, you have to like give them that employee handbook. You got to give them through all the training, you got to give them all the context of their work situation. It's a lot of context you need in order to make decisions and AI kind of needs it every single time. Unless you fine tune it and train it on all that data so it becomes part of its long term memory. Otherwise you have to load it with all the context and it has to be able to take it all in at once and then make it next to choice. So it sounds technical. But more and more the people who are crushing it with AI are giving it a ton of context. And before you could only give ChatGPT so much with, without before it started like screwing up where people will start to feel this the most. In my opinion, this is, this is the real key. This is where marketers will feel it. And I think my, because again there's, there's a huge amount of people. A lot of my peers on LinkedIn are like saying I'm done with chat GPT. That's it, I'm moving to Gemini or cloud. I'm going to figure it out. I'm like, okay, just wait until they update another thing. Because the one thing chat crushes it on is personalization. All that memory stuff. And guess what gets a lot better with all that. Once it got now that it has a larger context window, that's going to get way better. Wow. So if you haven't in a while, you might want to ask Chat Chat, like start testing because it's in my account. I don't know if it's at your account. The 5.4 thinking start throwing prompts you've had it at before to evaluate your past conversations. It should be way better. Why? Because it can go and find it better and handle more of the context of all those conversations better. See where it starts to become powerful.

6:54

I've, I've searched past conversations to kind of hey, based on past conversations, what, what do you, what is your evaluation? And it's like, sorry, I can't do that. It was like I, I, I mean I can remember what you told Me to remember but not go. Search past conversations. So now you're saying it will have a better time. Easier time doing that kind of.

8:57

You have to remember when it's searching past conversations, it's actually not searching the actual conversation. It never did. All the conversations are stored. Are they creates a whole separate piece, like a summary of the conversation with important details. So anytime it's going back into past memory or past conversations you've had, it's not actually searching the exact conversation, it's searching through the summaries created of the conversations. And my guess is they're. They're actively editing it every single time you have more conversations and it remembers, it has a longer summary of the more recent conversations. And the farther back it goes, the more it drops those. It starts shortening the summaries because why? Why would they have to do that? Because it only has so much freaking context. Right? So they had to figure out how to mechanically alter summaries and only include bits and pieces of all the conversations to kind of fit it. Shoehorn it into the smaller context window. Now that it has 10x the context window, well, you can imagine those summaries are going to be longer or more thorough, which means it's going to have an easier time remembering past conversations. So Chat GPT is going to continue getting better and it's continuing to crush on that personalization feature, which is why it's still kind of my daily driver. But I will say more and more, I am using Claude and Gemini regularly.

9:18

Gemini almost daily with both Gemini's personalization. Have you messed with that at all?

10:49

No, because they won't let it in. They won't let you have personalization in the workspace edition, which is what I pay for. I could just open up a free Gmail account and experiment with it there. But I'm just like, I just don't care enough. And I can imagine that it's just not as good because they just only released it November and you know, it takes a few cycles to get it right. Or chatgpt has been on this for a while now. Yep. So that's chatgpt5.4 thinking it should be in your account. If. If it's not yet, it will probably will be by the end of the day. Go and test it out. Throw some of the longer things you've given to ChatGPT and check. Check this again. Again. Suspect boost mostly. But the big thing, longer context. And I think where it really shines is not only in remembering your past account, but when you have it Go into research, do research for you on the web. That's going to be better.

10:55

I like that.

11:46

Oh, one other thing they did is that you know how it goes and thinks and then you can't talk to it.

11:47

Yeah.

11:52

Now you can, you can be like, hey, oh also, also while it's off, you can kind of like throw things at it while it's thinking and doing research for you. Yeah. So that's, that's an improvement too elsewise. In the news there were some other updates. ChatGPT didn't just upgrade its thinking model, it upgraded its instant model. It's like lowest performing model. Which is interesting because people are probably using this way more often if they think if you turn it on auto, this is the model you're actually using quite often. If you're a free account user, you're using this one all the time. And they bumped this one from chat GPT 5.2 to 5.3, which is interesting now that they have different numbers because it's usually they move to 5.4. They would have done that across the board, but not here apparently. So the instant model is now 5. 3. It's just spec boost everywhere. Faster, better, less cringy. It can handle emotional nuance much better. And Google also updated its what they call flashlight 3.1. It's small model that when you're doing the instant answers, this is the one. This isn't a big deal for pro users. People probably listening to this show but notice all your friends should probably notice an upgrade who are using the free models a lot. And if you're just using the auto and it's just giving you a fast answer, you'll notice those getting better too. So ChatGPT came out strong with like a big thinking model upgrade and an instant upgrade, which means chatgpt across the board got a lot better this today. How often are you using the thinking model versus the instant model?

11:53

I haven't clicked. I literally haven't clicked the thinking model in months and months and months.

13:24

Really? Geez, I feel like I do it all the time. I you just leave it on auto.

13:30

I leave it on auto and it has to do with my personality. I not a learner. So when I really want to understand something, the quick synopsis is what I'm looking for. I don't want the dissertation level research thing now. I have used it when I'm like I need to know exactly, like if I'm going to make this plan, what does it exactly look like? Tell me things in the past. What's going to happen in the future. Whatever. I have done it and it is helpful but my day to day life, there is not a time where I'm going, yes, I want to sit down and really understand it just it's more personality based.

13:36

So what about more web research? You wanted to look up something for

14:13

you

14:17

because that's where the thinking model really shines is when you want to go. You don't want to do a deep research project but you want it to go.

14:19

And that's funny.

14:24

Can you go search these things and find the best.

14:25

I switched to Gemini

14:28

and you use that thinking model. Okay.

14:31

Yes, because what the information that it pulls from Gemini. I used to go to Perplexity for that kind of research but I just went straight to the source and yeah, went to Gemini. So yeah, I, I, that's, it's funny the way your mental neurons work when you're like using different AI models that when I want to look, look up something on the website websites across many. I just go to Gemini and I find that its responses are more thorough and give me what I want not to the thinking level depth explanation. It's still a synopsis but it yeah

14:33

doesn't have that personalization fast and friend and friendly warm kind of thing going on.

15:11

Yeah, it's like fast, bold and beautiful. You know, on Gemini. So I'm like, yep. I just went down a few points in Dan's brain about how I use AI. He's like, you idiot.

15:17

No, I mean you're still using the thinking model. You're just doing Gemini for it. Yeah. So that makes more sense. I use auto all the time. There are many, many people who only use ChatGPT's thinking model, thinking that that's the best for every situation. I'm like, yeah, not really like some. I mean, yes. Is it better? Yeah, but sometimes it overthinks things and sometimes you just want it to give a its gut reaction or using having it not reason through things actually can be better for a lot of different things. So I'm using it all the time. All right, last in the news is Notebook LM came out with a video model and it's really good. Like it's not like the kinds of videos we've been talking about with seed dance and like action like action stars fighting each other. It's not that kind of video model. This is more like because remember we talked the last. I haven't covered Notebook LM. A lot, a lot of people are like raving fans of Notebook LM and I'M like, I don't get it. I like, I like it. But like, it's not something I'm using daily. I use it every once in a while for some things. So. But this video model is really cool. It can essentially, you know, it's got the. Let's just start with this. Like, the audio overview was something that blew everybody's mind about a year and a half ago. Right, with the two voices doing the podcast thing. Yes, They've essentially done that, dropped it down to one voice and it's a voiceover. But now it has essentially interact, not interactive, but B roll layered on top of it. So it's got strong captions at the bottom. And it's got like, kind of like those YouTubers who like lean a little bit too heavy on B roll. You ever seen that? Yeah. Or it's just a voiceover.

15:28

Faceless YouTube channels where it's only B roll.

17:10

Yeah, yeah. This, essentially that. Yeah. And the B roll face.

17:12

The guy doesn't have a face, right?

17:17

That's right. No face.

17:18

It's called Faceless YouTube.

17:19

Yeah, it's a. Yeah. So that's what Notebook LM can now do, which is impressive. If you haven't been there recently, go to Notebook LM video or just go look up Notebook LM video on X and just watch people's examples. It's impressive. Sometimes they're just like pan ins of photos, but sometimes they're like animated, like. Like motion graphics stuff that it kind of looks stock photo Y, which is why it really looks like the faceless YouTube videos. Because it looks like something that came from a stock video site of some kind of animation. But it's usually good as in it's the right animation explaining the concept, which is really cool. It makes me almost want to use it as a way to come up with interesting B roll that maybe I just cut out the voice voiceover and just put it in on top of my voice. I'm like, it would work because of the way they explain the concepts and the way it animates them. They're simple animations, but they're really good. But I can't wait till they combined like a Notebook LM video with like a, you know, like a seed dance where you actually have the A roll in front of it. Then it's just kind of game over for a lot of YouTubers that just teach and explain things like, so I'm

17:21

on NotebookLM and they have sources, they have a chat, like a chat bot. So you can ask about the resources that You've put in. And then they have something called Studio. And inside of Studio is where you can have the original thing, which is the audio overview. There's a video overview, something called a mind map, reports, flashcards, quiz infographs, slide deck, as well as a data table. So they have multiple forms of media inside of the Notebook lm, which is kind of fun. I'm gonna play with that.

18:37

Yeah. The infographic in particular is powerful. If you make it really complex, then it starts hallucinating stuff. But it does give you like a hey, this is beta warning on the more complex infographics. But the light and medium infographics it can make up are fantastic. Like, wow.

19:08

You click video overview and then it like, okay, what format would you like it? Explainer video or maybe a brief video. Choose your language, choose the visual style. Wow, that's crazy. It's. It's like a GUI system for your video. That's wild.

19:26

So as a marketer I'm trying to figure out like, what do I then do with this? And I'm just thinking it becomes supplemental to video. If you're already doing YouTube style videos, it is possible to make a whole nothing but a whole channel on just this one thing if you can find the right niche for it. This is more of like your solopreneurs who would do this. But people are doing it. Like for example, you know, the audio overview, the podcast thing. Well, somebody like went ham and essentially took all the Epstein, like made a podcast called Epstein Files and then just made audio overviews of all the news coming out. Because it's a lot of information. There's all this new stuff coming out all the time about it. So it's a whole podcast and guess what, it's a top performing podcast. It's just. But it's, I think it's confident that it's just Notebook LM and somebody's daily taking whatever news happened that day, dumping it into Notebook lm generating the audio overview and the pro.

19:43

The problem is that is that you can, you can move so fast and not really have any qc. By qc, I mean like quality check.

20:36

Oh for sure.

20:46

The information that's coming out.

20:47

That's why businesses would should be very weary of doing this because again, there's still hallucination. Again, it's still gonna get, it's gonna read something, take it as fact and report on it. But it might like have been totally. The source that it got might have been totally fabricated, you know, so that's the scary part of where we're heading but that kind of stuff's happening and you might be able to build something out of it. Kind of interesting, especially if it's not so serious like the Epstein files. It's maybe something on the lighter side,

20:48

I will say, but. But to have this, to make videos, good or bad, the fact that you can do it without being charged, because everyone had seen every video creation AI tool, they're like, here's 40 free credits, and it takes 39 credits to make one video. And then they're like, to get Premium, you need 600 more credits. It's 19.99amonth.

21:17

Gosh, that's true. It is nice.

21:40

You can't even test if AI video is good because they have a. Even the free. Sometimes you're like trying to make the free demo and it. You mess up and then it uses all the credits and you can't go and you're like, oh, I'm not going to pay to figure it out. That's what my experience has been. So even for marketers who are like, well, let me test the waters, where's the AI video creation from prompt to video? You have to, like, Pay, you know, $300 to test them all. It's annoying.

21:43

Yeah. I don't know. I mean, Gemini is part of Workspace and you can use that. Sora is free. And if you pay the $200 a month, you get the longer versions, which is pretty good.

22:08

$200 a month, bro.

22:20

I know, but you get a lot for $200 a month because it's when you're paying for API credits where it really starts getting expensive.

22:22

Right.

22:28

That's where people are breaking the bank on, like, claudebot right now. Or not cloudbot. Openclaw. Right. Open claw. Because, like, they're letting these agentic things go out, go haywire and do all this stuff, and then they're getting bills for thousands of dollars and API credits. So I guess that was another thing about chat GPT 5.4 thinking, is it as good as Claude Opus? No, but it's a lot cheaper and it's pretty comparable now. So that's another thing. People are competing on price on a lot of these things. But it is nice that NotebookLM is free or cheap. If you're just paying for like a Google workspace license, it's worth playing around with. Go take a look at it. It's worth taking a look at. If you're listening to this Today, look at NotebookLM video overview.

22:30

And it's so simple and fast. I literally said, give me the Geography of Israel. I just, that was my prompt. It gave me all these resources and it was like, do you want me to import it? I said, yes, import it. And then I went straight to studio, clicked what I wanted and boom, within literally 60 seconds.

23:15

You watching the video now?

23:29

It's, it's rendering or loading.

23:30

Well, we'll see how good it is maybe by the end of this recording.

23:33

The crazy thing was this style. It said, what style do you want this video in? Do you want it to be like real or do you want it to be cartoon? Do you want it to be paper cutout visuals? I'm like, seriously? I'm like, well, I was thinking of my kids. So I'm like, oh, paper cut up. So we'll see.

23:36

That's cool. Well, moving on to the sponsor, which of course is High Level. And I'm, I'm predicting that High Level is going to take some serious market share from HubSpot over the next couple of years. And I think it's going to be because of pricing, not because Zen High Level is cheaper, but because HubSpot is going to have a hard time breaking away from this charging per user thing because that's how HubSpot made its money. Oh, you add users, you add cost. But when you start having more AI employees, how then do you charge for the cost? Well, you have to pay for usage, but High Level is already in that model. High Level is already in a usage model and the usage fees are reasonable and cheap so that you could scale up. That's going to be the new model. It's 100 bucks, you know, base. And then however many text messages or emails you send, there's a cost to that. It's actually cheaper than almost everybody else charging you a freaking arm and a leg for it. But there, there's going to be up. They're probably just going to mark up whatever they have to pay for the APIs, for all the AI stuff because they don't have their own AI models. So that's going to be the future. So I'm thinking High Level will be way ahead because it's just cheaper and they're able to like take on the new pricing model before everybody else. Remember when software used to be something you had to like actually buy a disc of and then install on your computer? I remember doing that with Adobe back in the day. I used to, I spent $600 on the Adobe suite, just the graphic design suite. And that was a lot for a student back in the day. And you install it on your computer and then you got to keep it. But it took a long time for Adobe to update to the point where they were more SaaS where you pay every single month. Luckily, they didn't have any serious well, luckily for them, they didn't have any serious competitors come and steal a bunch of market share on the SaaS model other than Canva. But sometimes when pricing models change and distribution changes, the whole industry can get disrupted because the old model has a hard time fixing their pricing fast enough to adjust because you can't. It's kind of like changing the price while also putting in the new thing can be difficult because you can kill yourself financially if you do it too fast. So I can. I assure you HubSpot's having a lot of conversations about this now and high levels like, like they're like we got this and they're just going to keep going. So fun thing. Check out high level. Go to dances.com high level worth just taking a free trial, clicking around, find out how powerful it is you. If you're already paying for something like HubSpot, you'll probably want to move. Moving on to everyday AI. Trav, what do you got for us? How have you been using AI lately? I have.

23:49

Not a positive thing. I thought using AI to find flights would be. I haven't looked for a flight in a while, so I'm like, man, let me just try and use AI to like look for a flight with these open source. Not open source, but open LLMs. And it had, it didn't do great for me. I was like a little frustrated. It wasn't giving me the kind of detail. I still went to Google flights. I still ended up booking a flight through credit card points. Like I just, it was frustrating. So I'm like, does anyone else out there know a better practical way of using AI to find flights? Is that out yet? Anyone know?

26:17

I think that's kind of like the open claw kind of thing. You know, Chat GPT has these things called skills or tools where you might be able to hook it up to a. Something that can do that. I don't know if Expedia or Kayak or one of those tools or Priceline have like a, a skill for chat GPT to go and do that for, but if they did, that would probably be the way to do it. Well, I don't really use disappointment.

26:56

I was just disappointed it didn't use more resources to help me find a better flight. It was just kind of like, yeah, you could just cut. You could just go to Kayak and try Google Flights and maybe you could find something on Southwest. I was like, no, I gave you the dates, the times, the everything. Like, go find me the best flight.

27:22

No, that's where it needs to actually open up into agent mode and do it. That's that cross tab, cross browser website kind of work.

27:42

I wondered.

27:50

You can't just go and read it and figure it out. You have to actually pull the levers and push the buttons where it's still weak on that. OpenClaw is not. OpenClaw will go and do it, but OpenClaw's got other problems. But again, OpenClaw got acquired by OpenAI. So you know, over the next three, four, five months it's coming. And the guy who created OpenClaw by the way, has said like, I want to be so useful my mom can use it. He's a little older than us too, so I'm like, he like that says

27:51

mom's probably is in her 80s.

28:17

Yeah. So this week I built and launched another website with replit. It's so nice to be able to like my boss came to me and was like, hey, can we revamp this website? It's like a one page. It was like a one page website on GoDaddy's website builder, which you know is good garbage. You're like, oh my gosh, yes, I will rebuild this website. I'm like, kind of like this and I throw together a website. He's like, yeah, but more I'm like, great, repl it. Change this like this. Yeah, publish. It's done so fast and now anytime I need to update it or add new pages to it, it's just like nothing that it's just a few prompts away and the website's done. Um, I'm now planning a more substantial website that's more of a directory style website to build replit in the next month. So I'll report on that soon. But like replit for just building websites, micro sites, event sites, anytime you need a small, just like throw up a website kind of a thing, repl it. It's just the easiest one. Lovable. Also easy. But where repl it's nice is if you want to build like if you want to start getting into more substantial features for that website. Like I'm going to build a directory with like a CMS where people can log in and change their directory thing. Not a hard application to build. But normally I would have had to find some kind of WordPress plugin for that because drag and drop builders can't do that. Now I'm just like, repl it. Give me a login area. Repl it. Make it. So they can only edit these pages. Here's the fields they need to be able to fill out. Oh, repl it. Can you make it so every time something changes here they get an email? Yeah, you could just keep adding onto it. Wow. It just works really well.

28:20

Wow. I just used Gemini's image builder to make me a, like a lives. Not a live scoreboard, but a scoreboard that we're doing a food drive. So we're trying to collect as much food for a food pantry. So trying to get raised 3,000 items. But I want to show people that are seeing our booth if we've raised the 3,000. So I made a digital scoreboard that's going to sit on like a 75 inch TV by our table. And I used, I just threw at the graphics that we've already created for the whole event. And I said, I need a scoreboard where one side says collected, one side says goal. And then days left and then I need it in scoreboard font. Here's the imagery. Just whatever. It took me a couple times and then boom, it made it. And I'm like, can I update this number for next week? Yep. Can you change this number to this? So when I count the items again, it's like,

29:54

man.

30:49

Because I was looking for like a live scoreboard that I could put on the tv. And then, you know, apps, they go, oh, you have to pay $19 a month and it's charged yearly, so it's 280. I'm like, okay, I, yeah, perfect.

30:50

One thing I did that I'm now recommending to everybody is everybody's working. Or you either have a company or you're working in a company and that company has specific things going on. Like you're a specific industry selling to a specific type of person and you're a specific st of growth, whether you're pre million, 10 million, 30 million. Like how what you have to do as a marketer at every single one of those stages and then whatever industry and whatever audience you're selling to, drastically different on like what works at those different stages. Experience usually guides you in that. But even now, like the company I'm working for now is selling. I'm selling essentially supplements. And I know what we're doing has been done a hundred times, a thousand times. It's a creator LED brand selling custom made supplements. Like they go into the formulation, R D gets involved and we like have someone manufacture it. But like we're doing all the R and D. So I'm like, other people have done this. This is a normal influencer health space play. Surely someone's like, someone's been through the things that we're going through right now for the stage of revenue that we're at. So you can take all that information and then go to chat GPT and say, hey, do a deep research. Or you can even put it on thinking now and be like, find me all the best podcast YouTube videos where people are talking about who have been, who have had these kinds of, or in this selling to this kind of audience or in this kind of industry and are at this level or just past this level up. And now they're talking about what they did. Go and find me. All those case studies and all those stories, bro, it made me, I, I just went and added all those YouTube videos to like a watch list. And now while I work out, I'm just watching them all. It's so good. It's like having a personal playlist and the ability to go and just like watch what other creators have done. Maybe different audiences, but like same kind of business model as a marketer to be able to be like, what did they do? Oh, they did that with influencer marketing. Dang, that's freaking smart. You know, we could do that same thing too. It's so good. Everybody should try it for whatever industry. Wow. I'd say focus less on the audience and focus more on the industry and the business model and the revenue size. Like what's like the next marker for you? Usually if you're at 1, you're trying to get to 10 from 10, you're trying to get to 50, but there's a lot of differences between 10 and 50. So like someone who's 50 or 100, what did they do when they were at your level? Those are the things I can go and find for you. Wow. All right.

31:06

That's a perfect example of deep research because it's so specific and granular and then you just need all the resources coagulated, you know, it's good.

33:32

We did not do a poll of the week this week, so we're to skip that and move on to the viral post. It's a good one though, man. So, viral post from Zach Yatagari. I hope I'm saying that right. Yadigari. I don't know. Do you know how to pronounce his last name?

33:46

Yeah, that sounded right. Yadagari.

34:03

Okay. 19 year old kid sells his AI app for $50 million. Oh no. They say they broke. He actually didn't disclose how much he sold it for. But let's see. Let me just read the post to you because it's an amazing story. It says Calai, which is his Apple, has been acquired by my fitness pal. Henry and I started Cal AI as 17 year old high school students with one mission, make calorie tracking easier. With AI in just 18 months we've helped millions of people lose millions of pounds and we broke 50 million in ARR, which is a annual run rate along the way. So they were. So they probably got acquired for a lot more than 50 million if they had 50 million in reoccurring revenue on a SaaS business. It says we are an incredible. We are at an incredible inflection point in history where anybody can build a product that can improve lives and make millions. As founders, we get a lot of praise. The truth is that this would not have been possible without our incredible 30 person team. We are so proud of the team. And then he goes on to thank the team and says Calai will continue as a separate app from my fitness pal. The combined team will share resources, continue to help people achieve their fitness goals. This post had about 6.6 million views, which is funny. Have you heard of this app before? Nope. I have. I found it in June and I bought it. Actually paid money for it. It wasn't a lot. It was like $3 or something. Maybe, maybe it was more. I think I've spent like somewhere between 50 and 80 dollars for the year for it because I tested it out, it had a seven day little trial and it was really good because if you ever tried calorie tracking, it's like freaking hard because you have to go find it in a database and it's never right. Yeah. Or you have to manually or what? Yeah, you just take a picture of it.

34:05

Of your food.

35:55

Yeah. And it could be like literally. It's good. I literally took a picture of Thanksgiving dinner and it's like, oh, this is Thanksgiving dinner. Here's everything on the plate. Here's the macros of it all. I mean it was like 1600 calories. But you know, it was really good at figuring it out. Now it's not always right because some things look like other things and it might not know the exact ingredients on it. But you know, on a general level it's fantastic. Pick a picture. Bam. Calories included. And of course you have your database still you can go to and you can enter it in manually still and do all that stuff. But the ability to just take a picture and then bam, go into calorie counting mode was really smooth. Now it's becoming a standard thing across all fitness apps. But he was kind of one of the first big ones to hit it hard. Got me. I was paying for it. Wow.

35:56

You calorie counting, bro?

36:44

Yeah, for a while I was trying successfully for like three, four months. I was trying to, I was just experimenting just for funsies. I'm like, can I get to under 1800 calories but still take. Still hit 160 grams of protein? That's what I was doing for a few months. I was trying the hermosi thing to see. I'm like, how, how hard is this? And it took a while for me to dial it in, but I was able to get it, so. But I wouldn't have been able to do it without an app like this, actually just taking pictures and counting and tracking and figuring it out and honing and trying to get it done. Now I'm less focused on calories and I focused on other things, but it was helpful. The cool thing about this is that AI is making it easier to build this stuff. And AI's coming up with whole new ways of approaching tried and true things. Like my fitness pal has been around for forever. Tried and true. Nope. Now a new one comes in and AI is making it possible to disrupt. This is happening across the board. It will happen in every industry where some AI solution will come in and just change it enough to make it wildly better. The one problem with MyFitnessPal was logging it and this one, it was very similar. It just came in and fixed that one thing with an AI solution and that's coming all over the place. Is that a marketing thing? No, but just be aware that every industry, including the one you're probably working in, is probably going to have some AI startups coming and disrupting it. Your job as a marketer is to watch for those and help your company prepare for it. Or like just go quit and work for those guys. Because it's going to work. This is going to happen over and over again. There's going to be a level, a wave of disruption of incumbents from the new tools that AI provides.

36:46

Well, boom

38:25

pod out.

38:29