0:00

Support for the show comes from Toyota.

0:02

What do you get when you take quality

0:04

craftsmanship and reliable performance and mix it

0:06

with bold design and effortless sophistication? You get

0:08

a Toyota Crown. Whether it's sleek sedan

0:10

or an impressive SUV, the Toyota Crown family

0:12

has the car you've been searching for.

0:14

With a powerful exterior that makes you stand

0:16

out and a smooth ride that keeps

0:19

you grounded, you can learn more at Toyota.com/Toyota

0:21

Crown family. Toyota, let's go places. Oh,

0:24

it's such a clutch off -season pickup, Dave.

0:26

I was worried we'd bring back the

0:28

same team. I met those blackout motorized shades.

0:30

Blinds.com made it crazy affordable to replace

0:32

our old blinds. Hard to install? No, it's

0:35

easy. I installed these and then got

0:37

some for my mom. She talked to a

0:39

design consultant for free and scheduled a

0:41

professional measure and install. Hall of Fame's son.

0:43

They're the number one online retailer of

0:45

custom window coverings in the world. Blinds.com is

0:47

the goat. Shop Blinds .com right now and

0:49

get up to 45 % off select styles,

0:51

plus a professional measure. Rules and restrictions may

0:53

apply. So the only things I've seen is

0:55

number one, it was an AI model that

0:57

costs like a tenth of what open

0:59

AI's models cost and adjusts is good. Allegedly.

1:02

Allegedly. And then I heard that it

1:04

was just trained on open AI stuff and

1:06

not actually that impressive. Oh no, they

1:08

stole our stolen content. What

1:14

is up people of the internet? Welcome back

1:16

to another episode of the Waveform Podcast. We're

1:18

your hosts, I'm Marquez. And I'm David. And

1:20

that means that Andrew's not here. Andrew's out

1:23

this week. Ellis is also out. Adam's holding

1:25

down the board. Zuri is sleeping on the

1:27

corner like usual. It's going to be a

1:29

good time. It's been a couple of weeks,

1:31

so we do have a lot to talk

1:33

about. We got an exclusive look at Samsung's

1:35

Project Muhan headset. A lot of interesting thoughts

1:37

there. We're going to touch on the new

1:39

Deepseek R1 model that's kind of all of

1:41

the rage in the industry right now, but

1:43

also talk about Pebble. Can

1:46

I say nostalgia about Pebble? I think so. At

1:48

this point, I feel like I can use that

1:50

word. We'll talk to Pebble for a bit. And

1:53

also we're going to play another game of something

1:55

or nothing at the end. Love it. But

1:57

first of all, I'm going give a shout

2:00

out to piece of content. So nothing, that's

2:02

kind of, great headline, nothing, nothing, the

2:04

company, did a video on our YouTube

2:06

channel. That was almost too smooth. But

2:09

the video was, so they basically emailed me

2:11

a couple months ago and we're like, hey,

2:13

we're gonna make a video on like what

2:15

it would take to make your dream phone.

2:17

You know how you've had like a dream

2:20

phone in the past where you just

2:22

like Frankenstein together a bunch of parts,

2:24

like just give us one of those

2:26

and we'll make a video of it. And we'll

2:28

make a video of it. you know, some specs

2:30

and yeah, it'd be cool if you

2:33

can combine these sent it off. And then

2:35

two months later, they're like, hey,

2:37

we made the video. So I

2:39

watched the video and it was

2:41

pretty interesting. Basically, it's them breaking

2:43

down roughly what the costs are

2:45

of a smartphone. Most interestingly, the

2:47

bill of materials itself in different

2:49

components of a phone I go

2:51

into actually building it. So I'm

2:53

sure you've heard before that like,

2:55

yes, you buy a new phone

2:57

and it costs. $143 of materials.

2:59

And then everyone goes, well, they

3:01

could have charged less, or they

3:03

could have, you know, here's all

3:05

these other thoughts on the difference

3:07

between bill materials and final retail

3:09

price. But I gave them some specs, and

3:11

this is what they came up with. So

3:14

go watch the video if you haven't already,

3:16

but I'll just give you some top line

3:18

level spoiler alert stuff that I think is

3:20

interesting, and maybe you have a reaction to

3:22

this stuff. So I told them a huge

3:25

battery. We rounded it up a huge battery.

3:27

We rounded up. They said $13 per

3:29

battery. Your dream is 15 watt

3:31

wireless charging? Well, I didn't specify,

3:33

actually. I just gave them, I

3:36

said fast charging. Oh, okay. My

3:38

dream would be, give me a hundred

3:40

watts, 100, 100 watts of both,

3:42

but $13 per battery. Okay, which

3:44

doesn't seem that terrible, actually.

3:47

I wanted a 6.1 inch, 120

3:49

hertz, LTP, 40P, 6.1 inch being

3:51

on the small side, but everything

3:54

else basically being like a high

3:56

end. S-24 at ultra level display.

3:58

$35 per display. Okay, that's more

4:00

than I expected. Yeah, I mean this,

4:03

I guess, includes the glass and the

4:05

panel itself. The digitizer, all that.

4:07

Yeah, that's fair. 35 bucks. Triple cameras,

4:10

we basically, I just handed them like

4:12

S24 Ultra, roughly, but also, not really.

4:14

I would like a further tele photo

4:17

in the S24 Ultra House, but the

4:19

idea is triple cameras on the back

4:21

and pixel nine front facing camera and

4:24

80 dollars total for all the cameras.

4:26

$80 for the cameras. That seems... We

4:28

got a bunch of sensors. Yeah, a

4:31

bunch of sensors and then all the

4:33

glass and optics and whatever has OIS

4:35

in there. Right? Seems crazy that it's

4:38

that much higher than the display.

4:40

I always figured the display would be

4:42

just as expensive as the cameras. Yeah,

4:44

I imagine the plastic actual lenses are

4:47

not that expensive, but... You never know.

4:49

The sensors, Sony has a monopoly on,

4:51

so what are you going to do?

4:54

Fair. Yeah. Or does Samsung, well no,

4:56

those are Samsung sensors. Samsung makes some

4:58

of their own sensors. I forgot there's

5:01

Samsung ISO cell sensors. Yeah, so I

5:03

mean, I'm not tied to those particular

5:05

sensors, but yeah, roughly, roughly 80 bucks.

5:08

The storage, give me a terabyte of

5:10

fast storage and 16 gigs of

5:12

RAM, 90 dollars. Oh, for that. What?

5:14

Yeah. Yeah. Really? Yeah. Really? Oh, storage

5:17

and the RAM. Storage and RAM. Okay.

5:19

$90. That still seems like a lot.

5:21

It does. And you know, whenever you

5:24

go to upgrade, I guess it makes

5:26

sense whenever you go to upgrade a

5:28

phone and you try to double the

5:31

storage, then you see like real price

5:33

increments. It's not like you have to

5:35

pay $200 more dollars in bill of

5:38

materials to increase the storage, but... you

5:40

know, it is it is fast. Yeah,

5:42

it is a tiny little UFS 4.0

5:45

SST. We'll go with it. And

5:47

then a snap dragon aid elite. So

5:49

again, they said they couldn't reveal their

5:51

exact supplier costs, totally fairest, and the

5:54

M. Caulcom doesn't let them, but they

5:56

basically referenced that a typical high-end flagship

5:59

chip will cost about. $190 for them

6:01

in their volume levels. So

6:03

pretty pricey. I actually have some

6:05

more thoughts in my S24

6:07

Ultra review on the cost

6:09

of the Snapchat and 8 Elite.

6:12

We'll get there. But yeah, $190 for

6:14

the chip. And then some

6:16

other small things,

6:18

a motherboard, antenna,

6:20

electronics, haptics, other

6:22

small electronics miscellaneous,

6:24

$15. Packaging, cable, stickers,

6:27

box, $30 dollars. Wow. That's a

6:29

lot more than I thought. Design

6:31

materials, so if there's fancy stuff

6:33

you want to do with the

6:36

back glass or any decoration or

6:38

maybe some special paint or something

6:40

like that, $8. That would be

6:42

why we see so many insane

6:45

weird designs on the back of

6:47

funds. Not that expensive. You

6:49

do Dragon Ball Z for

6:51

$8? Exactly. Structural parts.

6:53

Buttons, licensing, software.

6:55

$29 per unit. And then rough total

6:58

bill of materials. Again, this isn't

7:00

all in cost to make the

7:02

phone, but the bill of actual

7:04

physical materials, $500. So that's for

7:06

my crazy, super high-end, dream phone

7:08

idea. $500 to build. But then, of

7:10

course, you have to add on R&D, all

7:12

the teams, the employee costs, and

7:15

licensing, and things that actually,

7:17

at one-time engineering costs to

7:19

make this bespoke motherboard, work with this

7:21

RAM, work with this storage, all that

7:23

stuff. And then, you know, it breaks

7:25

it down in detail in the video,

7:28

but it might cost 20 to 25

7:30

million dollars to develop a phone. That

7:32

was a very funny part of the

7:34

video, because they're like, all in all,

7:37

it should be about $500, but R&D

7:39

is about 20 million. Yeah. Okay, so

7:41

we gotta sell a lot of these.

7:43

You do have to sell a lot

7:45

of phones to sort of break even,

7:48

obviously. My favorite quote, was, they wrote

7:50

as one of the costs. purchasing

7:52

products from competitors to understand the

7:54

market better. And they didn't give an exact

7:56

cost for that, but I thought that

7:58

was interesting. Priceless. Interesting. happens in a

8:01

lot of industries and I just kind

8:03

of forgot that that's a thing you

8:05

have to budget in like car companies

8:07

just buy tons of other companies cars

8:09

yeah and then drive them around and

8:11

go huh I kind of like that

8:13

definitely more expensive for car companies than

8:15

smartphone manufacturers well in total costs but

8:17

maybe per in development relative to the

8:19

final cost of the product maybe it's

8:21

reasonably by five or six cars to

8:23

make a car you buy five or

8:25

six phones to make a phone. So

8:28

yeah, anyway, go check out that video.

8:30

It's pretty interesting and insightful. I don't

8:32

really see that sort of behind the

8:34

curtain number stuff very often from smartphone

8:36

manufacturers. So we'll check that link in

8:38

the description. Okay. But anyway, okay. So

8:40

I think the big interesting thing we

8:42

got to try this week with the

8:44

MK PhD video we put out Project

8:46

Muhan, Samsung's headset collaboration with Google and

8:48

Android XR. I thought it was very

8:50

interesting. Yeah, I have so many questions

8:53

about this. I have all the answers.

8:55

Okay. Yeah, so I've tried to put

8:57

as much as I officially know in

8:59

the video. We don't know things like

9:01

the exact release date or the exact

9:03

price. We don't know a couple final

9:05

details on things like field of view

9:07

and resolution. I can just tell you

9:09

like what I thought looking into the

9:11

headset. But most of what this experience

9:13

was was me getting to try it

9:15

and use Android XR and Gemini built

9:17

headset. And yeah, it looks a lot

9:20

like a Vision Pro. They clearly were

9:22

inspired by a lot of parts of

9:24

the Vision Pro, but they did improve

9:26

on some stuff, like the battery, even

9:28

though it is a battery tied to

9:30

the side and in your back pocket,

9:32

it's a removable cable, it's a usable

9:34

cable, it's USB type C, you can

9:36

pop on another battery, which is pretty

9:38

cool. Can you plug it into a

9:40

wall? Oh. That's a good question that

9:42

I didn't get an answer to. I

9:44

would imagine you can, but that didn't

9:47

get an official answer. Because it'd be

9:49

cool if you could plug it into

9:51

like a laptop as well, and then

9:53

you have lower latency of some sort

9:55

of screen expansion. True. That's a great

9:57

point. Yeah, the cable's long enough to

9:59

plug it into. laptop for sure. I

10:01

don't know about a wall but it's

10:03

a cool idea. Okay. Yeah, I mean

10:05

they they made this headset a while

10:08

ago they've been finalizing it. It

10:10

was near final so what

10:12

we saw and what I

10:14

put in the video is

10:16

probably basically what you're gonna

10:18

get when you eventually get

10:20

to buy one of these

10:22

things and they'll name it

10:24

and it'll come out and

10:26

it'll be a some unpacked

10:28

event later this year probably.

10:30

Oh, Android X-R, the platform

10:32

name, yes. That is the

10:34

name of the software officially,

10:36

yes. The Muhan is a code

10:39

name, it means infinity, and

10:41

it's a code for this

10:43

infinite canvas that you get

10:46

to play with, which is

10:48

kind of cool. But it's

10:50

not the name of the

10:53

product. Is reality really an

10:55

infinite canvas? Yes. Okay, so.

10:58

Something the Vision Pro does really

11:00

well is that when you're looking through

11:02

it, the frame rate is high enough

11:04

that it doesn't really feel like you're

11:06

looking through a screen. It kind of

11:08

just feels like you're walking around and

11:10

the quality is a little bit worse

11:13

than real life. How does that feel

11:15

through this headset? Yeah, okay. Couple

11:17

notes. One, the resolution of the display

11:19

was pretty good. I would say a

11:21

notch below Vision Pro, but it's pretty

11:23

crisp. The overall pass-through from

11:25

the cameras was good enough.

11:27

It wasn't stunning or amazing.

11:30

Vision Pro is definitely better.

11:32

When you look through Vision

11:34

Pro and pass through, you feel very

11:36

close to just looking through some

11:38

blurry glass or something. This was,

11:40

yeah, a notch below that for sure.

11:42

But they did something interesting, which is

11:44

they let you take off the bottom

11:46

of the eye seal, and that gives

11:48

you this peripheral vision where light is

11:51

coming in on the sides and the

11:53

bottom. And for some reason that

11:55

made it feel better because

11:57

your brain sort of fills in.

11:59

the bezels around the display with the

12:01

rest. So like if I'm looking at

12:03

a wall and I see through the

12:06

middle of the pass-through and then my

12:08

peripheral vision sees the outside of the

12:10

wall, my brain kind of fills in

12:12

the gaps, kind of merges them in

12:14

a way. Yeah, which is interesting. So

12:16

I prefer using it without the light

12:18

seal because of that. But generally decent,

12:20

B plus pass-through, I would say. Okay.

12:22

Yeah. I have a question. Yeah. So

12:24

we actually just dropped today just dropped

12:26

today an Apple Vision. One thing that

12:28

the developers were telling us is that

12:30

Apple did a good job at making

12:32

Swift easy to code in for iOS

12:34

apps and when they released the new

12:36

headset that you could just like kind

12:38

of like know what you're doing over

12:40

there. Yeah. Did they mention anything about

12:42

that with the Android version? Like native

12:45

coding? Yeah. So the details that I

12:47

got from the team are number one.

12:49

All Play Store apps are compatible. Off

12:51

the bat. No extra code, they will

12:53

work. Just like as a little window

12:55

pop up? Yeah. So I went to

12:57

the Play Store, I downloaded a regular

12:59

phone app, and you could just pull

13:01

it up and use it like a

13:03

normal phone app, change aspect ratio, scroll

13:05

around, it works. But there obviously would

13:07

be a benefit to having optimized apps,

13:09

so what they call spatial apps, and

13:11

a bunch of the built-in apps are.

13:13

Optimize and apparently it's a few lines

13:15

of code to update your app to

13:17

be Optimized for spatial and then you

13:19

can go crazy customizing and adding way

13:21

more stuff But like the YouTube app

13:23

for example was a spatial app It

13:26

shows to curve the window a little

13:28

bit it added these floating side panels

13:30

for related videos and comments and then

13:32

you could add a background for an

13:34

environment that you're in and then watch

13:36

YouTube videos and this like floating background

13:38

on a mountain Not as high fidelity

13:40

as Apple's just not as good-looking. Okay,

13:42

but cool feature So my understanding is

13:44

that it will be pretty easy to

13:46

make Spatial apps But if you don't

13:48

update your app at all it will

13:50

still work on the headset Do this

13:52

is the reason that they haven't made

13:54

a YouTube app for Vision Pro because

13:56

they want it to be exclusive to

13:58

an Android XR headset I wouldn't be

14:01

shocked. I mean, that's a pretty

14:03

huge thing, especially since they shut

14:05

down the third party Juno player.

14:07

My question is, what's a better

14:09

experience watching YouTube in the browser

14:11

on Vision Pro or a built-in YouTube

14:13

app on this Muhan headset? Right. Because

14:15

in Vision Pro Safari, they have a

14:17

special like window node for a video

14:20

player now. Yeah. So you can watch

14:22

a video full screen with a background

14:24

kind of exactly the same way you

14:26

do on the Muhan headset. It just won't

14:28

be curved. Right. But you can change the

14:30

size, make it huge, make it further away,

14:33

closer to you. It's similar. Yeah. I mean,

14:35

generally, I feel like a player, like an

14:37

app is going to have better UX design

14:39

as well than just the website in a

14:41

window. The second you have to move around

14:44

the actual browser, it's better on the Samsung

14:46

headset. Yeah. Because I don't want to

14:48

have to move around YouTube. Right. Oh.

14:50

Yes. Yes. I hear you're saying and

14:52

scrolling. Yes. Yes. I hear you're. I hear

14:54

you're saying. Yes. I hear you're saying.

14:56

Yeah, let's see, other questions. Do they

14:58

have any sort of group stuff you

15:00

could do? No, none of that was

15:03

demoed. No. I would assume no,

15:05

honestly. I think it took Apple

15:07

a while even to start building

15:09

these like shared experiences and share

15:11

play is like their magic sauce

15:13

that like works kind of cool

15:15

with certain apps. I had no

15:17

shared space experiences with like another

15:19

person in the headset interacting with

15:21

the same. Model is me.

15:24

So no personas? No

15:26

personas, crazy personas. That's

15:28

a good question. I

15:30

didn't make any video

15:32

calls. So I don't know what

15:34

that would like to the person.

15:37

Yeah, on the other side. Maybe

15:39

they just haven't. They just see

15:41

your eyes really up close. Yeah,

15:43

there was, I mean, I was

15:45

in there, I mean, I was

15:48

in there playing with the headset

15:50

for maybe two or three hours.

15:52

Take away was damn gemini. This might be the

15:54

most useful thing I've done with gemini just

15:56

because you can interact with it with your

15:58

voice it can actually poke into the UI

16:01

of what you're doing. And it seems like

16:03

it can actually move into spatial apps

16:05

and like use the UI for you inside

16:07

of spatial apps. But I could ask

16:09

Gemini to clean up my windows and if

16:11

I had six windows, it would line

16:13

them all up next to each other. That

16:15

was great. I could ask it to

16:17

open an app and open Google Maps and

16:19

show me this thing. And it would

16:21

do that for me. If I was looking

16:23

at something in pass through like a

16:25

painting on a wall, I could say who

16:27

painted this and it would just pull

16:29

up a web browser and Google something like

16:31

who painted this. So it was pretty

16:33

like hands free. Obviously it still has eye

16:35

tracking. It will still have controller support,

16:37

but the complete lack of having to do

16:39

anything with my hands and just talk

16:41

to the thing was surprisingly nice. That was

16:44

something I noticed for your video. So

16:46

you said that you basically just have to

16:48

put it on, do the eye alignment

16:50

thing, which takes like a second and it

16:52

automatically does hand tracking. Yes, hand tracking

16:54

is automatic. So you don't have to do

16:56

the whole vision pro thing of like,

16:58

look that way, look that way, look up,

17:00

look down, put your on. Or eye

17:02

tracking, you do have to do the looking

17:04

at the dots to dial in eye

17:06

tracking. So by default, eye tracking was actually

17:08

off and I was doing everything with

17:10

my hands. And so I'd point my pinched

17:12

fingers at the subject and select things,

17:14

which is a lot of arm movement. Obviously

17:16

in vision pro, you straight away set

17:18

up eye tracking and you never really do

17:20

that. You can do essentially the same

17:22

exact setup. It's like looking at a bunch

17:24

of different dots around the screen and

17:26

then it's like, all right, we can see

17:29

your eyes now. And then you can

17:31

do the same sort of stuff. Does the

17:33

eye tracking feel as precise as on

17:35

vision pro? Yes, yeah. Yeah, it's the same

17:37

magical look at something. It's highlighted instantly,

17:39

move around the UI with your eyes and

17:41

you can see like things like highlighting

17:43

as you look at them, it's pretty good.

17:45

Okay, so when vision pro came out,

17:47

it definitely felt like a very magical type

17:49

of technology. you think that this feels

17:51

like it's exactly caught up? If the vision

17:53

pro is an iPad pro, the Project

17:55

Mulhanna headset is a Samsung Galaxy Tab. So

17:58

if, yeah. Yeah, so if

18:00

you love the magic of this is a

18:03

super thin piece of hardware that is actually

18:05

a computer and is amazing that it can

18:07

do all these things on a display You

18:09

will feel the same thing with the Galaxy

18:12

tab There are certain things that Apple

18:14

does in their own ecosystem that

18:16

are another level of magic on

18:18

top of that whether it's share

18:20

play or shared experiences or whatever

18:22

apps that you're using that will not

18:24

quite be the same on the Google one,

18:26

but Google has Gemini, Google has YouTube,

18:29

Google apps built in, it has that

18:31

extra functionality to make up for maybe

18:33

not being as magical. Right. So it

18:35

is very functional. There's no way it

18:37

costs $3,500. I think this thing is

18:40

gonna be 2K tops, probably less. Right.

18:42

And I think that's gonna be a

18:44

big reason people will consider it over

18:46

a Vision Pro. But yeah, not as magical.

18:48

I think that's fair. Interesting. I guess the

18:51

relevant comparison could be like. chat GBT to

18:53

Gemini where because Google has all of these

18:55

Google apps that Gemini can tap into and

18:57

it already has all your data it's more

19:00

useful than just like a certain form. Yeah

19:02

in certain ways yeah it's and it's cool

19:04

and I still want to play around with

19:06

it again and try things like like searching

19:09

inside of apps like I wonder if I

19:11

could ask Gemini in a headset about an

19:13

old email and I wonder if it would

19:15

go into the email app and like search

19:17

through that and find it that would be

19:20

amazing. Because essentially it's just pulling up an

19:22

instance of Gemini Live, which is the conversational

19:24

thing that's on your phone already. So I

19:26

imagine whatever that can do, this can

19:28

do in the headset. And since it's multimodal,

19:30

I can take a picture on my phone

19:33

and it can search that. Well, the headset,

19:35

it's running past through all the time. Gemini

19:37

Live can see that. And so I can

19:39

just ask about something I see. So there's

19:41

a lot of potential, I think. And when

19:43

you were talking to Google Google, did they

19:45

have like a. Apple never really defined what

19:48

it thought the Vision Pro was best

19:50

suited for. It's true. And I feel

19:52

like it tried to figure that out

19:54

after it launched it. It was like,

19:56

this is great for media consumption, but

19:58

also kind of work and as they

20:00

have updated Vision Pro, they've updated the

20:02

categories that they found people were using

20:04

it for more, but it still isn't

20:06

that defined and I feel like that's

20:08

part of the reason it is not

20:10

doing well. Did Google have like a,

20:12

we want people to use this for

20:14

work, we want people to do this,

20:16

like do they have a positioning for

20:19

this thing? It was mostly centered around

20:21

Gemini. I think if you asked them,

20:23

they would mostly tell you like, like,

20:25

this is the best way to use

20:27

Gemini. You could use it on your

20:29

phone, you could use it on your

20:31

tablet, or you could just use it

20:33

on a pair of glasses or a

20:35

headset that you're wearing all the time

20:37

and it's always able to talk to

20:39

you. So I think their pitch would

20:41

be, it's the best way to use

20:43

Gemini. And that's, it reminds me of

20:45

your AI pin, like the part of

20:47

the video where you ask the IP

20:49

pin to name the car and then

20:51

you just pull out your S24 ultra

20:54

and it does it faster. In a

20:56

world where you're walking outside with this

20:58

headset on and a car passes in

21:00

front of you, you literally look at

21:02

the headset and go, what car, you

21:04

just look at the car and go,

21:06

what kind of car is that? And

21:08

Gemini can see the car you're looking

21:10

at and can tell you right away.

21:12

I had a clip in the video

21:14

where I just held up a book,

21:16

and the book had a picture and

21:18

in the bottom it said like the

21:20

name of the art, the name of

21:22

the photo and where it was taken,

21:24

and where it was taken. context, understood

21:26

that it was a location, opened Google

21:29

Maps, searched that location, and then took

21:31

me to that location in Google Maps.

21:33

So if it's able to do that

21:35

sort of parsing and understanding and carry

21:37

out actions into apps on my behalf,

21:39

I think that's the pitch. That would

21:41

be cool. I would like to see

21:43

what the limits of that are outside

21:45

of a contained space. Totally. Also, do

21:47

they have any sort of like... workspace

21:49

expansion stuff where they can expand your

21:51

laptop screen or anything like that. Did

21:53

not hear about that. Okay. Didn't hear

21:55

about that. Interesting. I would be curious

21:57

what happened. if there's any screen mirroring

21:59

or if we plug it into a

22:01

laptop, what happens? Because I didn't get

22:04

to do that either. I want to

22:06

see their first ad for this device

22:08

because I need to know what they're

22:10

trying to tell people to buy it

22:12

for. That's a great point. The ads

22:14

say a lot about what they imagine

22:16

people use it for. I did get

22:18

to connect a Bluetooth mouse and keyboard.

22:20

And there wasn't like a computer also,

22:22

so it was just me using the

22:24

mouse and keyboard to move around Android

22:26

XR apps. which is cool. But yeah,

22:29

the, you know, quickly going from like

22:31

pinching something and then typing on a

22:33

keyboard and then a mouse, all that

22:35

was, it was very fluid. Interesting. But

22:38

yeah, I'm curious about that first ad

22:40

too. Yeah. Yeah. Huh. I got two questions.

22:42

Yeah. One. Were you able to look

22:44

at a bright thing? and then back

22:46

at a not so bright thing. Yeah,

22:48

good question. So like you said, this

22:50

was very controlled environment, and in the

22:52

interest of getting as many shots for

22:54

our video as possible, we basically stayed

22:57

in this evenly well-lit room. Yeah. There

22:59

was a display in the room, I

23:01

could look at it, and I could look

23:03

around very quickly, and that seemed very fluid,

23:05

but I didn't really. I don't have any memory

23:07

or footage of looking at a super bright light

23:09

and then something dark. So I don't know if

23:12

it would handle it the same way the Vision

23:14

Pro. I kind of imagine it will though. Yeah,

23:16

just kind of trying to balance out exposure quickly

23:18

and hopefully not peak. Yeah. Okay, next question.

23:20

Do you think that this is like from what

23:23

you experienced and going back to what their first

23:25

ad is going to be? Is this basically just

23:27

a dev kit again, just like how the

23:29

Apple Vision Vision Pro was? Yeah, it's

23:31

basically a deaf kit. I think that's

23:33

accurate. Honestly, right now, they are using

23:36

this, this headset as a deaf kit.

23:38

And that's where you'll see it right

23:40

now before it's shipped. But yeah, it

23:42

feels like, okay, Vision Pro is out,

23:45

Quest 3 is out, there are headsets

23:47

that are out, there are headsets that

23:49

are out in the world, and if

23:51

we want any sort of advantage or

23:54

if we want people to use

23:56

our platform, we need people

23:58

to develop for it. them to

24:00

making magical amazing apps that

24:02

make you want to use it. Like

24:05

the Vision Pro problem still applies when

24:07

it's Android. Yeah. So I just really

24:09

need to know it applies but it

24:11

applies but it applies at half the

24:13

price probably probably hopefully less

24:15

maybe maybe less because even at

24:18

1750 this thing is way over.

24:20

Yeah because like a cheaper A-R-V-R

24:22

experience you get the meta the meta

24:24

quest. Yeah all the way down to

24:26

three 50-300 bucks yeah. Yeah, plus if

24:28

you can't plug this into a computer,

24:30

like Android apps or a... I'm

24:33

sure you'll be able to like mirror

24:35

your chrome tab or something, you know.

24:37

Because right now the Quest can already

24:39

have your computer display. I feel like

24:41

you need to have that feature in

24:43

this headset. But also if you're gonna

24:45

charge more than the Quest, then you

24:47

have to be able to do more

24:49

than the Quest. And the Quest has

24:52

a huge library of games. That's like

24:54

what they're really good at right at

24:56

right now. this kind of gets positioned as

24:58

yes you can work in it but also you're

25:00

you're just gonna get so much done with Gemini

25:02

all the time it's gonna be the super useful

25:04

thing that helps you all you're gonna want to

25:07

walk around the office and write emails while your

25:09

hands are in your pockets yeah but the problem

25:11

is this thing is so big just like the

25:13

Apple Vision Pro like how long until this

25:15

is in a pair of Ray bands like if they're

25:18

just gonna put Gemini in glasses yes okay

25:20

okay right Great point, and I think

25:22

this will be probably the last big

25:24

point for this. Android XR as a

25:26

platform can be in a bunch of

25:29

different form factors. So in this video,

25:31

I said that Project Muhan, think of

25:33

it as like the pixel or

25:35

the nexus or whatever of this

25:37

platform, meaning that it's an example for

25:39

a version of the hardware you can

25:42

put it on. Yeah. But there will

25:44

be, from other manufacturers, higher end versions,

25:46

lower end versions, different form factors. We're,

25:48

I think, maybe having a conversation about

25:51

PC's, maybe, was this last week on

25:53

the podcast? But like, desktop computers are

25:55

still a big thing that you put

25:57

on your desk and people will buy.

26:00

or lower-end components, but they'll just

26:02

run Windows again at the end

26:04

of the day. Right. And so,

26:06

AndroidXR will run on smart glasses,

26:08

too. But if it's on smart

26:10

glasses, you will have less total

26:12

functionality. You can't watch a movie

26:14

on them. You can't. See the

26:16

Gemini U. I is easily it'll

26:18

just be text-based instead of like

26:20

this whole animation thing So there

26:22

are pros and cons to different

26:24

form factors, but the idea is

26:26

they're introducing Android XR and then

26:28

hopefully the hardware ecosystem blooms and

26:30

the developer ecosystem Blooms right and

26:33

everything's blossoming at the same time

26:35

more of instead of trying to

26:37

crash the VR and the Smart

26:39

Glasses into each other you're just

26:41

saying we're just going to run

26:43

on on all the types of

26:45

platforms Yeah, all the types of

26:47

headsets. Exactly. Yeah, Android XR is

26:49

the base software layer for all

26:51

of the future AR, VR, X,

26:53

R, experiences that will come in

26:55

the form of headsets, classes, etc.

26:57

for Google. Interesting. Yeah. Okay. So

26:59

I'm very interested in where this

27:01

goes and how they position it

27:03

and all this stuff. I did

27:06

hear that they were planning to

27:08

release it in 2025 though. Yep.

27:10

So they wouldn't stop telling us

27:12

that. Okay! Every time we asked

27:14

about it, we're like, so this

27:16

year, and then Andrew would ask,

27:18

so actually this year, and then

27:20

they go, yeah, yeah, yeah, yeah,

27:22

yeah, a couple months. Yeah, so

27:24

they're, I think. You should have

27:26

asked them where the Galaxy Home

27:28

is. And I feel like. And

27:30

I feel like. You should have

27:32

asked them where the Galaxy Home

27:34

is. And I feel like. And

27:36

I feel like more comfortable more

27:39

comfortable more comfortable. Yeah. So, yeah,

27:41

Samsung, we'll see. 2025 allegedly. Okay.

27:43

And nothing about price. Nothing. Nothing

27:45

about price. Cool. All of the

27:47

price, all of my estimates would

27:49

come from, like if I was

27:51

betting, like we, we could probably

27:53

estimate in an educated way. Like

27:55

I looked at that display in

27:57

the field of view and the

27:59

material. and the battery, and I don't

28:01

know the bill of materials, but I

28:04

imagine from Samsung they would position

28:06

this in a way that they

28:08

could undercut Vision Pro, but still

28:11

be looked at as premium. Let's

28:13

take a market that has very few

28:15

people in it and make it even

28:18

smaller. Well Quest is popular.

28:20

That's true, but for gaming. Right.

28:22

Like you know what that's for. Nobody

28:24

knows what these other things are for.

28:27

Yeah. Well hopefully so that's it I

28:29

did play one game briefly I don't

28:31

even remember it was one it was I'm

28:33

gonna launch is like such an idiot I forgot

28:35

the name of this game it was like a

28:38

flat game with like a bunch it was

28:40

a house cross section with a bunch of flappy

28:42

bird oh was it a fallout shelter yes

28:44

that oh and I played that for 30 seconds

28:46

and I was like wow it works with

28:48

my hand control that's crazy where I could

28:50

have a mouse and did it feel intuitive at

28:52

all I don't know how to play that

28:54

game so I couldn't click things and I held

28:57

things down and I moved them around so

28:59

yeah I think he was pretty intuitive. Decided to

29:01

play mobile games on a thing I have to wear on

29:03

my face. I mean there's gonna be games that don't work

29:05

at all. Yeah, I don't I imagine

29:07

like a need for speed game is

29:09

not gonna probably just be in a

29:12

window, right? You gotta have controllers at

29:14

that point. Yeah, I think you need

29:16

controllers right. So is this built on

29:18

daydream or is this just a brand

29:20

new situation? We don't talk about that.

29:23

I don't think daydream. I don't think

29:25

Google even remembers. Do you remember cardboard?

29:27

Yeah, it's under my seat. Cardboard was

29:29

fire. Just put your phone right on

29:31

your face. I'm gonna guess when this

29:34

Samsung headset comes out. It's gonna retail

29:36

for... Nope. More. No, I

29:38

was gonna say 1599

29:40

or 1699. I'm guessing

29:42

like 699. All the

29:44

way to 699. I

29:46

think it'll be more

29:48

than Quest, but it'll

29:51

be way cheaper than

29:53

Vision. Wait, 699? 799.

29:55

799? Yeah. Not 1799.

29:57

I'm going with 1599.

29:59

Oh. Ooh, okay. Like

30:01

think about like an S24 Ultra

30:03

is already $12 .99. Yeah. But

30:05

everyone knows what to do with an S24

30:07

Ultra. That's my thing. Like in order to

30:10

get people to buy a thing, like people

30:12

buy the Quest, even though they're not totally

30:14

sure what they're gonna do with it because

30:16

it's 500 bucks. And it's like two games

30:18

they know they wanna play. And during Christmas,

30:20

it's 350. They wanna play Beat Saber. It's

30:22

a splurge moment. The Vision Pro is

30:24

almost never a splurge moment. Yeah,

30:28

I just, I think it's all about

30:31

positioning from Samsung. Like in the way

30:33

that they position their phones against other

30:35

phones, they have Do

30:37

they even position their phones? Well, they call this...

30:39

We know you're gonna buy it. Like they

30:41

have the S25 and the Plus and then the

30:43

Ultra which has to be more because it's

30:45

bigger and but it's not that much better but

30:47

it has to be more expensive. So it's

30:49

$12 .99. Yeah. That's because they know

30:51

that there's that market of people that will just

30:53

get the best thing no matter what. Right.

30:55

And I think they'll see other headsets come out

30:57

after this and then this one is the

31:00

premium one in that world. There will be $700

31:02

headsets that run Android XR. From Lenovo. But

31:04

I think, yeah, I think they're gonna look a

31:06

little different from this one. And I think

31:08

the one that looks like Vision Pro, I

31:11

think they're aiming it at people

31:13

who might have thought about buying a Vision

31:15

Pro. So somebody who has $3 ,000. You gotta

31:17

make your comeback, King. We're gonna get those

31:19

cheaper headsets but I think they're gonna go

31:21

right up well past 1 ,000. So that's

31:24

why I'm going $15 .99, $16. For some reason,

31:26

$16 .47 is in my head but that's

31:28

not a real price. $16 .40. Great year. Let's

31:30

go $15 .99 is my official price. I

31:32

guess I haven't touched it so I don't

31:34

know how the hardware quality is. There's glass

31:36

on the outside. The frame is metal, it's

31:38

not plastic. Remember, Quest is just plastic. And

31:40

I think this has metal, it had fans,

31:42

it had that Snapdragon chip inside. You

31:44

can hear them? No, I

31:47

didn't hear them. But it had fans,

31:49

it had like nice machined buttons. It

31:51

had the braided cable for the battery.

31:53

Like it was definitely supposed be premium. Yeah,

31:56

so, yeah. It looked like

31:58

a Vision Pro, it looked like. You had the top. of

32:00

it like on your head but above and it

32:02

has the same freaking button as vision pro. I

32:04

said in the video, and people got a mad

32:06

because I compared to vision pro so much, but

32:08

I was like, it's not a digital crown here,

32:10

it's a button. And they were like, why you keep

32:12

using Vision Pro as a reference? Well,

32:14

I wonder. Because Samsung did. That's why.

32:16

It looks like a vision pro. And

32:18

it's also the one that a lot

32:20

of people have seen. That video is

32:23

30 million views or something. Like people

32:25

have seen vision. Like people have seen

32:27

vision. Like people have seen vision. the

32:29

Android version of the thing you've already

32:31

said. This is we've got

32:33

Vision Pro at home. There

32:35

you go. For people that

32:37

want to buy it for

32:39

their kids. Do you want

32:41

to make an official price

32:43

guess? $9.99. $1.99. Oh, I

32:45

feel like it's going to

32:48

be more now. It is. It's

32:50

really not. I'm telling you. I'm

32:52

giving you a chance. OK. So,

32:55

David cause. and Damien

32:57

Henri were two French

32:59

Google engineers with their,

33:01

with their 20% innovation

33:03

time off, they made

33:06

something that was introduced

33:08

at Google I.O. 2014

33:10

developers conference, 2014. In fact,

33:12

everyone that attended walked

33:15

away with it. What was

33:17

it? I have an idea. I was

33:19

there, I think I went to that.

33:21

I think you did too. What was

33:23

the thing? Eleven years ago. I should

33:26

notice. When we were not old. When

33:28

we were re-lads. When we were re-lads.

33:30

All right, well we'll think about this

33:32

one. Answers will be at the end

33:34

like usual. We'll be right back. Support

33:48

for Wayform comes from Life 360. So we'll be

33:51

the first to tell you that not everything goes

33:53

to plan. Life comes with its own set of

33:55

curveballs for you to handle especially when it comes

33:57

to protecting your family. But while you can't fool.

34:00

your life with Life 360, you can

34:02

family proof your family. So Life 360

34:04

is a location sharing app that makes

34:06

family life easier. Knowing where everyone is

34:09

at any given time, making coordinating daily

34:11

routines and activities, abrease, and you can

34:13

attach a Life 360 tile tracker to

34:15

all of your family's important stuff and

34:18

then track that stuff down right within

34:20

the app. So whether it's your keys,

34:22

wallet, even the family pet, you can

34:25

get on a collar, it tends to

34:27

go missing. It tends to go missing.

34:29

in my life. Generally, my cameras

34:31

tend to go missing. Ooburs famously

34:33

tend to go missing with my

34:35

cameras. There was this one time

34:37

that I definitely left my camera

34:40

in an Uber and it was

34:42

driving around, but I knew that

34:44

it was in the Uber because

34:46

I had a tracker on my

34:48

camera. So now all my most

34:50

valuable cameras that care with me

34:52

a lot. they have trackers on

34:54

them. Perfect. Yeah. So this year

34:56

you can stay connected with location

34:58

sharing and stay coordinated with place

35:00

alert notifications for when someone arrives

35:02

or leaves a given location. So

35:04

you can family proof your family

35:06

with Life 360. Visit life 360.com

35:08

or download the app today and use code

35:10

wave to get 15% off. That's Life 360.com

35:13

code wave. Support for Wayform comes from

35:15

Hostinger. So making your presence felt on the internet

35:17

can be a challenge. Sir, you can try to

35:19

make all the viral TikTok videos you want, but

35:22

if you want to have a place where people

35:24

can interact with you and be able to check

35:26

out your product directly, you need a home base.

35:28

And not just one of those cookie cutter ones

35:30

that we've all used. This is your brand. It's

35:33

your aesthetic and your footprint on the internet. So

35:35

that's why you might want to check out what

35:37

you can make. with Hostinger. So whether you're looking

35:39

to create a website or an online store, you

35:41

can do it using Hostinger AI website builder. Just

35:44

describe your site, name it, and get a personalized

35:46

site in seconds. And we're not just talking

35:48

about the design, you'll also have the ability

35:50

to generate content that matches your style and

35:52

tone and tailor it to the length and

35:54

topic that you choose. And also be able

35:56

to launch your business with a free domain

35:58

and professional email included with your. plan for

36:00

the price of a cup of

36:02

coffee. Plus, Hostinger provides intuitive tools

36:04

that anyone can use and 24-7

36:06

support if you get stuck. So

36:08

take your big idea online today

36:11

with Hostinger. Head to hostinger.com/waveform and

36:13

start your website today. Use code

36:15

waveform for an extra 10% off. What

36:17

if there were a medicine that could

36:19

heal almost anything? That would be great.

36:21

But they didn't want you to

36:23

know about it. Sorry, who's they?

36:25

Some people are following me and

36:27

I brought my tortoise. From executive

36:29

producers Mike Judge and Greg Daniels

36:32

and co-creators Joe Bennet and Steve

36:34

Healy comes an animated comedy thriller

36:36

about what it takes to change

36:38

the world common side effects. Series

36:40

premieres with back-to-back episodes February 2nd

36:42

at 1130 PM on Adult Swim.

36:44

Next Day on Max. All right,

36:46

welcome back. We've got to talk about

36:49

this Deep Seek thing. Yeah. Now, here's

36:51

my relationship with the Deep Seek headlines.

36:53

Okay. I was working on the Project

36:55

Muhan video. Okay. And trying to distill

36:57

all of my thoughts on this headset

36:59

that I just tried and like over

37:01

a flight back and like writing it

37:04

and carving out all the footage and

37:06

color correcting and editing and animating all

37:08

that blah blah blah blah. And when

37:10

I uploaded it. Deep Seek had already gone

37:12

through an entire news cycle. Yeah. Of

37:14

it coming out, being at the talk

37:16

of the town, and then suddenly is

37:18

maybe not so hot. It's still in

37:20

the news cycle. It's definitely still a

37:22

new cycle. A lot. I need you

37:24

to explain what I missed. Okay. Yes.

37:26

Okay. So it turns out we all

37:28

missed this, actually, because the original big

37:30

Deep Seek news. Well. part of the

37:32

big deep sea news happened on Christmas

37:34

Day. So clearly a lot of us

37:36

were going to miss this. Oh wow.

37:38

Okay. That's my fault for not really

37:40

digging into the forums on the on

37:42

Christmas Day on Christmas Day for the

37:45

groundbreaking AI reports. Yeah. Actually, the only

37:47

things I've seen is number one, it

37:49

was an AI model that costs like

37:51

a tenth of what open AIs models

37:53

cost. Allegedly. Allegedly. And then I heard

37:55

that it was just trained on open

37:57

AIs stuff and not actually that impressive.

37:59

No, they still are stolen content.

38:01

So that's my entire familiarity. All

38:04

right. Yeah, high level overview. This

38:06

is an AI language model similar

38:08

to open AIs. GPT. Yes, it

38:10

is specifically a language model. They

38:13

also introduced an image model, but

38:15

the language model is the thing

38:17

that people are mostly talking about.

38:19

Yeah. So on Christmas Day, this company

38:21

called Deep Seek that is this

38:24

Chinese AI company released this model

38:26

called Deep Seek V3. V3 is

38:28

a foundation model. Came out on

38:30

Christmas. No one really talked about

38:32

it that much because it was

38:34

Christmas. The thing that is causing

38:36

a lot of stir in the

38:38

community is that it performs really

38:41

well. It performs very similarly to

38:43

things like GPT4 and like one

38:45

of Claude's best models, stuff like

38:47

that stuff like that. But the

38:49

probably the biggest thing that caused

38:52

the stock market to start tumbling

38:54

and all of these things to

38:56

happen is that reportedly it only

38:58

cost $5.6 million to train. I

39:00

think the headline I'm remembering now

39:03

is that Envideo lost like $600

39:05

billion of revenue. They lost half

39:07

a trillion and of market cap

39:09

in one day, which is a

39:11

whole Stargate. So because of these

39:13

headlines by Deep Seek. Yes. So

39:16

if you sort of extrapolate the

39:18

fact that this costs 5.6 million

39:20

dollars, most of these AI companies

39:22

basically are trying to create these

39:24

gates. Not Gates. They're trying to

39:26

create these walls on who can

39:29

compete with them by saying, oh,

39:31

it costs $100 million to train

39:33

a foundation model. So only these

39:35

giant companies like Google, an open

39:37

AI, and I guess, Claude, can

39:40

train their own models to be

39:42

competitive and meta, you know. So

39:44

it's, it's semi open source. It's

39:46

like open weights. So they released

39:48

this really big paper about how

39:51

they trained it. And the paper

39:53

basically... discloses all of these different

39:55

methodologies that they use that basically

39:57

take similar methodologies to what the

39:59

big companies use, but they take the

40:02

most efficient approach possible, right? I think

40:04

a lot of the companies like Open

40:06

AI and Meta and Google, their big

40:08

thing has been, oh, we're just gonna

40:11

throw more compute at it, because for

40:13

a while, these AI models scaled with

40:15

the amount of compute that they threw

40:17

at it. Eventually, they started kind of

40:20

slowing down, and now they're trying to

40:22

figure out different ways to make them

40:24

more effective, like for example. Open AIs

40:26

01 model is like this reasoning model

40:28

that takes multi-steps, it does multi-step reasoning

40:31

to think through problems, which has a

40:33

higher level of accuracy for actually solving

40:35

complex problems. But things like that 01

40:37

model are in Open AIs $200 a

40:40

month tier. So they have all these

40:42

insane gates, right, that stops people from

40:44

actually using them or competing. I don't

40:46

think I saw them saying something about

40:49

how. the usage of that tier was

40:51

so high that they were still losing

40:53

money per user. Yeah, which is crazy

40:55

how expensive this stuff is to be.

40:58

Right. So some of the efficiency things

41:00

that Deep Seek did to make this

41:02

a lot cheaper is they use techniques

41:04

like this technique called mixture of experts.

41:06

And what that is is basically if

41:09

you have a giant model, right, like

41:11

for example, this V3 model has 671

41:13

billion parameters, you can basically say, okay,

41:15

well I'm gonna take. 37 billion of

41:18

these parameters and specify it towards things

41:20

like code, right? Because when you feed

41:22

a model, a ton of data, you've

41:24

got a ton of coding data, you've

41:27

got a ton of just natural language

41:29

data, you've got a ton of data

41:31

about different topics. So it's not actually

41:33

efficient to do this like next word

41:36

processing thing if you have to run

41:38

through the entire model and find the

41:40

next token. It's more efficient to say.

41:42

Oh, I've got an expert on coding

41:44

in the model here, so that let's

41:47

only run a parameter of 37 billion

41:49

instead of 671 billion. Okay, sure. And

41:51

that's both more efficient and more effective.

41:53

So it can be cheaper, it can

41:56

be faster, and it can be just

41:58

as maybe even more efficient. more accurate.

42:00

Right. And it's also, yeah, it could

42:02

be more accurate. And it can also

42:05

be much cheaper to inference. So if

42:07

you're actually the end user using it,

42:09

it costs less per token because it

42:11

doesn't have to, you know, go through

42:14

the entire model to find what you're

42:16

looking for. Okay. So this V3 model

42:18

was already a big deal because of

42:21

how cheap it was. But the thing

42:23

that's causing the bigger stir is just

42:25

last week, they announced a new model

42:27

called R1. of rabbit R1 fame. Prove

42:30

it. Can't prove it. Cannot prove it.

42:32

But R1 is a reasoning model similar

42:34

to open AI's O1 model. And it

42:37

does the same multi-step reasoning that

42:39

the O1 model does, but it

42:41

does it out in the open,

42:43

which is very interesting. So the

42:45

difference is when you ask O1 a

42:48

question, it'll say thinking, thinking,

42:50

thinking, thinking, thinking, and then

42:52

just kind of gives you

42:55

the answer. Yeah. Yeah. That's what I do

42:57

usually. Whereas the interesting thing about this R1

42:59

model is when you ask it a question,

43:01

say like you give it some complex question

43:03

like, oh there are four blocks on top

43:05

of each other, one is green, it's in

43:08

the middle, one is red, and it's here.

43:10

If I were to move the green block

43:12

on top, can you tell me the order

43:14

of the blocks? This is an example from

43:16

the computer file YouTube channel that I just

43:19

ripped off. Now normally it would get

43:21

those things wrong because it's very difficult

43:23

to, you know. take all that information

43:25

and understand the physics and all that

43:27

stuff, but in a reasoning model, it's

43:29

similar to a human being who writes

43:32

down the problem, and you have to

43:34

work it out, right? Because you can't

43:36

just instantaneously come up with that answer.

43:38

And so it will, step by step,

43:40

tell you, okay, so theoretically, if you

43:43

were to move that block to the

43:45

top, then now the blue block is

43:47

below the red block, which means that

43:49

blah, blah, blah, blah, blah, blah, blah, blah.

43:51

R1 shows that step-by-step, which is very

43:53

interesting for a lot of people because

43:56

now it shows you how these models

43:58

are actually working, which is really... crazy.

44:00

That's I kind of like that. Yes.

44:02

Because it's not like sourcing it necessarily,

44:04

but it's kind of interesting to see

44:06

it and maybe catch a mistake in the

44:09

process as it happens. Right. That's cool. Right.

44:11

The other thing that is interesting about the

44:13

way that they trained this is that generally

44:15

the way that you train a lot of

44:17

these AI models with complex problems is you

44:19

would tell it the question. You would tell

44:22

it all of the steps in between working

44:24

out the answer and then you would tell

44:26

it the answer. And that would be... the

44:28

data that you feed into the model. So

44:30

you're just feeding like all of these

44:32

worked out questions into a model, which

44:34

is a lot more data, and it's

44:36

also not really learning how to do

44:39

learning in the broader sense of the

44:41

word, how to actually work out the

44:43

reasoning for these questions. It's just kind

44:45

of mapping. Here's the question, here's the

44:47

reasoning, here's the answer, and so it's less

44:49

likely to get those things correct, takes

44:51

a lot more data, etc. This time,

44:54

they're basically just giving it the

44:56

question. and they're giving it the

44:58

answer, and they're basically saying, you

45:00

get a cookie if you get it

45:02

right, and you don't get a cookie

45:05

if you get it wrong. And so

45:07

now, they're sort of training the model

45:09

to, instead of just mapping the question

45:11

to the workout, to the answer,

45:14

they're actually kind of training it

45:16

to actually do that reasoning itself.

45:18

And that is way cheaper to do as

45:21

well. So, yeah, the thing that

45:23

is crazy about all of this

45:25

is because it is so much

45:27

cheaper. And they did it also

45:29

on a bunch of invidia H800

45:31

GPUs, which are the older, less

45:33

fast GPUs that they were able

45:35

to buy before the export ban,

45:37

because currently under the Biden administration,

45:39

you are not, and Vidia cannot

45:41

sell H100s to China, because we want

45:43

to be the, you know, in

45:46

America, they want to be the

45:48

fastest, most brawniest, you know, whatever.

45:50

So they're saying, oh, we did

45:52

this on H800 GPUs. 10 to

45:54

50 times less GPUs, it only

45:56

took two months to train instead

45:58

of three months. that it took

46:01

to train the open AI model,

46:03

and the open AI model also

46:05

had way faster GPUs. So suddenly,

46:07

in order of magnitude less GPUs,

46:09

50X less GPUs, you can use,

46:11

and videos like, maybe people aren't

46:13

gonna buy as many of these

46:15

anymore. I see. That's why it

46:17

related to an video. So, okay,

46:19

so deep seek, it sounds like

46:21

it's a genuinely innovative, and useful,

46:23

much. more efficient, much less expensive,

46:25

both to use and to build,

46:27

model. Sounds like this is like

46:29

the Leotto mega of language models.

46:31

People are considering this a Sputnik

46:33

kind of moment. Sure. If by

46:35

people you mean Mark Andreessen. Well,

46:37

not much more than those. In

46:39

that it's like a new innovation

46:41

that's not from the US. in

46:43

that we kind of like yeah

46:45

a lot of US companies have

46:48

been like we're the only ones

46:50

in town and then out of

46:52

nowhere someone releases something that's tech

46:54

like better in some circumstances and

46:56

so it freaks everybody out so

46:58

what happens when you take this

47:00

efficient model and put it on

47:02

Google servers that are running like

47:04

a hundred thousand invidia GPUs Like

47:06

would it just be that much

47:08

better? So there is this paradox

47:10

that I cannot remember the name

47:12

of, that people bring up a

47:14

lot in regards to this, where

47:16

if you can run something more

47:18

efficiently and cheaper, more people are

47:20

going to want to use it.

47:22

So it scales with the efficiency.

47:24

So it's like, yeah, if you

47:26

have more servers, you're still going

47:28

to use those servers. you know,

47:30

it'll be cheaper to run. I've

47:32

heard the same thing. But more

47:34

people are going to use it.

47:37

Yeah, there's some paradox around it.

47:39

And that you just like fill

47:41

the demand with their savings. Yeah,

47:43

right. The other thing is that

47:45

theoretically now you could run this

47:47

on like a 40-90 or something.

47:49

You don't need like these super

47:51

computers anymore to be able to

47:53

run these models. And so if

47:55

you're like a university and you've

47:57

got a small cluster, but it's

47:59

not like an open. tailored to

48:01

specific tasks locally on your servers, then

48:03

you don't have to pay open AI

48:05

anymore, right? So this is the other

48:08

thing. Because the weights are open and

48:10

people are trying to re-engineer

48:12

it, and it's free to use, and the

48:14

research papers out there, it's kind

48:16

of open sourcing, this thing that

48:18

these few companies have kept closed

48:20

for a long time because they

48:22

want to maintain this monopolistic

48:25

kind of leadership stance.

48:27

Wow. Yeah. Okay. So that's the first half

48:29

of the new cycle. I can see

48:31

the reason for the hype and like

48:33

all of the movement and stock prices

48:36

and all the headlines and everything. But

48:38

then there was like another half of

48:40

it where it sort of cooled off

48:42

or at least started getting broken

48:44

down and exposed more for what

48:47

is actually happening. Is that also?

48:49

Yeah, I mean, people are finding

48:51

technicalities. So they said that the

48:53

last run that actually trained the

48:56

model. is the thing that cost $5.6

48:58

billion, or yeah, billion dollars. Million,

49:00

yeah, million, 5.6 million dollars, instead

49:02

of from zero. Yeah, instead of from

49:04

scratch. Like all the R&D hours, all

49:06

the other compute, all the data

49:09

collection, obviously that costs a lot more

49:11

money, so it's kind of like the

49:13

same thing as your bill of materials

49:15

thing, right? Yeah. It's like, what did

49:17

this actually cost full on? They

49:19

haven't released that, so it's a

49:21

little bit different. Has anyone asked

49:23

Carl pay? Yeah. Other thing that I

49:25

think is very funny is they're

49:28

basically saying that they are pretty,

49:30

there's this other thing that you

49:32

can do when your training model

49:34

is called distillation, where you basically

49:36

take your model and you take outputs

49:39

from a good model and you

49:41

use that as training data for

49:43

a new model. And they're thinking,

49:45

a lot of people are thinking

49:47

right now that they used outputs

49:49

from like GPT-40 as data. to train

49:51

this model, because you can

49:54

create a much smaller model

49:56

with better data, and if the

49:58

good data is in. the outputs

50:00

from another AI model, then you

50:02

can do that. And so a lot

50:05

of people are freaking out. But also

50:07

a lot of data. Exactly. Nice. So

50:09

a lot of people are freaking out

50:11

because they're like, oh my God, they

50:14

might have stolen open AIs outputs. But

50:16

it's like, boo-hoo. Yeah. I guess that

50:18

makes sense. Because at first when you're

50:20

saying that, I was like, OK, so

50:23

we're just making another GPT4. It's like,

50:25

like, what's the point. But it is

50:27

still cheaper. Giving you the same outputs

50:29

as GPT for it could have advantages.

50:32

So I guess I get it way

50:34

cheaper to run way cheaper to train

50:36

you could train like a a local version,

50:38

you could use your own local version.

50:41

Chain of thought reasoning is really cool.

50:43

Yeah, just a lot of things that

50:45

I think a lot of people were

50:47

not expecting. And then it also made

50:49

it seem like, okay, maybe we don't

50:52

actually need these supercomputer clusters the size

50:54

of Manhattan to be able to run

50:56

these things. Also the nuclear, the entire

50:58

nuclear industry kind of crashed on this

51:00

too, because people were investing in a

51:03

lot of nuclear companies because they were

51:05

thinking. we're going to scale up AI

51:07

clusters so much, we're going to need

51:09

nuclear power to run this. And now

51:11

they're like, actually you need 50 times

51:13

less GPUs than you thought. That's really

51:15

funny. And so now the nuclear industry

51:18

is like, uh-oh. That's really funny. I

51:20

mean, I am not the expert on

51:22

this, but in a far enough future

51:24

that I'm visioning and visioning of like

51:26

people using AI all the time. you

51:28

have to be able to do it

51:30

more efficiently. Like efficiency has to be

51:32

one of the focuses. We can't just

51:34

keep scaling up to just bajillions of

51:36

Japanese everywhere. Sure we can. I mean that's

51:39

what we've been doing. That's what in video once.

51:41

That's what we have been doing. And if you

51:43

follow the graph of a video stock price you

51:45

can see how we've done that. Exactly. But it

51:47

seems like yeah efficiency also has to eventually be

51:49

a focus in some of these. I'm glad that

51:51

it's at least started to at least started to.

51:53

highlight that. I think in most industries and

51:56

markets when more people have access to innovating

51:58

on a thing you get so much

52:00

innovation that the efficiency and

52:02

the value proposition just go

52:04

up like exponentially, but this

52:06

market is so closed. Like

52:08

you only have open AI,

52:10

Google, not even really Apple, you

52:12

know, Claude and Meta. That's sort

52:14

of it. There's a bunch of

52:16

small AI companies that have spun

52:18

up, but they're all basically using

52:20

open AI as their like foundation.

52:22

So yeah, I think this just

52:24

kind of shows that open source

52:26

is going to cause a lot of commotion.

52:29

Yeah, pretty big deal. I'm here for

52:31

it. Pretty big deal. Also, apparently if

52:33

you use the, there is now, there

52:35

is a Deep Seek app that you

52:37

can have that is now the number one

52:39

app on the app store. Sick, that

52:42

was fast. It will apparently not talk

52:44

about Tiananmen Square. Because it's

52:46

still a Chinese app. Yeah. So is it

52:48

going to get banned? That's so many things

52:50

that are unanswered about this. I would not

52:53

doubt it. if that got banned. So yeah,

52:55

if you run it locally, it will tell

52:57

you whatever you want. If you run it through the

52:59

app, which is kind of going through the

53:01

Chinese filtration systems, it will not

53:03

tell you whatever you want. I just want

53:05

to go sign up and it says, due

53:07

to large scale malicious attacks on our servers,

53:10

registration may be busy. Please wait and try

53:12

again. Yeah, they're basically saying they got didost.

53:14

Yeah, everybody's probably. They have a kind of

53:16

a magnifying glass on them right now, so

53:18

that doesn't really shockifying last on

53:20

them right now, so that doesn't really shocked.

53:22

things are happening. So that was a big

53:24

deal. Invideo stock price has since kind of

53:27

gone a little bit back up. There have

53:29

been many think pieces about, well, it costs

53:31

5.3 million to train, but what about all

53:33

the other money and what about this? And

53:35

oh, if they're using distillation of open AI

53:37

data, then the actual true cost is blah

53:39

blah blah blah. And it's like, OK, I feel

53:41

like that's a lot of hand waving personally. I

53:43

don't know. I think it just shows that if

53:45

you make something that if you make something open,

53:47

if you make something open, way more innovation

53:50

can happen. And also we're not

53:52

the only ones in the AI

53:54

race. Uh, pebble. Do you know

53:56

this name? The word pebble. And

53:59

it's not. Not the social

54:01

media network. Pebble brings

54:03

me back, because Pebble's the first

54:05

smartwatch I ever had, technically. It was

54:07

before the Moto 360. I had like a

54:09

red Pebble. I definitely did a

54:11

video on the thing too. A red Pebble with

54:13

a black band. And all it was was an

54:15

ink display on my wrist with notifications. I didn't

54:17

really want it to do too much else. It

54:19

was just so I could have less screen time.

54:22

I could just check my wrist and be like,

54:24

oh, I can see what time it is and

54:26

that I just got a text and I'm not

54:28

even gonna read it and there it is. That's

54:30

my Pebble experience. Yeah, so Pebble was one of

54:32

the most successful kick starters of all time. At

54:34

the time, it was the most. At the time,

54:36

I think it was the most. And then, but

54:38

they eventually went out of business, right? Yeah, okay,

54:41

so they raised a ton of

54:43

money. They raised $10 million at

54:45

their first launch when they expected

54:47

to raise $100 ,000. They came

54:49

really popular with nerds. They released

54:51

a bunch of new Pebbles throughout

54:53

the years. They lasted about four

54:55

years and then eventually Fitbit bought

54:57

Pebble. Fitbit bought Pebble. Fitbit bought

54:59

Pebble and they shut down the

55:01

development of the Pebble 2 and the

55:03

Pebble Time 2 and then they

55:05

refunded everybody. It was a whole thing.

55:07

And then, if you remember, Google

55:09

bought Fitbit. Oh, right. So there's just

55:11

been a bunch of distillation of

55:14

the IP of Pebble and

55:16

not a lot of people have used ePaper

55:18

displays since Pebble. So that's something

55:20

we didn't mention. It is a smartwatch,

55:23

but it is an ePaper display

55:25

smartwatch. Part of the kind of sell

55:27

of Pebble is that it lasted

55:29

seven days. It did very basic stuff.

55:32

It was very nerdy. It was a very

55:34

open ecosystem. You could develop for it. It was

55:36

this awesome community. You could see it in

55:38

sunlight. You could see it in a little

55:40

tiny three -pin charger. It was pretty solid.

55:42

I remember that now. The battery life

55:44

pings so long. Yeah. Now, something you might

55:46

not know is the founder of Pebble

55:48

is Eric Mitjikowski, which is the

55:50

same guy that made Beeper that we

55:52

talked about a number of months ago.

55:54

The whole world. Beeper was that

55:56

app that basically hooked into

55:58

iMessage and kind of - temporarily

56:00

disrupted Apple until they shut it down.

56:03

And I was completely wrong about whether

56:05

or not that would last. But you

56:07

know, here we are. So Eric has been

56:10

apparently rallying Google for a while

56:12

now to open source pebble because

56:14

they have not been doing

56:16

anything with it. Yeah, seems

56:18

like Google forgot. I bet

56:20

Google bought Fitbit and forgot

56:22

that they had pebble stuff

56:24

in there. Yeah, I think

56:27

they also forgot they bought

56:29

Fitbit, but you know, yeah.

56:31

So yeah, he's um, he

56:33

basically was able to rally

56:35

Google to open source all

56:37

of the pebble stuff and

56:39

because of this, he's bringing

56:42

pebble back could be kind

56:44

of neat. Like I don't know, I'm

56:46

still picturing the same general premise, which

56:48

is a smart roch type device on

56:50

my wrist with an e-ink display, a

56:52

week plus battery life, and like a

56:55

decently fast enough refresh that I can

56:57

get notifications, I can swipe them away,

56:59

maybe I can, you know, control media

57:02

on my wrist or something like that, like

57:04

basic stuff like that, I think that

57:06

would still be pretty cool. Yeah. Yeah.

57:08

I think that smart watches generally are

57:10

very overpowered for what a lot of

57:12

people use them for anyway. Yeah. And

57:14

just like on a lot of smartphones,

57:16

people would trade battery life for functionality

57:19

a lot of the time. I love that. So

57:21

Eric is basically starting a new company that is

57:23

going to use the pebble OS, but he

57:25

cannot call the company pebble, because Google

57:27

still owns the trademark. Of course. So

57:29

he basically put out this this website

57:31

with a sign-up forum where you can

57:33

sign up to get updates on when

57:35

new pebble stuff is coming out. Okay, yeah,

57:38

so they probably won't call it

57:40

pebble, but they'll probably have I'm

57:42

predicting now some really punny rip-off

57:44

name from pebble that reminds everyone

57:47

that pebble Existed. Yeah, I believe the the

57:49

website is called Repebble. There you go. So

57:51

there you go. You know, okay. I got

57:54

on a call with Eric a couple of

57:56

days ago to talk about this just to

57:58

kind of do some Q&A. You will be

58:00

able to build on top of the

58:02

new hardware software because pebbles can be

58:05

open sourced and everything that the new

58:07

company builds on top of on top

58:09

of the software they're putting back into the

58:11

open source project Yeah, which is cool, which

58:13

means you can build your own hardware basically

58:15

and like flash it to it That is

58:17

pretty cool. Yeah, so this is kind of

58:20

like an open source device They are going

58:22

to build new hardware, so I would love

58:24

to see what that looks like in 2025

58:26

because like you said the technology has gotten

58:28

much better Yeah, without the flat tire. Well,

58:30

they had the, they had the, they had

58:33

the pebble round. They had the pebble

58:35

round. And it was basically in the

58:37

shape of mode of 360. Bigger, please.

58:39

Bigger screen. Bigger screen. Okay. Other

58:41

cool thing, because it is open

58:44

source, it can run on basically

58:46

anything that has a micro controller.

58:48

So if you want to flash

58:50

your pebble OS to your fridge.

58:52

You could probably do that. Whoa. Someone will

58:55

probably do that. Someone's gonna do that.

58:57

Someone's gonna run doom on their fridge

58:59

on pebble OS, for sure. They are

59:01

planning on sticking to the core feature

59:03

set of the original pebble. Eric says

59:05

that not everyone wants all of the

59:08

features that you're getting from the Apple

59:10

watch and all stuff like that. So

59:12

that's exciting. He put out this blog

59:14

that was about the things that he

59:16

learned from having pebble fail in the first

59:18

place. And someone is working on a port

59:20

right now so that you can develop on

59:22

top of it. So it's pretty exciting. The

59:25

more you talk about this, the more

59:27

I'm picturing, like a pretty simple clean

59:29

smartwatch, like simple watch, simple

59:31

watch, like simple watch face, step tracker, shore,

59:33

maybe heart rate, I don't want too much

59:35

more than that, like keep it easy, keep

59:37

it like a week plus battery life.

59:39

I think a lot of people will

59:42

be into that. And it will probably

59:44

be cheaper because you don't have this

59:46

huge expensive, expensive, expensive LED. display,

59:48

it has to get super bright. Okay,

59:50

I'm in, yeah, I'm in. Eric had

59:52

told me he was going to CES to

59:55

talk to vendors and I was like,

59:57

what for? The vendors? Now I know

59:59

I. Yeah. I'm very curious to see

1:00:01

what it ends up looking like for

1:00:03

people to build on this thing because

1:00:06

the GitHub right now is like all

1:00:08

written in C which is very to me

1:00:10

difficult to understand. So I wonder

1:00:12

if there's going to be like

1:00:14

an easier way to write for this

1:00:16

platform. Yeah. Remember the play date? Yeah.

1:00:18

That thing is like the platform

1:00:21

to develop for it was a. custom

1:00:23

software that they released, right, is super easy

1:00:25

to use. Like a game builder. Yeah. So

1:00:27

I wonder if there's going to be something

1:00:30

like that. That'd be nice. Right now he

1:00:32

only has four part-time employees working on this.

1:00:34

So that's all you need. Yeah. I mean,

1:00:36

it's Google has on it too. Yeah. And

1:00:39

they forgot that those people were still

1:00:41

working for them. Yeah. Yeah, so I'm

1:00:43

very interested to see if the community

1:00:45

that loved Beble in the first place

1:00:47

like builds this back up if we

1:00:49

get some normies on this thing. I

1:00:51

think a lot of people are starting

1:00:53

to feel a little bit overwhelmed with

1:00:55

how much their technology is tracking them

1:00:57

and the overkill of everything right now,

1:00:59

you know, we've innovated past the amount

1:01:01

that we need. I think I think

1:01:03

the reason that Peble initially took off

1:01:05

and had like one of the most

1:01:07

successful Kickstarter ever is going to be even

1:01:09

more true now. I think people will still

1:01:11

like this. What was the reason in

1:01:14

your head? Well, it was a way

1:01:16

to use your phone a little less.

1:01:18

Like you would get the notification on

1:01:20

your wrist so you didn't have to

1:01:23

take your phone out your pocket. And

1:01:25

then the phone, the notification would say

1:01:27

like it's a text from someone or

1:01:29

whatever it's an email or whatever it

1:01:32

is. And then the phone, the notification

1:01:34

would say like it's a text from

1:01:36

someone or it's an email or

1:01:38

whatever it is. 49 smartwatch and

1:01:40

it could do all the things you

1:01:43

want the animations were also very nice.

1:01:45

Yeah Also this came out when the

1:01:47

iPhone 4s was the newest iPhone

1:01:49

so Back yeah throwback that nostalgia

1:01:52

too Yeah, I think so It's the

1:01:54

square one before they got round

1:01:56

and then eventually got

1:01:58

square again. I think 5s was the

1:02:00

first one I ever reviewed yeah

1:02:03

wow 5s well that's to

1:02:05

me that's nostalgia yeah

1:02:07

pre yeah wow anyway about all

1:02:09

for that we'll keep an eye on it

1:02:11

okay cool all right well we

1:02:13

should take another quick break we got

1:02:15

some something or nothing coming up but

1:02:17

before that trivia trivia

1:02:20

I

1:02:22

know

1:02:25

yeah slacking so at

1:02:27

CES 2018 Lenovo

1:02:29

launched a standalone headset

1:02:31

running on Google daydreams platform do

1:02:33

you remember what it was called

1:02:35

but the Lenovo headset was called

1:02:37

no q87 986 x close you

1:02:39

have to a thing pad in

1:02:41

front of the thing pad face the

1:02:44

thing pad elite Facebook face book

1:02:46

damn I have no idea I'll

1:02:48

be totally guessing at that one

1:02:50

yeah answers at the end though

1:02:52

we'll be right back this

1:03:08

episode is brought to you

1:03:10

by indeed when your computer breaks

1:03:12

you don't wait for it to magically

1:03:14

start working again you fix the

1:03:16

problem so why wait to hire

1:03:18

the people your company desperately

1:03:20

needs use indeed sponsor jobs to

1:03:22

hire top talent fast and

1:03:24

even better you only pay for

1:03:26

results there's no need to

1:03:28

wait speed up your hiring

1:03:30

with a $75 sponsored

1:03:32

job credit at indeed com/podcast

1:03:34

terms and conditions apply

1:03:37

welcome back so now we are

1:03:39

going to do a game that we call

1:03:41

something or nothing let's play a game let's

1:03:43

play game so the rules are simple I'm

1:03:45

gonna read a headline and you guys have

1:03:47

to tell me is this something that we

1:03:49

should care about or is this nothing

1:03:51

at all I got you yeah nothing

1:03:54

at all nothing at all all right

1:03:56

first at all iPhone se4 leak

1:03:58

shows single and no

1:04:00

dynamic island. Is this something

1:04:02

or is it nothing? Nothing. Nothing.

1:04:04

I'm going with nothing. It's like, the

1:04:06

one thing that's interesting about it is

1:04:09

it, usually the SE is in the

1:04:11

body of an old iPhone that's already

1:04:13

happened. Yeah. Which is why they can

1:04:15

make it so cheap. It's because it's

1:04:18

the part that they make already.

1:04:20

I agree. This one, slightly air quotes

1:04:22

new design, which is like, oh, they're

1:04:24

making like a new part for this thing.

1:04:26

Other than that. Single camera does not surprise

1:04:29

me. No dynamic rain. No dynamic island does

1:04:31

not surprise me. Yeah. So I'm going nothing.

1:04:33

Yeah, they're keeping the dynamic island for the

1:04:35

nicer phones. The SE is the phone I

1:04:38

was by my mom every seven years. So

1:04:40

I don't think she's gonna worry about

1:04:42

the dynamic island. You guys don't think

1:04:44

it looks beautiful? The dynamic

1:04:46

island? No. This is like a I

1:04:49

like how it looks. Yeah, there's

1:04:51

a week five almost. I feel

1:04:53

like this is gonna be Ellis's

1:04:55

next phone. Ohh. Mm. That's oh

1:04:57

because it's small because it's small

1:05:00

because it's small. It's not even

1:05:02

that small. It's not even that

1:05:04

small. It's not even that

1:05:06

small. Yeah, you're right. Isn't like

1:05:09

a six inch screen. It kind of

1:05:11

looks like an X-R. You mean

1:05:13

tenor? The tenor what they call

1:05:15

that could be cool. but still

1:05:17

probably still nothing. Still

1:05:20

nothing. All right. Next up,

1:05:22

Open AI launches GPT4 government

1:05:24

edition. Less than nothing. Is

1:05:27

this just gonna be a

1:05:29

repackaged GPT that's more expensive

1:05:32

because it's government?

1:05:34

Yeah. Cool. Cool. I think

1:05:36

it's also like specifically tuned

1:05:39

to government agencies. I think

1:05:41

it's just ways for Open AI to

1:05:43

milk the government. Yeah, so. The top

1:05:45

comment on the virtual article is, why

1:05:48

not Deep Seek? It's open source. And

1:05:50

the top replies, yeah, it's gonna be

1:05:52

banned. Yeah, it'll probably be banned within

1:05:55

a week. What's the over under that

1:05:57

Deep Seek is banned in a week?

1:05:59

No. It can't get it doesn't have

1:06:01

the clout yet for like that

1:06:03

guy in the Oval Office to do

1:06:06

anything about it It took like

1:06:08

a trillion dollars off the stock. That's

1:06:10

actually very Totally fair. I do

1:06:12

think that yeah, I think they need

1:06:14

Congress to be able to actually

1:06:16

ban it But I feel like they're

1:06:18

gonna ban it wait. We'll get

1:06:20

lemonade first Which will be banned first

1:06:22

lemonade 100 % deep -seek. No, nobody

1:06:25

remembers the lemonade. I got 20 ,000

1:06:27

people over there on lemonade Actually

1:06:30

true. I got 1700 on pixel -fed

1:06:32

You know what? I looked I looked

1:06:34

I have the 18th most popular photo

1:06:36

ever on pixel -fed. Wow What is it

1:06:38

a good photo? What is it? Oh,

1:06:40

I would say It's of Glacier National

1:06:42

Park. Oh, yeah, you can even break

1:06:44

top 10. I'm so disappointed. Wow ever

1:06:46

Adam I made my account two weeks

1:06:48

ago. No, a flex. That's a flex.

1:06:50

Yeah, let me fire. I respect that

1:06:53

anyway GPT for government edition. I'm going

1:06:55

nothing nothing nothing. All right. Yeah next

1:06:57

up Threads is officially getting ads something

1:06:59

everything everything Everything I called it. Didn't

1:07:01

I tell you guys like obviously threads

1:07:03

had their like, yeah you know moment

1:07:05

of Exploding onto the scene hundred million

1:07:07

users in five days, whatever it was

1:07:09

like everyone's Like the headlines are all

1:07:11

like this is gonna overtake Twitter. This

1:07:13

is the thing We finally have a

1:07:15

competitor and it's meta so meta's sneakily

1:07:18

in the background like yes Yes, come

1:07:20

to thread that doesn't add business, but

1:07:22

eventually yeah It's meta like we knew

1:07:24

that they were eventually gonna do ads

1:07:26

So we knew they were building in

1:07:28

the background waiting to find a good

1:07:30

time to turn it on and they're

1:07:32

gonna turn it on And it's gonna

1:07:34

be another meta property with ads and

1:07:36

what else is new. Yeah, but I'm

1:07:38

saying it's everything It is everything because

1:07:40

this is this is what they do

1:07:42

Yeah, and now they've acquired a lot

1:07:45

of users and then people will migrate

1:07:47

to blue sky, right? That's the next

1:07:49

move. Come on baby because blue sky

1:07:51

now think about blue sky 30 million

1:07:53

yesterday Just saying they pass 30 million

1:07:55

users blue sky. Well blue sky get

1:07:57

ads. Um, I have no idea 100%

1:08:00

because they are run by a

1:08:02

company. Obviously, they're run on that

1:08:04

open protocol, but they're still run

1:08:06

by a company. Correct. Service costs

1:08:08

money. Service costs money. These employees

1:08:10

cost money. So just blues guy

1:08:12

has to do that. You know,

1:08:14

blues guy doesn't even use AWS.

1:08:16

They run their own servers. That's

1:08:18

expensive. It's actually cheaper than AWS,

1:08:20

which is with their current value,

1:08:22

because AWS is expensive. I guess it

1:08:24

sells up with volume. So if you're.

1:08:27

30 million users are active on this thing,

1:08:29

it costs a lot. And they charge you

1:08:31

for convenience. But you still have to buy

1:08:33

those servers and like run over this, it

1:08:35

costs money. It costs money, but it is

1:08:38

technically cheaper. But yeah, I have been talking to

1:08:40

the Blue Sky Crew a lot recently

1:08:42

and they said they are thinking about the

1:08:44

right time to move the AT protocol

1:08:46

into the internet engineering task force. Which

1:08:49

if you watched our, hey, if you

1:08:51

watched our special episode, we all just

1:08:53

glazed over. Come on, secret history of

1:08:55

the internet, the IETF, they're the same

1:08:58

guys that do email. Shout out to

1:09:00

them. They have one of the nerdiest

1:09:02

sounding names of all time, but

1:09:04

I appreciate that. Daddy Vint. Yeah,

1:09:07

Daddy Internet. I was emailing with

1:09:09

Vint this weekend. It was crazy.

1:09:11

It was crazy. I was like,

1:09:13

like, I draw access it. Vint.

1:09:16

You mean Big V. Internet Zati.

1:09:18

Our father. Anyway, yeah, so Blue

1:09:20

Sky Public PLC wants to eventually

1:09:23

move the AT protocol into a

1:09:25

standards body because they want to

1:09:27

instill that trust because currently Blue

1:09:29

Sky PLC is the same company

1:09:32

that runs Blue Sky the platform

1:09:34

and the whole idea is that, you

1:09:36

know, they are not the only one

1:09:38

on the platform on the

1:09:40

protocol. So next one. Everything.

1:09:43

Everything. Everything. It is something

1:09:45

for sure. It is something in

1:09:47

the way meta moves. Do you think

1:09:49

this will matter? What ads? Yeah.

1:09:51

When ads hit, there will be

1:09:53

a another moment, another reckoning for

1:09:55

everyone on threads who has been

1:09:58

preaching the word of threads. and

1:10:00

I'll have to decide if they're cool with

1:10:02

it or not. Yeah, but most people haven't

1:10:04

been preaching to word of threats. They've just

1:10:06

been on Instagram and then they're like, what's

1:10:08

this? Click the button and now you have

1:10:10

a threat to count. Adam, you're gonna start

1:10:12

me ranting again. Well, there's that movement. There's

1:10:14

also the, I'm leaving Twitter movement. Yeah. And

1:10:16

then. When it gets ads, they have to

1:10:18

be like, I'm cool with that. That's

1:10:20

not what I hated about Twitter. That's

1:10:22

exactly my point. Everyone is used to

1:10:24

ads already. Yeah, exactly. Maybe it's nothing.

1:10:26

Maybe this was going to happen the

1:10:28

whole time. And it's like, it's just

1:10:30

going to be background. It's everything for

1:10:32

meta, because it'll probably boost their

1:10:34

revenue by a lot. Will it? Yeah. They

1:10:36

have WhatsApp and Instagram with two billion users.

1:10:38

Yeah. But a cool, 200, 200, 200, 200

1:10:41

million more isn't that bad more isn't that

1:10:43

bad. Yeah. Yeah. Yeah. All you need to

1:10:45

do, Blue Sky, if you want to make

1:10:47

all that money back, make very highly

1:10:49

customizable ads. Highly customizable profiles, where you

1:10:51

can have like a pink background and

1:10:54

a song that auto plays when you

1:10:56

go to my profile. Nice, I love

1:10:58

it. Next on the list is open

1:11:00

AI's new operator AI agent can do

1:11:02

things on the web for you So

1:11:05

did you read these articles about this?

1:11:07

No, I did not. Okay. I have

1:11:09

an understanding of what it sounds like

1:11:11

it's doing based on a headline But

1:11:13

yeah, maybe is there something deeper? Well,

1:11:16

okay. So operator is this new feature

1:11:18

that they have with the expensive model

1:11:20

that they have and basically the 200 dollar

1:11:22

one. Yeah, that's crazy. Basically, it's in research

1:11:25

preview mode right now. The idea is that

1:11:27

you can say like, I want to go

1:11:29

on a vacation. Can you book it for

1:11:31

me? And it'll go to like trip advisor

1:11:34

and then it'll. I still think this would

1:11:36

be a banger video. We should do that.

1:11:38

Having AI like a full trip

1:11:40

for you. I book a trip

1:11:42

to like Boston or somewhere nearby

1:11:44

and follow the itinerary. Yeah. So

1:11:46

they have like it's very weird.

1:11:48

They have specific partnerships right now

1:11:51

where it'll use specific websites and

1:11:53

services for all of these things

1:11:55

that it wants to do for

1:11:57

you, which feels very rabbit R1E

1:11:59

to me. if I do say

1:12:01

so myself. What did they call those

1:12:03

agents? Agents, yeah. Oh, same thing. Yeah.

1:12:05

It's basically an agent that goes and

1:12:08

does things for you, but you can

1:12:10

see it like clicking around. It was

1:12:12

funny, I believe Casey Newton did this,

1:12:15

and it booked it for like the

1:12:17

completely wrong date. Nice. Something

1:12:19

like that. And everyone that

1:12:21

has been using it, that has been

1:12:24

talking about it on Reddit, it

1:12:26

says that it's really bad. Like

1:12:28

it's, it barely works. doesn't work,

1:12:30

or if it works as well as rabbits, which

1:12:32

is not at all. Which I mean, okay,

1:12:34

it's opening eyes, so maybe I have to

1:12:37

give them benefit of the doubt, but like,

1:12:39

I don't do, I don't know. I also

1:12:41

like, I am, I am heavy, me personally

1:12:44

in the camp of I want to overthink

1:12:46

each individual detail of the things that I'm

1:12:48

doing. So instead of asking an AI agent

1:12:50

to buy something online for me. I want

1:12:52

to just double check it's going to

1:12:55

the right address and that I will

1:12:57

have the right shipping speed and all

1:12:59

that stuff. And if I'm booking a trip,

1:13:01

that's even more. Yeah, no way.

1:13:03

So I tend to think it's already

1:13:05

easy enough to do most of these

1:13:07

things online. The capabilities and the trust

1:13:10

level have to have to like merge

1:13:12

and both of those things, like the

1:13:14

capabilities are moving at like one mile

1:13:16

an hour right now and the trust

1:13:18

levels are like going the other direction.

1:13:21

Yeah. currently it's kind of kind of

1:13:23

nothing and I also don't really think

1:13:25

that this is the way that we're going

1:13:27

to be using AI agents in the future

1:13:29

so like Mark has said like I think

1:13:31

people just want to make their own

1:13:33

decisions I can just book a flight let

1:13:36

me look at the flights too like instead

1:13:38

of me going hey book a flight to

1:13:40

Boston for me for this weekend I want to

1:13:42

look at all the flights pick a time yeah

1:13:44

Which will, the one at noon is actually cheaper

1:13:46

than the one at, at 3 p.m. So let

1:13:49

me do that one instead. Now I'm gonna plan

1:13:51

something to do when I land at this time

1:13:53

because it's earlier, like I wanna, I wanna do

1:13:55

that. Yeah, if you can, if you can do

1:13:57

things like check the available flights and give me

1:13:59

off. Quicker than I could do it myself then that's

1:14:01

cool. Yeah, you know There's probably CEOs and stuff that are

1:14:03

like looking at the $200 a month price And they're

1:14:05

like yeah like book me a flight just do it I

1:14:08

don't care which one and they'll just do it and

1:14:10

they don't think about it And then I'll

1:14:12

get to the airport and they're like sir,

1:14:14

you're going to Paraguay And it's in three

1:14:16

months Yeah,

1:14:20

yeah, yeah, I'm going

1:14:22

with nothing for now with

1:14:24

nothing for now interesting

1:14:26

okay next on the list

1:14:28

iOS 18 .3 officially launches

1:14:30

which makes notification summaries

1:14:32

italic If this

1:14:34

ain't something I don't know what that's not

1:14:36

the official headline. That's what I wrote

1:14:38

oh Basically Apple was getting a lot of

1:14:41

flak because the notification summaries were saying

1:14:43

Crazy stuff and it just lying a lot

1:14:45

of the time because it was incorrect

1:14:47

about a lot of things and trying to

1:14:49

especially because it was trying to Group

1:14:51

multiple things into one sentence like multiple different

1:14:53

events into one sentence and

1:14:57

now When it has a

1:14:59

notification summary it is italic to

1:15:01

signify to the user that it's

1:15:03

in beta or whatever something

1:15:07

Really, I also think it's something shows

1:15:09

that Apple's listening to something They

1:15:12

just made it italic True

1:15:15

Do you think okay? This is my question

1:15:17

I have not talked to like a

1:15:19

normie and been like hey your

1:15:21

notification summaries Is it

1:15:23

more obvious that it's an

1:15:25

AI guess now because it's italic

1:15:29

because I Actually hate how it

1:15:31

looks now. Yeah, the italic look is

1:15:33

not not a good look It's like

1:15:35

it's it so it has to just

1:15:37

look different from a regular notification. That's

1:15:39

the point. Yeah, italic, right? So okay

1:15:41

Apple understands that

1:15:43

people have seen these AI

1:15:45

summaries and don't like them

1:15:48

and if we make them

1:15:50

look visually Distinct from a

1:15:52

normal text then or a

1:15:54

normal notification then that's better

1:15:56

than not. Yeah, so

1:15:59

that's That is something I

1:16:01

feel like it's something I feel like

1:16:03

it's something because this is a red

1:16:05

flag for just Apple intelligence in general

1:16:08

Yeah, I think yeah, summerizations are supposed

1:16:10

to be the easiest thing that AIs

1:16:12

can do right like hey take this

1:16:14

thing and summarize it Yeah, and this

1:16:17

isn't even the full Apple intelligence yet

1:16:19

like it's still rolling out. Yeah, so

1:16:21

yeah, if they're having problems with this,

1:16:23

how do I trust them to be

1:16:26

able to actually do a thing? in

1:16:28

three months. I really want them to

1:16:30

do that reach into apps thing. That's

1:16:32

supposed to be March. Yeah, I don't

1:16:35

even know if there's problems with this.

1:16:37

It's just not a useful app, not

1:16:39

a useful place to deploy summaries. Like

1:16:41

when I get a notification of like

1:16:44

one single email, I don't need you

1:16:46

to summarize the email. I'll just look

1:16:48

at the subject line. Like that's already

1:16:50

good enough. When I get three texts

1:16:53

in a chat with someone. I

1:16:55

don't know, it's three texts, like maybe

1:16:57

a summary is useful, but all I

1:16:59

really need to know is I have

1:17:01

three texts from this person. Context in

1:17:03

my brain is like, oh, I remember

1:17:05

what I was talking about with them.

1:17:08

That's the summary. I think there's some

1:17:10

value to like the urgency of the

1:17:12

text, you know what I mean? Like

1:17:14

if it says like, urgent request needed,

1:17:16

then I'm more likely to stop what

1:17:18

I'm doing and go deal with the

1:17:20

text than I am. Notifications are inherently

1:17:22

useful though. Like yeah we get a

1:17:24

lot of them and they're annoying but

1:17:26

they're there for a reason. So I

1:17:28

feel like adding this layer of friction

1:17:30

in front of the notification to like

1:17:33

summarize it that may or may not

1:17:35

be accurate is not really what people

1:17:37

want. They are funny though. To go

1:17:39

into these. Yeah they are tells there.

1:17:41

If I could just go and open

1:17:43

the app or like swipe down a

1:17:45

control center or something and then. choose

1:17:47

a summary notification than sure. But to

1:17:49

offer it like right off the bat

1:17:51

by default, it's like just another step.

1:17:53

Apple only really added AI do a

1:17:55

piece of shareholders anyway. Bar. Bar. This

1:17:58

is nothing. This is just more of.

1:18:00

of the nothing. Yeah, more of. Apple

1:18:02

intelligence is nothing. You know, you

1:18:04

know what happened? I tried

1:18:07

to use Apple intelligence on

1:18:09

on my Mac when it launched on

1:18:11

Mac and I used it on a on

1:18:13

an Apple note that was a whole

1:18:15

script of a video I had written

1:18:18

and it summarized it and then

1:18:20

I hit control command Z

1:18:22

because it was like, oh,

1:18:24

and I was just testing

1:18:26

the summarization feature. And then it

1:18:29

crashed. And then when I reopened

1:18:31

it, I could no longer command

1:18:33

see it. No, I lost the

1:18:35

entire script. Damn. So that was

1:18:37

fun. Damn. Don't test in production,

1:18:40

my friend. Oh my God.

1:18:42

Bad thing. So I should have copied

1:18:44

that no, but that's on me. Anyway,

1:18:46

is it? Yeah, no, it's not on

1:18:49

you. I'm gonna blame it on Apple.

1:18:51

It's on them. AI. is bad. Okay.

1:18:53

Intelligence is bad. Next up, David did

1:18:55

not win a ticket to the switch

1:18:57

to showcase in April and neither did

1:19:00

Adam. Everything. Yeah. This is everything. I

1:19:02

agree. Switch is so hype. Yep. Switch

1:19:04

is so hype. So hype. Dude. You

1:19:07

had to win a ticket to look

1:19:09

at it? Yeah. Right. Correct. And you

1:19:11

say you applied? Yep. For a chance.

1:19:14

Three chances to look. at the thing

1:19:16

that you might give the money for

1:19:18

later. No, no, I will give the

1:19:20

money for it. This is crazy. Life

1:19:23

is out of control. Dude, it is out

1:19:25

of control. Okay, so now that you're not

1:19:27

able to look at it, you're still gonna

1:19:29

buy it anyway. Yeah, probably. So you just

1:19:32

want to look at it just to look

1:19:34

at it. Yeah, so they basically, they were

1:19:36

gonna have three different showcase events where you

1:19:38

could like. play with it early but they had

1:19:40

to ticket it and they had to have a

1:19:43

raffle and so they had multiple sessions for New

1:19:45

York and Austin and Los Angeles wait are

1:19:47

you is this paid no no okay it's

1:19:49

a free event but this is just a

1:19:52

hype site this isn't this isn't Nintendo making

1:19:54

more money this is them just driving more

1:19:56

hype yeah people will show up with a

1:19:58

phone and make a 30-second video of it

1:20:01

on their phone and put it on

1:20:03

YouTube and shorts and tic-toc and like

1:20:05

look at the switch and it'll be

1:20:07

like oh my god I want the

1:20:09

switch and that'll be it yeah yeah

1:20:11

I'm taking my content strategy no that's

1:20:13

smart that's smart yeah I'm

1:20:15

tendo smart everything and everything

1:20:17

and the second and the

1:20:19

last one booms xb1 supersonic

1:20:21

passenger jet goes mock 1.1 I put

1:20:24

this in here. Yeah, so I'm assuming

1:20:26

you think it's something, I think it's

1:20:28

nothing. Oh boy, this is something. The

1:20:30

return of the, what was it called?

1:20:32

The Concord. The Concord. If you have

1:20:34

time for a tiny story time, I

1:20:36

can tell you the story. We have

1:20:38

nothing but time. Okay, so passenger

1:20:40

jets. And that's it. Thanks for

1:20:42

watching. No, the Concord was like

1:20:44

that super, super fast, super sonic

1:20:46

plane that did like cross country

1:20:49

and transatlantic flights in like no

1:20:51

time, because it was. twice as almost,

1:20:53

I think it was twice as fast.

1:20:55

It was much faster than it was,

1:20:57

so it was basically twice as fast

1:20:59

as normal passenger jet, much fast than

1:21:01

the speed of sound. The speed of

1:21:04

sound is like this benchmark of the

1:21:06

obviously supersonic travel. It's a benchmark for

1:21:08

speed for planes. Perfect. It's like 800

1:21:10

miles an hour, right? So the Concord that

1:21:12

crashed, you know, sort of retired that plane

1:21:14

and we went back to not supersonic travel.

1:21:17

But the thing that really hit me about

1:21:19

this about this months ago, but. What did

1:21:21

you used to do on a plane before

1:21:23

the Concord when you had a

1:21:26

six-hour cross country or transcontinental flight?

1:21:28

Nothing. You just sat there. There is

1:21:30

no internet. There is no TV. There

1:21:32

is nothing to do. You just sat

1:21:34

there. And so you were highly incentivized

1:21:36

to get the faster flight so you

1:21:38

could save more time, get to your

1:21:40

destination, and then start doing things again.

1:21:42

Now, if you have a six, seven,

1:21:44

eight-hour flight. you're on the internet, you're on

1:21:47

your phone, you can do email, you can watch

1:21:49

a movie, there's a whole bunch of stuff you

1:21:51

can do. So people kind of stop caring about

1:21:53

getting there twice as fast, and they have

1:21:55

slowly, gradually ramped up this like prototype of,

1:21:57

okay, we can go faster and faster than

1:21:59

they. got to like just under mock speed

1:22:01

and they just did a test run where

1:22:03

they went mock 1.1 so over the speed

1:22:05

of sound yeah their eventual goal is something

1:22:08

like mock 1.7 they want to be able

1:22:10

to go twice as fast as passenger jets

1:22:12

basically you can go anywhere in the world

1:22:14

from anywhere in the world up to 5,000

1:22:16

miles in like four hours now it's gonna

1:22:18

be more expensive it's the first one

1:22:20

it's gonna be pricey I think United

1:22:22

bought like 15 of these things and

1:22:24

they're gonna slowly start making them though

1:22:26

This is still in like the testing

1:22:28

phase of like they're okay validating we

1:22:30

can go mock 1.1 We still got

1:22:33

to go faster and make sure

1:22:35

it's safe But we have purchased

1:22:37

orders and when we start making

1:22:39

these bajillion dollar jets Yeah people

1:22:41

will be able to buy a

1:22:43

ticket and go from New York to

1:22:45

London in three and a half hours.

1:22:47

It's sick. Okay, but one thing we

1:22:49

live we record right next to an

1:22:51

airport. Yeah, when you go faster than

1:22:53

the speed of sound. Not ideal. One

1:22:55

of the downsides. Okay, but there

1:22:58

are certain slots of airspace that

1:23:00

are approved for supersonic travel. So

1:23:02

over the ocean, for example, or in

1:23:04

the corridor for going from New York

1:23:06

to London or certain paths where you

1:23:08

used to be able to go supersonic,

1:23:11

this is, they did this testing in

1:23:13

one of those corridors. So there are

1:23:15

places where it is acceptable to go

1:23:17

supersonic. Okay. And so I think the

1:23:19

idea of being able to get anywhere

1:23:22

in no time and probably also still

1:23:24

be scrolling your phone on the way

1:23:26

I think is pretty sick. Yeah, that's

1:23:28

why it's something to get to the

1:23:30

West Coast in two hours. Think about

1:23:33

that. That'd be crazy. Think about that.

1:23:35

The flight to California is typically five

1:23:37

and a half hours for us. Yeah.

1:23:39

What if it was two? In 20 years. Well,

1:23:41

when they finish making the jets and they start

1:23:43

shipping them and the tickets come down to price

1:23:45

because they've got a bunch of jobs out there.

1:23:47

So that's 60 years. So that's 60 years. And

1:23:49

then, yeah. I'm not getting on that first flight

1:23:51

is what I'm saying. No, oh, no, no, no,

1:23:54

no. Nobody's on the first flight. I don't want

1:23:56

to be, none of you are allowed to be

1:23:58

on the first. But the idea is it will

1:24:00

be. something in 10, 20 years when we

1:24:02

are back to these super sick. Like the

1:24:04

jets we fly on now are so old.

1:24:06

Yeah. They're not even necessarily old. They're

1:24:08

just the same technology as forever ago

1:24:10

and we don't really, we're more concerned

1:24:13

with like, I mean we're cramped back

1:24:15

there, the legroom's not that great, the

1:24:17

internet's not that great, the TV is

1:24:19

not that great. Yeah. Like we would

1:24:21

like to do some upgrading. So I

1:24:23

think it's something, I'm excited. Why not

1:24:25

just make trains. Because it would

1:24:27

take oil. I have to train

1:24:30

to California channel my inner Ellis

1:24:32

here and say bullet trains exist.

1:24:34

These are different bullet trains exist

1:24:37

to certain land to land destinations.

1:24:39

Yes. Not in the US. But

1:24:41

not to London. Not yet. Last

1:24:44

rate to London. Now we're bad

1:24:46

attitude. Big tunnel. Big tunnel. And

1:24:48

you can put your Tesla on a

1:24:51

little track. Yes. And then it'll go,

1:24:53

it'll, it'll, moo. There can be, all

1:24:55

I'm saying is there can be tears

1:24:57

of travel, or like, there will still

1:25:00

be regular flights like this, but there

1:25:02

will also be, this will be an

1:25:04

expensive ticket, and this will be

1:25:06

like, okay, you splurge, you take

1:25:08

the bullet train instead of the

1:25:10

regular train, and you get there in

1:25:12

half the time. And it's just an

1:25:15

available option. I have a proposition for

1:25:17

you. They call the rich people on

1:25:19

the Concord. They come to you and

1:25:21

they say, hey. You can be the first person

1:25:23

on this flight and get exclusive video rights

1:25:26

on everything. I don't want to be first.

1:25:28

I don't want to be first. I will

1:25:30

I so so they're doing these test flights.

1:25:32

They're there. It's like a Navy pilot and

1:25:34

like they're going in this specific

1:25:37

specific corridor and like slowly ramping up in

1:25:39

speed. That's all great. Don't put me on

1:25:41

those. Then they're going to get to like

1:25:43

mock 1.7 and they're going to be like

1:25:45

okay. We can do this at this at

1:25:47

this speed and it's safe. Now we can it's

1:25:49

almost like car testing like when you

1:25:52

see the top speed runs from like

1:25:54

the Hennessy venom F5 or some random

1:25:56

car that's like you have to do the top

1:25:58

speed run in both directions with all

1:26:00

the safety equipment and the professional driver, once

1:26:02

you get past that stage, then you can

1:26:04

ship them to regular people and regular drivers.

1:26:07

And so you can put real passengers on

1:26:09

it and not worry too much. By the

1:26:11

way, I'm just tangent city. These bowings can

1:26:13

go faster, just like our cars can go

1:26:15

faster. Yeah, but they're just, they operate at

1:26:17

a certain speed in certain jet ways because

1:26:20

it's most efficient for fuel. Yeah. And sometimes

1:26:22

I'll get on a plane. This thing is

1:26:24

just gonna say screw it, screw it. and

1:26:26

just book it. No, it's going to be

1:26:28

able to go even faster, but it's going

1:26:31

to go at Mach 1.7 because that's what's

1:26:33

safe. Mock 4, let's go. Sometimes I get

1:26:35

on a flight and we'll be like 30

1:26:37

minutes late from the gate and the pilot

1:26:39

will go, we're going to try to make

1:26:41

up some extra time in the air. And

1:26:44

you're going to try to make up some

1:26:46

extra time in the air. And you're like,

1:26:48

we're going to try to make up a

1:26:50

jet stream, some down, some down, and we're

1:26:52

actually, and we're actually, and we're actually, and

1:26:55

we're actually, we're actually, we're actually going to

1:26:57

get you, we're actually going to get you,

1:26:59

and we're actually going to get you, and

1:27:01

we're actually going to get you, we're actually

1:27:03

going to get you, we're actually going to

1:27:06

get you, we're actually going to get you,

1:27:08

we're going to get you, we're going to

1:27:10

get you, we're going to get you, but

1:27:12

going faster in any vehicle takes more fuel.

1:27:14

So they have to be more efficient. So

1:27:16

this is a plane that's gonna be capable

1:27:19

of going like, whatever, Mach 2 or something,

1:27:21

and it's gonna go Mach 1.7, and it's

1:27:23

gonna be great. Well, I for one, I'm

1:27:25

excited to be one of the first to

1:27:27

get on one of the super sonic jets

1:27:30

from boom technologies. Boom XB1. Come into an

1:27:32

airport near you. Hard pass. We'll see. We'll

1:27:34

see. We'll see. Okay. That's all I got.

1:27:36

Figure out those trivia questions. Figure out. I

1:27:38

think you mean... Where are you going with

1:27:40

this? Hell Adam, the answers I already know.

1:27:43

I think I have one of them. I

1:27:45

don't know. Question one. David Kaz and Damien

1:27:47

Henri were two French Google engineers with their

1:27:49

20% innovation time off project. They made something

1:27:51

that was introduced at Google IO 2014 developers

1:27:54

conference. In fact, all attendees walked away with

1:27:56

one of these things. What was it? all

1:28:03

attendees um

1:28:05

well I wasn't there so I can't

1:28:08

confirm that all of them walked out

1:28:10

with it but did they get you

1:28:12

freaking nexus devices that I know yeah

1:28:14

now you got like a water bottle

1:28:16

and a t -shirt yeah disgusting

1:28:18

although

1:28:21

pretty great water bottle two and a half trillion

1:28:23

dollar company they give me a water bottle

1:28:25

and the same one as last year to flip

1:28:27

them and read what do we got I

1:28:31

wrote Google cardboard but I added

1:28:33

a diagram to the bottom I

1:28:35

also added Google cardboard and added

1:28:37

a diagram to the bottom so

1:28:40

we should go by whose diagram

1:28:42

is better well first of all are we

1:28:44

right I was gonna say you are both okay

1:28:46

good right and I

1:28:49

have three dimensions to my

1:28:51

diagram oh okay though

1:28:53

oh Marquez wins because come on our

1:28:55

3d you need three dimensions hold on

1:28:57

here although cardboard is 3d now

1:28:59

suckers was literally just a cardboard box

1:29:01

you put your phone in but

1:29:03

you you have to fold it yourself

1:29:05

that was what was sick about

1:29:07

it yeah it was pretty and it

1:29:09

wasn't I think like it was a

1:29:12

Verizon promo or something like somebody did

1:29:14

it with the packaging of the box

1:29:16

that you're new like nexus device came

1:29:18

in what was cardboard that's interesting yeah

1:29:20

that's hilarious oh yeah that's right yeah

1:29:22

yeah yeah it was awesome all right

1:29:24

next question okay going along

1:29:26

the same theme at CES 2018

1:29:28

Lenovo launched a standalone

1:29:30

headset running on Google

1:29:32

Daydreams platform what was it

1:29:34

called I

1:29:36

don't know if you could tell but I have

1:29:39

complete faith that Android XR is gonna be a

1:29:41

thing that's why I keep bringing these review

1:29:43

questions I'll give

1:29:45

you a

1:29:47

hint it was

1:29:49

the Lenovo

1:29:52

Mirage what Mirage

1:29:54

what there's

1:29:56

another word there

1:30:00

Mirage face headset. I don't know.

1:30:02

Flipperman Reed, what do we got?

1:30:04

The Lenovo Mirage boom. Dang it.

1:30:06

Foot air. Mirage air. Marquess just

1:30:09

wants to fly in that airplane.

1:30:11

I can tell. The correct answer

1:30:13

was the Lenovo Mirage solo. Cool.

1:30:15

Wow, that's depressing. Yeah, it's a

1:30:17

little sad. Yep. Because you're the

1:30:19

only one who can see the

1:30:21

thing. You want to be more

1:30:23

lonely? They were the only ones

1:30:25

that we're realistic about it. That's

1:30:27

true. I'm solo. Yeah, this is

1:30:29

the solo. So, you know what

1:30:32

you're getting into. No friends. This

1:30:34

is like the Aesus Republic of

1:30:36

gamers. You have a lot of

1:30:38

friends now. Yeah, that's what it's

1:30:40

called. Republic. Republic. Democratic Republic of

1:30:42

gamers. Well, I'm glad I got

1:30:44

a plan out of that. Yeah,

1:30:46

that was fun. Hey, let us

1:30:48

know in the comments. Would you

1:30:50

get on a flight? That goes

1:30:52

faster than the speed of sound.

1:30:54

If the company was called boom.

1:30:57

If they offered you a flight,

1:30:59

if you're on the first one,

1:31:01

it's free, it's free. You're on

1:31:03

the first plane and they go,

1:31:05

hey, pick a destination, we'll fly

1:31:07

you there for four years. Oh

1:31:09

my gosh. What? I just remember

1:31:11

to dream I had like a

1:31:13

two weeks ago. Was it flight

1:31:15

related? No, we'll eat, kind of.

1:31:17

I had a flight related dream

1:31:20

recently. It was about going to

1:31:22

the moon. And when it came

1:31:24

up to my turn, I had

1:31:26

like the worst nerves ever and

1:31:28

something went wrong with the system

1:31:30

and I was like, I'm fucking

1:31:32

out of you. You are not

1:31:34

putting me on the, they're like,

1:31:36

yeah, we had some engineering errors,

1:31:38

it's gonna be an extra 10

1:31:40

minutes, I was like, nah! That's

1:31:42

a extra 10 minutes, I was

1:31:45

like, nah! That's a fair choice.

1:31:47

But Brandon was already in space.

1:31:49

Oh no. All right, all our

1:31:51

witches in the comments, please let

1:31:53

us, please let us, let us

1:31:55

know. We've heard produced by Adam

1:31:57

Malia. We're part of the Vox

1:31:59

Media Podcast Network and our intro-autro

1:32:01

music is Bye Vane. And so

1:32:06