0:00

There's so much sex appeal to building new

0:02

things. And in 10 years, we'll have this

0:04

really cool, new autonomous thing, drone, walking humanoid,

0:06

that's going to solve all of these problems.

0:08

But the problem is, there's a lot of

0:10

issues going on today. And so the approach

0:13

to solving and using specific robots for specific

0:15

jobs is actually just to earn the right

0:17

to begin building really cool robots that are

0:19

able to do more interesting things. But you

0:21

got to get the business model right. And

0:24

the business model has to incentivize and make

0:26

a CEO or CFO give a f*** about

0:29

how useful these industry 4.0 principles

0:31

and tools are. Because right now, that's not

0:33

true. I see this time and time again,

0:35

where like, I won't name the AI companies,

0:37

but these AI companies come in and say,

0:39

well, completely turn on your head the way

0:41

you're operating your entire business. And they'll come

0:43

in for some contract that ends up expiring

0:46

because it just did not produce. And that's

0:48

the problem. You think you have all the

0:50

information and data, but you're building your AI

0:52

and your solutions off of ground truth that's

0:54

actually not ground truth. This

0:57

Week in Startups is brought to you by Vanta.

1:00

Compliance and security shouldn't be a

1:02

deal breaker for startups to win

1:04

new business. Vanta makes it easy

1:06

for companies to get a SOC

1:09

2 report fast. Twist listeners can

1:11

get $1,000 off

1:13

for a limited time

1:15

at vanta.com/twist. Dev

1:17

squad. Most dev agencies

1:20

only offer developers. Why?

1:22

Because product management is hard. Get

1:25

an entire product team for the cost

1:27

of one US developer plus 10% off

1:31

at devsquad.com/twist. And

1:33

HubSpot Podcast Network. HubSpot Podcast

1:36

Network presents a new podcast

1:38

called The Next Wave. The

1:40

Next Wave is your personal

1:42

chief AI officer bringing fresh

1:44

takes, industry insights, and a

1:46

trustworthy perspective on. Just

1:51

search for The Next Wave on YouTube

1:53

or in your favorite podcast app.

1:56

That's The Next Wave. All

1:58

right. Welcome Back to another episode. Alright,

2:04

welcome back to Can You Survive This podcast.

2:07

Before we get started, just want to let

2:10

everyone know I got a coupon code for

2:12

manscaped.com. This. Dread

2:14

of oh My. God. What? If.

2:17

A I Plus Robotics gets put together

2:19

and we have the Terminator films. The

2:21

truth is autonomous robots are coming and

2:23

they will have say I built in.

2:25

You've probably seen figure or what he

2:27

wants working on with optimism or a

2:30

Tesla. This is going to change the

2:32

world in my police and today we

2:34

have a company that's been working on

2:36

it for a little one's called Seco

2:38

Robotics. We have the Ceo hear his

2:40

name is Jay. Looser area. So.

2:43

Tell me about the robots you're

2:45

building and for those of you

2:48

not watching this weekend start on

2:50

you tube. Or. The video

2:52

version of spot of I didn't go over to you

2:54

to decide been as reserves hit the subscribe but you'll

2:56

find this video they are under the video sap. And.

2:59

You can actually see what we're talking about

3:01

here. Yeah, tell me what you're building with.

3:03

second visit me on i'm really excited says

3:05

to diving on more him or top It's

3:07

a robotics and artificial intelligence sound effects the

3:09

world but some sort of in college my

3:12

several a robotics company out of college when

3:14

I saw firsthand sexually The Seat and how

3:16

are the physical world of realign? Every single

3:18

day collapses and is always there for you.

3:21

And says is happening to power plants were

3:23

I got to see firsthand where. Are

3:25

plant was having his massive shutdowns and

3:27

the best way to stop it was

3:29

sitting a human is when these environments

3:31

and trying to predict when he's built

3:33

structures in in particular Spoiler was gonna

3:35

fail. And. Sad to mess with

3:37

the do that was sending it a human to

3:40

the environment. The same year I'd that gone there,

3:42

some had fallen, died doing this job. Sturdy.

3:45

dangerous, under, And. To

3:47

be talked about as I go to walk on

3:49

a robot in college to solve. For.

3:51

The the k of Critical instructs

3:53

we care. as we rely on

3:55

so deeply to. liberalize every

3:57

single day so eleven years after that

3:59

Now here I am, still

4:01

working on the same critical mission

4:04

of protecting and helping to build

4:06

a new infrastructure, but more intelligently.

4:08

So these are purpose-designed robots

4:10

to do very specific tasks.

4:13

You're not taking the approach that

4:15

Elon's taking a Tesla or the

4:18

figure robot is taking of specifically

4:20

a humanoid robot. These are robots

4:22

that are designed for a specific

4:24

function like climbing up and inspecting

4:27

a building, correct? And

4:29

the reason I missed was it seems like

4:31

we don't care that deeply or at least know

4:33

that much about the built world that we rely

4:35

on. And that was the thought when

4:37

I was in college, hey, we go over a

4:39

bridge every day, hey, we rely on power plants,

4:42

we rely on manufacturing facilities, ships

4:44

to carry supplies all around the world. How

4:46

do we know if those things are going to

4:49

be around or there for us? Is it the right

4:51

assumption to believe that the bridge I'm crossing is going

4:53

to be structurally sound and not going to collapse? So

4:56

that's where you started the journey. You said,

4:58

hey, infrastructure is the ideal

5:01

customer profile for your startup and

5:03

for this product, robotics. Your

5:06

customer is essentially infrastructure

5:09

and specifically infrastructure in the United

5:11

States, which for whatever reason, we

5:13

seem to have not allocated enough

5:15

resources towards. Yeah, it's, you

5:17

know, in 2013 when I was in

5:19

college designing this first robot, you

5:22

know, I've read this report, it's a 3.34% of

5:24

GDP around the world was spent on fighting

5:26

corrosion. I was like, wow, that's a crazy

5:28

three and a half trillion dollar number. I

5:31

wonder what that is. And you look into, you know,

5:34

these, these interesting reports that show the US

5:36

is that like a D grade in terms

5:38

of its infrastructure and, you know,

5:40

it costs, you know, $1 trillion just to keep it there

5:43

and not to just not to improve it, but just to keep it there.

5:46

Maintain it. Yeah, maintain it. And

5:48

that's, you know, irregardless of building new things.

5:50

So it started with the critical industries and

5:52

infrastructure, but it was mostly this like thought

5:55

that was, wow, we, we seemed like, we

5:57

seem like we talk as if we know a

5:59

lot. and have a lot of data about how

6:01

the built world works and how to make it

6:04

better. But that's actually like very far from true

6:06

for the physical world, for example. We don't know

6:08

if a concrete structure like a bridge is gonna

6:10

be sounded and gonna be there and

6:12

how long will it last. You know, the bridges and

6:14

infrastructure that we rely on was not built for the

6:16

kind of like traffic and loads that we currently are

6:18

demanding today. So

6:21

we're collecting the infrastructure on top of it. It

6:24

was built for, you know, the

6:26

Golden Gate Bridge was built at a time when a certain

6:28

amount of vehicles would go

6:30

over it, a certain amount of weight of

6:32

those vehicles and obviously, you know, we've induced

6:34

a lot more usage of that with a

6:37

lot heavier vehicles. So maybe

6:39

you could show us in

6:41

SportsCast, one of these robots doing inspections.

6:43

And I know that you're not just

6:45

doing infrastructure. You've got energy defense, manufacturing

6:48

other robots and other verticals you're flying

6:50

in. But I would love to see what

6:53

these robots are and then get into,

6:55

you know, the business model because

6:58

it is this week in startups of how you make

7:00

money with these robots, yeah. Yeah,

7:02

absolutely. You know, when I was in college, looked

7:04

around and like I was describing, there seemed

7:07

to be like this world that, you know,

7:09

technologists and startups like didn't really pay that

7:11

much attention to. It's the

7:13

world of energy. It's the world of manufacturing.

7:15

It's the world of defense and

7:18

public infrastructure. And, you know,

7:20

I saw this like up close and personally with the

7:22

power sector. And it was just this idea of, man,

7:24

we don't really have that much data on the built

7:26

world and thus it

7:28

makes it really hard to know and understand like

7:31

how to predict how it's gonna perform. And what

7:33

you're showing on the screen here is

7:35

the Golden Gate Bridge, I assume a

7:37

nuclear reactor. And then it looks

7:40

like a really either another

7:42

type of bridge and inspectors

7:44

literally repelling up and

7:47

down. Yeah, that's exactly right.

7:49

Which is dangerous and I'm sure quite expensive. I don't know

7:51

what those individuals get paid, but they're

7:53

getting paid half as much as they should. What

7:56

if a person get paid to repel off of

7:59

a nuclear power plant? The door for the Golden Gate

8:01

Bridge. What are they make? One hundred bucks an hour. Fifty

8:03

bucks an hour, You know? You must, yeah,

8:05

it's it's a balance air defense new

8:07

level, and it's about in between, like

8:10

thirty and seventy bucks an hour that's.

8:12

Oh yeah, it's price is even precisely an

8:15

hour is sixty thousand dollars a year this

8:17

time such as it acid since alone and.

8:20

Yeah, overtime yellow higher for yeah, It's exactly

8:22

right at this is a circle tank. And

8:24

for example at our oil and gas refineries things

8:26

like the Cnt like this world you know

8:28

like most most like folks who are starting to

8:31

acknowledge companies are in about a tray i

8:33

like have never set foot on are at a

8:35

refinery or don't really know the first thing about

8:37

looks for for and materials science or which from

8:39

what. What? Does the hundreds of different

8:42

as a throws in his and steals or

8:44

instead of concrete but these are really important

8:46

and not just like to predicts and and

8:48

ends. Answer that were nice. Not.

8:51

Suffering from some sort of test or failure

8:53

which actually as environmental as well as the

8:55

a functional and implications but also how do

8:57

you actually. Modulate a

9:00

your operating the infrastructure to actually

9:02

get more at as the other

9:04

see the power plant or the

9:06

refinery and while also reducing the

9:08

amount of greenhouse emissions that are

9:10

being there being released by this

9:12

was a company because whenever the

9:14

success or failure who pipeline guess

9:16

what lot of like explosion leads

9:18

to unfiltered carbon emitted right into

9:20

the atmosphere and like the worst

9:22

normal. He. Recorded environmental incident was

9:24

the Nord stream pipe is quoting for

9:26

example our in of these are these

9:28

deepwater horizon events so he of ensuring

9:30

that you know how discuss I feel

9:32

is actually a really important as you

9:34

would serve as a zero and and

9:36

those things but yet so the story

9:38

was ten years ago in college and

9:40

basically across as a we're problem of

9:42

our plants have any shutdowns and some

9:44

had died. Listen a

9:46

strong cel seem to make all the difference for be

9:49

to be start up. But. If you're

9:51

gonna hire said he need to let them

9:53

son and he said slow them down with

9:55

compliance hurdles like sat through. What? Assassin

9:57

both and. Company the full custody.

10:00

in the cloud needs to be SOC 2 compliant.

10:02

If you don't have your SOC 2 tight,

10:04

your sales team can't close major deals. It's

10:06

that simple. But thankfully, Vanta makes it real

10:08

easy to get and renew your SOC 2

10:11

compliance. On average, Vanta customers are compliant in

10:13

just two to four weeks. Without

10:15

Vanta, it takes three to five months. Vanta

10:17

can save you hundreds of hours of work

10:19

and up to 85% on compliance costs. And

10:21

Vanta does more than just SOC 2. They

10:23

also automate up to 90% compliance for

10:26

GDPR, HIPAA, and more. So here's your call to

10:28

action. Stop slowing your sales team down and use

10:30

Vanta. Get $1,000 off at vanta.com. That's

10:34

vanta.com. For $1,000 off your SOC 2. A

10:38

charming backstory if there ever was one. Show

10:40

us the robot. Okay, all right. Show us

10:42

the robot. We wanna see this thing in

10:44

action. Okay, so. Every founder's

10:47

got a charming story. You have a

10:49

PR team that helped you craft that or that's the

10:51

authentic story? This is me. No, this is me. The

10:53

authentic story, okay. Yeah, so. I

10:55

believe you. The first robot was one that was

10:57

climbing up a wall and

10:59

gathering information, visual and ultrasonic.

11:02

Basically what we're looking at, what

11:04

I was looking at 10 years ago was what

11:07

is the structural integrity of the pressure vessel?

11:10

And how do you ensure that you're understanding, just

11:12

like you're doing a CAT scan

11:14

or doing a sonogram, you use high

11:16

frequency sound wave to look inside of a material without

11:18

needing to open it up and destroy the surface. So

11:21

what we're seeing on the screen is

11:23

a robot that's about the size of a pool

11:25

cleaner with a tether. And

11:27

it's zipping up and down some pipes. And

11:31

it's open, so you can see all

11:33

the innards of it. It looks like an insect crawling

11:36

along the pipes. That's the size of a pool cleaner,

11:38

like I said. Am I about right? The

11:40

robot's, yeah, it's about the size of a briefcase. I

11:42

mean, it comes in a couple different forms, but

11:45

basically. How is it? Yeah,

11:48

because it's upside down and it's

11:50

gripping. Is it suction cups,

11:52

magnets? What is it doing there? So

11:55

it's climbing up surfaces, whether it be

11:57

an outside of a ship or let's

11:59

say a price. some sort of like piping or

12:01

a dam even, we'll use

12:05

neodymium rare earth magnets arranged in a

12:07

hallbark array. And that maximizes pull

12:09

force into a surface to allow for

12:11

payloads to be added onto the robots.

12:14

And they're collecting different kinds of data layers. One

12:16

of the data layers, for example, is ultrasonic

12:19

data layer that's looking at

12:21

what's the structural integrity, corrosion,

12:23

erosion of the surface

12:25

to get generalized idea of what is the

12:27

health of this? Just like you would do

12:29

like a picture of a belly using a

12:31

sonogram test for pregnancy. Wow, so do

12:34

humans do this when they're climbing

12:36

up and down? We saw them rappelling. Do they have

12:38

some device that they do this manually with? Yeah,

12:41

they do. So basically our

12:43

savior today is Joe, Joe in

12:45

a rope. But

12:47

basically it's these guys. These guys are

12:50

our best defense. The

12:52

guys that are hanging off of ropes

12:54

or climbing on scaffolding around JLGs. And

12:56

they're armed with single probes that

12:59

you use some gel, you squirt the gel on

13:01

a surface, let's say on kilometers of pipeline, you'd

13:03

squirt gel every 10 meters, every one meter depending

13:05

on the criticality. And then you use the ultrasonic

13:07

sensor and you record the

13:10

waveform. And then because you know, if

13:12

you understand the speed of sound through

13:14

that material, you can actually understand what's

13:16

the thickness of that material. And

13:18

then you could use, and then you

13:20

record that down on a piece of paper or

13:22

in an Excel sheet. And basically that's the way

13:24

that we understand how the field war works. And

13:26

you're taking an example, but

13:29

you're taking continuous so you have

13:31

the full picture. It's possible

13:33

in fact probable that

13:35

the humans are going to miss most

13:38

issues. Am I correct that they're

13:40

gonna miss most or some? They're

13:43

gonna miss a fair amount or

13:45

there's actually human errors that relates to interpreting

13:47

the waveforms. But these

13:49

are the kinds of techniques that you either are or

13:51

are not using. So visuals one is like, hey, this

13:53

thing looks like it's leaking, that's bad. Or

13:56

around like welds, like you have to

13:59

do sort of. of welds on

14:01

critical pipelines, for example, and you're using

14:03

x-rays. Interpreting the x-rays

14:05

is actually pretty difficult and it's

14:07

also super dangerous because you're using something that can

14:10

cause cancer if you're not

14:12

appropriately operating it. So

14:14

we actually will put on the robot

14:17

something called phased array which is basically

14:19

ultrasound with just like hundreds

14:21

of different sound waves going

14:23

into like a very small area. You can

14:25

apply basically these different payloads onto the robots

14:27

to look at corrosion, look at

14:31

generalized erosion, but then you can also add

14:33

other kinds of information. You can use electromagnetics

14:35

to look at what's the

14:37

damage overtop of

14:39

some substrates that you

14:42

have to remove like some sort of

14:44

insulation. So anyway, what you're trying

14:47

to solve for the customer is how do you reduce the

14:49

downtime or

14:51

the time I'm spending not making my product.

14:54

And so that's what you're trying to first

14:56

help the customer understand is how do you

14:58

ensure that you're solving this problem of

15:01

ensuring that there's not going to be some cash off a consent, but

15:04

limiting the amounts of time you're not making

15:06

your product. So the robots are

15:08

going into these like missile silos, for

15:10

example, or on top of light decks

15:12

on destroyers. It's climbing inside

15:14

of power plants at boilers. It's

15:17

going on to dams using suction

15:19

and adhesion and there,

15:21

but basically we've gone from like what you just

15:24

saw in terms of the robot climbing up a

15:26

wall, looking at corrosion and erosion. And

15:28

we would now like combine that into a

15:30

bunch of different robots, some

15:33

of which are doing this climbing, some of which are

15:35

using just like drones

15:37

that are looking at using photobametry

15:39

to understand what is like in

15:42

general, let me do

15:44

a quick analysis of potential damage

15:46

areas over like large geographical area

15:48

or maybe integrating like a walking

15:51

dog or and then you can

15:53

use six sensors to continually monitor.

15:57

This reminds me of the Pranovo full

15:59

body. scan, which a

16:01

lot of doctors will say, hey, you don't

16:03

need it. It's going to cause you to

16:06

find things, nodules, little things, growth in your

16:08

body. You're not going to know what they

16:10

are and you might panic and get anxiety.

16:12

And I'm like, well, wait, but what if

16:14

it is something and you

16:16

live longer because you found, you know,

16:18

some God forbid cancer or tumor early

16:21

or something with your brain? I

16:23

would much rather have that. Therefore, you

16:26

are going to inspect these things and

16:29

have an image in time. And

16:31

then you can look for the deltas and what's

16:33

changed between the two imaging. So

16:35

if you were to do this

16:38

every year on the Golden

16:40

Gate Bridge, what would be the frequency that the

16:42

Golden Gate Bridge or, you know,

16:45

a submarine should have this done to it? So

16:48

we're actually working on, I'm in Pittsburgh, Pennsylvania

16:50

right now, which is where

16:53

I started the company, did three and

16:55

a half years of boot shopping it

16:58

down to like a hundred bucks meg account, ended up

17:00

choosing to go to YC if it was to an

17:02

acquisition offer, went out to

17:04

California against all investors, like desires came back

17:06

to Pittsburgh, close to customers, was able

17:08

to grind closer to there. In Pittsburgh, though, it's

17:10

interesting, you know, there's so many bridges. It was

17:12

where we, you know, 69% of the

17:14

world's steel was built here. And

17:17

now it's kind of reinventing itself in terms of this robotics

17:19

and AI hub. But what's exciting

17:22

is actually it's actually a really great

17:24

state as it relates to the

17:26

political support to try and utilize

17:28

technologies like GECOs to do things

17:31

like create the most

17:33

sophisticated bridge evaluation

17:35

infrastructure process. There's like, you know,

17:38

we're working with the governor actually on an initiative

17:40

with bridges. But to answer your question, you want

17:42

to look at how often you got to inspect

17:44

a bridge, you want to be able to look

17:46

at a bridge, you'd want to look

17:48

at it with a deep scan, like we would do like

17:50

a full health, like here's exactly what's going on with the

17:53

entire bridge. Maybe like once three or

17:55

five years. You don't want to like look

17:57

at it every year. You actually go though, like once you

17:59

understand. the general

18:02

health, similar to how you would do

18:04

with a human, you would then use

18:08

six sensors that are enabled by Wi-Fi or

18:10

5G. And then those are

18:12

constantly updating a digital twin. And

18:16

that actually like this... Explain what a digital twin

18:18

is for people. So

18:20

digital twin is represented in software, it's

18:22

three-dimensional, you can manipulate it, but it

18:24

needs to update itself. So it needs

18:27

to be continually updating with

18:29

information, whether information is the health of the

18:31

asset or how the asset is performing. So

18:33

an example for a bridge might be a

18:35

real world example you've come across might be?

18:38

Yes, a real world example is a tank.

18:41

So a tank at, let's say, a pulp and

18:43

paper manufacturing place, a place where we all get

18:45

our paper. So

18:47

we have a really big contract with this

18:49

company that's interested in extending the useful life

18:51

of their tanks. But then what they

18:54

want to do is instead of the tank is 20 years

18:56

old, it's past its useful

18:59

life, so we have to build a new one. And we come in

19:01

and say, actually, you don't need to. We'll

19:03

take this tank plus 50 to other tanks

19:05

that look similar to this. And

19:08

we'll tell you how to make it last 10 years longer or

19:10

even 20 years longer. We'll tell you exactly what to repair. And

19:12

Jason, we're actually... Because we now have this

19:15

information on the

19:17

health and structural integrity of the world's...

19:19

Some of those will put critical infrastructure,

19:21

like 500,000 assets. That's

19:24

where we use AI. Hey,

19:26

take a moment. Picture the ultimate all-star team for

19:28

your startup. You got that image? Maybe you're thinking

19:30

about the Avengers, huh? Maybe the X-Men? Wolverine? Cyclops?

19:34

Iron Man? Well, let's be realistic here. Until

19:36

you've raised your Series A or B, you're

19:38

going to need help building this A team,

19:40

right? You may not have all the resources

19:42

to do it. Finding talent, managing all these

19:44

timelines, and ensuring quality is a huge task

19:46

for anybody building a product. And then, scouting

19:48

for the perfect all-in-one 10X developer is going

19:50

to take a lot of time. Well, there's

19:52

somebody out there who does this for a

19:54

living. Skip the hassle and power up with

19:56

Dev Squad, okay? They're like Professor X from the

19:59

X-Men. They just know... where all those great

20:01

mutant developers are. Those really powerful ones. DevSquad

20:03

offers a complete product team loaded with top-tier

20:05

talent straight from Latin America. Same time zone

20:07

as us, right? Your team is going to

20:09

have two to six full stack developers, depending

20:12

on what you need, a technical product manager,

20:14

keep everything tight, and a bunch of specialists.

20:16

Maybe you need UI and UX design. Maybe

20:18

you need DevOps. How about quality assurance? They're

20:20

all going to be synced up in your

20:22

time zone and going to cost 75% less

20:25

than the equivalent team in the USA. We

20:27

all know there's an incredible arbitrage. There's talent

20:29

everywhere in the world. And now that

20:31

we all work online, remote, hybrid,

20:33

whatever it is, you know that

20:35

you can integrate this amazing DevSquad

20:38

team into your team. If you're

20:40

looking for seamless collaboration and the

20:42

flexibility of monthly payments without long-term

20:44

commitments, it's time to take action.

20:46

Head over to devsquad.com/twist and snag

20:48

an exclusive 10% off your engagement.

20:50

That's devsquad, devsquad.com/twist. It's time to

20:52

squad up and transform your startup

20:54

with the all-star team you deserve.

20:57

So let's take a look at a digital twin then. I'm

20:59

assuming you can show me one or... Yeah, because

21:01

this is fascinating. You inspect this container,

21:04

right? And let's say it's got... You

21:06

didn't give the exact example of what

21:08

you would continuously monitor in the digital

21:10

twin, but I'm assuming maybe there's some

21:13

area where you think it might get

21:15

fractured or be compromised. And so

21:17

you put a sensor on that permanently?

21:19

Yeah. That sends a

21:21

contiguous, continuous reading

21:23

to let you know if it's getting worse. And at what

21:25

point you think it's going to explode or crack

21:28

or fail, which would be the equivalent

21:30

of like in a human body, just monitoring

21:33

some tumor that might be benign or

21:35

might not be benign. Am I correct?

21:37

Am I framing here?

21:39

That's right. But you also... Okay, so this orients

21:42

towards the business model too. It's like, we

21:44

started building robots and they're really exciting and

21:46

cool. What we ended up finding was that just

21:48

building robots and using a robot as a service, we

21:51

don't actually sell the robots. We're going out to site

21:53

with the robots and getting data and then giving it

21:55

to customer. Got it. What we

21:57

ended up finding was that that wasn't... That

22:00

wasn't a model that actually oriented towards

22:02

value creation. So we were

22:04

creating outsized returns, in some cases, like

22:06

nine and even like, there's

22:08

even a case of a half a billion dollar

22:11

value creation because we stopped this like crazy explosion

22:13

at this refinery, the biggest refinery in the US,

22:16

because they had gotten faulty information from the guy on

22:18

the road, Joe. What we ended

22:20

up doing was, you know, charging them like a couple

22:22

hundred thousand bucks for this. And that

22:24

was crazy because how outsized the value creation was. So

22:27

what we ended up also finding was there was

22:29

lack of an ability to take action on the

22:31

data to improve how

22:33

the customer was operating their

22:35

assets. So let me walk

22:38

you through that. So what we'll do is we'll

22:40

have a suite of different robotics that we

22:42

offer and work with the customer

22:44

to try and solve for a problem. In this case,

22:46

it was how do I manage, you know, 50 of

22:48

my most important fresh vessels and tanks for this customer.

22:50

And about five years ago, we started developing Canalleaver,

22:52

which is our enterprise software. So when a customer

22:55

buys Gecko, they're buying enterprise

22:57

software and that's called Canalleaver. And what they're

22:59

getting from that is a solution

23:01

oriented towards a very specific problem for

23:03

that customer set. So we actually not only

23:06

had to become experts on robotics and

23:08

AI and software, we also had experts

23:10

on like our customer's actual problem, both

23:12

upstream and downstream. You know, I'll take

23:14

you through an example of a... Which

23:16

I'm assuming you got by asking

23:18

them questions in customer

23:20

interviews and saying, well, what are

23:22

you going to do with this data? We've

23:24

now given you the data and they told you,

23:26

hey, well, we need to make this decision when

23:29

to retire this tank. And then it

23:31

began, your business becomes not

23:33

selling a robot or selling an inspection.

23:36

Your business is now extending the life of

23:38

tanks. That is one of the

23:40

important value outcomes. Yes. And

23:43

but it also was like, in the beginning, like

23:45

I had to like spend all my days and

23:47

time at the customer sites and just like living

23:49

and understanding their problems, like better than

23:51

they could. And did that for

23:53

not just power plants, but you know, many

23:55

factory facilities, like the places are making steel

23:57

or places are making aluminum. places

24:00

that are refining oil that are

24:02

operating a hydroelectric dam. We

24:05

have to end up going into these

24:07

industries and understanding exactly what they're

24:09

trying to produce. From our first

24:11

principles, what goes into both the

24:13

good and poor outcomes. And then

24:15

also understanding where they are

24:18

getting value and get full value, like from a

24:21

regulation standpoint. It gets really complicated.

24:24

So for this customer, we... We call this technique,

24:26

by the way, in the business of bear-hugging. So

24:28

when you have a customer who's like a key

24:30

customer, you give them that big bear-hug, which

24:33

means you get on location, you spend time

24:35

with them. I learned this from a company

24:37

where investors in colddensity.io that does people-caunting. And

24:40

when you are on site, you will overhear

24:43

things. And you're

24:45

going to have the customer, just

24:47

through the course of hanging out with them, give

24:49

you insights that you're not going to get in

24:52

a 20-minute customer interview. You might get them, but in

24:55

all likelihood, just hanging out at the facility or

24:57

maybe having even a drink or having lunch with

24:59

folks, at some point, you're going to have these

25:02

epiphany moments. Yeah? Then that's what happened for you?

25:04

Yeah, 100%. But also, you can codify that into

25:06

a business model. So one

25:09

of our... One of our... So our series A

25:11

investor was Founders Fund. Trey

25:14

Stevens was our partner and board member there. But what

25:16

was interesting is while I was... We went to... We

25:18

took a trip to the UAE in 2020. It was

25:20

literally right before COVID. We

25:25

almost got stuck, actually, in Oman,

25:27

I think it was. We ended up

25:29

just talking through Palantir in

25:32

the early days and how he's helped set

25:34

up the Palantir office in the Middle East.

25:36

We ended up talking deeply about forward-deploy engineering.

25:39

And I was like, wow, this is

25:41

so cool that they have

25:43

taken this approach forward-deploy engineering as you

25:45

send out your engineers on deployments as

25:48

an implementation team with the software you

25:50

sold. And then you work

25:52

alongside customers to understand their problems, to help

25:54

create the software modules that

25:56

are oriented towards the solutions that the

25:58

customer is actually trying to... us all

26:00

because in reality, when you deal with these

26:02

industries, they're so complex, these problems are so

26:04

complex, and they are so hesitant to either

26:07

communicate or even to talk about the different

26:09

problems that exist in these Manhattan-sized

26:12

environments, like the other size of refineries,

26:14

the size of Manhattan. And

26:16

so there's so many different things,

26:21

variable frequency drives that need to be looked at

26:24

and wrench turned in this way and all these nuances.

26:27

And this is actually one of the big

26:29

issues as well is that these people that are

26:31

relying every single day are completely...

26:33

It reached

26:35

this point of phasing out, whether

26:38

they're dying or retiring, and there's a huge

26:40

knowledge gap. So anyway, I

26:42

began to think about what if

26:45

actually we took the early learnings

26:47

of forward deploying roboticists in

26:49

combination and concert with forward

26:51

deployed software engineers to actually

26:53

build a vertically integrated stack

26:56

of data collection of various

26:58

types and a lot of it. So

27:00

we call them data layers, and then pull

27:02

all that into a single source

27:05

of truth, a data warehouse, and

27:07

then deliver the modules and software

27:09

to solve customer problems, but do

27:12

so located actually

27:14

alongside customers because

27:17

you have to convert someone into using a

27:19

different system, you actually have to help build

27:21

it alongside of them. Yeah. And

27:23

the current system was probably pencil and

27:25

paper, pictures, and stuff scattered across

27:28

disparate systems, I'm assuming. Yeah.

27:31

That's right. And inconsistent per site.

27:33

So Marathon, they might have seven

27:35

refineries and each of those refineries

27:37

operates completely differently because they're both

27:39

producing $20 billion each or something

27:41

like that. You didn't show us the

27:43

actual digital twin. Let's get it. Make sure

27:46

we show that. Yeah. Of

27:48

course. Yeah. It's so fascinating what

27:50

you're doing. It's easy in an interview like this to

27:52

get sidetracked into all the different nuggets

27:55

of what you're discovering as a founder, but I did want

27:57

to see the digital twin concept. Okay. Let's

27:59

do it. Or like, okay, what we're probably trying to

28:01

solve. Well, we're trying to solve for increasing

28:04

life extension or understanding like how to fix

28:06

like 50 tanks and manage 50 tanks. All

28:08

right, sounds good. So the outcomes we

28:10

were able to do this solve or I'll just skip

28:12

that to the end. Basically, customer will

28:14

say like, okay, customer, I need your metadata

28:16

as it relates to the structures that we're

28:18

gonna go out and try to evaluate. So

28:20

they'll send us the metadata and then we'll

28:22

incorporate that instead of cantilever as we build

28:24

out their profile. And so

28:27

we're delivering using just drawings,

28:29

what is a very rudimentary digital twin.

28:32

And so this is an example of

28:34

a 3D representation of a tank using

28:36

the dimensions of the customer. Then you

28:38

send out your robot fleet. And

28:41

so the robots go out there and they're climbing

28:43

all over these structures and they're trying to evaluate

28:45

what is the health of this tank. And

28:48

doing so as quickly as possible, while the tank is

28:51

actually in operation to shut the thing down. And then

28:53

you understand what the health of that structure is. This

28:55

one was pretty bad. Red

28:57

is good. For example, green is bad.

28:59

And it looks really pixelated because as

29:01

the robot's climbing, it's pulsing the area

29:04

it's climbing over hundreds

29:06

of times every single inch. And then you can

29:08

either look at what's the mean in

29:10

terms of how structurally sound or how healthy

29:12

that area is, like an inch by inch

29:14

grid. Or you can look at the data

29:17

in other ways. But you wanna label all

29:19

that data set because it'll be very helpful

29:21

as it relates to what kind of corrosion

29:23

is going on. And

29:26

as a root cause. And as a tank shell, the

29:28

bottom 25% of the tank is green. Yeah,

29:31

so the bottom is actually super healthy is what you're

29:33

seeing. So the red is bad. But green is healthy.

29:35

Red is bad. Yeah. All

29:38

right, AI is moving really fast and

29:40

we're seeing it in all aspects of

29:42

our lives in business, personally. And you

29:44

know what? It's okay to be confused

29:46

by all of this. I

29:48

found a great podcast for you. That's

29:51

perfect at helping you navigate all things AI.

29:53

It's called The Next Wave. And it's brought

29:55

to you by the House Fat Podcast Network,

29:58

the host fat wolf and Nathan Land. both

30:00

have an insane level of AI

30:02

expertise because they've both founded and

30:04

invested in AI companies. So think

30:06

of the next wave as your

30:09

personal chief AI officer. That's right.

30:12

Serving up fresh perspectives, industry insights,

30:14

tools, tips, and tricks on

30:16

how to use AI to grow your

30:18

business. That's the most important part, tactical

30:21

advice to actually put AI to use

30:23

in your business. Check out the episode,

30:25

The Large Language Model Race with Peter

30:27

Wang, founder of the Noron, and he

30:29

covers the latest AI developments, trends,

30:32

and research, including a fresh take

30:34

on where AI and your creativity

30:36

collide. Its impact on industries like

30:38

music and entrepreneurship, the balance

30:40

between innovation and ethics, and

30:42

the rising capabilities of LLMs.

30:45

So here's your call to action. Super easy. I

30:48

want you to go search for

30:50

the next wave on YouTube or

30:52

your favorite podcasting app and subscribe

30:54

and discover how to leverage AI

30:56

to unlock new opportunities for your

30:58

business. That's what you buy, the

31:00

HubSpot YouTube Network. So the

31:02

base is healthy, but the top

31:04

75% for some reason is really weathered. I

31:07

don't know if that goes by wind or

31:09

the ground is painted or covered. What would

31:11

cause that weird pattern? What we found is

31:13

it's very correlated with how you operate the

31:15

tank. So this is a

31:17

really important tank, and they were typically filling their tank

31:20

halfway up. And this is a

31:22

very common method in fuel storage

31:24

tanks, for example, if you're filling your fuel

31:26

storage tanks, if you're exon 40% of the way up, because

31:29

you just don't want everything to collapse

31:31

and you just view a bunch of oil

31:34

into a river somewhere, which is actually not

31:36

good because you're actually destroying the asset in

31:39

a certain way and you're reducing the capacity you can

31:41

run at and how much oil you can store.

31:44

So they were scared of a disaster. So

31:47

they made a decision that increased the likelihood of a

31:49

disaster. So good intent, bad outcome.

31:52

That's right. For one of

31:54

our customers, Adnok, which is

31:56

the UAE's national

31:58

oil company. We just send a $30

32:00

million deal with them. But

32:03

the big thing they're trying to solve for is how do I produce more

32:05

oil per day, barrels per day,

32:07

go from 3.5 million to 5 million barrels

32:09

per day. But what

32:11

we're trying to also show is that

32:13

you can reduce the potential

32:15

carbon emissions while also increasing your throughput

32:18

if you just operate your assets more

32:20

intelligently. So we'll get into exactly what

32:22

that means. But you don't

32:24

have to actually build new things necessarily,

32:27

which is a huge deal. Then you

32:29

send in robots that can do evaluations,

32:31

again, while the tank is in operations

32:33

in a submersed way.

32:35

And so again, you're looking at what's the

32:37

structural integrity of the floor because the floor

32:39

is actually one of

32:41

the most compromised areas. And so the green, again, is

32:44

good, the red is bad, and you want to try

32:46

and evaluate where to fix things. And

32:48

then you use lidar to look at what's the

32:50

depression of the tank because it's really heavy. And

32:53

so it'll begin to depress in

32:55

certain locations, which can also lead to a

32:57

bunch of issues. So lidar is really

32:59

important. And there's different

33:01

kinds of rules and regulations that different

33:04

bodies set out. And then we'll create

33:06

these repair plans. And so what that's

33:08

trying to do is help

33:10

the customer understand how much carbon to deploy and

33:12

then how many years you get from the life

33:14

extension. So what we'll do is we'll use the

33:16

data that we're collecting from this tank, the other

33:19

50 tanks at that site, and then also the

33:21

thousands of other assets that look just like this

33:23

to try and evaluate where are the areas that

33:25

you want to fix right now to

33:28

extend the useful life of the asset. But

33:30

then once you do that, you have to send

33:32

it back to the real world. So it goes

33:34

physical, digital, physical. So it has to be an

33:36

output where there's an action being taken from the

33:39

insights and a digital twin is actually helping to

33:41

be used by folks that are welding

33:43

and doing repairs, for example, or

33:46

trying to make an actual functional decision around

33:49

how to operate the asset, in

33:51

this case, the tank. But then

33:53

we'll install fixed sensors that are taking

33:55

the digital twin every day. These are

33:57

those black little circles around the compromised

33:59

area. One of the rings of this, they

34:02

give it like a barrel, is compromised.

34:04

So you're putting sensors on it that

34:06

tell you what. It'll look

34:08

at the structural integrity. So

34:10

it'll ping you every day because corrosion actually is...

34:13

Funny enough, it doesn't happen linearly. It

34:15

happens typically in these

34:17

weird moments of large decay over short

34:19

periods of time. Typically that's related to

34:21

what kind of chemical is in the,

34:23

let's say, into the tank that may

34:25

be out and abnormal. Maybe there's some

34:27

sort of, this is really windy

34:30

and rainy that month, or there's a lot of

34:32

sodium in the air and that's causing a lot

34:34

of increased corrosion, if you like, by

34:36

a gulp, for example. It also... Like

34:38

people say, how did it happen? Slowly

34:41

then all at once. It's

34:44

slowly degrading and then some event happens and

34:46

all at once it gets compromised is generally

34:49

aviation, bankruptcy, and

34:52

structural failures all seem to go in that direction.

34:55

Slowly then all at once. That's exactly right.

34:58

And there are some root causes that you can

35:00

begin to understand. So you might get a shipment of

35:03

new fuel or a new chemical

35:06

and that might be actually

35:08

compromised in the quality of that. This

35:10

is actually an issue as well with you manufacturing

35:12

new things. Windows are falling

35:14

over in Germany right now because of poor steel quality.

35:18

So you end up having

35:20

this issue where you might

35:23

be getting really inefficient process for

35:25

some reason. You can actually tell when the

35:27

inefficiency is occurring or when there's an imbalance

35:29

of chemicals in your processing

35:33

batch that you want to be

35:35

able to react to. And that's something that's not

35:38

predictable. It's reactive. People started coming to

35:40

you saying, hey, we're installing this new

35:42

thing. We want to have a

35:44

day one inspection so we

35:47

have a benchmark. And so if it was

35:49

installed improperly, we can... Before

35:52

we make the final payment to the construction

35:54

group, we want you to do

35:56

the inspection of the work done. Has that started to

35:58

happen? and actually has

36:01

happened with the 132 billion

36:04

dollar Columbia class nuclear sub

36:06

program as well as

36:09

other sub works. We're working with

36:11

the Navy actually on new

36:13

builds and manufacturing. What's happening is...

36:15

So you're inspecting them when they're in dry dock. So inspecting

36:18

them while they're being built actually. Oh, so

36:20

what ended up happening was... So

36:22

on the government side and the defense side, scheduled

36:25

adherence is a really big problem as well as if you're

36:27

paying 132 billion dollars for new

36:29

subs, 12 new ones, you want to make sure your

36:31

tech dollars are actually being used for building

36:34

good things with high integrity. So we're

36:36

actually building out digital twins of the

36:38

sub as it's being constructed and looking

36:40

at the quality of welds

36:43

because that just recently caused like a six

36:45

month delay in a process where they had

36:47

to take sections of the sub apart. So

36:49

anyway, this is like priority one right now

36:52

on the manufacturing of new sub side. And then we

36:54

also do some work with the

36:57

Navy. We've got a pretty

36:59

large... You have

37:01

a digital twin of the sub? You have a digital twin

37:03

of the battleship? I'm not sure

37:05

I can show you, but we're doing

37:07

this as well with... Actually,

37:09

it's an interesting program. I can show you

37:11

real quick, but we're beginning to

37:13

use the same kind of philosophy as

37:16

it relates to concrete. So what we're

37:18

doing for the US Air Force is

37:21

we're sending our robot, a

37:23

different form factor robot with leveraging

37:26

the stack that we've developed to climb up nuclear

37:28

missile silos. So there's 450 in the

37:30

US and what's going on right

37:32

now with the Sentinel program is about $125 billion

37:36

are being deployed to upgrade

37:38

the Cold War era nuclear

37:41

deterrent system of these

37:45

ICBMs in silo. But

37:48

what's happening is the concrete's been into

37:50

decay and crumble is actually causing these.

37:53

In Oklahoma, there's this oxidation explosion, actually one

37:55

of the ICBM missile silos, which you don't

37:57

want oxidation explosions inside of a nuclear chamber.

38:00

So, we're... Yeah, generally speaking, explosions

38:02

plus nuclear weapons, not a good combination.

38:05

Not a good combination. So, we got

38:07

so sourced on work. It'll be

38:09

about 250 million dollar project.

38:12

But it's... But what you're looking at is, we

38:14

want to understand there's a steel liner and then

38:16

a concrete liner, like five feet of concrete. You

38:18

want to understand what's the structural integrity of the

38:20

concrete and like where all the dam...

38:23

where all the issues are occurring and then what's

38:25

the structural integrity of the steel. And then

38:27

you can figure out, okay, now I can create a plan to fix

38:29

all this stuff. Amazing. So,

38:31

anyway, there's these interesting applications and

38:33

there's new built history for it.

38:37

And you're... Back to the example. Yeah,

38:40

no, it's incredible. And I know

38:42

you have in the deck the destroyer

38:44

as well. Yeah. And I guess looking

38:46

at the hull of that, is all

38:49

of this going to culminate in permanent

38:52

robotics and permanent sensors being put

38:54

onto these things? Or is that

38:56

cost prohibitive in some way? Or

38:58

just too bulky and

39:01

too much maintenance in and of itself?

39:03

Because based on what you're learning, maybe

39:06

the sensors should just be built into everything. In

39:08

the same way, I know this is like a

39:10

minor analogy here, but, you know, like

39:12

air tags, the act of finding your stuff

39:15

is going to be built into other devices. I

39:17

think like the Apple remote controls have air tags

39:19

built into them essentially. And so it does seem

39:21

to me that based on all your learning and

39:23

the stuff, man, they should just be

39:26

putting these sensors in a lot of different

39:28

places permanently or have these robots permanently installed because

39:30

the robots currently have an inspector working with them,

39:32

correct? They have to be supervised. They're not

39:34

like... We're not at the point where like these

39:36

robots just exist in a little cabin and

39:38

go up every week and inspect

39:40

and go back like a droid in Star Wars, right?

39:42

We're not at that level yet. No, and I don't

39:44

know if you need to be actually. You want to

39:47

do what you said, which is once

39:49

you build something, you want to understand what's the health

39:51

of that structure. So, Jacob, what I believe is like

39:53

in the next like five to seven years, you won't

39:55

be able to build new things without first initiating the

39:57

health of that asset on its construction.

40:00

And you create a digital twin that

40:02

is able to be updated as well with sensors that

40:04

you build into these structures, especially structures that are really

40:06

important like the nuclear sub. And what

40:08

you want to do is instead of the sub saying it's going

40:10

to last 40 years, it's going to last

40:12

50 years, that actually can be doubled. So you want to

40:14

be able to, you want to make sure that you can

40:17

create something that can be updated every time

40:19

that you're doing some turnaround. And then eventually,

40:21

maybe even don't need to spend 18 months

40:23

in dry dock to do an evaluation of

40:26

the health of the structure, which is currently the state of a

40:28

lot of our Navy. So a third

40:30

of our Navy is currently in dry dock, trying

40:33

to do its maintenance cycles, which

40:35

means that a third of our Navy is

40:37

not out there patrolling the seas and ensuring

40:39

that conflict is being deterred. So it's actually

40:41

a pretty large problem that Secretary of Navy

40:44

Del Toro and I have talked about like

40:46

a lot actually is the schedule adherence and

40:48

also understanding what is the

40:50

state of the structures as we build them. And

40:53

how do we ensure that we're having, creating these living models

40:57

of these glass sets. You could

40:59

cut that dry dock time in all of these

41:01

cases down by 50%. You

41:04

think ultimately 90%? Well, the

41:07

goal should be actually like don't spend

41:09

any time if you can help it

41:11

in dry dock. Oh, continuous monitoring. So

41:13

that if you're inspecting a battleship or

41:15

submarine, you could have it at

41:18

the surface of the submarine. Obviously, the ship is

41:20

already at the surface. You could have underwater robots

41:22

inspecting the hull while it's out in the ocean.

41:24

Yeah, you can, I mean, there's like a 2% gain

41:27

in efficiency if you can like scrub a hull while

41:29

you're like, what will a ship

41:31

is going from like one place to the other. It's because

41:33

of like, you know, barnacle and build up on the hull

41:35

of a ship. There's a bunch of

41:37

things you can do to improve the efficiencies of

41:40

existing critical infrastructure. But

41:42

I think the big thing we should be orienting to

41:44

is like, how do you not be so reactive and

41:47

incentivize a model which is currently incentivized for

41:49

time and materials, right? So whenever you're doing

41:52

whenever you have like these large maintenance crimes,

41:54

they are incentivized to have the maintenance cycle

41:56

last as long as this is capacity to

41:58

last. Right. I'll show you

42:00

the outcome. Exactly. The longer it's in

42:02

dry dock, the more the meter is running. Right.

42:05

You are a big threat to maintenance

42:07

companies because you'll tell them you

42:10

only need to do maintenance on this 20 percent,

42:12

the other 80 percent is fine. Well

42:14

I think it's not a threat. It's

42:17

orienting the outcome towards improved

42:20

performance. And

42:22

so it's a power

42:25

by the hour was like the old Rolls Royce model.

42:27

It's like how do you get

42:29

paid for the amount of uptime you're producing. That

42:31

should be the orientation. That should be the incentive.

42:36

How do you keep this thing in service,

42:38

not how much service do you do.

42:41

And that's really hard to do because then you have an incentive the other

42:43

way. Hey, we got to keep this thing

42:45

flying and maybe you put something up

42:47

there in the air that shouldn't be flying and should

42:50

be in dry dock and should be inspection. And

42:52

what you're trying to do is get to the truth. And

42:54

the truth shall make you free. If you actually

42:57

have the truth, you don't need to... If

42:59

you can get to ground truth, your first principles, you're

43:01

going to not have to try to game an incentive.

43:03

But also, yeah, exactly. But when you build new

43:05

things, like think about it, this is the way

43:07

I get excited about. When

43:09

you build new things, you

43:12

want to be able to learn from the experience of

43:15

these billions of iterations

43:17

of that thing being in use every

43:19

single day. We don't do that right now. We don't... We

43:22

can model as much as we want about how to

43:24

build the best sub or how to build the best

43:26

destroyer or how to build the best refinery or new

43:28

hydrogen conversion power plant. But

43:30

we haven't learned from what's the

43:33

impact of the equipment in operations

43:35

and use. That's what we

43:37

have to figure out because you can't build new

43:39

infrastructure unless you're learning from how the old ones

43:42

are working. And this is why it's so important

43:44

to... We get... There's

43:46

so much sex appeal to building new

43:48

things. And in 10 years,

43:51

we'll have this really cool new autonomous

43:53

thing, drone, walking, humanoid, that's

43:55

going to solve all of these problems. But the problem is, there's

43:57

a lot of issues going on today. And

44:00

so the approach to solving and using

44:02

specific robots for specific jobs is actually

44:04

just to earn the right to begin

44:07

building really cool robots that are

44:09

able to do more interesting things. But

44:12

you got to get the business model right.

44:14

And the business model has to incentivize and

44:16

make a CEO or a CFO give a f*** about

44:19

how useful these industry

44:21

4.0 principles and tools are. Because

44:24

right now, that's not true. You

44:26

can hire... I see this time

44:28

and time again where I won't

44:31

name the AI companies, but these AI companies

44:33

come in and say, well, completely

44:35

turn on your head the way you're operating your

44:37

entire business and they'll come in for some contract

44:40

that ends up expiring because it did not produce.

44:42

And that's the problem. It's like you think you

44:44

have all the information and data, but you're building

44:47

your AI and your solutions

44:49

off of ground truth that's actually not ground truth.

44:52

There's actually a low amount of integrity. If

44:55

you're not interrogating the data all the way to the

44:57

ground level. And so for us, we are

44:59

building AI and software,

45:02

but off of data sets that

45:04

robots and smart centers are

45:06

actually collecting. In order to affect some

45:08

large business outcome, even in cashflow is

45:11

what we orient to, or it

45:13

could be scheduled adherence, or it can be

45:15

environmental impacts. But you have to be able

45:17

to interrogate the impact from the solutions

45:19

all the way down to what's actually causing

45:21

the change. And for us, it's very clear,

45:24

like if you can start with the

45:26

core foundation of what is the health of

45:28

everything, of the viability structures, then what we've actually

45:30

found is that we don't have to ask our

45:32

customers for data sets. They'll give it to us.

45:35

And so the end of that case study I was

45:37

going to show you is those

45:39

50 tanks, we actually were able to extend the

45:41

useful life of that one tank by

45:44

10 years and scrap it at an

45:46

$8 million CapEx expense. And we were

45:48

able to, across the 50 assets, the

45:50

site said, and did an analysis

45:53

that predicted, because of a

45:55

modulation in fill heights, we were

45:57

able to actually impact gross margins.

45:59

by about 4%. And

46:02

so there's like these, it has to be

46:04

oriented towards the outcomes. So the customer has

46:06

to buy that outcome. They can't

46:08

buy the robots. And we'll be happy

46:11

to use and integrate and we do

46:13

other sorts of robots because I don't want to build all the robots.

46:16

But it has to be again oriented towards

46:18

the big outcome and problem for

46:20

the customer. Otherwise, like, you know, it's not going to get

46:22

funded that way. Yeah, I mean, we are. There's

46:25

a use for doing deep tech. And

46:28

there's a use for just trying to make

46:30

things work in the world. But at

46:32

a certain point, they have to solve a problem. And

46:35

I think that's, I think what you learned was,

46:37

you know, the robot was one way to solve

46:39

the problem. But the sensors is another

46:41

way to do it, right? And the sensors being on

46:43

there and wow, it was

46:45

incredible progress you've made. Let's talk a little

46:47

bit about AI. We'll open the aperture here.

46:50

As you collect all this data, and over

46:52

the next 10 years, you'll have

46:54

systems fail, you'll have things you got right, you'll have

46:57

things you got wrong. You know, weird

46:59

things will happen, random things will occur, ships

47:01

will lose power and run into bridges, all kinds

47:04

of events are going to occur. And you're

47:06

going to be collecting all this

47:08

data about these things. And then

47:10

AI will be able to process all that and

47:12

maybe give us some insights. When do you think

47:14

you'll start having insights

47:17

powered by AI a human wouldn't

47:19

have gotten to in a reasonable amount of

47:21

time? And what do you

47:24

think the insights might look like? What might you figure

47:26

out collecting all this data and then,

47:28

you know, running algorithms,

47:30

machine learning, other things against the new? We've

47:33

collected now, I know, data

47:35

sets on the health and

47:37

structural integrity of over 500,000 other world's most critical assets.

47:40

And what we're doing is we're capturing this immense

47:42

amount of information as it relates to what is

47:44

going on as it relates to why do things,

47:46

where things damage? Why

47:48

are they, what kind of damage mechanism is occurring?

47:52

And Also building out machine

47:54

learning to interpret what is a sound

47:56

wave attenuation indicative of what kind of

47:58

like issues. As we've

48:00

been able to train actually on and

48:02

what is causing certain some damage mechanisms

48:04

has he been liebling for the past

48:06

eleven years. Ah, and.

48:09

So it's like it's that the I that

48:11

we've the where we all we believe there

48:13

are very very much in is like less

48:15

be masters and very excellent at being the

48:17

best in the world and understanding why things

48:19

palm are damaged. And. What kinds of

48:21

materials with and repair techniques were kinds

48:24

of an input says relates to when

48:26

what kind of variables in our leading

48:28

certain and death mechanisms. Of. To

48:30

be oh begin to inform and and

48:33

and form what to expect as it

48:35

relates to outnumber decks when something with

48:37

fail. I. Would have the and then

48:39

also on how to use Pc efficiency, movie

48:42

thermal efficiency or thats throughput or even you

48:44

know a guy a motor efficiency detecting one

48:46

of those about to fail and how to

48:48

make it may make sure you're adjusting at

48:51

to be optimum of se we want to

48:53

treat efficiencies. I'm offers dislike or information and

48:55

data sets that we have that know enough

48:57

does which is so I'll be able to

49:00

go back to the people who manufactured these

49:02

tanks be integration firms that actually install them,

49:04

the construction companies and say you know at

49:06

what we've seen when you're here. Within one

49:08

hundred miles or fifty miles of coastline, salt

49:10

water is x, y, and z. Z.

49:13

Thanks to be built in the following way for

49:15

this is what happens in extreme heat. This is

49:17

what happens extreme cold. This is what happened from

49:19

sun damage, In there might be silly things

49:21

like a certain quotas. Some seal.

49:24

In. One area my father problems and you

49:27

can even start ab testing this right you

49:29

can. You could. Tell. This person fifty

49:31

tell ya say we were a multi very

49:33

tested. We want to have. Ten

49:35

layers put on these five layers but on three

49:38

letters from his a Mutt These other ones to

49:40

be any say that I'm I'm coming up with

49:42

stupid ideas here. But. Just there's no

49:44

bad idea when it comes to testing.

49:46

He extending the life of critical infrastructure

49:48

had. And yeah, that's right

49:51

to be super powerful. Have you started to

49:53

give manufacturers like notes or installation people notes

49:55

like on how to do things better? From

49:57

the get we've not opened up arms. An.

50:01

Ad Products or Services you it says

50:03

helping improve the alien process and what

50:05

materials to choose otherwise. But. Some

50:08

but yeah the of your correct in assuming like

50:10

us were headdress. As well as assuming

50:12

Ucas lot of other stuff, all insurance companies.

50:14

Do. Iraq is ensuring all these assets

50:16

search during The downtime from his assets is

50:18

what he did as instructed by inserting as

50:21

you it's to on the care instead. Of

50:23

adopting this kind of. And

50:26

these kinds of tools because the insights.

50:28

Of the our well as my arm billion

50:30

dollars as and for second taken care of

50:32

your and as being informed by John are

50:34

up to have a lot of you could

50:36

measure it. He. And you could

50:38

manage it. And you can ensure everybody

50:41

says if you measure it you can manage it. Managing.

50:44

It in a lot of cases me, Juri and actually

50:46

Free Burgoyne This one who brought you guys up on

50:48

a recent all and pod and with my invited you.

50:51

Are muted metro Mile which was also measuring how

50:53

many miles are you doing. We should only charged

50:55

for that. And then you start thinking about Tesla.

50:57

They have a driving school. Another storm?

50:59

Yeah. But. No. One.

51:02

Of the people who was driving my car is

51:04

quite an aggressive driver has times and ah it

51:06

was like whoa. You're driving pretty close to the

51:09

person in front of you, has a distance of

51:11

has the speed and you know, zip zip zip.

51:13

And you could just make insurance for people

51:16

who are zippy in their cars and people

51:18

who are slow motion. They'll let the right

51:20

lane. And. You could just right

51:22

size insurance. What you're saying is hey with

51:24

these tanks If we're expecting them and were

51:26

doing he does is Sam. The.

51:28

Mediation. Yeah, maybe we have

51:30

a different insurance profile than somebody who does none of

51:32

them. And of openly sensors on here

51:34

from we have a different on the currents. That's

51:37

exactly what happens in journalism by the when

51:39

I first started my first magazine. You're

51:42

like, do you had to fact checking. Do

51:44

you quotes with the folks who did

51:46

it? Do you record your cause? He.

51:48

Only went through all the stuff and I'll say.

51:51

Oh wow this is really interesting. Oh my god

51:53

is is mad I google going to make different

51:56

levels of insurance based on your fantasy. Some.

51:58

Media and for a similar notice. You.

52:00

Know if you are. I. Don't know and

52:02

I'll leave know. Some people like Alex Jones think

52:04

like an extreme example who likes as conspiracy theories

52:06

of whatever. Like. Yeah. Uninsurable.

52:09

and then you go to people like

52:11

ceremony or times and. If. You've ever

52:13

been in a New York's I'm sorry. Like the they check

52:15

the fax. The. They call you and and

52:17

from the close Now. Yorker, Abbey's

52:19

and my since. I haven't had a New

52:21

York Times accepted chef, but I have to.

52:23

New York Versa. Have had Vanity

52:26

Fair chat so contrast as a really good

52:28

job of that and the insurance. I think

52:30

what's our being proportional to the effort you

52:32

put into getting your too fast for us

52:34

in here it's you know the effort you

52:36

put into getting your senses. Cricket has have

52:39

insurance companies are circling with the advent of

52:41

the result some and they've coming down by

52:43

we we may see to sell to the

52:45

approach of. That. We're really focused on

52:47

like helping improve the seat of Art or

52:49

part the smurfs large problems and or and

52:51

or insisted I station will be more incident

52:53

those kinds of models that used to were

52:55

talking about as which to oh yeah I'm

52:57

an insurance. And at some point in the

52:59

future. But. Right now it says be a

53:01

wouldn't want to build up the infrastructure and a

53:04

good architecture to begin. Implementing

53:06

this like or is an issue for for

53:08

know. Said was like talk. On.

53:11

Ends in a in a pragmatic way is also trying

53:13

to me to customers were there and I'm in

53:15

a lot of these allowed his customers have a hard.

53:18

Have a hard time in a very

53:20

adverse to technologists. Software

53:22

and robotics like those coming in because it

53:24

just have not seen. The. Impact towards

53:26

like helping them were up by every the

53:28

fires of they fight every single day. And.

53:31

So they're not actually that willing to give

53:33

you much. Information. To

53:35

help you build a good products that. I

53:37

think this is like. This is why the

53:39

death of so many. You know we're not drunk.

53:41

Companies are about as companies are software companies. As

53:44

occur in this sector and because one. Venture

53:47

capitalists have no idea what these sectors and

53:49

what they're talking about. And. So like. And

53:51

we're very lucky because we're this like Pittsburgh

53:53

provides company focus on and Aziz are all

53:56

the wrong things. Back into the sixty one

53:58

wants to? I see a twenty six. Yes.

54:00

In 2024, now everybody's got the

54:02

bug right after they've seen what's

54:04

happened with Tesla and SpaceX. Absolutely.

54:06

And that also led you up to some

54:08

of these boring industries or real

54:11

world industries might be worth going

54:13

after. Yeah. And it was also

54:16

Uber and Airbnb were also real world businesses. I remember

54:18

when they were raising their funding, people were like, I

54:20

don't want to be in a real world business. Too

54:22

dangerous. What if somebody trashes your apartment? It's like,

54:25

well, people trash hotels every

54:28

weekend. Yeah. Kind

54:30

of what hotels are for, at least amongst

54:33

addicts and rock stars is we're trashing them. And

54:36

hotels have figured out how to deal

54:38

with a trash hotel room. They just throw everything in

54:40

and they charge the person money for trashing it. The

54:42

end. Yeah. Part of the game. Again,

54:45

here in Pittsburgh, there's so

54:47

many robotics companies that start night all the time

54:49

and it's began because it's not because they are

54:51

really dumb at building great robots and solutions. They're

54:53

actually really smart. But the problem is, what are

54:55

the robots useful for? Yes.

54:58

Yes. And that's like the big issue. And

55:00

that's why I spend so much time trying to dig in with

55:05

the customers in an embedded way. Because

55:07

if you aren't embedded... Yeah. The bear hug is so critical.

55:10

If a founder gets anything out of our tower

55:12

together, it's the bear hug works.

55:14

Being on location. And it was a

55:16

famous story. I think Paul Graham told of telling or Jebia

55:20

told it on this podcast, the co-founder of Airbnb

55:22

said, all the customers were

55:25

in New York and Paul Graham told

55:27

them, go to New York. And he said,

55:29

all the places with good pictures, get

55:31

rented. The places without pictures don't get rented. He said,

55:33

go to New York and take pictures and get a

55:35

quick camera. And they literally bought a

55:37

digital S-R and started taking great pictures,

55:39

I think. Literally, Brian

55:42

and Joe took the pictures themselves as

55:45

the co-founders. And this is like the closer

55:47

you get to the customers, the closer you

55:49

get to the truth. It should seem obvious,

55:51

but it's scary to talk to customers for

55:53

some introverted builders, engineers, whatever. You just

55:55

got to be right there at their desk, sitting

55:58

side by side with them, The

56:00

prom together at any particular did. Some customers don't want

56:02

that, right? but you only need one or two to

56:04

say yes and then they get the benefit. So if

56:06

you're on the customer side of this. If.

56:09

You at the start up in bed with you. And.

56:11

Black in bed, in bed with any view.

56:13

It better start up the your company. you

56:16

get all the gains. Years before

56:18

your competitors. So if you're in a big company

56:20

embed, those, start and take a risk with them.

56:22

And. Help them build a future with you. That's what you

56:24

were able to do. With. A so brilliant

56:26

and existing that that they get support

56:28

from a society and to is like

56:30

in a very regulated environment where especially

56:32

this one a monopoly monopolies a play

56:34

with their be the government's that sector

56:36

or a be the energy sector. Like

56:39

changes like yet very disincentivize. See don't

56:41

want so I changed for you your

56:43

you know you're maintaining so the that

56:45

you go boom ends up kill people

56:47

and sit down. You know a refinery

56:49

as making fifty million bucks That. So.

56:52

That's actually not that intuitive. Or

56:55

accepted for folks to do.

56:58

And. Say hey kid, come on,

57:00

give your best shot that when I

57:02

was effective was actually in the power

57:04

sector and specifically the fossil fuel cell

57:06

receptor. Woodward is like. L

57:08

like I need. I need help because. Ah,

57:11

young, I get less funding of that,

57:14

less people have less expertise, and my

57:16

demand is actually like pretty. I still.

57:18

And some and like I'm having shutdowns of

57:20

my our plants fifty percent of the year.

57:23

Because pressure vessels as keep exploding. And

57:25

only way to look at this is what's at

57:27

stake in our I saw Founders Like how much

57:29

is at stake here and if you're doing. The.

57:33

Ah family trip planning out every time we

57:35

get pitches in like every hundred and one

57:37

or we don't have one is a meeting,

57:39

an app that take your group chat and

57:41

much a plan a trip in an app.

57:44

And you're like. Not a lot at stake. And.

57:46

The solution of doing it. And. I message

57:49

room. Whenever what's app you're

57:51

into. A worse I just signed

57:53

the second is. Like. there's not

57:55

much at stake here like splitting the bell

57:57

it's like it's one hundred hours and mexican

58:00

you have to split it four ways, nobody cares, it's not enough

58:02

at stake. Then you start looking at, hey,

58:05

getting to space, putting stuff in space,

58:07

space acts a lot at stake, self-driving,

58:10

getting from point A to point B, that's

58:12

actually a lot at stake in Uber's

58:14

business or Airbnb's, like I'm going on vacation, I

58:16

need a place to stay, there's a lot at

58:18

stake there. And what you found is, if

58:21

one of these things fails, that's a half

58:23

billion dollars. And insurance companies,

58:25

people lose their jobs at the

58:27

company, people get sued, that's

58:30

a lot at stake. And as

58:33

you said in that one example, you extend that one tank,

58:35

you save eight million bucks. And you probably

58:37

made, what, 800,000 off that customer or

58:40

80,000 off that customer? Yeah, a

58:42

little more than that, but yeah, it's a- A little more than 80

58:44

or a little more than 800? A

58:47

little more than 800. Okay, so essentially, if

58:49

you made a little more, you

58:51

were 15% of the cost of the other

58:53

reality. So

58:56

they got 85% of the benefit, you got 15. Pretty

58:59

happy. Fall apart. Yeah. That's

59:03

where I think technology is at its best, when the customer

59:05

gets the bulk of the gain and the company gets a

59:07

small portion of the gain, it makes it a no brainer. Yeah,

59:11

and also you have to understand that these

59:13

sectors are trying like hell to figure out

59:15

how to adopt technology and

59:17

not be sold a bag of goods that

59:20

is false. And

59:22

so there has to, what you end

59:24

up, you end up have to do is create a

59:26

model that, very clearly you can

59:29

backtrack into, where is the value creation happening?

59:31

But also, how do I sort through like

59:33

the 10 to 20 different options

59:36

for robotics and drones and AI

59:39

companies? That's really tough for

59:41

these large organizations and they really just want someone

59:43

to come in and solve a bunch of their

59:45

problems. And so if you can create an environment

59:48

where you can bring in and vet technology, you

59:50

can vet different kinds of robotics

59:53

tools, sensors that

59:55

are enabled by some smart technology,

59:57

you end up putting together the

59:59

different pieces. that make up some large outcome that you're

1:00:01

trying to solve for the customer. Packaged

1:00:03

though in a software

1:00:05

that helps to centralize decision making

1:00:08

and very clearly articulates where

1:00:10

the value creation is coming from and

1:00:13

you can interrogate how the decisions were

1:00:15

made and what inputs led to the

1:00:17

improved outcome. So you actually need to

1:00:19

help, in order to have a

1:00:21

lot more startups enter the sector, you need

1:00:23

to actually create a model that very clearly

1:00:25

articulates what the product market fit needs to

1:00:27

be or what the problem you have to

1:00:30

solve needs to be or the data layer

1:00:32

that needs to be added to the stack,

1:00:34

needs to end up looking like or what

1:00:36

kind of information you end up collecting that's

1:00:38

not currently out there. That

1:00:42

actually is pretty, we're going to do this

1:00:44

with now it happens in other robotics companies

1:00:46

where we're like, hey, come under our contracts

1:00:48

and we really love the solutions that you're

1:00:50

building. You can come under our

1:00:52

contracts and add these different solutions. Wonderful. And then you

1:00:55

can say, hey, you've got a great drone, a walking

1:00:57

drone, underwater drone, we don't have it. Yeah,

1:00:59

we'll plug it in. Here's the API. Let's

1:01:01

rock. And this is where I see

1:01:03

the humanoid robots going. It's like these are really

1:01:05

complicated problems to solve and the data that the

1:01:07

robots collect in the real world is interesting, but

1:01:10

it's not actually super valuable to some customer that's

1:01:12

trying to solve. How do

1:01:14

I increase the efficiency of

1:01:16

my batch process of making a roll

1:01:19

of steel? So the

1:01:21

robots can do interesting tasks and can

1:01:23

actually observe interesting data in the real

1:01:25

world and get information that's not previously

1:01:27

available. However, what is the use of

1:01:29

that information as it relates

1:01:31

to solving some large outcome for a client? So

1:01:34

that's how I'm excited about humanoids and walking

1:01:36

dog robots because that offers different data layers,

1:01:39

but they're one of a couple different

1:01:41

data layers. There'll be a thousand flowers

1:01:43

that are going to bloom in robotics. These little ones

1:01:46

to carry your burritos from point A to point B.

1:01:48

If you just watched Star Wars or any modern

1:01:50

science fiction, you're going to see a range of

1:01:53

robots. And you know, science

1:01:55

fiction authors and directors and

1:01:57

creatives, they really do think about

1:01:59

human use. cases and

1:02:01

sure enough, these little robots that would

1:02:03

scurry past Darth Vader's feet look

1:02:06

just like the ones that are delivering burritos in a

1:02:08

lot of major cities. And sure, we'll

1:02:10

have a C3PO, we'll have an R2D2, we'll have

1:02:13

everything in between. And the

1:02:15

bomb, the build of materials on

1:02:18

your robots is a little bit high because you have

1:02:20

some, I think, some really intense sensors. So

1:02:23

those look like those could be tens

1:02:25

of thousands of dollars, I assume, in

1:02:27

terms of the bomb? Yeah, it's like

1:02:29

upper, it's like close to six figure

1:02:32

is what we're at. But it's not, we're not optimizing

1:02:34

for the bomb. But yeah, that's right. But when

1:02:36

you look at the general robot, optimus or

1:02:38

figure or some of these, the bomb on

1:02:40

those is going to be what do you think? If

1:02:43

you had to pick a number five years from now, what's

1:02:46

the build of materials? And then we could

1:02:48

extrapolate pricing for consumers after that. What

1:02:51

do you think like a functional

1:02:53

robot that could walk your dog, or

1:02:56

I don't know, do your dishes or I

1:02:58

don't know, tidy up around the house or work in a

1:03:01

factory? What do you think without the specialized sensors, what do

1:03:03

you think the bomb on one of those is going to

1:03:05

be? It's going to be interesting because you also

1:03:07

like to think about what kind of certifications like the

1:03:09

robot has to have or come under. But yeah, I

1:03:11

think it'll end up being, it's going to be hard

1:03:14

for me to imagine it's below 40 in

1:03:16

five years. I think it's actually

1:03:18

a lot higher than that. And I think-

1:03:20

Because there was early on, but ultimately you think a

1:03:22

$40,000 bomb? Yeah, but ultimately

1:03:24

I think a $40,000 bomb makes sense in the next like

1:03:26

seven years. That's higher than I thought. I thought it was

1:03:29

going to be more like 20 or something. Yeah, in the

1:03:31

next seven years, I think it'll end up going down basically

1:03:33

just based on what kind of volume. So I'm not assuming

1:03:35

like in seven years, a lot of volume. If there's a

1:03:37

lot of volume, then I'd probably estimate it's

1:03:40

closer to the 20K. I think it'll get

1:03:42

to 10. It'll get cheaper than that

1:03:44

coming out of China. Absolutely. Yeah. So

1:03:46

40 when they launch, 10

1:03:49

ultimately when they're commoditized, everything

1:03:51

in between. And what are the

1:03:53

major costs you think in that robot? What are

1:03:55

the top two or three costs that you're going

1:03:57

to need? The actuators. Oh.

1:04:01

I. Get computers Also gonna be like expensive a

1:04:03

much rather be dealt with. he does it have

1:04:06

like the equivalent of. You. Know a

1:04:08

h one hundred powering it or does it

1:04:10

have like a math book and it's connected

1:04:12

to the net? Yards. Like a very

1:04:14

interesting lesson. Yeah. The it'll be there

1:04:16

be a lot of robots that like a certain robots

1:04:18

that was more b o to go into certain environments.

1:04:21

that's like certified for explosion for. News.

1:04:23

Like those are expensive. He and. He

1:04:25

goes to get a certification to ensure that

1:04:28

they like won't like com us present new

1:04:30

satellite comes to mind for the actuators our

1:04:32

what make the move their arms there wouldn't

1:04:34

be the equivalent of your joints essentially in

1:04:36

there and the the pulley system to move

1:04:38

things around those are not cheap right now.

1:04:41

Yeah. And like find dexterity is like those

1:04:43

are England tricky the hands yeah we we

1:04:45

had a company rude and I that was

1:04:47

picking strawberries and with the mit hand area

1:04:49

in some of these m I t hands

1:04:51

or so and credible what they are. tip

1:04:53

of the when the we have passing I.

1:04:56

Couldn't. Have coffee cups and me by latte and

1:04:58

putting foam on them and we thought it would crush

1:05:00

the top and how does it do it? Cuts are

1:05:03

easy like. We're. Working with berries

1:05:05

and. Will. For your work toward

1:05:07

bearing on strawberries and. Off

1:05:09

and raspberries like return a fragile berries

1:05:11

often stems is not an easy task

1:05:13

when you think about it. And.

1:05:16

But I guess in some ways of his. And

1:05:18

then he did. You could actually see these. Being.

1:05:21

Rented. For. Ten bucks a

1:05:23

day. Twenty bucks a day, You know, And bucks

1:05:25

a day is thirty. six hundred ozzy. Or Twenty

1:05:27

bucks a day. Seven thousand a year. For.

1:05:30

Twenty bucks a day is what people spend

1:05:32

our lunch now. So twenty bucks a day

1:05:34

to have robots? Pretty dope the my mind.

1:05:37

I think maybe less about like

1:05:39

the commercial landed said the be

1:05:42

a C implications. And. Most is

1:05:44

because like. The amount you've

1:05:46

dispense on making the getting that lasts

1:05:48

ten percent of robotics and the last

1:05:50

time. As good as

1:05:53

a nice it's he had the as cases are

1:05:55

just so hard. expensive so. In. My

1:05:57

opinion is is more a line see

1:05:59

like what. Evaluate feeding from the robots and then

1:06:01

if he can create a lot of value in charge

1:06:03

a lot and he can. Send. Can justify

1:06:05

like a large amounts of spend on onset

1:06:07

making some really cool robotics. I think that's

1:06:10

the key that most roboticists haven't actually saw

1:06:12

for is what is the what is that

1:06:14

are you feeding and how much me second

1:06:16

of a ph on what you do that

1:06:19

the never a vicious cycle of being able

1:06:21

to optimizers robotics to the really cool things

1:06:23

bow and I think that's. That's

1:06:25

the that least we're approaching. It's them because

1:06:28

it allows you know. Bible. Enormous

1:06:30

user when and where they seemed good at picking

1:06:32

markets, where to send the first. Human.

1:06:34

Right Robotics to to maximize the

1:06:36

business model. Soldiers. Wilders.

1:06:40

Sodas, Army soldiers

1:06:42

come to mind to be think about how much money

1:06:44

we put into a soldier. I. Mean I had

1:06:47

a friend who is agree barrage like I'm like of us

1:06:49

he told me was like a family and our ass at

1:06:51

he said the seals are like a twenty million dollars at

1:06:53

each. You. Know. I've. Been on

1:06:55

are they wanted of training? You know? I.

1:06:57

Think robotics will not to use in warfare

1:06:59

and my says like some large afloat. And.

1:07:02

Then it'll be like. That.

1:07:04

We have bigger problems than robotics anything

1:07:06

or any problems and yeah a but

1:07:08

I think it's like he can't some

1:07:10

like a robotics and to some village

1:07:12

because like the edge cases right it's

1:07:14

like there's so much potential issue and

1:07:16

then like and you're dealing with the

1:07:18

colors pr problem or darn it was

1:07:20

memory I think it's or thing in

1:07:22

a d where the costs of where

1:07:24

the human exposure in the cost of

1:07:26

potentially having like a large issue it

1:07:28

is a like are some the ocean

1:07:30

violation or wasn't like that survive minus

1:07:32

go to like. What? Is the most

1:07:34

and. Fifty. One of them, isn't it

1:07:36

dangerous species like the one I see my head I was

1:07:38

going to because it's. A and that's

1:07:40

the most. One was dangerous Shards illicit sex

1:07:43

is A for years and sounds and wars

1:07:45

of years Wow that's his are under it's

1:07:47

It's like that people get be like hundreds

1:07:49

of thousands of dollars. And. To

1:07:51

get a job I think like five hundred k and

1:07:53

was like the you're going rate for like a undersea

1:07:56

while there but like you like law disease like so

1:07:58

sad. Yeah,

1:08:00

trees are dangerous. Rufors,

1:08:02

construction workers, truck drivers, miners. I don't

1:08:04

think roofers just because it's, again, I

1:08:07

don't know how many people know this.

1:08:09

Firefighters also very dangerous, running into burning

1:08:11

buildings. I think, but I think it's like, what

1:08:13

is the cost? Like

1:08:15

what is the value you can create, like from the

1:08:17

information that the robots are collecting? I think, your

1:08:21

mind's going more to like labor, which I

1:08:23

think makes sense, but I think I'm more

1:08:25

interested in like what kind of- That's where

1:08:27

chat GPT is mined when, I just did

1:08:29

a chat GPT, what's the most dangerous professions.

1:08:31

Logging's up there. I mean, you think about

1:08:33

it, logs falling everywhere and like heavy duty

1:08:35

machinery with chainsaws and blades. And if you

1:08:38

ever seen those logs like rolled down a hill, man,

1:08:40

you're dead. You get hit by one of those. I

1:08:43

think that it'll end up just being oriented towards,

1:08:45

yeah, maybe on the labor side, they're still welding,

1:08:47

but I actually think it's more just like walk

1:08:49

down to that refineries. It's, you know, it's like

1:08:52

welding and perfecting the weld, speeding up the time

1:08:54

of the weld, but it's more oriented towards like

1:08:56

what kind of new information am I getting in

1:08:59

a way that helps to improve the

1:09:03

overall state of a, let's say like of the organism that

1:09:08

is like making a roll

1:09:10

of steel or paper products

1:09:12

or refining petroleum or making power. That's

1:09:14

one of the things. Well,

1:09:16

it comes to mind as a good first step too, right? I

1:09:19

think that's what Elon thinking is. If I get them to work in the

1:09:21

Tesla factory on repetitive tasks and a control

1:09:23

of the system, I think that's what Elon is doing.

1:09:26

He's doing tasks in a controlled zone. So I'm assuming, no

1:09:28

treatments to get run into. Right. What I'm

1:09:31

thinking more of is like what kind

1:09:33

of information is being collected by the robot that

1:09:35

helps improve an overall process that a human, you

1:09:37

know, is

1:09:39

not constantly streaming data somewhere, right? It's

1:09:42

constantly streaming to your head. And that's never gets. That's an

1:09:44

example of that. That's an example

1:09:46

of it. So you're walking, let's say like you're

1:09:48

walking down a refinery and you're looking and

1:09:50

trying to listen to

1:09:52

the kinds of noises that might be indicative of some

1:09:54

like leak somewhere, or like you're looking at, you

1:09:56

know, you're trying to like look at like temperature transmitters

1:09:58

and see if there's like. some kind of inconsistency

1:10:00

of temperature that like dilute to something going

1:10:03

boom. You can use like things like thermal

1:10:05

cameras, you can use things like, you

1:10:07

can use things like LIDARs as well to like constantly update

1:10:09

like what is the, what's the process

1:10:13

of refining petroleum.

1:10:17

You can begin to like incorporate different kinds

1:10:19

of pieces of information that can tell

1:10:21

you how efficient is your facility operating

1:10:23

at and then update whatever model

1:10:25

you're using and change the way that you're

1:10:28

actually operating to the facility because ultimately what

1:10:30

they would learn in the field that could

1:10:32

be incorporated into a better process that you're

1:10:34

saying. We know very little about

1:10:36

what's going on in the

1:10:39

real world and so like robots or like

1:10:41

cars that are going around with LIDARs spending all

1:10:43

the time. It's like very interesting information and data

1:10:45

that we've been monetized in like interesting ways that

1:10:48

we don't even know, hear about or talk about.

1:10:50

But it's that same sort of.

1:10:52

No population density. They know how

1:10:54

popular Broadway is in your town at

1:10:57

one o'clock on a Sunday. And

1:11:00

measuring emissions is a good thing too. It's like if you're

1:11:02

like walking, if you have a robot walking around and

1:11:05

doing tests of how much CO2 is coming

1:11:07

out of my stack. It's

1:11:11

like these are like interesting, different

1:11:14

kinds of information that can drive certain

1:11:16

sorts of large outcomes. Maybe it's like some sort of

1:11:18

premium you can get from the inflation reduction act or

1:11:20

something like that. It's fascinating. I think it's gonna be

1:11:22

like a brave new world. Can't wait for these things

1:11:25

to come out. All right, listen,

1:11:27

gecko robotics. Another

1:11:29

overnight success, 11 years in the making.

1:11:32

Congratulations, keeping us safe. I

1:11:35

mean, I was just thinking about that building in

1:11:37

Miami. Remember the pool and that building collapsed in

1:11:39

Miami? Yeah. Man, they had

1:11:41

just, and they kind of knew

1:11:43

that it was messed up, but they just didn't take it

1:11:45

seriously. They didn't inspect it. Man, you

1:11:47

get a couple of those happening. And there

1:11:49

are other countries where the building standards are

1:11:51

not like the US and that happened in

1:11:53

the US. Man, I don't know

1:11:56

what developing nations now are getting rich and

1:11:58

have a lot of buildings maybe

1:12:00

when they weren't as rich and there

1:12:03

weren't as much regulation going back and figuring out

1:12:05

hey these buildings built in you know

1:12:07

I'm thinking of emerging countries that are now

1:12:09

we don't use the term

1:12:11

first and third world anymore but frontier markets

1:12:13

turning into emerging markets turning into primary

1:12:16

markets they're gonna need to inspect some of

1:12:18

that previous infrastructure and make sure it's well

1:12:21

what yeah that's it there's interesting stats like there's 700 there's 1700

1:12:23

17,555 bridges in New York and I think six

1:12:29

was the latest are not in need of

1:12:31

immediate repairs it's like this

1:12:33

stuff's old and and

1:12:35

the various arrows and it was

1:12:38

rusted and gross that's like

1:12:40

one of the premier bridges in New York it's

1:12:42

really sad to see and then also just like

1:12:44

you don't think about it in the US you

1:12:46

can reduce row does

1:12:49

a finishing study 18% is

1:12:51

the as the reduction in

1:12:53

US emissions by 2030 if you can stop

1:12:56

critical assets from failing and exploding within the

1:12:59

oil and gas manufacturing sector so like these

1:13:01

are like pretty interesting you know

1:13:03

connection points into how important it is to understand

1:13:05

the health of the built world that

1:13:08

most people don't think about yeah and it's

1:13:10

that's a hard one to sell on unless there's

1:13:12

just been something terrible that's happening on an infrastructure

1:13:14

basis and people are highlighted to because people don't

1:13:17

want to talk about the reality of another

1:13:19

BP oil spill in the Gulf or another bridge

1:13:21

collapsing it's just it's dark to think about it

1:13:23

but great that there are people like you out

1:13:25

there solving these problems so the rest of us

1:13:28

can feel safer great job Jake's wish you continued

1:13:30

success and we'll see you all next time on

1:13:32

this weekend startups bye bye