0:00

The DOD is a black box and

0:02

I say that meaningfully because we're trying

0:04

to fix it We want to

0:06

overwhelm the system with how much

0:08

better it can be The US Army

0:11

should never trade blood for blood in

0:13

first contract and it should always be

0:15

blood for an iron and it should

0:17

always be their blood In 2025 when

0:19

people think of government speed or

0:22

innovation aren't typically words that come

0:24

to mind, but this was not

0:26

always the case In the early

0:28

40s, we built the Pentagon in

0:30

16 months. Shortly after, the Manhattan

0:32

Project took three years. When the

0:34

Soviets launched Sputnik in 57, it

0:36

took a mere 84 days for

0:38

America to respond with Explorer 1,

0:40

kicking off the space race. In

0:42

the 60s, Kennedy asked for a

0:44

man on the moon, and the

0:46

United States Apollo program answered within

0:48

the decade. And here's the thing. The

0:50

very roots of our technological bronze

0:52

as a country are deeply rooted

0:54

in the federal government. But in

0:56

recent decades a gulf has been

0:59

developing between Silicon Valley and Washington

1:01

DC. But a renewed interest is

1:03

developing and founders building toward the

1:05

national interest. But many have no

1:07

clue how to navigate the black

1:10

box that is government procurement. This

1:12

is not a technology problem. This

1:14

is a culture and a process

1:16

problem that we can apply technology

1:18

to. So in today's episode, recorded live

1:21

at our third annual American Dynamism Summit

1:23

in the heart of Washington DC, we

1:25

bring in two chief technology officers from

1:27

two of the most important institutions in

1:30

the nation. Alex Miller is the CTO

1:32

for the chief of staff of the

1:34

army, and Justin Finelli is the CTO

1:36

for the Department of the Navy. Listeners

1:39

of this podcast may know what a

1:41

CTO does in the private sector, but

1:43

what does this mean in the public

1:45

sector, especially when... The CTO position within

1:48

government came 30 years after it did

1:50

private sector, right? And even later than

1:52

that, after defense. Joining Alex and

1:54

Justin is our very own A16Z,

1:57

go-to-market partner focused on American dynamism,

1:59

Layla Hay. Together, the three discuss

2:01

important topics like how we accelerate, shifting

2:03

from planning years ahead to software speed.

2:05

The entire way we think is predict

2:08

the future three to five years out,

2:10

land the perfect shot, get a trick

2:12

shot, and go. And it just doesn't

2:14

work. It does not work. Also, the

2:17

reality of resources on the ground. It

2:19

is shocking to me how much stuff

2:21

we buy for soldiers that no one

2:23

would accept if I handed it to

2:26

you in your everyday life. And importantly,

2:28

whether our leaders have the leeway to

2:30

fix what's broken. We discuss all this

2:32

more, including the role of culture in

2:34

bridging the divide between Silicon Valley and

2:37

DC, plus the changes ushered in by

2:39

the new administration. So do we need

2:41

to rip up the system and start

2:43

over? Or are we already on the

2:46

path to a solution? Listen in to

2:48

find out. As a reminder, the content

2:50

here is for informational purposes only. Should

2:52

not be taken as legal, business, tax,

2:55

or investment advice, or be used to

2:57

evaluate any investment or security. and is

2:59

not directed at any investors or potential

3:01

investors in any A16Z fund. Please

3:03

note that A16Z and its affiliates

3:06

may also maintain investments in the

3:08

companies discussed in this podcast. For

3:10

more details, including a link to

3:12

our investments, please see a16c.com/disclosures. I'm

3:14

the CTO for the Chief of

3:16

Staff of the Army and that

3:18

is a big title because the

3:21

Army is a big place. In

3:23

my day job it's a lot

3:25

of education. It's I would say

3:27

probably 60-40 talking to people about

3:29

what technology can do for a

3:31

mission and then 40% actually hands-on

3:33

like putting my engineering cap back

3:36

on and actually doing things that

3:38

moved the ball forward. I would

3:40

say very similarly to a company.

3:42

What I do as the CTO

3:44

for the chief is think about

3:46

what the North Star should be

3:48

for the army in terms of

3:51

how we leverage technology how we

3:53

employ it and then what our

3:55

missions are that can leverage it

3:57

and then actually go forward and

3:59

helping people get there while companies

4:01

have boards of directors and presidents

4:03

and all these different things. We

4:06

also have a board of directors

4:08

that is Congress. We also have

4:10

stakeholders that are the American people. However,

4:12

most companies don't have 60 years of

4:14

policy and rules that were written generally

4:17

by enthusiasts and visionaries but implemented

4:19

by minimalists and what I mean

4:21

by that is risk minimalists. So

4:23

most of our day job is just going

4:25

through educating, educating. having a unique understanding of

4:27

what our job is, what the mission is,

4:30

and having done it to make that connection

4:32

happen. And what did you do before you were

4:34

the CTO? Before I was the science and

4:36

technology advisor for the Army G2, and that

4:38

is the deputy chief of theft that runs

4:41

intelligence. So I was the guy doing all

4:43

the technology for Army military intelligence. I got

4:45

to do some pretty cool stuff work with

4:48

some really interesting people, and that was where

4:50

all my downrange time was as an intelligence

4:52

professional. Justin, how about you. Justin Finelli,

4:54

Department of Navy, so Navy and Marine

4:56

Corps, Chief Technology Officer, just like Alex

4:59

said, Navy is a big place, oceans

5:01

are two-thirds of the world. We want

5:03

to cover as much ground as possible,

5:05

and the CTO position within government came

5:07

30 years after it did private sector,

5:09

right? And even later than that, after

5:12

defense, this is really a late acknowledgement

5:14

or a nod that commercial technology

5:16

can be a disruptive force. positively

5:18

for the government and that we

5:20

can go much faster, but we

5:23

need digital to do that. And

5:25

so to Alex's point, finding innovations

5:27

from non-traditional partners and at the

5:29

edge and scaling those much quicker

5:31

and tying it to some sort

5:33

of divestment because that money is

5:36

locked in that three-year period where

5:38

we're spending on existing tech. We're

5:40

out there scouting very similar to

5:42

how private sector would do. We're

5:44

working with hundreds of internal organizations

5:46

in some cases with different kind

5:49

of sub-optimized or very specific goals.

5:51

How do we make that a

5:53

reconciled space where we can just

5:55

bring the biggest outcomes to the

5:57

hardest problems? And so that allows.

5:59

piece to bring ultimately outcomes overmatch is

6:02

our goal. So you both mentioned industry,

6:04

you both mentioned startups. My colleague Ryan

6:06

and I actually just wrote a blog

6:08

post about selling into the DOD because

6:10

there are these paths that startups can

6:12

pursue. We all know about SBIR, STTR,

6:14

there are great organizations like DIYU and

6:16

AF works that are out there looking

6:18

to help pave these paths. But for

6:21

a lot of startups, working with the

6:23

Army or the Navy still feels like

6:25

a black box. Can you walk us

6:27

through what kind of pathways are evolving

6:29

and new opportunities for industry to engage

6:31

with your organizations? Let's talk lexicon first

6:33

and I know it's a little bit

6:35

bureaucratic but I think it's important. So

6:37

startups generally focus on like the R&D

6:39

side so they're taking a concept up

6:42

through some type of development into that

6:44

golden mentally viable product phase and the

6:46

P is really important because it's not

6:48

mentally viable prototype which I think a

6:50

lot of DOD thinks about in terms

6:52

of like hey that sort of looks

6:54

like it's useful but what startups are

6:56

thinking about is hey what's that first

6:58

product that I can productize and get

7:00

to market and get to market for

7:03

that phase. The D-I-U-U-S, the works, whether

7:05

it's F-works or softworks or Eagle works,

7:07

like all of these different organizations, that's

7:09

exactly where we want them to be.

7:11

The reason I say that is because

7:13

the Do-O-D is a black box, and

7:15

I say that meaningfully because we're trying

7:17

to fix it. that black box was

7:19

set up on purpose because in the

7:22

60s and 70s coming out of the

7:24

Cold War and all the way up

7:26

through the 90s we had all these

7:28

different things the Packard Commission for how

7:30

you think about buying things the Klinger

7:32

Cohen for how you do information technology

7:34

goldwater nickels on how you connect mission

7:36

with buying things and all of that

7:38

is still the rules that we live

7:40

under and everything since then's been duct

7:43

tape and bubble gum. So what we

7:45

are trying to do is instead of

7:47

MacGyver in our way through that, which

7:49

all of us have spent a lot

7:51

of time doing, we are trying to

7:53

like sort of domicly cut through all

7:55

of it, get rid of all of

7:57

this extra bureaucracy, work with the hill

7:59

so that it's not a black box.

8:02

It's like the 12-step program. If you

8:04

can't admit there's a problem, you can't

8:06

actually solve the problem. That's sort of

8:08

the left side of the equation. The

8:10

right side is what the Army is

8:12

doing is just bringing industry in and

8:14

saying, hey, we know we have a

8:16

problem, we want to help, whether it's

8:18

helping us write requirements better so that

8:20

we know what's available, so we're not

8:23

trying to divine that from nothing, into,

8:25

hey, how do we bring consortiums together?

8:27

I love the new software strategy from

8:29

the sector because it tells us you

8:31

use other transaction authorities, like use the

8:33

things that Congress has given you and

8:35

do it aggressively. So putting consortiums of

8:37

teams together with industry to say, I'm

8:39

not going to pretend like I know

8:41

everything. We've done that for 50 years.

8:44

No more. You tell us what's available

8:46

and we'll tell you the mission and

8:48

how we can apply to that. We

8:50

want more game changers in this space.

8:52

And so when we have barriers, it

8:54

stops outcomes. And so in general we

8:56

have more companies than we've ever seen

8:58

before who want to work on national

9:00

security. There's a in general recognition that

9:03

this world can be a dangerous place

9:05

and the more secure we are, the

9:07

better we are at this, the more

9:09

overmatch there is, the more that we

9:11

can have peace through strength. And this

9:13

is in my opinion probably the best

9:15

alignment that we'll have in our lifetimes.

9:17

This is potentially a century anomaly for...

9:19

actually being able to die vest, it's

9:21

very hard to turn things off and

9:24

then bring new players in and bring

9:26

new waves of ultimately positive disruption through

9:28

and through. And so that translation piece

9:30

has been very expensive and so product

9:32

market fit is slow and at times

9:34

we turn away companies that would have...

9:36

brought in breakthroughs, but only superior military

9:38

technology can credibly deter war. And the

9:40

fact that we have more people interested

9:42

in supporting that mission, what does the

9:45

go-to-market strategy look like? And so Alex

9:47

talked on, hey, we have some of

9:49

these front doors. We don't want a

9:51

hundred front doors. That's confusing. So how

9:53

does streamlining that look like? If you're

9:55

going to completely breakthrough and change the

9:57

game and you're selling something that we

9:59

do. don't buy right now, you're swimming

10:01

upstream. And so Andrew has made it

10:04

through and fought that fight. SpaceX has,

10:06

Palantir has, but number one, we want

10:08

that to be more straightforward. In that

10:10

particular case, the PEOs are not your

10:12

first customer. Sorry, Program Executive Offices. Maybe

10:14

we can just quickly talk about what

10:16

a DOD program office is because I

10:18

think that's another black box challenge for

10:20

startups. I think a lot of companies

10:22

know that you generally want to be

10:25

on that journey. You know, you get

10:27

some R&D dollars, you work with DIA,

10:29

you get an OTA, but ultimately startups

10:31

want to get to a program of

10:33

record. Yeah. Can you talk about exactly

10:35

what that is and what that is?

10:37

program executive offices. They buy at scale.

10:39

These PEOs are made up of program

10:41

management offices, and then you have hundreds

10:43

of those. Navigating that to figure out

10:46

where to sell is a scale game.

10:48

and it's been too hard. So one

10:50

thing that we've started to do within

10:52

Department of Navy, we have 18 of

10:54

those 75, is we've said, wouldn't it

10:56

be easier if we made data-driven decisions

10:58

and started treating these buys not as

11:00

sacred program of record, but let's buy

11:02

capability per dollar? And so our move

11:05

towards military moneyball has been to open

11:07

up the aperture through converting programs of

11:09

record and program offices to portfolios. And

11:11

so we've shifted. PEO Digital and I

11:13

believe that we'll shift even more of

11:15

those program executive offices across DOD to

11:17

that. Now that's for a Horizon One

11:19

capability. So if something improves the way

11:21

that we're doing business, if we already

11:23

have it budgeted, then you can go

11:26

directly to them. And there's actually an

11:28

index online that shows, I think Steve

11:30

Blank put it out, that says here's

11:32

what each PEO does, here's where they

11:34

are, here's where you can talk to

11:36

within them. The goal there because they

11:38

have fixed budgets, because they have fixed

11:40

budgets. is finding something again to turn

11:42

off because no one has runway to

11:44

do three years of piloting until you

11:47

transition something. If you are doing a

11:49

Horizon 2 capability, we're working. on a

11:51

breakthrough, but it's existing technology or there

11:53

is a budget line. We have now

11:55

more value of death funds than we

11:57

used to. We have small business innovative

11:59

research, but that's more of a door

12:01

prize. You need to get to that

12:03

operations and maintenance long tail funding. We

12:06

have something called at fit. Accelerating for

12:08

procurement for innovative technology. There you go,

12:10

right? It allows some breakthroughs to get

12:12

pulled all the way through. There's just

12:14

not. enough of those, right? There's just

12:16

not enough funding and not enough awards.

12:18

And so the point is, if you

12:20

can shrink the valley of death, you're

12:22

doing this in the Horizon 2 and

12:24

Horizon 1 space. For three, you probably

12:27

have to do still like some guerrilla

12:29

marketing and advertising, pull in the DiU,

12:31

pull in the requirements writers to make

12:33

sure that you're not doing all the

12:35

translation yourself. So what if we didn't

12:37

do all that? The cool thing is

12:39

everything that Justin has just said is

12:41

a workaround. It is a workaround to

12:43

a system that is fundamentally broken. The

12:45

cool thing is, PEOs are not actually

12:48

in statute anywhere. They don't exist in

12:50

law. We made them up because the

12:52

rule is that the SEC deaf gets

12:54

to determine what that looks like. So

12:56

as we think through what the future

12:58

is... I don't want to have to

13:00

know, like that acronym soup, why would

13:02

we put that on anybody willingly? The

13:04

only reason it exists is because the

13:07

process doesn't work. Just a short story

13:09

on how this works for us. So

13:11

Justin talked about, we build these requirements

13:13

and forever they've been super gold-plated, they're

13:15

this thick, they're every word known to

13:17

man because people are afraid that if

13:19

you don't get it in the requirement,

13:21

you can't move to this program record

13:23

record. Thousands of pages. What if we

13:25

put that on. because it is not

13:28

actually what we have to do and

13:30

I'll give you an example of that

13:32

for countering unmanned systems and there's lots

13:34

of startups in this and I know

13:36

that a lot of your founders are

13:38

not in the threat headspace that I'm

13:40

in right now so think about it

13:42

you're at an NFL stadium or a

13:44

soccer stadium. and you see a quadcopter,

13:46

like in my world, that's a horrifying

13:49

prospect because that means that it's intruded

13:51

into your airspace. You can't do anything

13:53

about it. It's really hard to bring

13:55

them down. Now think about thousands of

13:57

those, and that is the real threat

13:59

that we're under. Well, all of our

14:01

requirements writers, they're not operations, they're not

14:03

the guys getting shot at, they're just

14:05

trying to do the right thing, and

14:08

you write these sort of black box

14:10

requirements. I found one the other day

14:12

as I was going through, as I

14:14

was going through the other day as

14:16

I was going through, as I was

14:18

going through, as I was going through,

14:20

as I was going through, as I

14:22

was going through, If you took the

14:24

combination of statistics that they put in

14:26

there, like your operational availability, your operational

14:29

accuracy, your operational reliability, which you ended

14:31

up with with a system that only

14:33

had to work 51% of the time,

14:35

what person that only had to work

14:37

51% of the time, what person at

14:39

what company is going to go, yeah,

14:41

I'm okay with my engineers building something

14:43

that only works 51% of the time

14:45

or if you flip it, I'm okay

14:47

with wasting 49% of my resources. I

14:50

just want to blow up all of

14:52

that. Anything that's below law, we should

14:54

be able to reshape. No startup should

14:56

ever have to go, I want to

14:58

be a programmer record, because what that

15:00

really turns into is a 30-year-long set

15:02

where Microsoft might have been thinking, office

15:04

is going to be around forever. But

15:06

inside of them, they still let product

15:09

managers and program managers who were making

15:11

sure that Excel was really good. Please

15:13

don't stop making Excel good. The entire

15:15

world runs on Excel. But they're not

15:17

thinking, oh, if I just get it

15:19

here, I can slow down, I can

15:21

take my foot off the pedal, which

15:23

is what happens in programs of record.

15:25

As soon as you make it, as

15:27

soon as you've got that long-term budget,

15:30

there's a perverse incentive just to maintain

15:32

and create inertia. So my, I guess,

15:34

theory of the case is I'm going

15:36

to try to blow up all of

15:38

that. The question is like, how do

15:40

we increase yields, right? the government has

15:42

piloted, DOD has piloted at some point,

15:44

right? And so in this particular case,

15:46

on requirements, I worked with a 4,000

15:48

page requirements document. I'm fairly certain no

15:51

one read... Do you read it? No

15:53

one read all of it cover to

15:55

cover. You look at reference it like

15:57

the dictionary, but it's a little bit

15:59

longer. In this particular case, we said,

16:01

hey, we're going to use capability need

16:03

statements because this isn't going to get

16:05

what we want anyway. And so we

16:07

fast-tracked an 18-month process in three months

16:10

with a couple pages between C&S's capability

16:12

and need statements. and top level requirements.

16:14

The software acquisition pathway allows us to

16:16

do that. And so we already have

16:18

examples where the right thing is working

16:20

way better than the old thing. It's

16:22

a K-curve. This is a matter of

16:24

scaling what's working better. And the only

16:26

way we do that, because there is

16:28

some risk aversion, is. emphasize that the

16:31

outcomes and the tradeoffs are for our

16:33

warfighter. And so when we have the

16:35

best software developers in the world where

16:37

we have the best capability coming in,

16:39

let's not cut it down with a

16:41

thousand paper cuts, 10,000 paper cuts, let's

16:43

make decisions based on the impact that

16:45

they're making. We're doing more of that,

16:47

but it's pockets of excellence. That's huge.

16:49

So there's this acknowledgement about the black

16:52

box. and then a solution to managing

16:54

the black box, sounds like there's also

16:56

major changes happening when it comes to

16:58

capabilities and being able to say we

17:00

need to experiment because the pace of

17:02

technology is so fast. We can't write

17:04

a requirement for a capability that's going

17:06

to be developed three years from now

17:08

because of the pace of software development,

17:11

AI. all of these things, it's moving

17:13

so fast that you need to be

17:15

more agile. Especially with horizontal capabilities, like

17:17

artificial intelligence, like how would we pull

17:19

in artificial intelligence across 75 PES? Like

17:21

we need to name enterprise services. And

17:23

so defense innovation unit actually has a

17:25

digital on-ramp where they're trying to streamline

17:27

that so that we can make buys

17:29

based on impact. The chances that... a

17:32

bunch of horizontal organizations were going to

17:34

do the same thing, was always high.

17:36

A quick example is we had edge

17:38

compute ashore. We had a hyperconverged infrastructure,

17:40

small box that we could dive. invest

17:42

a billion dollar purchase and just move

17:44

to this in every way more capable

17:46

solution. And we said, what about that

17:48

for a float? And one of the

17:51

captains of a ship said, let's put

17:53

it on mine, let's figure out how

17:55

this works. We put it there. He

17:57

deployed to the Red Sea. It's on

17:59

other ships right now. This is enterprise

18:01

moving to Edge, specifically a float in

18:03

a case that we didn't need to

18:05

go down two different procurement paths. There

18:07

was another example, actually recently. where we

18:09

had an electronic warfare solution and we

18:12

heard that the army was going to

18:14

go out and do a request for

18:16

proposal on something that was really similar

18:18

with the same long requirements document that

18:20

we had. We said can you just

18:22

paint this brown and throw it on

18:24

a Humvee and that's exactly what they

18:26

did. They did it in three months

18:28

for 300K and then that went to

18:30

scaled fielding. So these things are happening.

18:33

It's just a matter of can you

18:35

make the best thing the normal thing.

18:37

And Justin said it's treating the value

18:39

as the actual goal. So everyone has

18:41

heard this PM's love off of this

18:43

concept of cost schedule and performance, right?

18:45

They want to make sure they hit

18:47

their timelines, at the cost that they

18:49

were given, and performance to me, is

18:52

like, hey, it meets my time lines,

18:54

at the cost that they were given,

18:56

and performance to me is like, hey,

18:58

it meets my warfighting needs, but performance

19:00

is actually what's written in this requirements

19:02

document, which they might not have actually

19:04

had any hand in doing. everyone has

19:06

permission to do the right thing. So

19:08

as you think about venture funding all

19:10

the way up through series, X, Y,

19:13

or Z, where a lot of the

19:15

companies that work with actually still live

19:17

in those spaces, they need to have

19:19

a front door, but what we're doing

19:21

in the Army is going, hey, if

19:23

you're ready, let's go to the field

19:25

right now. And I say this a

19:27

little bit appropriately. The Army is the

19:29

service that can do that because you

19:31

can move around the edges of some

19:34

of your ships. The Air Force can

19:36

move around the edges of some of

19:38

the aircraft. The capital assets, those are

19:40

fixed. What I have the luxury in

19:42

the Army is we can modify our

19:44

organizations, we can modify our kit, we

19:46

can tailor it to missions. So that's

19:48

what we've been doing with some of

19:50

our brigades that have been in Europe

19:53

and in the Pacific and frankly going

19:55

in and out of the Middle East

19:57

still. And this is an open invite

19:59

for all of the founders and all

20:01

of the companies. If you have things

20:03

that work, we have a front door.

20:05

We want to try to get you

20:07

in here. We want you to be

20:09

able to engage with Army Futures Command.

20:11

We want you to be able to

20:14

engage with our units directly because. Only

20:16

the soldiers will actually be able to

20:18

tell you, yes, this works at a

20:20

time of crisis. So I'm super excited

20:22

because I remember during the global war

20:24

on terror, we had quick reaction capabilities

20:26

and as a customer I could go,

20:28

I need this right now. And somebody

20:30

would go, cool, I will get you

20:32

that. In 30 days, I had kit

20:35

in Afghanistan that was working for me,

20:37

which told me it was legal, it

20:39

was moral, ethical, we could do it.

20:41

But we never actually went back and

20:43

fixed the system. We just moved off

20:45

to a side chain and everything was

20:47

a side quest. Now we're saying, hey,

20:49

we got to fix the main storyline.

20:51

Everybody's like, oh, we don't know how

20:54

to do that. This is a period

20:56

about scaling what actually works. And so

20:58

I'm excited about what Alex is doing,

21:00

that transforming in contact in execution is

21:02

a big deal because it's tech informed.

21:04

If you read Origins of Victory, they

21:06

say it's not about who necessarily just

21:08

has the best tech. It's how they're

21:10

adapting. to that tech and using that

21:12

as advantage. And so the way that

21:15

we're measuring outcomes is we're now using

21:17

outcome driven metrics to figure out do

21:19

you move the needle? Not is this

21:21

the best tech or how does this

21:23

fit into our current operating model? How

21:25

does this transform us? And so we

21:27

are more open and listening to and

21:29

working with and pulling through more of

21:31

those horizon three capabilities than we were,

21:33

but we need. out comes overmatch. We

21:36

need it to kick the ass of

21:38

whatever's there right now, so we can

21:40

show that difference. It's defensible, and we

21:42

can move off of the old and

21:44

onto the new. You've mentioned overmatch a

21:46

couple times. Can you double click into

21:48

that and share what exactly that means?

21:50

It's very hard to make a change

21:52

for a 15% improvement, unless it's something

21:55

that you're doing a lot. We want

21:57

to make the case that there are

21:59

orders of magnitude, right? Most recently, we

22:01

have a thousand different identity management solutions.

22:03

Just a thousand? So we've said naval

22:05

identity service is the enterprise system. That

22:07

service is what's moving forward. That opens

22:09

the door to repurposing some of that

22:11

money, repurposing some of that attention to

22:13

harder problems. And so to this point,

22:16

we are looking for AB testing that

22:18

shows that... A is no longer a

22:20

viable option and there's much more value

22:22

from B. If a company comes in

22:24

like Sironic comes in and they said

22:26

we have tests, we use rescale, we

22:28

use the cloud, and we show that

22:30

based on running this for 100,000 reps

22:32

that would have taken 10 years, we

22:34

can fast forward this and that is

22:37

proven, it was proven on paper, we

22:39

didn't have to put that into an

22:41

exercise, we can now with a lot

22:43

less risk reduction, but that is an

22:45

easier AB decision because they've translated it

22:47

into our language, which is outcomes. I'll

22:49

give an army example. So we have

22:51

this concept of mission command. I don't

22:53

want to go into like army terms

22:56

because it gives us this false sense

22:58

of security like we're doing something unique.

23:00

We had 17 programs a record in

23:02

this mission command enterprise and each one

23:04

did similar things in the fact that

23:06

each one of them had a map.

23:08

Humans like pixels were really good at

23:10

seeing pixels, they make us happy. So

23:12

we looked at maps. Well each one

23:14

of those came with its own mapping

23:17

server. and its own way to do

23:19

tiles, and it's a way to deliver

23:21

the maps, and it's a way to

23:23

store the maps. And about 2015-16, everybody

23:25

went, hey, this Google Earth thing is

23:27

pretty legit. We should probably just try

23:29

to use that. And it took five

23:31

years for people to go, oh, we

23:33

can just start sharing KMLs and KMCs

23:35

with each other and start sharing it.

23:38

Well, by then, you had really, really

23:40

good software companies that had said, cool.

23:42

We're going to give you these enterprise

23:44

services. I don't care what map you

23:46

use. I don't care who provides it.

23:48

You just tell it. You point us

23:50

to it and we'll serve it to

23:52

you. And all of us kind of

23:54

went, yeah, I want that. But we

23:57

couldn't because we had 17 programs around.

23:59

own with their own vestigial architecture and

24:01

we had to do a bunch of

24:03

architecture archaeology and we had a bunch

24:05

of people who were protecting their rice

24:07

bowls in terms of those programs and

24:09

it's only just now in 2025 are

24:11

we actually trying to hack and slash

24:13

and kill some of these systems to

24:15

go no if you want to look

24:18

at the map just look at the

24:20

map if you've got some weird unique

24:22

data like we do fires like we

24:24

do artillery like all those unique things

24:26

that are unique to us as a

24:28

military We'll figure out a way to

24:30

get those on the map, but we,

24:32

the Department of Defense, are certainly not

24:34

in 2025, who is building the best

24:36

maps, we're not building the best data

24:39

platforms, we're not building the best API

24:41

gateways. So let's not pretend it, let's

24:43

just get all those pieces together and

24:45

collapse all of these other vestigial organs.

24:47

I cannot describe how excited I am,

24:49

because we're doing it, we're doing it

24:51

live, and I love it. It's leverage

24:53

play, like the reuse piece. It's interesting

24:55

because I'd never thought about it from

24:58

this angle, but it's not unlike venture

25:00

capital where we're making bets on companies.

25:02

It's not about can this company move

25:04

the needle by 10x? Because to your

25:06

point, there's this switching cost. So we're

25:08

investing in 100x solutions. And then you

25:10

all need to be able to validate

25:12

that this new solution is going to

25:14

provide 100x outcome. And you also need

25:16

to be able to replace and sunset

25:19

capabilities as well. So it sounds like

25:21

our organizations are not that dissimilar. they're

25:23

not and what I appreciate is I've

25:25

learned what venture looks at and how

25:27

you grade and how you assess risk

25:29

is you do not expect every company

25:31

to be good at everything and that

25:33

is a failure that we as a

25:35

department of defense and really anybody who's

25:37

done federal acquisition falls into the trap

25:40

of where if you came up and

25:42

you won a program I expect you

25:44

to deliver every part of that top

25:46

to bottom even the stuff you're not

25:48

good at well what that does is

25:50

it creates an amalgamation of things that

25:52

people aren't good at and you can

25:54

actually multiply those tolerances across lots of

25:56

systems until you get something that looks

25:59

like, oh, there's a lot of not

26:01

good stuff here. Well, what we're trying

26:03

to do now is just figure out.

26:05

I'm a Lego person. Like how do

26:07

you find the right Lego pieces and

26:09

how do you put them together so

26:11

that if Justin Corp is really really

26:13

good at serving data? I get him

26:15

serving data. Layla Corp is really really

26:17

good at you're doing UI. I don't

26:20

make you do the things you're not

26:22

good at because frankly your engineers are

26:24

focused and excited on doing the things

26:26

you're good at and I can put

26:28

all these things together and that's what

26:30

we expect in our everyday lives. It

26:32

is shocking to me how much stuff

26:34

we buy for soldiers that no one

26:36

would accept if I had it to

26:39

you in your everyday life. And that

26:41

is a mindset that we are changing.

26:43

And I love the secretary is on

26:45

board, the chief is on board, all

26:47

of these senior leaders within the army

26:49

are just really going, hey common sense,

26:51

like what's the value and what's the

26:53

opportunity cost of making really dumb decisions

26:55

and how do we stop doing that?

26:57

Yeah. And one of the unlocks there

27:00

is, they're acknowledging, the Secretary of the

27:02

Secretary of the Use one component that

27:04

is higher performing than others. Should we

27:06

have a fully different artificial intelligence stack

27:08

for edge compute for sensor integration than

27:10

we do with no leverage as an

27:12

enterprise use case? No, of course not,

27:14

right? What does the governance look like

27:16

versus what does the compute look like?

27:18

We need. economies of scale. And so

27:21

we're working together more with our vendor

27:23

partners and even like signaling to some

27:25

of the investors to say, here is

27:27

our hard problem, but can we land

27:29

the plane on buying them? Can we

27:31

land the plane in sustainable cost? Whether

27:33

it's in a program executive office or

27:35

not, everyone should have to sing for

27:37

their dinner every year, but the idea

27:40

of bringing those breakthroughs in and then

27:42

again zambonying out the legacy capabilities so

27:44

we can unplug them for the first

27:46

time in a long time. These work

27:48

together within the valley of death like

27:50

this is actually happening within our budget

27:52

cycle if we do it well. It

27:54

sounds like you guys are really trying

27:56

to break down a lot of these

27:58

silos with the program offices. There are

28:01

75 of them, there are all these

28:03

different programs, and you all are thinking

28:05

about how can we find some of

28:07

these horizontal capabilities that are relevant, not

28:09

just within the Navy or the Army,

28:11

but across the services, to be able

28:13

to find economies of scale for your

28:15

organizations, and that also make it easier

28:17

for startups to partner with you all

28:19

and scale with you. Am I hearing

28:22

that right? Yes, and one of the things

28:24

that I've been really big on is we need

28:26

the enterprise to do enterprise things.

28:28

and get us out of the business of doing

28:30

things that are common to all. In the

28:32

intelligence community, we had this concept of

28:34

services of common concern and what that

28:36

means is it is something that everyone

28:39

should be able to use, therefore don't

28:41

let every individual agency or organization do

28:43

it themselves. It's sort of a common

28:45

sense on the face of it thing.

28:47

However, in the combat side of the department,

28:50

we've had a really hard time adopting. I'll

28:52

give a couple examples. So the CDA. and

28:54

D-I-U and some of these other organizations have

28:56

said, hey, we're going to pull capabilities up

28:58

and deliver it in a common way for

29:00

everyone. That's great for things that are common.

29:02

So that when Justin and I go back

29:04

to our offices, if we need to share

29:06

data, we can do it in a common

29:08

way. For things that are unique to me,

29:10

or unique to the Air Force, or unique to

29:13

the Space Force, or the Coast Guard and Marines,

29:15

whoever, if it's really, really, really, really unique, like

29:17

unique, like can only be done by them and

29:19

should only be done by them. That's where we

29:21

want to focus on because that's the niche. But

29:24

I will keep hampering on this. There's option

29:26

for startups and small companies in all

29:28

of that because Palantir was an inkytel

29:30

investment. Like we found them as a

29:33

very small company and now it's driving

29:35

a lot of the way we do warfare.

29:37

So even those big companies will go, I'm

29:39

not good at this, but that company, that

29:41

really small company, they're really small company, I

29:44

want them on my team. We want to

29:46

unlock. all of that and we want to

29:48

enable it and I'm trying not to be

29:50

super super technical just because this is not

29:52

a technology problem. This is a culture and

29:55

a process problem that we can apply technology

29:57

to. I had a chance last week to

29:59

sit with all of the bigLLM providers.

30:01

And I loved the conversation because they

30:03

were ruthlessly honest with me and with

30:06

each other in that, hey, all of

30:08

these things are available, whether it's llama

30:10

or claw, or claw, or claw, or

30:12

chat, GPT, or even deep sync, they're

30:15

all available. What makes them unique and

30:17

good and competitive is the unique data

30:19

sets that their customers have and the

30:22

unique ways that they're going to use

30:24

them. That's us. That's my job to

30:26

be able to be able to go.

30:29

Startups can't solve a fundamental deficiency in

30:31

us not being able to describe our

30:33

problems, but they can help us identify

30:35

where those problems are. This is why

30:38

it's really important to be close to

30:40

the warfighter. If you are a startup

30:42

CEO... Don't hire sales people before you

30:45

understand the ecosystem. Move to San Diego.

30:47

Spend nine months near the group that

30:49

you are trying to positively enable and

30:51

empower to do their job 10 or

30:54

100 times better than they are right

30:56

now. So that proximity piece is real,

30:58

especially for unique or specific problems. And

31:01

then more often than not, there is

31:03

some aspect that should be evaluated as

31:05

dual use. From a shared services perspective,

31:07

Alex already said it. We're doing more

31:10

common. and similar than dissimilar from industry.

31:12

In some cases we're not making the

31:14

tough decisions to scale what's working fast

31:17

enough. And so if we have more

31:19

shared services at the department level, how

31:21

can we do autonomy together? as services?

31:24

How can we make sure that that

31:26

is well integrated from the ground up

31:28

and we're not like looking at this

31:30

as an integration problem because we made

31:33

separate decisions three years later? From an

31:35

artificial intelligence perspective, right now we're leveraging

31:37

an Air Force and Army and Navy

31:40

solution. Can we single up for common

31:42

enterprise problems in one place and then

31:44

other places? Can we have narrow models

31:46

where we share and all of them

31:49

feed into some of the more unique

31:51

problems like command and control? And so

31:53

we're starting to. architect in that way

31:56

and make decisions based on what's available

31:58

not based on what we've done. Now

32:00

one kind of an older comment you

32:03

made was we have some things in

32:05

common with VC. Sure, from a funnel

32:07

perspective that would be the ideal state.

32:09

One of the tricky parts for us

32:12

because we are such a big organization

32:14

is finding people who think that way.

32:16

and fanning the flames and empowering them.

32:19

And so every organization has latent risk

32:21

takers even if they're conditioned to not

32:23

take risk. I mean, the acquisition community

32:25

teaches us that lowering risk, reducing risk,

32:28

is the goal. So essentially like playing

32:30

not to lose. It doesn't really keep

32:32

pace in many, many cases with the

32:35

current trajectory of technological advancement. And so

32:37

we have unleashed people, Mavericks, who are

32:39

willing to work against their self-interest to

32:42

bring breakthrough capabilities through. I know Justin

32:44

and I could sit here and talk

32:46

really deep tech all day long and

32:48

we can go into the nuts and

32:51

bolts. And we will. There's two things,

32:53

the silos and they just put something

32:55

in my brain. I want everyone who

32:58

listens or see this to understand we

33:00

haven't missed the signal. The department hasn't

33:02

missed the signal. In 1968, something fundamentally

33:04

changed the way that we do buying

33:07

in the government and it has not

33:09

stopped changing since then. In 1968, there

33:11

was a program put directly in the

33:14

United States. budget and it was the

33:16

Minuteman 3 program. That was the first

33:18

time that a program by name was

33:20

put into the budget that we had

33:23

to spend money against. Before that, there

33:25

were just these big pots of money

33:27

like the army had vehicles and ammunition

33:30

and soldiers and All of the money

33:32

was just, hey, do what you need

33:34

to do to run the army. A

33:37

portfolio? A portfolio. A portfolio. Yeah, absolutely.

33:39

And then in 1968, pam, Minuteman 3

33:41

is in there, and I get it

33:43

from a national security perspective. You want

33:46

to make sure that money goes into

33:48

a long-term project and you want to

33:50

make sure that it's successful because we

33:53

were the only people who were going

33:55

to do that. Now, today, the army

33:57

is only... 21% of the budget. We

33:59

are everywhere. We, the Army, are everywhere.

34:02

And again, this is going to sound

34:04

a little parochial, but we're on the

34:06

border. We're the Global Response Force, we're

34:09

the immediate response force, we're the Homeland

34:11

Defense Brigade, we're in Haiti, we're in

34:13

Africa, we're in CENTCOM, we're in PAC.

34:16

We are everywhere. 21% of the budget,

34:18

and less than 50% of that, do

34:20

we have any flexibility to spend? to

34:22

change how we're spending. Even less than

34:25

that, there are thousands of budget line

34:27

items that are directed to us in

34:29

the National Defense Authorization Act, everywhere that

34:32

tell us exactly how we're going to

34:34

spend that money. So as we're talking

34:36

about earlier, that three-year spend cycle, what

34:38

that really means for us is we

34:41

have to predict the future three to

34:43

five years out, hope for the best,

34:45

and then work in the year of

34:48

execution to move money around the edges

34:50

to try to do advanced technology. Otherwise,

34:52

what happens is... We give money to

34:54

our labs and they try to do

34:57

the right thing and they try to

34:59

pull advanced technology in and they try

35:01

to like land This transition this magical

35:04

transition where it goes from something that

35:06

works in a lab to something that

35:08

works in the field in one go

35:11

And we know that that's not real

35:13

because technology evolves so fast where you

35:15

could go from something that is a

35:17

concept to someone writing it up and

35:20

doing it in software space in days

35:22

to weeks in our entire system, the

35:24

entire way we think is predict the

35:27

future three to five years out, land

35:29

the perfect shot, get a trick shot,

35:31

and go. And it just doesn't work.

35:33

It does not work. This is all

35:36

the more reason if someone does have

35:38

a capability that is so much better

35:40

than what we have in the field

35:43

right now, like we have stocked up

35:45

the pipeline, but the data and the

35:47

stories associated with that, we want it

35:50

to be. undeniably good so that we

35:52

can then make the case. People come

35:54

to me and they say, isn't Zero

35:56

Touch AI better than how you're doing

35:59

unified endpoints right now? Yes, of course

36:01

it is. And we are trying to

36:03

make the case that this will save

36:06

us money, we need a little bit

36:08

of capital. so that we can get

36:10

the op-ex right by the total cost

36:12

of ownership and the impact of this

36:15

from a mission perspective is best for

36:17

those who are protecting us right? Here

36:19

are ways that we can lower the

36:22

tail on ineffective things but we have

36:24

to aim at the exact solution that

36:26

we have even if it's a manual

36:29

process. And if we can find several

36:31

things, we can find enough room until

36:33

all the workarounds are turned into scaled

36:35

solutions that Alex is talking about, we

36:38

can still pull that through. But if

36:40

you can make the A to B

36:42

not close, blow it away, that's what

36:45

we want to see. And so we

36:47

use world-class alignment metrics to show how

36:49

a divestment is possible through a game-changing

36:51

investment. And that's a really big change.

36:54

We want that to be... 90% of

36:56

the portfolio, you know, like, is that

36:58

really a big percent of the, like,

37:01

they're, each is, right? When we still

37:03

talk about, the IU is, like, less

37:05

than 1% of the budget, like these

37:07

innovation activities are still very small. So

37:10

it is a good signal that we're

37:12

doing them. The impact of scaling, what's

37:14

working. once in a lifetime opportunity and

37:17

imagine how much more secure that our

37:19

country and the world would be if

37:21

we were making the exact ruthless highest

37:24

impact decisions across the board. So I

37:26

want to double-click into the culture piece.

37:28

Is that doorlocked? To protect us. The

37:30

world I'm in, venture capital, we're in

37:33

the business of risk. Like we're all

37:35

in on risk and that's how we

37:37

think about the world. Obviously with the

37:40

Pentagon. that's different. You all are making

37:42

a lot of strides to change that,

37:44

but I'd love your perspective on, like,

37:46

how do we get closer together between

37:49

Silicon Valley and Washington and get that

37:51

risk appetite up a little bit? We're

37:53

assessing risk on different things, and it's

37:56

at a necessity. I'm using big we,

37:58

there's a lot of folks in the

38:00

department who actually understand venture whether or

38:03

not they came from that, or they've

38:05

just been working with the venture ecosystem

38:07

for a long time. facing example. Hinketail's

38:09

ratio is 25 outside dollars for every

38:12

federal dollar spent and then dual use

38:14

takes care of the rest of that.

38:16

And so this is one example where

38:19

we've proven this model. What we have

38:21

to work through and Justin said it

38:23

really elegantly earlier, the people who take

38:25

risk, their job, and I don't want

38:28

to disparage anybody because they are hardworking

38:30

Americans trying to do the right thing,

38:32

their job is to minimize the risk

38:35

to the government. And the way that

38:37

they measure risk is the potential for

38:39

waste, fraud, or abuse of dollars. So

38:41

what they're trying to say is, hey,

38:44

if I give you a dollar, do

38:46

I get a dollar of value out

38:48

of it? That's a good metric. What

38:51

they don't do is if I give

38:53

you a dollar and I take 10

38:55

years to spend that dollar, is it

38:58

still worth a dollar? So what we

39:00

are trying to do is figure out,

39:02

hey, I'd rather lose a dollar right

39:04

now. then spend that 10 years because

39:07

that is also super valuable time that

39:09

we could be trying to figure out

39:11

what all of the players in the

39:14

startup landscape and all of the players

39:16

in the venture landscape and frankly all

39:18

the because once you sell your first

39:20

product you're not to start up your

39:23

business and all the businesses that are

39:25

available how do they get us the

39:27

best war winning capabilities right now? Maybe

39:30

it sounds cheesy. But I mean it

39:32

the risk calculus we're not comparing the

39:34

same things and what would help is

39:37

that those venture folks talk to our

39:39

leadership continuing to engage because we're gonna

39:41

work up they need to work laterally

39:43

to go hey the way your measure

39:46

risk is fundamentally incoherent with where technology's

39:48

going. If a quarterback was measured on

39:50

not making mistakes versus throwing touchdowns or

39:53

winning games the performances would be a

39:55

hundred percent different right and so if

39:57

your goal is to. hedge against anything

39:59

happening than you're fighting against inflation as

40:02

opposed to against disruptive technology. We used

40:04

to have something called cost as an

40:06

independent variable. We make anything an acronym,

40:09

so cave. And so this was to

40:11

say, like, how do you back in?

40:13

and make sure that you're spending the

40:16

budget and then optimizing for that. If

40:18

we had speed as an independent variable,

40:20

save, and we measured based on outcomes,

40:22

then I think that looks pretty different.

40:25

That looks more like executing. And so

40:27

when we say, hey, unleashed folks, what

40:29

do you have? They're not bringing 20%

40:32

better capability. They're saying, hey, we're doing...

40:34

cross domain solution right now through a

40:36

bunch of different program offices. The army's

40:38

doing that as a service. Can we

40:41

jump on cross domain even if it'll

40:43

affect or make some people that created

40:45

something in-house upset? Yes. For the sailor

40:48

and for the Marine, yes. I just

40:50

want people to understand who's listening, how

40:52

ridiculous this is because cross domain as

40:54

a service is an API for binary

40:57

XML between two networks. That's all we're

40:59

talking about. Now there's national security implications,

41:01

but... That is a solved problem and

41:04

that we are fighting over. We want

41:06

to keep solved problems solved, get higher

41:08

up the stack in general. I think

41:11

it's academically interesting to resolve the same

41:13

problem. That's for school. And so we

41:15

have marine innovation units that is a

41:17

way to keep really talented Marines in

41:20

the reserves who have venture jobs as

41:22

their full-time job. We're using them to

41:24

scout technologies. We have folks in Europe

41:27

right now working with some of the

41:29

startup founders to say, hey, here is

41:31

a breakthrough. What would scaling this look

41:33

like? Well, it would mess up our

41:36

program of record plan. That's not what

41:38

we're optimizing for. We have contracting officers

41:40

sometimes say. Well, I want to do

41:43

fewer contracts. Well, I want to win.

41:45

So then there are really good contracting

41:47

officers who get that. And short of

41:50

changing the way that we fuel incentives,

41:52

measuring how much better, 25 times better,

41:54

the get after it folks are in

41:56

different organizations, that K-curve fuels us, and

41:59

it actually gives us more leverage to

42:01

do more of this. And so ultimately,

42:03

send your success stories. if you are

42:06

doing this really well, and you're already

42:08

in and you've made it, because we

42:10

want to double down on this flywheel

42:12

of this can happen. We don't want

42:15

two successes a month. We want to

42:17

overwhelm the system with how much better

42:19

it can be. If you're not ready,

42:22

when the window of opportunity opens, sometimes

42:24

you're waiting a really long time. And

42:26

so we put cape. mobile virtual network

42:28

operator company start up in Guam and

42:31

then we needed them. So like hey

42:33

the pilot was ready for actual use.

42:35

If we are out hands-on testing with

42:38

the best stuff then we have an

42:40

opportunity to scale in a significant way

42:42

much shorter than any of these timelines

42:45

you hear about whenever you read about

42:47

the government. Very optimistic and this is

42:49

really exciting. It feels like we're in

42:51

a lightning in the bottle moment where

42:54

there's a lot about to happen. I'd

42:56

love to just get your quick take

42:58

on. the future the year ahead what

43:01

you both are most excited about? I'm

43:03

excited about the idea of we don't

43:05

have to prove that these things work

43:07

anymore so if we scale half of

43:10

what's working well right now I mean

43:12

the software acquisition pathways we have gotten

43:14

more permission like significant permission to do

43:17

the right thing now how many people

43:19

do it and what does that look

43:21

like and what wins do we get

43:24

out of that and so ultimately I

43:26

am excited about participating in the race

43:28

to have the best technology solutions and

43:30

the companies that are offering the best

43:33

breakthroughs to be fielded into that operations

43:35

and maintenance pot and getting that to

43:37

the production to our war fighters hands

43:40

sooner. So I'm excited about the best

43:42

chance I've ever seen in the next

43:44

six to 12 months to get capability

43:46

that wasn't under government contract into full-scale

43:49

production. I'm actually really excited about a

43:51

lot of things. This year, we are

43:53

going to put the consortium for the

43:56

Army's next generation command and control into

43:58

its program phase. So right now it's

44:00

been in software pathways. Army Futures Command

44:02

has been piloting this. We have a

44:05

consortium and I keep using that term

44:07

deliberately of just great companies who are

44:09

going here's what technology can bear and

44:12

I was at the National Training Center

44:14

out in the middle of the Mojave

44:16

Desert last week and I crawled into

44:19

an M1A2 set B3 tank the coolest

44:21

you know eight-year-old Alex was very excited

44:23

because I was in a tank commander

44:25

but that tank commander he was running

44:28

the same socks. We have taken a

44:30

20-year-old tablet that's mounted in that Abrams

44:32

that's really hard to move, and we've

44:34

given him everything on his tack device,

44:36

which is an Android app on an

44:39

Android phone, and now he can command

44:41

and control his tank company. So we're

44:43

going to take that next generation, command

44:45

and control, and to get that next

44:47

generation, command and control, and to get

44:49

that on contract and start moving forward

44:51

and get that fielded to one of

44:53

our divisions this year. Ukraine and what's

44:56

happening over there and we're actually going

44:58

to with fifth core and with UCRF

45:00

we're actually going to work on how

45:02

we scale up adding non-conventional sensors

45:04

non-conventional compute how do we

45:07

do automation so that if you were driving

45:09

and it doesn't matter what part of the

45:11

world you're in if you're an American warfighter

45:13

and there is a drone threat there is a

45:15

bunch of automated decision mechanisms that are

45:17

just going hey I sent something in

45:19

the electromagnetic spectrum here's what you need

45:21

to do and you don't have to

45:24

worry about it In the IED fight, they taught us something when you're going

45:26

to combat right into this, they said, down, out, up. You have to look down

45:28

to make sure you're not stepping in IED. You have to look out to make

45:30

sure you're not going to walk into an IED. You have to look up to

45:32

make sure that you're not getting shot at it. Now, people have to look and

45:34

make sure that you're not getting shot at it. You have to look up to

45:36

make sure that you're not getting shot getting shot getting shot getting shot getting shot

45:38

at. You're not getting shot getting shot getting shot getting shot getting shot getting shot

45:40

getting shot at getting shot getting shot getting shot getting shot getting shot getting

45:42

shot getting shot getting shot getting shot getting shot getting shot

45:44

getting shot getting shot getting shot getting shot getting shot getting

45:46

shot getting shot getting shot getting shot getting shot getting shot

45:48

getting shot getting shot getting shot getting shot getting shot getting

45:50

shot getting shot getting shot getting shot getting shot getting shot

45:52

getting shot getting shot getting shot getting shot getting shot getting

45:54

shot getting shot getting shot getting The US Army should never

45:57

trade blood for blood in first contract and it should always

45:59

be blood for an iron. and it should always be

46:01

their blood. And we have to make sure

46:03

that we can actually get the autonomy pieces

46:05

together. And it's not one company. It's not

46:08

Miller Corp. That's going to solve autonomy. It's

46:10

how do we get the right tool chain

46:12

in place from Sim all the way through

46:14

V&V, and then how do we get the

46:17

right actions and the right libraries, and then

46:19

how do we get the right actions and

46:21

the right libraries, and how do we get

46:23

ourselves off of this sort of archaic

46:25

blockers that have stopped us? Like I said,

46:28

it's a with our shields or on it

46:30

type of phase. We are going to

46:32

body block our way through this to

46:34

do the right thing because we

46:36

can't not do it. Laura. I

46:38

wanted to ask you guys a

46:40

couple lightning questions. I could talk

46:43

to you guys all day. Okay.

46:45

Very quickly. Red flag, green flag.

46:47

What do you love startups

46:49

doing? What do they need to stop

46:52

doing when they come talk to you?

46:54

Red flag. Red flag. I would

46:56

say at this. and commercial solution offerings,

46:58

don't bang your head against the

47:00

wall, find top level requirement or something

47:03

easier to work with and people who

47:05

get outcomes. Green flag, keep building the

47:07

product that you started on the path to

47:09

build, have your North Star and then work

47:11

with us to figure out how to do,

47:13

you know, roof shots as you're on your

47:16

way to your Moonshot rather than trying

47:18

to tailor it to us. The DOD, we

47:20

are going to be a better customer, I

47:22

wholeheartedly believe that, but if you had a

47:24

vision. Like go after your vision and then

47:26

get in the hands of our war fighters

47:28

to help tailor it. Any red flags?

47:30

Stop trying to use the Department

47:33

of Defense as your only means of

47:35

revenue. And I mean that because if

47:37

you get complacent with the Department of

47:39

Defense, then you are no longer actually

47:41

on the quest that you were on.

47:43

And it's just not good. We want

47:45

to make sure that we have the

47:47

best war winning dual use technology. And

47:50

if we become the only provider to

47:52

your technology, it's neither. the goal of

47:54

partnering with industry. You are just a

47:56

part of us at that point. And keep

47:58

imagining a safer world. We want the

48:00

brightest minds to help every aspect

48:02

of this. The goal here is

48:04

to just be so effective that

48:07

we are getting as much return

48:09

in investment as possible and that

48:11

translates to protecting those who are

48:14

protecting all of us every day.

48:16

Technology is the advantage and if

48:18

we have the best of America's

48:20

talent working on these types of

48:23

problems, our hardest problems like

48:25

we've fixed the alignment problem.

48:28

What book should every

48:30

defense-based founder read? 100-year

48:32

marathon by Michael Pillsbury.

48:34

What is the number one

48:37

misconception about your service that you

48:39

would want to dispel? This is

48:41

not your granddaddy's DOD anymore.

48:43

We will be more flexible.

48:45

We will be more open-minded.

48:47

And if you bring data... to

48:49

us, then we will work our butts

48:52

off to make sure that if the

48:54

policy is the problem, then that policy

48:56

will get accepted. We've gotten 12 exceptions

48:59

to policy because we brought data and

49:01

said this is hurting us. And so

49:03

these sub-optimizations that we've always had, we

49:05

can do better than that, and we

49:07

will. I'm proud of the Army. I

49:09

make a joke all the time. The

49:12

Navy and the Army are the only

49:14

two services in the Constitution. We've been

49:16

doing this in 1775, and everyone knows

49:18

where the Army brings. Unemotionally, when the

49:20

world dials 911, the phone rings at

49:22

Fort Bragg. It doesn't ring at Naval Station,

49:24

San Diego, it doesn't ring at Anderson Air

49:26

Force Base, and I love my Navy and

49:29

Air Force brothers. The phone rings at Fort

49:31

Bragg, North Carolina, when the world dials 911.

49:33

So the biggest mis-conception that I would

49:35

tell people is, we haven't missed this.

49:37

We're in the fight still, we're in

49:39

the fight globally, and we're going to

49:42

transform in contact and continue to transform

49:44

and contact, because we've been doing this

49:46

in 1775. And we are teammates and

49:48

we are working together on these problems

49:50

and we're going to start using that

49:52

buying power to make more effective purchases

49:54

because when the president calls for the

49:56

carriers before they call brag that we

49:59

need to be fully integrated and

50:01

teamed. We fight joint every day.

50:03

Love that. And then last one,

50:05

fun one, favorite military movie. Top

50:07

Gun Maverick. That happened on CVN

50:09

72, which has one of those hyperconverged

50:11

infrastructure stacks that we talked about. It

50:13

has to. So we're moving out. Is

50:15

that how they did the post-processing? I

50:18

am a big fan of World War

50:20

II. I used to watch it with

50:22

my dad. This is going to sound corny.

50:24

I love Starship Troopers. I know it's a

50:26

bad adaptation of the book. It is the

50:29

mobile infantry. It is just people on the

50:31

ground just getting it done. So in the

50:33

spirit of doing things a little differently, I'll

50:35

go to Starship Troopers. Thank you both so

50:37

much for your time. So great to know

50:40

that there are folks like you working at

50:42

the highest levels of government to help

50:44

make some of these changes. All right,

50:46

that is all for today. If you

50:48

did make it this far, first of

50:50

all. We put a lot of thought

50:52

into each of these episodes, whether it's

50:54

guests, the calendar, Tetris, the cycles with

50:56

our amazing editor Tommy, until the music

50:58

is just right. So if you'd like

51:01

what we've put together, consider dropping as

51:03

a line at ratethis podcast.com/A16Z. And let

51:05

us know what your favorite episode is.

51:07

It'll make my day, and I'm sure

51:09

Tommy's too. We'll catch you on the

51:11

flip side.