B (2:51)

You know, I'm kind of comparing those images. I will say, Gemini, your family looks more like you.

A (2:59)

Yes.

B (3:01)

Though the picture looks almost too on the nose.

A (3:05)

Right.

B (3:06)

It feels a little too perfect. Where the other one, like you went.

A (3:10)

To a studio and they proposed you perfectly. And she's like, now hold still. Everybody smile. Click.

B (3:16)

Right.

A (3:16)

It feels like that. Right? And which is what the prompt said to do, you know.

B (3:20)

But the other One, Gemini, is ChatGPT did do a good job.

A (3:23)

It's like, look, I know what you said, but. But I'm gonna. I'm gonna make it. I'm gonna give you what you want there. Is that about Gemini or Chat GPT? It's still a little bit better dealing with the abstract where Gemini is a little bit more literal. Like you have to know what you want and then describe it. And if you know how to describe what you want, it's going to give it to you. Chat GPT is going to. It takes some more intuitively figure out what you want.

B (3:49)

And it does. Pretty good at it, except it's missing.

A (3:52)

Your leg, which you can't tell really, until you pay really close attention.

B (3:57)

Your leg, your leg could be just propped up, you know, which would be weird because why would you do that? But anyways, yeah, it's hard.

A (4:08)

The cool thing is, and this was the thing that I walked away with. This is a photo. I had a friend in there too, that's been staying with us. That's why I don't have a family of six. And there's seven people in this image. That's seven faces, seven sets of clothes to replicate. None of the models couldn't come close to doing that six weeks ago. None of them. Until Nano Banana Pro came out. Nobody could deal with six faces, dude. It struggled with two. This is why I had. I haven't had you in the thumbnails because it struggled to get my likeness and your likeness right. It would cross. It would just screw up on One of them. So I just had me in the thumbnails, but now you're in every thumbnail.

B (4:46)

Just make me. Some random black guy was like, oh, yeah. I don't know when I.

A (4:52)

But it's good to know. Even though it still doesn't quite capture likeness, it's so much closer than what it was that I. I mean, I was obviously using thumbnails before, but now the thumbnails are even better. This is a great tool because I feel like now these image tools are more than just image tools. They're essentially full on design tools, 100%. It can lay out text, it can lay out color palettes. You are now the art director. AI is now the designer. But it's.

B (5:17)

But it's like a really good designer who's, like, hyped up on caffeine and just keeps submitting projects that are not. You're like, don't. No, no. Everything in the text has to be aligned, not just the top part. Yeah, it's like. It's like perfect. Oh, but. But then if you try to be like, can you go back and tweak it? It's like. Like this. And you're like, no, not.

A (5:43)

I know I did it last night. This is something I was gonna share in the everyday AI section. But one of the things that I love about these new image models is for interior design, you could take a picture of a wall, and I had my mom hold up a paint swatch on the wall that I got from Home Depot that's up. And I said, hey, take the paint swatch on the wall, because now if it's aligned on the wall, you kind of know how the lighting's gonna hit it. So it can take a pretty accurate representation of what that color looks like expanded on the whole wall. The first pass it did. It did beautifully. Except at the very bottom of the wall, it looked like a paintbrush had, like, just taken it down. You could see the paint strokes, but it was like white for the last foot of the wall. I was like, did you forget to finish the job? Like, you didn't see. What are you, 11 year old? Who did. Who thought you finished and didn't. That's what my. My kids do anyway. I'm like, hey, put away the dishes, and they leave a stack of bowls on the counter almost all the way there. But no, you didn't follow through all the way. That's AI sometimes.

B (6:37)

Well, I. But. But what do you do with that? You just tell it to try and redo it again and again. And again, sometimes I just said, hey.

A (6:45)

Take the paint down the rest of the way. And then I was like, hey, now put this mirror back on there. And I did a bunch of different things to see. I'm like, okay, now put some beadboard on this wall. What does that look like? And they all look very realistic. And now I can imagine what that wall looks like because I'm. I'm experimenting with some very bright colors. My wife makes my wife really nervous. But it is what it is. Seeing is believing. Right now, you could take a picture of anything in your house and be like, hey, make it more rustic. Hey, make it more this. And you could do it. As a marketer, I gotta imagine that's fun for, like, if I owned a store, like, if I had a storefront, you could take a picture of the storefront and be like, hey, create a window display that feels wintry and runs with our current campaign of everyday lights. Or something like that. You just throw it in there and it will dream up something. And then you could be like, huh, that looks good. We could build that.

B (7:29)

Wow, I love that. I've had to. When creating things like that, I still have to lean on it sometimes to write the prompt for me. I'll say, this is. This is what it looks like. This is what I want it to look like. Write the prompt that an AI, you know, generative fill can do. And it always does better when it writes the prompt versus me trying to figure out what language to use.

A (7:53)

You know, I imagine I haven't tried this yet, but I'm sad that I didn't get to try this before this episode. I'll have to revisit this because I'm going to be digging into this world a lot. I imagine if you took a screenshot of your website, you could be like, okay, like, how could I make this website more whatever you want it to be? And then it gives a text description of what it would change, and you say, like, great, do it. And then it can render a whole website, or not a whole website, but a snapshot of what it could look like. Then you could take that snapshot and run it through something like any of the AI coders, and it'll build it for you. That's what I think we can do. Or at least it'll get close. Probably can't create the images that it has on there. You'd have to figure out how to.

B (8:31)

Or if you already have a web developer, you could just pass that screenshot along and say, I want this. This is what my vision is.

A (8:37)

Yeah, but we don't marketers now have the ability to design. I wonder. I don't know. You're doing. You do design work with it. For me, it's hard to not know. Cause I came up as a graphic designer. I know the language, I know how to describe it. I know that. I read a whole book on graphic design history. Like, I know how to describe what I want. For the most part, I'm like, how has it been for you though? You. You pretty. You work your way around and get what you want out of it, right?

B (9:02)

Yeah. I mean, recently we have something going on where there's an entire campaign. It's a certain feel, it has a lot of imagery. Someone said, hey, we need a one sheet to send out to all of our volunteers and I would love it to be an image or a designed image, if possible. So it was in a big group chat with a bunch of leaders and someone was like, did you, you know, did you see what they said? They need all this information on a thing. It's the end of the week. How are we going to get this? And I was like, hold. So I just used ChatGPT. I was driving. I was literally driving with my kids after getting a car wash. And I was like, here's all the details. Put it in a bullet point format. I took all that and I said, okay, I need you to put this on a one sheet that I can text out as an image. The design quality. Like, here's what the design is. Think Snowy Village. Simple colors. Red, light blue and white. 2D. It's flat. Everything's flat. So I need you to create a prompt that will create this image with this text. First try. It made nearly exactly what my designer had already created in terms of the imagery for the campaign.

A (10:09)

Did you feed it an image ahead of time? She designed. No. This is where ChatGPT is good because it kind of. It's just more intuitive. It knows what you want.

B (10:19)

I'm like, it. What it put out. If you saw that her. Her design and what it spit out without any. Not seeing it at all. I was like, no way. But it was because I had the right language.

A (10:31)

Mm.

B (10:32)

I had snowy village 2D flat. Gave it. Gave it the colors. I even one time I gave it colors. I said, listen, our. The spectrum or range of colors that we use for our branding is. Is the shade of blue from light to medium to dark. So it's just like all the blues and some white. Create a 12 page training manual. Here's all the content. I need a 12 page training manual designed but here's all the written stuff. Just make it, bro. With that little bit of, you know, descriptor, descriptors for your like branding guide or whatever. It did a great job. I was like, oh, you have to give it a little bit detail. If you're just like, make me a training manual that's designed very well. It's not, you'll get frustrated because it, it doesn't fit what you're talking about. You need to give it.

A (11:22)

Somebody likes vague. Nobody likes vague direction.

B (11:25)

Right. Neither does chat.

A (11:27)

So if you listen to this about this too.

B (11:30)

About this too. Chat GBT did better. I, I was running same prompt Jim. I, this is what I do now when I create images. I go Gemini and Chad gbt. Same prompt, same prompt. Gemini could not, could not do it. We're usually, usually Gemini slays at that stuff.

A (11:46)

Yeah, there's a, I'm going to circle back around to which one I'm going to be using more in a second, but I have to move on to the second piece of news, which Google. It's funny because Google, I think Google was like holding this back. They're like, I'm going to hold this back. We're going to wait to see what ChatGPT releases and then we're going to do this the day after to try to steal some thunder. And they went literally with that phrase, thunder. Because the day I think the day like later in the day after they did the image thing, all the Google account just posted lightning bolts. Lightning bolts as a tease of what was coming out the next day. And sure enough, they released their flash model 3.3, which people are like, okay, like it's not a thinking model, it's not the best of. But what you don't really know is that Google's Flash 2.5 model is one of the most popular models for developers to call on the APIs because it's fast, it works well and it's cheap. It's cheap which when you're running lots of calls against the API, it adds up for your company. If you're an AI company or you're just calling a lot of these things. Three is so good that it's just a step behind all the frontier thinking models. And this is their fast cheap model, like 50 cents, a hundred tokens or something like, or 100k tokens or something like that. It's just like stupid cheap. So that's not going to hit marketers hard because a lot of us like I don't even use the Flash model in Gemini. I just keep it on the thinking model because their thinking model is pretty fast. So I'm like, I don't have to wait long. I just keep it on thinking with chat gbt. I'll swap back and forth between thinking and their, their, their fast model because sometimes I know their fast model will get the answer just fine. I don't need it to think about it. It's an easy answer. It knows, or I already know it knows. But this Flash model is going to be a really big deal. And I think it's going to be a big deal because this next year on this show, I'm going to be covering a lot more on vibe coding. I think it's something that all marketers need to start doing, and I think we need to start doing it sometime. It doesn't have to be January, but sometime this year. This needs to be a major school in your tool belt. You do not have to know how to code in order to do this, but you might have to learn some of the language of code in order to give it better instructions.

B (13:59)

Right?

A (13:59)

It's just part of the thing. Just like with the image maker, like, you have to learn how to speak the right way in order for you to. You actually have to know what you want. As far as a website or a little mini app, I'm actually, somebody else did this recently and I'm thinking about doing the same thing. But he's rebuilding his site and all of his client sites and now on Replit instead of using WordPress. And this is why the Gemini Flash thing's so important, because for little, these little mini apps and maybe use replit, maybe you don't maybe use your Gemini and just use something out of the box and. Or chatgpt, Codex or cloud code, whatever it is, I think a lot of us are going to start developing little mini apps. Instead of going and buying it, we're just going to build it ourselves. And I think it's possible. I'm like, I can imagine taking an RSS feed from my podcast and just automatically creating blogs from the podcast automatically on the website. But why not just stop with one? Why not make three blog posts for every podcast episode? Take the transcript and write it on the perspective of a cmo, a solopreneur. And like I have mult. Like I said at the beginning of the show, there's multiple different audiences for this podcast. What if I had an article customized for every single major person listening to this podcast? Like, we took the News. And then ChatGPT reoriented it for the Freelancer Solopreneur Fortune 500 Marketer, Mid Size Marketing Director. With the team of five. You know that starts to help, right? You can automate that stuff. You can't find a WordPress plugin to do that, but Replik can probably build it. But that's what's gonna come. That's gonna be awesome. But that 2.5 or that Flash 3 model is gonna be a big difference for all of us who want to start dabbling with vibe coding this next year. Which is why Google. So Google swung back. Google had the last punch and ChatGPT's last punch wasn't that strong, to be honest. I think their image model has gotten back up to Nano Banana. It's like it's right there, but where are we at now? It was a freaking dog fight. This year Google finally woke up, started actually taking things seriously. Larry Page, the founder of Google, went back into being full time and working on AI, kind of like for Amazon. Yeah, this is why Google's waking back up. Somebody told him, I think went to Larry Page, like, dude, why are you retiring when the most exciting thing in tech of the last century is happening right now? You're just going to sit on the sidelines and watch that bugged him and he got back in the game and we're now starting to see the fruit of it because they're, he's not, he's, he's not, he's holding people accountable and pushing the ship forward and they're shipping stuff, right? Good. So we'll see what they do.

B (16:42)

It's hard not to get sleepy and fat and happy when you're number one. I mean, Google's been number one in so many different areas for so long. And then Chat GPT and just came across and was like, oh, they might kill them just on search functionality, which is the number one way people are using Chat GPT. Right. They're just using it as a, they're using it as a, you know, hyper.

A (17:10)

Search engine, I would say confidently, and this is the title of the podcast, so it's no secret, but I think Google won the year as far as the stuff that came out with the punches. They, they, they put the punches they hit OpenAI with. I mean they were the first to come out with speaking video, way ahead of everybody else with vo. They're got one of the best models in the game. They're Flash. They finish strong with Nano Banana Pro Crushing. They have a bunch of side projects the way it's integrating into their ecosystem, like AI in their Google spreadsheets is now freaking amazing. If you haven't been in there and played around with calling AI and individual cells or even like I did a podcast, you can go find it. If you're listening to this, just go and find the one about Google sheets a couple of weeks ago. But you can do all this stuff with AI and spreadsheets now. If you're a big, heavy spreadsheet user, that's gonna be a game changer. I go in there every once in a while to build stuff, but it's nice that AI is gonna be in there with me. And the fact that you can have one ecosystem and collaborate and share gems, it doesn't have projects yet. I was expecting Google might add that at the end of the year. So I'm expecting early Q1. They'll probably add that. Cause that would be a game changer. Right.

B (18:18)

Can we. Let's talk about how. When they. The fact that they've won the year. When. When was it last where we were going, oh, cute, Gemini, cute. I mean, compared to ChatGPT was like the king king. And it was like, oh, we can. You could see Meta trying to catch up, Gemini trying to catch up. You hear Amazon is trying to create something in the background. Microsoft is trying to get something going. And we're like, oh, these cute little. And then all of a sudden, spring.

A (18:45)

Of this year, I remember specifically saying on this show, was it like April? I was like, who uses Gemini? Does anybody use Gemini? I think I've met all of one person that confidently is like, I love Gemini one. And I said this in spring of this year.

B (19:02)

Right. So I'm like, what? I mean, impressive that. I mean, Google's resources is unbelievable. But the fact that they were able to jump up this quickly, I mean, Grok Elon was putting a ton of resources in there, but then they just fell off. It was kind of like Chat, GBT and GROK were the closest players. Grok's still not out of the running.

A (19:25)

We were comparing, watching Grok carefully because it looked like they were going to catch up because of the rate of change they were seeing.

B (19:31)

Right.

A (19:32)

But they're struggling with the image model and the video model. They're like far behind. They caught up in text, but then they're struggling across other parts of the platform. And it's no longer just a game of how strong is your text game. It's also. It's all these other models of voice and which we don't talk nearly enough about voice is kind of plateaued for a year and a half now. Yeah. But it's going to be a bigger deal than text is. So there's like, there's the text, there's the voice, there's the image, there's the video, and then there's all the collaborative features inside. And then there's the one thing that I think separates chatgpt above them all. That to me still means Chat GPT is in the win. Because this one thing they do well makes it that. If I. If I had to choose between Chat GPT and Gemini right now, I'd pick Chat GPT if I can. If I wasn't allowed to touch anything else for 2026, I'm still committed to chat GPT.

B (20:26)

Wow. That's a big statement.

A (20:29)

Now I'm probably going to use both. I'm probably going to be like 70 chat GPT 30 Gemini. Because there are just some things that just Gemini crushes it on. So I'm going to be using both.

B (20:39)

Right.

A (20:40)

Especially since I already pay for it because I already pay for the Google workspace or whatever. So it's like, well, it's there. I might as well use it for the things it does well in. Right. But the reason why I think ChatGPT is still winning the game and I didn't realize how important this was until I was using both often, is its personalization.

B (20:57)

Right.

A (20:57)

Same Gemini doesn't even have a field in it that lets you give it custom instructions that gets applied to every chat.

B (21:06)

Funny. Why do they not?

A (21:07)

You're like, give me. Give me at least one field so that can. I can load it with a few basic things that I wanted to remember over and over again. So that memory, no instruction sets, no memory, it can't reference past chat.

B (21:21)

It says it, I can go chat forever and always remember this about me. And it goes, yeah. And you see it go, memory updated. God, that's a good feeling.

A (21:29)

And it's good. Like, at first it is good, and then they got better at it. And then with the most recent model, the 5.1 and the 5.2, I find it's actively pulling back things that we've talked about before into recent conversations without me even asking for it. And I think that's the thing that makes ChatGPT better. It doesn't even have to have the strongest benchmarks right now because it's more personalized to me. Right. And that freaking matters.

B (21:55)

Agreed.

A (21:55)

It's just like having a teammate in human relationships. The people you've been working well with. Over time you work better and faster with them because you've gelled more and everybody knows that intuitively when you've had a co worker you've been working hard alongside a while, you can get that stuff done better and faster with them versus new people. It's kind of like that with ChatGPT. It knows you, you can, you can gel, you can. It kind of picks up on what you're putting down faster and it can help you move along more quickly at the rate you need to go versus Gemini. You're having to re explain it every single time you're starting fresh. It's more powerful, but it's not as intuitive. It's like Android versus iPhone. Android phones are powerful, but I find myself second guessing, getting confused in between the interfaces, you know, where the iPhone is just more intuitive and easy to use and nicer. That's how, that's how OpenAI and Gemini feel to me.

B (22:46)

Good analysis. I agree.

A (22:50)

You're still, you're still chatgpt all the way, right?

B (22:53)

I mean, not all the way, but I, I'm like 90 10. I only go to Gemini when I'm looking for some image creation, design, need for work or I happen to use Gemini because I know Googling the thing about the thing I'm looking for will be better, you know, I mean, it's not close.

A (23:14)

Gemini's got some work to do here. I think they used to say content is king. I'm thinking now context is king for the next two years. And Google has the ability to pull on a much larger context because it has all your access to Google Drive and yeah, it has a bigger context window. It can actually pull in more information into a single inference call. But they're not quite tapping into it yet. Right now ChatGPT is dominating that part. So it makes Chat GPT better. So I'd say Chat Google definitely won the battle in 2025, but chat GPT is still ahead in the war, right? But their margins are getting smaller, that's for sure, than they were last year. So let's move on to the sponsor of the week. Usually I talk about high level. Today I want to talk about Tele Tella. It's a fantastic little software app that I use to record solo videos. If you've ever watched one of my videos on YouTube or on Spotify, where I go back and forth and I share my screen and has like that little vertical video, it's a tele video. I love it because I replaced loom with it like years ago because it gives you the ability to record your camera and your screen and then deal with which one you want to swap back and forth with afterwards. It also has all the AI tools to actually auto edit. I just be like find mistakes. It finds a little mistakes, cuts them and I just make sure and they're usually pretty accurate. It cuts down on like maybe 60% of my edit time. Of all the little mistakes I have to go and edit, it just like chops them all out and then I have to go and tweak them a little bit. But it just makes it easy. It even has an auto layout button that's kind of cool. But if you're in the market of trying to get off a loom or finding something better, check out tela. Go to dances.comtela and get a free trial. Test it out for yourself. But if you're looking for something where you can record awesome, professional, amazing solo videos that are like screen sharing your screen presentation, doing a demo, or you want to use it as the replacement for loom, you got to chella. It's much better than Ling these days. Let's move on to everyday AI. Moving on to you, bro. What have you been using AI to improve your life these days?

B (25:11)

Well, I already shared the one that I had one sheet sent out to my team and got that created. Can we just talk maybe not about what I accomplished, but how quickly I did it in about 15 minutes. So think about needing, oh, I need to send the text to the designer. The designer has to get it in and plug it in and maybe they'll hit it first try because they already got the design made. But that takes hours if not days.

A (25:35)

To get the back and forth time.

B (25:37)

So now it's just like so fast, which I so appreciate that time span. And my designer right now is overwhelmed with the amount of work she has to do right now. Over. It's like, I don't want to put another thing that I need quickly because who wants to? She's got projects she's working on. It's like, hey, stop everything you're doing. Give this to me in the next 15 minutes. No, that's awful. So the fact that I could help and just take something light, that's not going to mess up the continuity of the brand, the branding for said event, you know? Love that.

A (26:14)

I've been reading a book that's it's, it's pretty accessible, I get it. But there's like deeper meanings behind every chapter. It's a C.S. lewis's science fiction trilogy. And I'm reading the last book, and it's pretty heady. There's some stuff in it. I'm like, wait, why? What's going on? Like, I know what he said, but, like, what's actually going on? So every time I finish a chapter, I go to chat GPT and be like, okay, tell me what's actually going on in chapter three. And then it gives me, like, the cliff notes version of chapter three. And it's fun because I can actually have a dialogue with chat GBT about what C.S. lewis meant. C.S. lewis meant by this or that, and we can actually have a fun conversation. I get way more out of the book. It's the first time I've read a book. It's. It's science fiction, but obviously it's CS Lewis. The guy is very philosophical and has a lot of layered meanings into it and what he's trying to actually communicate. So it's been really fun to actually dive deeper into material as, like, a cliff notes type substitute. But now it's personalized for me because I can dive deeper into it and be like, well, what about this angle? What about that? We can have a conversation about every single chapter that way.

B (27:15)

You want to hear another one? I was looking through my different recent search histories. There's, like, I met this guy who met him a Mason, a freemason, and they have, like, a secret language. I was using chatgpt to, like, figure out how they talk to each other without knowing, But I won't share that on here because people get weird about that kind of stuff.

A (27:35)

You're like, chatgpt, help me decode the secret language of this cult.

B (27:39)

Yeah, that's literally what it was doing. I was like, okay, I don't. I don't know if I shared it last week. Did I share it last week where I talked about creating a one sheet for a game that I played for this Christmas for a staff Christmas party. And I just had Gemini make. I just was like, I need a Christmas game. I want to see who can list out the most Christmas songs. So just put ultimate Christmas song challenge. Design it Christmas themed. And I need listed one to a hundred with lines next to it for a space to write in. Chad GPT couldn't even get to 100. It got to, like, 19 and then hallucinated after that. Gemini got from 1 to 100, but then anywhere between 87 and 100. The numbers were a little bit, like, mixed up, but it got there. The person who won, she's like, one of the smartest people on our staff. She wrote 107 Christmas songs all by herself.

A (28:35)

Goodness.

B (28:36)

She's like, she remembers all the numbers to PI. You know, she can go like a hundred digits to PI kind of person. Yeah.

A (28:42)

Fantastic memory.

B (28:44)

Anyways, I like designing one sheets.

A (28:47)

I guess something I use AI all the time for is to find things I don't know how to describe or I can't remember the name of. Right. Google. Always a struggle. AI never a struggle. I was like, hey, I used to use the software that would like track people as they came into coffee shops. And then as soon as they put in their email to like ungate the WI fi, it would like be able to correlate that with every time they had ever been in that coffee shop. What was that software? Because I couldn't find it. I couldn't even remember what was called. They're like, oh, you mean this? It used to be called this, but they changed their name and rebanded back in 2020. Wow. That's why you're having a hard time finding it. But it's most for sure the software. This is exactly what they do. I'm like, huh? Everybody listening to this is like, what software can do that? Yes, yes. Many pieces of software can track ear. It's called the imac address on your phone and you can't change it. It's like having cookies without being able to remove the cookies. Your phone has it and every wi fi can track it. Sorry, it's just reality of what tech. If you're in California, they have privacy protections against that. But anywhere else, people can track how many times you've been in a store. It's just through WI fi and your phone being on. Wow. Good for marketing, bad for consumers. Oh, well. But using AI to find.

B (30:00)

Do you even know don't track if I've been in your store.

A (30:04)

All right, let's go to the poll of the week where I asked people who they thought was the winner, which AI company won in 2025. And of course, I had all the usual suspects of Anthropic, Claude, Google, Gemini, xaa Grok and Open A in chat. And the results were as follows. Anthropic scored a 16%. You're like, okay, Google Gemini was the winner, of course, at 48%. Xai Grok had 2%. Good for Grok.

B (30:30)

Wow, 2%.

A (30:32)

OpenAI ChatGPT 34%. So at least for these 166 votes, it was a pretty strong consensus that Google had won the year. Now I should follow up with this and be like, yeah, but who do you think's ahead and see where people are at with it. Like, which tool are you using? But I think ChatGPT is still people's primary tool. Kind of like what the people that are wrestling with it right now. I feel like a lot of them are still probably siding towards Chat gbt though more and more are starting to find they like Google better. I've ever just heard this morning that a friend of mine was saying, you know what? I think I underestimated Claude. I'm using Claude more now. It's just better. Like, okay, for me it's still chatgpt.

B (31:17)

People are gonna mad that I haven't been really mentioning Anthropic. I don't even mention them. But they got 16% of the vote, which is a big deal.

A (31:26)

People really like Claude. Yeah, but. And it's funny because I always get someone in the comments. It's like one or two people that are like, what about Perplexity? I'm like, there's always one or two of you. The Perplexity people, they never make the list because they don't really have their own AI. They're just an aggregate of all of them. I'm like, I don't. I use Perplexity every once in a while, but it's pretty rare. Claude's pretty good. Claude crushes it on code and writing. Hence is why I'm working on a book right now and I'm doing it all with Claude. I wow, subscription to Claude for a month and I'll use it and unsubscribe because I don't need three. I don't need three things to pay for. I'm already paying for Gemini and Chat gbt. I don't need three. And for this one use case where I'm writing a freaking book. Sure. Then it's. It's better to use the model that's the best at writing. All right, moving on to the viral post. Great post from Flavio Adamo.

B (32:21)

Oh yeah, this is great.

A (32:23)

But it was a side by side comparison of Chat GPT versus Gemini's Nano Banana. It was like two images, very detailed. This is the prompt that he gave to both image models just to do a side by side comparison. And it was a highly detailed aerial view of Times Square on New Year's Eve, packed with crowds, glowing billboards, confetti in the air, and a realistic reflections on wet asphalt at night. This is a challenging image to make. And when you look at them side by side, to me, one looks more realistic. Oh, 100%. The other one looks more idealistic.

B (32:58)

Yes.

A (32:58)

That's how I describe them.

B (32:59)

100%.

A (33:01)

How do you. How did they hit you?

B (33:04)

I actually didn't know which one was which. I should have known which one was which. But, yeah, one of them hits really well. The. The.

A (33:14)

The.

B (33:15)

Well, it's funny, as I'm looking at the images.

A (33:18)

Okay.

B (33:19)

I don't know if you know this, but obviously Gemini did a better job with the realism with the. The reflections. It just looks. It just looks more real.

A (33:30)

Yeah.

B (33:31)

Totally screwed up Times Square. That's not Times Square. That.

A (33:35)

It's just a different angle of Times Square.

B (33:38)

No, it is not just a different angle. That is not how Times Square looks. As many times I would know.

A (33:43)

I've never been there. I just figured it was looking at a different angle.

B (33:46)

It's completely distorted. It has two of that main pillar screens shifted. It's not an L shape. It is just one long corridor.

A (33:55)

And.

B (33:56)

And Gemini nailed it. That's exactly what Times Square looks like. Or chatgpt nailed it. In terms of what the shape. How does it feel? How does the street widen and narrow in between different places? Like, that's exactly what it looks like.

A (34:11)

Interesting.

B (34:12)

And Gemini destroyed what. What Times Square looks like.

A (34:16)

So it got the big picture wrong. Even though all the details look more realistic and real versus chatgpts on the details. It's kind of fuzzy and almost. Almost painterly looking. Even though it has this, like, blurry.

B (34:29)

Kind of like famous taxi New York painting where it's reflecting. Yep.

A (34:33)

Huh. And. But it looks like Times Square instantly. It's got Happy New Year on the billboard. You're looking straight down at the big pillar where the ball drops. It's got a warmer tone. It looks more celebratory. And the other one, if you saw it by itself, you wouldn't know that that's New Year's Eve. It could just be a big protest of some kind with a bunch of people on the street, because you can't really see the confetti, but in chat GPTs, you actually can see the confetti. Right. And it looks like New Year's Eve. It's got a big stage out in front. Yeah. So I thought it was a clear winner, but now after talking to you about it, I can see how both have their strengths. And it's probably more neck and neck than I thought. Yeah.

B (35:13)

At first I was like, oh, that one's clearly. And then I went, wait a minute. What am I looking at? This is not even Close. Anyways.

A (35:23)

So there you have it. Honestly, I'm probably going to be using more. I'm probably going to be using Chat GPT for images just because I enjoy using Chat GPT more. And every once in a while, I'll probably use Nano Banana Pro just because I like the liberties Chat GPT takes with making images.

B (35:41)

Only problem is, it takes forever.

A (35:43)

They sped it up, though.

B (35:45)

Did they?

A (35:46)

They sped it up and you can ask for an image and then ask for a second one while it's rendering so you can get three. I used to have to open up three tabs in ChatGPT and then repaste a prompt to be like, hey, make this. Hey, make this.

B (35:58)

How do you do that? How do you do that? How do you get it to.

A (36:01)

That's what I used to have to do. Now it even tells you if you're making an image because they just updated it this week. It'll be like, hey, do you want to make another image while you're waiting for this one?

B (36:08)

Oh, it does.

A (36:09)

It's still slower than Nano Banana, but at least it's four times faster than it was because it was painfully slower.

B (36:14)

Painful.

A (36:15)

Now it's faster. Not quite as fast as you'd like it to be, but you can at least move on in the conversation or generate more images all at once.

B (36:24)

It's funny because Gemini would make something in like 20 seconds. Gemini was taking. Or ChatGPT was taking five minutes sometimes.

A (36:35)

Yeah, painful.

B (36:36)

Painful. Anyways, man, fat lady hasn't sung yet, huh?

A (36:42)

No, we are still going. I think the best comment that I had on the post when I asked who's winning, they're like, hopefully nobody and everybody.

B (36:50)

Yes.

A (36:50)

Because the more these guys compete, the better everything, Everything gets for us.

B (36:54)

Right? Right.

A (36:55)

The more open AI and Chat GPT continue to stick to their pricing. Models don't raise prices because if they become. If anybody gets too far ahead, then they raise prices and nobody. All of us consumers who actually use this stuff don't win. So please compete it out, which is.

B (37:09)

Concerning when we'll reap the rewards. Netflix is buying Warner Brothers and HBO Max. I'm like, oh, no, this isn't good. No, it's the same thing. If all AI is one consolidation, it's.

A (37:21)

Like, oh, yeah, so may they all continue to win and us continue to win even more from their competition.