1
00:00:00,000 --> 00:00:03,285
Sergey Litvinenko: Koop is an
insurance platform for robots

2
00:00:03,285 --> 00:00:06,682
and autonomous vehicles. Robot
insurance. Sounds a little bit

3
00:00:06,682 --> 00:00:09,855
bizarre to some people. I mean,
robots are supposed to be

4
00:00:09,855 --> 00:00:13,140
perfect. They're not going to
make any mistakes. That's not

5
00:00:13,140 --> 00:00:16,537
true. This is going to be the
first time in the human history

6
00:00:16,537 --> 00:00:20,045
where robots will become like a
daily occurrence to millions of

7
00:00:20,045 --> 00:00:23,163
people, I think autonomous
vehicle ownership is going to

8
00:00:23,163 --> 00:00:25,000
compete with like home
ownership.

9
00:00:25,000 --> 00:00:29,815
Audrow Nash: You said robots are
becoming an asset class. Tell me

10
00:00:29,815 --> 00:00:31,000
more about that.

11
00:00:31,000 --> 00:00:35,000
Sergey Litvinenko: Two parts to
the answer.

12
00:00:35,000 --> 00:00:39,000
Audrow Nash: Alright, hi Sergey,
would you introduce yourself?

13
00:00:39,000 --> 00:00:42,298
Sergey Litvinenko: Hey, Audrow,
this is Sergey from Koop here,

14
00:00:42,298 --> 00:00:44,000
joining you for the second time.

15
00:00:44,000 --> 00:00:47,764
Audrow Nash: Yeah, hell yeah.
Excited to talk again, it's been

16
00:00:47,764 --> 00:00:51,529
quite some time. Tell me about
Koop, just for those who aren't

17
00:00:51,529 --> 00:00:55,232
familiar, what do you guys do?
And then I'd love some company

18
00:00:55,232 --> 00:00:58,814
details too, like funding and
how large you guys are, where

19
00:00:58,814 --> 00:01:01,000
you're based, these kinds of
things.

20
00:01:01,000 --> 00:01:05,106
Sergey Litvinenko: For sure.
Koop is an insurance platform

21
00:01:05,106 --> 00:01:09,211
for robots and autonomous
vehicles. We work with all kinds

22
00:01:09,211 --> 00:01:13,812
of use cases from robot arms and
traditional industrial robots to

23
00:01:13,812 --> 00:01:18,059
mobile robots that you can see
on the streets, to drones, to

24
00:01:18,059 --> 00:01:22,306
self-driving cars, and all kinds
of automated equipment like

25
00:01:22,306 --> 00:01:26,341
agriculture, bulldozers,
construction, all of that stuff.

26
00:01:26,341 --> 00:01:30,588
It's a really exciting journey
because we get to see lots of

27
00:01:30,588 --> 00:01:34,906
stuff that's happening within
robotics and actually put money

28
00:01:34,906 --> 00:01:38,516
to it. So we act as an
underwriter and we price the

29
00:01:38,516 --> 00:01:43,046
insurance that we underwrite. So
we see lots of things happening

30
00:01:43,046 --> 00:01:47,506
behind the scenes. We've been at
it since 2021 and based in New

31
00:01:47,506 --> 00:01:51,824
York City, a team of 10 people
keeping it very efficient. And

32
00:01:51,824 --> 00:01:56,000
we raised two rounds of funding,
pre-seed and a seed. Thank

33
00:01:56,000 --> 00:01:56,319
Audrow Nash: I think from I'd
like to get into kind of like

34
00:01:56,319 --> 00:01:56,661
how it all works. I'm impressed
you guys are such a small team.

35
00:01:56,661 --> 00:01:56,991
because like you have a report
that you came out with. You're

36
00:01:56,991 --> 00:01:57,332
insuring a lot of the robotics
industry. but I guess one of the

37
00:01:57,332 --> 00:01:57,679
things that strikes me with what
you're saying now, I don't know

38
00:01:57,679 --> 00:01:57,971
if you had as much of an
autonomous vehicles focus. Is

39
00:01:57,971 --> 00:01:58,301
that new? Or Has that been kind
of one of the things you guys

40
00:01:58,301 --> 00:01:58,616
have been doing all along, or
are you like associated with

41
00:01:58,616 --> 00:01:58,935
Tesla insurance or anything
else? I see your shirt with the

42
00:01:58,935 --> 00:01:59,000
Tesla on it.

43
00:01:59,000 --> 00:02:14,000
Sergey Litvinenko: You. Yeah, Ha
yeah. we are not associated with

44
00:02:14,000 --> 00:02:26,923
Tesla, but it's not stopping us
from insuring autonomous

45
00:02:26,923 --> 00:02:41,692
vehicles. I can talk more about
Cybercabs in particular a little

46
00:02:41,692 --> 00:02:44,000
bit later.

47
00:02:44,000 --> 00:02:45,000
Audrow Nash: Cool.

48
00:02:45,000 --> 00:02:49,221
Sergey Litvinenko: When we
started the company, we focused

49
00:02:49,221 --> 00:02:53,733
on robotics in general. We look
at robots as an emerging asset

50
00:02:53,733 --> 00:02:57,881
class, as an emerging risk
class, and we wanted to be use

51
00:02:57,881 --> 00:03:01,957
case and vehicle type agnostic.
And we worked with a few

52
00:03:01,957 --> 00:03:06,396
autonomous vehicle companies. We
still work with them. And it

53
00:03:06,396 --> 00:03:10,836
happened to be that the growth
in non-public growth use cases

54
00:03:10,836 --> 00:03:12,000
was much higher,

55
00:03:12,000 --> 00:03:13,000
Audrow Nash: Oh, yeah.

56
00:03:13,000 --> 00:03:16,382
Sergey Litvinenko: Partly
because of liability, because

57
00:03:16,382 --> 00:03:20,195
when you are on a public road
and you get into an accident, it

58
00:03:20,195 --> 00:03:23,884
could be a very challenging
situation. And when you are in a

59
00:03:23,884 --> 00:03:27,574
constrained environment, it
becomes a lot simpler. So if you

60
00:03:27,574 --> 00:03:31,448
look at our customer growth, the
non-public road customers were

61
00:03:31,448 --> 00:03:34,953
just much higher growth than
public road customers. But I

62
00:03:34,953 --> 00:03:38,581
think that's going to change in
the near future. And that's

63
00:03:38,581 --> 00:03:42,148
where Cybercabs come into the
picture. Not only Cybercabs,

64
00:03:42,148 --> 00:03:45,899
there are also Waymos and other
players that are coming up in

65
00:03:45,899 --> 00:03:49,589
the US. But I think autonomous
vehicles are about to finally

66
00:03:49,589 --> 00:03:53,155
have their moment and hoping
Koop will see exciting growth

67
00:03:53,155 --> 00:03:55,000
coming from that area as well.

68
00:03:55,000 --> 00:03:58,461
Audrow Nash: Cool. Very
exciting. Yeah, I want to talk

69
00:03:58,461 --> 00:04:02,498
about that quite a bit. So, give
me a high-level perspective of

70
00:04:02,498 --> 00:04:06,215
how it works. What I remember
from last time is you have a

71
00:04:06,215 --> 00:04:10,060
good way of logging behavior of
the robots, and then you use

72
00:04:10,060 --> 00:04:14,097
that to assess risk and failure
modes and things like that, and

73
00:04:14,097 --> 00:04:17,942
you use that to underwrite it to
figure out how much risk is

74
00:04:17,942 --> 00:04:21,851
involved for any robot so that
you can make, like, insurance,

75
00:04:21,851 --> 00:04:25,824
figure out how much you have to
charge them for insurance. So,

76
00:04:25,824 --> 00:04:29,605
it's a good bet for you, and
they feel covered in this kind

77
00:04:29,605 --> 00:04:33,706
of thing. But tell me about the
process more, like, I guess, how

78
00:04:33,706 --> 00:04:37,423
does it all work, and how do you
guys service the robotics

79
00:04:37,423 --> 00:04:38,000
industry?

80
00:04:38,000 --> 00:04:39,888
Sergey Litvinenko: Of course.
Two parts to answer that

81
00:04:39,888 --> 00:04:42,021
question. First, why do you even
need this whole thing in the

82
00:04:42,021 --> 00:04:44,119
first place? Robot insurance.
Sounds a little bit bizarre to

83
00:04:44,119 --> 00:04:46,322
some people. I mean, robots are
supposed to be perfect. They're

84
00:04:46,322 --> 00:04:48,000
not going to make any mistakes.
That's not true.

85
00:04:48,000 --> 00:04:57,000
Audrow Nash: Love it. Yeah,
yeah. Mm-hmm.

86
00:04:57,000 --> 00:05:00,427
Sergey Litvinenko: You buy robot
insurance for two reasons.

87
00:05:00,427 --> 00:05:03,913
Number one, when you have a
deployment, liability is written

88
00:05:03,913 --> 00:05:07,631
into the deployment contract and
insurance is usually a big part

89
00:05:07,631 --> 00:05:10,942
of it. So your customer wants
you to buy insurance. So if

90
00:05:10,942 --> 00:05:14,427
something goes wrong, there is
somebody paying for that. And

91
00:05:14,427 --> 00:05:17,797
they don't want to be the ones
paying for that. That's the

92
00:05:17,797 --> 00:05:21,224
first thing. And the second
thing, because you just want to

93
00:05:21,224 --> 00:05:24,710
protect yourself, you as the
robotics developer operator, in

94
00:05:24,710 --> 00:05:27,963
case, let's say you have
thousands of robots and you get

95
00:05:27,963 --> 00:05:31,622
into an accident and there is a
lawsuit, insurance protects you

96
00:05:31,622 --> 00:05:34,817
from liability. So it's
contractually imposed. And also

97
00:05:34,817 --> 00:05:38,303
it's like management or board
mandate that you have to have.

98
00:05:38,303 --> 00:05:41,672
Those are the two reasons why
people buy it. So how do you

99
00:05:41,672 --> 00:05:44,577
actually price the robot
insurance? That's a great

100
00:05:44,577 --> 00:05:47,946
question. I'll tell you that at
the end of the day, when a

101
00:05:47,946 --> 00:05:51,199
robotics company buys an
insurance product from Koop, it

102
00:05:51,199 --> 00:05:54,743
looks like a policy that has
many different pages. And that's

103
00:05:54,743 --> 00:05:58,228
what essentially you're paying
for. And the way we construct

104
00:05:58,228 --> 00:06:01,830
that policy and rate that policy
is unique to our underwriting

105
00:06:01,830 --> 00:06:05,315
and pricing process. So when it
comes to underwriting is all

106
00:06:05,315 --> 00:06:08,801
about saying, hey, if robotics
company comes to me, should I

107
00:06:08,801 --> 00:06:11,996
back it? Should I back it with
my insurance product? So

108
00:06:11,996 --> 00:06:15,714
something goes wrong, I'm going
to pay for that. And we'll do it

109
00:06:15,714 --> 00:06:19,315
algorithmically, meaning that we
have a model where we collect

110
00:06:19,315 --> 00:06:22,917
certain inputs from the company,
and then it spits out whether

111
00:06:22,917 --> 00:06:26,054
this risk is acceptable or not
acceptable with certain

112
00:06:26,054 --> 00:06:29,365
conditions. There is also some
human judgment involved in

113
00:06:29,365 --> 00:06:33,025
situations where maybe the model
is lacking some data. It could

114
00:06:33,025 --> 00:06:36,452
be some exotic use case like
autonomous nuclear submarines.

115
00:06:36,452 --> 00:06:39,822
Not a lot of data about that. So
you have to involve human

116
00:06:39,822 --> 00:06:43,481
underwriters. And then once you
pass the underwriting, you have

117
00:06:43,481 --> 00:06:46,909
to do pricing. And pricing
actually tells you, OK, how much

118
00:06:46,909 --> 00:06:49,000
should we charge this company?
Yeah.

119
00:06:49,000 --> 00:06:53,134
Audrow Nash: And just
explicitly, what is the output

120
00:06:53,134 --> 00:06:57,904
of underwriting? Like, I'm not
that familiar with insurance,

121
00:06:57,904 --> 00:07:02,515
and I suspect some people are
not. And I know this is your

122
00:07:02,515 --> 00:07:06,808
day-to-day walk, but what is
underwriting? What is the

123
00:07:06,808 --> 00:07:08,000
function of it?

124
00:07:08,000 --> 00:07:08,417
Sergey Litvinenko: It's very
simple. Underwriting is all

125
00:07:08,417 --> 00:07:08,872
about evaluating the risk and
saying whether we want to cover

126
00:07:08,872 --> 00:07:09,326
that risk or not. You actually
encounter underwriting all the

127
00:07:09,326 --> 00:07:09,773
time. Let's say you want to
mortgage. You have to go through

128
00:07:09,773 --> 00:07:10,160
underwriting. You open a credit
card. You go through

129
00:07:10,160 --> 00:07:10,592
underwriting, get a car loan.
You go through underwriting.

130
00:07:10,592 --> 00:07:10,957
Similar thing. You have to
provide Koop with some

131
00:07:10,957 --> 00:07:11,374
information so we can make a
decision whether we want to

132
00:07:11,374 --> 00:07:11,851
cover you or not. And that's the
whole underwriting process. The

133
00:07:11,851 --> 00:07:12,000
end result of that--

134
00:07:12,000 --> 00:07:18,843
Audrow Nash: Mm-hmm. For
everything. Yeah. Mm-hmm. And

135
00:07:18,843 --> 00:07:26,953
it's basically, it's gonna be a
ratio of how much you can charge

136
00:07:26,953 --> 00:07:34,810
and how much you expect to pay
out, this kind of thing. Is it?

137
00:07:34,810 --> 00:07:40,766
Because you're saying an
incident could be this

138
00:07:40,766 --> 00:07:48,749
expensive, but it's this likely,
and you have to make sure that

139
00:07:48,749 --> 00:07:56,986
it makes financial sense for you
all to be involved with. Is that

140
00:07:56,986 --> 00:07:58,000
correct?

141
00:07:58,000 --> 00:07:59,656
Sergey Litvinenko: That's more
in pricing what you're

142
00:07:59,656 --> 00:08:00,000
describing.

143
00:08:00,000 --> 00:08:02,000
Audrow Nash: Ah, okay.

144
00:08:02,000 --> 00:08:05,688
Sergey Litvinenko: You are
talking about the actuarial

145
00:08:05,688 --> 00:08:09,307
model that says, for the
customer that we are able to

146
00:08:09,307 --> 00:08:13,200
cover within our rule set, how
much we should charge them

147
00:08:13,200 --> 00:08:16,000
versus how much insurance
they're buying.

148
00:08:16,000 --> 00:08:18,274
Audrow Nash: Yeah, it's not
pricing yet. We're not there

149
00:08:18,274 --> 00:08:20,914
yet. But you're seeing if it's a
good bet effectively for you all

150
00:08:20,914 --> 00:08:23,025
to be involved with, or, like,
what is the output of

151
00:08:23,025 --> 00:08:24,000
underwriting, I suppose.

152
00:08:24,000 --> 00:08:25,105
Sergey Litvinenko: Yeah. In our
use case-- sorry, in the way we

153
00:08:25,105 --> 00:08:26,105
build our model, it pretty much
spits out whether this is

154
00:08:26,105 --> 00:08:27,228
acceptable, whether this risk is
declined, or it's accepted with

155
00:08:27,228 --> 00:08:28,193
certain conditions. So it's
literally-- that's what the

156
00:08:28,193 --> 00:08:29,281
platform spits out. And the way
the platform evaluates this is

157
00:08:29,281 --> 00:08:30,000
primarily-- it's both on
quantitative and

158
00:08:30,000 --> 00:08:31,035
non-quantitative data. So we
could look at some data coming

159
00:08:31,035 --> 00:08:32,158
from the API, maybe coming from
the documents that they provide,

160
00:08:32,158 --> 00:08:33,246
the customers I mean, but would
also look at the use case. For

161
00:08:33,246 --> 00:08:34,298
example, if we're working with
drones, we would also look at

162
00:08:34,298 --> 00:08:35,404
how the drone segment performs
and maybe what particular things

163
00:08:35,404 --> 00:08:36,000
they're hauling or what
particular

164
00:08:36,000 --> 00:09:12,000
Audrow Nash: Mm-hmm.

165
00:09:12,000 --> 00:09:16,006
Sergey Litvinenko: Missions
they're on. And there are rules

166
00:09:16,006 --> 00:09:20,148
built in that, hey, if you are
doing, let's say, anything too

167
00:09:20,148 --> 00:09:24,426
dangerous, we would decline this
for certain reasons. I'll give

168
00:09:24,426 --> 00:09:28,704
an example like areas where we
have either very high conditions

169
00:09:28,704 --> 00:09:32,370
or like nearly automated
declinations would be medical

170
00:09:32,370 --> 00:09:34,000
robots, surgical robots.

171
00:09:34,000 --> 00:09:37,000
Audrow Nash: Oh, interesting.

172
00:09:37,000 --> 00:09:39,447
Sergey Litvinenko: Yeah, the
reason being there is that it's

173
00:09:39,447 --> 00:09:41,893
not even about the robot itself,
it's more about the medical

174
00:09:41,893 --> 00:09:44,299
practice, is that it's highly
regulated. And as a result, a

175
00:09:44,299 --> 00:09:46,950
mistake can cost a lot of money,
including lots of litigation. So

176
00:09:46,950 --> 00:09:49,315
unless the robot has like tons
of data that is like nearly

177
00:09:49,315 --> 00:09:51,476
perfect, which none of the
robotics companies that we

178
00:09:51,476 --> 00:09:54,045
encountered in that space have,
we decided to pass just because

179
00:09:54,045 --> 00:09:56,411
if a robot makes a mistake and
everybody gets sued. So our

180
00:09:56,411 --> 00:09:58,857
underwriting is built around
this knowledge of how liability

181
00:09:58,857 --> 00:10:01,263
works within different pockets
of robotics deployments. And

182
00:10:01,263 --> 00:10:03,791
then it spits out a decision and
say, hey, this is acceptable,

183
00:10:03,791 --> 00:10:06,116
this is not acceptable or
acceptable with conditions. And

184
00:10:06,116 --> 00:10:08,359
then there is pricing. And
pricing says, okay, how much

185
00:10:08,359 --> 00:10:10,724
should we charge for this risk
based on the inputs that we

186
00:10:10,724 --> 00:10:13,171
have? And that's where it's all
quantitative. And we look at

187
00:10:13,171 --> 00:10:15,495
robotics risks through the
standardized model. That means

188
00:10:15,495 --> 00:10:17,819
that we can compare one risk
against the other, or we can

189
00:10:17,819 --> 00:10:20,429
develop the baselines, which is
what the In Bots We Trust report

190
00:10:20,429 --> 00:10:21,000
was all about.

191
00:10:21,000 --> 00:10:51,000
Audrow Nash: Yeah. Mm-hmm.

192
00:10:51,000 --> 00:10:54,420
Sergey Litvinenko: And then we
say, okay, we're going to look

193
00:10:54,420 --> 00:10:57,839
at the risk performance, which
primarily comes from the robot

194
00:10:57,839 --> 00:11:01,146
telemetry, we're going to look
at the environment that they

195
00:11:01,146 --> 00:11:04,454
operate in. So we know how close
they are to other people's

196
00:11:04,454 --> 00:11:07,705
property or other humans on a
job. And then we're going to

197
00:11:07,705 --> 00:11:11,293
look at your safety and security
certifications. And those three

198
00:11:11,293 --> 00:11:14,488
things get encoded into the
final rate that our customers

199
00:11:14,488 --> 00:11:17,796
pay. And that rate could be
either at the company level, it

200
00:11:17,796 --> 00:11:20,711
could be at the deployment
level, it could be at the

201
00:11:20,711 --> 00:11:22,000
individual robot level.

202
00:11:22,000 --> 00:11:24,000
Audrow Nash: So cool.

203
00:11:24,000 --> 00:11:25,000
Sergey Litvinenko: So for
example, you come to me,

204
00:11:24,000 --> 00:11:26,783
Audrow Nash: That's so cool you
can do it at the robot level. So

205
00:11:26,783 --> 00:11:27,000
cool.

206
00:11:27,000 --> 00:11:27,502
Sergey Litvinenko: Yeah, I'll
give you an example. Well,

207
00:11:27,502 --> 00:11:28,076
humanoids are a thing right now.
Everybody's talking about them,

208
00:11:28,076 --> 00:11:28,561
hoping to see lots of
deployments next year. Today, we

209
00:11:28,561 --> 00:11:29,099
can, a humanoid company can come
to Koop, and we're going to

210
00:11:29,099 --> 00:11:29,655
issue a humanoid coverage on a
per humanoid basis. We're going

211
00:11:29,655 --> 00:11:30,193
to list every single humanoid
like it's a fleet of cars, and

212
00:11:30,193 --> 00:11:30,749
you'll be able to track the rate
for each, even if it's across

213
00:11:30,749 --> 00:11:31,000
different customer sites. So

214
00:11:31,000 --> 00:11:56,000
Audrow Nash: Mm-hmm. It's
amazing.

215
00:11:56,000 --> 00:11:59,573
Sergey Litvinenko: It becomes it
becomes very, it's just very

216
00:11:59,573 --> 00:12:03,088
cool to open your insurance
policy and see a fleet of robots

217
00:12:03,088 --> 00:12:06,544
in there. That's where that's
where we had it. And we just,

218
00:12:06,544 --> 00:12:10,000
you know, very glad to be at the
forefront of it, you know.

219
00:12:10,000 --> 00:12:13,766
Audrow Nash: Yeah. Tell me
about, so is there any, like

220
00:12:13,766 --> 00:12:17,875
it's a, I don't know, like we
have a Tesla and the Tesla has

221
00:12:17,875 --> 00:12:22,052
Tesla insurance and one of the
things that I think is so cool

222
00:12:22,052 --> 00:12:26,435
is it adjusts our premium month
to month based on how well we've

223
00:12:26,435 --> 00:12:30,407
been driving by its estimate. Is
there anything that makes

224
00:12:30,407 --> 00:12:34,516
estimating on a like robot by
robot level, like instead of a

225
00:12:34,516 --> 00:12:38,693
fleet level or a company level
or whatever, is there anything

226
00:12:38,693 --> 00:12:42,665
that makes it harder? Like, is
that more of a challenge? I

227
00:12:42,665 --> 00:12:46,499
imagine like a lot of the
traditional, like a lot of the

228
00:12:46,499 --> 00:12:50,882
other automotive companies, they
do it by the year and they look

229
00:12:50,882 --> 00:12:54,580
at a lot of like demographic
info on you and your past

230
00:12:54,580 --> 00:12:58,209
driving record and things like
that. But it's just so

231
00:12:58,209 --> 00:13:01,907
interesting to have very fine
resolution to individual

232
00:13:01,907 --> 00:13:05,537
entities to insure. Like, I
guess it's more of a data

233
00:13:05,537 --> 00:13:09,303
problem and if you solve the
data problem then it's not

234
00:13:09,303 --> 00:13:13,343
intrinsically more difficult
than anything else. Or tell me

235
00:13:13,343 --> 00:13:17,589
about going to the like single
entity level and the challenges

236
00:13:17,589 --> 00:13:18,000
there.

237
00:13:18,000 --> 00:13:22,476
Sergey Litvinenko: Yeah, yeah,
it's called unit of exposure.

238
00:13:22,476 --> 00:13:27,250
When you drive your Tesla, each
mile that you drive is a unit of

239
00:13:27,250 --> 00:13:31,876
exposure. So theoretically, each
mile that you go, an accident

240
00:13:31,876 --> 00:13:36,352
can happen, right? in So if you
are operating, if you have a

241
00:13:36,352 --> 00:13:40,753
wheeled use case, let's say
self-driving cars, mileage is a

242
00:13:40,753 --> 00:13:45,528
perfect unit of exposure. That's
why we do this and other people

243
00:13:45,528 --> 00:13:49,780
do that too. In robots, it
becomes a little bit different

244
00:13:49,780 --> 00:13:53,808
because, sure, mileage is one
thing, but if you deploy

245
00:13:53,808 --> 00:13:58,359
humanoid, let's say, humanoid,
like the their mileage is kind

246
00:13:58,359 --> 00:14:00,000
of useless, right? But

247
00:14:00,000 --> 00:14:05,000
Audrow Nash: Mileage is just
kind of irrelevant. Mm-hmm.

248
00:14:05,000 --> 00:14:06,432
Sergey Litvinenko: What's
important for humanoids, it

249
00:14:06,432 --> 00:14:08,054
would be how many hours they're
in service and how many jobs

250
00:14:08,054 --> 00:14:09,000
they perform, yeah, exactly. So
for

251
00:14:09,000 --> 00:14:14,000
Audrow Nash: Yeah, operating
time. Mm-hmm.

252
00:14:14,000 --> 00:14:17,351
Sergey Litvinenko: Lots of those
non-public road use cases, we

253
00:14:17,351 --> 00:14:18,000
track hours.

254
00:14:18,000 --> 00:14:19,000
Audrow Nash: Cool.

255
00:14:18,000 --> 00:14:21,692
Sergey Litvinenko: And then
based on those hours, you are

256
00:14:21,692 --> 00:14:25,385
able to estimate how much you
charge for the risk. And it

257
00:14:25,385 --> 00:14:29,336
really depends, I'll tell you
that we have two models. We can

258
00:14:29,336 --> 00:14:33,028
do hourly model, we can do
non-hourly model, you can kind

259
00:14:33,028 --> 00:14:34,000
of convert one.

260
00:14:34,000 --> 00:14:36,452
Audrow Nash: What are the
models? I'm sorry, I missed it.

261
00:14:36,452 --> 00:14:38,000
You can do what model or what
model?

262
00:14:38,000 --> 00:14:38,625
Sergey Litvinenko: I mean, when
we calculate the rate for robot

263
00:14:38,625 --> 00:14:39,250
deployments, we can do an hourly
model where you, as a robotics

264
00:14:39,250 --> 00:14:39,864
company, you come and tell us,
hey, we're gonna have this many

265
00:14:39,864 --> 00:14:40,469
robots estimated to perform this
many jobs, running this many

266
00:14:40,469 --> 00:14:41,055
hours, and we can use that to
develop a rate, like per hour

267
00:14:41,055 --> 00:14:41,630
rate, and then the more hours
you perform, the more you're

268
00:14:41,630 --> 00:14:42,225
gonna pay. Alternatively, we can
do it in a prepaid fashion,

269
00:14:42,225 --> 00:14:42,790
like, okay, here is a rough
estimate, but we're not gonna

270
00:14:42,790 --> 00:14:43,375
charge you on a, let's say,
monthly basis, you just prepaid

271
00:14:43,375 --> 00:14:44,000
for the whole thing and off you
go. But we are in a world where

272
00:14:44,000 --> 00:15:12,540
Audrow Nash: Oh, hourly, yeah,
yeah. Mm-hmm. That's really

273
00:15:12,540 --> 00:15:15,000
cool.

274
00:15:15,000 --> 00:15:17,130
Sergey Litvinenko: AI is
replacing physical work.

275
00:15:17,130 --> 00:15:19,565
Physical work is measured
through the number of jobs you

276
00:15:19,565 --> 00:15:21,783
complete, the output you
produce, but for insurance

277
00:15:21,783 --> 00:15:24,261
purposes, usually hourly based.
And that's what we use as

278
00:15:24,261 --> 00:15:25,000
exposure of unit.

279
00:15:25,000 --> 00:15:28,230
Audrow Nash: Mm-hmm. Very cool.
I like moving to such a granular

280
00:15:28,230 --> 00:15:31,309
perspective because it's like,
it almost approaches, it makes

281
00:15:31,309 --> 00:15:34,388
physical work approach like API
usage kind of thing where you

282
00:15:34,388 --> 00:15:37,467
pay actually by the, yeah, yeah,
yeah. But it's just, it's so

283
00:15:37,467 --> 00:15:40,394
interesting because it's so
efficient because you pay what

284
00:15:40,394 --> 00:15:41,000
you're using

285
00:15:41,000 --> 00:15:48,000
Sergey Litvinenko: That's a good
comparison.

286
00:15:48,000 --> 00:15:52,221
Audrow Nash: And it, and also if
you are getting paid per hour

287
00:15:52,221 --> 00:15:56,374
with your robots it's very nice
if your insurance costs scale

288
00:15:56,374 --> 00:16:00,732
linearly with your usage. Like,
that's very efficient, not a lot

289
00:16:00,732 --> 00:16:04,000
of waste, this kind of thing.
Super, super cool.

290
00:16:04,000 --> 00:16:07,786
Sergey Litvinenko: Quick comment
here. We also look into the

291
00:16:07,786 --> 00:16:11,888
digital AI world because the way
those models perform for digital

292
00:16:11,888 --> 00:16:15,612
tasks is very interesting. In
the physical world, it's much

293
00:16:15,612 --> 00:16:17,000
more complicated. But,

294
00:16:17,000 --> 00:16:21,046
Audrow Nash: Do you mean like,
Grok or other, do you mean like,

295
00:16:21,046 --> 00:16:24,000
AI agents for doing stuff or
what do you mean?

296
00:16:24,000 --> 00:16:24,199
Sergey Litvinenko: Yeah,
exactly. So AI agents, but let's

297
00:16:24,199 --> 00:16:24,420
say so people measure AI agents
through tokens. How much tokens

298
00:16:24,420 --> 00:16:24,640
they consume and what kind of,
how many tokens they produce and

299
00:16:24,640 --> 00:16:24,853
how much it costs. It's probably
gonna be like the comparison

300
00:16:24,853 --> 00:16:25,000
here is that, okay, in the
physical world,

301
00:16:25,000 --> 00:16:44,000
Audrow Nash: Oh, that's so cool.
Mm-hmm. Yep.

302
00:16:44,000 --> 00:16:47,143
Sergey Litvinenko: People will
track hours and useful output

303
00:16:47,143 --> 00:16:50,233
per hour. So it's gonna be
similar to the digital world and

304
00:16:50,233 --> 00:16:53,062
I think it's just exciting to
think that there is some

305
00:16:53,062 --> 00:16:55,000
commonality between those two
things.

306
00:16:55,000 --> 00:16:59,000
Audrow Nash: Convergence. Mm.

307
00:16:59,000 --> 00:17:02,548
Sergey Litvinenko: Still a long
way to go, but maybe AI

308
00:17:02,548 --> 00:17:06,484
eventually is just gonna be not
digital AI, physical AI, it's

309
00:17:06,484 --> 00:17:07,000
just AI.

310
00:17:07,000 --> 00:17:12,567
Audrow Nash: Hahaha That'd be so
cool. Yeah, what a thing that'll

311
00:17:12,567 --> 00:17:17,706
be. Okay, hell yeah. So then, so
we underwrite it, yes or no

312
00:17:17,706 --> 00:17:23,273
decision, do we move on? Then we
price it, what's a fair cost for

313
00:17:23,273 --> 00:17:28,583
the risk of covering? And then
how do you get data from robots

314
00:17:28,583 --> 00:17:33,637
to understand their behavior?
How do you deal with upgrades

315
00:17:33,637 --> 00:17:38,690
and changing behavior for the
robots? How does all this how

316
00:17:38,690 --> 00:17:43,486
does all this fit in? Like,
where, how does this work, I

317
00:17:43,486 --> 00:17:44,000
guess?

318
00:17:44,000 --> 00:17:47,455
Sergey Litvinenko: The most
valuable data comes from the

319
00:17:47,455 --> 00:17:51,096
API. So we can hook up to, it
really depends on a customer,

320
00:17:51,096 --> 00:17:54,675
but usually customers have a
place where all the data goes

321
00:17:54,675 --> 00:17:58,377
to. And they can let us access
part of the data that we tell

322
00:17:58,377 --> 00:18:02,203
them over an API. So whenever
let's say they have a deployment

323
00:18:02,203 --> 00:18:05,782
and then they send the data to
the cloud, whenever they do

324
00:18:05,782 --> 00:18:09,484
that, part of the data hits us.
And then we can see how on a

325
00:18:09,484 --> 00:18:13,372
robot by robot basis, what kind
of usage they're doing and what

326
00:18:13,372 --> 00:18:17,197
kind of behavior they're doing,
okay? And within the behavior,

327
00:18:17,197 --> 00:18:20,838
we're really looking at things
that are could indicate some

328
00:18:20,838 --> 00:18:24,355
problematic behavior. It could
be safe stops, it could be

329
00:18:24,355 --> 00:18:27,996
unusual behavior that would
derive based on telemetry where

330
00:18:27,996 --> 00:18:30,772
the vehicle accelerated,
decelerated or robot

331
00:18:30,772 --> 00:18:34,413
accelerated, decelerated or
turned improperly. And when you

332
00:18:34,413 --> 00:18:38,362
look at a lot of data, there are
some of those corner cases that

333
00:18:38,362 --> 00:18:42,311
you can catch. And that data is
very useful because it gives you

334
00:18:42,311 --> 00:18:45,952
lots of precision in terms of
what's happening. And then as

335
00:18:45,952 --> 00:18:49,531
you gain enough knowledge, you
can see how it shapes up by

336
00:18:49,531 --> 00:18:52,677
different use cases. For
example, autonomous mobile

337
00:18:52,677 --> 00:18:56,195
robots, AMRs deployed within
industrial facilities is our

338
00:18:56,195 --> 00:18:59,835
most popular use case. And there
you can see how those AMRs

339
00:18:59,835 --> 00:19:02,797
behave themselves across
different sites, across

340
00:19:02,797 --> 00:19:06,438
different deployment types and
scale as well. So that's the

341
00:19:06,438 --> 00:19:09,955
first thing. And then that thing
can be supplemented with

342
00:19:09,955 --> 00:19:13,781
anything that customers might
send us. They might send us some

343
00:19:13,781 --> 00:19:17,236
additional files with data,
maybe that was not available

344
00:19:17,236 --> 00:19:21,000
over API, they just throw it to
us and we take it from there.

345
00:19:21,000 --> 00:19:33,000
Audrow Nash: Awesome. Huh.

346
00:19:33,000 --> 00:19:36,758
Sergey Litvinenko: The second
part is the environment. We have

347
00:19:36,758 --> 00:19:40,636
a certain way how we encode the
environment is that it primarily

348
00:19:40,636 --> 00:19:44,455
comes through a pretty detailed
survey that we receive from the

349
00:19:44,455 --> 00:19:47,909
customer. A customer gives us
lots of details about where

350
00:19:47,909 --> 00:19:51,364
they're gonna deploy the robot.
And then all of that gets

351
00:19:51,364 --> 00:19:54,939
translated into the certain
parameters and variables within

352
00:19:54,939 --> 00:19:58,333
that equation that we built. And
then the final thing is

353
00:19:58,333 --> 00:20:02,212
certifications. It's safety and
security certifications. We love

354
00:20:02,212 --> 00:20:04,818
seeing robot safety
certifications, be that

355
00:20:04,818 --> 00:20:08,152
functional safety, be that
mobile robot safety, which I

356
00:20:08,152 --> 00:20:11,848
think ISO has something coming
up, also on the humanoid side.

357
00:20:11,848 --> 00:20:15,303
And also company-wide safety
standards, especially around

358
00:20:15,303 --> 00:20:18,879
cyber. So when you combine all
of that, it's like one giant

359
00:20:18,879 --> 00:20:22,333
equation. It spits out the rate
that we should charge the

360
00:20:22,333 --> 00:20:26,030
company and then as deployments
change or the nature changes,

361
00:20:26,030 --> 00:20:29,000
let's say the robot data
changes, the environment

362
00:20:29,000 --> 00:20:32,697
changes, or the certification
changes, that's what drives the

363
00:20:32,697 --> 00:20:33,000
rate.

364
00:20:33,000 --> 00:20:37,994
Audrow Nash: Mm-hmm. Hmm. Very
cool. I wonder. I feel like we

365
00:20:37,994 --> 00:20:42,825
have so many interesting things
to talk about with AI, with

366
00:20:42,825 --> 00:20:47,819
cyber. What a crazy thing. But
with this approach, if you see

367
00:20:47,819 --> 00:20:52,158
the behavior of the robots
changing, more errors in a

368
00:20:52,158 --> 00:20:57,316
certain way after some update,
you can identify that the errors

369
00:20:57,316 --> 00:21:01,737
have happened and maybe
something's different and then

370
00:21:01,737 --> 00:21:06,404
revisit your risk and then
change the rate on them. Like,

371
00:21:06,404 --> 00:21:11,561
one thing, a while ago with the
Tesla, it was like when updates

372
00:21:11,561 --> 00:21:16,801
would roll out, for a few days,
the autonomous driving wasn't as

373
00:21:16,801 --> 00:21:22,041
good. Seems to, we don't notice
it now, but it was funny because

374
00:21:22,041 --> 00:21:27,117
it was like, "What the hell did
it do?" And it'd be like, "Oh,

375
00:21:27,117 --> 00:21:32,111
we just recently updated." And I
guess you could probably see

376
00:21:32,111 --> 00:21:36,450
that with other robots too. So
yeah, you can look for

377
00:21:36,450 --> 00:21:41,608
regressions in behavior because
you have the telemetry data and

378
00:21:41,608 --> 00:21:43,000
things like this.

379
00:21:43,000 --> 00:21:44,000
Sergey Litvinenko: That is
correct.

380
00:21:44,000 --> 00:21:45,000
Audrow Nash: Very interesting.

381
00:21:45,000 --> 00:21:47,963
Sergey Litvinenko: Also one
thing, not only tracking the

382
00:21:47,963 --> 00:21:50,979
software releases and then
performance after the software

383
00:21:50,979 --> 00:21:54,259
release, but sometimes customers
run two types of deployments.

384
00:21:54,259 --> 00:21:57,328
There's a test or R&D deployment
and there is a production

385
00:21:57,328 --> 00:22:00,767
deployment. And then in the test
deployment, you see all kinds of

386
00:22:00,767 --> 00:22:03,783
crazy stuff because people put
this thing to use to test,

387
00:22:03,783 --> 00:22:05,000
right? And we actually,

388
00:22:05,000 --> 00:22:09,000
Audrow Nash: Yeah. So you can
insure those differently

389
00:22:09,000 --> 00:22:09,240
Sergey Litvinenko: Yeah, and
actually we do not, I mean,

390
00:22:09,240 --> 00:22:09,498
people do want to insure their
R&D deployments, but they are

391
00:22:09,498 --> 00:22:09,738
pretty cheap compared to
production deployments. Because

392
00:22:09,738 --> 00:22:10,000
the worst thing that can happen
in the R&D deployment is that

393
00:22:09,000 --> 00:22:23,000
Audrow Nash: With this. So cool.

394
00:22:23,000 --> 00:22:26,401
Sergey Litvinenko: There's gonna
be very limited outside

395
00:22:26,401 --> 00:22:30,227
liability to the company, right?
And you probably do that under

396
00:22:30,227 --> 00:22:33,506
specific constraints, maybe in
specific places. But in

397
00:22:33,506 --> 00:22:37,150
production, it's real humans,
it's third parties, and that's

398
00:22:37,150 --> 00:22:41,036
what drives the rate. So we are
also able to separate pricing by

399
00:22:41,036 --> 00:22:43,951
deployment, test versus
production, which people

400
00:22:43,951 --> 00:22:47,413
appreciate because it makes
sense to them that, well, why

401
00:22:47,413 --> 00:22:50,692
should I pay bad rate for
something that's supposed to

402
00:22:50,692 --> 00:22:53,000
behave badly because I put it to
test?

403
00:22:53,000 --> 00:22:55,097
Audrow Nash: Yeah. That makes
sense. Man, I love the

404
00:22:55,097 --> 00:22:57,556
granularity that you're able to
do. That's so incredible. One

405
00:22:57,556 --> 00:23:00,016
thing that really strikes me,
especially with you all being a

406
00:23:00,016 --> 00:23:02,435
team of 10 people, the variation
that you must see and like,

407
00:23:02,435 --> 00:23:04,815
"Okay, this one company that
you're insuring, you get their

408
00:23:04,815 --> 00:23:06,992
API data this one way, this
other company you get it a

409
00:23:06,992 --> 00:23:09,331
different way, then they just
dump some data on you that's

410
00:23:09,331 --> 00:23:11,589
useful for things, all the
different certifications, all

411
00:23:11,589 --> 00:23:14,048
the different just... It's so
high mix in some sense. Tell me

412
00:23:14,048 --> 00:23:16,629
a bit about that because that is
a very interesting thing that I

413
00:23:16,629 --> 00:23:18,000
assume is a big challenge.
Mm-hmm.

414
00:23:18,000 --> 00:23:29,148
Sergey Litvinenko: Mm-hm. Well,
in order, so just to put it into

415
00:23:29,148 --> 00:23:38,729
the perspective, over the past
few years, I think we've

416
00:23:38,729 --> 00:23:45,000
touched, by touched, I mean,
insured

417
00:23:45,000 --> 00:23:46,000
Audrow Nash: Insured.

418
00:23:46,000 --> 00:23:50,694
Sergey Litvinenko: And pay
claims for 20 to 25% of the US

419
00:23:50,694 --> 00:23:53,000
robotics companies by count.

420
00:23:53,000 --> 00:23:54,000
Audrow Nash: It's incredible.
Thank you.

421
00:23:54,000 --> 00:23:56,741
Sergey Litvinenko: So fourth of
the market. And that's what gave

422
00:23:56,741 --> 00:23:59,353
us the ability to create this
report. And I can tell you that

423
00:23:59,353 --> 00:24:01,880
when we started Koop, lots of
things were done by hand. But

424
00:24:01,880 --> 00:24:04,192
now everything, almost
everything is automated. And by

425
00:24:04,192 --> 00:24:06,633
automated, I mean the data
intake, the data analysis, the

426
00:24:06,633 --> 00:24:08,988
underwriting part, and the
pricing part. The only thing

427
00:24:08,988 --> 00:24:11,729
that requires human input would
be the claims, because depending

428
00:24:11,729 --> 00:24:14,341
on who's involved, you might
have to deal with like five, six

429
00:24:14,341 --> 00:24:16,611
different parties at a time. And
that requires manual

430
00:24:16,611 --> 00:24:19,095
communication over email and
phone. But everything else is

431
00:24:19,095 --> 00:24:21,793
automated. Now, if let's say we
give an API to the company, and

432
00:24:21,793 --> 00:24:24,405
they're unable to hook it up,
and they just dump the data, we

433
00:24:24,405 --> 00:24:26,760
develop the data analysis
pipeline, where it just takes

434
00:24:26,760 --> 00:24:29,458
whatever files we have and puts
them into the right format that

435
00:24:29,458 --> 00:24:31,000
we need for the model to take it
in.

436
00:24:31,000 --> 00:24:54,000
Audrow Nash: Mm-hmm. Hmm.
Buckets. Mm-hmm.

437
00:24:54,000 --> 00:24:58,972
Sergey Litvinenko: So that's why
each customer that comes into

438
00:24:58,972 --> 00:25:03,943
Koop could be underwritten and
priced nearly instantly. And if

439
00:25:03,943 --> 00:25:08,995
10 customers come to Koop after
this interview, I'll be able to

440
00:25:08,995 --> 00:25:11,000
do them in 10 minutes. So

441
00:25:11,000 --> 00:25:12,000
Audrow Nash: That's amazing.

442
00:25:12,000 --> 00:25:12,406
Sergey Litvinenko: Just one by
one. We do have human in a loop,

443
00:25:12,406 --> 00:25:12,794
so we review what the machine
speeds out. But we do not have

444
00:25:12,794 --> 00:25:13,000
to actually do analysis by hand.

445
00:25:13,000 --> 00:25:22,280
Audrow Nash: Hell yeah. Yep.
Huh. Yes. How much efficiency

446
00:25:22,280 --> 00:25:29,000
gain have you gotten from all of
the LLMs,

447
00:25:29,000 --> 00:25:30,000
Sergey Litvinenko: The why
efficiency.

448
00:25:29,000 --> 00:25:33,261
Audrow Nash: Different AI models
and whatnot? Like, do you guys

449
00:25:33,261 --> 00:25:37,319
use Claude Code for everything?
Or like, do you have... Just

450
00:25:37,319 --> 00:25:41,377
tell me about how AI has changed
stuff for you and if you're

451
00:25:41,377 --> 00:25:43,000
using it and everything.

452
00:25:43,000 --> 00:25:46,925
Sergey Litvinenko: It has been
magical. I'm pretty sure our

453
00:25:46,925 --> 00:25:51,049
experience is similar to yours
and the rest of the industry is

454
00:25:51,049 --> 00:25:54,974
that the velocity, it's like,
you know, vibe coding has bad

455
00:25:54,974 --> 00:25:59,032
connotation to it. But I think
when you do it correctly, it's

456
00:25:59,032 --> 00:26:02,891
like the secret weapon that you
wish you only had. We both

457
00:26:02,891 --> 00:26:06,882
increase the velocity of the
product development, especially

458
00:26:06,882 --> 00:26:10,873
with some recent changes that we
made and the quality of the

459
00:26:10,873 --> 00:26:14,000
product. By product, I mean the
platform. Okay.

460
00:26:14,000 --> 00:26:17,000
Audrow Nash: Nice. Mm-hmm.

461
00:26:17,000 --> 00:26:20,307
Sergey Litvinenko: The insurance
product itself, like the huge

462
00:26:20,307 --> 00:26:23,720
piece of paper, also has gotten
much more dynamic because now we

463
00:26:23,720 --> 00:26:26,653
can almost algorithmically
change parts of the coverage

464
00:26:26,653 --> 00:26:29,000
depending on what the customers
are seeking.

465
00:26:29,000 --> 00:26:31,000
Audrow Nash: So cool. Mm-hmm.

466
00:26:31,000 --> 00:26:34,004
Sergey Litvinenko: And same
applies to the rate. If a

467
00:26:34,004 --> 00:26:37,404
customer comes and we need to
adjust the rate outside of our

468
00:26:37,404 --> 00:26:41,031
pre-existing boundaries, it can
be done with a couple of clicks.

469
00:26:41,031 --> 00:26:44,488
So on the back end, so whatever
the customers are not seeing,

470
00:26:44,488 --> 00:26:48,059
it's been dramatic improvement.
And now that we're able to spit

471
00:26:48,059 --> 00:26:51,572
out the insurance products and
quotes really fast, people like

472
00:26:51,572 --> 00:26:54,973
that. They're like, oh, I would
rather work with Koop to get

473
00:26:54,973 --> 00:26:58,203
this whole thing done in a
couple of hours than work with

474
00:26:58,203 --> 00:27:01,603
another carrier that's gonna get
it done in weeks, sometimes

475
00:27:01,603 --> 00:27:02,000
months.

476
00:27:02,000 --> 00:27:06,194
Audrow Nash: For months. Mm-hmm.
Oh, yeah, yeah. Crazy. Man, what

477
00:27:06,194 --> 00:27:08,000
a cool thing. Yeah, it is...

478
00:27:08,000 --> 00:27:09,573
Sergey Litvinenko: How about
you? Are you on the same page,

479
00:27:09,573 --> 00:27:10,000
AI good for you?

480
00:27:10,000 --> 00:27:13,653
Audrow Nash: Oh, yeah, amazing.
I think it's the best time to be

481
00:27:13,653 --> 00:27:16,849
a builder. Like, anyone who
makes stuff, just incredible

482
00:27:16,849 --> 00:27:19,988
now. I don't know, for
everything, too. It's like, it's

483
00:27:19,988 --> 00:27:23,412
wild. I always... I'm like,
throwing whatever I can at it to

484
00:27:23,412 --> 00:27:26,893
see if it does well with it. And
some of the things that have

485
00:27:26,893 --> 00:27:30,318
been too hard in the past, like,
Opus 5.5 is now able to do.

486
00:27:30,318 --> 00:27:33,000
Like, it's just so wild. It's
such a cool time.

487
00:27:33,000 --> 00:27:34,000
Sergey Litvinenko: Mm-hmm. A
hundred percent.

488
00:27:34,000 --> 00:27:39,053
Audrow Nash: Uh-huh. How... So,
do you... I guess you can use AI

489
00:27:39,053 --> 00:27:43,237
very effectively to kind of
build systems to automate

490
00:27:43,237 --> 00:27:47,895
things. You can maybe use it for
some of the judgments that

491
00:27:47,895 --> 00:27:52,711
you're doing about, like, put
this here, put this here. Like,

492
00:27:52,711 --> 00:27:57,289
tell me about some of the
specific ways you're using AI in

493
00:27:57,289 --> 00:28:02,263
your whole platform, I suppose.
Like, how are you guys actually

494
00:28:02,263 --> 00:28:04,000
using it specifically?

495
00:28:04,000 --> 00:28:07,880
Sergey Litvinenko: Yeah, so a
part of the data intake and

496
00:28:07,880 --> 00:28:11,691
analysis would be a good
example. Whenever customers are

497
00:28:11,691 --> 00:28:15,503
not using the API and they
literally just throw whatever

498
00:28:15,503 --> 00:28:19,723
they have at us, the AI... We
have built a pipeline that takes

499
00:28:19,723 --> 00:28:23,806
all the data and then reads
through the whole thing and then

500
00:28:23,806 --> 00:28:27,754
finds what we were looking for
and formats in a way that's

501
00:28:27,754 --> 00:28:30,000
compatible with the rating
model.

502
00:28:30,000 --> 00:28:31,040
Audrow Nash: Mm-hmm. Hell yeah.
Yeah, that's such a good use

503
00:28:31,040 --> 00:28:32,150
case for it. Where you can find
the specific things to fill them

504
00:28:32,150 --> 00:28:33,260
in. That feels like, even for a
while now, like, I don't know, a

505
00:28:33,260 --> 00:28:34,301
year or two years, something
like that. Finding the parts to

506
00:28:34,301 --> 00:28:35,358
fill in, to make it... Like,
doing the interfacing? It's been

507
00:28:35,358 --> 00:28:36,000
phenomenal. That's such a great
idea.

508
00:28:36,000 --> 00:28:41,407
Sergey Litvinenko: Previously
you had 100%. And Mm-hmm. And I

509
00:28:41,407 --> 00:28:46,460
got to tell you the data
ingestion pipeline that we built

510
00:28:46,460 --> 00:28:51,335
where you transform whatever
you've been given into the

511
00:28:51,335 --> 00:28:56,476
format that you need, that
itself could be a company, like

512
00:28:56,476 --> 00:29:01,972
a separate company. I'm pretty
sure somebody's already working

513
00:29:01,972 --> 00:29:07,645
on it. So that has been the game
changer. That's probably a year

514
00:29:07,645 --> 00:29:08,000
ago.

515
00:29:08,000 --> 00:29:11,193
Audrow Nash: Oh, I'm sure. I
have a friend working on

516
00:29:11,193 --> 00:29:13,000
something similar. Yeah, yeah.

517
00:29:13,000 --> 00:29:16,702
Sergey Litvinenko: So that part
has been probably the best

518
00:29:16,702 --> 00:29:20,532
efficiency gain for us by far.
There is another cool feature

519
00:29:20,532 --> 00:29:24,362
that we have where we actually
turn your deployment contract

520
00:29:24,362 --> 00:29:28,319
language into the specific tasks
and control you need to do in

521
00:29:28,319 --> 00:29:32,277
order to make sure everything is
in a good shape. I'll give an

522
00:29:32,277 --> 00:29:36,234
example. So let's say you are
deploying with an enterprise and

523
00:29:36,234 --> 00:29:40,128
most likely that enterprise has
a procurement department that

524
00:29:40,128 --> 00:29:43,894
wants you to do things in a
certain way. So they send you a

525
00:29:43,894 --> 00:29:47,915
contract and then contract has
like a... it's like an operating

526
00:29:47,915 --> 00:29:51,936
agreement. I'll give an example
like if you are maybe deploying

527
00:29:51,936 --> 00:29:55,766
on some network, your robots,
that network will send you the

528
00:29:55,766 --> 00:29:58,894
agreement that you have to
operate by. And I mean

529
00:29:58,894 --> 00:30:02,532
historically people used to do
it. They would take like a

530
00:30:02,532 --> 00:30:06,489
Sharpie and they would take the
piece of paper and just go and

531
00:30:06,489 --> 00:30:10,000
you know mark everything. But we
built a pipeline where

532
00:30:10,000 --> 00:30:13,000
Audrow Nash: Oh, so tedious,
yeah.

533
00:30:13,000 --> 00:30:14,614
Sergey Litvinenko: Our system
analyzes the whole agreement,

534
00:30:14,614 --> 00:30:16,229
understands different clauses,
then organizes them and then

535
00:30:16,229 --> 00:30:17,871
looks for the existing controls
that you have on a platform,

536
00:30:17,871 --> 00:30:19,430
links them up together and
automatically tells you, "Hey,

537
00:30:19,430 --> 00:30:21,127
you're already meeting let's say
50 out of the 60 controls and

538
00:30:21,127 --> 00:30:22,851
here are the 10 requirements you
need to pay attention to." And

539
00:30:22,851 --> 00:30:24,547
that way you like know exactly
what you need to be focused on.

540
00:30:24,547 --> 00:30:26,134
That comes from our compliance
module which we now have in

541
00:30:26,134 --> 00:30:27,721
addition to insurance. But
people like that because I mean

542
00:30:27,721 --> 00:30:29,391
you hook up Koop with Claude
MCP, you upload that contract to

543
00:30:29,391 --> 00:30:31,005
your Claude and roughly in a
couple of minutes you're going

544
00:30:31,005 --> 00:30:32,619
to know exactly what you're
missing and then you can assign

545
00:30:32,619 --> 00:30:34,289
that to the people on your team
and off you go. Without AI it

546
00:30:34,289 --> 00:30:35,000
would have taken you weeks

547
00:30:35,000 --> 00:31:06,000
Audrow Nash: Mm-hmm. Cool.
Mm-hmm. Forever.

548
00:31:06,000 --> 00:31:07,000
Sergey Litvinenko: To get it
done. Like literally weeks.

549
00:31:07,000 --> 00:31:11,351
Audrow Nash: And you'd have your
expert on it, and it's just,

550
00:31:11,351 --> 00:31:15,774
what a thing. Yeah, it blows my
mind how, Just how much things

551
00:31:15,774 --> 00:31:20,125
have changed with that kind of
thing. Because they're... It's

552
00:31:20,125 --> 00:31:24,119
like... That was a whole...
Like, you would study it and

553
00:31:24,119 --> 00:31:28,613
work on it, and have the whole
thing in your head forever to go

554
00:31:28,613 --> 00:31:33,036
mark all the things, and that
would take forever. To link them

555
00:31:33,036 --> 00:31:36,745
up to requirements, to the
contract, and everything.

556
00:31:36,745 --> 00:31:40,882
Just... And now it does it in,
like, minutes. This kind of

557
00:31:40,882 --> 00:31:45,233
thing. Just so crazy. That seems
like a really good use case,

558
00:31:45,233 --> 00:31:49,370
too. Yeah, I think finding
relationships between different

559
00:31:49,370 --> 00:31:53,650
data in different places is just
phenomenal. Very cool. Hell

560
00:31:53,650 --> 00:31:57,573
yeah. Now, Has it helped? I
guess has AI helped with...

561
00:31:57,573 --> 00:32:01,354
Like, it strikes me...
Certifications are detailed. I

562
00:32:01,354 --> 00:32:05,562
haven't been too involved with
much for certifications, but

563
00:32:05,562 --> 00:32:09,842
I've done some, like... I don't
know... Simple certification

564
00:32:09,842 --> 00:32:13,979
trainings for some of the space
work I was doing when Open

565
00:32:13,979 --> 00:32:18,117
Robotics was doing it. And,
They're long. There's a lot to

566
00:32:18,117 --> 00:32:22,396
know. There's a... Like, people
study the specifications and

567
00:32:22,396 --> 00:32:26,819
everything. You being in a bunch
of different industries, it's

568
00:32:26,819 --> 00:32:31,313
like you have all these reports.
They're iterating and changing

569
00:32:31,313 --> 00:32:35,593
over time. I assume AI is very
helpful for staying on top of

570
00:32:35,593 --> 00:32:39,302
that, too, and seeing how you
have to relate the new

571
00:32:39,302 --> 00:32:43,368
certification to... Tell me
about that. Tell me about the

572
00:32:43,368 --> 00:32:47,647
certification space. And just,
like, keeping up with all the

573
00:32:47,647 --> 00:32:51,000
changing regulation around...
Well, everything.

574
00:32:51,000 --> 00:32:53,281
Sergey Litvinenko: Mm-hmm. Yeah.
Certifications are very

575
00:32:53,281 --> 00:32:55,643
important in our underwriting
process because not even the

576
00:32:55,643 --> 00:32:58,086
fact that you have all those
controls in place, but the fact

577
00:32:58,086 --> 00:33:00,000
that your company as the
organization made this

578
00:33:00,000 --> 00:33:02,281
investment and now has everybody
operating under certain

579
00:33:02,281 --> 00:33:04,480
principles. That's really
important because, you know,

580
00:33:04,480 --> 00:33:06,801
lots of people don't realize
that you can have, you know,

581
00:33:06,801 --> 00:33:09,285
even the safest robot out there,
and then somebody hacks your

582
00:33:09,285 --> 00:33:11,810
email and gets access to your
system, and then the whole thing

583
00:33:11,810 --> 00:33:14,050
goes down. The trouble comes
from the places you expect

584
00:33:14,050 --> 00:33:16,127
least. I mean, that's why
hackers are still hacking

585
00:33:16,127 --> 00:33:18,000
because they keep finding new
vulnerabilities.

586
00:33:18,000 --> 00:33:38,000
Audrow Nash: Mm-hmm. Mm... Yeah.
Mm-hmm.

587
00:33:38,000 --> 00:33:40,500
Sergey Litvinenko: And
certifications is probably the

588
00:33:40,500 --> 00:33:43,000
best way how to try to protect
yourself holistically.

589
00:33:43,000 --> 00:33:43,913
Audrow Nash: Cool. I like that
approach. Or that way of looking

590
00:33:43,913 --> 00:33:44,000
at it.

591
00:33:43,000 --> 00:33:47,490
Sergey Litvinenko: We are a big
fan of robot safety. We believe

592
00:33:47,490 --> 00:33:51,267
that there are standards out
there that have a lot of

593
00:33:51,267 --> 00:33:55,543
credibility like ISO 10218 or
what is it, R15.08. Those have

594
00:33:55,543 --> 00:33:58,892
great adoption, great
reputation. So whenever a

595
00:33:58,892 --> 00:34:02,598
customer has those, we
definitely mine those for the

596
00:34:02,598 --> 00:34:06,802
controls that we're interested
in for the underwriting. But

597
00:34:06,802 --> 00:34:10,508
even if you have that
certification, you get like an

598
00:34:10,508 --> 00:34:14,000
automated discount, okay,
because we already know

599
00:34:14,000 --> 00:34:20,000
Audrow Nash: Mm-hmm.

600
00:34:20,000 --> 00:34:23,444
Sergey Litvinenko: What they
have. So that definitely helps.

601
00:34:23,444 --> 00:34:27,175
That saves money. Also, it might
help with underwriting if, let's

602
00:34:27,175 --> 00:34:30,677
say, you are a, I would say,
conditional use case that we are

603
00:34:30,677 --> 00:34:34,178
on the fence about, but you have
a certification in place, we

604
00:34:34,178 --> 00:34:37,278
will say yes, okay. So
certifications are helping with

605
00:34:37,278 --> 00:34:39,000
both pricing and underwriting.

606
00:34:39,000 --> 00:34:43,000
Audrow Nash: Cool. Mm-hmm.

607
00:34:43,000 --> 00:34:47,111
Sergey Litvinenko: Something
that we're interested in, I

608
00:34:47,111 --> 00:34:51,296
think, so there is a lots of
rumors about what's going to

609
00:34:51,296 --> 00:34:55,187
happen with digital AI
regulations. They're trying to

610
00:34:55,187 --> 00:34:59,298
do some federal stuff. I mean,
we'll see what happens. I

611
00:34:59,298 --> 00:35:04,070
definitely think the physical AI
world will not be able to escape

612
00:35:04,070 --> 00:35:08,402
regulations. You'll have to do
them. And a great example is

613
00:35:08,402 --> 00:35:12,807
obviously European Union, EU. I
think starting January 2027,

614
00:35:12,807 --> 00:35:16,991
they plan to introduce EU
machinery regulation. I haven't

615
00:35:16,991 --> 00:35:21,690
read the body of the regulation
yet, but the premise is that any

616
00:35:21,690 --> 00:35:25,801
machine that runs any kind of
machine learning where the

617
00:35:25,801 --> 00:35:29,839
machine can perform any task or
a job using any kind of

618
00:35:29,839 --> 00:35:34,024
algorithm will have to be
certified. And I think that's a

619
00:35:34,024 --> 00:35:38,649
sign where everything is going,
is that these machines, robots,

620
00:35:38,649 --> 00:35:43,054
especially robots running any
kind of neural nets, will have

621
00:35:43,054 --> 00:35:46,725
to pass some form of
certification in order to get

622
00:35:46,725 --> 00:35:51,203
into the, I mean, probably will
not be needed for pilots, but

623
00:35:51,203 --> 00:35:55,535
for production and scale 100%.
And I'm watching closely how

624
00:35:55,535 --> 00:36:00,013
well the EU regulation will be
taken. I hope it will be taken

625
00:36:00,013 --> 00:36:04,198
in favorably, but don't forget
it's also a burden for the

626
00:36:04,198 --> 00:36:07,942
companies developing those
robots. And that's where

627
00:36:07,942 --> 00:36:12,494
hopefully we'll be able to help
by making it easier and trying

628
00:36:12,494 --> 00:36:16,238
to save them money via
insurance. That's the way to

629
00:36:16,238 --> 00:36:18,000
recoup their investment.

630
00:36:18,000 --> 00:36:22,524
Audrow Nash: Interesting. What's
your read on it? Do you think it

631
00:36:22,524 --> 00:36:26,491
will help? Do you think it
will... I mean, I've talked to

632
00:36:26,491 --> 00:36:30,597
founders where they're just
going to skip Europe because of

633
00:36:30,597 --> 00:36:35,121
the regulations for this kind of
thing. More regulations probably

634
00:36:35,121 --> 00:36:39,505
makes it even harder for this.
What is your... What's your take

635
00:36:39,505 --> 00:36:43,751
on it? Do you think it will go
favorably, as you've said? And

636
00:36:43,751 --> 00:36:47,927
by favorably, do you mean that
entrepreneurs will accept the

637
00:36:47,927 --> 00:36:52,311
burden of the certifications and
then they proceed that way and

638
00:36:52,311 --> 00:36:56,000
the market kind of stays safe
for this kind of thing?

639
00:36:56,000 --> 00:37:01,066
Sergey Litvinenko: So first of
all, if you look at our, like,

640
00:37:01,066 --> 00:37:06,215
do those things matter? If you
look at our data that we shared

641
00:37:06,215 --> 00:37:11,197
within the In Bots We Trust
Report, companies that have some

642
00:37:11,197 --> 00:37:15,765
form of safety control, safety
plan and certifications,

643
00:37:15,765 --> 00:37:20,000
experience less claims. So I
mean, we don't have...

644
00:37:20,000 --> 00:37:23,121
Audrow Nash: Is it causality, do
you think? Do you think it's

645
00:37:23,121 --> 00:37:26,191
causal or... Like, I wonder if
it's just like they're better

646
00:37:26,191 --> 00:37:29,312
run, therefore they have these
certifications, therefore they

647
00:37:29,312 --> 00:37:32,433
have fewer claims? Or do you
think it's... I don't know, it's

648
00:37:32,433 --> 00:37:35,605
like that whole military thing
where it's like a ship that has

649
00:37:35,605 --> 00:37:38,879
the little details polished, has
all its maintenance sorted out,

650
00:37:38,879 --> 00:37:42,000
this kind of thing. Do you think
it could be similar to that?

651
00:37:42,000 --> 00:37:44,122
Sergey Litvinenko: Probably, I
think so. I think so, yeah. If

652
00:37:44,122 --> 00:37:46,104
you are able to achieve a robot
safety certification, you

653
00:37:46,104 --> 00:37:48,296
probably have everything else in
order. Yep. So it might not be

654
00:37:48,296 --> 00:37:50,000
fair to attribute this just to
the certification,

655
00:37:50,000 --> 00:37:56,000
Audrow Nash: Everything else
taken care of?

656
00:37:56,000 --> 00:37:58,638
Sergey Litvinenko: But if you
run the numbers and you look at

657
00:37:58,638 --> 00:38:00,800
the companies that had
accidents, that didn't have

658
00:38:00,800 --> 00:38:02,962
accidents, the ones that didn't
have accidents had

659
00:38:02,962 --> 00:38:04,000
certifications in place.

660
00:38:04,000 --> 00:38:07,633
Audrow Nash: Yeah. And also,
from your perspective, as an

661
00:38:07,633 --> 00:38:11,458
insurance company, it doesn't
matter what's causal. You kind

662
00:38:11,458 --> 00:38:15,410
of want to lump people into the
like the lowest risk from your

663
00:38:15,410 --> 00:38:19,044
perspective as possible. You
don't necessarily care about

664
00:38:19,044 --> 00:38:22,741
causality as... Like, it's too
hard to find, it's too many

665
00:38:22,741 --> 00:38:26,566
resources to find, it's not
worth it. But you see it as just

666
00:38:26,566 --> 00:38:30,390
better if these companies have
this certification because in

667
00:38:30,390 --> 00:38:33,705
your observation, these
companies have fewer claims,

668
00:38:33,705 --> 00:38:36,000
this kind of thing. Is that
correct?

669
00:38:36,000 --> 00:38:39,221
Sergey Litvinenko: Correct. We
look at the evidence, at the

670
00:38:39,221 --> 00:38:42,387
real world experience, and if
the experience tells us that

671
00:38:42,387 --> 00:38:45,717
that's working, we update our
priors and we incorporate that.

672
00:38:45,717 --> 00:38:48,720
Obviously, statistical
significance is the thing, and I

673
00:38:48,720 --> 00:38:52,159
would like to test that on all
the robotics companies globally,

674
00:38:52,159 --> 00:38:55,325
but I think, you know, our
sample size is getting there. I

675
00:38:55,325 --> 00:38:58,000
think we'll get to statistical
significance soon.

676
00:38:58,000 --> 00:39:02,296
Audrow Nash: Yeah. But it
also... The thing that's

677
00:39:02,296 --> 00:39:06,850
interesting is it really doesn't
matter. Like, for an

678
00:39:06,850 --> 00:39:11,575
organization that's doing
insurance, like for Koop, you

679
00:39:11,575 --> 00:39:16,902
are navigating so many complex
spaces. You're going to pick up

680
00:39:16,902 --> 00:39:21,714
on things that increase or
decrease risk. And just... If

681
00:39:21,714 --> 00:39:26,955
you find that there's a p-value,
like a small p-value, and it

682
00:39:26,955 --> 00:39:32,368
ends up reducing the likelihood
of claims and things, then it's

683
00:39:32,368 --> 00:39:34,000
a good thing to do.

684
00:39:34,000 --> 00:39:35,000
Sergey Litvinenko: Exactly.

685
00:39:34,000 --> 00:39:37,376
Audrow Nash: Just... Like, it's
like... You move up a level of

686
00:39:37,376 --> 00:39:40,535
abstraction, you don't worry
about causality, and you say,

687
00:39:40,535 --> 00:39:43,965
"Across the board, we see these
things." And that's how we make

688
00:39:43,965 --> 00:39:45,000
our determinations.

689
00:39:45,000 --> 00:39:48,050
Sergey Litvinenko: Yep, Mm-hmm.
Yep, that's what insurance is

690
00:39:48,050 --> 00:39:49,000
all about. Exactly.

691
00:39:49,000 --> 00:39:50,000
Audrow Nash: Yeah. It's cool.
It's

692
00:39:50,000 --> 00:39:51,765
Sergey Litvinenko: You see,
you're becoming an insurance guy

693
00:39:51,765 --> 00:39:52,000
with me.

694
00:39:52,000 --> 00:39:55,857
Audrow Nash: I feel like it's a
space that most people don't

695
00:39:55,857 --> 00:39:59,907
find interesting, I think, but
it's very interesting. It's like

696
00:39:59,907 --> 00:40:01,000
once you start to

697
00:40:01,000 --> 00:40:05,000
Sergey Litvinenko: Mm-hmm. Yeah.

698
00:40:05,000 --> 00:40:09,129
Audrow Nash: See the details of
it, it's really cool work you're

699
00:40:09,129 --> 00:40:12,871
doing, and you're enabling so
much to happen by doing this

700
00:40:12,871 --> 00:40:16,355
work that most people wouldn't
consider. Which is just

701
00:40:16,355 --> 00:40:17,000
wonderful.

702
00:40:17,000 --> 00:40:17,711
Sergey Litvinenko: You know,
Warren Buffett made a fortune on

703
00:40:17,711 --> 00:40:18,422
GEICO. That was his big thing,
and he realized that insurance

704
00:40:18,422 --> 00:40:19,098
was an attractive business. I
really hope that in the 21st

705
00:40:19,098 --> 00:40:19,716
century, it's going to be a
similar story, with robot

706
00:40:19,716 --> 00:40:20,322
insurance, when you start
shipping billions of those

707
00:40:20,322 --> 00:40:20,939
things, and you have to
financially protect them. Not

708
00:40:20,939 --> 00:40:21,615
only them, but also humans
around them, and I think it's a

709
00:40:21,615 --> 00:40:22,000
revolution. It's going to be
fun.

710
00:40:22,000 --> 00:40:27,635
Audrow Nash: Yeah. Mm-hmm. Yeah.
And especially if there are

711
00:40:27,635 --> 00:40:33,083
billions of robots out there,
like, you guys are in such a

712
00:40:33,083 --> 00:40:38,718
good position to ride that whole
wave, too. So if Tesla does

713
00:40:38,718 --> 00:40:44,166
deliver on making a million
humanoids, or whatever by some

714
00:40:44,166 --> 00:40:49,050
date, or figure, or any
companies, even if it's just

715
00:40:49,050 --> 00:40:54,779
autonomous mobile robots, it's
like you ride that wave, which

716
00:40:54,779 --> 00:40:56,000
is excellent.

717
00:40:56,000 --> 00:41:04,000
Sergey Litvinenko: Mm-hmm.

718
00:41:04,000 --> 00:41:08,233
Audrow Nash: What do you
think... Like... I guess, what's

719
00:41:08,233 --> 00:41:12,764
your, Where do you think it's...
Like, I guess... What do you

720
00:41:12,764 --> 00:41:17,220
think the rate of growth is for
this? I'm sure that you all,

721
00:41:17,220 --> 00:41:22,048
having 25% of the market, are...
I mean, that's incredible growth

722
00:41:22,048 --> 00:41:26,578
for you all. But then you're
looking at the total addressable

723
00:41:26,578 --> 00:41:31,183
market growing. Like, what do
you see about the robotics space

724
00:41:31,183 --> 00:41:32,000
in general?

725
00:41:32,000 --> 00:41:34,722
Sergey Litvinenko: Well, to
begin with, robots is a small

726
00:41:34,722 --> 00:41:37,779
market. if you Even if you loop
in drones and industrial robots,

727
00:41:37,779 --> 00:41:40,501
still a small market. In the
U.S., you can probably track

728
00:41:40,501 --> 00:41:43,272
anywhere between 800 to 1200
robotics companies. In China,

729
00:41:43,272 --> 00:41:45,755
there is more, but in China,
there are lots of parts

730
00:41:45,755 --> 00:41:48,430
manufacturers, okay? So, the
ecosystem is still early. I

731
00:41:48,430 --> 00:41:51,296
would say, well, it's fair to
say that Koop is a function of

732
00:41:51,296 --> 00:41:54,257
that industry. So, we are going
to be a derivative, and if the

733
00:41:54,257 --> 00:41:57,218
industry goes through the roof
and becomes like multi-trillion

734
00:41:57,218 --> 00:41:59,893
dollar, which I think it
eventually will, we're going to

735
00:41:59,893 --> 00:42:02,854
follow that. In the meantime,
we're just gaining market share.

736
00:42:02,854 --> 00:42:04,000
That's what we're doing.

737
00:42:04,000 --> 00:42:11,170
Audrow Nash: Yep. Mm-hmm. Hell
yeah. Yep. Because you have the

738
00:42:11,170 --> 00:42:18,571
two hills to climb. You have the
pie getting bigger overall, but

739
00:42:18,571 --> 00:42:21,000
then you have growing

740
00:42:21,000 --> 00:42:25,000
Sergey Litvinenko: And, yeah.

741
00:42:25,000 --> 00:42:26,371
Audrow Nash: Part... You have a
growing part of that existing

742
00:42:26,371 --> 00:42:27,000
market, which is quite cool.

743
00:42:27,000 --> 00:42:30,530
Sergey Litvinenko: Yeah. I think
one... So, historically, if you

744
00:42:30,530 --> 00:42:33,950
look back, robots grew in places
that you didn't see or didn't

745
00:42:33,950 --> 00:42:37,370
hear about. It's factories, you
know? So, if you go about your

746
00:42:37,370 --> 00:42:40,790
daily life, you don't really see
the robots. I think, starting

747
00:42:40,790 --> 00:42:44,155
probably next year, we're going
to see... This is going to be

748
00:42:44,155 --> 00:42:47,464
the first time in the human
history where robots will become

749
00:42:47,464 --> 00:42:50,609
like a daily occurrence to
millions of people, and that's

750
00:42:50,609 --> 00:42:53,532
not going to be through
humanoids just yet. Humanoids

751
00:42:53,532 --> 00:42:57,007
right now are still at the demo
stage. It's going to be through

752
00:42:57,007 --> 00:42:58,000
self-driving cars.

753
00:42:58,000 --> 00:43:03,000
Audrow Nash: Mm-hmm. Mm.

754
00:43:03,000 --> 00:43:04,158
Sergey Litvinenko: Right now,
you go to SF to Austin, they're

755
00:43:04,158 --> 00:43:05,259
everywhere, and I think Waymo
and Tesla are about to flood

756
00:43:05,259 --> 00:43:06,361
this market with their CapEx
money, and that's going to be

757
00:43:06,361 --> 00:43:07,557
the first time in human history
where robots are going to be as

758
00:43:07,557 --> 00:43:08,658
easy as getting a latte at
Starbucks. You know? It's going

759
00:43:08,658 --> 00:43:09,000
to be like normal.

760
00:43:09,000 --> 00:43:21,000
Audrow Nash: Yeah, probably
true.

761
00:43:21,000 --> 00:43:25,036
Sergey Litvinenko: And I'm super
excited about this. I think

762
00:43:25,036 --> 00:43:28,735
that's going to be the moment
when we'll see other form

763
00:43:28,735 --> 00:43:32,973
factors come through the door as
well, delivery robots, drones,

764
00:43:32,973 --> 00:43:36,000
humanoids, and we'll see what
else we'll get.

765
00:43:36,000 --> 00:43:40,822
Audrow Nash: Mm-hmm. Yeah, it's
a really cool time. I do feel

766
00:43:40,822 --> 00:43:44,853
like we're on the cusp of
everything. And it's also

767
00:43:44,853 --> 00:43:49,517
like... It was coming. Like, it
was feeling inevitable five

768
00:43:49,517 --> 00:43:54,260
years ago. Like, it's coming.
It's getting there. But I feel

769
00:43:54,260 --> 00:43:58,844
like with AI, it's coming so
much faster. And with all the

770
00:43:58,844 --> 00:44:03,508
emphasis on compute and things
like this, so people can get

771
00:44:03,508 --> 00:44:08,330
good dynamic models of robots
and things like this. And maybe

772
00:44:08,330 --> 00:44:12,361
some, like, I don't know,
language action models or

773
00:44:12,361 --> 00:44:17,183
vision, whatever, action models.
It's coming faster, which is

774
00:44:17,183 --> 00:44:22,163
really, really cool. Let's see.
Do you have any guess on, like,

775
00:44:22,163 --> 00:44:27,222
growth percentage per year kind
of thing? Or, like, I'm sure you

776
00:44:27,222 --> 00:44:32,123
look at your total addressable
market and whatnot. Do you have

777
00:44:32,123 --> 00:44:37,024
any thoughts about that or any,
like, any concrete feelings on

778
00:44:37,024 --> 00:44:39,000
where the market's going?

779
00:44:39,000 --> 00:44:42,909
Sergey Litvinenko: Yeah, well,
in my previous answer, I hyped

780
00:44:42,909 --> 00:44:46,883
the market. Now, I can tell you
what we're actually seeing. So

781
00:44:46,883 --> 00:44:50,087
to put into perspective,
whenever people, whatever

782
00:44:50,087 --> 00:44:53,868
companies come to Koop, they
typically have to tell us what

783
00:44:53,868 --> 00:44:57,457
they expect to see 12 months
from now. So we call it the

784
00:44:57,457 --> 00:45:00,917
projected performance and
projected exposure. And they

785
00:45:00,917 --> 00:45:04,378
say, well, you know, in 12
months, we're going to have

786
00:45:04,378 --> 00:45:08,223
1,000 robots. And right now, we
have 100. If you look at the

787
00:45:08,223 --> 00:45:12,389
difference between how much they
were projecting to how much they

788
00:45:12,389 --> 00:45:15,977
actually delivered, the
difference is 5 to 6x. So people

789
00:45:15,977 --> 00:45:20,143
say, oh, we're going to do 5,000
robots, they end up doing 1,000.

790
00:45:20,143 --> 00:45:24,244
So that means that the ambition
is there. People definitely want

791
00:45:24,244 --> 00:45:27,000
to grow, but deployments are
hard. Exactly.

792
00:45:27,000 --> 00:45:31,186
Audrow Nash: Mm-hmm. Reality's
hard. Yeah, manufacturing's

793
00:45:31,186 --> 00:45:34,000
hard. Updates are hard. Lots of
things.

794
00:45:34,000 --> 00:45:37,980
Sergey Litvinenko: So I think,
yeah, physical world is hard.

795
00:45:37,980 --> 00:45:41,695
And I really hope that,
especially with the way hardware

796
00:45:41,695 --> 00:45:45,477
gets iterated and how AI models
behave themselves on that

797
00:45:45,477 --> 00:45:49,457
hardware, I really hope that gap
is going to contract in the

798
00:45:49,457 --> 00:45:53,371
coming years. And people will
say, oh, we're planning to do

799
00:45:53,371 --> 00:45:57,351
1,000 robots, but we ended up
doing 2,000 robots. Right now,

800
00:45:57,351 --> 00:46:01,000
it's a long way to get there,
but hopefully soon. Thank

801
00:46:01,000 --> 00:46:07,031
Audrow Nash: Yeah. Do you have
any, so you mentioned the

802
00:46:07,031 --> 00:46:08,000
Cybercab.

803
00:46:08,000 --> 00:46:09,000
Sergey Litvinenko: You.

804
00:46:09,000 --> 00:46:13,946
Audrow Nash: How, because I'm
near Austin. yeah, yeah. And we

805
00:46:13,946 --> 00:46:18,811
see them all the time. what,
that to me has gone from, like,

806
00:46:18,811 --> 00:46:23,595
nothing to a bunch of them,
which is so cool. how does that

807
00:46:23,595 --> 00:46:27,730
fit in this? Are they beating
your expectations for

808
00:46:27,730 --> 00:46:32,432
deployments? Are they still
going slow? Like, I guess, how

809
00:46:32,432 --> 00:46:36,000
does that fit in this from your
perspective?

810
00:46:36,000 --> 00:46:37,139
Sergey Litvinenko: Cybercabs is
probably the reason why I'm

811
00:46:37,139 --> 00:46:38,356
super excited for 2027. I think
2027 is going to be pivotal for

812
00:46:38,356 --> 00:46:39,515
them. And if you look at the
chart of Waymo's deployment and

813
00:46:39,515 --> 00:46:40,712
Cybercab deployment, Waymo has
been the AV leader for the past

814
00:46:40,712 --> 00:46:41,813
few years, and they took a very
measured, safety-oriented

815
00:46:41,813 --> 00:46:42,972
approach. And that was great.
They helped bring the industry

816
00:46:42,972 --> 00:46:44,112
to the level where it is today.
But I think Tesla is such a

817
00:46:44,112 --> 00:46:45,000
different beast that they now
can afford to do

818
00:46:45,000 --> 00:47:12,000
Audrow Nash: Hell yeah. Mm-hmm.

819
00:47:12,000 --> 00:47:12,529
Sergey Litvinenko: With their
manufacturing ability and the

820
00:47:12,529 --> 00:47:13,058
way they run FSD with camera
only. Yeah, you could see that

821
00:47:13,058 --> 00:47:13,605
they ramped it. Waymo was going
like this, and then Tesla was

822
00:47:13,605 --> 00:47:14,000
like this. I think that's going
to continue.

823
00:47:14,000 --> 00:47:25,000
Audrow Nash: That's the
difference to me. Yep.

824
00:47:25,000 --> 00:47:29,108
Sergey Litvinenko: The reason
I'm excited about Cybercabs for

825
00:47:29,108 --> 00:47:33,215
Koop is actually something that
our company has been thinking

826
00:47:33,215 --> 00:47:37,054
about since 2021. Right now,
most of the Waymo's fleet, I

827
00:47:37,054 --> 00:47:41,094
believe, and Tesla's Cybercab
fleet, they owned by Tesla and

828
00:47:41,094 --> 00:47:45,000
Waymo. They own them, they
operate them, they insure them.

829
00:47:45,000 --> 00:47:47,425
Audrow Nash: Oh, consumers
buying them. Is that what you're

830
00:47:47,425 --> 00:47:48,000
excited about?

831
00:47:48,000 --> 00:47:53,000
Sergey Litvinenko: So consumers,
like,

832
00:47:53,000 --> 00:47:54,000
Audrow Nash: Or businesses, say.

833
00:47:54,000 --> 00:47:57,690
Sergey Litvinenko: Possible.
Consumers possible, maybe at

834
00:47:57,690 --> 00:48:00,992
some point. But in the short
term, it's going to be

835
00:48:00,992 --> 00:48:04,941
businesses buying self-driving
cars. Fleets, the way you have

836
00:48:04,941 --> 00:48:08,567
trucking fleets or used to have
taxi fleets, you'll have

837
00:48:08,567 --> 00:48:12,322
third-party fleets that will
own, maintain, charge, clean,

838
00:48:12,322 --> 00:48:15,818
update those vehicles, and
they're going to be off the

839
00:48:15,818 --> 00:48:19,897
balance sheet of Tesla, and it's
going to be an opportunity for

840
00:48:19,897 --> 00:48:23,846
hundreds of thousands of fleets
all around the world to scale

841
00:48:23,846 --> 00:48:27,472
autonomous vehicles. And the
moment when Tesla announced

842
00:48:27,472 --> 00:48:31,356
Cybercab and they posted that
form, they said, "Hey, help us

843
00:48:31,356 --> 00:48:35,371
build the robotaxi network, fill
out the information, and then

844
00:48:35,371 --> 00:48:39,255
who are you? Are you a fleet
operator or like some ecosystem

845
00:48:39,255 --> 00:48:43,334
or something else?" When I saw
this, I was like, "Okay, we were

846
00:48:43,334 --> 00:48:47,477
right about this five years ago.
That's the way it's going. It's

847
00:48:47,477 --> 00:48:51,038
going to be fleets doing
self-driving cars." And that's

848
00:48:51,038 --> 00:48:54,729
exactly where Koop can help
because Koop already actually

849
00:48:54,729 --> 00:48:58,548
has products for third-party
autonomous vehicle fleets. And

850
00:48:58,548 --> 00:49:02,109
they're all going to need a
non-trivial way of how that

851
00:49:02,109 --> 00:49:05,929
insurance is going to be done
for them. So I have very high

852
00:49:05,929 --> 00:49:09,878
hopes for Cybercabs, and I think
it's just going to be a game

853
00:49:09,878 --> 00:49:13,180
changer, not only for
transportation, but also as a

854
00:49:13,180 --> 00:49:17,000
new business opportunity and
obviously for players like us.

855
00:49:17,000 --> 00:49:21,311
Audrow Nash: Mm-hmm. Hell yeah.
Yeah, I'm undecided, like, being

856
00:49:21,311 --> 00:49:25,623
near Austin. It's like, there's
a bunch of them. And I suspect a

857
00:49:25,623 --> 00:49:29,799
lot of people around here will
be buying them too. And so it's

858
00:49:29,799 --> 00:49:34,111
like, it'd be interesting, like,
buy one or two or something and

859
00:49:34,111 --> 00:49:38,220
start sending it out and seeing
if you get rides. But I guess

860
00:49:38,220 --> 00:49:42,531
the demand for transportation is
really high. I wonder, I guess,

861
00:49:42,531 --> 00:49:46,708
I wonder if it'll be like, a few
hubs are absolutely saturated

862
00:49:46,708 --> 00:49:50,952
with them. Or, like, go out to
middle of nowhere where there is

863
00:49:50,952 --> 00:49:54,994
a dense population. Like, I
mean, just not like Austin or SF

864
00:49:54,994 --> 00:49:58,766
or wherever. But, I don't know.
I'm looking at it like a

865
00:49:58,766 --> 00:50:02,000
business opportunity. It's an
interesting thing.

866
00:50:02,000 --> 00:50:05,623
Sergey Litvinenko: Huge business
opportunity. Yeah, I mean,

867
00:50:05,623 --> 00:50:09,246
insurance is a derivative of
that, but people who are going

868
00:50:09,246 --> 00:50:12,684
to be doing just Cybercabs,
whether they're going to own

869
00:50:12,684 --> 00:50:16,368
them or operate them, there are
already people coming to us,

870
00:50:16,368 --> 00:50:19,991
looping funds from different
individuals to start an LLC to

871
00:50:19,991 --> 00:50:23,000
do that. And it's just going to
be great. I think

872
00:50:23,000 --> 00:50:27,000
Audrow Nash: Mm-hmm. Cool.

873
00:50:27,000 --> 00:50:30,278
Sergey Litvinenko: Autonomous
vehicle ownership is going to

874
00:50:30,278 --> 00:50:33,333
compete with like home
ownership. It's going to be just

875
00:50:33,333 --> 00:50:36,000
very profitable, very stable and
high liquidity.

876
00:50:36,000 --> 00:50:39,970
Audrow Nash: Yeah, because I
guess one thing is that you can

877
00:50:39,970 --> 00:50:43,940
sell it, should you need to. You
can move it somewhere else.

878
00:50:43,940 --> 00:50:48,042
Like, if you're in Austin, you
can go out to, I'm in Lockhart,

879
00:50:48,042 --> 00:50:52,011
you can go to Lockhart, you can
go to Bastrop, you can go to

880
00:50:52,011 --> 00:50:55,981
Oklahoma, you can go to, like,
you can just push them out to

881
00:50:55,981 --> 00:51:00,017
different places because it may
be saturated in one area, but

882
00:51:00,017 --> 00:51:04,119
you could keep moving them and
that will keep, like, until the

883
00:51:04,119 --> 00:51:07,824
whole U.S. is saturated, which
is a big country. And so,

884
00:51:07,824 --> 00:51:11,000
there's probably a tremendous
opportunity there.

885
00:51:11,000 --> 00:51:16,000
Sergey Litvinenko: Yeah. Yeah.

886
00:51:16,000 --> 00:51:16,846
Audrow Nash: Hell yeah. So,
you're gonna own a fleet of

887
00:51:16,846 --> 00:51:17,000
Cybercabs.

888
00:51:17,000 --> 00:51:17,510
Sergey Litvinenko: One thing, I
I'm definitely thinking about

889
00:51:17,510 --> 00:51:17,996
this. Look, just to give it a
try. I mean, look, right now

890
00:51:17,996 --> 00:51:18,523
there is no real infrastructure
in place where you can do it in

891
00:51:18,523 --> 00:51:19,000
a hassle-free way, but I think
people will figure it out.

892
00:51:19,000 --> 00:51:34,000
Audrow Nash: Yeah. Yeah. Yeah.
Yeah.

893
00:51:34,000 --> 00:51:38,356
Sergey Litvinenko: But you know,
it's like just talking about the

894
00:51:38,356 --> 00:51:42,510
opportunity, take whatever Uber
and Lyft do, then take mileage

895
00:51:42,510 --> 00:51:45,995
driven in your personally owned
cars and take public

896
00:51:45,995 --> 00:51:47,000
transportation.

897
00:51:47,000 --> 00:51:49,000
Audrow Nash: Mm-hmm. Sum it up.
Yeah.

898
00:51:49,000 --> 00:51:49,673
Sergey Litvinenko: So the whole
thing is going to be so

899
00:51:49,673 --> 00:51:50,433
autonomous vehicles will come
for the whole thing. And because

900
00:51:50,433 --> 00:51:51,167
they're going to be so safe and
pleasant to use, it actually

901
00:51:51,167 --> 00:51:51,951
going to expand the pie because
more people will want to do them

902
00:51:51,951 --> 00:51:52,000
too.

903
00:51:52,000 --> 00:51:56,549
Audrow Nash: Yeah. Yeah. Yeah, I
know, like, a number of people

904
00:51:56,549 --> 00:52:01,170
don't like Uber because of the,
like, experience with the driver

905
00:52:01,170 --> 00:52:05,069
often, this kind of thing, or
have, like, one or a few

906
00:52:05,069 --> 00:52:09,256
negative experiences and then
kind of, like, write it off,

907
00:52:09,256 --> 00:52:13,661
which is interesting. So then,
like, with this, it's like you

908
00:52:13,661 --> 00:52:18,209
alone in a clean car, ideally,
and then it just takes you where

909
00:52:18,209 --> 00:52:22,614
you want. And, like, I mean, we
don't have a Cybercab, but we

910
00:52:22,614 --> 00:52:27,162
have the Tesla, and it's just,
like, the Tesla Y, and it's just

911
00:52:27,162 --> 00:52:31,422
amazing. Like, it drives 98% of
our miles for us. It's just

912
00:52:31,422 --> 00:52:32,000
bonkers.

913
00:52:32,000 --> 00:52:33,000
Sergey Litvinenko: I would say
outside of your

914
00:52:33,000 --> 00:52:36,000
Audrow Nash: So.

915
00:52:36,000 --> 00:52:39,520
Sergey Litvinenko: Chatbot that
you use, I mean, let's call it

916
00:52:39,520 --> 00:52:43,153
an AI system that you use day to
day, be that Claude or ChatGPT,

917
00:52:43,153 --> 00:52:46,559
driving in a self-driving car is
the second most intimate AI

918
00:52:46,559 --> 00:52:49,000
experience after working with
your chatbot,

919
00:52:49,000 --> 00:52:49,776
Audrow Nash: Yeah. I could say
that, too. Hell yeah. What a

920
00:52:49,776 --> 00:52:50,000
cool thing. Okay.

921
00:52:50,000 --> 00:52:55,000
Sergey Litvinenko: In my view.
Yeah. Brave new world.

922
00:52:55,000 --> 00:52:58,980
Audrow Nash: Yeah. Making me
excited about the idea of that.

923
00:52:58,980 --> 00:53:03,291
Let's see. So, one thing that I,
well, so, first, I mean, there's

924
00:53:03,291 --> 00:53:07,536
so many things I'd like to talk
about, but you have your In Bots

925
00:53:07,536 --> 00:53:11,781
We Trust report. We've mentioned
it a few times. Want to tell me

926
00:53:11,781 --> 00:53:14,964
about it? Just kind of
high-level what were your

927
00:53:14,964 --> 00:53:18,811
findings, and then, like, what
are you thinking behind the

928
00:53:18,811 --> 00:53:21,000
report? What's your big
takeaway?

929
00:53:21,000 --> 00:53:25,163
Sergey Litvinenko: In September
we published a report called In

930
00:53:25,163 --> 00:53:29,392
Bots We Trust 2026. This report,
which is roughly 35 pages long,

931
00:53:29,392 --> 00:53:33,225
summarizes our experience
underwriting physical AI for the

932
00:53:33,225 --> 00:53:36,595
last five years. And we
published different tables,

933
00:53:36,595 --> 00:53:40,427
actuarial tables, charts,
looking at the capacity, looking

934
00:53:40,427 --> 00:53:44,458
at the losses, looking at the
performance. And it's the first

935
00:53:44,458 --> 00:53:48,687
report, I believe, in the world
that does that in a more or less

936
00:53:48,687 --> 00:53:52,520
standardized way and actually
shows people what's working,

937
00:53:52,520 --> 00:53:56,022
what's not working in the
robotic space. So I'm super

938
00:53:56,022 --> 00:54:00,317
proud of it because I think it's
a original piece of work. We got

939
00:54:00,317 --> 00:54:04,348
great feedback for it and we're
going to hopefully publish it

940
00:54:04,348 --> 00:54:06,000
every year going forward.

941
00:54:06,000 --> 00:54:13,000
Audrow Nash: Hell yeah.

942
00:54:13,000 --> 00:54:17,129
Sergey Litvinenko: So like I
mentioned earlier, we touched

943
00:54:17,129 --> 00:54:21,186
roughly 20-25% of the US
robotics market and that gave us

944
00:54:21,186 --> 00:54:24,817
the credibility to go and
publish the results to it

945
00:54:24,817 --> 00:54:28,803
published. But some key
takeaways is that, first of all,

946
00:54:28,803 --> 00:54:32,861
robots as a risk class is doing
great. From the moment we

947
00:54:32,861 --> 00:54:37,132
started to where we are, the
cost of robot liability dropped

948
00:54:37,132 --> 00:54:41,475
69%. So what does it mean? It
means that companies are paying

949
00:54:41,475 --> 00:54:45,247
less year over year for
protecting themselves against

950
00:54:45,247 --> 00:54:49,447
robot mistakes. It means that
the robots are getting better

951
00:54:49,447 --> 00:54:53,505
and the companies behind those
robots are getting better.

952
00:54:53,505 --> 00:54:57,563
Second, deployments are hard.
Even though the industry is

953
00:54:57,563 --> 00:55:01,051
growing, it's not as growing
according to its own

954
00:55:01,051 --> 00:55:05,322
expectations, primarily because
people are taking a measured

955
00:55:05,322 --> 00:55:09,593
approach. Maybe they would want
to 10x their fleet, but just

956
00:55:09,593 --> 00:55:13,651
because of the deployment
friction, they cannot. Or maybe

957
00:55:13,651 --> 00:55:16,000
there is limited customer
demand.

958
00:55:16,000 --> 00:55:18,744
Audrow Nash: What do you mean,
"measured approach"? I don't

959
00:55:18,744 --> 00:55:20,000
know what that means. Okay.

960
00:55:20,000 --> 00:55:24,580
Sergey Litvinenko: Well, let's
say if you have a customer, a

961
00:55:24,580 --> 00:55:29,466
pilot, and you are signed up to
deploy, let's say, 10 inspection

962
00:55:29,466 --> 00:55:34,122
robots to inspect some oil and
gas facilities, maybe you wish

963
00:55:34,122 --> 00:55:38,702
you could deploy 100 robots
there, but due to the customer's

964
00:55:38,702 --> 00:55:42,977
internal demand or internal
politics, you are unable to,

965
00:55:42,977 --> 00:55:47,405
even if you can. So I think
that's why companies just like

966
00:55:47,405 --> 00:55:50,000
subject to whatever their
customer

967
00:55:50,000 --> 00:55:51,000
Audrow Nash: We'll do.

968
00:55:50,000 --> 00:55:50,920
Sergey Litvinenko: Imposes on
them. Lots of those customers,

969
00:55:50,920 --> 00:55:51,825
don't forget they are either
enterprises or the government.

970
00:55:51,825 --> 00:55:52,684
So even if you are able,
sometimes you're constrained by

971
00:55:52,684 --> 00:55:53,635
the contract that you signed. So
that's why measured approach,

972
00:55:53,635 --> 00:55:54,540
meaning that you have to pace
yourself, measure yourself to

973
00:55:54,540 --> 00:55:55,000
whatever you're allowed to do.

974
00:55:55,000 --> 00:56:01,667
Audrow Nash: Mm-hmm. Oh, I see.
You're limited, effectively.

975
00:56:01,667 --> 00:56:08,667
Yeah. One thing that I've heard
in One thing that I've heard on

976
00:56:08,667 --> 00:56:12,000
this "measured approach" idea,

977
00:56:12,000 --> 00:56:18,000
Sergey Litvinenko: Yeah, another
thing. Yeah.

978
00:56:18,000 --> 00:56:20,428
Audrow Nash: A lot of companies
that are big will try

979
00:56:20,428 --> 00:56:23,314
everything. Like, they'll go and
they'll buy one of each of all

980
00:56:23,314 --> 00:56:26,017
the different robot startups
that are trying to solve their

981
00:56:26,017 --> 00:56:28,766
thing, and they'll just, like,
kind of not commit to anyone.

982
00:56:28,766 --> 00:56:31,423
But I've heard this over and
over, where it's like, people

983
00:56:31,423 --> 00:56:34,126
can get in the door, they're
with this big company, they're

984
00:56:34,126 --> 00:56:37,058
psyched about it, and then that
one big company is, like, really

985
00:56:37,058 --> 00:56:39,669
dragging their feet or not
wanting to renew, even if they

986
00:56:39,669 --> 00:56:42,555
did show results for this kind
of thing, because it's harder to

987
00:56:42,555 --> 00:56:44,846
get. It's like they have
innovation initiatives or

988
00:56:44,846 --> 00:56:47,686
something, and so they go try
something for it, but they don't

989
00:56:47,686 --> 00:56:50,618
really have a serious intention
of heavy adoption of it. I don't

990
00:56:50,618 --> 00:56:53,000
know how much of that is, like,
a measured approach.

991
00:56:53,000 --> 00:57:05,000
Sergey Litvinenko: Oh yeah.

992
00:57:05,000 --> 00:57:07,000
Audrow Nash: But maybe some of
it.

993
00:57:07,000 --> 00:57:09,769
Sergey Litvinenko: Probably I
would say in our customer base

994
00:57:09,769 --> 00:57:12,492
probably most of our customers
are experiencing that. yeah,

995
00:57:12,492 --> 00:57:13,000
yeah, yeah.

996
00:57:13,000 --> 00:57:17,677
Audrow Nash: Yeah, that's crazy.
What so What's the big

997
00:57:17,677 --> 00:57:22,525
bottleneck with it? It's the
company dragging their feet.

998
00:57:22,525 --> 00:57:27,457
It's, like, because I'm sure
that's a frustration of these

999
00:57:27,457 --> 00:57:32,900
companies. It's a frustration to
you, because you're tied to the

1000
00:57:32,900 --> 00:57:38,088
growth of the industry. How do
we get over that? Like, what's

1001
00:57:38,088 --> 00:57:42,000
the plateau buster for that
measured approach?

1002
00:57:42,000 --> 00:57:44,498
Sergey Litvinenko: So if you
have to realize that in our

1003
00:57:44,498 --> 00:57:47,219
report most of the companies
that we deal with they are going

1004
00:57:47,219 --> 00:57:49,896
after big enterprises like you
said. Those are typically the

1005
00:57:49,896 --> 00:57:52,572
enterprises that have the budget
to try something like this.

1006
00:57:52,572 --> 00:57:54,000
They're like the first adopters.

1007
00:57:54,000 --> 00:58:00,000
Audrow Nash: Mm-hmm.

1008
00:58:00,000 --> 00:58:04,233
Sergey Litvinenko: And just by
being the first adopter you just

1009
00:58:04,233 --> 00:58:08,264
don't know lots of things and
you give it a try and then you

1010
00:58:08,264 --> 00:58:12,363
make a decision and it takes
years. I think one way to escape

1011
00:58:12,363 --> 00:58:16,529
this is to try to build towards
the world that has a different

1012
00:58:16,529 --> 00:58:20,627
buyer. That would be an example
of that would be self-driving

1013
00:58:20,627 --> 00:58:24,726
cars. In the self-driving car
space your customer is a person

1014
00:58:24,726 --> 00:58:29,026
with a cell phone that needs to
go from point A to point B. It's

1015
00:58:29,026 --> 00:58:33,326
not an enterprise that's buying
a corporation that's buying from

1016
00:58:33,326 --> 00:58:37,357
you. So that's why I think the
world has already spoken that

1017
00:58:37,357 --> 00:58:41,657
they like self-driving cars over
human driven taxis or Ubers and

1018
00:58:41,657 --> 00:58:45,487
just a matter of supply to do
that and then the demand is

1019
00:58:45,487 --> 00:58:49,451
going to be there and all of a
sudden you know becomes very

1020
00:58:49,451 --> 00:58:53,617
very good business. A similar
thing could be with humanoids if

1021
00:58:53,617 --> 00:58:57,917
they go into let's say homes but
there we have to think about it

1022
00:58:57,917 --> 00:59:00,000
twice because the risk is high.

1023
00:59:00,000 --> 00:59:04,000
Audrow Nash: The risk is high.

1024
00:59:04,000 --> 00:59:07,565
Sergey Litvinenko: So we'll see.
I think we need to see what

1025
00:59:07,565 --> 00:59:11,189
Tesla is going to do. So that's
why you know as and why can't

1026
00:59:11,189 --> 00:59:14,933
you ship a fully able humanoid
to a human home today? Well it's

1027
00:59:14,933 --> 00:59:18,320
because it's not ready yet. It's
either the hardware, the

1028
00:59:18,320 --> 00:59:21,944
software, probably both that
just cannot do everything is too

1029
00:59:21,944 --> 00:59:25,628
heavy or maybe not too safe to
deploy. So it's a matter of the

1030
00:59:25,628 --> 00:59:29,193
technology of the product. So
take it a couple of iterations

1031
00:59:29,193 --> 00:59:33,055
on the product side then make it
affordable enough for humans and

1032
00:59:33,055 --> 00:59:36,917
then you're going to escape that
enterprise bottleneck but that's

1033
00:59:36,917 --> 00:59:40,304
going to be only a handful of
companies to begin with and

1034
00:59:40,304 --> 00:59:43,988
they're probably going to split
the market in the near future.

1035
00:59:43,988 --> 00:59:47,553
Everybody else will still have
to fight this enterprise game

1036
00:59:47,553 --> 00:59:50,821
and there's just no escaping.
It's just hoping that the

1037
00:59:50,821 --> 00:59:54,564
enterprise get better but also
just make sure that your product

1038
00:59:54,564 --> 00:59:57,000
is stellar so they cannot say no
to that.

1039
00:59:57,000 --> 00:59:58,000
Audrow Nash: Mm-hmm. Yeah, my
impression...

1040
00:59:58,000 --> 01:00:00,837
Sergey Litvinenko: It's a very
high level overview but that's

1041
01:00:00,837 --> 01:00:02,000
kind of the key takeaway.

1042
01:00:02,000 --> 01:00:06,185
Audrow Nash: Yeah, my impression
with, like, from talking to

1043
01:00:06,185 --> 01:00:10,231
polymath robotics, who do a lot
of these autonomous mobile

1044
01:00:10,231 --> 01:00:14,555
vehicles, you probably know them
with all of it, but they were

1045
01:00:14,555 --> 01:00:18,671
saying how... or one of my
takeaways from one of our recent

1046
01:00:18,671 --> 01:00:22,926
talks was, it's very hard to
build a robotic startup in a lot

1047
01:00:22,926 --> 01:00:27,180
of areas with big enterprise,
because you can make some money

1048
01:00:27,180 --> 01:00:31,366
solving their specific problem,
but it's probably not enough

1049
01:00:31,366 --> 01:00:35,272
money to justify, like, a
venture capital funded startup

1050
01:00:35,272 --> 01:00:39,457
for this kind of thing. Like,
the markets are kind of small,

1051
01:00:39,457 --> 01:00:43,572
and so my read from that was
kind of like, there's a lot of

1052
01:00:43,572 --> 01:00:46,850
opportunity, but they're
probably like internal

1053
01:00:46,850 --> 01:00:51,245
efficiency gains within larger
companies versus startups. Like,

1054
01:00:51,245 --> 01:00:55,221
it's hard to get into them. I
wonder, I mean, I guess you

1055
01:00:55,221 --> 01:00:59,336
could look at, like, robotic
arms as being the original get

1056
01:00:59,336 --> 01:01:03,591
into big companies robots, but I
haven't seen... and I guess,

1057
01:01:03,591 --> 01:01:07,985
like, KUKA, warehouse logistic
robots, is another good example,

1058
01:01:07,985 --> 01:01:12,170
but I don't know, it's a hard
road map to get into these big

1059
01:01:12,170 --> 01:01:16,286
companies, and it's expensive,
although, actually, you know

1060
01:01:16,286 --> 01:01:20,331
what might be very cool? AI,
helping robotics companies do

1061
01:01:20,331 --> 01:01:24,098
things for cheaper, so they
don't need as much venture

1062
01:01:24,098 --> 01:01:28,423
funding, so they don't have to
have such a large market after,

1063
01:01:28,423 --> 01:01:32,398
because I've seen so many
companies fail, like, that seem

1064
01:01:32,398 --> 01:01:36,374
very promising right when
they're getting to their Series

1065
01:01:36,374 --> 01:01:40,769
B, because they don't make sense
for Series B investors because

1066
01:01:40,769 --> 01:01:45,023
their addressable market is too
small, and I wonder if making

1067
01:01:45,023 --> 01:01:46,000
the cost lower

1068
01:01:46,000 --> 01:01:55,000
Sergey Litvinenko: Yeah. Yeah.
Yeah.

1069
01:01:55,000 --> 01:01:56,419
Audrow Nash: Could lead to a
heyday for robotics companies,

1070
01:01:56,419 --> 01:01:57,838
because you can, instead of
having a hundred million dollar

1071
01:01:57,838 --> 01:01:59,162
company, you can have a ten
million dollar company that

1072
01:01:59,162 --> 01:02:00,533
solves some real problems, this
kind of thing, or maybe a

1073
01:02:00,533 --> 01:02:02,000
hundred million instead of a
billion, say. What do you think?

1074
01:02:02,000 --> 01:02:07,977
Sergey Litvinenko: Mm-hmm. Yeah.
I definitely resonate with you

1075
01:02:07,977 --> 01:02:13,479
when you say that there might be
a misfit between robotics

1076
01:02:13,479 --> 01:02:18,887
companies, their cycles, and
what venture capitalists are

1077
01:02:18,887 --> 01:02:24,485
seeking. 100%, you know, the if
I I've never been a venture

1078
01:02:24,485 --> 01:02:30,651
capitalist, but my understanding
is that you have like five years

1079
01:02:30,651 --> 01:02:36,059
to get as close as you can to
100 million, maybe now it's

1080
01:02:36,059 --> 01:02:39,000
billion in revenue or valuation

1081
01:02:39,000 --> 01:02:41,000
Audrow Nash: Yeah, probably.

1082
01:02:41,000 --> 01:02:42,000
Sergey Litvinenko: And just, you
know,

1083
01:02:41,000 --> 01:02:42,172
Audrow Nash: Hey, you want a
ten, you want a hundred X, at

1084
01:02:42,172 --> 01:02:43,000
least, this kind of thing, in
five years.

1085
01:02:43,000 --> 01:02:47,470
Sergey Litvinenko: Yeah, and
then exit. You know, that's, I

1086
01:02:47,470 --> 01:02:48,000
mean, y

1087
01:02:48,000 --> 01:02:50,000
Audrow Nash: Depending on the
stage of investing, but yeah.

1088
01:02:50,000 --> 01:02:53,163
Sergey Litvinenko: That's, you
know, by definition, those have

1089
01:02:53,163 --> 01:02:55,000
to be like viral products,
unicorns.

1090
01:02:55,000 --> 01:02:58,000
Audrow Nash: Unicorns.

1091
01:02:58,000 --> 01:03:00,011
Sergey Litvinenko: In the
physical world, it's a little

1092
01:03:00,011 --> 01:03:02,314
bit different, but at the same
time, you could see companies in

1093
01:03:02,314 --> 01:03:04,544
the defense space like Field AI,
Shield AI getting like crazy

1094
01:03:04,544 --> 01:03:06,628
valuations. You see robotics
companies like Standard Bots

1095
01:03:06,628 --> 01:03:08,639
recently raised a lot of money,
became a unicorn, Gecko

1096
01:03:08,639 --> 01:03:10,796
Robotics, and those companies
were at it for a while. And I

1097
01:03:10,796 --> 01:03:13,099
think maybe the robotics story
is different where you grind for

1098
01:03:13,099 --> 01:03:15,475
five to seven years, you're just
trying to figure out what clicks

1099
01:03:15,475 --> 01:03:17,815
and then when it clicks, that's
when you get to venture. But the

1100
01:03:17,815 --> 01:03:19,753
question becomes, how do I
survive this five to seven

1101
01:03:19,753 --> 01:03:21,690
years? Yeah. Do you have
bootstrap or, you know, do I

1102
01:03:21,690 --> 01:03:24,030
have some other type of funding?
I honestly don't know. I got to

1103
01:03:24,030 --> 01:03:26,297
tell you, all of the customers
that we work with, they're like

1104
01:03:26,297 --> 01:03:28,381
very early stage bootstrapped,
but everybody else venture

1105
01:03:28,381 --> 01:03:30,684
funded. So right now, it seems
like venture is the only way for

1106
01:03:30,684 --> 01:03:32,000
robotics. Nobody else will touch
it.

1107
01:03:32,000 --> 01:03:37,128
Audrow Nash: Yeah, but that's
hard to bootstrap. Yeah, that's

1108
01:03:37,128 --> 01:03:42,593
the awkward space. Yeah, it's an
interesting thing. I don't know,

1109
01:03:42,593 --> 01:03:47,889
I don't have a good solution to
this either. It's like take out

1110
01:03:47,889 --> 01:03:53,270
a HELOC or whatever the loan is
on your house to fund stuff, but

1111
01:03:53,270 --> 01:03:58,735
it's obviously not the good way.
Ideally, if you can do something

1112
01:03:58,735 --> 01:04:03,695
revenue producing early and
avoid investment is kind of the

1113
01:04:03,695 --> 01:04:07,898
thing, but robotics is
expensive, manufacturing is

1114
01:04:07,898 --> 01:04:13,111
expensive, volume drives costs
down. My thought is once we get

1115
01:04:13,111 --> 01:04:17,819
to standardized platforms for
things, we'll start to see

1116
01:04:17,819 --> 01:04:22,863
change. Like, I was imagining if
you have like a really good

1117
01:04:22,863 --> 01:04:27,403
humanoid supplier for a
reasonable price, you get your

1118
01:04:27,403 --> 01:04:32,447
humanoid robots to go do some
service like cleaning homes or

1119
01:04:32,447 --> 01:04:37,239
machine, I don't know, like
doing anything, any task that

1120
01:04:37,239 --> 01:04:42,199
needs to be done, but you could
focus on just that specific

1121
01:04:42,199 --> 01:04:47,243
task, but you are not investing
as much in the manufacturing

1122
01:04:47,243 --> 01:04:52,708
because you can buy it almost as
a service. That to me seems like

1123
01:04:52,708 --> 01:04:57,500
an exciting future robotics
business model, but we're not

1124
01:04:57,500 --> 01:05:02,460
there yet for most platforms,
maybe some for logistics, but

1125
01:05:02,460 --> 01:05:07,000
that's already such a comp
competitive space, I think.

1126
01:05:07,000 --> 01:05:11,254
Sergey Litvinenko: Well, Audrow,
you definitely have some

1127
01:05:11,254 --> 01:05:15,657
foresight because probably in
the last, I call it like two,

1128
01:05:15,657 --> 01:05:20,433
let's call it three weeks, we've
seen a dozen of humanoid rental

1129
01:05:20,433 --> 01:05:22,000
companies come to us.

1130
01:05:22,000 --> 01:05:23,000
Audrow Nash: Cool. Super

1131
01:05:23,000 --> 01:05:26,444
Sergey Litvinenko: Doing exactly
what you said, they have a

1132
01:05:26,444 --> 01:05:29,770
little bit of money to get a
humanoid and then they start

1133
01:05:29,770 --> 01:05:33,389
performing the service with that
humanoid right away. Manually

1134
01:05:33,389 --> 01:05:36,774
operate it and then they make
some money and then they use

1135
01:05:36,774 --> 01:05:38,000
that to continue R&D.

1136
01:05:38,000 --> 01:05:41,498
Audrow Nash: Cool. What a thing.
Yeah, I think that'll be more

1137
01:05:41,498 --> 01:05:44,940
and more popular, and humanoid
might be the good platform for

1138
01:05:44,940 --> 01:05:48,213
that, but there might be other
ones, but it's hard to make

1139
01:05:48,213 --> 01:05:51,655
something general enough that
they can produce it in bulk for

1140
01:05:51,655 --> 01:05:54,928
cheap. And so, I don't know,
what's your thought on how it

1141
01:05:54,928 --> 01:05:56,000
proceeds from here?

1142
01:05:56,000 --> 01:05:59,957
Sergey Litvinenko: There are two
ideas which I think robotics

1143
01:05:59,957 --> 01:06:03,914
founders should explore. First,
government funding, there are

1144
01:06:03,914 --> 01:06:07,481
grants out there that you can
tap into. They're kind of

1145
01:06:07,481 --> 01:06:08,000
painful.

1146
01:06:08,000 --> 01:06:10,182
Audrow Nash: I think it heavily
depends on the area, right? Like

1147
01:06:10,182 --> 01:06:11,000
defense, yeah, for sure.

1148
01:06:11,000 --> 01:06:13,000
Sergey Litvinenko: That's true.
That's true, yeah.

1149
01:06:13,000 --> 01:06:13,871
Audrow Nash: Maybe construction,
maybe things like But that. Or

1150
01:06:13,871 --> 01:06:14,728
but what do you think? Because
that would be heavily dependent

1151
01:06:14,728 --> 01:06:15,585
on the area you're going for,
and that's a very different type

1152
01:06:15,585 --> 01:06:16,000
of startup in my mind. Mm-hmm.

1153
01:06:16,000 --> 01:06:20,019
Sergey Litvinenko: Yeah. If
you're able, meaning that if

1154
01:06:20,019 --> 01:06:24,253
you're within DoW or working
with any other agencies, there

1155
01:06:24,253 --> 01:06:28,630
are opportunities out there but
you have to figure it out. If

1156
01:06:28,630 --> 01:06:32,793
you're not doing government,
then you're not going to have

1157
01:06:32,793 --> 01:06:37,314
access to it. Another way which
I think to be able to pull this

1158
01:06:37,314 --> 01:06:41,619
off, you have to be incredible
at enterprise is that you can

1159
01:06:41,619 --> 01:06:46,284
find a customer, and I think you
alluded to this earlier, and you

1160
01:06:46,284 --> 01:06:50,303
can excite them so much that
they're going to prepay the

1161
01:06:50,303 --> 01:06:54,537
contract. And You can use that
to fund your development and

1162
01:06:54,537 --> 01:06:58,412
then ship a pilot to them. It's
like knowing how those

1163
01:06:58,412 --> 01:07:02,000
enterprises operate, it has to
be like either some

1164
01:07:02,000 --> 01:07:09,000
Audrow Nash: Hell yeah.

1165
01:07:09,000 --> 01:07:11,992
Sergey Litvinenko: Insane
relationship with the customer

1166
01:07:11,992 --> 01:07:14,985
or some insane ability to
convince them to buy something

1167
01:07:14,985 --> 01:07:16,000
that doesn't exist.

1168
01:07:16,000 --> 01:07:19,339
Audrow Nash: I think that's
where internal startups are

1169
01:07:19,339 --> 01:07:23,043
really good, but the equity and
stuff, it's vastly different,

1170
01:07:23,043 --> 01:07:26,443
but it's like if you're at
Caterpillar or something, and

1171
01:07:26,443 --> 01:07:29,843
you propose an initiative, and
they fund it for a while,

1172
01:07:29,843 --> 01:07:33,061
because they do big R&D
projects, but that's possibly

1173
01:07:33,061 --> 01:07:36,886
successful, in my opinion, for
this kind of thing. But it's not

1174
01:07:36,886 --> 01:07:40,286
the same appeal. Like you don't
get the equity of like a

1175
01:07:40,286 --> 01:07:43,868
startup, I think, for a lot of
these internal things. So it

1176
01:07:43,868 --> 01:07:47,754
doesn't attract the same type of
people, I think, unfortunately.

1177
01:07:47,754 --> 01:07:50,000
But maybe companies will be
flexible.

1178
01:07:50,000 --> 01:07:52,000
Sergey Litvinenko: I've seen
zero examples like this.

1179
01:07:52,000 --> 01:07:55,460
Audrow Nash: Yeah, that's what
I'm saying. Because the

1180
01:07:55,460 --> 01:07:59,112
population who's doing the
startups just goes and does it

1181
01:07:59,112 --> 01:08:02,444
because they're not at
Caterpillar. And I don't know

1182
01:08:02,444 --> 01:08:06,224
Caterpillar specifically, but
I'm just thinking big company

1183
01:08:06,224 --> 01:08:09,812
kind of thing. Yeah, I've seen
zero of those two. But in

1184
01:08:09,812 --> 01:08:13,721
theory, it seems like it might
work, but I don't know. But it

1185
01:08:13,721 --> 01:08:17,693
also wouldn't be the same kind
of rewarding inequity of like a

1186
01:08:17,693 --> 01:08:20,000
regular startup, this kind of
thing.

1187
01:08:20,000 --> 01:08:21,538
Sergey Litvinenko: You're not
getting 100% of equity in that

1188
01:08:21,538 --> 01:08:22,000
case, most likely.

1189
01:08:22,000 --> 01:08:26,094
Audrow Nash: No, you're just
keeping your W2 kind of thing

1190
01:08:26,094 --> 01:08:30,259
and running an internal startup.
That's really interesting,

1191
01:08:30,259 --> 01:08:34,424
though. What do you think? So
what's your take on humanoids

1192
01:08:34,424 --> 01:08:38,800
with it? You were saying, okay,
there are demos so far. You're

1193
01:08:38,800 --> 01:08:42,894
starting to see companies that
are trying to use them as a

1194
01:08:42,894 --> 01:08:47,129
service, which is really cool.
And I think we'll see more of

1195
01:08:47,129 --> 01:08:51,576
that. What's your thoughts like
a few years out in this kind of

1196
01:08:51,576 --> 01:08:52,000
thing?

1197
01:08:52,000 --> 01:08:55,957
Sergey Litvinenko: Well, I
follow lots of robotics people

1198
01:08:55,957 --> 01:09:00,052
on X and there is a debate going
on about the humanoid form

1199
01:09:00,052 --> 01:09:04,148
factor. It makes sense and
sometimes it doesn't make sense.

1200
01:09:04,148 --> 01:09:08,035
It's all over the place. So I
really don't know how well

1201
01:09:08,035 --> 01:09:12,547
humanoid is going to be received
or how that form factor is going

1202
01:09:12,547 --> 01:09:16,642
to be, how successful it's going
to be. And we honestly are

1203
01:09:16,642 --> 01:09:20,876
agnostic to it. If humanoids
succeed, we'll happily work with

1204
01:09:20,876 --> 01:09:24,000
it. We already have humanoid
customers. Okay?

1205
01:09:24,000 --> 01:09:26,000
Audrow Nash: You take whatever
succeeding. It doesn't matter.

1206
01:09:26,000 --> 01:09:30,846
Sergey Litvinenko: Exactly. So
we are agnostic, but I would say

1207
01:09:30,846 --> 01:09:35,385
I think we'll have to see how
big companies like Tesla pave

1208
01:09:35,385 --> 01:09:40,308
the way. And if Tesla is able to
set an example of how it can be

1209
01:09:40,308 --> 01:09:45,000
done and they have massive R&D
budget, then hell yeah, it can

1210
01:09:45,000 --> 01:09:49,538
be done. I heard some rumors
that they are, I think they're

1211
01:09:49,538 --> 01:09:54,462
on track to start deploying them
as early as 2027. Somebody even

1212
01:09:54,462 --> 01:09:56,000
said like late 2026.

1213
01:09:56,000 --> 01:09:59,000
Audrow Nash: I'm not sure.

1214
01:09:59,000 --> 01:10:01,953
Sergey Litvinenko: So I'm hoping
we'll be able to see something

1215
01:10:01,953 --> 01:10:04,625
from them and they'll pave the
way. But I personally have

1216
01:10:04,625 --> 01:10:07,438
nothing against the humanoid
form factor. I think it's going

1217
01:10:07,438 --> 01:10:08,000
to be great.

1218
01:10:08,000 --> 01:10:09,000
Audrow Nash: Me neither.

1219
01:10:09,000 --> 01:10:13,000
Sergey Litvinenko: As long as
it's safe and doesn't kill me,

1220
01:10:13,000 --> 01:10:14,000
Audrow Nash: Yeah, those are my
requirements, too. Yeah.

1221
01:10:13,000 --> 01:10:16,486
Sergey Litvinenko: Which goes
back to liability, which goes

1222
01:10:16,486 --> 01:10:19,736
back to liability. In all
seriousness, a very practical

1223
01:10:19,736 --> 01:10:23,577
example would be, sure, I have a
humanoid in the house and then I

1224
01:10:23,577 --> 01:10:26,768
have a child. And if that
humanoid slips and trips and

1225
01:10:26,768 --> 01:10:30,314
falls on a child, it's going to
be a big problem. So even in

1226
01:10:30,314 --> 01:10:33,918
situations like that, it's like
in self-driving cars, you can

1227
01:10:33,918 --> 01:10:37,286
get to 99.9% reliability, but
the remaining nines is what

1228
01:10:37,286 --> 01:10:39,000
matters. Same with humanoids.

1229
01:10:39,000 --> 01:10:43,136
Audrow Nash: Yeah. What do you
think? So from your extremely

1230
01:10:43,136 --> 01:10:46,927
amazing position being an
insurance, where do you think

1231
01:10:46,927 --> 01:10:51,063
you'll see humanoids roll out
first? Like I suspect the home

1232
01:10:51,063 --> 01:10:55,336
is going to take a while just
due to safety things. That to me

1233
01:10:55,336 --> 01:10:59,679
seems likely unfortunate because
I would love to have them tidy

1234
01:10:59,679 --> 01:11:04,021
up. It's been real hard to find
a cleaner, but it would be just

1235
01:11:04,021 --> 01:11:08,226
wonderful. But I think the home
will be a difficult spot just

1236
01:11:08,226 --> 01:11:12,637
due to safety reasons and things
like this. But maybe commercial

1237
01:11:12,637 --> 01:11:16,428
places, maybe back, like behind
in like a, say, if they

1238
01:11:16,428 --> 01:11:20,702
waterproofed them. It's like a
kitchen of a restaurant or just

1239
01:11:20,702 --> 01:11:24,907
running dishes or like where do
you think that they'll emerge

1240
01:11:24,907 --> 01:11:28,698
for first? Because it's obvious
they'll be used in like

1241
01:11:28,698 --> 01:11:32,971
logistics as they already are
with agility and they'll be used

1242
01:11:32,971 --> 01:11:37,107
in manufacturing to some level.
But where do you think we'll

1243
01:11:37,107 --> 01:11:40,967
start to see them in public
domain, say, or like regular

1244
01:11:40,967 --> 01:11:44,000
life outside of manufacturing
and logistics?

1245
01:11:44,000 --> 01:11:47,947
Sergey Litvinenko: Yeah. So,
robot underwriter speaking, it's

1246
01:11:47,947 --> 01:11:51,894
manufacturing logistics, that's
it. We, everything that we've

1247
01:11:51,894 --> 01:11:55,647
seen so far, that's those two
use cases, we have seen some

1248
01:11:55,647 --> 01:11:59,724
pilots come through for home. We
did not do them because it was

1249
01:11:59,724 --> 01:12:03,541
just, it was just, didn't make
sense for us. It was just no

1250
01:12:03,541 --> 01:12:07,747
data, bad environments, startups
with no certifications. Sorry, I

1251
01:12:07,747 --> 01:12:11,759
can't do it. It's too much. So,
yeah, commercial use case. So,

1252
01:12:11,759 --> 01:12:15,253
we are going back to that
enterprise bottleneck, where

1253
01:12:15,253 --> 01:12:19,071
they're going to get those
humanoids, they're going to test

1254
01:12:19,071 --> 01:12:23,018
them, they're going to think
about it, and then come back for

1255
01:12:23,018 --> 01:12:26,641
another pilot, maybe with
another vendor. It is going to

1256
01:12:26,641 --> 01:12:28,000
take a while. Mm-hmm.

1257
01:12:28,000 --> 01:12:32,371
Audrow Nash: Mm-hmm. Do you
think, like, internal logistics,

1258
01:12:32,371 --> 01:12:36,597
maybe logistics moves from
warehouses to, like, my grocery

1259
01:12:36,597 --> 01:12:40,823
store stocks itself at night
kind of thing with humanoids?

1260
01:12:40,823 --> 01:12:45,340
And I could see the form factor
being a little bit better for,

1261
01:12:45,340 --> 01:12:49,711
well, I guess the floors are
flat and I don't know, I'm just

1262
01:12:49,711 --> 01:12:53,864
wondering. Agility made a
pragmatic argument for why they

1263
01:12:53,864 --> 01:12:58,381
use a humanoid for moving totes.
But I wonder if that argument

1264
01:12:58,381 --> 01:13:02,606
would hold for, like, grocery
stores or anything like this

1265
01:13:02,606 --> 01:13:07,196
where it's, like, stocking small
items. I don't know. My belief

1266
01:13:07,196 --> 01:13:11,786
is it's probably a lot easier to
just use a base, like a mobile

1267
01:13:11,786 --> 01:13:16,012
base with an arm for a lot of
that related things. But you

1268
01:13:16,012 --> 01:13:19,727
also might benefit from
decreased cost if humanoids

1269
01:13:19,727 --> 01:13:24,026
start scaling from this kind of
thing. But I don't know. It

1270
01:13:24,026 --> 01:13:28,324
might be very far away. I'm
just, I'm trying to think where

1271
01:13:28,324 --> 01:13:30,000
we might see him first.

1272
01:13:30,000 --> 01:13:31,000
Sergey Litvinenko: Yeah, yeah.
possi

1273
01:13:31,000 --> 01:13:34,000
Audrow Nash: Maybe Defense is
the next big driver.

1274
01:13:34,000 --> 01:13:37,361
Sergey Litvinenko: Actually,
possible defense possible, but

1275
01:13:37,361 --> 01:13:40,949
again, in defense, we do have a
good number of customers there.

1276
01:13:40,949 --> 01:13:43,000
It's like the warfare is
changing to

1277
01:13:43,000 --> 01:13:45,000
Audrow Nash: Yep. Jones.

1278
01:13:45,000 --> 01:13:48,109
Sergey Litvinenko: Non-combat
operations, where you don't have

1279
01:13:48,109 --> 01:13:51,167
people fighting, now you have
machines over the air fighting,

1280
01:13:51,167 --> 01:13:54,276
or cyber warfare. So, it's like
hybrid war. So, I mean, if you

1281
01:13:54,276 --> 01:13:57,234
got humanoids going against
other countries' humanoids, and

1282
01:13:57,234 --> 01:14:00,393
it's just like the robot fight,
it'll be pretty entertaining to

1283
01:14:00,393 --> 01:14:03,000
watch, as long as humans are
intact, don't get hurt.

1284
01:14:03,000 --> 01:14:06,896
Audrow Nash: It's a wild thing.
And I'm know, not sure. you

1285
01:14:06,896 --> 01:14:10,000
Yeah. Just who with the biggest
factories wins.

1286
01:14:10,000 --> 01:14:13,871
Sergey Litvinenko: But we'll
see. I actually, yeah, I haven't

1287
01:14:13,871 --> 01:14:17,678
seen a humanoid defense use case
at all. Only factories. But

1288
01:14:17,678 --> 01:14:21,168
look, in a limit, if you make
some assumptions that the

1289
01:14:21,168 --> 01:14:25,102
hardware is cheap, the software
is amazing, and the robots can

1290
01:14:25,102 --> 01:14:28,972
do everything the humans can,
then yeah, sure, humanoids will

1291
01:14:28,972 --> 01:14:32,970
take over the whole thing. But I
think we need to get there yet

1292
01:14:32,970 --> 01:14:35,000
first. It's a technical problem.

1293
01:14:35,000 --> 01:14:39,557
Audrow Nash: It's a long way.
Yeah, yeah. And safety and all

1294
01:14:39,557 --> 01:14:43,658
that. Okay. One thing you
mentioned that I thought was

1295
01:14:43,658 --> 01:14:47,911
really interesting was about
cyber. Like, tell me about,

1296
01:14:47,911 --> 01:14:51,481
because, like, what I'm
thinking, you mentioned

1297
01:14:51,481 --> 01:14:55,962
cybersecurity for robotics
companies. and What I'm thinking

1298
01:14:55,962 --> 01:15:00,443
is, like, all of the incredible
vulnerabilities released in

1299
01:15:00,443 --> 01:15:05,000
October, this kind of thing.
Like what a month that was for.

1300
01:15:05,000 --> 01:15:08,873
It's just like, oh, the
cybersecurity people are so

1301
01:15:08,873 --> 01:15:13,127
stressed and everything is
crazy. So tell me a bit about

1302
01:15:13,127 --> 01:15:17,911
what you're thinking about cyber
security in general and how do

1303
01:15:17,911 --> 01:15:21,861
you ensure that. And I don't
know. Tell me what your

1304
01:15:21,861 --> 01:15:23,000
perspective is.

1305
01:15:23,000 --> 01:15:27,206
Sergey Litvinenko: So for
robotics companies,

1306
01:15:27,206 --> 01:15:33,000
unfortunately, there is nothing
special about them being Yeah.

1307
01:15:33,000 --> 01:15:38,000
Audrow Nash: It's still just
software.

1308
01:15:38,000 --> 01:15:41,279
Sergey Litvinenko: And it's
through social engineering,

1309
01:15:41,279 --> 01:15:45,035
through nothing like no physical
tampering that we experienced,

1310
01:15:45,035 --> 01:15:47,896
but definitely social
engineering is the biggest

1311
01:15:47,896 --> 01:15:51,413
thing, and some ransomware
attached to it. You just, people

1312
01:15:51,413 --> 01:15:54,812
get access to the internal
systems, and then demand money

1313
01:15:54,812 --> 01:15:58,210
You just have to properly
protect yourself. Like cyber is

1314
01:15:58,210 --> 01:16:01,965
essential. I believe that you,
other than using good tools, you

1315
01:16:01,965 --> 01:16:05,602
know, and having your employees
receive proper training, just

1316
01:16:05,602 --> 01:16:09,119
invest a little bit of money
into just protecting yourself.

1317
01:16:09,119 --> 01:16:12,577
Get like a certification, it
could be like SOC, ISO, maybe

1318
01:16:12,577 --> 01:16:15,975
something else. It does reduce
the probability of getting

1319
01:16:15,975 --> 01:16:19,433
hacked. Like the data shows
that. Obviously, if you're, if

1320
01:16:19,433 --> 01:16:22,652
you have, let's say, a
certification, and you're still

1321
01:16:22,652 --> 01:16:26,467
reckless as a company, it's not
going to help, but if you have a

1322
01:16:26,467 --> 01:16:29,925
certification, and you actually
follow the policy that you

1323
01:16:29,925 --> 01:16:33,085
install, it does reduce the
chance of getting hacked.

1324
01:16:33,085 --> 01:16:36,602
Specifically, robotics, we
actually think, so we're working

1325
01:16:36,602 --> 01:16:40,179
on one project, which we'll
probably release next year. It's

1326
01:16:40,179 --> 01:16:43,935
going to be exciting. It's going
to be like attached to Koop, a

1327
01:16:43,935 --> 01:16:47,512
separate initiative that we
have, which will be testing lots

1328
01:16:47,512 --> 01:16:51,208
of things in the physical space
when it comes to robots. And I

1329
01:16:51,208 --> 01:16:54,844
do think that we have yet to see
a massive cyber, robot cyber

1330
01:16:54,844 --> 01:16:58,421
event, primarily because there
are not that many robots, but

1331
01:16:58,421 --> 01:17:01,879
maybe a couple of years ago,
people were hacking Teslas, I

1332
01:17:01,879 --> 01:17:05,635
believe, you know. And if Tesla
got hacked, you know, you could

1333
01:17:05,635 --> 01:17:09,033
hack lots of things, because
Tesla is pretty good when it

1334
01:17:09,033 --> 01:17:11,000
comes to cyber, as far as I
know.

1335
01:17:11,000 --> 01:17:27,000
Audrow Nash: Mm-hmm. That's
horrifying. Mm-hmm.

1336
01:17:27,000 --> 01:17:30,312
Sergey Litvinenko: So, delivery
robots, self-driving cars,

1337
01:17:30,312 --> 01:17:33,338
drones, you need to make sure
that the, it's properly

1338
01:17:33,338 --> 01:17:36,764
protected. Thankfully, you know,
since it's not going to be,

1339
01:17:36,764 --> 01:17:40,019
since they're going to be
autonomous and not connected to

1340
01:17:40,019 --> 01:17:43,387
any places that you can take
control over it remotely, that

1341
01:17:43,387 --> 01:17:46,756
helps, but at the same time,
maybe you can do some physical

1342
01:17:46,756 --> 01:17:50,296
interference with the vehicle
and get access to it somehow. We

1343
01:17:50,296 --> 01:17:53,380
are learning as we go, but we
already do provide cyber

1344
01:17:53,380 --> 01:17:56,691
coverage that does extend to
those robots, meaning that it

1345
01:17:56,691 --> 01:18:00,231
just, that area is not clearly
defined yet in terms of how the

1346
01:18:00,231 --> 01:18:03,657
claims are going to look like.
But we do write that coverage

1347
01:18:03,657 --> 01:18:04,000
today.

1348
01:18:04,000 --> 01:18:08,061
Audrow Nash: Huh. Yeah.
Interesting. What do you think

1349
01:18:08,061 --> 01:18:12,499
the space is going to look like?
So, you were saying, we're

1350
01:18:12,499 --> 01:18:17,086
hitting a tipping point. There
will be lots of--I guess Tesla

1351
01:18:17,086 --> 01:18:21,674
will probably do a good job with
its cyber security, with the

1352
01:18:21,674 --> 01:18:26,337
Cybercabs. I don't know. But do
you think we're going to have,

1353
01:18:26,337 --> 01:18:30,850
like, a--I don't know, like a
crazy October, bugs, whatever,

1354
01:18:30,850 --> 01:18:35,362
where we're hacking physical
systems? Do you think, like, in

1355
01:18:35,362 --> 01:18:39,950
the next five, ten years, we're
going to see, like, one, that

1356
01:18:39,950 --> 01:18:44,387
there's a lot of robots, and
two, that all of a sudden they

1357
01:18:44,387 --> 01:18:48,975
get hacked, and who knows what
that would look like? But what

1358
01:18:48,975 --> 01:18:53,187
do you think? What are you
preparing for, thinking about

1359
01:18:53,187 --> 01:18:57,549
with this? Or is it just, like,
we'll see? We'll see as it

1360
01:18:57,549 --> 01:18:58,000
comes.

1361
01:18:58,000 --> 01:19:02,126
Sergey Litvinenko: Yeah. I think
most people will take the

1362
01:19:02,126 --> 01:19:06,751
approach, will see, because they
just don't want to deal with it.

1363
01:19:06,751 --> 01:19:11,304
There needs to be a company like
Koop or somebody like Koop that

1364
01:19:11,304 --> 01:19:15,431
will put all those robots to
test and try to actually hack

1365
01:19:15,431 --> 01:19:16,000
them for

1366
01:19:16,000 --> 01:19:17,000
Audrow Nash: Ah.

1367
01:19:17,000 --> 01:19:20,000
Sergey Litvinenko: Safety
purposes. to the

1368
01:19:20,000 --> 01:19:21,524
Audrow Nash: I'm excited to see
about your future initiative. Is

1369
01:19:21,524 --> 01:19:22,000
it in any--it--yeah,

1370
01:19:22,000 --> 01:19:25,672
Sergey Litvinenko: Yeah. To
develop some benchmarks, you

1371
01:19:25,672 --> 01:19:26,000
know.

1372
01:19:26,000 --> 01:19:26,373
Audrow Nash: Yeah. Okay,
awesome. I would probably--I

1373
01:19:26,373 --> 01:19:26,796
would love to talk again when
that initiative gets launched,

1374
01:19:26,796 --> 01:19:27,000
because it sounds quite cool.

1375
01:19:27,000 --> 01:19:31,066
Sergey Litvinenko: It'll be
pretty cool. It'll be pretty

1376
01:19:31,066 --> 01:19:34,696
cool. Yeah, yeah, for sure.
Yeah, yeah, well, I'll

1377
01:19:34,696 --> 01:19:38,761
definitely reach back to you
when it's live. So I think,

1378
01:19:38,761 --> 01:19:42,972
again, maybe not in the next
five years, but once we start

1379
01:19:42,972 --> 01:19:47,400
reaching scale, that's going to
be self-driving cars, robots,

1380
01:19:47,400 --> 01:19:51,611
deliveries, and eventually
humanoids. There is going to be

1381
01:19:51,611 --> 01:19:56,040
a cyber event with robots, 100
percent. It's going to happen.

1382
01:19:56,040 --> 01:20:00,396
And the reason why it's going to
happen is that because it's

1383
01:20:00,396 --> 01:20:04,461
going to be new so people
wouldn't really even know what

1384
01:20:04,461 --> 01:20:08,672
to protect against and then
somebody would hack some robot

1385
01:20:08,672 --> 01:20:13,028
and do something with it. 100
percent. So 100 percent, sure.

1386
01:20:13,028 --> 01:20:17,602
It just, you know, hopefully it
will be contained and hopefully

1387
01:20:17,602 --> 01:20:22,030
it's going to happen, you know,
with less exposure, not a car

1388
01:20:22,030 --> 01:20:25,806
going 50 miles an hour on a
highway. But yeah, yeah,

1389
01:20:25,806 --> 01:20:29,000
something like that will 100
percent happen.

1390
01:20:29,000 --> 01:20:32,444
Audrow Nash: Mm. Yeah, what a
world we're going to be in. It's

1391
01:20:32,444 --> 01:20:35,833
just so interesting, all this
technology advancing, and where

1392
01:20:35,833 --> 01:20:38,833
it goes, and the new risks, and
everything. Like, it's

1393
01:20:38,833 --> 01:20:41,778
both--it's a sharp sword, I
guess. There's, like, the

1394
01:20:41,778 --> 01:20:45,056
amazing with it, and there's
also the risks associated with

1395
01:20:45,056 --> 01:20:48,389
it, which, I guess, is the whole
point of good insurance for

1396
01:20:48,389 --> 01:20:49,000
everything.

1397
01:20:49,000 --> 01:20:52,815
Sergey Litvinenko: Yeah, yeah,
yeah, yeah, exactly. That's why

1398
01:20:52,815 --> 01:20:56,446
you buy it and that's why it
helps you just, you know, feel

1399
01:20:56,446 --> 01:20:57,000
at peace.

1400
01:20:57,000 --> 01:21:01,560
Audrow Nash: Uh-huh. Yeah.
With--so, moving on a little bit

1401
01:21:01,560 --> 01:21:05,811
from cyber stuff, or somewhat
related, you mentioned AI

1402
01:21:05,811 --> 01:21:10,293
agents. Are you going to be
getting into the insurance, or

1403
01:21:10,293 --> 01:21:14,853
have you already gotten into
insuring AI agents for things,

1404
01:21:14,853 --> 01:21:19,877
or how do you think that'll look
from an insurance point of view?

1405
01:21:19,877 --> 01:21:24,205
Like, will it be individuals
insuring themselves against

1406
01:21:24,205 --> 01:21:29,151
agents doing something? Like, I
use Grok for my finances, and it

1407
01:21:29,151 --> 01:21:33,711
sells my house for a bag of
chips, or something ridiculous.

1408
01:21:33,711 --> 01:21:38,503
Like, do I get insurance for,
like, ridiculous events where it

1409
01:21:38,503 --> 01:21:42,986
behaves weird, or what do you
imagine insurance might look

1410
01:21:42,986 --> 01:21:46,000
like in this--for AI agents, I
suppose?

1411
01:21:46,000 --> 01:21:50,216
Sergey Litvinenko: Haven't
entered the AI agents and

1412
01:21:50,216 --> 01:21:54,838
services industry with our
insurance program yet. We have

1413
01:21:54,838 --> 01:21:58,730
three verticals. We have
robotics and autonomous

1414
01:21:58,730 --> 01:22:03,919
vehicles. We have a defense. For
those two, we do both insurance

1415
01:22:03,919 --> 01:22:08,784
and compliance. It comes as a
package, okay? And then for AI

1416
01:22:08,784 --> 01:22:13,649
agents and services, right now
we have a compliance offering

1417
01:22:13,649 --> 01:22:18,514
because those companies need
that, but insurance is still in

1418
01:22:18,514 --> 01:22:23,703
the development. It's, you know,
it's not that bad when it comes

1419
01:22:23,703 --> 01:22:28,243
to insuring AI agent mistakes. I
actually think that the

1420
01:22:28,243 --> 01:22:33,108
opportunity there is probably
not as big as people think, in

1421
01:22:33,108 --> 01:22:34,000
my opinion.

1422
01:22:34,000 --> 01:22:37,106
Audrow Nash: Mm. Just, they're
quite reliable in general, and

1423
01:22:37,106 --> 01:22:40,264
so they're not likely to do-- I
mean, maybe it'll, like, email

1424
01:22:40,264 --> 01:22:43,421
everyone something, that'll be
kind of embarrassing, but it is

1425
01:22:43,421 --> 01:22:45,000
not life-destroying, and thus--

1426
01:22:45,000 --> 01:22:47,813
Sergey Litvinenko: It's not a
claim of – yeah, it's not like

1427
01:22:47,813 --> 01:22:48,000
an –

1428
01:22:48,000 --> 01:22:49,000
Audrow Nash: Yeah, it's not
claimable.

1429
01:22:49,000 --> 01:22:53,614
Sergey Litvinenko: Exactly so,
sure. Let's say my agent makes a

1430
01:22:53,614 --> 01:22:57,641
mistake and unless it's a –
unless you have like a real

1431
01:22:57,641 --> 01:23:02,109
financial loss attached to it,
it's going to be hard to claim

1432
01:23:02,109 --> 01:23:06,576
against that. So, okay, I email
wrong people and maybe I tell

1433
01:23:06,576 --> 01:23:10,823
them something wrong. As long as
I leak somebody's private

1434
01:23:10,823 --> 01:23:15,437
information that I get sued for,
you know, I think I'm going to

1435
01:23:15,437 --> 01:23:18,000
be fine. I'm more concerned
about –

1436
01:23:18,000 --> 01:23:19,445
Audrow Nash: Yeah, because it
has to be with multiple

1437
01:23:19,445 --> 01:23:21,000
entities. That's the point of
insurance for this. Mm-hmm.

1438
01:23:21,000 --> 01:23:25,430
Sergey Litvinenko: Yeah. I'm
more concerned about – there are

1439
01:23:25,430 --> 01:23:29,860
a couple of things I'm concerned
about. First of all, you get

1440
01:23:29,860 --> 01:23:34,000
hacked and then somebody uses
your agent to do something.

1441
01:23:34,000 --> 01:23:35,000
Audrow Nash: Oh. Yeah.

1442
01:23:35,000 --> 01:23:36,249
Sergey Litvinenko: That's the
first thing. And the second

1443
01:23:36,249 --> 01:23:37,389
thing, you use powerful AI
models to just, you know,

1444
01:23:37,389 --> 01:23:38,682
penetrate different systems. And
that's less of an AI agent

1445
01:23:38,682 --> 01:23:39,932
insurance. That's more of a
cyber insurance. So, I'm more

1446
01:23:39,932 --> 01:23:41,225
interested how cyber is going to
evolve with AI agents than

1447
01:23:41,225 --> 01:23:42,584
necessarily insuring Grok making
a wrong call on a job that it

1448
01:23:42,584 --> 01:23:43,000
was supposed to do.

1449
01:23:43,000 --> 01:23:48,198
Audrow Nash: Yep. Yeah. I see.
Yeah, that makes a lot of sense.

1450
01:23:48,198 --> 01:23:52,653
Where? So, I guess, from the
cyber--where do you think

1451
01:23:52,653 --> 01:23:57,521
cyber's going for all this? I
guess robots will probably be

1452
01:23:57,521 --> 01:24:02,472
hacked at some points, but,
like, what's your read on it? Is

1453
01:24:02,472 --> 01:24:07,175
it, um--is, like, software
engineering going to turn into

1454
01:24:07,175 --> 01:24:11,713
cyber protection? Like, we're
all going to become cyber

1455
01:24:11,713 --> 01:24:16,416
engineers, or, like, how do
you--where do you think we're

1456
01:24:16,416 --> 01:24:21,201
going with all this? Like,
because the AI--it's enabling a

1457
01:24:21,201 --> 01:24:26,069
lot of new things with all of
this, and, I mean, it's crazy

1458
01:24:26,069 --> 01:24:30,112
finding, like, super old
vulnerabilities in Linux

1459
01:24:30,112 --> 01:24:33,000
distributions and things like
this.

1460
01:24:33,000 --> 01:24:36,772
Sergey Litvinenko: Mm-hmm. From
our perspective, we're already

1461
01:24:36,772 --> 01:24:40,727
seeing how – so, when a robotics
company comes to Koop, the first

1462
01:24:40,727 --> 01:24:43,830
policy they buy is usually
general liability, which

1463
01:24:43,830 --> 01:24:47,480
includes product liability.
That's something that you get in

1464
01:24:47,480 --> 01:24:51,313
order to protect third parties
within your company. And then if

1465
01:24:51,313 --> 01:24:55,146
you ship a product and you have
no control over the product, it

1466
01:24:55,146 --> 01:24:58,918
also covers that if something
goes wrong with the product. And

1467
01:24:58,918 --> 01:25:02,325
then, because most of our
robotics companies have a RaaS

1468
01:25:02,325 --> 01:25:05,854
model, they also have to buy
robotics errors and omissions

1469
01:25:05,854 --> 01:25:09,504
and bundle it with cyber because
sometimes if something goes

1470
01:25:09,504 --> 01:25:13,398
wrong, you don't know, was it by
software or was it like a cyber

1471
01:25:13,398 --> 01:25:16,623
issue on board. And I think
cyber and tech errors and

1472
01:25:16,623 --> 01:25:20,334
omissions combined, there is
going to be an essential package

1473
01:25:20,334 --> 01:25:24,167
that you probably will get from
day one and it's going to be as

1474
01:25:24,167 --> 01:25:28,000
big if not bigger than general
liability for robotics industry.

1475
01:25:28,000 --> 01:25:35,000
Audrow Nash: Mm-hmm.
Interesting.

1476
01:25:35,000 --> 01:25:37,927
Sergey Litvinenko: Probably – I
mean, definitely the same thing

1477
01:25:37,927 --> 01:25:40,854
for AI agents, primarily because
there is nothing about general

1478
01:25:40,854 --> 01:25:43,827
liability that you have to worry
about. It's all about cyber and

1479
01:25:43,827 --> 01:25:46,708
errors and omissions. So it's
just going to become the default

1480
01:25:46,708 --> 01:25:49,635
standard line that you have to
have. And I'm not saying that AI

1481
01:25:49,635 --> 01:25:52,469
agents don't need insurance.
They do need insurance. I'm just

1482
01:25:52,469 --> 01:25:55,117
saying that I don't think it's
going to be as bad as some

1483
01:25:55,117 --> 01:25:56,000
people think. Yeah.

1484
01:25:56,000 --> 01:26:00,805
Audrow Nash: It's a smaller
market. Yeah. Yeah, I mean, it,

1485
01:26:00,805 --> 01:26:05,772
like, doing some dumb trade for
me or something is, like, the

1486
01:26:05,772 --> 01:26:10,251
big example that I can think of,
or, like, leaking some

1487
01:26:10,251 --> 01:26:15,300
information. But the--it's very
interesting. Do you think that

1488
01:26:15,300 --> 01:26:19,860
the cyber industry will be
bigger? Like, it'll be--it'll

1489
01:26:19,860 --> 01:26:24,583
blow up and be absolutely
enormous for insurance? Like, it

1490
01:26:24,583 --> 01:26:29,713
very well might be. It very well
might eclipse robotics in this

1491
01:26:29,713 --> 01:26:34,599
near time with, um--like, could
you imagine if Koop ever was

1492
01:26:34,599 --> 01:26:39,567
insuring vibe-coded software as
a service or other, um--other

1493
01:26:39,567 --> 01:26:44,453
businesses that, like, want
insurance against cyber attacks?

1494
01:26:44,453 --> 01:26:46,000
This kind of thing.

1495
01:26:46,000 --> 01:26:48,930
Sergey Litvinenko: It's on the
roadmap. The thing is, well when

1496
01:26:48,930 --> 01:26:51,674
we if we were to look at this
opportunity, we would look at

1497
01:26:51,674 --> 01:26:54,186
this from the AI agent
perspective. So those are still

1498
01:26:54,186 --> 01:26:56,977
bots and you can get telemetry
from those bots. You can know

1499
01:26:56,977 --> 01:26:59,767
which environment they operate
in, which kind of access they

1500
01:26:59,767 --> 01:27:00,000
have.

1501
01:27:00,000 --> 01:27:06,743
Audrow Nash: Yep. Oh, cool. So
it fits your existing model.

1502
01:27:06,743 --> 01:27:08,000
Yeah, yeah.

1503
01:27:08,000 --> 01:27:11,661
Sergey Litvinenko: Yeah. And
then certifications. So we can

1504
01:27:11,661 --> 01:27:15,571
like insure those agents against
liability. But then we'll also

1505
01:27:15,571 --> 01:27:19,108
have to understand the whole
cyber posture of the company

1506
01:27:19,108 --> 01:27:22,707
because the prop the claims are
going to arise from agents

1507
01:27:22,707 --> 01:27:26,183
leaking information that you're
supposed to protect, the

1508
01:27:26,183 --> 01:27:30,154
third-party information. So that
part is exciting. And we do the

1509
01:27:30,154 --> 01:27:33,691
similar thing in the physical
world. Just in the physical

1510
01:27:33,691 --> 01:27:37,291
world, the liability is much
more particular there than in

1511
01:27:37,291 --> 01:27:40,890
the digital world. Digital, it's
all cyber. Physical, it's

1512
01:27:40,890 --> 01:27:43,000
damaging people or hurting
people.

1513
01:27:43,000 --> 01:27:49,000
Audrow Nash: Mm-hmm.

1514
01:27:49,000 --> 01:27:50,242
Sergey Litvinenko: So I
definitely think there is like a

1515
01:27:50,242 --> 01:27:51,639
way to do this in the physical
world, in the digital world. And

1516
01:27:51,639 --> 01:27:52,947
then another convergence might
happen is that if the models

1517
01:27:52,947 --> 01:27:54,233
behind those AI agents start
working on robots like Astra,

1518
01:27:54,233 --> 01:27:55,475
working on robot arms and
showing good results, then you

1519
01:27:55,475 --> 01:27:56,583
have to understand Astra
performance from start to

1520
01:27:56,583 --> 01:27:57,958
finish. And then it could apply
to robots, to robot use cases,

1521
01:27:57,958 --> 01:27:59,000
to browser use, to computer use
and everything.

1522
01:27:59,000 --> 01:28:17,000
Audrow Nash: Mm-hmm.

1523
01:28:17,000 --> 01:28:17,518
Sergey Litvinenko: That's
exciting. We're not there yet. I

1524
01:28:17,518 --> 01:28:17,991
think it's the place that we
want to explore with the

1525
01:28:17,991 --> 01:28:18,571
research initiative that we have
coming up. And hopefully it will

1526
01:28:18,571 --> 01:28:19,000
materialize in new products
later down the road.

1527
01:28:19,000 --> 01:28:25,715
Audrow Nash: Mm-hmm. Cool. Hell
yeah. Yeah, it seems like an

1528
01:28:25,715 --> 01:28:32,319
emerging space to me. and so
that's a--that's a cool thing.

1529
01:28:32,319 --> 01:28:38,810
let's see. I wonder, one thing
that we talked about at the

1530
01:28:38,810 --> 01:28:45,861
beginning, or you and I before,
you said robots are becoming an

1531
01:28:45,861 --> 01:28:52,689
asset class. Tell me more about
that. Like, I don't even know

1532
01:28:52,689 --> 01:28:59,628
what it means, really. I think
asset class, like, real estate,

1533
01:28:59,628 --> 01:29:05,000
but I don't understand the
implications of that.

1534
01:29:05,000 --> 01:29:08,730
Sergey Litvinenko: I'll two
parts to the answer. First part,

1535
01:29:08,730 --> 01:29:12,584
if you today, and I'm going to
take the insurance angle, today

1536
01:29:12,584 --> 01:29:16,065
if you go to big insurance
companies and you fill out an

1537
01:29:16,065 --> 01:29:19,857
application to get coverage, you
cannot select robotics as an

1538
01:29:19,857 --> 01:29:23,462
industry. It's just not on the
list. It's not in the North

1539
01:29:23,462 --> 01:29:26,384
American industry
classification. I don't think

1540
01:29:26,384 --> 01:29:28,000
they have robotics at all.

1541
01:29:28,000 --> 01:29:34,000
Audrow Nash: Mm-hmm. Mm.

1542
01:29:34,000 --> 01:29:36,250
Sergey Litvinenko: And that's
crazy because robotics is an

1543
01:29:36,250 --> 01:29:38,734
industry. It has applications in
various use cases, but it is an

1544
01:29:38,734 --> 01:29:41,139
industry. And nobody has tried
to look at robots as one thing.

1545
01:29:41,139 --> 01:29:43,196
People take autonomous vehicles
as kind of one class,

1546
01:29:43,196 --> 01:29:45,369
self-driving cars, then they
look at drones as the other

1547
01:29:45,369 --> 01:29:47,813
class, industrial robots as the
third class, mobile robots. But

1548
01:29:47,813 --> 01:29:50,102
nobody tried to look at them as
one thing, as robots. And I

1549
01:29:50,102 --> 01:29:52,508
think we're the first company to
do this. And we already share

1550
01:29:52,508 --> 01:29:54,875
the results, and results are
pretty good. So when I say robot

1551
01:29:54,875 --> 01:29:56,000
is a new asset class, is that

1552
01:29:56,000 --> 01:30:18,000
Audrow Nash: Mm-hmm.

1553
01:30:18,000 --> 01:30:20,360
Sergey Litvinenko: If you
combine the all of them together

1554
01:30:20,360 --> 01:30:22,883
and you see how the deployments
are going and that we're going

1555
01:30:22,883 --> 01:30:25,162
to have millions and billions of
robots eventually, that

1556
01:30:25,162 --> 01:30:27,563
ecosystem will need lots of
help. Insurance, so that's risk

1557
01:30:27,563 --> 01:30:29,963
management, making sure that
those robots can get deployed.

1558
01:30:29,963 --> 01:30:32,446
Obviously, third-party safety
certifications, that's going to

1559
01:30:32,446 --> 01:30:34,928
be a huge thing. Financing, so
do you buy the robot or do you

1560
01:30:34,928 --> 01:30:37,491
lease the robot? Securitization,
if I have million robotaxis in

1561
01:30:37,491 --> 01:30:39,974
the US and they have predictable
service and predictable cash

1562
01:30:39,974 --> 01:30:42,293
flow, you can build like
literally financial products out

1563
01:30:42,293 --> 01:30:44,775
of it. So when I say robot is an
asset class, it's more about

1564
01:30:44,775 --> 01:30:47,217
thinking about second and third
derivative of that industry,

1565
01:30:47,217 --> 01:30:49,577
what other things are going to
come out of it. And I think

1566
01:30:49,577 --> 01:30:52,181
financial products and services,
which we're a part of, is going

1567
01:30:52,181 --> 01:30:54,663
to be a huge enabling thing like
it was with cars. Now, every

1568
01:30:54,663 --> 01:30:56,983
single car company is also a
finance company because they

1569
01:30:56,983 --> 01:30:58,000
want, they help people to

1570
01:30:58,000 --> 01:31:21,000
Audrow Nash: Mm-hmm.

1571
01:31:21,000 --> 01:31:24,557
Sergey Litvinenko: Get car
loans. So it's Ford Financial,

1572
01:31:24,557 --> 01:31:28,425
Hyundai Financial. And I think
robots will see a similar thing

1573
01:31:28,425 --> 01:31:32,419
where you'll be able to finance
robots, you'll be able to insure

1574
01:31:32,419 --> 01:31:36,100
them, get all kinds of services
on top of them. And at some

1575
01:31:36,100 --> 01:31:39,969
point, robotics will become the
biggest industry in the world.

1576
01:31:39,969 --> 01:31:43,837
And that's why we need to start
looking at robots and tracking

1577
01:31:43,837 --> 01:31:47,269
like robot deployments
population across many different

1578
01:31:47,269 --> 01:31:50,951
use cases as one asset class,
not a separate asset classes.

1579
01:31:50,951 --> 01:31:54,695
That is going to be more, it
starts to make more sense that,

1580
01:31:54,695 --> 01:31:58,376
okay, the robots are really
coming just in different shapes

1581
01:31:58,376 --> 01:31:59,000
and forms.

1582
01:31:59,000 --> 01:32:03,377
Audrow Nash: Mm-hmm. Yeah, it's
all the same technology backbone

1583
01:32:03,377 --> 01:32:07,617
in some sense, but they've come
in, I mean, arms and humanoids

1584
01:32:07,617 --> 01:32:11,583
and everything else. Very
interesting. Do you think, So, I

1585
01:32:11,583 --> 01:32:15,618
guess... It makes... I guess I
still, I don't fully get the

1586
01:32:15,618 --> 01:32:19,995
implications with It sounds like
robots are coming. I agree with

1587
01:32:19,995 --> 01:32:23,619
that. From a financial
perspective, you're looking at

1588
01:32:23,619 --> 01:32:27,449
the industries that'll pop up
around it. It's like who's

1589
01:32:27,449 --> 01:32:31,621
selling the shovels in the gold
rush kind of thing. There's a

1590
01:32:31,621 --> 01:32:35,861
whole bunch of... infrastructure
and things that robotics will

1591
01:32:35,861 --> 01:32:39,759
need that will need to be
tailored to robotics companies.

1592
01:32:39,759 --> 01:32:44,067
Is that kind of the big point?
The big point is... we can shell

1593
01:32:44,067 --> 01:32:48,239
sell You can sell... There's
gonna be a lot of opportunity to

1594
01:32:48,239 --> 01:32:52,000
sell shovels to robotics
companies. This kind of thing.

1595
01:32:52,000 --> 01:32:54,796
Sergey Litvinenko: Yep. The
Yeah, to be specific, the

1596
01:32:54,796 --> 01:32:57,962
implication is that there is a
wave of products that will be

1597
01:32:57,962 --> 01:33:00,863
built on top of the robotics
industry, and specifically

1598
01:33:00,863 --> 01:33:04,082
talking financial products that
people are not thinking about

1599
01:33:04,082 --> 01:33:07,141
right now. And I think that
financial layer will be a huge

1600
01:33:07,141 --> 01:33:10,201
enabler. And in order to make
this layer work, you need to

1601
01:33:10,201 --> 01:33:13,314
think about robots as an asset
class, and then it will make

1602
01:33:13,314 --> 01:33:14,000
sense to you.

1603
01:33:14,000 --> 01:33:18,559
Audrow Nash: Ah. Yep. You think
about it holistically like you

1604
01:33:18,559 --> 01:33:23,118
all are doing, and then it makes
it so that robotics companies

1605
01:33:23,118 --> 01:33:24,000
can leverage

1606
01:33:24,000 --> 01:33:30,000
Sergey Litvinenko: Exactly.
Yeah. Okay.

1607
01:33:30,000 --> 01:33:32,720
Audrow Nash: All of what's being
built out specifically for

1608
01:33:32,720 --> 01:33:35,624
robotics companies. Okay, I like
that idea a lot. Hell yeah. so

1609
01:33:35,624 --> 01:33:38,528
starting to wrap up, I see that
my toddler is really wanting to

1610
01:33:38,528 --> 01:33:41,433
hang out. Yeah, but what are you
looking at looking out for the

1611
01:33:41,433 --> 01:33:43,000
next three to five years for
Koop?

1612
01:33:43,000 --> 01:33:51,000
Sergey Litvinenko: It's time.
Okay.

1613
01:33:51,000 --> 01:33:55,662
Audrow Nash: And perhaps the
robotics industry overall. Where

1614
01:33:55,662 --> 01:34:00,325
are you thinking we're going?
Like what do you think three or

1615
01:34:00,325 --> 01:34:03,000
five years from now will look
like?

1616
01:34:03,000 --> 01:34:06,654
Sergey Litvinenko: I think the
first noticeable viral use case

1617
01:34:06,654 --> 01:34:09,836
that the humanity will
experience will be self driving

1618
01:34:09,836 --> 01:34:13,490
cars. I think that set in stone
is going to be cars everywhere

1619
01:34:13,490 --> 01:34:17,085
in every major city and that's
going to be the first physical

1620
01:34:17,085 --> 01:34:20,680
AI experience that people will
say, Oh, shoot, the robots are

1621
01:34:20,680 --> 01:34:24,157
here. That's probably give it
one more year just to warm up

1622
01:34:24,157 --> 01:34:27,810
for Tesla and then Tesla Waymo
and other players, they're just

1623
01:34:27,810 --> 01:34:31,582
going to go nuts. Okay. They've
proven the technology now it's a

1624
01:34:31,582 --> 01:34:35,000
business problem. And I'm really
hoping that shortly after

1625
01:34:35,000 --> 01:34:43,000
Audrow Nash: Mm-hmm.

1626
01:34:43,000 --> 01:34:47,633
Sergey Litvinenko: Humanoids and
all other form factors will

1627
01:34:47,633 --> 01:34:52,344
follow. talk Talking about 2030s
or so and I really hope that

1628
01:34:52,344 --> 01:34:57,131
we'll see about 2027 but I hope
people like if you have money,

1629
01:34:57,131 --> 01:35:01,996
you'll be able to get a humanoid
into your home in 2028. That'd

1630
01:35:01,996 --> 01:35:03,000
be great. And

1631
01:35:03,000 --> 01:35:05,000
Audrow Nash: That'd be so cool.

1632
01:35:05,000 --> 01:35:09,025
Sergey Litvinenko: From that
moment, I think it's only gonna,

1633
01:35:09,025 --> 01:35:13,117
it's not going to go like viral
like ChatGPT adoption was it's

1634
01:35:13,117 --> 01:35:17,076
going to be steady, but it's
going to be certain growth. And

1635
01:35:17,076 --> 01:35:20,772
I hope that by call it 2030 to
2035 humanoid robots self

1636
01:35:20,772 --> 01:35:24,995
driving cars drones everywhere,
it just going to become a normal

1637
01:35:24,995 --> 01:35:29,284
thing, like, definitely children
will be like yeah, haven't these

1638
01:35:29,284 --> 01:35:31,000
things always been around.

1639
01:35:31,000 --> 01:35:35,235
Audrow Nash: Yeah. I know. It's
wild. Never learn to drive and

1640
01:35:35,235 --> 01:35:39,197
take it. I just, like our one
and a half year old, I can't

1641
01:35:39,197 --> 01:35:42,817
imagine she's going to grow up
with, like, LLMs being

1642
01:35:42,817 --> 01:35:46,915
unbelievable. Like, she never
knew it before time, this kind

1643
01:35:46,915 --> 01:35:51,150
of thing. And it's very unlikely
she's going to learn to drive

1644
01:35:51,150 --> 01:35:55,180
for anything but recreation,
like this kind of thing. Like,

1645
01:35:55,180 --> 01:35:56,000
what a time.

1646
01:35:56,000 --> 01:36:00,000
Sergey Litvinenko: Yeah, yeah.
What a time indeed.

1647
01:36:00,000 --> 01:36:03,506
Audrow Nash: Okay, hell yeah.
Well, this was a lot of fun. I

1648
01:36:03,506 --> 01:36:07,130
appreciate you coming back for a
second time, and you guys are

1649
01:36:07,130 --> 01:36:09,000
doing some really exciting work.

1650
01:36:09,000 --> 01:36:12,824
Sergey Litvinenko: Thank you,
Audrow. Great to be back and

1651
01:36:12,824 --> 01:36:15,000
hoping for round three next
year.

1652
01:36:15,000 --> 01:36:20,000
Audrow Nash: Hell yeah. Alright.
Bye, everyone. Alright, I'm
