Ep 656: Inside Gemini 3: What’s new and what it unlocks for your business with Google's Logan Kilpatrick

This is the Everyday AI show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business and everyday life.

The day is here. Not that anyone's been waiting for Gemini 3, but it is out, it is live, and it is very impressive. So on today's Fast and Furious show, we're going to be talking with Logan Kilpatrick and going over exactly what is new in Gemini 3. So let's skip the normal chit chat you guys know. If you want to more, go to our website, your everyday AI dot com. All right, so let's bring on Logan Kilpatrick to the show. Logan, thanks for joining again. But tell us real quick what's new inside Gemini 3.

Yeah, Jordan, thanks for. Thanks for having me. I think today's. Today's a special day, a very long time coming. You know, Gemini 3 is now available across our sort of suite of products, across our. The suite of sort of ecosystem products, which is really exciting. And the cool thing is it's state of the art across, you know, pretty much every dimension that we have the ability to measure right now. And I think folks talk a lot about benchmarks, but I think you actually really feel this progress as you use a bunch of these product experiences. And I think, like, one of those is AI Studio for vibe coding. The other one, the Gemini app, other products, you can really sort of see the nuance of how much better this model really is across a bunch of different tasks. So, yeah, I think we should build some stuff. I'd be happy to show some examples of what Gemini is capable of, but the sort of headline is we have the world's most capable model and it's available to everyone to try and build with across again, AI Studio, the Gemini app in our APIs, which is super exciting.

Yeah, I will ask you to show us under the hood, but real quick on that because you brought up benchmarks. And this is one thing I was kind of shocked by because even though Gemini 2.5, right. It's been out since, I think, March and on a lot of kind of different leaderboards, third party analytics, it was still kind of the top model. Yet you guys coming out with a brand new one, you know, explain a little bit about that, you know, mindset shift of going past just benchmarks.

Yeah, it's a great question. And I do think that it is cool that Gemini 2.5 Pro is really still holding its own. I mean, there was definitely things where sort of the ecosystem had surpassed the capabilities, and I think one of those is, like, agentic coding, for an example. And 2.5 Pro was. Was obviously really good on some stuff, but 3.0 really delivers on sort of like getting it back to the state of the art. And yeah, I think this going past the benchmarks era, I think is really important. And this for the, for us, this launch is super exciting beyond just having a really great model, because it's one of the first times that we've made the model on day one available across this, like, huge suite of different products. And really, it's about, like, we want people to use the model. We're building these models to benefit humanity and to, like, have people actually use them. So in order to make that true, you need to be able to use the model easily. So, yeah, there's a huge amount of, like, underappreciated product work, but also a lot of infrastructure work to enable this. Shipping models across Google on day one is not easy to do. We have an incredibly large user base and many different places and. Yeah. So shout out to our infrastructure teams who have been working sort of night and day for many, many weeks to make this launch happen. Um, it is, it is not easy.

Yeah.

And.

And that is one that stuck out to me. I'm like, oh, it's available via AI mode, like today.

Right.

Rolling out to search, which is huge. So, yeah, Logan, maybe we'll have you grab the screen and kind of show our audience around a little bit. So, like, so much that's new. The, the. The Google anti gravity, the, the generative interfaces. Right. But yeah, let's go ahead. We're going to bring on your screen here. Let's go ahead and grab it there. So, uh, walk us through, uh, especially for our podcast, have on your screen and what you're going to show us here.

Yeah, so I'm in, I'm in AI Studio. If you go to AI Studio, build. One of the things that we're showcasing with Gemini 3 is just how capable the model is for Vibe coding. And I think historically people have sort of, you know, Vibe coding has a mixed reputation depending on who you are and what space you're looking in. But I think with 3, with Gemini 3 Pro, you really start to feel that, like, if you've tried Vibe coding before and it sort of like, didn't work or, you know, you weren't super impressed, give it another try. With this model, the sort of, like, design decisions and the sort of complexity of what the model is able to build is really just. It feels like Magic. So when you come into, you'll get dropped into build mode. You can put in whenever your idea is. You can click on Feeling Lucky and sort of have us come up with an idea for you. You can also go to the gallery and in our new gallery we're sort of showcasing all these really great examples of Gemini 3 Pro in action. Sort of the design decisions, the models made from a, from a landing page perspective. All of these immersive and visual 3D worlds and games is crazy. Like you can go in and play this Sky Metropolis game and it'll actually just, you know, literally it's completely vibe coded. You can go in and play and the people sort of start showing up in the, in the world. It's crazy. So the model is able to. And again this was like not the intent of this was not to go and cherry pick a bunch of like, you know, the best possible example we could come up with. Like this was made by somebody on our team with like a very minimal amount of work and just like showcases like how approachable this model is at being able to build stuff. And again, if you want to, if you're an AI studio, you want to sort of keep building with Gemini 3, you like one of these sort of gallery applets, you can go to the left hand chat and just like add features. So I could say add in some additional building thing or whatever it is. So I think it's, it's really cool to just see the breadth of what the model is able to do. Like I think, I don't know if folks see this like voxel art and the bouncing ball examples and all these other things. Like really just my main push for everyone is be ambitious with what, with what you want from the model. Like this model, I feel like I'm always using AI and I'm like trying to come up with like the simplest thing for it to go and build for me. And I think the mindset shift with Gemini 3 Pro is like be as ambitious as you can possibly be because this model can really help you build stuff that you probably wouldn't have otherwise been able to come up with.

So yeah, you know what you're showing us now inside of AI Studio. This model is available everywhere. But you know, it seems like especially since 2.5, the usefulness and the utility of Google AI Studio, which I'm sure was maybe originally built for developers, it's now I think great for dev curious and even for non technical people. So maybe Logan, I think when people log in and they're exploring Gemini i3 Pro inside AI Studio. If they're a dev, the light bulbs go off and they're like, yes, like I'm good, I don't need to sleep for non devs. If they go into AI studio now using Gemini 3, where should they start? Right, like just your average everyday, non technical business leader. They go in here, they're like, oh my gosh, there's so many capabilities.

Yeah, it's a great question. I think vibe coding is the place to start right now. And, and part of this, and like a great sort of example of this actually in practice is um, I was yesterday, I was last night I was sort of looking through the benchmark results for Gemini 3 and I was like, I mean not that. Yeah. And I was like, oh, this is pretty boring. It's just like a big table of numbers. Um, so I took the, I took a screenshot of the results and I stuck it into AI Studio and I was like, build me an interactive experience that sort of brings these numbers to life and like makes it more, in some cases more digestible and makes it a little bit more interactive for me to see. So this literally took all the data and lets me sort of go through and I can like compare different models if I want to compare just 2, 5 and 3 and then I can switch to different categories and stuff like that. So it's a good example of like, you know, if you're, if you're sort of just at the beginning of this journey of figuring out what AI is capable of and you know your intent is not to like build a product, you can just take this mantra of Gemini 3 helping you bring anything to life. I feel so viscerally like just take a picture of some data or take a picture of art on your wall or whatever the random thing is, put it into, into AI Studio and ask the model to help bring that thing to life for you and you sort of, it's, it blows me away every time what the model is able to do. It's just, it's so impressive and I do think it delivers on that like bring anything to life mantra.

Yeah. And I'm glad that you chose this as the example because this is one that I love telling non technical people because we have access to so much data.

Right.

Even if you're not, you know, spending hours in Google Analytics or Search Council every day.

Right.

Everyone has access to so much data. You know what is, is maybe one new advancement specifically going from 2, 5 to 3 that helps with something like this, like what you're showing on the screen, you know, building this very, you know, bespoke visual dashboard with. With all of these metrics. What's that big jump that's really gonna make this even more useful?

Yeah, I think it's on the like, model capability side, which I think translates to what you see in the product experience. It's sort of twofold. It's the multimodal understanding that the model has. And obviously, like so much of the world that we all experience as humans is multimodal. And you see pictures and all these other things and we're digesting things in a lot of cases, like predominantly visually. And the model just has this like really nuanced understanding of visuals and can sort of reason and not just understand, but like reason deeply about the context that's there. Combine that with state of the art coding capabilities, and that's where you sort of get this magic. And that's why I think this, I think, you know, Vibe coding was like the. The word of the year through from one of the. One of the main dictionaries that that's out there. But I think actually Gemini 3 is like, further accelerates this because, you know, again, some of the. There were some examples that worked well before, but there's a lot of things you just couldn't do. And we actually have these in some of the blog posts where you sort of do this side by side. You can maybe take a technical paper that you don't understand about, you know, quantum mechanics or something like that. You can stick it in a studio and say, hey, I have no idea how this concept works. Can you take this technical paper and help bring it to life for me so that I can really understand the nuance of what's happening here and build interactive visualizations and it'll just digest all that content and like, build you that on demand experience, which feels so special.

All right, so a very quick word from our very relevant sponsor for today.

This podcast is sponsored by Google. Hey, folks, I'm Amar, product and design lead at Google DeepMind. We just launched a revamped vibe coding experience in AI Studio that lets you mix and match AI capabilities to turn your ideas into reality faster than ever. Just describe your app and Gemini will automatically wire up the right models and APIs for you. And if you need a spark hit. I'm feeling lucky. And we'll help you get started. Head to AI Studio, build to create your first app.

AI is so fast. Our ad from yesterday already has to be updated for today. But you know, Logan, I remember one of the first times that we talked on this show, I was kind of asking you like, hey, what's next? What's the next frontier? And I remember one thing you said is, you know, hey, proactive agents. So I know we don't have time to go into every single new release, but can you talk a little bit about what's new on the agentic side? Because, I mean, you have Gemini Agent, Anti Gravity, give us a quick rundown?

Yeah, no, that's a great question. So foundationally, Gemini 3, from an agentic perspective, the model is really, really good at tool calling. This is, this is one of the areas where I think like we had the most gap to close and I think 3.0 actually closes like much, if not all of that gap, which is really exciting. And I think when, when our, when, when the team sort of as we were developing Gemini 3, the intent was like, hey, let's make sure we have product experiences that really bring this like, tool calling capability to life. And I think you mentioned two of them and there's actually a bunch more as well. But the sort of two main ones are in the Gemini app. There's now an agent mode that's sort of rolling out to, I think it's ultra customers to begin with, and then sort of pushes the frontier of like what you would expect from a personal assistant that can sort of agentically do work on your behalf. So we can do things like triage your inbox and other stuff. And I think the reception so far from a bunch of our internal testing and external has been like, really, really positive, which is awesome to see. And then the other one is Google Anti Gravity, which is our new sort of agentic developer coding platform also built on top of Gemini 3.0. And the folks on the Anti Gravity team have really been trying to push the frontier of like, how do you change the way that building software works as a developer? And so you still have some of the familiar stuff like an IDE and such, but maybe, maybe you actually start with this sort of like agent first experience instead of going into an ID into like an actual code editor first. And I think there's something really magical there and what they've come up with. And yeah, it's been, it's been fun to see. It's also, it's like, you know, the internal reception inside of Google, which I think has a very, very high bar for like professional developer stuff is, has been really cool to see. So I think if you're, if you're sort of a professional developer, your company has A lot of developers, I think Anti Gravity could be a really, really great thing to explore for you.

Right. Because I know everyone has internally been dogfooding this for a while. I'm sure what's going to change the most, even for you and maybe even your colleagues, how you all work, because I think that's something our audience can learn from.

Yeah, I think there's this ambition piece which I think continues. And I feel like if I reflect back as somebody who was. Who sort of formally trained as an engineer and sort of was a software engineer, but no longer professionally write software is the good thing for the world, I sort of feel like AI helped me be more ambitious. And I feel like when the sort of, like, AI 1.0 era hit, like, I could really be like 10x more ambitious and I could really go and tackle, like, difficult engineering problems by myself. And I feel like it just. That continues to sort of, like, ratchet up. And I'm like, now I just need to. I continue to go through this mental reset of reminding myself I need to go and push more. I need to ask the model to do more. And one of my favorite things now, which I would encourage folks to do, is as I'm vibe coding, my favorite prompt now is and Gemini 3 hits it out of the park. Is I just, what I ask for whatever I'm trying to do in that prompt, and then I say add five new features. And it is this really like the model. It like, barely adds any latency and the model just, like crushes it. And there's, you know, a lot of stuff is useless, but I think, like, one in five. I'm like, this is a great. This is a great feature for our product. Why are we not doing this? And so it's been a ton of fun to, like, see that and have the model actually become, like, more of a thought partner in building and less of something that, like, I need to prescribe every detail in order to get any amount of value out of it, which I think is historically what you would have to do with a lot of these models.

All right, so today's show is going to be a fast one. But, Logan, as we wrap up, what would you say is the one most important take away from the Gemini 3 release for companies that are trying to take the absolute most that the model has to offer and grow their company?

Yeah, I think Gemini 3 can truly help you bring anything to life, whether it's an idea or an experience or a product or whatever it is, but it requires you to push it So I think push the model as far as you can, be as ambitious as you can. And we're doing our best to make sure that you have models that can be a partner and sort of show up and be ambitious with you. So hopefully folks, enjoy the model. If you have feedback, things that don't work well or just like broadly product feedback stuff, please send it to us. The team's working super hard to make sure we're scaling up, compute and all the other stuff with how much demand there is. So please reach out if you. If you need things from us.

Yeah, and Logan's not kidding. I have seen him and the team responding at like 1am and squashing bugs. So, Logan, thanks so much for taking time out of your day to join the Everyday AI show. We really appreciate it.

Thank you. Bye, Jordan. See ya. All right.

Thanks, y'. All. So, Logan, we will be wrapping up everything else in today's newsletter. Thank you for tuning in. Hope to see you back tomorrow and every day for more Everyday AI. Thanks, y'.

All.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic. Visit your everydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.