A (2:45)

What are they really leaning into? What are they focusing on? And so they're saying state of the art reasoning, deep multimodal understanding. We're going to talk about that. That's the ability to really consume audio and video in addition to text and get insights. Vibe coding. Vibe coding has been a theme all year and you've had Claude and ChatGPT really lean into coding. You've had Replit and Lovable, lots of players in that Vibe coding space. Gemini has now entered the arena, so to speak, and you've got much more powerful agentic capabilities. And they're making this all available in the Gemini app and on gemini.google.com and so we're off to the races here with this new model and I want. Kieran, your initial thoughts.

B (3:30)

I've been using it to kind of Vibe code some legion magnets for marketeers. I think that's one of the core use cases that our audience will be interested in and Vibe coding in general. So many months ago on this show there was like a little feature that Google had launched as part of a package of features and it was not that widely covered and it was the ability to dynamically create user interfaces. And what we posited then was at some point that was going to be visible in search and it would completely change the landscape of search because now you would have custom built user experiences for every search. Like you actually are building in some way lightweight apps to return the answers of search and drastically changes how you acquire traffic from sites like Google. And so AI mode is now, I think, default for most people. It's default for me. As part of this launch, I now get that as the default mode to search. And here I want to give you an example of this. And so this is their chief science officer. He wants to talk about how we're building on top of Gemini 3, which is the latest model. We're going to talk about how they're building AI mode search on Gemini 3. And because Gemini 3 is so good at coding, it's generated generative UI experiences. And this is what we talked about. So here's an example of how you might learn a complex topic like RNA polymer works, whatever, whatever.

A (4:53)

I don't even know how to read this, let alone understand.

B (4:55)

So you still have to choose the model here. They don't have the same issues as ChatGPT, but like this is basically like an ebook, right? In the past we do this and.

A (5:03)

Then we've got this long, detailed scientific journal. And they show me how this works. It's creating this composable ui.

B (5:11)

Right. It's starting to build an interactive UI to show you.

A (5:17)

This is mind blowing.

B (5:18)

Unbelievable, right? This is the future of search. When it has thynk. Right. You can see he asked for the thinking model. It will actually return lightweight apps, not blue links.

A (5:30)

Yes.

B (5:30)

And that is bananas because the lightweight app is an incredible way to keep you on Google. You never have to go to any website and be able to monetize that demand for Google. Right.

A (5:44)

And you can see where this is going, right? Like that composable sidebar could have ads. Even these animation could have ads. Yeah. Like, it's going to be awesome, I actually think, in the future for marketers, but we're in this weird transitionary period right now. I love, like the demo you're doing here, like the tables, the visualization.

B (6:02)

It's so cool.

A (6:03)

Like, there's been a couple experiments and we talked about this previously. There was an experiment called Learn about, which basically was this composable UI experiment. And they basically taken the best of that and put it into AI mode. Now that the model's good enough to actually do it fast enough and at the level of quality they want. I would say one other thing, Karen. I think we should take a victory lap. When AI coding started, we said there were going to be two big themes and this was like, I think like 18, 24 months ago.

B (6:32)

Right.

A (6:32)

We said, hey, we're going to move to this era of personal software where you can have this kind of disposable software that you use for use cases that you wouldn't previously use software for because it would have been too expensive. And then we said, you're going to have these composable interfaces in real time and it's going to have this very personalized experience, not just for search, but for anything. And we're starting to see that come through with exactly. This is the first, like real scale, like production version of that. And I think it is the perfect, like microscope into the future of the Internet.

B (7:10)

I agree. I think that again, was not something that was widely covered and now actually is front and center in how you search within the future. And I think this ability to get interactive apps is going to be a huge part of the Internet and Google's experience in terms of like how I feel about Gemini 3 in early kind of stages of playing around with it. I'll just show you some of the things I was doing very, very late last night. I think it's like really, really, really really good.

A (7:38)

Professional opinion. Really, really, really, really good.

B (7:40)

Yeah, like I can see me gravitating a lot of my usage actually over to.

A (7:44)

I'm thinking about making the switch from chat GPT as my primary.

B (7:48)

I feel like this is a huge step up and so this is like a really simple app again if I was thinking about this as a marketeer. So what this basically does is it will custom create you a marketing strategy. Now I'm going to just tell you I just did a single prompt. This is not a final app. Let me take a brand. Who do we want to do a quick one for? Okay, let's just do HubSpot actually public company and let's just do CAC is too high paid advertising CAC is too high and so what it will do is it will go off and analyze a bunch of websites and a bunch of data and then come back with a custom marketing plan now as a first version and given it a single prompt. Like even these kind of things are pretty cool right? Like even the way it's showing like a progress bar these small details. I know it's easy to understate these. This is kind of what you do in like good on boarding right? Like you tell the user hey don't worry it's going to come back. It's just going to take a certain amount of time so they don't like click off. I actually was pretty impressed and anything I built of the small details of how good it was in the small details. So this is what it outputs. Now are there improvements we can make to this? Yes. So it basically says optimize paid advertising campaigns by deeply integrating with HubSpot CRM data for hyper segment personalized and creative and landing pages. That's very true. Basically one of the ways you can do it is use HubSpot CRM to enrich the data platforms to improve the personalization of the ads landing page and everything else.

A (9:20)

Yes.

B (9:21)

And shift budget to complement stronger and chat focus on paid efforts on high intent lower funnel conversions and smart retargeting. And then it goes into this where it shows you some key strengths of HubSpot. It tells you a top opportunity and it pulls out all the sources how it's come up with this and then the critical gap challenge for high paid advertising tax suggests a potential gap in alignment Pay strategy with HubSpot's core inbound methodology, blah blah blah. And so then it has like these opportunities. So it would list that the strategic opportunities and actually prioritizes them by Impact and effort, which again is really good. I actually didn't tell it to do that. Goes into competition. So look at this. Pulls in traffic. SEO keywords reviews.

A (10:01)

This is really good.

B (10:02)

Social followers.

A (10:03)

When did you build this?

B (10:05)

11Pm last night. Then I built this one at. I built this one at quarter past 11 last night.

A (10:13)

That's true.

B (10:13)

So roadmap. Look at this. Build you, your marketing team, a roadmap to start instrumenting all those things now. I love this. Look at this. Yeah, it gives me the methodology and it's showing me all the sources.

A (10:27)

Dude, that's sick.

B (10:28)

Look at all the queries it ran.

A (10:29)

That's wild. Wild. It's taking like hours of searching and just compiling it.

B (10:34)

Oh my God. So I was going to play around with this and actually create a Little app and charge 499 a report for a little test. One thing I would say they should really do and this is why they're kind of still going to seed the market to like a lovable. What really someone like me would like to do is just easily deploy that app and integrate payments.

A (10:51)

Yes.

B (10:51)

And just do it all from within Google.

A (10:53)

They tell you to do it on replit.

B (10:54)

Oh do they?

A (10:55)

They have this replit integration. You probably missed it because if you go too fast they have like a lightweight like little toggle to repl it. It's not even a full time toggle, it's like a little feature flag basically.

B (11:05)

Wow.

A (11:06)

To go and do it in replit.

B (11:07)

As a replan investor.

A (11:12)

That'S the only investment anybody has that's doing well right now. But that's, that's another story for another day.

B (11:18)

I can show you the other one that I think is actually cooler, please. So when I started building was like in a customer intelligence platform at 11pm last night. But I was like, okay, what's a cool product for marketeers? One of the ones would be I can plug in who I'm looking for and it would plot out where those people are. This is really for startup founders to get my first a hundred to a thousand customers. So let me give you a quick example I was playing around with. Okay, I'm basically a founder and I've built a app that helps men who want to start practicing meditation, Right?

A (11:49)

Yep.

B (11:49)

So my wellness founder is looking for men who want to start meditation and then I'll click this. And again, this is. I thought this was like really cool. So this is a basically kind of trying to do customer intelligence for founders to try to get to a thousand customers. And so what it's basically doing is giving you an executive summary of that audience. And so basically say, hey, this is like men often busy under stress, curious about meditation, but face internal, external barriers. And it goes in and it describes the audience really well.

A (12:19)

Yeah, right.

B (12:19)

It gives me the lcp. Then it finds all of their mentioned pain points and it tells me where they're talking about those pain points. Literally.

A (12:28)

Look at, like dope, man.

B (12:30)

Look at this. My mind wanders a lot and I can't remain mindful. It's hard to stay focused while meditating. I struggle with consistency and make meditation a diet habit. I don't know where to start or worry about doing meditation. Right. These are all real quotes.

A (12:41)

It's never been easier to build a company and to start a company.

B (12:44)

And it shows me where it's extracted them from. So I can just like build the app and go in here and say, I've solved these problems.

A (12:49)

Yeah. Then go to take the app into those channels and get feedback.

B (12:52)

And look, it's got buying signals. It's building me a whole intense platform and I can just start a new research. This one I'm going to build out as well. And I'm going to charge 499 for report.

A (13:04)

You definitely should. I think lots of people would pay that. It's very dope.

B (13:07)

Pay me a cup of coffee and get intelligence on tap. I'm a marketer. That's a strap line already. I just thought of that.

A (13:14)

Oh, man. Look, you're in the zone. You're in the zone. He is Vibe coding and marketing in all the same time. We'll be right back to the show, but first we talk all the time about Google Gemini. Google Gemini is under the radar, one of the most valuable tools in a marketer's toolbox today. And that is exactly why we at HubSpot dropped our ultimate Google Gemini guide. And this is going to be packed with exactly what marketers need to know to transform their research, transform their marketing strategy, scale their content production, everything you're going to want to need to know to do marketing better with this guide. And there's. There's no beginner basics. These are detailed workflows. We're giving you the prompts, we're giving you everything you need to go and use Gemini to get the most out of that for your marketing. And if this is something that you're prioritizing, you're not going to want to miss this guide. You got to check it out. You can click the link in the description below. Now back to today's Show. Okay, so I think that was a great breakdown of vibe coding. Karen, I want to share with everybody. I think the most under the radar aspect of Gemini. Are you ready for this? I don't even think you've gotten here yet. So one of the things that the Gemini account posted about was they've got a free plan for college students. But lost in there is this first sentence. Kieran, with advanced reasoning and the ability to analyze hour long videos, Gemini is perfect for students. Did you know that it could analyze up to hour long videos?

B (14:46)

What? So you could basically give it a YouTube link or like a video file.

A (14:50)

Oh, wait a second, Kieran. I'm going to show you. Let's go. We have this awesome show here, everybody. And by the way, if you haven't hit like and subscribe, please do that now. So this is the show we did with Sabrina Romanov. Awesome episode. If you've not watched that on Content Automation, go check that out. So I went there and I was like, oh, I would like to see if it could be a good coach for you and I, Kieran, and tell us how to be better hosts of this podcast.

B (15:14)

Oh, wow.

A (15:15)

So I took this link and I went over and I said, hey, can you analyze this video? Because I didn't know what it could do. And it said, sure, sure, I can analyze video.

B (15:26)

Great.

A (15:26)

And I said, can you download this YouTube link and analyze the video and give us feedback as podcast hosts? And all I did was that prompt and paste the link in and it gave us a detailed breakdown of our hosting.

B (15:40)

That is awesome.

A (15:41)

What you did brilliantly. Keep doing this. The show don't just tell format, acting as the audience surrogate, and it called out specific wins. Do you see this? When Sabrina showed the NAN workflow, which looked like a complex spaghetti chart to a novice, you immediately paused to ask, is this harder to learn than lovable? And what would you recommend for a beginner?

B (16:03)

Oh my God, that is so good. I'm gonna create a AI coach for podcasters in charge. 4.99 Report like. Like I'm the biggest show.

A (16:16)

Very much literally blew my brain wide open. I got more. It basically is a great coach and it gave us production notes. It even embedded the video down below. So like, even the UI is really good. And then I was like, hey, what are three things we should improve? Use the sportscaster technique. The issue this episode was highly visual. While great for YouTube, you likely alienated your audio only listeners on Spotify and Apple who had no idea what an NAN node look like. The fix you need to narrate the action more explicitly. Don't assume the audience sees what you see. Come the hell on.

B (16:54)

That is really good. We should do that. That's like a great tip.

A (16:58)

We're going to do it. It's great. But like, hunt for the breaking point. Like, these are like, these are not fluffy piece of advices, right? Like, demo went perfectly. The app built in five minutes. The automation ran smoothly. This creates a happy path bias where the audience thinks it's magic to fix. You need to be the skeptic, force the guest to show you where the tech fails.

B (17:19)

Can you ask it to give you the perfect 30 second clip from the show that we can use in a social media post?

A (17:27)

For what kind? Let's be specific.

B (17:28)

Let's say for LinkedIn. So it has to be highly engaging and educational or say highly engaged and educational or a spicy take and see what it does. And if it does this really well, then I'm retiring. But I have so many ideas about this. I hadn't thought about how good it is at analyzing video.

A (17:47)

Oh, dude, just, just, just freaking wait, man, look, what I also had it do is I had to build me a prompt to be an ad analyzer so that like you're a creative director giving an analysis of any ad. Oh. And so basically a complete prompt of hook analysis, narrative script structure, psychological triggers, scoring system, everything. Okay. And then I just gave it a random ad. In this case, I gave it this Ed Sheeran Heinz ad, which is just Ed Sheeran talking about ketchup. Okay.

B (18:21)

I mean, when you're a billionaire, you do not need to do ads for tomato sauce.

A (18:26)

But anyway, first of all, it's ketchup, not tomato sauce. Okay, I know, but look at this. It gave us a full hook. And now, like, this is really good stuff. Pacing, conversational and slow. Invites the viewer to lean in and listen to the story. This is how you know it's not looking at a transcript also, by the way. Yeah, yeah. And it did this while we were on the show. It is so fast.

B (18:48)

I know this could be a really dumb question. Does it know it knows the visuals.

A (18:52)

Let me do this. What was the most compelling frame of this?

B (18:57)

Yeah, I want to know how well it knows the visual.

A (18:59)

And let's see what it does here. I'm going to go back to our other.

B (19:02)

Okay. All right.

A (19:03)

Here are two perfect 30 second clips tailored for LinkedIn. One is a spicy take. One is pure education. And so it doesn't edit the clip out for you, Kieran, but it tells you what it is. And the start and end points, it.

B (19:14)

Validates people that are overwhelming. So it gives you the start and it gives you the end. So my advice for beginners is to be completely honest is to ignore AI and automation. That is actually really good. Holy smokes. That is actually perfect.

A (19:26)

All right, let's go to the next one. The function I think about is like, skill times clarity equals the leverage I get out of AI.

B (19:34)

Oh, my gosh.

A (19:35)

All right.

B (19:35)

My mind is these are the best use cases we've done in a while.

A (19:38)

I think we have about like 10 Gemini three shows ahead of us, folks.

B (19:43)

I am going to create an app and I'm going to charge 4.99 per video clips.

A (19:48)

Okay. Oh, back to Ed. Based on the created analysis and narrative of the ad, the most compelling frame occurs at 52nd mark. The hero reveal of the ketchup bottle. So hold on, I'm going over here because this is a minute and 19 second video. It's saying this is the most compelling frame. That's insane that it can do that, right?

B (20:11)

That isn't so. It actually does understand the visuals.

A (20:14)

It understands them to the level where they can pick out the frame.

B (20:18)

Why does it say that's the most compelling frame?

A (20:19)

Okay, here's the breakdown of why this frame wins. It describes the shot. The shot where Ed Sheeran reaches into his backpack and pulls out a standard plastic bottle of Heinz ketchup.

B (20:29)

Oh, my God.

A (20:29)

Holding it up against the backdrop of a dimly lit, ultra expensive fine dining restaurant. Why this is the most compelling color theory.

B (20:37)

Oh, my gosh.

A (20:38)

This is bright saturated Heinz red cuts through the posh color palette of the warm golds, browns and blacks texture class. The cheap plastic industrial design of the bottle sits in direct opposition to the crystal chandeliers, white tablecloth, and fancy vegetables.

B (20:55)

Hitting the fancy food is a close second. Holy.

A (20:58)

We can go look at that real quick. So then it's saying the 12 second mark. It was basically just talking about fancy food.

B (21:06)

So good.

A (21:07)

And did I blow your mind?

B (21:08)

That is an incredible use case. I'm kind of blowing my mind on how good those clips were.

A (21:13)

Imagine having it go and find the best hundred ads of the year. Find the best frame in each of those ads and doing an analysis across all of them to understand why those frames work and basically then have it use that formula to then build a net new ad in VO3.

B (21:30)

Yeah, yeah. We should vibe build an ad live because we can use VO3. So all we need to do is do what you said, then build a prompt Gemini Gem to be our VO3 prompt and then have that create 8 second clips. Create the ad and then create the ad.

A (21:47)

I want to give the folks at Google a little round of apps.

B (21:49)

So good.

A (21:49)

This is a step function. Better, I will say specifically from like a marketing and business growth perspective than the other models right now.

B (21:56)

Yeah.

A (21:57)

There's a new set of use cases that are unlocked and imagine integrating these into tools. Like obviously we're biased at HubSpot. Like it's going to be fricking awesome. Awesome, awesome.

B (22:08)

I have so many $4.99 miniature apps I'm going to go and build this week. Be prepared.

A (22:14)

I love that you just decided you're just going to five code apps and sell them for the price.

B (22:19)

You and I were whatsapping about something. We couldn't. We couldn't sleep when we were trying to solve a problem. And then the next night I couldn't sleep and I was thinking about building 499 apps and I was like, okay, I'm going to build a whole suite of 499 apps because it's going to be Fiverr. It's going to be a penny cheaper than Fiverr.

A (22:36)

Oh, Kieran. And it's four nine nine apps. Okay. We are going to be back with several more Gemini shows. If you have specific ideas of what you want us to COVID leave them in the comments below. Or if you have your favorite Gemini 3 use case, tell us what that is in the comments below. Shout out to the folks at Google. This has been awesome. We are going to be back with lots more. But I hope you found this to be a super helpful episode and get you into Gemini 3. See you all on the next episode. Sam.