The Man Behind Every Google AI Tool Showed Me What's Coming Next

Most people are not even using 1% of

what Google has built.

While the whole world has been focused

on a damn chatbot, Google has been

building something much much larger. I'm

talking about a tool where you can enter

100 pages of notes and it converts all

of that into a cinematic video in just

60 seconds. A tool where you can just

enter one sentence and can convert the

whole thing into a fully working app in

just a few minutes. And there's one tool

where you can just enter your website's

URL and in one click, it will generate

the whole week's marketing on the go.

Now, here is what is different about

this video. You know that I don't

usually show up like this. It's usually

my AI clone that creates all the content

for me. Of course, my thoughts are real,

but for this video, it had to be

different. So, I flew to meet someone

very interesting. Meet Josh Woodward. He

is the person behind Google Gemini. He's

also the person who runs the Google

lamps team that built Notebook LM. And

of course, he's also the person behind

Google AI Studio. In fact, Time magazine

just rated him as one of the top 100

most influential AI figures in the

world. We didn't just talk about it. I

asked him to show me things live in

front of a camera. And if you're

wondering, this is some basic tutorial.

Guess what? It's not. Josh ended up

speaking about something that is still

being built and most people in the world

haven't heard about it. And you know

what? You're going to hear all about it.

So, let's get started.

>> Okay, Josh. First of all, thank you very

much for doing this. Very excited. Good

to see you in flesh

>> because I've been seeing you on Twitter.

>> Right. No,

>> it's great to meet you in person.

>> Absolutely. So, first question that I

have is you have a very interesting job,

right? You run the Gemini app, the

Gemini team, and you also run the labs.

>> Right. Right. Right. My question is what

keeps you awake? Is that the Gemini team

that keeps you awake or the labs?

Because you guys have been shipping at a

velocity which is unprecedented. I don't

have a different word for this.

>> Yes. Yes. Yes. Uh I actually sleep

pretty well right now. Uh I mean both

are extremely fun. They're fullon for

sure.

>> Um but no, they they don't keep me up.

There's uh there is a lot to ship

though. Um I feel like it's one of the

best parts of the job right now. every

week there's some new capability and

it's how do you kind of turn it into a

product or a feature and get it out. So

>> look one thing that is happening is

every time you ship a new labs product

every time you ship a new feature inside

of Gemini

>> there is one narrative that propels

across the internet and that is oh PLI

is going to replace marketers oh LIA is

going to replace a lot of music people

right anti-gravity is going to replace a

lot of engineers so there is this

narrative built across AI that's going

to replace a lot of people and

everything around it so my big question

that I have for you is to all those

people who are worried what is one line

of advice that you have for everyone

>> and what do you think

>> people should do to win in the AI first

world? Yeah. Well, it's interesting

because how we think of a lot of these

tools, they are tools actually. And so a

lot of what we do, one of our values on

both the teams as we talk about how we

co-create. So whether if it's Pomelli,

we actually are working with marketing

and creative directors on that product.

Notebook LM that came from working with

authors. Um anti-gravity flow for

filmmaker. You can kind of go down the

list, right? Uh so I think that would be

how we'd think about it at a high level

is more how do we get this tool in the

hands of either a professional who

generally becomes way more prolific

>> or these tools are incredible about

really lowering the barriers where

people can sort of do things they never

could before.

>> Yeah.

>> So you mentioned LIA and music. I'm

terrible at music but I can kind of make

stuff with LI. Do you know what I mean?

It's not going to win any awards or

anything but it's fun for my friends. I

would be surprised though.

>> No no no at all. I mean, the other thing

I would say though is um with a lot of

this stuff, my best advice, you said

kind of one line of advice, you just got

to try it

>> um all the time and see kind of where

that frontier is because that frontier

keeps moving out and it's jagged, right?

And so that would be the way I would say

is try these tools, see if they fit in

your workflow today. In some cases, they

might, in some cases, they won't yet. U

but that would be the main main advice.

>> So basically playing with AI tools,

orchestrating them is the

>> Yeah, I think that's answer. I think the

word you just said is very good, which

is kind of play. Yeah.

>> And I think that's how we try to

approach a lot of things, whether it's

on Gemini or Labs or AI Studio, is kind

of this mindset of play.

>> And you're trying to figure out like

what is this thing good for? You know

what I mean? In some cases, you're like,

"No, this is bad." In some cas,

you and you're trying to figure out kind

of in between.

>> No, that makes a lot of sense. In fact,

one other question that comes quite a

lot for people is every time they think

about, hey, you have to play with these

AI tools and you have to be on top of

it. And you know, there's a lot of VIP

coding that people have been doing, lot

of agent building like Opel has been

doing that really well. Today, Gems has

the power of building an agent as well.

I think Google AI studio was initially

built for developers. Now, I use that

quite a lot.

>> Yeah.

>> Right. To wipe code using build and then

anti-gravity dropped. So there is this

also sentiment especially for people who

can't code

>> saying that hey if I want to win the AI

race if I want to win in this AI first

world I need to be technical I need to

be a developer what is your sense on

that how can someone who's not technical

as well have the fair shot at winning

>> versus technical people as well

>> yeah it's interesting what we're seeing

right now folks that are maybe more

professionally trained as a software

engineer they are wielding you know 5 10

15 agents at a time doing incredible

things but I think the other part that's

happening at the exact same time is sort

of the future of software development

really democratizing in really

interesting ways and I mean you talked

about AI studio one of the most common

things happening right now inside Google

people take a screenshot of their

product say Gmail and you may have a

marketing person or a UXer or a PM or

you know anybody on the team that's not

quote unquote technical drop that into

AI studio start adding features and sort

of you know hacking on it in that way

Well, we're not at a point yet where you

can hit a button and that goes to

production, but we're not far. And I

think what's interesting is kind of

thinking about even like how the

relationships between the different

functions is changing. And at least for

us, a lot of our teams, there's this

blurring of people who are like part

engineer, part G P, part UXer. Do you

know what I mean? And those people are

oftentimes kind of creating the most

interesting kind of breakthrough

products.

>> I've been seeing my designers kind of

shipping products. Yeah. uh my kind of

building that's going into production,

>> right? Engineers,

it's insane, right? It's absolutely

insane. It's happening live.

>> It is. It is. It's kind of unfolding in

front of us. Yeah.

>> No, look like uh I think you guys have

been shipping like crazy. Uh the most of

the time that I would like to spend

going forward is for me to see stuff on

how you use some stuff.

>> Yes, let's do that.

>> Okay, that sounds good. Well, um we can

maybe cut to the demo here and I thought

we could maybe just start. you've been

on this site and I'm sure a lot of your

your listeners have too, but this is the

Google lab site. Yeah. As far as I know,

it's the only Google site with a hot

pink banner at the top, but it's kind of

got all our stuff. So, you know,

Pomelli, CC, Mixboard, Opal, you can

kind of go through and kind of check it

all out. Absolutely. And um anyway, so

everybody can check that out. It's

really fun. But maybe what I thought we

could start with is Pomelli.

>> I don't know if you noticed this

internet has been filled with videos

about Pomelli. It has.

>> People have been loving it.

>> Yeah. Fun fact, the Pamelli tweet,

number one Google Labs tweet in history.

>> Insane.

>> Number four tweet in Google's history.

>> So, it really took off.

>> Do you guys think of it as you guys are

building it? Because I also wanted to

understand how you guys think about labs

projects and how do you run the team as

well, but we'll get to it.

>> Yeah. Well, maybe Pomelo is an

interesting example. So, we can do the

demo is like as you're starting on

something like this, we're very

interested in how hard it is for small

businesses to create marketing assets

and collateral and all the social stuff

and everything. So, it kind of started

with a deceptively simple idea and it

was an experiment. It was basically like

I'll show you this next tab. This is uh

an organization in my hometown in

Oklahoma City, OKC, called Resto OKC.

And the idea was, could you just take

the URL restoc.org

and drop it into a tool, hit a button,

and start getting creative assets.

That's where it started.

>> Can a tool be a marketing team?

>> Well, we didn't even start there. We

basically were like, can you even just

extract interesting things? So, here's

what you do. You drop the tool in and

you get this thing called your business

DNA. And you can see with there's the

URL, here's the font, here's the color

palette, here's some of the assets from

the website. And we're like, "Okay,

that's kind of interesting because it's

like one little step, you get all this

value." And then we're like, "Okay, what

if you could make campaigns out of

this?" And it just starts creating.

Again, you look at all this stuff.

You're going to have like the font, the

sort of sort of assets. It kind of fits.

This organization actually does a lot of

kind of food service and donations in

the community. And then a few weeks ago,

we dropped this feature called photo

shoot. And this was the one that really

popped.

>> Yeah. It's all over the internet. Yeah.

And the idea here was like, can you take

assets from your sort of, you know,

website in this case, select an image.

Here's all the assets off the web page.

We'll pick this like head of lettuce or

something like this. And you can say

like, yeah, it looks good. Um, and then

basically what you can do is kind of put

these in templates. So, I kind of zoomed

ahead here so you can see it. But now

it's like, here's your piece of lettuce

on all these interesting backgrounds and

all this stuff. You can download and

take it wherever you want, right? And

that's kind of how this thing gets

started. And now people are starting to

use it like crazy. Yeah.

>> Um everywhere. Uh it launched in 160

countries. So it's now spread out.

>> India launch was wild.

>> Yeah, it it was I agree. We were

watching the the dashboards of the TPUs

very closely. But this is an example of

how something starts. Again, we're

co-creating it with marketing people.

Yeah, we think of it as a tool. We don't

think of it as replacing people because

this is actually something a lot of

small businesses don't have access

>> to photography like this or the time or

the setup. So it kind of creates this

ability to kind of give someone sort of

superpower in this case. Um so that's

one example. Um another one that's kind

of interesting on the design side is a

project called Stitch. You see this one?

Okay. Yeah. Yeah. Cool. So this one's

fun. The idea behind this is what if you

could kind of describe kind of an

interface you wanted and you can see

here it could be an app or a website.

I've kind of was working on one earlier.

This is like a discover Oklahoma City

again back to my hometown here. But you

can see it's kind of an infinite canvas

kind of idea. Nice little touch by the

way design. See the little spotlight

that traces. Okay. So here's how it

works. You're kind of in this thing and

you can go in and literally just come in

and edit these things. Drag it around.

You can describe sort of changes you

want to make. And then we just added

this feature recently which is sweet.

You right click on this thing and make

an instant prototype. Now you see this

arrow just got added that stitch almost

like the AI's in the board with you

starting to make things. It's going to

go off and use the Gemini 3 model. Here

it comes to turn this thing into

actually a clickable prototype. So if I

click in here, boom, here's like your

trip plan. You can go back. You can go

to some of your saved places. So again,

this whole thing, what's also so cool

about this is any of these screens, you

can rightclick and just grab the code

and it'll pull this up and you've got

all the front-end code ready to go. It

also

>> So while Stitch is actually a design

first tool, it's in a way being built

using code.

>> Exactly.

>> So it's VIP coding but with a VIP design

interface.

>> It's like an inception thing here.

That's right. That's right. Yeah. And I

think what people what we're finding

here, you mentioned this earlier, it's

like who do who's using Stitch? Yeah,

>> this is a whole category we think of

vibe design.

>> People that are using it are not

designers are using it, but you also

have a engineer

>> maybe who never could have maybe done

front-end code as a backend engineer

that comes to it. You're having people

take this into AI studio and then start

doing things. So, you're getting all

these really interesting combinations.

>> Now, you can do that to anti-gravity as

well.

>> That's right. That's right. That's

right. You can drop it in there, too.

Um, so this is kind of where Stitch is

right now. Kind of where it's going.

It's kind of interesting. Um, maybe I'll

just keep going. Is that okay? We just

have a lot of tabs lined up for you.

>> Let's go for it. So, this is um this is

notebook LM. This is one of our first

Google Labs products that sort of went

viral with the podcast. I think the

things that I'm very excited about where

this is going next. So, we've added

obviously the ability to do deep

research right in Notebook LM. So,

here's like a trip to like planning a

trip to Hawaii. You can do deep research

and grab 66 sources.

And of course, you can do your normal

Q&A with it. But there's some new things

that have come out that are pretty fun.

So, one is I'm sure people have seen the

infographics. Yeah.

>> One tap. This is pretty cool because

Hawaii obviously several different

islands. I love these little snapshots

of each island uh that you can kind of

check out. So, that's kind of fun. The

newest thing though is actually these

cinematic video overviews.

>> Oh boy. So, what's really cool before I

play it is imagine 66 sources. In this

case, Gemini is going to basically act

as kind of an assistant director, put

together a script with kind of like

drama and plot and everything. It's

going to come up with a visual style

which you can customize if you want and

then put it all to kind of audio and

time it all up. So, do you want to see a

little

>> put a little uh audio on here and I'll

make it bigger. Now, this is pretty

cool. This was like a felt style it just

came up with. This is the ocean. Check

this out. When you picture a trip to

Hawaii, you probably imagine this. Empty

white sand beaches, perfect surf, and

total isolation from daily stress. This

equation shows the financial reality of

that dream. A standard 7-day trip for

two people, factoring in place, lodging,

and

>> now I'm going to pause for just a second

because all of these are actually coded

in the background. So Gemini is like, I

should make an interactive chart that

builds this bar chart,

>> writes the code, and then drops it in,

which is a kind of a cool.

>> So it's basically converting an

animation into a video.

>> That's it.

>> And the animation is So the beauty of

this world is everything is written

using code.

>> There you go. There you go. So as we go

through it, we won't play the whole

thing, but I want to play you one more

little section, but this I wanted to

call out just because I thought you'd

like that

>> at 250.

>> Okay. So this is an expensive trip that

it's

>> I'm sure you can afford it.

Now check this out. Then it's going to

go through

>> blur the Hawaiian islands together into

one identical flawless paradise behind

those filtered photos. Each of the four

main islands features radically

different weather patterns.

>> Okay, so here's another example. We take

like a map, convert it to code, and kind

of do this overlay and start connecting

it this way. So when we saw I remember

the first time I saw this, I was like,

whoa, this is new. Um, but it's kind of

again this blurring and mixing of in

this case some code, some video, some

nana banana all coming together. And I

think what's amazing and kind of what's

made notebook such a fun project is just

it helps you understand anything. So

whether you're planning a trip or you're

a student or you're a teacher or you're

a knowledge worker, whatever, you just

drop in the sources and then you can

start making stuff with it.

>> I think this is a perfect tool for

people when they say right uh every

person can have a AI tutor. Yes,

>> I think this is it.

>> That's how a lot of people are using it.

That's exactly right. Yeah, quizzes in

this thing are super popular

>> and so is the podcast. So, I will tell

you how I use this, right? Like I'm

geeking on AI and every single day I

wake up with 100 updates.

>> Yeah.

>> So, I' I have found a way to ingest all

the updates into this. It converts that

into a podcast. And when I'm in the gym,

I'm listening to the notebook podcast.

>> Nice.

>> Right now, I have a request for you.

>> Okay.

>> Okay. You guys shipped the CLI for

Google Workspace. Yes. Yes.

>> Can you please add notebook element to

that as well?

>> One of the top

>> so that I can automate it.

>> One of the top requests either CLI or

MCP for notebooks. Okay. Feature request

noted.

>> Now people have been hacking this round.

I actually do it by the way today by

using browser use.

>> Yeah. Yeah.

>> Right. But I'm like can you please

note it? Noted. Okay. Two more for you.

That's okay. Let's do a little Genie and

then we can jump over to Gemini. So, um,

Genie, I think, is one of those projects

where you kind of have to see it to

believe it. And I remember exactly where

I was when I saw the first one. I was

just like, "What? I can't believe

possible."

>> So, um, you know, the thing about Genie

is kind of a world building kind of

tool. It's very experimental right now.

It's very limited in terms of even just

how many people we can have on it.

What's kind of fun is we built all these

little worlds that you can try. The cat

one's kind of fun, right? Yeah. I think

the other thing that's kind of fun about

Genie is um you can bring your own,

right? So that's pretty fun, too. This

one's kind of one of my new favorites.

I've been using this one f recently.

This is like, you know, kind of a a

parrot or whatever flying through the

Amazon. Now, as this world gets built,

it takes about 20 25 seconds. You'll see

the keys coming on here. It's kind of

old school gaming. Spacebar to jump, you

know, WD to turn. You can change the

camera angle.

>> GTA V city is bad.

>> That's right. That's right. And I think

what's been fun is just watching because

at the end of this it'll record it and

you can share the video and so there's a

whole other kind of side of Genie which

are people sharing these worlds.

>> I tried to put myself in one of these

scenes as well.

>> Oh nice. Is it any good?

>> It was pretty good actually. It's better

than what I expected.

>> Okay, so here's our bird. He's Oh wow,

he's flying pretty fast. So we're going

to try to make it through this tunnel

and I think Yeah, there we go. Now we're

on the other side. I'll hit spaceb bar

and I'll kind of he'll start climbing.

Try to get to the top of this whatever

little ruins

>> like a balan. Check the shadow. See the

shadow right over there, which is

incredible. Yeah, it looks feels like

we're somewhere like anchor watt or

something, right? Um, now I think what's

amazing about this is every frame is

being generated

>> on the fly.

>> On the flying. So, it's insane. When I

turn this, I'm going to kind of steer

him a little bit. That jungle's there.

If we came back and flew through this

again, that jungle may not be there,

right? So, it's all got it.

>> It's all generated in sort of truly in

the generative sense. Every pixel is

kind of streamed in this way. Now, the

other thing you'll notice, which is kind

of fun, you see at the top there's this

bar that's getting dangerously close to

the end.

>> This our world is about to collapse

here. So, right now, the genie worlds

only last 60 seconds and then they're

gone. Poof. So, what'll happen is it

gets to the end here. Oh, I've never

actually seen this little river back

here. This is interesting. Okay. See how

you just froze and now our world's like

collapsing on us. We do that partly to

keep the consistency. That's about as

long as it'll last. Partly because we

don't want to melt too many.

>> Yeah, that's what you don't want to melt

your GP.

>> That's right. That's right. So anyway,

this is um Genie. We're super excited

because um clearly this is a little bit

of a toy right now in terms of the

experiment we've built, but it shows

kind of the potential we think of

gaming, the potential maybe of

entertainment, potential even of

robotics and embodied AI because as

you're building these worlds, you're

getting a feel for how to navigate them.

So we'll see a lot more from this in the

coming years.

>> Yeah. The first thing that came to my

mind when I saw Genie is I miss my old

hometown. The way it looked. Oh yeah,

>> right. And I believe tech like this will

enable me to preserve the place the way

it is on a 3D

>> and then I could actually go back.

>> Yeah. Yeah. Yeah.

>> Right. So it's like Google Earth.

>> Yeah.

>> But 10 years behind, 15 years behind.

And I preserve that forever for my kids

to see, for my grandkids to see.

>> Yeah. Yeah. Sort of captured in that

world.

>> So I can see a I can see a product I'm

sure you guys are thinking about it. A

mix of Genie plus Google Earth.

>> Yes. Yeah. Right.

>> Well, you know, it's so funny. Um, back

in California in Mountain View, there

used to be this big Google Earth

display. You'd walk in, it was

wraparound screens, okay?

>> You had a little joystick. Guess what

the number one thing everyone would do?

They'd go to their hometown. That's it.

That's it. And they would kind of zoom

in. It was satellite view.

>> I had done that, too. I mean, the first

time when I came across Google Earth, I

would be like, "Can I find my home?"

>> Your oldest.

>> Yeah. But this is cool cuz you could

actually like walk the streets or, I

don't know, become a parrot and fly

around or something. I think you guys

should have shipped this four, five

years before so that GTFI could have

made it earlier.

>> Oh man. Yeah, I don't know where that's

a different thing, man.

>> No, but we have done VIP design. We've

done VIP code. Now we are basically VIP

gaming.

>> Yeah, we've got world building.

>> This is God mode.

>> Let me uh let me show you one last bit

on just Gemini. We can wrap here, I

guess. Like so I think Nano Banana

obviously um thank you to India was one

of the the first countries where it took

off was just unbelievable. I would say

I'll show you some various things, but I

mean one thing with Nano Banana 2 is

obviously the text rendering just sets

it apart. This is really fun viral trend

is like nulling things, you know,

separating it out into the component

parts. Um, but this has just become a

thing. I think we'll see nano banana

type experiences for video, for music,

for voice, like this idea to just

conversationally edit feels so natural,

right? It's how it should work. It's

like talking to a designer who's right

next to you.

>> Yeah, that's right. And being able to

kind of iterate on things. Goes back to

some of the things we said about control

and sort of that precision. So anyway,

we're really excited about all the the

nano banana growth and other usage. This

one's kind of fun. You know, I threw in

something. Congrats on the T20, by the

way. Happened yesterday when we were

filming this, I guess. U but you can

take um something in Gemini now and just

kind of create music um very easily. And

actually, maybe I'll just show you real

quick how that comes up. It's super easy

because you just drop into Gemini and

just say like create music and you've

got all these templates to choose from,

right? And so here's one. One of my

favorite ones is this thing called bad

music.

So you could throw in something simple

like India Champions T20, right? 96

runs. Uh and then it'll just come back

with something like this.

>> We are the Titans, the thunder of

victory. The trajectory of planetary

alignment has blessed this sporting

endeavor.

>> Okay, so it's going to go on for 30

seconds and just like talk about how

great the victory was and everything,

but I think this is again 30 seconds.

Eventually, these will get longer.

Eventually, you'll have more of a

control kind of part to it. Um, just a

couple weeks ago, we announced we've

acquired this company, Producer AI. It's

very good at this kind of back and forth

iterative music. Um, so I'm very excited

about where music's going to go and part

of it entertainment and fun. I mean,

this song's never going to win awards or

anything, but it's more like

>> that one WhatsApp share that you do with

your friends.

>> That's it. You drop it in, you know, the

group chat with your buddies and it's

really fun. U, the other thing we're

seeing actually teachers and students

are using it. You know, sometimes you

can learn by a song or kind of the way

kind of a melody works and so that's a

whole other different branch of this

which is really interesting. Kind of

like a learning side to it. So that's um

that's Lia and kind of music in Gemini.

Um you had asked about personal

intelligence. I wanted to show you this

one. So a couple weeks ago um I lo like

lost my goggles for swimming because I

was like oh man what's happening? So I

had I basically did this search. What

are the most comfortable prescription

goggles you can buy? Um and what's cool

about this is so Gemini starts

structuring this. It's very neat, right?

Top overall, best custom fit, all of

these images. This is actually hooked up

to Google's shopping graph. Literally

billions of items where you can see you

get like the ratings, the pricing. If I

click on one of these, it'll actually I

can go out and buy this. It's all kind

of neatly integrated. So that's kind of

like the base experience, which is like

awesome. And then you can kind of scroll

down and it's like, oh man, what's my

prescription? I have no idea. And then

here it goes. Because I've opted into

personal intelligence. You last went to

the optometrist on February 11th, 2021.

I should probably go back. That's been a

while.

>> Is that pulling from your emails?

>> This is pulling from my email and a

photo of the prescription I took a

picture of.

>> That is insane.

>> It's incredible, right? And again, this

is a feature you have to opt into. So,

this won't just happen. You've got to

turn it on. I turned it on and then you

can see like right eye, there it is.

Left eye, there it is. And it's like,

oh, by the way, you also have on these

Warby Parker glasses you bought in

August, which is these are the ones I'm

wearing right now. Um, but this starts

to show the examples of like how

powerful it is if you want to connect

this stuff and it's how we see Gemini

being this personal proactive powerful

assistant that's working for you. And in

this case, I was like, I would have

looked through my email or my gallery

forever to find this.

>> It's so hard to find these kind of

things. So, it's like instant retrieval.

The other thing we're finding people are

using it a lot for I'll show you this

kind of fun was um remind me about my

past visits to India before this trip.

And so here it is like here's when you

were in Delhi, Bangalore, Mumbai,

Hyderabbad, you went to Matura and

Jaipur and then what's kind of fun is

like oh show me my photos from Jaipur.

It's like bam there it is. Um, so it's

kind of this incredible ability with

personal intelligence to really make AI

yours and I think that's kind of the

promise we see with something like this.

So this is um in the US today and

rolling out very soon. Um, the feedback

on this has been very positive. Um, so

we're really excited to get this out.

>> How far are we Josh from AI literally

recommending our stuff

>> proactively rather than us prompting it?

For example, here you're talking asking

about prescription, right? Of course,

this has to be prompted and not know.

But it could be a bunch of things.

>> Oh, yeah.

>> It could be proactive. For example, you

have a flight.

>> Yeah.

>> And you haven't booked your your seat.

>> Yeah. Yeah. Yeah.

>> Or the car like a taxi on the other side

or whatever, rental car, whatever.

You're reading the road map.

>> Yeah. Okay.

>> Um part of when we think of these three

Ps of Gemini, personal, proactive,

powerful, the proactive is all about

this idea of like why do I always have

to be the one talking to AI? Like

there's times you want it to talk to

you, right? and you want it to bring you

things. And I think ultimately for us, I

think what'll be like true success for

Gemini is like how much time does it

save you in your day? Absolutely.

>> How much money does it save you? And

these are the things you start to think

about on the proactive side that are

really interesting. So, this is a little

glance into the future of where that's

going.

>> Hey, look, thanks for shipping

everything that you've been shipping.

Yeah.

>> All right. I think uh behind the

curtains all of us have been like, you

know, cheering because you guys have

been shipping like crazy and we love it.

>> Great. Well, thanks for all the feedback

too and using it.

>> Now, for the last segment, what I wanted

to do is because we have like a couple

of minutes left.

>> I wanted to do something really quick.

We're going to do a quick rapid fire.

>> Yeah.

>> Right. If you're up for it,

>> I'm going to put you in soup.

>> I'll try to I'll try to. And let's see

where it gets to.

>> Sounds good.

>> According to you, what is the biggest

lie in AI right now?

>> The biggest lie?

>> Yeah.

>> Oo, that's interesting.

I would say like there's a lot of talk

of like is there too much compute? And I

all I can say from my side with the

products I'm working on is we just need

more

>> more TPUs.

>> I mean TPUs, GPUs, anything. Uh that the

the demand right now for this stuff is

so high. So I don't know if it's a lie

or not or maybe but that's the first

thing that came to mind.

>> Sounds good.

>> Yeah.

>> Uh I' I've read that 25 to 30% of code

at Google is written by AI already.

>> Yeah.

>> What happens when it gets to 90%.

>> A good question. I mean, one thing we're

seeing right now is things like

anti-gravity is the amount of kind of

PRs and just code for review that's

coming in is becoming the real

bottleneck. Yeah.

>> And so, um, and it also is introducing

interesting types of kind of almost like

fatigue when you have to review so much.

Do you know what I mean? It's like in

your your business, you're just getting

pelted with things. So, that's probably

what happens.

>> Orchestrator dilemma.

>> The orchestrator dilemma. Yeah. Yeah.

Well said.

>> All right. Beautiful. Uh, how do you

define AGI and ASI? What is your

definition? Because there is no defined

definition.

>> I've kind of given up on trying to

define so I'm not trying to get out of

the question, but there's so many

different definitions. I mean, I think

the things I'm watching right now with

all these different products and the

capabilities at the frontier are what

are the things that are surprising me?

>> What is the last thing that's surprising

you? Well, I'm still I think we're still

living through like a real

internalization of like how much you can

hand off to these models and that they

can do in these longunning tasks. And

I'm seeing stuff almost every week. I

feel like since Christmas and New Year's

like I thought the frontier was here,

but it's a little further.

>> The last three months have been just

wide.

>> It's totally changed, right? Yeah. I

feel like we're in another acceleration.

>> Awesome. One skill that will be more

valuable in the AI first world according

to you.

>> That's interesting. You know, a lot of

people are saying taste right now. I

think that's true. I think another one

though I would say is um

>> I think some notion of like being able

to reinvent yourself

>> to live in the AI first world

>> to live in the AI first world to

understand these tools. I think

sometimes

>> maybe sometimes people's identity gets

tied to a certain job that they do and I

think um

>> people that are kind of able to kind of

reinvent themselves and change with it

is going to be maybe one of the top

skills. I've been creating a lot of

content talking about everyone becoming

AI generalist. AI first problem solvers.

Solve problems, think AI first.

>> Yes,

>> that's how you win.

>> I think that's wise. I think that's

right.

>> Awesome. What's actual thing about

agents that nobody's talking about?

>> Oh, that's interesting. I would expect

us to be having many more conversations

about like the audit trails of agents.

There's some interesting startups that

are starting to be in this space. But I

think because agents have only really

started working in the last 60 days or

so. There's like a set of kind of stuff

around the harness and being able to

kind of understand what did the agent do

that I feel like is kind of

underexplored right now.

>> Makes sense. What are the top AI tools

and workflows that you use every day?

>> Oh man, a lot of the ones I showed you

today. Um I mean across our teams

everyone's in kind of Gemini AI studio

there's I'm in more and more meetings

where you know if you get any kind of

deck it's something created in notebook

but most of the time it's a prototype

you can play with and it maybe is a

stitch thing or maybe um so that's a

whole set of things I also try to play

with a lot of just other stuff I'm

>> very interested every Monday I kind of

write a note to the team of here's all

the cool stuff I saw in AI um and maybe

it sounds a little bit like the podcast

you're listening to in the gym I'm just

like trying to stay on top of it all. Uh

but yeah, I'm really trying to learn

right now because it's changing so fast.

>> So you also are learning every single

day.

>> Every day. Every day.

>> Beautiful. Are there any things that you

use that the world has not seen yet?

>> Oh yes, quite a few.

>> And I know about them. I can turn the

cameras off if you want to.

>> I mean the one at some point I can't

wait to get it. It's been out to India a

little bit I think but is um a project

in Google Labs called Google Beam. Have

you seen this thing?

>> No.

>> Okay. So imagine like a 55 65 in

>> Oh yeah. The screen the the 3D the 3D

screen.

>> Yeah. This team is so ambitious. It's

amazing team. They're trying to solve

teleportation.

>> Yeah.

>> So it's it's a really really good one uh

to go after and I think I think we've

cracked it. So those devices will start

shipping later this year. We're making

them with HP. So it's going to be fun.

>> Beautiful. Thank you so much Josh. This

was super fun.

>> Thank you.

>> I'm sure a lot of people are going to

learn a lot and I will see you soon.

>> Sounds great. Thanks for the time.

>> Thank you.

>> Yeah, I see you.

>> Good. Is Rishi super pissed at us

already?

>> Yes. Yes.

>> Have you been cutting the time? Like

time up. Time up.