Silicon Valley Girl: AI, Tech and Career Growth - Google I/O 2026: A New Era of Work Just Started | Josh Woodward

Episode Date: May 22, 2026

Josh Woodward runs Google Labs, AI Studio, and the Gemini app — basically every AI product Google ships. Google just launched Gemini Spark, the AI agent that works for you 24/7 in the background.In this episode, Marina sits down with Josh during Google I/O to talk about:— Why we're moving from "doing" to "directing"— Why human judgment and taste are becoming more valuable, not less— Why voice is about to become the dominant way we interact with AI— How his small teams ship viral products in two weekends— What Google actually shipped at I/O 2026: Gemini Spark, 3.5 Flash, Ask YouTube, smart Gemini glasses with Gentle Monster, and Docs Live— Whether Google is really losing the AI raceTopics: Google Gemini, AI agents, future of work, voice AI, Notebook LM, AGI, Josh Woodward, Google Labs, Silicon ValleyMore from the Silicon Valley Girl: Newsletter:⁠⁠⁠⁠⁠⁠⁠⁠ https://siliconvalleygirl.beehiiv.com/subscribe?utm_source=spotify&utm_medium=video&utm_campaign=futureproof-sub&utm_content=guest-name&utm_term=⁠Josh WoodwardInstagram: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.instagram.com/siliconvalleygirl/ ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠YouTube: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://www.youtube.com/@SiliconValleyGirl⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠LinkedIn: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠linkedin.com/in/marinamogilko⁠⁠⁠⁠⁠

This spring, denim gets a softer, lighter update. Introducing Old Navy's drapey denim wide leg, a new fit that moves with you. It's everything you want denim to feel like for summer. Easy, breathable, and effortlessly cool. With a fit that creates natural movement and a wide leg that feels modern, not overwhelming. Plus, that signature, wait, for this price, moment. Old Navy's drapey denim wide leg. The amount you're going to be able to get done is unbelievable.

What used to take a few seconds is just going to be like, boom, boom, boom, boom. This is Josh Woodward, head of Gemini. Google just launched Gemini Spark, an AI agent that works for you 24-7 in the background. That was a project we started called Personal Intelligence. If you're a Gmail user, if you use Google Doc, sheets, slides, calendar, all that stuff just works. Helping us free up time in our calendar. So we can do things we want to do, either together with our family, hobbies and whatnot.

So if agents do a lot, agents do a lot of stuff instead of us. Is this officially a new era of work coming? Let's get into it. Josh, welcome to Silicon Valley Girl. I'm so excited to finally be chatting with you. I know, yeah. Congrats and all the success.

You've been with Google for almost 20 years? Well, yeah, 16 and a half, I think. Yeah, I started as an intern. Crazy. Okay, so you've seen a lot of Google IOs. Can you rank this one in terms of significance? Yeah. I think this is a big one for a couple of reasons.

One, it's the first time we're really starting to shift more into this agentic era. And I think that'll be a big shift for a lot of our products. You saw some of it yesterday. Oh, yeah. But there's a lot more cooking we have in Google Labs coming to. So that's one. I think the other, this Gemini Omni model is another one. It's a very big shift in how you start to think about any kind of input in and the output

that comes out of it. Yeah. So it should be fun. And then at the end, it was a bit, you know, Google I.O. was such a flood of news. But Tordaia and Demis actually mentioned a couple of our science products that we're working on. And I'm very excited about that and where that's going to go.

What was your most exciting update yesterday? Well, I can't wait for people to play with Gemini Spark. I think it's going to be fun. It's rolling out this week to testers and next week to Pro? Yeah, to the Google AI Ultra members in the US. And then we're going to scale it up from there. You know what was my favorite update? What?

When you were demoing this voice? feature like you're talking to your computer yeah but instead of just transcribing it actually went to your google drive went to your gmail found all the necessary information folded together and created this beautiful email with like a table inside and like this is nice this is the next step of voice yeah is that rolling out yeah that's soon that'll be in the coming weeks exactly we're playing with it right now it's so fun um and i think what's amazing to me the first time i remember it was two weekends ago we were hacking on this on the team and And they were like, do you think we could get this in Google I.

It's like it's a little tight to get something in. We snuck it in. So this feature is two weeks old? Yeah, this feature was two weekends ago. And it's amazing because you select all your files, either that could be in drive or on your desktop, and then you just talk to it. And it can go through and understand what's in the PDF, what's in the image, and then create it all. And then it also enhances the text.

Like I intentionally made a mistake on stage to like change the day. And it understands. understands, exactly. Yeah. So this thing is happening now that I wanted to touch upon. So a lot of people are using different tools, right? And we've seen something similar in like, I don't know, Kodak's Cloud. Yeah, yeah.

With these features rolling out, what would you tell to people who are thinking if they need a switch? Like, why would they switch to Gemini? Yeah, yeah. Well, I think the Gemini, you're going to get a few things. And if you're thinking about Gemini Spark in particular first, that is deeply integrated in the Google environment and ecosystem. You don't even need connectors. That's right.

I've been saying this for somebody who's like, please start involving the cloud. Yeah, so I think, you know, whether you're a Gmail user, use Google. Dot, sheets, slides, calendar, all that stuff just works. Yeah. And it's very good. Second thing, we actually have virtual machines that we can spin up in the background on Google Cloud. So I think where all this is going is you won't just have one or two tasks. have hundreds of tasks and you can imagine we can do that in parallel and just the amount you're

going to be able to get done is unbelievable so that's two right up the back the third i would say is we're really trying to push the limits on the types of things that spark's going to be able to create for you so what i showed on stage was a google doc or a google slide deck for our neighborhood block party but imagine this has got manna banana it can make images it's got omni it can make videos it's got lyria it can make songs. And that's the kind of stuff that no one else has, that sort of generative media suite. And so I think the ability of this thing over time is just going to be way more versatile. But we're starting out. It's very early. It's a beta. There will be a lot of things when we add NCP connections in a few weeks. We'll add agentic payments with Google Pay and wallet.

YouTube analytics by any chance. Oh, that's good feature request maybe. The thing is as a YouTuber. I can't really use Gemini to talk to my analytics because it's not connected. So I have to either download a CSV file or use computer agentic features to go and take screenshots. Yeah, yeah. Is there a killer feature that everyone should try once this rolls out so that they're convinced? Because I feel like a lot of people, they're hearing words, agentic, you know, agent running in a cloud. But like, as a mom, what does that mean for me? What can I try and do to be blown away by this? Yeah, yeah. Well, I think just to take that example, maybe outside of work as a mom, like you said, it is so good at any kind of just,

a lot of people who have been doing user research are like, I have all these digital chores I have to go do. So it might be like helping you remember things that you might have forgot. It's helped me remember a lot of deadlines for our kids. Like you've got to return that library before the end of the school year. I'm like, how many things am I supposed to do for both my kids? That's right. Yeah, yeah. So that's right. So anything around that, I would say also it's very good at I finding, and my wife and I use it this way, for helping us free up time in our calendar.

And so we can do things we want to do either together with our family, hobbies and whatnot. One of the things I would say for your work life, for a lot of people out there, ask it, what are the three meetings I should cancel this week? It's very good at that. Oh, it's a good prompt. Yes. So that's one that's really interesting. I'm using it more and more when I started, I was setting up reoccurring.

schedules. And so one of my favorite ones right now were huge NBA basketball fans for the Oklahoma City Thunder. And they're in the playoffs right now. And so I just get like all the news about the team. But it's inviting it to me as if it's like a diehard sports fan for the Thunder. So there's all kinds of different ways. But I think anything around like help me not forget something. Yeah. Help me free up time. Help me follow an interest I'm really passionate about. It's excellent and all that stuff. Yeah. And you also released Docs live yesterday when you can talk to your Google Doc.

And you know what it got me thinking? I would be slowly moving to voice first apps. Do you notice that internally? Because I started talking to my computer all the time. Now you're releasing this feature, which not only allows you to transcribe where you're saying, but have some smart thinking behind the scenes. Do you think the shift is happening now when we're using more voice? I think so.

We see it in our Gemini usage stats as well. Do you have any numbers? Well, not to share here, but I will say that like there are certain countries now where it's tipped and voices become just a dominant way. A lot of people are interacting. It's natural, right? It's natural. It's faster.

It's also now we're at a point where you can kind of ramble and the model can clean it up for you. We've also now got it where the model's hooked in. They can do all the tool calling. It can generate the images. So it's, and then the stuff I showed yesterday with the dialects is really fun. Oh, yeah. So you can have it sort of talk in any number of different ways.

So all that stuff is coming, you know, in the next, we're talking weeks here. This is not. Are you one of those media strategy people clicking through slides, scrolling spreadsheets? Yes? Good. This is for you. Because on Spotify, there's an audience that's different. Locked in.

Loyal, invested. They're called fans. Fans don't just listen to music. They feel seen by it. Like it belongs. to them. So when your brand shows up on Spotify, that's who you're talking to. And you're right next to artists like me, Lizzo. So, are you ready to talk to fans? Spotify advertising. You're among fans. Talk to me about the shift that's happening. So what I'm noticing as a user is that a lot of

companies were optimizing for software engineering. Like, let's use AI to speed up with software engineers. I'm seeing this huge shift to knowledge workers. Like, let's now help knowledge workers to experience the same magic that it's AI has done to software engineering. Is that the priority at Google right now? Yeah, there's big focus on it. And I think what's interesting is kind of taking the lessons we've learned over the last few months and quarters with coding and applying some of that to knowledge work. One of the early projects I worked on on all of this, which is still going a lot of Snowbook L.M.

Oh, yeah, my favorite. Whenever I'm prepping for like a scientific podcast, I go to Notebook LAM, upload all the articles. they send me, like, listen to a podcast about this phenomenon. And I think that was one of the first ones we realized, if you can make it really easy for people to assemble all that context, like all those papers in your case. Yeah.

And then either hit a button, it's a podcast, hit a button, it's a slide deck, it's a mind map, whatever you want. That was kind of our first, like, glimpse into where this is all going. And I think, like, where notebooks's going to go, and even you'll see it in Gemini and other stuff, is, like, that ability to just describe or talk about what you want. It's like, I need to understand this, go grab all these things, make this, this and this. And it's just going to happen.

And so I think for knowledge workers, it's very exciting. But also it is a different shift because you're almost going to a point where you're orchestrating much more about the outcome or the deliverable you want instead of the how to get there in some cases. Exactly. And so we on the team talk about as you're moving from doing to directing. And that's like a big shift. That's right. Yeah. We've having thought on the team, you know, in companies like Google, you would do like manager days or leadership days.

You do manager training. And that would only be for the people managers, like a certain percentage of your whole team. Now we're imagining. We may need that for everybody. Because you may be managing these different sort of like agents and others. Absolutely. And you're launching so many products that are helping people do that. Yeah, we're trying to.

But for the past three years, I was sad to hear that people were saying, like, Google is kind of losing this AI race because people are using like all the other tools. you know, Chad CPP was the first with the LLMs when it comes to search. Yeah. How do you see the situation changing now? Yeah, well, I mean, it's very dynamic. It's very fast moving. It is.

As you know, you have people on your show too. I think, first of all, it's fun. I think competition, a lot of us on the team, we like that. Because you get this, it sharpens you. You know what I mean? So that's one, I would say. The other thing I would say, though, is that there's a lot of things Google has, And in some cases, it's had for years.

And what's fun for someone like me who's been around here a while is like, how do you kind of put these together in new and interesting ways? And so, you know, we talked earlier about you don't have to use connectors. That was a project we started called personal intelligence. And it started from this idea, what if you could just tap one button and all your stuff could be connected? And I wouldn't mean that everybody wants to do it. We made it opt in.

So it's like you can choose. But for a lot of people, they're like, yes, I want that. Yeah. And so I feel like it's really fun because we're in this stage we're getting to reimagine not only kind of existing products, but combine them in new ways to make new products. Yeah. And that's just super fun to do. And also like fun fact, my dad started using AI like a few months ago when he came here who was visiting. Yeah. And he does like he teaches MBA. So he needed to do some like table on weights of different like experts, etc. Yeah. So as chat GPT, he was like, Okay, we asked Claude.

He needed some follow-up questions, but he was happy with the result. We asked Gemini. First, prompt and he was so happy. Oh, nice, great. I had to give him a phone that's powered by Gemini so you can run all these things. And like this is what makes me sad, right? It's still, like if we look at B2B usage and Tropic is leading it.

Yeah, yeah. What would you say will make people make this shift? Yeah. Well, I think there's a few things. I think one, the core model, a lot of the early adopters choose wherever that is. And you can even look back over the last like six months, how people have jumped from different models to different models. I know, we're jumping every week. Yeah, yeah, yeah.

So there's a lot of that. Now, that's happening kind of at the edge, the really early adopters. I think for most people in the world, they hear AI, and it means all kinds of things. For some it may be excitement, like more for you and me. For others, it may mean fear, uncertainty or what is this thing? And so I think how we try to think about it is like, almost the, if you almost think of a tagline, it would be like, look what Gemini. can do for you because it's not about look what AI can do for you it's about for your life yeah your problems your daily annoyances the things that make you grumpy how's it going to help and i think that's how

we're trying to think about all these features as like and it may be start very small like for your dad that one thing may be like whoa that leads to another thing and another thing so we're trying to build out these very almost bite-sized things that you'd be able to turn to your dad or i could turn to a friend and say, hey, look what this can do for you. And that's kind of how we're imagining. So over the years, you kind of learn sometimes you get lucky and something goes incredibly viral on one thing. But often it's this accumulation of small things that add up.

And cause people to be like, oh, that thing, I use it for this. Yeah, I use for this, is for that. Maybe I should. That's right. That's right. Do you think Japanese personality is important? Because we know AIs have their different personality. How would you describe Japanese personality?

Oh, that's a great question. I mean, hopefully it's helpful. Hopefully it's something that you can trust. It's like comes across. We see this in the user research. People think it's factual. It's accurate. It's precise. Not to agreeable. Yeah, that's right. We also hope that it's concise. It doesn't ramble on and on. It gets to the point. We also want it to be kind of warm, sort of friendly, but not too friendly. Do you know what I mean? And I think one of the things we've also been exploring is ways we can let people steer the personnel. There's maybe something that comes out of the box by default. But then some people do want it, in certain cases, we'll see this where they'll be like, be really harsh on my ideas, like poke holes and everything. And so we want the model to have some degree of steerability too. But I do think, yeah, for us, we don't think of it, though, as, you know, a friend or something you're trying to fall in love with or anything like that. It's a tool.

And it's something that, again, at the end of the day, hopefully it's something that's useful and helpful. Do you have any tips for someone who's just starting? using Gemini. And you mentioned, some people ask it to disagree with everything. You're like, is there anything that you have personally? Because for example, I have this MD file called Personal Constitution. It talks about all my principles. So I would have loaded to every AI that I'm using so it learns my principles.

Or my tone of voice or like a person's dossier because I use AI to write a lot of text. So I don't want it to come up with facts. Do you have your own personal productivity setup that you would recommend to everyone? Oh, that's interesting. This sounds good, by the way. Yeah, I love it. Because then AI really becomes a strategic partner versus just an answer machine. Okay, that's what I was the same. The biggest shift I feel like years ago I made was instead of just going to it one-off,

actually trying to build that context. Like you use MD files, it sounds like. How do you create this context? Like, how do you store it? Yeah. So, I mean, I have certain files like you've got now, like something like Gemini Spark. These are all now skills that are in there. I've turned on personal intelligence.

So it knows kind of my backstory of all my gym. email and calendar and drive. The other thing that I do is it's very good if you ask it. So one thing I try to do pretty regularly is I'll ask it, what are the things I'm doing that I should no longer be doing? Yeah. Or what are the patterns you see me using in Gemini that I should just not do? I can make better.

And so I almost use it as a mirror in that way to sort of reflect. That's interesting. I've also, I've got tons of notebooks, which now I'll sink across Gemini too, which are awesome. So I've got ones who's like everything I've ever written that I feel like is like my best writing. Not for the style necessarily, although it takes on some of that,

but also just the ideas. I read a lot of books. Just going back on that. So when you have all your newsletters inside a notebook and you ask Geminiq question, it can reference those newsletters because it's one whole ecosystem. That's right.

Do you have to specifically point it out, like go to my notebook or it knows? Yeah, usually today how it works is you can just hit the plus icon and just add that notebook to your prompt or if you go into the notebook itself, it takes all that context automatically. I wonder if I'm just thinking

from the user perspective, is it too much context because it has access to Gmail, you know, calendar, notebook. Is there, like, do you need to kind of narrow it down and say only take these things? Like, what would... Yeah, we're actually experimenting with this right now.

So in a product like Notebook LM, as you know, you can change which sources it uses to sort of shape the answer. But I don't know if the UI paradigms for all this are really have settled yet. Because I think that's one of the things we're finding too is these models love context, but then finding the right context. Finding the right context, not eating all the tokens while. Because I've been using Gmail since 2015, right?

Yeah, exactly. Exactly. Yeah. We all get too many emails. That's right. But I think what you'll find model over model and even this three point flash model, way better at these things. We put it through all these tests.

Can it retrieve the right stuff? Can it synthesize the right stuff? It's very good at this. And I think that'll just keep getting better, too. Okay. So you would recommend, so you put your best newsletters? Yeah, that's what I do. I put that in.

I have another one where I read a lot. So any of the notes or books, you know, I take lots of notes. I use read-wise for Kindle highlights. I sync all that. So you can almost imagine my own collection of like experts, their quotes, things that have shaped how I think. That's a whole other area. I put it as well.

Yeah. Okay. So I feel like we live in an area when everyone has to organize their personal context. Yeah. Because it's so important to launch your agents because they have to have a proper base on what they're acting. If you and I were having coffee after this episode, I'd certainly pull up my phone and say, look what I tried this week. It changed how my tea works.

This is what I do all the time. To share these kinds of things that don't fit into the podcast, I started my own newsletter. Every week, I read about AI tools, strategies, and experiments I'm running in my own business with real numbers, real results, templates that you can use, and also honest mistakes. If you want to be in the loop, the link is waiting for you in the description. Okay, let's talk about AGI. I've heard that word so many times yesterday on stage, and I'm like, oh, because last year, I remember Dem is talking with Sergey here at Google I.

Yeah, yeah. And they were, like, kind of predicting, like, maybe in five years. yesterday was like, we made another step towards AGI with this model. Yeah. What is AGI for me? Like, for a normal person, what does that mean? Yeah, and at this point it's been thrown around so much. It has lost, I think, a lot of its meaning in some ways.

To be honest with you, I don't think about it all the time all that much. There's a spectrum of people who talk about it every time they get on a stage. I'm probably more on the other side of it. I think at some level, it's probably like, is there an app or some sort of experience, software you can probably talk to that is able to kind of answer almost like superhuman questions for you. And I try to think always the best products make you feel something. And so to me, I think less about how many years away and things are, but more probably what

are the feelings or the experience you would get interacting with something that's that good? And it probably would feel something like, man, it just like saves me so much time. It's like mental release. Or even I imagine, I'm hoping it'll be fun. It'll be able to kind of bring up things and insights and connect dots that you wouldn't otherwise do. So I tend to think of it a little bit more that way, I guess, from like a product perspective, as opposed to some of these maybe bigger thoughts and timelines and things, specific definitions.

How do you think it's going to change our workflow? Because you're releasing so many features for workflows? Will the workflow exist at all if AI is capable of so many things? I happen to think it still will, actually, because I think there's still going to be a huge premium. And I'm actually sympathetic to the argument. Actually, things like human judgment, human taste, these things are actually going to become more valuable than they have been even in today or in the past. But then also AI can learn your taste. This is what I'm trying to do within my team right now.

Yeah, yeah. Teach AI my taste. So we're collecting all my feedback that I'm giving to my team. And like, let's teach AI that so it can help me make decisions. Yeah, yeah. I think that may still happen and probably will. we do things similar where you can almost simulate like a meeting before the meeting ever happens and you can almost like anticipate the feedback you'll get maybe that's what you're doing

on your team. The other thing I would say is I do think on the workflows point we are seeing like areas where AI today can get you maybe to like a really good first draft and maybe let's assume it'll get you to a second third or fourth draft it'll get better but there's still something about creating, and I'm finding in small teams of people, it's just so fun. And it's hard to imagine, I don't know, I wouldn't want to give that up. And you could imagine there will be like single person, giant companies or whatever, all the predictions. But to me, it's like life is about the memories and the people you're doing it with too. And so it feels like hopefully we're going to put a lot of power tools on the table and let small groups of people just create amazing stuff.

Yeah. What would you say to someone who's starting their career now and worry about AI. I mean, the advice I continue to try to even give myself is just every week be on the edge of these tools and try to figure out, like, build an intuition for what they're really good at and what they're not good at. And then the important thing is what's like almost possible because in another maybe rev or two of the models that may become possible. So I think that's probably it's a new skill set.

I think there's a lot of soft skills and maybe like each. you around some of this too is like how to deal with uncertainty, how to invent things, how to adapt to things. So there's a whole whole set of things even beyond just the core tool usage, I would say. Yeah. Can we learn some of your culture as well? Because you're rolling out so many products. Like, as you mentioned, this voice feature, you coded it like two weeks ago and it was on stage yesterday. Yeah.

How are you able to manage all of that, let alone like having four kids, I can't imagine that. Is there a principle? Is there a life hack? Oh, I wish there was a life hack. If you find it, let me know. I mean, I'd say it's a lot of things. I mean, part of it, I think just as an individual, I try to think about what are the things in life that I really want to prioritize.

And some of them are at work, but a lot of them are outside work. I mean, life's way more than just work, right? So I think that's just like a personal thing. I think at work, I think it's a lot about small teams with these tools. can do remarkable things. And so there's been a big shift, a lot of our Google Labs projects. They start with like five, six people maybe,

and you kind of are hunting for like what's an interesting problem to solve that will really make a difference in a certain set of people's lives. And then you kind of just let them cook. And we try to minimize all the reviews and all this stuff you could imagine. And I think that's, to me,

the best thing people say is like, oh, this is like, I'm empowered. We built this thing in two weekends. You know what I mean? To me, that's like we're hitting the mark. And then what the other part of this is you'll build a lot of stuff. Some of that maybe you get lucky. It becomes an AI studio or a notebook L.M or Google Flow. A lot of it, though, doesn't work. What's the percentage of the things that don't work. I wish I would know. I mean, what I will say is it usually takes us three, four, five times on an idea before we realize that's it or just ditch it. Yeah. What's the metric that you're looking for?

You know, it's funny, we don't look as much at, like, retention or dashboards. I always kind of tell the teams, and I try to go with them on a lot of these, you see it in people's eyes. When they use the thing, do their eyes light up? Or are they like, recoil? If you're an entrepreneur launching something, how do you try to sit down with a customer and just watch them do it? Go to the coffee shops, go to the student unions, go out in the real world. Really, I'm always trying to keep our teams as little as time possible. in conference rooms, thinking and theorizing about stuff.

It's like, take it out here, go talk to those two guys, see what they say. Yeah. And that's when you really know how far off you are, from your sort of lofty concepts to like reality. That's fascinating. And so, and then you just see people's faces light up and this is how you know you're going to continue working on it. Yeah, you're trying to develop an intuition of like, because the biggest thing is you make something no one wants. That's the biggest risk. And so you're trying to

So you're trying to find, like, did that strike a nerve with her or with him? And like, how you quantify that? Because you're in a large organization also. Exactly. Exactly. Go to people and say, oh, I saw their eyes light up. We're going to like a billion. Yeah, yeah.

Well, you take it step by step. You don't go from like zero to a billion. But I do think it is, it's very countercultural to a very data-driven place, like a Google or other big tech companies or even startups, right? You're looking for like, what's that? Yeah. And we'll get to that. But it's kind of that early stage.

One of the biggest things is doing like all big things start small. And we're very lucky at a place like Google. You saw Sundarck showed that slide with all these huge products with a billion users and things. But actually all those things originally started going to coffee shops or showing it to your friends and seeing like, is this something people are going to like? And that's how Notebook LM started. That's how FLOW started. That's how AI Studio started. So now, you know, they've got millions and millions of people.

But it all kind of starts, people forget that it starts in things. It starts very small. And then you'll get to what are our Dow's? What's our D7 retention? What's our AAR? All that stuff comes. But you've got to get out of the first phase, which is like, does anyone even care? About what you're working on?

What would you say to an entrepreneur who's working right now building these ideas five times? He tried, nothing's working. What's the mindset they should be adopting? Yeah, that's always the hardest part we have too is like you believe in your idea. You're on your third or fourth or fifth iteration. I think sometimes the cash just runs out, and that's when you know. In our case, sometimes what I've noticed is the core team, those like four, five, six people, they often know before I do if there's anything here.

And so sometimes you just have to face that moment. I also like to be, when we're doing this with our teams, like, so clear of like, all right, well, let's try it. We have one more shot. And everyone kind of gives it like one more shot. a lot like if you're playing on a sports team or something, you know, you kind of put it all out on the field and you kind of see if it works or not. Sometimes you're just more like, if I text take another step, then it's going to work and you're like, I don't know when to stop. And you know,

it's funny. Sometimes I know we've stopped things too early. Sometimes they go too long. I think that's the other thing I try to remind the teams as like it's a lot more art than science at that stage. It's a lot more gut and sort of like feel. Yeah. What's the next big AI trend you're predicting? That's good. Well, you're seeing maybe. Yeah, I mean, you hit on one. I'll think if there's others too, but I was spent a couple of minutes on voice. Oh, boy.

I think we're on the cusp of like a big shift to voice. So that one immediately comes to mind. The other one, I am not sure people have fully internalized, even at Google, is how fast these models are going to get and how quickly. So 3.5 flash is already way faster than what we had before. But even the demo we showed, Sundarck showed with the dinosaur. or the Dino game, 1500 tokens per second. This is coming.

This is not theoretical. Like we have this running at Google internally right now. And when you can stream that fast, it changes everything. We've seen that over time with search. Every time we make a feature faster, more people use it. So speeds a feature.

And I think that's one part of these models that's like just around the corner, that's gonna be incredible. Okay, so we're gonna be iterating on so many ideas, I guess, really fast. Well, practically, that's how going to happen is what used to take like send, wait a few seconds, is just going to be like boom, boom, boom, boom.

And that'll be amazing for all the image, video, audio stuff. It'll also be amazing for anything with agents or anything with multi steps because it'll just compress. Yeah. And so I think some research accelerating. That's the thing. Yeah. Then you apply that to something like our science tools or other things.

You got all this discovery that's just waiting to happen. So I'm very excited about that. I think that'll be fun. Yeah. My last question is, your daughters are five, six years old. Mine are five and six. Yeah, yeah. It's almost five.

Yeah. What do you think their future is going to look like? Oh, it's so hard to imagine right now. Is it going to be totally different? Or do you ever think it's just going to be the same things will have in a different way, but it's going to be the same life? Yeah, I mean, it's interesting because I think in some ways human nature is remarkably resilient, right? For its. Yeah.

Both like virtues and vice. Exactly. So some of that, I think. predict will be the same but I do think like our kind of almost human computer interaction feels like it will be way different and I don't know you talk to your parents you mentioned your dad earlier I think about my parents or my wife and then it's like you hear them talk about when they were in the workforce and suddenly like spreadsheets showed up and you know for us it's like

oh yeah of course we worked there's like no big deal but I was like I wonder if there'll be feelings like that where our like our girls will be like yeah of course you just talk to this thing and it does all this magic and it's like done in like 200 milliseconds yeah so i i do wonder i mean you sit in a way mode today and you think about i can't believe i'm in this car driving on the highway and so i think there'll be many of these shifts i'm very excited though for them you know uh to be able what it's going to do for what they can learn about what's the main principle you're trying to instill in I think for them, I'm trying to teach them a lot about like you can both create anything you can think of, but you can also, if something doesn't make sense to you,

just tell it to describe it and something you're really interested in. One of our girls is so into soccer right now. She can't wait for the World Cup. Oh, wow. So there's a lot of like lessons from even like kindergarten where we're just kind of translating it into like, oh, what if Ronaldo was doing it this way? You know what I mean? And like that's the kind of stuff that it just feels like it'll be like at their fingertips.

It's pretty fun to think about. Fascinating. Yeah, yeah. Thank you so much, Josh. Thank you. I'm looking forward to all of the updates from you because whenever you release something that means I'm using it in my workflow, because I love where you're building. Well, thank you and keep the feedback coming.

We'll follow up on your feedback request. Oh, thank you so much. Thank you. All right, thanks.