1
00:00:00,000 --> 00:00:05,594
Stefan Seltz-Axmacher: Like the
best robot in the world today,

2
00:00:05,594 --> 00:00:11,008
is less valuable for its owner
than like an average computer

3
00:00:11,008 --> 00:00:12,000
was in 1960

4
00:00:12,000 --> 00:00:14,680
Audrow Nash: Alright. Hi,
everyone. Let's start with

5
00:00:14,680 --> 00:00:17,000
intros. Stefan, would you
introduce yourself?

6
00:00:17,000 --> 00:00:21,455
Stefan Seltz-Axmacher: Yep. I'm
Stefan Seltz-Axmacher. I'm CEO

7
00:00:21,455 --> 00:00:25,623
and co-founder of Polymath
Robotics. And, the lesser good,

8
00:00:25,623 --> 00:00:29,000
co-host of automated, another
robotics podcast.

9
00:00:29,000 --> 00:00:32,000
Audrow Nash: Alright, Ilia,
would you introduce yourself?

10
00:00:32,000 --> 00:00:35,362
Ilia Baranov: And I'm Ilia
Baranov co-founder and CTO here

11
00:00:35,362 --> 00:00:38,667
at polymath. And, I'm the
silent, straight man partner of

12
00:00:38,667 --> 00:00:40,000
the, Automated podcast.

13
00:00:40,000 --> 00:00:44,244
Audrow Nash: Nice. All right.
Hell, yeah. So tell me about

14
00:00:44,244 --> 00:00:48,341
polymath. Let's do a quick
review. Stefan, you will jump

15
00:00:48,341 --> 00:00:49,000
in. Yeah.

16
00:00:49,000 --> 00:00:52,425
Stefan Seltz-Axmacher: Of
course. Yeah. So, polymath,

17
00:00:52,425 --> 00:00:56,109
basically what we've done is
we've take taken the kind of

18
00:00:56,109 --> 00:00:59,922
commodity hard but unsought but,
purchasable point to point

19
00:00:59,922 --> 00:01:03,735
navigation stack that every
robotics team has to build over

20
00:01:03,735 --> 00:01:07,742
and over and over again, and we
package it up as software that

21
00:01:07,742 --> 00:01:11,814
can be deployed across any off
highway robot. So we're we don't

22
00:01:11,814 --> 00:01:15,756
want to be a vertical robotics
company. We don't even want to

23
00:01:15,756 --> 00:01:19,117
be the only autonomy
organization in the world. What

24
00:01:19,117 --> 00:01:22,801
we really want to do is we want
to take the parts of your

25
00:01:22,801 --> 00:01:26,420
robotics stack that you've
probably built three, 3 to 18

26
00:01:26,420 --> 00:01:30,362
times and make it so that you
can just integrate ours instead

27
00:01:30,362 --> 00:01:34,111
of rebuilding your own arm.
We're on tractors, bulldozers.

28
00:01:34,111 --> 00:01:37,924
We're too heavy for the DoD. As
neat updates from last time

29
00:01:37,924 --> 00:01:41,866
around the show. We're on more
different types of robots than

30
00:01:41,866 --> 00:01:46,003
we have people on our team. And
we can do everything from all of

31
00:01:46,003 --> 00:01:50,074
the software from point to point
navigation all the way down to

32
00:01:50,074 --> 00:01:54,081
ripping out individual modules
like a certifiable safety layer

33
00:01:54,081 --> 00:01:57,765
or localization or obstacle
detection or any one of those

34
00:01:57,765 --> 00:02:01,707
things that you don't want to
build and isn't really core to,

35
00:02:01,707 --> 00:02:03,000
to your own program.

36
00:02:03,000 --> 00:02:06,000
Audrow Nash: Hell yeah. Ilia,
anything to add?

37
00:02:06,000 --> 00:02:09,697
Ilia Baranov: No I mean Stefan,
basically summarized it. It's,

38
00:02:09,697 --> 00:02:13,156
it's been it's been a journey
to, to build the team out. I

39
00:02:13,156 --> 00:02:16,734
think we're overall doing well.
Probably not as well as we'd

40
00:02:16,734 --> 00:02:19,000
love to do in our wildest
dreams, but.

41
00:02:19,000 --> 00:02:21,777
Stefan Seltz-Axmacher: I don't
think that it's going to be

42
00:02:21,777 --> 00:02:24,840
possible in a robotics. I don't
think there's a physical area is

43
00:02:24,840 --> 00:02:27,665
physically possible in robotics.
Do as well as your wildest

44
00:02:27,665 --> 00:02:28,000
dreams?

45
00:02:28,000 --> 00:02:30,000
Audrow Nash: Yeah, I guess it's
not like that.

46
00:02:30,000 --> 00:02:31,000
Ilia Baranov: Or any tech
business

47
00:02:31,000 --> 00:02:34,755
Audrow Nash: Some crazy unicorn
startup. But you guys are also,

48
00:02:34,755 --> 00:02:38,093
I mean, trying not to have a
vertical, trying to be more

49
00:02:38,093 --> 00:02:41,132
software focused where you
integrate real well with

50
00:02:41,132 --> 00:02:44,887
hardware. Yep. It's like that to
me from our last conversation.

51
00:02:44,887 --> 00:02:48,642
It's like it seems like a super
good, like lifestyle startup in

52
00:02:48,642 --> 00:02:49,000
a way.

53
00:02:49,000 --> 00:02:52,410
Stefan Seltz-Axmacher: Yeah,
it's definitely a nicer life

54
00:02:52,410 --> 00:02:56,000
than having to ship engineers to
west Nowheresville, on five

55
00:02:56,000 --> 00:02:59,470
hours notice to try to fix
something. But essentially, you

56
00:02:59,470 --> 00:03:02,880
know, a thing that's really
interesting about this market

57
00:03:02,880 --> 00:03:06,410
and, and kind of, describe it
more towards roboticists. But

58
00:03:06,410 --> 00:03:10,120
after my first robotics startup,
I went and hung out with some

59
00:03:10,120 --> 00:03:13,650
SaaS companies, and I didn't,
like, realize how much easier

60
00:03:13,650 --> 00:03:17,239
their life was. And it's not
just the lack of hardware. They

61
00:03:17,239 --> 00:03:20,769
feel like a lot of people in
robotics are like, okay, cool.

62
00:03:20,769 --> 00:03:24,120
Yes, this is harder than
DocuSign, but like, I gotta buy

63
00:03:24,120 --> 00:03:27,889
actuators, I gotta buy sensors.
There's regulatory stuff. Maybe

64
00:03:27,889 --> 00:03:31,479
it's like, that's actually not
the why. Like SaaS is easier.

65
00:03:31,479 --> 00:03:35,308
SaaS is easier because like most
people don't have to build most

66
00:03:35,308 --> 00:03:39,077
of the stuff. And like if you're
building a website, you should

67
00:03:39,077 --> 00:03:42,368
not really be doing backend
stuff unless you're back at

68
00:03:42,368 --> 00:03:45,838
website. You should not really
be innovating on collecting

69
00:03:45,838 --> 00:03:49,726
credit card payments. You should
not really be innovating on like

70
00:03:49,726 --> 00:03:53,436
99% of the code base of that,
that website or that that online

71
00:03:53,436 --> 00:03:56,786
service. Almost everything
should be clear code at least

72
00:03:56,786 --> 00:04:00,197
until you're the millions of
dollars that are. Whereas in

73
00:04:00,197 --> 00:04:03,487
robotics, kind of there's
features that you've built in

74
00:04:03,487 --> 00:04:06,778
your last company, there's
features that, disrupt their

75
00:04:06,778 --> 00:04:10,308
thesis about. There's features
that some company just did a

76
00:04:10,308 --> 00:04:13,718
really cool marketing video
showing off that they can do,

77
00:04:13,718 --> 00:04:17,368
which leads everyone else to
think that it is as easy to say,

78
00:04:17,368 --> 00:04:20,957
make a humanoid walk up, grab a
doorknob, twist it, and walk

79
00:04:20,957 --> 00:04:24,547
through the door and gently
close the door behind them. When

80
00:04:24,547 --> 00:04:28,256
in reality, if you were starting
with, if I sent you a Unitree

81
00:04:28,256 --> 00:04:31,846
humanoid today, it would still
be a decent amount of work to

82
00:04:31,846 --> 00:04:35,316
get even to the point where you
could do that, even if the

83
00:04:35,316 --> 00:04:39,085
hardware was out of the problem.
And that's because of, of kind

84
00:04:39,085 --> 00:04:42,974
of this, this lack of being able
to, to take components, from one

85
00:04:42,974 --> 00:04:46,863
robotics program to another. And
that's kind of what we're hoping

86
00:04:46,863 --> 00:04:48,000
to partially solve.

87
00:04:48,000 --> 00:04:52,471
Audrow Nash: Yeah, I totally
agree with that. And that's why

88
00:04:52,471 --> 00:04:56,941
a lot of web dev stuff feels so
much faster because you, you

89
00:04:56,941 --> 00:05:01,188
don't have to write everything.
Yeah. There's really good

90
00:05:01,188 --> 00:05:05,733
libraries for thorny specific
problems like payments or other

91
00:05:05,733 --> 00:05:07,000
things like this.

92
00:05:07,000 --> 00:05:10,319
Ilia Baranov: Or you just go, or
you just go to the vibe coding.

93
00:05:10,319 --> 00:05:13,378
Right. Which Stefan has been
experimenting with. Yeah. Some

94
00:05:13,378 --> 00:05:14,000
things come.

95
00:05:14,000 --> 00:05:17,546
Stefan Seltz-Axmacher: Oh. But
yeah. Like, but like, I mean, I

96
00:05:17,546 --> 00:05:21,092
think like and that's the, like
the gap between where robotics

97
00:05:21,092 --> 00:05:24,753
actually is and like, where the
people outside of robotics think

98
00:05:24,753 --> 00:05:28,242
it is, is only kind of seeming
to get bigger as time goes on.

99
00:05:28,242 --> 00:05:31,731
Because people when you look at
a web app, when you look at a

100
00:05:31,731 --> 00:05:35,105
SaaS application, any feature
that you can you see, you can

101
00:05:35,105 --> 00:05:38,537
assume is commodity and easily
replicable. The more money is

102
00:05:38,537 --> 00:05:41,797
cut into robotics, the more
features people have seen and

103
00:05:41,797 --> 00:05:45,000
they've applied that same
heuristic to it. And that's a.

104
00:05:45,000 --> 00:05:47,000
Audrow Nash: Really interesting.

105
00:05:47,000 --> 00:05:49,000
Stefan Seltz-Axmacher:
Interesting thing.

106
00:05:49,000 --> 00:05:53,390
Audrow Nash: Yeah. So I agree
with you. And I feel like it's

107
00:05:53,390 --> 00:05:57,488
hard I mean like seeing the
Unitree robot do like a side

108
00:05:57,488 --> 00:06:02,244
flip. Yeah. And stuff. It's like
pretty hard not to project a lot

109
00:06:02,244 --> 00:06:04,000
of competence onto that.

110
00:06:04,000 --> 00:06:07,714
Stefan Seltz-Axmacher: I saw you
make I saw an interesting

111
00:06:07,714 --> 00:06:11,365
headline. And, and I said, I
know this is a headline or a

112
00:06:11,365 --> 00:06:15,271
Reddit post, but someone saying
Unitree is a scam. I bought a

113
00:06:15,271 --> 00:06:18,921
Unitree and I can't do some
function, and I don't know if

114
00:06:18,921 --> 00:06:22,828
some function is walk through
the walk up to the door, grab a

115
00:06:22,828 --> 00:06:26,862
doorknob, and walk on through. I
don't know if some function is

116
00:06:26,862 --> 00:06:30,704
do a backflip. I don't know if
some function is whenever you

117
00:06:30,704 --> 00:06:34,803
get an obstacle course. But they
bought this thing thinking I've

118
00:06:34,803 --> 00:06:38,773
seen a Unitree do a cartwheel.
If I buy a Unitree, it can do a

119
00:06:38,773 --> 00:06:42,744
cartwheel. Because that's how
everything else works. It's only

120
00:06:42,744 --> 00:06:46,458
that's losers in robotics where
you see a function happen,

121
00:06:46,458 --> 00:06:50,365
you're like, cool, I'll go buy a
$40,000 robot. And if I do a

122
00:06:50,365 --> 00:06:54,143
really good nine months of hard
work, I can replicate that.

123
00:06:54,143 --> 00:06:56,000
Yeah, this saw sort of. Yeah.

124
00:06:56,000 --> 00:07:00,066
Audrow Nash: For sure. So with
this, you're saying everybody

125
00:07:00,066 --> 00:07:04,336
else does this? I understand it
because or like the way I think

126
00:07:04,336 --> 00:07:08,605
of it. And I'd like to see Ilia
your perspective, and then jump

127
00:07:08,605 --> 00:07:12,739
back to you, Stefan. But the,
the way I understand it is that

128
00:07:12,739 --> 00:07:16,873
robotics is very early. We don't
have standards yet. We still

129
00:07:16,873 --> 00:07:21,075
haven't understood all the hard
problems you have. ROS and ROS

130
00:07:21,075 --> 00:07:24,532
solves some of the hard
problems, but there's still

131
00:07:24,532 --> 00:07:27,852
plenty of hard problems,
especially in the, like,

132
00:07:27,852 --> 00:07:32,122
behavioral space. Which is what
you guys are doing. But, how do

133
00:07:32,122 --> 00:07:36,527
how do you think of this? Do you
think that's correct? That it is

134
00:07:36,527 --> 00:07:40,593
because we're still early or
it's just harder or what do you

135
00:07:40,593 --> 00:07:41,000
think?

136
00:07:41,000 --> 00:07:44,736
Ilia Baranov: Yeah. We kind of
go back and forth on this topic

137
00:07:44,736 --> 00:07:47,930
a lot. And if you really think
about it, robots as an

138
00:07:47,930 --> 00:07:51,606
industrial manipulators, go back
to around the same time that

139
00:07:51,606 --> 00:07:55,041
heavy ion computers started
popping up. You know, you had

140
00:07:55,041 --> 00:07:58,716
the original hydraulic arms in
auto plants come up around the

141
00:07:58,716 --> 00:08:02,091
same time as, as computers are
really getting going. But

142
00:08:02,091 --> 00:08:05,707
computers have accelerated so
much faster and made so much a

143
00:08:05,707 --> 00:08:09,322
deeper impact on society than
robots. And it's a question of

144
00:08:09,322 --> 00:08:12,576
why, right? And in a lot of
ways, I think a big job of

145
00:08:12,576 --> 00:08:16,071
computing and of UI is to get
squish, the weirdness of the

146
00:08:16,071 --> 00:08:19,627
world, into a very specific
framework that the computer can

147
00:08:19,627 --> 00:08:23,423
deal with. So like a screen with
this type of web browser, with

148
00:08:23,423 --> 00:08:26,978
this kind of data or whatever,
right. And a lot of work was

149
00:08:26,978 --> 00:08:30,474
done on here's a communication
protocol like TCP, here's a

150
00:08:30,474 --> 00:08:33,788
standard for wiring, here's a
standard for CPUs so that

151
00:08:33,788 --> 00:08:37,223
everybody can enter inter
process. Whereas robotics never

152
00:08:37,223 --> 00:08:40,597
really had that push. And so
industrial arms made by one

153
00:08:40,597 --> 00:08:43,851
company were drastically
different than industrial was

154
00:08:43,851 --> 00:08:47,407
made by another company. And up
until very recently, if you

155
00:08:47,407 --> 00:08:51,022
bought like a ABB arm and were
an expert in installing them,

156
00:08:51,022 --> 00:08:54,638
you'd know next to nothing about
installing a Kuka arm. Like

157
00:08:54,638 --> 00:08:58,555
basically you're useless. And in
fact, you had to unlearn a bunch

158
00:08:58,555 --> 00:09:02,231
of stuff because that's the way
they do things. And it's only

159
00:09:02,231 --> 00:09:05,967
really in the last, let's call
it ten, 15, maybe 20 years that

160
00:09:05,967 --> 00:09:09,763
we started to have this kind of
standardization layer appear in

161
00:09:09,763 --> 00:09:13,439
robotics. If I had to guess, I
think one is because you don't

162
00:09:13,439 --> 00:09:16,572
have this nice screen box, you
can put a lot of your

163
00:09:16,572 --> 00:09:19,887
interaction and code in you.
There wasn't a good way to

164
00:09:19,887 --> 00:09:23,141
standardize on physical
interaction with the world. If

165
00:09:23,141 --> 00:09:26,997
you even just think of grippers.
Right. Like, there's no there's

166
00:09:26,997 --> 00:09:30,432
no good way to say this gripper
will work in every single

167
00:09:30,432 --> 00:09:34,108
condition because it's either
too expensive or too fragile or

168
00:09:34,108 --> 00:09:37,543
not precise enough. Right?
There's always some trade offs

169
00:09:37,543 --> 00:09:41,279
where some like monitors like
we've had monitor standards from

170
00:09:41,279 --> 00:09:44,654
us for 50 years. Right? Like
there's no there's no magic

171
00:09:44,654 --> 00:09:48,571
there. And so I think it's taken
time to get the fundamental tech

172
00:09:48,571 --> 00:09:51,885
energy efficient enough and
complex enough and powerful

173
00:09:51,885 --> 00:09:55,742
enough to the point where we now
have enough extra compute power

174
00:09:55,742 --> 00:09:58,935
to add on this kind of
standardization layer. Because

175
00:09:58,935 --> 00:10:02,250
anytime you standardize
something, anytime you abstract

176
00:10:02,250 --> 00:10:06,167
something, and we do that a lot,
you inevitably pay a performance

177
00:10:06,167 --> 00:10:09,722
cost. Like when I was working at
Amazon on the little Astro

178
00:10:09,722 --> 00:10:13,157
robot, the Lego model of it
right here, we had to squeeze

179
00:10:13,157 --> 00:10:16,531
every micro out of the CPU
because there was a small CPU

180
00:10:16,531 --> 00:10:19,906
and we had to do a lot of
complicated things. And so you

181
00:10:19,906 --> 00:10:23,461
just don't have the space to do
a lot of affordances and or

182
00:10:23,461 --> 00:10:26,956
abstraction layers and those
kind of things. And now we're

183
00:10:26,956 --> 00:10:30,873
starting to. Right, and I think,
again, we're seeing a little bit

184
00:10:30,873 --> 00:10:34,308
of this race of end to end
neural models, and we can talk

185
00:10:34,308 --> 00:10:37,984
about that in a second, kind of
again, being very custom in a

186
00:10:37,984 --> 00:10:41,599
sense of like, here's a camera
input, here's the output, and

187
00:10:41,599 --> 00:10:45,095
magic happens in the middle and
you don't know what it is.

188
00:10:45,095 --> 00:10:48,590
Battling off against more
flexible applications, which can

189
00:10:48,590 --> 00:10:51,000
then be transported from robot
to robot.

190
00:10:51,000 --> 00:10:54,444
Audrow Nash: Yeah I agree with
you. So where do you think it's

191
00:10:54,444 --> 00:10:58,000
going. Ilia. And then Stefan I'd
love to hear your thoughts too.

192
00:10:58,000 --> 00:11:01,230
Ilia Baranov: Yeah. Good
question. I think I still, you

193
00:11:01,230 --> 00:11:04,813
know, looking at the history of
technology and looking at the

194
00:11:04,813 --> 00:11:08,337
history of the internet and
computers in particular, I think

195
00:11:08,337 --> 00:11:11,861
almost universally the most open
source, most comprehensible

196
00:11:11,861 --> 00:11:15,032
standard wins and the most
flexible one wins in almost

197
00:11:15,032 --> 00:11:18,674
every case. Right? So I think
that a lot of the big neural end

198
00:11:18,674 --> 00:11:22,198
to end neural net models, which
are starting to come out for

199
00:11:22,198 --> 00:11:25,487
machine learning, if they're
open source and adjustable,

200
00:11:25,487 --> 00:11:29,246
that's a good first step. But I
think especially in robotics, in

201
00:11:29,246 --> 00:11:32,887
this kind of idea of embodied
AI, you can't, for example, take

202
00:11:32,887 --> 00:11:36,470
a Unitree demo, flip and put it
onto an Atlas robot and like,

203
00:11:36,470 --> 00:11:39,994
you know, install suit up until
the Atlas kickflip and like,

204
00:11:39,994 --> 00:11:43,753
there you go. Goes that doesn't
exist, right? And I think that's

205
00:11:43,753 --> 00:11:47,218
hampering progress. So I think
there's still going to be an

206
00:11:47,218 --> 00:11:50,801
underlying flexibility layer
that we still have to build. And

207
00:11:50,801 --> 00:11:53,972
that's part of what polymath
does. And other groups in

208
00:11:53,972 --> 00:11:57,555
introspection, in communication,
in configuration management,

209
00:11:57,555 --> 00:12:00,668
these are all things that
computers have more or less

210
00:12:00,668 --> 00:12:03,957
solved. And robot robots
haven't. And so I think there's

211
00:12:03,957 --> 00:12:07,657
a lot of growth there. And then
I think on top of that layer is

212
00:12:07,657 --> 00:12:11,063
these kind of big end to end
models, which tend to be very

213
00:12:11,063 --> 00:12:14,000
specific per task or per robot
or per environment.

214
00:12:14,000 --> 00:12:19,000
Audrow Nash: Okay. Stefan, what
do you think?

215
00:12:19,000 --> 00:12:23,047
Stefan Seltz-Axmacher: So
looking at, yeah, kind of the

216
00:12:23,047 --> 00:12:27,020
broader history, like, like
related, but more from the

217
00:12:27,020 --> 00:12:31,214
business sense. Early crappy
computers, like 50s era, 70s

218
00:12:31,214 --> 00:12:35,481
era, type of computers, which
could be a massive amount of

219
00:12:35,481 --> 00:12:39,822
hardware expense, but create
really, really big savings and

220
00:12:39,822 --> 00:12:44,531
really, really big upside, like
a single piece of hardware could

221
00:12:44,531 --> 00:12:48,431
save you tens of millions of
dollars. If that if that

222
00:12:48,431 --> 00:12:52,845
hardware is changing your supply
chain or accounting for how

223
00:12:52,845 --> 00:12:57,554
you're spending money and you're
a company the size of IBM or GM

224
00:12:57,554 --> 00:13:01,674
or whatever. And, and so the,
the ROI was pretty big and

225
00:13:01,674 --> 00:13:05,868
pretty early. And that put
pretty easy pressure on, like,

226
00:13:05,868 --> 00:13:10,136
hey, let's drive down compute
costs. So not just GM can do

227
00:13:10,136 --> 00:13:14,771
this, but also companies who are
only worth $10 billion and now

228
00:13:14,771 --> 00:13:19,112
companies are only worth a
billion, and so on and so forth.

229
00:13:19,112 --> 00:13:23,600
Robotics has more problems than
just build a compute. There's

230
00:13:23,600 --> 00:13:28,088
also batteries. There's also
actuators. There's also sensing.

231
00:13:28,088 --> 00:13:32,576
There's that big long list of
other things. And it is hard to

232
00:13:32,576 --> 00:13:36,623
come up with a robot, a robot
that individually saves a

233
00:13:36,623 --> 00:13:41,258
company more than $5 million a
year. Like the best robot in the

234
00:13:41,258 --> 00:13:45,967
world today is less valuable for
its owner than, like an average

235
00:13:45,967 --> 00:13:50,603
computer was in 1860. Probably
even in just like, non inflation

236
00:13:50,603 --> 00:13:54,870
adjusted dollars. The math
probably doesn't make sense. So

237
00:13:54,870 --> 00:13:59,211
like the pressures have been
phenomenal on inflation. Yeah.

238
00:13:59,211 --> 00:14:03,699
Like even if you don't include
inflation like probably one of

239
00:14:03,699 --> 00:14:08,040
those IBM mainframes in 1960
probably cost $10 million. And

240
00:14:08,040 --> 00:14:12,602
for GM, save them $150 million.
And like there's no robot that

241
00:14:12,602 --> 00:14:14,000
that is comparable.

242
00:14:14,000 --> 00:14:17,026
Ilia Baranov: We have a good
example that, we can come back

243
00:14:17,026 --> 00:14:18,000
to in mining space.

244
00:14:18,000 --> 00:14:22,212
Stefan Seltz-Axmacher: So, so
like the, the initial pressure

245
00:14:22,212 --> 00:14:26,354
for robots to get better, just
like the math didn't make as

246
00:14:26,354 --> 00:14:30,777
much sense. In the 60s and the
70s and 80s, early 90s, and even

247
00:14:30,777 --> 00:14:35,059
in the US, I think a really cool
thing happened for robotics,

248
00:14:35,059 --> 00:14:39,201
which is that a bunch of people
made a crap ton of money on

249
00:14:39,201 --> 00:14:42,992
software, on web apps, on
microservices, on organizing

250
00:14:42,992 --> 00:14:47,344
databases of information to make
them quickly retrievable. And

251
00:14:47,344 --> 00:14:51,556
many of those people thought,
hey, I'm a cool tech person. I

252
00:14:51,556 --> 00:14:55,558
got money to burn. Obviously a
cool tech person should be

253
00:14:55,558 --> 00:14:59,559
working on robots. Definitely.
And that poured money into

254
00:14:59,559 --> 00:15:03,490
things like the various
self-driving car programs. And I

255
00:15:03,490 --> 00:15:07,913
think that led to a, a race, a
series of races where people put

256
00:15:07,913 --> 00:15:11,704
so much money into this market
that, you know, compute

257
00:15:11,704 --> 00:15:15,916
continued to get better and
cheaper as it naturally did, but

258
00:15:15,916 --> 00:15:19,988
ruggedized compute became
commodity. Sensor modalities got

259
00:15:19,988 --> 00:15:24,270
wider and more reliable, more
usable. And they became cheaper

260
00:15:24,270 --> 00:15:28,131
and cheaper and cheaper as
massive amounts of, frankly,

261
00:15:28,131 --> 00:15:32,062
dumb money got poured in.
Because everyone assumed that,

262
00:15:32,062 --> 00:15:36,345
like, Moore's Law would work
just as quickly for self-driving

263
00:15:36,345 --> 00:15:40,206
as it did for, you know,
microchips in the 90s. And, we

264
00:15:40,206 --> 00:15:44,558
now we're in a position where
the components that you need for

265
00:15:44,558 --> 00:15:48,419
robotics have never been
cheaper. There's, there's been

266
00:15:48,419 --> 00:15:52,842
10 to 15 years of press release
propaganda about how robots are

267
00:15:52,842 --> 00:15:57,194
imminently going to be here. So
my hunch would be there's been

268
00:15:57,194 --> 00:16:01,617
more, more measurable interest
in robotics in the last 15 years

269
00:16:01,617 --> 00:16:05,829
than the last 100 years put
together. Probably like, and way

270
00:16:05,829 --> 00:16:09,760
more for us. An interesting
thing. It's kind of happened

271
00:16:09,760 --> 00:16:14,113
where everyone assumes there's
more robots than there are, and

272
00:16:14,113 --> 00:16:18,606
everyone assumes that should be
more purchasable than it is. So,

273
00:16:18,606 --> 00:16:22,748
you know, a big part of what
polymaths businesses, frankly,

274
00:16:22,748 --> 00:16:27,030
is there's these big industrial
companies who have said, hey,

275
00:16:27,030 --> 00:16:31,523
cool, if you can drive a car in
San Francisco, I want to drive a

276
00:16:31,523 --> 00:16:35,875
vehicle in a Roomba like pattern
in a weird part of a mine. 80

277
00:16:35,875 --> 00:16:40,158
hours a week, nonstop. Who do I
have to write a check to make

278
00:16:40,158 --> 00:16:41,000
that happen?

279
00:16:41,000 --> 00:16:43,000
Audrow Nash: It's like no one
exists.

280
00:16:43,000 --> 00:16:46,591
Stefan Seltz-Axmacher: Yeah, and
that's what we're doing. A

281
00:16:46,591 --> 00:16:50,122
polymath, functionally speaking.
And that's the that's the

282
00:16:50,122 --> 00:16:53,957
operations that we're doing. But
I think like we the components

283
00:16:53,957 --> 00:16:57,426
for robotics have gotten
cheaper. The interest has gotten

284
00:16:57,426 --> 00:17:01,322
to a fever pitch. And now it's a
matter not of like, let's argue

285
00:17:01,322 --> 00:17:04,487
about which theoretical
technology might be the most

286
00:17:04,487 --> 00:17:08,139
exciting ten years from now. I
think we're actually now in a

287
00:17:08,139 --> 00:17:11,609
how do we as a market build
products that seize this like

288
00:17:11,609 --> 00:17:14,835
incredible intersection of
conditions to make robots,

289
00:17:14,835 --> 00:17:18,000
frankly, a, a relevant industry
at the global stage?

290
00:17:18,000 --> 00:17:21,923
Audrow Nash: Yeah. I think one
thing I'm curious about your

291
00:17:21,923 --> 00:17:25,646
perspective on and then Ilia, I
want to hear your mining

292
00:17:25,646 --> 00:17:29,303
example, but the also I think
that I see a lot is labor

293
00:17:29,303 --> 00:17:33,359
shortages across things, which I
think is creating additional

294
00:17:33,359 --> 00:17:37,481
pressure. Yeah. On the factors
you are describing now. So it's

295
00:17:37,481 --> 00:17:41,604
like it's hard to find labor.
People are aging out. Yeah. What

296
00:17:41,604 --> 00:17:43,000
do you think of that?

297
00:17:43,000 --> 00:17:47,198
Stefan Seltz-Axmacher: There's.
So labor shortages became pretty

298
00:17:47,198 --> 00:17:51,134
popular to talk about I think
like post Covid. I think those

299
00:17:51,134 --> 00:17:55,070
of us in the white collar world
realize that, we desperately

300
00:17:55,070 --> 00:17:59,203
need the people, the blue collar
world to do stuff for, for our

301
00:17:59,203 --> 00:18:03,402
little tip of the pyramid to be
so pleasant. But in reality, the

302
00:18:03,402 --> 00:18:06,944
sorts of industries that we
primarily serve, you know,

303
00:18:06,944 --> 00:18:10,880
mining quarries, oil and gas,
all the, logistics, railroads,

304
00:18:10,880 --> 00:18:15,078
all of them have been plagued by
labor shortages for the last 50

305
00:18:15,078 --> 00:18:18,621
years in, like, pretty
structural, big ways. So, like,

306
00:18:18,621 --> 00:18:22,229
kind of fundamentally thinking
about manufacturing, sku

307
00:18:22,229 --> 00:18:24,000
diversity is generally bad.

308
00:18:24,000 --> 00:18:25,000
Audrow Nash: What is that? I
don't know what that is.

309
00:18:25,000 --> 00:18:28,871
Stefan Seltz-Axmacher: Yeah. SKU
like, SKU like individual types

310
00:18:28,871 --> 00:18:32,742
of products that you have. Like
when, when, when the iPhone came

311
00:18:32,742 --> 00:18:36,310
out, there was initially, you
know, it only works with, one

312
00:18:36,310 --> 00:18:39,818
radio network and there was
three battery sizes that were,

313
00:18:39,818 --> 00:18:43,205
sorry, three memory sizes. And
that was all your options

314
00:18:43,205 --> 00:18:46,773
because that was a orders of
magnitude cheaper supply chain

315
00:18:46,773 --> 00:18:50,402
to manage orders of magnitude
cheaper engineering process to

316
00:18:50,402 --> 00:18:54,031
manage. Lots and lots and lots
of advantages of we sell, you

317
00:18:54,031 --> 00:18:57,780
know, Henry Ford, you can have
your car in any color you want,

318
00:18:57,780 --> 00:19:01,409
as long as it's black. Right?
In, in industries like mining,

319
00:19:01,409 --> 00:19:05,038
however, labor shortages are so
acute that there's been like

320
00:19:05,038 --> 00:19:08,667
massive explosions of vehicle
diversity. And like, there are

321
00:19:08,667 --> 00:19:12,477
vehicles that you fundamentally
different things like you can't

322
00:19:12,477 --> 00:19:15,985
replace an excavator with a
bulldozer, you can't replace a

323
00:19:15,985 --> 00:19:19,191
bulldozer with a haul truck.
Like those are those are

324
00:19:19,191 --> 00:19:22,940
different things. Totally. But
what I mean is like caterpillar

325
00:19:22,940 --> 00:19:26,630
sells something in the range of
150 to 250 different types of

326
00:19:26,630 --> 00:19:30,319
bulldozer for, different engine
sizes, different blade sizes,

327
00:19:30,319 --> 00:19:34,069
different configurations. All of
that means and like in total,

328
00:19:34,069 --> 00:19:37,516
caterpillar has something like
800 different varieties of

329
00:19:37,516 --> 00:19:38,000
vehicle.

330
00:19:38,000 --> 00:19:40,000
Audrow Nash: That's crazy. Yeah.

331
00:19:40,000 --> 00:19:45,000
Stefan Seltz-Axmacher: Like GM,
my 100 B has like 45.

332
00:19:45,000 --> 00:19:47,000
Audrow Nash: Yeah. Wildly
different.

333
00:19:47,000 --> 00:19:48,000
Stefan Seltz-Axmacher: Yes.

334
00:19:48,000 --> 00:19:49,000
Audrow Nash: Magnitude
different. Okay.

335
00:19:49,000 --> 00:19:52,788
Stefan Seltz-Axmacher: The, and
what happens, we have all this

336
00:19:52,788 --> 00:19:56,576
diversity is every part becomes
more bespoke, which means more

337
00:19:56,576 --> 00:20:00,058
expensive. What happens is every
fleet ends up being more

338
00:20:00,058 --> 00:20:03,907
diverse, which means, you can't.
You have fewer interchangeable

339
00:20:03,907 --> 00:20:07,756
parts. Which means there's parts
that, like, are really hard to

340
00:20:07,756 --> 00:20:08,000
get.

341
00:20:08,000 --> 00:20:11,791
Audrow Nash: Just to make sure I
understand the reason that it

342
00:20:11,791 --> 00:20:15,032
becomes more bespoke is it
becomes more bespoke, more

343
00:20:15,032 --> 00:20:18,824
custom to make it so that you
have things that are more suited

344
00:20:18,824 --> 00:20:22,309
to the job so that you can
maximize the people that there

345
00:20:22,309 --> 00:20:25,000
are because of the labor
shortages. Is that.

346
00:20:25,000 --> 00:20:27,990
Stefan Seltz-Axmacher: Correct?
I need to basically have the

347
00:20:27,990 --> 00:20:30,831
biggest bulldozer that I can fit
in these major operating

348
00:20:30,831 --> 00:20:34,020
conditions, because I can never
get more people, and my mind can

349
00:20:34,020 --> 00:20:36,910
have a slightly smaller maximum
size bulldozer than yours.

350
00:20:36,910 --> 00:20:39,850
Again. So you have a wild you
have a different model than I

351
00:20:39,850 --> 00:20:40,000
do.

352
00:20:40,000 --> 00:20:43,213
Audrow Nash: So how has that
trend been changing over time?

353
00:20:43,213 --> 00:20:46,589
Has it been equally bad for the
last you said like 50 years or

354
00:20:46,589 --> 00:20:49,911
something? We've known these
labor shortages or has it gotten

355
00:20:49,911 --> 00:20:51,000
worse? Or how is it?

356
00:20:51,000 --> 00:20:54,111
Stefan Seltz-Axmacher: It seems
like you there are only more

357
00:20:54,111 --> 00:20:57,015
OEMs entering these markets,
selling only more different

358
00:20:57,015 --> 00:20:58,000
types of equipment.

359
00:20:58,000 --> 00:21:01,000
Audrow Nash: So that implies
it's getting worse and worse.

360
00:21:01,000 --> 00:21:02,000
Stefan Seltz-Axmacher: And like
these, it's.

361
00:21:02,000 --> 00:21:03,000
Audrow Nash: More and more
bespoke.

362
00:21:03,000 --> 00:21:06,882
Stefan Seltz-Axmacher: And these
models, you know, if you're if

363
00:21:06,882 --> 00:21:10,517
you're John Deere, you want to
sell people a new tractor as

364
00:21:10,517 --> 00:21:14,152
quickly as possible. So the sale
of replacement parts gets,

365
00:21:14,152 --> 00:21:17,479
sunsetting fairly frequently.
Like, we know if we have

366
00:21:17,479 --> 00:21:21,237
organizations that are building
in-house tractors because, it

367
00:21:21,237 --> 00:21:25,118
has a meaningful difference to
their bottom line, to be able to

368
00:21:25,118 --> 00:21:29,000
amortize the, the vehicle frame
for 20 years as opposed to ten.

369
00:21:29,000 --> 00:21:30,000
Audrow Nash: Wow. That's wild.

370
00:21:30,000 --> 00:21:33,869
Stefan Seltz-Axmacher: And all
of this is because lack of

371
00:21:33,869 --> 00:21:37,738
having people. Yeah. And when
you have all of these weird

372
00:21:37,738 --> 00:21:41,674
things, like going back to,
like, the parts that you can't

373
00:21:41,674 --> 00:21:45,339
replace, the ultra class haul
trucks have these really

374
00:21:45,339 --> 00:21:49,615
gigantic, you know, 3 or 4 meter
in, 3 to 5m in diameter tires,

375
00:21:49,615 --> 00:21:53,891
that are 50 to $100,000 a pop. I
have heard stories of publicly

376
00:21:53,891 --> 00:21:58,032
traded companies worth hundreds
of billions of dollars buying

377
00:21:58,032 --> 00:22:01,968
spare tire tires from known
criminals because it meant the

378
00:22:01,968 --> 00:22:05,905
vehicle got back to work weeks
earlier than it would have.

379
00:22:05,905 --> 00:22:09,977
Yeah, it's wild. And that's
because you have a weird special

380
00:22:09,977 --> 00:22:13,982
vehicle with a weird, special
part if, in fact, you're just

381
00:22:13,982 --> 00:22:18,054
using a truck like you see on a
highway, you go to the local

382
00:22:18,054 --> 00:22:22,466
equivalent of Jiffy Lube and buy
replacement tire. The reason you

383
00:22:22,466 --> 00:22:26,471
can't do that is because you
can't get enough people to, to

384
00:22:26,471 --> 00:22:30,000
drive enough trucks to do that
in that size vehicle.

385
00:22:30,000 --> 00:22:33,979
Audrow Nash: And so and so you
think with robots, if robots are

386
00:22:33,979 --> 00:22:37,705
able to take over some of the
labor. Yeah, we can have more

387
00:22:37,705 --> 00:22:41,305
general robots. And thus
economies of scale kick in, cost

388
00:22:41,305 --> 00:22:42,000
comes down.

389
00:22:42,000 --> 00:22:45,478
Stefan Seltz-Axmacher: It kind
of becomes a what is the most

390
00:22:45,478 --> 00:22:48,609
efficient size for hauling
material out of a mine? And

391
00:22:48,609 --> 00:22:52,261
efficient starts to look like.
Okay, where can we get the most,

392
00:22:52,261 --> 00:22:54,000
the most cheapest spare parts?

393
00:22:54,000 --> 00:22:57,000
Audrow Nash: Yep. For
maintainability and things.

394
00:22:57,000 --> 00:22:59,696
Stefan Seltz-Axmacher: And that
also, in turn, makes it easier

395
00:22:59,696 --> 00:23:01,000
to make these things electric.

396
00:23:01,000 --> 00:23:05,079
Ilia Baranov: Yeah. There's also
the spares consideration here.

397
00:23:05,079 --> 00:23:08,770
Where, where when you have a
critical piece of equipment.

398
00:23:08,770 --> 00:23:12,784
This is common in farming a lot.
Is that your crop that you're

399
00:23:12,784 --> 00:23:16,475
growing is not only your crop,
but you're in a geographic

400
00:23:16,475 --> 00:23:20,424
location where your neighbors
are growing the same crop, too,

401
00:23:20,424 --> 00:23:24,504
and your harvest season is all
at the same time. So your demand

402
00:23:24,504 --> 00:23:28,000
on your harvesters is like 0000
thousand percent zero.

403
00:23:28,000 --> 00:23:30,364
Audrow Nash: There is everyone's
one widget is selling at the

404
00:23:30,364 --> 00:23:32,690
same time because everyone has
peak demand at the same time.

405
00:23:32,690 --> 00:23:33,000
Exactly.

406
00:23:33,000 --> 00:23:37,044
Ilia Baranov: Exactly. And so as
a farmer, it's actually a smart

407
00:23:37,044 --> 00:23:41,025
investment not only have one $2
million harvester, but have two

408
00:23:41,025 --> 00:23:44,753
$2 million harvesters just in
case the first one breaks and

409
00:23:44,753 --> 00:23:48,291
it's that size and that cost
because your most expensive

410
00:23:48,291 --> 00:23:52,209
thing is your person driving it.
So if you don't need that and

411
00:23:52,209 --> 00:23:56,000
instead have ten $100,000 ones,
if one of them breaks, yeah.

412
00:23:56,000 --> 00:23:58,766
Audrow Nash: Doesn't matter if
one is out of commission. So

413
00:23:58,766 --> 00:23:59,000
what?

414
00:23:59,000 --> 00:24:01,000
Stefan Seltz-Axmacher: Yeah.

415
00:24:01,000 --> 00:24:04,557
Audrow Nash: Yeah. This to me,
sounds a lot like what Electric

416
00:24:04,557 --> 00:24:08,057
Sheep is doing where they have
really small mowers as opposed

417
00:24:08,057 --> 00:24:11,213
to really big ones that were
created by labor shortages

418
00:24:11,213 --> 00:24:14,770
because they wanted one person
to maximize their surface area.

419
00:24:14,770 --> 00:24:15,000
Yep.

420
00:24:15,000 --> 00:24:18,742
Ilia Baranov: Exactly, exactly.
It's the same thing. I'll go

421
00:24:18,742 --> 00:24:22,670
back to that example. You know,
the size of this equipment and,

422
00:24:22,670 --> 00:24:26,599
and the labor issues also kind
of caused this interesting story

423
00:24:26,599 --> 00:24:30,029
where we were talking to a
mining group about deploying

424
00:24:30,029 --> 00:24:33,708
autonomy on some of their ultra
class haul trucks. And, you

425
00:24:33,708 --> 00:24:37,075
know, we did the math on the
business case of how much

426
00:24:37,075 --> 00:24:40,443
they'll save their improved
reliability, those kind of

427
00:24:40,443 --> 00:24:44,372
things. And it came up to some
number, I forget what it is, but

428
00:24:44,372 --> 00:24:47,988
let's call it like the $5
million a year make up a number.

429
00:24:47,988 --> 00:24:51,917
They talked to us, talk to us,
and then they came back and they

430
00:24:51,917 --> 00:24:55,908
said, you know what? We're going
to put this on pause because we

431
00:24:55,908 --> 00:24:59,837
did the math that if we just put
a circuit in the thing to turn

432
00:24:59,837 --> 00:25:03,703
off the engine, if it's idling
like an idler, same thing you'd

433
00:25:03,703 --> 00:25:07,507
have in a car. Auto start, auto
stop, we'll save 30 dollars a

434
00:25:07,507 --> 00:25:10,999
year. Of fuel because the
drivers have been idling their

435
00:25:10,999 --> 00:25:14,866
vehicles, charging their phone.
Yeah, and air conditioning, of

436
00:25:14,866 --> 00:25:18,358
course. And so, you know,
there's, there's those kind of

437
00:25:18,358 --> 00:25:22,099
efficiencies where all of the
size and complexity has caused

438
00:25:22,099 --> 00:25:25,966
these weird distortions, where
you need this really heavy fuel

439
00:25:25,966 --> 00:25:29,520
for very specific, large
equipment that has to be shipped

440
00:25:29,520 --> 00:25:33,386
and parts and assembled on site
and lives its entire work life

441
00:25:33,386 --> 00:25:37,128
on site. I it's buried on site,
like all these kind of very,

442
00:25:37,128 --> 00:25:40,807
very weird, complicated things
just to reduce the amount of

443
00:25:40,807 --> 00:25:44,487
people at the tip. So if you
have a supply chain, you know,

444
00:25:44,487 --> 00:25:47,979
and do things have similar
concerns, right. The furthest

445
00:25:47,979 --> 00:25:51,533
deployed base or your furthest
deployed mines is the same

446
00:25:51,533 --> 00:25:55,212
problem. Every nut and bolt that
has to get there has to go

447
00:25:55,212 --> 00:25:58,892
through three stops to get
there, and getting more and more

448
00:25:58,892 --> 00:26:02,820
expensive every time. And so the
more you can reduce the amount

449
00:26:02,820 --> 00:26:06,500
of people you have at your
furthest point, the smaller your

450
00:26:06,500 --> 00:26:10,179
overall it gets. And so every
one driver we can save, we're

451
00:26:10,179 --> 00:26:14,045
saving probably ten ish people
that are moving food around and

452
00:26:14,045 --> 00:26:17,849
providing lodgings and doing air
and refueling. And so and we

453
00:26:17,849 --> 00:26:21,528
don't you know, Stefan and I
don't suspect that we'll get a

454
00:26:21,528 --> 00:26:25,270
fully like fully autonomous as
in no human beings, you know,

455
00:26:25,270 --> 00:26:28,762
there's no factories, for
example, they're coming on the

456
00:26:28,762 --> 00:26:32,379
lights out factories where
literally the overhead lighting

457
00:26:32,379 --> 00:26:36,308
is off because no human stepped
foot in it. This is happening a

458
00:26:36,308 --> 00:26:40,112
lot in China, for example, for
cell phone production. I think

459
00:26:40,112 --> 00:26:44,040
that's further along than what
outdoor mining robotics is right

460
00:26:44,040 --> 00:26:47,470
now, where you still need a
human to do maintenance and

461
00:26:47,470 --> 00:26:51,149
refueling and change the tires
and fix the gears. You know,

462
00:26:51,149 --> 00:26:54,891
that got jammed, for example,
all those things, you'll still

463
00:26:54,891 --> 00:26:58,445
need some level of human
support, but every one person we

464
00:26:58,445 --> 00:27:02,000
can strip increases the value of
their asset drastically.

465
00:27:02,000 --> 00:27:05,434
Audrow Nash: Yeah, the mine is
that pyramid you're mentioning.

466
00:27:05,434 --> 00:27:08,757
That's a really good metaphor
for it because like, you raise

467
00:27:08,757 --> 00:27:12,135
the top a little bit and then
the like, there's a significant

468
00:27:12,135 --> 00:27:15,458
side added which goes all the
way down. Yup. Support. People

469
00:27:15,458 --> 00:27:18,505
need support. People need
support. People like H.R. all

470
00:27:18,505 --> 00:27:20,000
the things and support and.

471
00:27:20,000 --> 00:27:21,000
Ilia Baranov: Supply cost.

472
00:27:21,000 --> 00:27:22,000
Audrow Nash: Complexity.

473
00:27:22,000 --> 00:27:25,658
Ilia Baranov: Yeah. But some of
these really far minds you know

474
00:27:25,658 --> 00:27:28,968
just to give you an idea,
especially in Australia, are so

475
00:27:28,968 --> 00:27:32,568
far that not only are the people
flown in and the spare parts,

476
00:27:32,568 --> 00:27:35,877
but there's fuel flown in.
Right. They fly in jet fuel to

477
00:27:35,877 --> 00:27:39,303
refuel the planes that are
flying. It's crazy. Right? Yeah,

478
00:27:39,303 --> 00:27:40,000
it's a hook.

479
00:27:40,000 --> 00:27:42,067
Audrow Nash: And they were like,
that's the cheapest way to do

480
00:27:42,067 --> 00:27:44,000
it. I think that's the most
efficient thing we thought of.

481
00:27:44,000 --> 00:27:46,476
Ilia Baranov: Yes. Wow. Yes.
Yeah. They're not they're not

482
00:27:46,476 --> 00:27:48,738
stupid right. No of course you
don't. You don't run a

483
00:27:48,738 --> 00:27:51,000
multi-billion dollar business by
not being efficient.

484
00:27:51,000 --> 00:27:54,614
Stefan Seltz-Axmacher: But it's
there's also like some

485
00:27:54,614 --> 00:27:58,630
interesting like game theory
that there's some impacts where

486
00:27:58,630 --> 00:28:02,713
if you're going to set up a mine
that size, it has to be that

487
00:28:02,713 --> 00:28:06,929
size like you have to you have
to rip the whole piece of ground

488
00:28:06,929 --> 00:28:10,811
apart, things that that extend
the life of the mine by six

489
00:28:10,811 --> 00:28:14,291
years but might be less
environmentally destructive.

490
00:28:14,291 --> 00:28:18,441
Just don't cut it. Whereas if
the labor becomes more fungible,

491
00:28:18,441 --> 00:28:22,524
if that that stops being the
stopping point. You can be a lot

492
00:28:22,524 --> 00:28:25,000
more strategic about where you
go at.

493
00:28:25,000 --> 00:28:27,818
Ilia Baranov: And you could be
more efficient generally.

494
00:28:27,818 --> 00:28:30,887
Resources gold, for example,
there's no you know, people have

495
00:28:30,887 --> 00:28:34,057
this image of like a guy with a
pickax knocking nuggets of gold

496
00:28:34,057 --> 00:28:36,925
off the wall, but that that
doesn't exist anymore. Right?

497
00:28:36,925 --> 00:28:40,044
It's like the geologist has told
us that there's one ounce per

498
00:28:40,044 --> 00:28:41,000
metric ton. We can.

499
00:28:41,000 --> 00:28:42,000
Audrow Nash: Find it.

500
00:28:42,000 --> 00:28:46,000
Ilia Baranov: Here. Or more like
0.1oz per metric ton.

501
00:28:46,000 --> 00:28:47,000
Audrow Nash: Yeah. And then
process it.

502
00:28:47,000 --> 00:28:50,419
Ilia Baranov: So yeah. You don't
find it, you just. Yeah. You

503
00:28:50,419 --> 00:28:53,895
just process metric ton after
metric ton after metric ton. And

504
00:28:53,895 --> 00:28:57,482
you need a certain size for that
again, because of that pyramid.

505
00:28:57,482 --> 00:29:00,845
But maybe there's another site
that is two ounces per metric

506
00:29:00,845 --> 00:29:04,096
ton, but it's only 50,000oz
total. And so there's no point

507
00:29:04,096 --> 00:29:07,684
in building a site there because
you're exhausted before you can

508
00:29:07,684 --> 00:29:11,103
ramp up, that all becomes more
possible with more automation,

509
00:29:11,103 --> 00:29:12,000
with less labor.

510
00:29:12,000 --> 00:29:15,884
Audrow Nash: Yeah, I think
that's a very cool. I mean, that

511
00:29:15,884 --> 00:29:19,768
does seem like generally how
technology diffuses. It's like

512
00:29:19,768 --> 00:29:23,455
it's starts super expensive,
very difficult. And then it

513
00:29:23,455 --> 00:29:27,536
becomes commoditized over time.
And then you get like web apps

514
00:29:27,536 --> 00:29:31,420
where you have payments as a
service and stuff like that is

515
00:29:31,420 --> 00:29:33,000
kind of the end of that.

516
00:29:33,000 --> 00:29:36,386
Ilia Baranov: Yeah. And I think,
I think for us, we're on this

517
00:29:36,386 --> 00:29:39,498
curve right now where we're,
we're, we're trying to build

518
00:29:39,498 --> 00:29:42,720
kind of a standard. I hate to
use the term operating system

519
00:29:42,720 --> 00:29:45,724
because it's a little bit
overused. Everybody has their

520
00:29:45,724 --> 00:29:49,164
own OS. But like we're trying to
build the basic autonomy layer

521
00:29:49,164 --> 00:29:52,167
and all the hardware abstraction
to interface with your

522
00:29:52,167 --> 00:29:55,498
autonomous machine, such that
then you can do it and go build

523
00:29:55,498 --> 00:29:58,829
the really fun apps, and then
either the client themselves or

524
00:29:58,829 --> 00:30:02,160
the or other developers like
Stefan is a good example on fuel

525
00:30:02,160 --> 00:30:05,000
versus yield usage. Which you
can't do, right? Yeah.

526
00:30:05,000 --> 00:30:08,783
Stefan Seltz-Axmacher: Like a
lot of what I, what I would but

527
00:30:08,783 --> 00:30:12,690
I often jokingly referred to as
like MBA porn about autonomy is

528
00:30:12,690 --> 00:30:16,535
not is pretty far disconnected
from point to point navigation.

529
00:30:16,535 --> 00:30:20,070
So like if the, the, the,
example Ilia was calling out is

530
00:30:20,070 --> 00:30:23,853
in a mind, you have a workspace
where you're getting material

531
00:30:23,853 --> 00:30:27,698
from, you have a process or you
need to dump the material. You

532
00:30:27,698 --> 00:30:31,357
can imagine that there's two
routes in between them. One is

533
00:30:31,357 --> 00:30:35,016
more, one is faster, but less
fuel efficient. Yeah, there's

534
00:30:35,016 --> 00:30:37,000
slower, but more fuel efficient.

535
00:30:37,000 --> 00:30:39,324
Audrow Nash: And so it's an
optimization problem to see

536
00:30:39,324 --> 00:30:40,000
which is better.

537
00:30:40,000 --> 00:30:43,729
Stefan Seltz-Axmacher: Not
today. And today you just drive

538
00:30:43,729 --> 00:30:47,714
the way you drive. Whereas in a
half day of work, a reasonable

539
00:30:47,714 --> 00:30:51,700
developer could code up language
that says, you know, based on

540
00:30:51,700 --> 00:30:55,557
the prevailing price of, or
based on the prevailing price of

541
00:30:55,557 --> 00:30:58,000
fuel, will route vehicles
accordingly.

542
00:30:58,000 --> 00:31:01,562
Ilia Baranov: And like or even
less you in less than two

543
00:31:01,562 --> 00:31:05,314
routes. It's just like your
engine has an efficiency point.

544
00:31:05,314 --> 00:31:09,258
And so if you drive 10% slower
and try getting an operator to.

545
00:31:09,258 --> 00:31:12,820
right, like that's like no
maliciousness. It's just as a

546
00:31:12,820 --> 00:31:16,000
human, you're not going to be
you're not gonna be.

547
00:31:16,000 --> 00:31:17,000
Audrow Nash: Yeah. And you'll be
impatient.

548
00:31:17,000 --> 00:31:21,000
Ilia Baranov: Two to oh, fuel
prices higher. 10% lower.

549
00:31:21,000 --> 00:31:22,000
Audrow Nash: Feel like I just
want to get it done.

550
00:31:22,000 --> 00:31:25,420
Ilia Baranov: Like even that
speed adjustment. Yeah. Is not

551
00:31:25,420 --> 00:31:28,957
is not possible. Whereas is if
you have an autonomy layer and

552
00:31:28,957 --> 00:31:32,319
you have a simple way to command
it that becomes basically

553
00:31:32,319 --> 00:31:34,000
trivial. Like that MBA could.

554
00:31:34,000 --> 00:31:36,000
Stefan Seltz-Axmacher: my code
with ChatGPT

555
00:31:36,000 --> 00:31:39,636
Ilia Baranov: Use ten lines of
code. Basically I've like look

556
00:31:39,636 --> 00:31:43,212
up fuel price, look up oil price
or price, subtract, set the

557
00:31:43,212 --> 00:31:45,000
speeds and see which is lower.

558
00:31:45,000 --> 00:31:46,000
Audrow Nash: Pick one.

559
00:31:46,000 --> 00:31:49,874
Stefan Seltz-Axmacher: And I
think like our kind of thesis

560
00:31:49,874 --> 00:31:53,883
right now about like what's
going on in the world. There's a

561
00:31:53,883 --> 00:31:57,891
lot of attention being sucked up
by humanoids, by foundation

562
00:31:57,891 --> 00:32:02,032
models for robots. None of that
matters compared to that. What

563
00:32:02,032 --> 00:32:06,241
we just talked about like a fuel
optimization, like what I just

564
00:32:06,241 --> 00:32:10,049
described for a company like
BHP, which is a $280 billion

565
00:32:10,049 --> 00:32:13,990
publicly shared mining company
that might individually give

566
00:32:13,990 --> 00:32:17,931
them 5 to $15 billion more year
profit, more profit a year,

567
00:32:17,931 --> 00:32:19,000
like, wow. Like.

568
00:32:19,000 --> 00:32:20,000
Audrow Nash: Did you say
billion?

569
00:32:20,000 --> 00:32:24,959
Stefan Seltz-Axmacher: Yes.
What? Yeah. Like humanoids don't

570
00:32:24,959 --> 00:32:30,000
matter. I'm like, this is not
like it's cool, it's neat, but.

571
00:32:30,000 --> 00:32:34,267
Ilia Baranov: Well, as asterisk.
Asterisk if BHP if BHP wants to

572
00:32:34,267 --> 00:32:38,200
increase their share price by
being a tech forward company,

573
00:32:38,200 --> 00:32:42,133
I'm sure they'll buy one of
every humanoid. Yeah. Make them

574
00:32:42,133 --> 00:32:46,333
dance a little bit, have their
stock price raise and that'll be

575
00:32:46,333 --> 00:32:49,000
enough. But the actual main
drivers are.

576
00:32:49,000 --> 00:32:53,565
Stefan Seltz-Axmacher: And like
we're in a state now where like

577
00:32:53,565 --> 00:32:57,841
for industry the things that
unlock that sort of efficiency

578
00:32:57,841 --> 00:33:02,116
hacking is, is now relatively
trivial. It's now like pretty

579
00:33:02,116 --> 00:33:06,464
straightforward, where you could
really just have a bunch of

580
00:33:06,464 --> 00:33:10,884
product manager types of people,
vibe code out solutions that

581
00:33:10,884 --> 00:33:14,000
drive very, very large gains in
efficiency.

582
00:33:14,000 --> 00:33:19,345
Audrow Nash: Since I am unsure
if this is terrifying or what,

583
00:33:19,345 --> 00:33:24,253
but that's crazy, I do. I see
what you're saying. What a

584
00:33:24,253 --> 00:33:29,686
thing. Yeah, so but it's really
interesting to put robotics in

585
00:33:29,686 --> 00:33:34,593
this context or a lot of
robotics applications that they

586
00:33:34,593 --> 00:33:39,325
are creating. Not that much for
savings. Whereas these

587
00:33:39,325 --> 00:33:44,495
efficiency gains might be,
hugely impactful, especially for

588
00:33:44,495 --> 00:33:48,000
these large companies. Yep.
That's wild.

589
00:33:48,000 --> 00:33:49,000
Ilia Baranov: But you need the
first.

590
00:33:49,000 --> 00:33:52,931
Audrow Nash: Because the
robotics give you the monitoring

591
00:33:52,931 --> 00:33:55,000
and the ability to actuate it.

592
00:33:55,000 --> 00:33:58,308
Ilia Baranov: And that's the
hill that a lot of robotics

593
00:33:58,308 --> 00:34:01,793
companies well enable in the
control. Oh yeah. So, so again

594
00:34:01,793 --> 00:34:05,337
another example by the way in
agriculture is another example

595
00:34:05,337 --> 00:34:09,058
in agriculture is like five ish
ten years ago. There's a ton of

596
00:34:09,058 --> 00:34:12,543
like egg data providers. We'll
fly a quadrotor around with.

597
00:34:12,543 --> 00:34:15,850
That was super popular and it
was great until it hit the

598
00:34:15,850 --> 00:34:19,453
barrier of like, okay, but I
have to do with you, like, okay,

599
00:34:19,453 --> 00:34:22,820
I know that this particular
point in the field is getting

600
00:34:22,820 --> 00:34:26,364
over watered. Want me to do
about it? Right. And so you need

601
00:34:26,364 --> 00:34:30,144
to actually close the loop with
a system that actually take that

602
00:34:30,144 --> 00:34:33,629
data into account and actually
change your behavior of your

603
00:34:33,629 --> 00:34:37,409
planting your why. Yeah. Because
remember. And so the same thing

604
00:34:37,409 --> 00:34:38,000
that it's.

605
00:34:38,000 --> 00:34:40,570
Audrow Nash: Just it's like
quantified self stuff where

606
00:34:40,570 --> 00:34:43,000
you're just measuring but you're
not changing stuff.

607
00:34:43,000 --> 00:34:46,952
Ilia Baranov: But you're not
changing anything And so again

608
00:34:46,952 --> 00:34:50,569
like it's you need you need
intelligent actors. And so

609
00:34:50,569 --> 00:34:54,455
again, this is kind of like AI
agents is the buzzword. But

610
00:34:54,455 --> 00:34:58,541
basically it's like you need
drivers in seats of devices that

611
00:34:58,541 --> 00:35:02,627
are better responded to data
enough that they can do 100% per

612
00:35:02,627 --> 00:35:06,177
minute of their speed, or
they're watering or they're

613
00:35:06,177 --> 00:35:10,129
planting or whatever, which
humans do, and that's the piece

614
00:35:10,129 --> 00:35:11,000
to unlock it.

615
00:35:11,000 --> 00:35:14,975
Stefan Seltz-Axmacher: So we
work with we're working with a

616
00:35:14,975 --> 00:35:18,816
pretty large AI conglomerate
that, you know, is large and

617
00:35:18,816 --> 00:35:22,589
secretive. And if I, if I, if I
named them, you wouldn't

618
00:35:22,589 --> 00:35:26,362
recognize them. But yeah, I
social name them, they, they

619
00:35:26,362 --> 00:35:30,539
harvest a specialty crop, and
they have, you know, a window of

620
00:35:30,539 --> 00:35:34,851
time three to 3 to 6 months, to
get all of it off the field. And

621
00:35:34,851 --> 00:35:38,826
every year, they do not know
whether they have three months

622
00:35:38,826 --> 00:35:43,004
or, you know, five and a half.
They don't know that on January

623
00:35:43,004 --> 00:35:46,979
1st. It depends on how rain
happens in the first part, like

624
00:35:46,979 --> 00:35:49,000
in the growing season, the of.

625
00:35:49,000 --> 00:35:51,000
Audrow Nash: Changes during that
time.

626
00:35:51,000 --> 00:35:54,500
Stefan Seltz-Axmacher: Changes.
Changes on a daily basis

627
00:35:54,500 --> 00:35:55,000
changes.

628
00:35:55,000 --> 00:35:57,878
Audrow Nash: So you want to
estimate basically and you want

629
00:35:57,878 --> 00:36:00,707
to do the intelligent thing
based on. Now how likely it is

630
00:36:00,707 --> 00:36:03,000
that not you have to get it out
quicker versus.

631
00:36:03,000 --> 00:36:06,446
Stefan Seltz-Axmacher: I mean
like they like they're a super

632
00:36:06,446 --> 00:36:09,835
cool company and super data
driven do a lot of really neat,

633
00:36:09,835 --> 00:36:13,568
analysis stuff. But like, as the
season happens, they'll start to

634
00:36:13,568 --> 00:36:16,957
get a sense for how long it's
going to be until crops start

635
00:36:16,957 --> 00:36:20,633
rotting in field. And we'll make
decisions on the fly about like

636
00:36:20,633 --> 00:36:24,194
we're going to speed all of our
harvesters up by 25%, or we're

637
00:36:24,194 --> 00:36:27,928
going to drop a by 10%, or we're
going to do this, or we're going

638
00:36:27,928 --> 00:36:31,489
to do that because that has a
very substantial yield, impacts.

639
00:36:31,489 --> 00:36:34,993
They also like have to jump from
one field to another because

640
00:36:34,993 --> 00:36:37,979
while they have hundreds of
thousands of acres under

641
00:36:37,979 --> 00:36:41,368
cultivation, those hundreds of
thousands aren't all next to

642
00:36:41,368 --> 00:36:42,000
each other.

643
00:36:42,000 --> 00:36:43,000
Audrow Nash: Over. Yeah.

644
00:36:43,000 --> 00:36:46,382
Stefan Seltz-Axmacher: So
there's this kind of big, you

645
00:36:46,382 --> 00:36:50,072
know, traveling salesman, multi
level optimization game that

646
00:36:50,072 --> 00:36:54,008
they have to play. And like some
part of that is yell at guys on

647
00:36:54,008 --> 00:36:57,390
the phone with limited with
limited language, to change

648
00:36:57,390 --> 00:37:00,896
their operations in real time.
And that is the thing that

649
00:37:00,896 --> 00:37:04,647
robots can very quickly do and
like, we think it's reasonable

650
00:37:04,647 --> 00:37:08,275
to think that that could lead to
10 to 20% more, more crops

651
00:37:08,275 --> 00:37:11,842
successfully getting off the
field and also like even like

652
00:37:11,842 --> 00:37:15,471
down to the minute where that
equipment might be in a field

653
00:37:15,471 --> 00:37:18,853
and it might leave that field
before the field is fully

654
00:37:18,853 --> 00:37:22,604
harvested. I because they have
to go to another field. But if

655
00:37:22,604 --> 00:37:26,233
you took in the data of the
truck that's going to move that

656
00:37:26,233 --> 00:37:29,861
equipment from one field to
another, and you took them that

657
00:37:29,861 --> 00:37:33,366
location to, to do a live,
estimated time of arrival, you

658
00:37:33,366 --> 00:37:37,302
could theoretically be operating
until just the moment that that

659
00:37:37,302 --> 00:37:40,992
truck arrives. You could change
the road that you harvest to

660
00:37:40,992 --> 00:37:44,866
make you be as close as possible
to. Where are you going to get

661
00:37:44,866 --> 00:37:48,556
loaded into that truck, when it
arrives? Because that itself

662
00:37:48,556 --> 00:37:52,000
might mean they get 2 to 5% more
products off the field.

663
00:37:52,000 --> 00:37:55,000
Audrow Nash: Which makes a huge
impact in terms of total sales.

664
00:37:55,000 --> 00:37:58,588
Stefan Seltz-Axmacher: If your
costs and your fixed costs are

665
00:37:58,588 --> 00:38:01,941
the same, but you suddenly
increase revenue by 4% through

666
00:38:01,941 --> 00:38:05,000
some data science shenanigans,
that's the ball game.

667
00:38:05,000 --> 00:38:07,000
Ilia Baranov: Yeah, that's your
entire profit margin. Basically.

668
00:38:07,000 --> 00:38:10,677
Audrow Nash: That's wild. Or
it's I haven't thought about

669
00:38:10,677 --> 00:38:11,000
this.

670
00:38:11,000 --> 00:38:13,000
Stefan Seltz-Axmacher: Yeah.

671
00:38:13,000 --> 00:38:16,094
Audrow Nash: It's such an
interesting way to view

672
00:38:16,094 --> 00:38:19,692
everything, and it makes a lot
of sense to me. And yet so

673
00:38:19,692 --> 00:38:22,786
there's tons of these big
companies and these big

674
00:38:22,786 --> 00:38:25,943
companies have very large
inefficiencies. And it's

675
00:38:25,943 --> 00:38:29,731
partially just because people
are in the loop and people are

676
00:38:29,731 --> 00:38:33,392
not you can't control people
that well. Yeah. Robotics has

677
00:38:33,392 --> 00:38:36,738
covered with the or like the
quadrotor stuff you were

678
00:38:36,738 --> 00:38:40,590
mentioning where they do a bunch
of sensing. It's like, okay,

679
00:38:40,590 --> 00:38:44,251
sensing is not the only thing
you need, because I remember

680
00:38:44,251 --> 00:38:47,850
this huge wave of quadrotor
companies that had monitoring

681
00:38:47,850 --> 00:38:51,701
and they almost entirely have
gone out of business, you know,

682
00:38:51,701 --> 00:38:55,489
and I think that is exactly why
what you're saying, which is

683
00:38:55,489 --> 00:38:59,088
they couldn't change anything
easily. And so this kind of

684
00:38:59,088 --> 00:39:02,939
thing, when you add robots and
the robots make it possible to

685
00:39:02,939 --> 00:39:06,917
identify these things, you can
make better. And then to quickly

686
00:39:06,917 --> 00:39:09,000
put in practice whatever is
best.

687
00:39:09,000 --> 00:39:10,000
Stefan Seltz-Axmacher: Yep.

688
00:39:10,000 --> 00:39:12,000
Audrow Nash: That's incredible.

689
00:39:12,000 --> 00:39:15,315
Stefan Seltz-Axmacher: Yeah. I
mean, like this AI group does

690
00:39:15,315 --> 00:39:18,630
like studies on how much more
efficiently they can harvest a

691
00:39:18,630 --> 00:39:21,724
field if they eliminate left
turns. But yeah, humans for

692
00:39:21,724 --> 00:39:22,000
that.

693
00:39:22,000 --> 00:39:23,000
Audrow Nash: Yeah.

694
00:39:22,000 --> 00:39:25,980
Stefan Seltz-Axmacher: Yeah,
yeah. Then like, it's, but like,

695
00:39:25,980 --> 00:39:30,222
if you're, if you're ups and you
want to try that out, you need a

696
00:39:30,222 --> 00:39:33,811
lot of data before you can
slowly roll it out. And then

697
00:39:33,811 --> 00:39:37,269
there's a lot of change
management to get rid of left

698
00:39:37,269 --> 00:39:41,054
turns. Whereas yeah, if you
like, like we, we for example,

699
00:39:41,054 --> 00:39:44,773
were bidding on a program. I
think this is obscure enough

700
00:39:44,773 --> 00:39:48,754
and, and, and unspecific enough
that it's fine to talk really

701
00:39:48,754 --> 00:39:52,473
openly about. So we're bidding
on a program to automate a

702
00:39:52,473 --> 00:39:56,192
vehicle in an asphalt lab. And
basically they put asphalt

703
00:39:56,192 --> 00:39:59,912
outside in a ring, and, and
drive a truck with about your

704
00:39:59,912 --> 00:40:02,000
weight over it. This particular.

705
00:40:02,000 --> 00:40:04,000
Audrow Nash: They look at tire
wear or what do they do.

706
00:40:04,000 --> 00:40:06,000
Stefan Seltz-Axmacher: Think
primarily where on the asphalt.

707
00:40:06,000 --> 00:40:09,211
Audrow Nash: The road wear on
the asphalt. Yeah, that's a

708
00:40:09,211 --> 00:40:12,761
frickin lariats to just make a
circle of asphalt and then drive

709
00:40:12,761 --> 00:40:14,000
on it until there's a.

710
00:40:14,000 --> 00:40:18,367
Stefan Seltz-Axmacher: Yeah. So
the driving super variable. Yep.

711
00:40:18,367 --> 00:40:22,665
The people are variable. There's
a truck that's been driving on

712
00:40:22,665 --> 00:40:26,691
this lab, 40 hours a week for
the last 40 years. Right. But

713
00:40:26,691 --> 00:40:30,375
like, like if you want to
experiment, in that lab, you

714
00:40:30,375 --> 00:40:34,742
need to write a test plan, that
says, okay, for the next for the

715
00:40:34,742 --> 00:40:38,904
next six months, I want you to
drive at 45 miles an hour, and

716
00:40:38,904 --> 00:40:43,203
do a hard braking event every 15
minutes or something. And then

717
00:40:43,203 --> 00:40:46,000
you need to coach the person to
actually.

718
00:40:46,000 --> 00:40:48,000
Audrow Nash: Do that.

719
00:40:48,000 --> 00:40:52,018
Stefan Seltz-Axmacher: Whereas,
if you, if that vehicle suddenly

720
00:40:52,018 --> 00:40:56,100
autonomous, you could, you could
operate that testing environment

721
00:40:56,100 --> 00:40:59,804
like you would medical testing
equipment where you could be

722
00:40:59,804 --> 00:41:03,760
driving 24/7 there's a blizzard
going on and you wouldn't want,

723
00:41:03,760 --> 00:41:07,339
you know, no humans should risk
their life. Driving in an

724
00:41:07,339 --> 00:41:10,855
asphalt lab during a blizzard, I
think is a easy, easily

725
00:41:10,855 --> 00:41:14,873
agreeable statement. Okay. But,
like, you could have the vehicle

726
00:41:14,873 --> 00:41:18,139
do erratic braking, in a
blizzard nonstop. You could

727
00:41:18,139 --> 00:41:21,780
have, I could be running one
experiment, where I get every

728
00:41:21,780 --> 00:41:25,485
raining day of the year, and you
could be doing a different

729
00:41:25,485 --> 00:41:28,875
naturalist experiment where
we're driving differently,

730
00:41:28,875 --> 00:41:32,392
during daylight hours for
Aurora's experiment. And maybe

731
00:41:32,392 --> 00:41:36,033
Ilia is doing yet a different
experiment about how long it

732
00:41:36,033 --> 00:41:39,863
takes for there to be a change
in the coefficient of friction

733
00:41:39,863 --> 00:41:43,505
after it's finished raining.
And, and it's starting to dry

734
00:41:43,505 --> 00:41:47,021
out. And like, those three tests
can very much be run in

735
00:41:47,021 --> 00:41:50,789
parallel. Those three tests can
very much be orchestrated by

736
00:41:50,789 --> 00:41:54,870
piece of software. They can't be
orchestrated by a person yelling

737
00:41:54,870 --> 00:41:56,000
at a truck driver.

738
00:41:56,000 --> 00:41:58,952
Audrow Nash: Yeah. Very
interesting. Yeah. Because, I

739
00:41:58,952 --> 00:42:01,848
mean, basically what I'm
imagining is you can tell a

740
00:42:01,848 --> 00:42:04,967
truck driver, you can tell a
person to do these kinds of

741
00:42:04,967 --> 00:42:08,531
things. Yeah. The thing is, the
compliance will be somewhat low.

742
00:42:08,531 --> 00:42:11,705
Yeah. And you'll see a lot of
noise introduced because of

743
00:42:11,705 --> 00:42:15,047
that. Whereas you can have the
robot just follow it and then

744
00:42:15,047 --> 00:42:18,166
you can more easily. It's like
the standard deviation of

745
00:42:18,166 --> 00:42:21,173
different outcomes, you see
decreases, it becomes much

746
00:42:21,173 --> 00:42:24,069
pointier, much smaller
distribution or much narrower

747
00:42:24,069 --> 00:42:27,578
distribution. Yeah. Okay. That's
super cool. I love the idea of

748
00:42:27,578 --> 00:42:31,087
being able to command tests like
that. So you can test. And you

749
00:42:31,087 --> 00:42:34,484
can also, especially if it's
trivial to set up the tests, you

750
00:42:34,484 --> 00:42:38,104
can run more of them. So you get
more data points, you understand

751
00:42:38,104 --> 00:42:41,000
more things and then you can
optimize significantly.

752
00:42:41,000 --> 00:42:44,445
Stefan Seltz-Axmacher: I, I felt
comfortable saying, yeah, what

753
00:42:44,445 --> 00:42:47,836
we're bidding for is because to
my knowledge, there's like 100

754
00:42:47,836 --> 00:42:51,281
of these labs in America. Like,
I, I probably didn't say nearly

755
00:42:51,281 --> 00:42:54,781
anything that could lead you to
guess where were benefiting that

756
00:42:54,781 --> 00:42:55,000
far.

757
00:42:55,000 --> 00:42:57,583
Ilia Baranov: Well, we know it's
not in Florida because of the

758
00:42:57,583 --> 00:42:58,000
blizzards.

759
00:42:58,000 --> 00:43:01,980
Stefan Seltz-Axmacher: but like,
there's in Florida, for example,

760
00:43:01,980 --> 00:43:05,714
other than at least 4 or 5 of
these sorts of facilities where

761
00:43:05,714 --> 00:43:09,510
it's a test track for investing,
blah, blah, blah. And I and a

762
00:43:09,510 --> 00:43:13,306
you don't need as many of them
if you can get 150 hours a week

763
00:43:13,306 --> 00:43:16,000
of testing, as opposed to 35 or
40 for sure.

764
00:43:16,000 --> 00:43:20,145
Ilia Baranov: But I think I
think coming back to the meta

765
00:43:20,145 --> 00:43:24,727
point, though, like, I think I
think the storyline here of what

766
00:43:24,727 --> 00:43:28,873
we're doing is a lot of what I
would kind of in quotation

767
00:43:28,873 --> 00:43:33,018
marks, but basic autonomy of
moving a vehicle from A to B

768
00:43:33,018 --> 00:43:37,164
safely and successfully, and
doing so in a way that makes

769
00:43:37,164 --> 00:43:41,455
business sense and that seems
boring in a lot of ways. Like

770
00:43:41,455 --> 00:43:45,964
all of that to most people. We
talk to you though, like, yeah,

771
00:43:45,964 --> 00:43:50,400
of course like Waymo does it
like y, you know, how hard is it

772
00:43:50,400 --> 00:43:54,400
to do it in a, in a space or,
you know, the DARPA Grand

773
00:43:54,400 --> 00:43:58,764
Challenge did it ten years ago.
What's the problem? And it's

774
00:43:58,764 --> 00:44:03,273
similar kind of discussion to,
you know, one, one computer ran

775
00:44:03,273 --> 00:44:07,564
once, like there is a D.O.D.
computer in the 70s that could

776
00:44:07,564 --> 00:44:11,491
run a simulation. So there's
this kind of trickle down

777
00:44:11,491 --> 00:44:12,000
effect.

778
00:44:12,000 --> 00:44:15,855
Stefan Seltz-Axmacher: I played
Grand Theft Auto once. Why can't

779
00:44:15,855 --> 00:44:17,000
The matrix be real?

780
00:44:17,000 --> 00:44:20,208
Ilia Baranov: Yeah. Yeah.
Exactly. That's a similar

781
00:44:20,208 --> 00:44:24,233
discussion. Right. So I think I
think tier two point, they asked

782
00:44:24,233 --> 00:44:27,881
at the start of the thing, I
think there's a lot of effort

783
00:44:27,881 --> 00:44:31,404
and hype right now in AI and
humanoids and those sort of

784
00:44:31,404 --> 00:44:35,429
things. But I think if you look
back on history, the lag between

785
00:44:35,429 --> 00:44:39,392
that being interesting and that
mean, being a significant chunk

786
00:44:39,392 --> 00:44:42,788
of the market is much bigger
than people think, almost

787
00:44:42,788 --> 00:44:46,436
always. Right? Like the lag
between, you know, smartphones

788
00:44:46,436 --> 00:44:50,084
appearing as a concept and
smartphones being common, which

789
00:44:50,084 --> 00:44:54,047
was one of the fastest ever, was
still like six years. I'd have

790
00:44:54,047 --> 00:44:57,947
to go fast seven for everything,
right? Right, right. But tell

791
00:44:57,947 --> 00:45:01,783
that to a VC like that's the
fastest tech ever, right? Here's

792
00:45:01,783 --> 00:45:05,620
my humanoid, which costs 30 to
$120,000. The lag you're going

793
00:45:05,620 --> 00:45:09,520
to have from this to in every
home is probably in the order of

794
00:45:09,520 --> 00:45:13,482
12 to 20 years. So please invest
$1 billion a year for the next

795
00:45:13,482 --> 00:45:17,256
20 years and maybe will be one
of the winners versus kind of

796
00:45:17,256 --> 00:45:21,156
Stefan and I have taken a much
more cash efficient approach of

797
00:45:21,156 --> 00:45:24,553
like, where are the most
expensive pain points that we

798
00:45:24,553 --> 00:45:28,578
can automate for the most amount
of money, with the least amount

799
00:45:28,578 --> 00:45:32,352
of technical effort and risk now
because we're lazy decision

800
00:45:32,352 --> 00:45:36,000
because like, that's how you
actually run and efficiently.

801
00:45:36,000 --> 00:45:39,034
Audrow Nash: I view it as like,
why do you need to make life

802
00:45:39,034 --> 00:45:41,916
harder? Yeah, just do the
valuable thing. Like, yeah, you

803
00:45:41,916 --> 00:45:45,000
don't need to play life on hard
mode. It's already difficult.

804
00:45:45,000 --> 00:45:46,916
Stefan Seltz-Axmacher:
Especially when you, when you,

805
00:45:46,916 --> 00:45:48,000
when your life is with robots.

806
00:45:48,000 --> 00:45:52,651
Audrow Nash: Yes. Totally. Yeah.
Let's see. Yeah. Ellie, I really

807
00:45:52,651 --> 00:45:57,159
like your point with all that. I
wonder. Let's see, I guess. So

808
00:45:57,159 --> 00:46:01,524
with all of this, what I think
is a lot of times people focus

809
00:46:01,524 --> 00:46:05,674
on the real sexy applications
and things like this, or the

810
00:46:05,674 --> 00:46:09,753
sexy demos. But I've been
believing increasingly that the

811
00:46:09,753 --> 00:46:13,975
majority of the value is in
really unsexy tasks, and people

812
00:46:13,975 --> 00:46:18,555
don't get as excited for it. But
the thing is, it's like budget.

813
00:46:18,555 --> 00:46:22,848
That's where you can make the
biggest gains. Yeah, for these

814
00:46:22,848 --> 00:46:26,927
kinds of things. And like start
to realize the promise of

815
00:46:26,927 --> 00:46:28,000
robotics. Yeah.

816
00:46:28,000 --> 00:46:32,079
Stefan Seltz-Axmacher: So and I
think there's a challenge right

817
00:46:32,079 --> 00:46:36,029
now because in a way you can
think about there being like two

818
00:46:36,029 --> 00:46:39,785
different robotics markets.
There there's companies that I

819
00:46:39,785 --> 00:46:43,929
think of as like hyper robotics
where the, the, the, the company

820
00:46:43,929 --> 00:46:47,750
exists because they're doing
cool stuff, the money, because

821
00:46:47,750 --> 00:46:51,635
the thought is they could do
more cool stuff eventually. You

822
00:46:51,635 --> 00:46:55,326
know, cool stuff has to be
valuable, right? So if we wait

823
00:46:55,326 --> 00:46:59,470
long enough, it will eventually
do something. And then there's a

824
00:46:59,470 --> 00:47:03,420
it's a thought. Yeah, it's a
thought. It's a thesis. And then

825
00:47:03,420 --> 00:47:06,917
there's companies who are
building, you know, robotics

826
00:47:06,917 --> 00:47:10,737
companies similar to how you
would build a web app business

827
00:47:10,737 --> 00:47:14,364
or a SaaS company or an
e-commerce site where it's like,

828
00:47:14,364 --> 00:47:18,249
here's a, here's a, a problem
that the market has that I can

829
00:47:18,249 --> 00:47:22,328
reasonably build a solution for
with a seed round. And now that

830
00:47:22,328 --> 00:47:26,408
I built a solution for them, I
can get them to pay enough money

831
00:47:26,408 --> 00:47:30,552
that there's a line of sight to
where I take in more money every

832
00:47:30,552 --> 00:47:34,308
year than I spend. And there's
this. And I'm not just, you

833
00:47:34,308 --> 00:47:38,258
know, talking my own book here.
There's a couple companies in

834
00:47:38,258 --> 00:47:42,143
the solar market, doing that.
There's um, is a great example

835
00:47:42,143 --> 00:47:46,222
of that. And I, there's a bunch
of that stuff, and, and I think

836
00:47:46,222 --> 00:47:49,784
that's taken me a while to
realize is that jumping from

837
00:47:49,784 --> 00:47:53,734
hype over to realistic is like
fraught with peril. And, and I

838
00:47:53,734 --> 00:47:56,000
don't know of many, many,
robotics.

839
00:47:56,000 --> 00:47:58,000
Audrow Nash: Companies that have
successfully made it.

840
00:47:58,000 --> 00:48:01,963
Stefan Seltz-Axmacher: Yeah.
Because like, if you're worth

841
00:48:01,963 --> 00:48:05,925
$10 billion off of hype, like,
like cruise, for example, I

842
00:48:05,925 --> 00:48:09,888
think at peak was worth $30
billion from hype. But when no

843
00:48:09,888 --> 00:48:14,261
one was willing to say that our
worth $40 billion, that all they

844
00:48:14,261 --> 00:48:18,292
had to fall back on was a robo
taxi business that was maybe

845
00:48:18,292 --> 00:48:20,000
making $4 million a year.

846
00:48:20,000 --> 00:48:21,000
Audrow Nash: Yeah, super low.

847
00:48:20,000 --> 00:48:23,182
Stefan Seltz-Axmacher: That $4
million a year or that $1

848
00:48:23,182 --> 00:48:26,648
million a year was never going
to cover their 3000 engineers.

849
00:48:26,648 --> 00:48:30,000
Huge burn rate, half $1 billion
a quarter burn or whatever.

850
00:48:30,000 --> 00:48:32,711
Audrow Nash: What do you think
of, like, Waymo with this? All

851
00:48:32,711 --> 00:48:35,467
right. Because they're I mean,
they're still around. Yeah. And

852
00:48:35,467 --> 00:48:38,000
I hear people like them quite a
bit. And I think they're.

853
00:48:38,000 --> 00:48:42,878
Stefan Seltz-Axmacher: Waymo's
really cool. Go fund it. I'm

854
00:48:42,878 --> 00:48:48,004
really I'm really rooting for
it. But let me let me put you in

855
00:48:48,004 --> 00:48:53,130
a scenario right. You are the
CFO of alphabet, right? You have

856
00:48:53,130 --> 00:48:58,421
an equity plan with 1% or so of
alphabet. So I haven't looked at

857
00:48:58,421 --> 00:49:03,382
I don't, I don't like, follow
public companies that closely,

858
00:49:03,382 --> 00:49:08,591
but one, I think they're worth
$1 trillion, maybe two. Easy $10

859
00:49:08,591 --> 00:49:13,551
billion, of an equity incentive
plan. If, you know, the hype

860
00:49:13,551 --> 00:49:18,429
around robotaxis is, is up and
down. But let's say some bad

861
00:49:18,429 --> 00:49:23,224
thing happens to Waymo. Public
bad thing happens. And that

862
00:49:23,224 --> 00:49:28,020
brings down the Waymo stock 5%,
even if it's just 5% for a

863
00:49:28,020 --> 00:49:33,146
month. So that means a loss of,
$50 billion in value. And it's

864
00:49:33,146 --> 00:49:37,610
right around the time when
you're being asked to write

865
00:49:37,610 --> 00:49:41,000
another 5 to $10 billion check
for Waymo.

866
00:49:41,000 --> 00:49:44,077
Audrow Nash: Yeah. You just saw
a 50 billion slide. You're like,

867
00:49:44,077 --> 00:49:46,000
even more than that, this kind
of thing.

868
00:49:46,000 --> 00:49:50,934
Stefan Seltz-Axmacher: If you
Audrow the CFO of alphabet, take

869
00:49:50,934 --> 00:49:55,869
that moment to decide. You know,
I'm going to cut Waymo. We're

870
00:49:55,869 --> 00:50:00,803
going to publicly announce it.
The stock's going to go down to

871
00:50:00,803 --> 00:50:05,578
5% and go up another 10%. The
company alphabet will suddenly

872
00:50:05,578 --> 00:50:09,000
be worth $150 billion more. And
I as Audrow

873
00:50:09,000 --> 00:50:10,000
Audrow Nash: I’m responsible.
Yeah.

874
00:50:09,000 --> 00:50:13,293
Stefan Seltz-Axmacher: No no no
no no nothing that screw your

875
00:50:13,293 --> 00:50:17,657
responsibility who you are you.
Yeah. As Audrow, we'll make an

876
00:50:17,657 --> 00:50:21,880
extra $1.5 billion. Yeah. Would,
Audrow would you keep Waymo

877
00:50:21,880 --> 00:50:26,103
alive for as cool as it is and
how impactful could be in how

878
00:50:26,103 --> 00:50:30,326
many lives it could save? Four
and $1.5 billion. Could I pay

879
00:50:30,326 --> 00:50:33,000
you $1.5 billion a couple? Look,
look.

880
00:50:33,000 --> 00:50:36,614
Audrow Nash: I see how the
incentives go. Yeah, yeah.

881
00:50:36,614 --> 00:50:39,000
That's crazy. Yeah, it makes
sense.

882
00:50:39,000 --> 00:50:41,540
Stefan Seltz-Axmacher: I, I
think it's an amazing product. I

883
00:50:41,540 --> 00:50:43,741
think it's really cool
technology. I really hope the

884
00:50:43,741 --> 00:50:46,450
situation I just described never
comes to a pass, but I have yet

885
00:50:46,450 --> 00:50:47,000
to play that.

886
00:50:47,000 --> 00:50:49,000
Audrow Nash: Game clear with
anyone. This kind of thing.

887
00:50:49,000 --> 00:50:52,780
Stefan Seltz-Axmacher: Who would
keep it personally alive. For

888
00:50:52,780 --> 00:50:56,073
what if I offer them $1.5
billion to be the person who

889
00:50:56,073 --> 00:50:59,000
signed a piece of paper to do
less work to exit?

890
00:50:59,000 --> 00:51:01,000
Audrow Nash: Yeah, I know, to do
less work. Yeah. That's wild.

891
00:51:01,000 --> 00:51:03,000
Stefan Seltz-Axmacher: Yeah.

892
00:51:03,000 --> 00:51:04,000
Audrow Nash: What a thing.

893
00:51:04,000 --> 00:51:08,274
Stefan Seltz-Axmacher: Yeah. So
I really hope that I'm wrong. I

894
00:51:08,274 --> 00:51:12,344
really hope it's alive. But I
think the challenge with these

895
00:51:12,344 --> 00:51:16,143
like hype programs and the
challenge with these like big

896
00:51:16,143 --> 00:51:19,467
companies is they look
incredibly stable from the

897
00:51:19,467 --> 00:51:23,808
outside because very few people
are in the rooms where like it's

898
00:51:23,808 --> 00:51:27,743
like, hey guys, it's not it's
not just that audio make 1.5

899
00:51:27,743 --> 00:51:31,745
billion Ilia Also make 1.5
billion and Stefan and I'll make

900
00:51:31,745 --> 00:51:36,154
it easy 2.2 billion and I can be
pretty good with that situation.

901
00:51:36,154 --> 00:51:40,292
Yeah for sure. And yeah like
we've seen, we, you know, aliens

902
00:51:40,292 --> 00:51:44,227
have been around, for long
enough to see a bunch of really

903
00:51:44,227 --> 00:51:48,365
good robotics companies with
really cool tech died because of

904
00:51:48,365 --> 00:51:52,571
assholes in finance who can get
a good bonus check. And that's

905
00:51:52,571 --> 00:51:56,573
just a scary reality, which
means, like, the faster you can

906
00:51:56,573 --> 00:52:00,575
get to a self-sustaining robot,
the more realistic it is to

907
00:52:00,575 --> 00:52:02,000
actually have robots.

908
00:52:02,000 --> 00:52:05,941
Audrow Nash: Yeah, yeah, it has
to make sense financially. And

909
00:52:05,941 --> 00:52:09,564
it's very hard to jump that
chasm from hype to practical.

910
00:52:09,564 --> 00:52:13,568
Yeah. For this kind of thing. Do
you, I'm interested in both of

911
00:52:13,568 --> 00:52:17,508
your perspectives. Do you think
any of the companies will make

912
00:52:17,508 --> 00:52:21,322
it out of the humanoid bubble?
Like, will we see any winners

913
00:52:21,322 --> 00:52:25,008
from it, or will we? Well, I
mean, will Tesla quietly chug

914
00:52:25,008 --> 00:52:28,631
away at it and solve it because
they have the whole Tesla

915
00:52:28,631 --> 00:52:32,000
business? Or what do you think?
Maybe start with Ilia

916
00:52:32,000 --> 00:52:34,876
Ilia Baranov: Well, I mean, I
think you got to you got to

917
00:52:34,876 --> 00:52:38,055
define what you mean by winner,
right? Like, is Atlas is Boston

918
00:52:38,055 --> 00:52:41,032
Dynamics a winner? Because I
eventually got traded around a

919
00:52:41,032 --> 00:52:43,000
bunch of times and ended up at
Hyundai.

920
00:52:43,000 --> 00:52:47,239
Audrow Nash: Say they make that
they jump from the they jump

921
00:52:47,239 --> 00:52:51,054
across that chasm from hype to
reality. Where they are

922
00:52:51,054 --> 00:52:54,587
providing value that is
justifying the cost of the

923
00:52:54,587 --> 00:52:56,000
businesses business.

924
00:52:56,000 --> 00:52:59,134
Stefan Seltz-Axmacher: Well, I
think I think also what you're

925
00:52:59,134 --> 00:53:02,115
really trying to say is they
make more money. I will well,

926
00:53:02,115 --> 00:53:05,146
there be any company who makes
more money selling humanoids

927
00:53:05,146 --> 00:53:07,870
than they spend as an
organization. Exactly. Not like

928
00:53:07,870 --> 00:53:10,696
his Ilia retort was Boston
Dynamics has made a bunch of

929
00:53:10,696 --> 00:53:13,830
money, and a bunch of investors
have made a bunch of money in

930
00:53:13,830 --> 00:53:17,119
Boston Dynamics without actually
shipping robots at scale. Which

931
00:53:17,119 --> 00:53:20,253
should be fair, is a great trip
that I a great trick that I'd

932
00:53:20,253 --> 00:53:22,000
love to be able to pull off.
Yeah.

933
00:53:22,000 --> 00:53:25,843
Ilia Baranov: Yeah. No, I mean,
you know, I, I don't know the

934
00:53:25,843 --> 00:53:29,435
internal numbers for something
like Unitree, but they're,

935
00:53:29,435 --> 00:53:33,404
they're one of the few who are
seemingly shipping and volume at

936
00:53:33,404 --> 00:53:37,122
reasonable prices. And I doubt
their I doubt they're wildly

937
00:53:37,122 --> 00:53:40,713
profitable, but I also doubt
that they're losing money on

938
00:53:40,713 --> 00:53:44,305
every Unitree robot they ship.
It's like they're probably

939
00:53:44,305 --> 00:53:47,959
because, you know, the trick
they also do is like, I don't

940
00:53:47,959 --> 00:53:51,929
remember exactly, exactly. But
like the humanoid robot, there's

941
00:53:51,929 --> 00:53:53,000
the 30,000 rises.

942
00:53:53,000 --> 00:53:54,000
Audrow Nash: Significant.

943
00:53:53,000 --> 00:53:56,555
Ilia Baranov: And if you
actually do the one with it.

944
00:53:56,555 --> 00:53:59,909
Yeah. Like the model that
actually has sensors and

945
00:53:59,909 --> 00:54:04,000
compute. It's like 120, right?
Like the 30,000 is a RC burns.

946
00:54:04,000 --> 00:54:07,000
Audrow Nash: Out when you look
at it.

947
00:54:07,000 --> 00:54:11,701
Ilia Baranov: Right. So, so, so,
you know, I think if I had to

948
00:54:11,701 --> 00:54:16,403
guess, you know, on the $30,000
model, they're spending 28 and

949
00:54:16,403 --> 00:54:21,104
on the $120,000 model, they're
spending 100 and like something

950
00:54:21,104 --> 00:54:23,000
110 maybe. Right. So low.

951
00:54:23,000 --> 00:54:24,000
Audrow Nash: Margins maybe.

952
00:54:24,000 --> 00:54:27,477
Ilia Baranov: They're making
money. Have they made back the

953
00:54:27,477 --> 00:54:30,955
money they invested in it.
Probably not yet. But again I, I

954
00:54:30,955 --> 00:54:34,668
just don't see the industry as a
company that's willing to just

955
00:54:34,668 --> 00:54:38,145
burn money for no reason. So
they're probably the closest I

956
00:54:38,145 --> 00:54:41,682
think all the other kind of
startups are deep, deep, deep in

957
00:54:41,682 --> 00:54:44,923
the red, including Tesla
Optimus. Right. Like those are

958
00:54:44,923 --> 00:54:48,460
all into until your robot shows
up on Robot Shop and you can

959
00:54:48,460 --> 00:54:51,819
swipe your credit card and buy
now, like I don't. I don't

960
00:54:51,819 --> 00:54:54,000
suspect that they're quite there
yet.

961
00:54:54,000 --> 00:54:55,000
Audrow Nash: Make sense.

962
00:54:55,000 --> 00:54:58,659
Stefan Seltz-Axmacher: I have,
I, so someone told me that they

963
00:54:58,659 --> 00:55:02,024
knew of someone doing the
strategy, but I have a strategy

964
00:55:02,024 --> 00:55:05,742
that I think would work in the
humanoids which I think of as, I

965
00:55:05,742 --> 00:55:09,224
call it the minimally viable
humanoid. And to be clear, you

966
00:55:09,224 --> 00:55:12,825
have to have you have to be big
enough to build two robots to

967
00:55:12,825 --> 00:55:16,071
support the strategy. So, your
first robot is your cool

968
00:55:16,071 --> 00:55:19,848
humanoid, and your cool humanoid
has legs. It has hands. It has,

969
00:55:19,848 --> 00:55:23,390
head. It looks like a head.
Maybe there's little bit of face

970
00:55:23,390 --> 00:55:26,813
LEDs. There's a bunch of sensors
in it. There's a bunch of

971
00:55:26,813 --> 00:55:30,000
compute on board. It can do
really, really cool stuff.

972
00:55:30,000 --> 00:55:32,000
Audrow Nash: Yeah, it's the race
car. Yeah.

973
00:55:32,000 --> 00:55:35,416
Stefan Seltz-Axmacher: For sure.
Cost $3 million. But it's in

974
00:55:35,416 --> 00:55:38,776
every press release. The, the
robot along side of it, is the

975
00:55:38,776 --> 00:55:42,080
minimally viable humanoid which
is on wheels. Might even be

976
00:55:42,080 --> 00:55:45,384
tethered. Sure. There's probably
a sensor cluster, but that

977
00:55:45,384 --> 00:55:48,968
sensor cluster isn't necessarily
around. It's honestly in a head

978
00:55:48,968 --> 00:55:52,160
shape thing. It's kind of
wherever the sensors should be.

979
00:55:52,160 --> 00:55:55,632
There might be two arms there.
I'd only be one. The arms might

980
00:55:55,632 --> 00:55:58,992
have fingers. They might have
something. It looks more like,

981
00:55:58,992 --> 00:56:00,000
you know, gripper.

982
00:56:00,000 --> 00:56:01,000
Audrow Nash: Like a pincher.

983
00:56:01,000 --> 00:56:05,308
Stefan Seltz-Axmacher: Yeah.
And, and it's probably, you

984
00:56:05,308 --> 00:56:09,846
know, building materials is
probably in the 35th. The 20 to

985
00:56:09,846 --> 00:56:14,154
$40,000 range. And the whole
point of it is it is easily

986
00:56:14,154 --> 00:56:18,846
deployable in smaller and non
massive factories, to do things

987
00:56:18,846 --> 00:56:22,923
like machine tending or, or
deep, deep palletizing or

988
00:56:22,923 --> 00:56:27,615
getting things out of a box or
into boxes and kind of rapidly

989
00:56:27,615 --> 00:56:31,846
be moved from one task to
another with maybe with maybe

990
00:56:31,846 --> 00:56:36,385
only two hours of setup. And I
think I've heard a company I

991
00:56:36,385 --> 00:56:40,000
wanted to say 1x might be doing
this, but they.

992
00:56:40,000 --> 00:56:43,489
Ilia Baranov: Went the other
way. They went the other way,

993
00:56:43,489 --> 00:56:47,158
they built this wheeled one and
now they're building the full

994
00:56:47,158 --> 00:56:48,000
humanoid look.

995
00:56:48,000 --> 00:56:51,616
Stefan Seltz-Axmacher: Yeah. And
I think that is that is a

996
00:56:51,616 --> 00:56:55,169
strategy where, like, people
will show up to buy the cool

997
00:56:55,169 --> 00:56:58,971
looking one and then when like,
yeah, the cool looking one is

998
00:56:58,971 --> 00:57:02,587
great, $2 million and has 20
minutes run time. The wheeled

999
00:57:02,587 --> 00:57:06,452
one, will fit your application
and is economical compared to a

1000
00:57:06,452 --> 00:57:10,379
human. Yeah, by one of the cool
ones. We love. You buy one, but

1001
00:57:10,379 --> 00:57:12,000
by 50 of the wheeled ones,

1002
00:57:12,000 --> 00:57:15,893
Ilia Baranov: I just I also
think, like on a meta level, the

1003
00:57:15,893 --> 00:57:19,850
whole idea of humanoids sort of
is, is sort of like a fallacy

1004
00:57:19,850 --> 00:57:23,159
that in my mind, I feel I'm
probably making a wrong

1005
00:57:23,159 --> 00:57:27,116
prediction here. But like, you
know, I looked at 1X video and

1006
00:57:27,116 --> 00:57:31,268
like promo and they're like, 1x
Thank you. And like the humanoid

1007
00:57:31,268 --> 00:57:34,902
is moving a vacuum like a
specific a Dyson, of course, a

1008
00:57:34,902 --> 00:57:38,794
Dyson around somebody's house.
And they're like, we're going

1009
00:57:38,794 --> 00:57:41,000
for home robots. Do you know
what.

1010
00:57:41,000 --> 00:57:43,000
Stefan Seltz-Axmacher: Home
people have?

1011
00:57:43,000 --> 00:57:46,018
Ilia Baranov: It doesn't look
anything like a humanoid. It

1012
00:57:46,018 --> 00:57:49,088
costs 100th the price and it
works just great. I have three

1013
00:57:49,088 --> 00:57:52,470
of them in my house right. Like.
Right. And it's the whole, like,

1014
00:57:52,470 --> 00:57:55,488
you know, we live in a world of
horse drawn carriages. And

1015
00:57:55,488 --> 00:57:58,506
somebody is like, I'm going to
make a more efficient horse

1016
00:57:58,506 --> 00:58:01,836
drawn carriage. It's going to be
a mechanized horse instead of a

1017
00:58:01,836 --> 00:58:05,062
car. Right. Like, I think that's
the fallacy people fall into.

1018
00:58:05,062 --> 00:58:08,392
And there's this thing like, oh
yeah, but you know, the humanoid

1019
00:58:08,392 --> 00:58:11,774
human world is built for humans.
And like, it's sorta like, yeah,

1020
00:58:11,774 --> 00:58:15,000
I know House has stairs, but it
also has perfectly flat floor.

1021
00:58:15,000 --> 00:58:18,823
Stefan Seltz-Axmacher: I
actually rarely intersects with

1022
00:58:18,823 --> 00:58:23,260
the part of the world that, that
was built for humans and like so

1023
00:58:23,260 --> 00:58:27,561
to be clear, what do you mean?
Yeah. Humans were not humans did

1024
00:58:27,561 --> 00:58:31,589
not come about, in houses with
level floors and heating and

1025
00:58:31,589 --> 00:58:35,821
protection from the elements and
whatever humans came about in

1026
00:58:35,821 --> 00:58:40,122
forests. Well, regardless of how
you think we came about there.

1027
00:58:40,122 --> 00:58:44,149
So it's I, you know, pick a
particular side, but regardless

1028
00:58:44,149 --> 00:58:48,314
of it, you need legs for walking
in mud and sand and dirt and

1029
00:58:48,314 --> 00:58:52,000
arms for picking up, you know,
picking. You're saying.

1030
00:58:52,000 --> 00:58:53,000
Audrow Nash: It doesn't have to
fit the.

1031
00:58:53,000 --> 00:58:56,000
Stefan Seltz-Axmacher: Circular
spear? Yeah. You need. Yeah.

1032
00:58:56,000 --> 00:58:59,022
Audrow Nash: I live the
constraints that made us aren't

1033
00:58:59,022 --> 00:59:01,000
what made the or that a robot
needs.

1034
00:59:01,000 --> 00:59:02,778
Stefan Seltz-Axmacher: I think
there's a room who rarely

1035
00:59:02,778 --> 00:59:03,000
leaves.

1036
00:59:03,000 --> 00:59:05,000
Ilia Baranov: There's a reason
wheels never in.

1037
00:59:05,000 --> 00:59:08,432
Stefan Seltz-Axmacher: My life
is between a series of flat

1038
00:59:08,432 --> 00:59:11,805
surfaces. Like my house is a
series of flat surfaces. The

1039
00:59:11,805 --> 00:59:15,000
stairs are some more flat
surfaces. It's flat surface.

1040
00:59:15,000 --> 00:59:18,122
Audrow Nash: I feel like that is
how most robotics way to. Yeah,

1041
00:59:18,122 --> 00:59:19,000
but it looks like.

1042
00:59:19,000 --> 00:59:22,916
Stefan Seltz-Axmacher: Which I
am. I it outside of my ten

1043
00:59:22,916 --> 00:59:27,382
minute walk between my house and
our office. I am never more than

1044
00:59:27,382 --> 00:59:31,229
30ft from an outlet. Ever.
That's just not a thing in my

1045
00:59:31,229 --> 00:59:35,489
life. Yeah, I never outside of
cell phone range. I'm not a big

1046
00:59:35,489 --> 00:59:39,679
hiker, but also I don't need to
buy a $10,000, device to hike

1047
00:59:39,679 --> 00:59:44,076
with. That's. Theoretically, if
I'm hiking, I should not be like

1048
00:59:44,076 --> 00:59:46,000
bringing a computer on legs.

1049
00:59:46,000 --> 00:59:49,412
Ilia Baranov: And I think I
think I think again, like, the

1050
00:59:49,412 --> 00:59:52,941
meta point is like, I think
there'll be more robots in homes

1051
00:59:52,941 --> 00:59:56,118
over time. And there are
already. Right. And the robot

1052
00:59:56,118 --> 00:59:59,000
vacuum clear makers are making
units with arm to.

1053
00:59:59,000 --> 01:00:00,000
Audrow Nash: Like, move simple.

1054
01:00:00,000 --> 01:00:03,302
Ilia Baranov: Things. The
ability to climb stairs right,

1055
01:00:03,302 --> 01:00:06,781
right. And like and I think the
home robots again the astro

1056
01:00:06,781 --> 01:00:10,319
being an example but like a
bunch of them I don't think need

1057
01:00:10,319 --> 01:00:13,799
to be human shaped to accomplish
their task. And so I think

1058
01:00:13,799 --> 01:00:17,572
humanoids are very cool looking,
but I think it's I think it's a

1059
01:00:17,572 --> 01:00:20,993
fallacy, honestly. I think it's
a fallacy until technology

1060
01:00:20,993 --> 01:00:24,000
progresses a few more orders of
magnitude and then.

1061
01:00:24,000 --> 01:00:27,734
Stefan Seltz-Axmacher: Maybe,
you know, like, like triple

1062
01:00:27,734 --> 01:00:31,403
clicking on this a little bit.
And this is kind of a new

1063
01:00:31,403 --> 01:00:35,268
thought. So this might sound
really stupid. A humanoid to a

1064
01:00:35,268 --> 01:00:39,068
general purpose is a cool
general idea, and it makes a lot

1065
01:00:39,068 --> 01:00:43,130
of sense if you can only have
one thing right. If you can only

1066
01:00:43,130 --> 01:00:47,060
have one computer, it makes
sense for that computer to be in

1067
01:00:47,060 --> 01:00:51,188
your living room. It makes sense
to have multiple user accounts

1068
01:00:51,188 --> 01:00:55,184
so everyone in the family can
log into it. It makes sense for

1069
01:00:55,184 --> 01:00:58,853
it to have a bunch of central
things. There's a bunch of

1070
01:00:58,853 --> 01:01:02,718
things like that that go along
with that. And humanoids are

1071
01:01:02,718 --> 01:01:05,928
more significantly more
expensive than the family

1072
01:01:05,928 --> 01:01:09,924
computer was back in the 90s.
What's happened since then? And

1073
01:01:09,924 --> 01:01:13,789
I'm just looking around myself
because I am an embarrassing

1074
01:01:13,789 --> 01:01:17,786
person. With what we're about to
talk about, we have infinite

1075
01:01:17,786 --> 01:01:21,978
computers. We can have infinite,
like, looking around me, I have

1076
01:01:21,978 --> 01:01:26,171
two smart speakers here. I have,
an Apple Watch on another Apple

1077
01:01:26,171 --> 01:01:30,233
Watch right there, a phone, a
laptop here, another laptop over

1078
01:01:30,233 --> 01:01:34,033
there, an Apple TV, which is
functionally another computer

1079
01:01:34,033 --> 01:01:38,029
right over there, a smart TV
that can also stream things. And

1080
01:01:38,029 --> 01:01:42,025
I'm just looking at like the
thing, I have a Kindle. You have

1081
01:01:42,025 --> 01:01:46,218
been sitting on my desk. This is
one of like four Kindles I own.

1082
01:01:46,218 --> 01:01:50,345
I have an iPad. I have all the I
have all these things. Yeah. I

1083
01:01:50,345 --> 01:01:54,473
cannot imagine a world where a
single $10,000 desktop computer,

1084
01:01:54,473 --> 01:01:58,600
would be better than the series
of devices I just listed off as

1085
01:01:58,600 --> 01:02:02,596
being within eight feet of me.
And similarly to Eliza's point

1086
01:02:02,596 --> 01:02:06,658
about, you know, roombas, for
the, for intelligent mechatronic

1087
01:02:06,658 --> 01:02:10,785
devices that I might want in my
household, it is hard for me to

1088
01:02:10,785 --> 01:02:14,847
imagine a series of applications
I'd want done by a robot that

1089
01:02:14,847 --> 01:02:18,647
would add up to something I'd
spend 5 to $10,000 on, but I

1090
01:02:18,647 --> 01:02:22,250
can't imagine buying 15 $200
robots one by one over the

1091
01:02:22,250 --> 01:02:26,181
course of six years. Anything
one $5,000 robot that does all

1092
01:02:26,181 --> 01:02:30,242
of my, like, eat applications
that I care about a robot doing,

1093
01:02:30,242 --> 01:02:34,239
that's probably never going to
happen. That's like probably a

1094
01:02:34,239 --> 01:02:38,366
nonstarter with my wife. Like,
maybe, maybe if something really

1095
01:02:38,366 --> 01:02:42,231
cool happens, a massive amount
of money comes my way. Like,

1096
01:02:42,231 --> 01:02:44,000
maybe I can blow ten grand.

1097
01:02:44,000 --> 01:02:47,645
Audrow Nash: Yeah, but it's just
that. Yeah, yeah, it's blowing

1098
01:02:47,645 --> 01:02:51,000
that money on it. Yeah. The
thing, because of the novelty.

1099
01:02:51,000 --> 01:02:54,019
Ilia Baranov: Yeah. And, and I
think, I think again, like

1100
01:02:54,019 --> 01:02:57,143
humans. And we fall into this
too, by the way. We fall into

1101
01:02:57,143 --> 01:03:00,374
this fight like, all the time.
Stefan and I had a discussion.

1102
01:03:00,374 --> 01:03:03,287
Right. Probably a year ago now
on, on our podcast where

1103
01:03:03,287 --> 01:03:06,146
basically we're like, oh, you
know, I want, I want the

1104
01:03:06,146 --> 01:03:09,377
classic, like, go get me a beer
from the fridge robot. And it

1105
01:03:09,377 --> 01:03:12,819
was like, okay, so you need like
an a robot with an arm that goes

1106
01:03:12,819 --> 01:03:15,838
and opens the fridge and like
finds manipulates, gets the

1107
01:03:15,838 --> 01:03:19,174
beer, brings it back, whatever.
And I was like, Stefan or like,

1108
01:03:19,174 --> 01:03:22,405
and this wasn't in the house, by
the way. This is this is for

1109
01:03:22,405 --> 01:03:25,000
work. So it makes a little bit
more sense, right?

1110
01:03:25,000 --> 01:03:27,591
Stefan Seltz-Axmacher: Just
because we drink in an office

1111
01:03:27,591 --> 01:03:28,000
facility.

1112
01:03:28,000 --> 01:03:30,000
Audrow Nash: Oh well, very good.
to clear up

1113
01:03:30,000 --> 01:03:31,000
Stefan Seltz-Axmacher: Yeah.

1114
01:03:31,000 --> 01:03:34,354
Ilia Baranov: But like you know,
and, and he was like,

1115
01:03:34,354 --> 01:03:37,894
overcomplicated this whole
thing, I was like, Stefan, you

1116
01:03:37,894 --> 01:03:41,000
could just have a vending
machine. Like a vending.

1117
01:03:41,000 --> 01:03:42,000
Stefan Seltz-Axmacher: Machine.

1118
01:03:41,000 --> 01:03:44,489
Ilia Baranov: Has all the
management and stuff one at a

1119
01:03:44,489 --> 01:03:48,360
time. Yeah, yeah. No no no no no
no. My point is, like, I, my

1120
01:03:48,360 --> 01:03:52,356
point is like a Roomba a Roomba
or an Astro or whatever. Like a

1121
01:03:52,356 --> 01:03:56,163
robot. I can summon somewhere
and a vending machine. Will do

1122
01:03:56,163 --> 01:04:00,033
everything you want. Like you
drive the robot under the ramp.

1123
01:04:00,033 --> 01:04:03,840
You internet call the vending
machine because they're almost

1124
01:04:03,840 --> 01:04:07,393
all internet connected now
anyway. It dispenses the beer

1125
01:04:07,393 --> 01:04:10,882
you want, it falls into the
thing, and it delivers your

1126
01:04:10,882 --> 01:04:14,689
slightly shaken beer to your
room for a fraction of the cost

1127
01:04:14,689 --> 01:04:18,622
of any arm, or any manipulator,
or any sensing or any humanoid

1128
01:04:18,622 --> 01:04:22,619
you could possibly buy. Right?
Like even cheaper than a Unitree

1129
01:04:22,619 --> 01:04:23,000
robot.

1130
01:04:23,000 --> 01:04:27,089
Stefan Seltz-Axmacher: Easily.
Our other our other version of

1131
01:04:27,089 --> 01:04:31,313
this, is because people always
like the idea of robots for, the

1132
01:04:31,313 --> 01:04:35,201
dishwasher, for taking the
dishwasher. And I have exciting

1133
01:04:35,201 --> 01:04:39,089
news for you. Although there's a
robot you can buy. That's

1134
01:04:39,089 --> 01:04:43,112
awesome. And for any of the
listeners, for $400, purchasable

1135
01:04:43,112 --> 01:04:47,000
at nearly any, any big box
store, buy a second dishwasher.

1136
01:04:47,000 --> 01:04:48,000
Audrow Nash: Yeah, totally.

1137
01:04:48,000 --> 01:04:51,362
Stefan Seltz-Axmacher: Yeah.
Suddenly, if you put all your

1138
01:04:51,362 --> 01:04:54,841
dirty plates in one, take all
your stuff, you just flop back

1139
01:04:54,841 --> 01:04:58,377
and forth, just flopping back
and forth and look at that. You

1140
01:04:58,377 --> 01:05:01,971
know, you I just saved you. You
95% of your dish. Put it away,

1141
01:05:01,971 --> 01:05:04,000
with technology and then you
don't.

1142
01:05:04,000 --> 01:05:06,400
Audrow Nash: Need space for
storing them. Even better. Yeah.

1143
01:05:06,400 --> 01:05:07,000
More efficient.

1144
01:05:07,000 --> 01:05:09,096
Stefan Seltz-Axmacher: You
already have space for storing

1145
01:05:09,096 --> 01:05:11,301
your plates. So you just take
one of your cabinets. Attorney

1146
01:05:11,301 --> 01:05:12,000
general dishwasher.

1147
01:05:12,000 --> 01:05:14,000
Audrow Nash: Yeah.

1148
01:05:14,000 --> 01:05:15,000
Ilia Baranov: That's a little.
That's a little.

1149
01:05:15,000 --> 01:05:17,000
Stefan Seltz-Axmacher: Yeah.
What's funny, though, is that.

1150
01:05:17,000 --> 01:05:18,000
Ilia Baranov: I think the main.

1151
01:05:17,000 --> 01:05:19,598
Stefan Seltz-Axmacher: Thing and
having a big piece of metal go

1152
01:05:19,598 --> 01:05:21,000
around my house. Yeah, for
10,000.

1153
01:05:21,000 --> 01:05:22,000
Ilia Baranov: Exactly.

1154
01:05:22,000 --> 01:05:24,000
Stefan Seltz-Axmacher: And scare
them of Jesus out of my dog.

1155
01:05:24,000 --> 01:05:27,782
Ilia Baranov: This one X group.
Like I think again, looking at

1156
01:05:27,782 --> 01:05:31,197
their press release recently,
they're like, we're really

1157
01:05:31,197 --> 01:05:34,918
targeting like we decided to
target the home market as, like,

1158
01:05:34,918 --> 01:05:38,821
why there's no I can't think of
a reason other than you tried to

1159
01:05:38,821 --> 01:05:42,481
go after the industrial market.
Realize that it doesn't make

1160
01:05:42,481 --> 01:05:46,140
sense and now have to pivot to
get more fun. You could be in

1161
01:05:46,140 --> 01:05:49,861
more hype because like in every
other respect, the industrial

1162
01:05:49,861 --> 01:05:53,582
market is so much more valuable
and so much more forgiving of

1163
01:05:53,582 --> 01:05:57,119
failures, like an industry
that's testing a new technology

1164
01:05:57,119 --> 01:06:00,779
will be okay if you drop a part
somebody is testing in their

1165
01:06:00,779 --> 01:06:04,682
home and you drop a plate. We'll
put a negative review on Amazon

1166
01:06:04,682 --> 01:06:07,000
instantly. Like it with no
hesitation.

1167
01:06:07,000 --> 01:06:08,000
Audrow Nash: Drop my plate.
Right.

1168
01:06:08,000 --> 01:06:10,703
Ilia Baranov: Like there's no
reason to go after that market

1169
01:06:10,703 --> 01:06:13,000
unless you just can't. Yeah,
unless you just can't.

1170
01:06:13,000 --> 01:06:16,512
Audrow Nash: Yeah. Okay. I
wonder the thing that was

1171
01:06:16,512 --> 01:06:20,361
interesting to me. I haven't
been in touch with them in a

1172
01:06:20,361 --> 01:06:24,548
while since I interned. Really.
I mean, I saw them a few other

1173
01:06:24,548 --> 01:06:28,803
times since, but, when I started
with one acts as an intern. We

1174
01:06:28,803 --> 01:06:32,787
were going after eldercare for
this kind of thing. And that

1175
01:06:32,787 --> 01:06:37,109
application, it was like remote.
Someone would beam into it and,

1176
01:06:37,109 --> 01:06:40,891
have it microwave food or
something like this. Which the

1177
01:06:40,891 --> 01:06:44,875
prices of assisted living are
very high. Yeah. So that that

1178
01:06:44,875 --> 01:06:48,927
made some sense to me. I like
the mission there, but then it

1179
01:06:48,927 --> 01:06:52,844
was surprising to see them
pivoting a few times. And maybe

1180
01:06:52,844 --> 01:06:56,964
it is a funding thing. I don't
know, but it is funny how much

1181
01:06:56,964 --> 01:07:01,083
you can simplify things with
better design than just throwing

1182
01:07:01,083 --> 01:07:05,068
a very expensive robot at it. A
parallel that I have for my

1183
01:07:05,068 --> 01:07:08,849
experience is, for the first
year of this podcast, I was

1184
01:07:08,849 --> 01:07:13,036
trying very heavily to use AI
tools for helping me find clips,

1185
01:07:13,036 --> 01:07:17,088
helping me, do editing, like the
script will, it'll pick the

1186
01:07:17,088 --> 01:07:21,073
person to be in the shot and
things like this. And so I was

1187
01:07:21,073 --> 01:07:25,327
trying so hard to get AI tools
to work for this, and then I got

1188
01:07:25,327 --> 01:07:29,379
so frustrated with it that I
just hired someone. And they're

1189
01:07:29,379 --> 01:07:33,431
not that expensive. They're
absolutely wonderful. And it's a

1190
01:07:33,431 --> 01:07:37,551
huge like the quality is far,
far higher. And it doesn't have

1191
01:07:37,551 --> 01:07:41,738
like weird artifact drastically
better for this kind of thing.

1192
01:07:41,738 --> 01:07:45,857
And similarly, yeah, of course
hiring like so you could buy a

1193
01:07:45,857 --> 01:07:49,909
say it goes down to $5,000,
$5,000 robots. Or you could just

1194
01:07:49,909 --> 01:07:53,826
hire a housecleaner for this
kind of thing, and they'll be

1195
01:07:53,826 --> 01:07:57,000
excellent and custom work and
things like that.

1196
01:07:57,000 --> 01:08:01,249
Stefan Seltz-Axmacher: There are
services that are like Uber for

1197
01:08:01,249 --> 01:08:05,033
people to show up at your
parents house to microwave food

1198
01:08:05,033 --> 01:08:08,884
for them. Yeah. And those
services are not that expensive.

1199
01:08:08,884 --> 01:08:13,000
I yeah, like the services like
approach like $10 a visit. Wow.

1200
01:08:13,000 --> 01:08:14,000
Audrow Nash: That's crazy.

1201
01:08:14,000 --> 01:08:18,617
Stefan Seltz-Axmacher: Which and
like maybe I'm rounding down too

1202
01:08:18,617 --> 01:08:22,808
aggressively, but not $100.
Yeah. The math for that being a

1203
01:08:22,808 --> 01:08:26,715
better outcome than by a
humanoid is, like, pretty, not

1204
01:08:26,715 --> 01:08:31,048
incredibly hard. For that to
work out, I think, like, there's

1205
01:08:31,048 --> 01:08:35,239
a, a hard thing in robotics that
that happens frequently is

1206
01:08:35,239 --> 01:08:39,430
roboticists are a specialized
skill set who generally don't

1207
01:08:39,430 --> 01:08:44,048
want to go work on a SaaS app or
not or not for cats or whatever.

1208
01:08:44,048 --> 01:08:47,812
So it's kind of like I have
robotics, where can I put

1209
01:08:47,812 --> 01:08:52,003
robots? With a little bit less
thought to like, what is the

1210
01:08:52,003 --> 01:08:56,621
best solution for the problem at
hand? And I think, like, I don't

1211
01:08:56,621 --> 01:09:00,243
think any of us here are
bearish. Is there value in

1212
01:09:00,243 --> 01:09:04,648
robotics? I don't think any of
us here think like AI tools are

1213
01:09:04,648 --> 01:09:09,123
net negative, but like the all
of us have, all three of us have

1214
01:09:09,123 --> 01:09:13,030
gotten to play and use AI and
machine learning and deep

1215
01:09:13,030 --> 01:09:17,008
learning for like I mean, I
think at least a decade plus

1216
01:09:17,008 --> 01:09:21,128
each, it's really cool. But
like, the word AI feels like a

1217
01:09:21,128 --> 01:09:25,248
stretch. Yeah. And that almost
makes us seem like the most

1218
01:09:25,248 --> 01:09:29,226
negative people in the world
about these incredibly cool

1219
01:09:29,226 --> 01:09:33,701
technologies. I think like, my
view of, of broader AI is, is AI

1220
01:09:33,701 --> 01:09:38,247
is going to be as big of a deal
so that over the next decade as,

1221
01:09:38,247 --> 01:09:42,438
like microservices were in the
20 tens. And like, there's a

1222
01:09:42,438 --> 01:09:46,345
bunch of people for whom that's
like, I don't know what

1223
01:09:46,345 --> 01:09:50,749
microservice architecture is. I
don't know what that is. Okay,

1224
01:09:50,749 --> 01:09:55,224
cool. And those are the people
who are probably not making good

1225
01:09:55,224 --> 01:09:57,000
AI investments right now.

1226
01:09:57,000 --> 01:09:58,000
Audrow Nash: Yeah, I think so.

1227
01:09:58,000 --> 01:10:01,995
Stefan Seltz-Axmacher: Whereas
instead, like, yeah, AI is going

1228
01:10:01,995 --> 01:10:05,483
to make every company 10 to 20%
efficient, not 98% more

1229
01:10:05,483 --> 01:10:09,478
efficient. But like 10 to 20 and
10 to 20 is pretty meaningful.

1230
01:10:09,478 --> 01:10:13,410
Substantial. Yes. Like massive a
huge deal. But no, no, no one

1231
01:10:13,410 --> 01:10:17,151
has a job anymore. We don't roll
video editors anymore. And

1232
01:10:17,151 --> 01:10:21,146
humanoids will do all physical
labor for us in every situation.

1233
01:10:21,146 --> 01:10:24,000
Yeah, which is a bummer because
it'd be cool.

1234
01:10:24,000 --> 01:10:26,000
Ilia Baranov: Yeah. It's, you
know.

1235
01:10:26,000 --> 01:10:27,000
Audrow Nash: Go ahead, Ilia.

1236
01:10:27,000 --> 01:10:30,561
Ilia Baranov: I, I want to
wrench us in another direction.

1237
01:10:30,561 --> 01:10:34,246
Just one funny comment, right?
Yeah. Go ahead. One challenge

1238
01:10:34,246 --> 01:10:37,746
for your listeners and viewers
is, next time you do see a

1239
01:10:37,746 --> 01:10:41,553
humanoid ad, check if they show
how this thing charges and how

1240
01:10:41,553 --> 01:10:44,868
long its battery life is.
Because I think, and I could

1241
01:10:44,868 --> 01:10:48,307
easily be wrong because I
haven't seen every single one,

1242
01:10:48,307 --> 01:10:52,114
but I think I haven't seen one
yet. That shows itself plugging

1243
01:10:52,114 --> 01:10:55,000
itself in to recharge. And that
is a trick that

1244
01:10:55,000 --> 01:10:57,987
Audrow Nash: 1x does. I've seen
a video of one Zeus, robots

1245
01:10:57,987 --> 01:10:59,000
charging themselves.

1246
01:10:59,000 --> 01:11:01,776
Ilia Baranov: Oh, awesome.
Awesome, because that is a trick

1247
01:11:01,776 --> 01:11:03,000
PR2 did a decade way back.

1248
01:11:03,000 --> 01:11:04,000
Audrow Nash: Yeah.

1249
01:11:04,000 --> 01:11:07,030
Ilia Baranov: Right. Right. And
so like for the full like,

1250
01:11:07,030 --> 01:11:10,425
usability cycle, how do you make
sure that the thing can actually

1251
01:11:10,425 --> 01:11:13,769
charge itself a little? Because
again, even your Roomba fails to

1252
01:11:13,769 --> 01:11:16,694
dock, you know, every time.
Yeah. Right. So anyway, I'll

1253
01:11:16,694 --> 01:11:20,037
stop harping on that. But like,
if you're, if you're getting the

1254
01:11:20,037 --> 01:11:22,963
itch to buy a humanoid, check
that fact. How long is the

1255
01:11:22,963 --> 01:11:25,000
battery life and does it charge
itself?

1256
01:11:25,000 --> 01:11:28,239
Audrow Nash: I think the battery
life thing is a very interesting

1257
01:11:28,239 --> 01:11:31,279
thing, because they can plug
themselves in to some level. One

1258
01:11:31,279 --> 01:11:34,319
x it was the wheeled base, one
that I've seen. I haven't seen

1259
01:11:34,319 --> 01:11:37,359
their, bipedal one. But it's an
interesting thing because the

1260
01:11:37,359 --> 01:11:40,000
battery life might be super
short. I don't know, the.

1261
01:11:40,000 --> 01:11:43,960
Stefan Seltz-Axmacher: Biggest
thing that I've yet to ever hear

1262
01:11:43,960 --> 01:11:47,920
anyone be able to justify to me,
even in passing, is my legs. I

1263
01:11:47,920 --> 01:11:51,000
have. I have yet to hear an
argument for a while.

1264
01:11:51,000 --> 01:11:54,214
Ilia Baranov: Like best one I've
heard is stairs. Like for home

1265
01:11:54,214 --> 01:11:56,000
use stairs. Yeah, like that's
lots.

1266
01:11:56,000 --> 01:11:58,652
Stefan Seltz-Axmacher: Of stairs
in the middle of my airplane

1267
01:11:58,652 --> 01:11:59,000
factory.

1268
01:11:59,000 --> 01:12:02,590
Ilia Baranov: Or for industrial,
I think industrial. Really?

1269
01:12:02,590 --> 01:12:06,000
Yeah, I get that for people are
excited about home. Yeah.

1270
01:12:06,000 --> 01:12:09,667
Stefan Seltz-Axmacher: Well,
yeah. Well, which is why

1271
01:12:09,667 --> 01:12:13,748
computers first came to the
home. Like the first place that

1272
01:12:13,748 --> 01:12:17,000
anyone had a computer was in
their living room,

1273
01:12:17,000 --> 01:12:18,000
Ilia Baranov: To store it. Store
recipes. Right.

1274
01:12:18,000 --> 01:12:20,000
Audrow Nash: Was out there.
Yeah. Yeah, that was like.

1275
01:12:20,000 --> 01:12:21,000
Ilia Baranov: Store recipes.

1276
01:12:21,000 --> 01:12:25,025
Stefan Seltz-Axmacher: You know,
besides computer, you know, you

1277
01:12:25,025 --> 01:12:29,050
don't need a living room if you
have a computer to, to print out

1278
01:12:29,050 --> 01:12:31,000
cards with what your recipe is.

1279
01:12:31,000 --> 01:12:34,377
Audrow Nash: And but so, just to
be clear, just so we're not

1280
01:12:34,377 --> 01:12:37,697
losing anyone, computers first
went into businesses, right?

1281
01:12:37,697 --> 01:12:41,130
For, like, Excel, though. Yeah,
and this kind of thing. Yeah.

1282
01:12:41,130 --> 01:12:44,563
And then the applications came
later as the costs drove down.

1283
01:12:44,563 --> 01:12:47,714
So I suspect that that's
probably the good metaphor. Not

1284
01:12:47,714 --> 01:12:51,316
starting with the home as you're
suggesting. Yeah. What are you.

1285
01:12:51,316 --> 01:12:54,636
So, segueing a bit what are you
thinking about? All the LLM

1286
01:12:54,636 --> 01:12:58,126
stuff and vibe coding and things
like this. Let's see. Stefan,

1287
01:12:58,126 --> 01:13:00,883
you've been playing with it.
Yeah. What were your

1288
01:13:00,883 --> 01:13:04,260
impressions? I like, I've done
it a bit, and it's very good,

1289
01:13:04,260 --> 01:13:07,693
but you can hang yourself on the
complexity very quick. Yeah.

1290
01:13:07,693 --> 01:13:10,000
I'm. You don't structure things
well, so.

1291
01:13:10,000 --> 01:13:13,297
Stefan Seltz-Axmacher: I'd say
I'm probably three hours into

1292
01:13:13,297 --> 01:13:16,374
playing around with it. So this
is this is a relatively,

1293
01:13:16,374 --> 01:13:19,835
relatively light, the thing that
I'm trying to do right now is,

1294
01:13:19,835 --> 01:13:22,967
have it vibe coded application
that lives behind an email

1295
01:13:22,967 --> 01:13:25,000
address. Like a back end
application.

1296
01:13:25,000 --> 01:13:27,000
Audrow Nash: Emails are hard.
Okay.

1297
01:13:27,000 --> 01:13:29,586
Ilia Baranov: Let's. But by the
way, what's your definition,

1298
01:13:29,586 --> 01:13:32,000
Stefan? Of a vibe coding. So
we're all on the same page.

1299
01:13:32,000 --> 01:13:36,762
Stefan Seltz-Axmacher: I I've,
I've been able to do hello World

1300
01:13:36,762 --> 01:13:41,600
and a little bit beyond that in
Python and C plus before. So I'm

1301
01:13:41,600 --> 01:13:46,212
trying to describe a function
that needs to happen. And it be

1302
01:13:46,212 --> 01:13:50,747
coded for me. And the function
I'm trying to is the function

1303
01:13:50,747 --> 01:13:55,283
I'm trying to get to is take a
zoom summary of a sales call,

1304
01:13:55,283 --> 01:13:59,819
which is, let's say, 2000 words
long. Forward it to an email

1305
01:13:59,819 --> 01:14:04,279
address, that then, runs it
against ChatGPT to turn it into

1306
01:14:04,279 --> 01:14:07,000
five bullets so that after our
call.

1307
01:14:07,000 --> 01:14:10,000
Audrow Nash: That's fairly
straightforward. Yeah.

1308
01:14:10,000 --> 01:14:16,104
Stefan Seltz-Axmacher: I've been
using, windsurfer kite surfer

1309
01:14:16,104 --> 01:14:22,207
windsurf. It will not compile.
Well, will not turn on, and has

1310
01:14:22,207 --> 01:14:28,016
struggled to be able to adjust
prompts. Is the status of my

1311
01:14:28,016 --> 01:14:29,000
code base.

1312
01:14:29,000 --> 01:14:33,890
Audrow Nash: Cursor is actually
pretty sophisticated. For things

1313
01:14:33,890 --> 01:14:38,169
I would suspect you'd be able to
be successful with that

1314
01:14:38,169 --> 01:14:42,983
application and cursor. You use
a good model like I find claude

1315
01:14:42,983 --> 01:14:47,568
3.7 now. Yeah. To be the best
model for these things, but, I

1316
01:14:47,568 --> 01:14:52,000
would think you'd be able to
achieve that. Yeah. For sure.

1317
01:14:52,000 --> 01:14:55,173
Stefan Seltz-Axmacher: And,
like, to be clear, I've. My

1318
01:14:55,173 --> 01:14:58,750
three hours is a series of 30
minute increments with, hey, go,

1319
01:14:58,750 --> 01:15:02,442
go write a complex email to this
person about a complex subject.

1320
01:15:02,442 --> 01:15:06,077
So I've hardly been flow state
developing, let alone flow state

1321
01:15:06,077 --> 01:15:07,000
vibe developing.

1322
01:15:07,000 --> 01:15:09,465
Audrow Nash: That makes sense.
Yeah. Ilia, what do you think?

1323
01:15:09,465 --> 01:15:11,000
Have you played around with it
at all?

1324
01:15:11,000 --> 01:15:14,281
Ilia Baranov: Yeah, I mean for
sure. I mean that also, you

1325
01:15:14,281 --> 01:15:17,561
know, to Stefan's point that you
still probably in the top

1326
01:15:17,561 --> 01:15:20,842
quartile of people who've
touched software before and have

1327
01:15:20,842 --> 01:15:24,348
some concept of compiling. Yeah,
you know, that sort of thing,

1328
01:15:24,348 --> 01:15:27,686
which is like a still a very
small piece of the population.

1329
01:15:27,686 --> 01:15:31,305
And to step outside of a bubble
and remember that, like the vast

1330
01:15:31,305 --> 01:15:34,869
majority of humanity has no idea
how the magic in the box works

1331
01:15:34,869 --> 01:15:36,000
to, to run software.

1332
01:15:36,000 --> 01:15:39,123
Audrow Nash: People that are not
robotics people or like software

1333
01:15:39,123 --> 01:15:41,958
people have not even tried
ChatGPT yet that I've talked to,

1334
01:15:41,958 --> 01:15:44,889
which is that, like, you just
talk to general people. They're

1335
01:15:44,889 --> 01:15:47,964
like, I've heard about it, but I
haven't tried it, haven't tried

1336
01:15:47,964 --> 01:15:50,895
it, which is just unbelievable
because it's been a year and a

1337
01:15:50,895 --> 01:15:52,000
half or whatever. Yeah.

1338
01:15:52,000 --> 01:15:54,813
Ilia Baranov: So you're already
talking to like a person who

1339
01:15:54,813 --> 01:15:57,719
through experience is in a very
small slice of the population.

1340
01:15:57,719 --> 01:15:58,000
Right.

1341
01:15:58,000 --> 01:16:01,000
Audrow Nash: Definitely. And
using to especially.

1342
01:16:01,000 --> 01:16:04,891
Ilia Baranov: Yeah. Well yeah, I
mean I mean, you know, I, I code

1343
01:16:04,891 --> 01:16:08,543
semi-regularly usually to the
chagrin of my engineering team,

1344
01:16:08,543 --> 01:16:12,255
whatever I have to do anything
and generally I use it, for two

1345
01:16:12,255 --> 01:16:16,146
reasons. One is if I'm trying to
put together any kind of complex

1346
01:16:16,146 --> 01:16:19,439
documentation or email or
something where I'm trying to

1347
01:16:19,439 --> 01:16:23,211
get a very large thought across
the first thing I do like, I've

1348
01:16:23,211 --> 01:16:26,683
almost replaced mind mapping
with prompting, where I'll be

1349
01:16:26,683 --> 01:16:28,000
like, I want to write.

1350
01:16:28,000 --> 01:16:29,000
Audrow Nash: Almost.

1351
01:16:29,000 --> 01:16:31,629
Ilia Baranov: Yeah, almost.
Yeah, like I want to write.

1352
01:16:31,629 --> 01:16:34,593
Unless is very esoteric, right?
But like, I want to write, you

1353
01:16:34,593 --> 01:16:37,700
know, object oriented programing
for robots, like, that's, that's

1354
01:16:37,700 --> 01:16:40,377
the title of my talk making One
up. Right. What should I

1355
01:16:40,377 --> 01:16:43,340
consider? And it'll list out a
bunch and then I'll pick like 5

1356
01:16:43,340 --> 01:16:46,208
or 6. Right. Yeah. Okay. That
sounds interesting. I might go

1357
01:16:46,208 --> 01:16:49,124
with that. Let's talk a bit more
about that. Build a skeleton

1358
01:16:49,124 --> 01:16:51,992
basically, or a mind map. And
then from that, that I'm, that

1359
01:16:51,992 --> 01:16:54,000
I'm writing in myself, I'm doing
whatever.

1360
01:16:54,000 --> 01:16:56,000
Audrow Nash: Right. Then you're
back to your mind map. Yeah.

1361
01:16:56,000 --> 01:16:59,353
Ilia Baranov: Right. Right. And
like and like that mind map

1362
01:16:59,353 --> 01:17:02,650
phase. I could be sitting,
drinking coffee and just trying

1363
01:17:02,650 --> 01:17:06,288
to think of ideas from thin air
for like 30 minutes. Instead, it

1364
01:17:06,288 --> 01:17:09,869
takes me three. Right? And like,
skips that phase and then it's

1365
01:17:09,869 --> 01:17:13,449
sort of narrows my focus. And it
will usually miss stuff that I

1366
01:17:13,449 --> 01:17:16,803
do when I talk about. So I'm
kind of hybridizing it with my

1367
01:17:16,803 --> 01:17:20,270
own thoughts. But it takes me
that 30 minutes. That's, that's

1368
01:17:20,270 --> 01:17:23,794
like half the usage of ChatGPT
for me. The other half, or that

1369
01:17:23,794 --> 01:17:27,375
kind of like the other half. I
was recently doing kind of like,

1370
01:17:27,375 --> 01:17:30,899
camera distortion and projection
to get a top down view from a

1371
01:17:30,899 --> 01:17:34,423
camera. And that was like, oh,
what libraries exist and how do

1372
01:17:34,423 --> 01:17:37,719
they behave and what's the
easiest? Oh yeah. That's great.

1373
01:17:37,719 --> 01:17:41,243
Gave me some ideas. Right. And
it's like, oh, here's an OpenCV

1374
01:17:41,243 --> 01:17:44,483
function. It takes these as
arguments. I'll go back, I'll

1375
01:17:44,483 --> 01:17:47,780
try it out, I'll code it, I'll
hit an error, I'll be like,

1376
01:17:47,780 --> 01:17:51,133
okay, what the heck is this
error? And that again, it saves

1377
01:17:51,133 --> 01:17:54,373
this iteration time a little
bit. So it takes the initial

1378
01:17:54,373 --> 01:17:57,954
development that usually takes,
you know, 45 minutes to 2 hours

1379
01:17:57,954 --> 01:18:00,000
down to ten minutes to half an
hour.

1380
01:18:00,000 --> 01:18:01,000
Audrow Nash: Totally.

1381
01:18:01,000 --> 01:18:03,826
Ilia Baranov: That's pretty
valuable. I find it becomes less

1382
01:18:03,826 --> 01:18:06,417
and less valuable taking from
that to a final code of a

1383
01:18:06,417 --> 01:18:09,149
complex piece of software, I
think it breaks more and more

1384
01:18:09,149 --> 01:18:11,880
and more and more to the point
where it's useless. But the

1385
01:18:11,880 --> 01:18:14,000
really early stuff, it speeds up
quite a bit.

1386
01:18:14,000 --> 01:18:17,335
Audrow Nash: Yep. Yeah. I think
like de-risking things

1387
01:18:17,335 --> 01:18:21,103
technically very quickly because
you can get a prototype real

1388
01:18:21,103 --> 01:18:24,932
fast is fantastic. You could see
which libraries work. You can

1389
01:18:24,932 --> 01:18:28,453
run into things that weren't
documented. Well, this super

1390
01:18:28,453 --> 01:18:32,344
valuable, in my experience. But
yeah, it's an interesting world

1391
01:18:32,344 --> 01:18:35,000
we're heading into. Yeah. With
all of this.

1392
01:18:35,000 --> 01:18:39,451
Stefan Seltz-Axmacher: I mean, I
think, like, we, we've we fooled

1393
01:18:39,451 --> 01:18:43,560
around with some stuff. So our
architecture, basically has a

1394
01:18:43,560 --> 01:18:47,943
common autonomy core nestled in
between two hardware abstraction

1395
01:18:47,943 --> 01:18:52,325
layers that separates us and are
given sensors for a vehicle, us

1396
01:18:52,325 --> 01:18:56,297
and the, the actual drive by
wire of a vehicle. And so the

1397
01:18:56,297 --> 01:19:00,543
most the most effort, the most,
dev work for bringing up a new

1398
01:19:00,543 --> 01:19:04,446
robot for us. Where do you type
of robot is building that

1399
01:19:04,446 --> 01:19:08,418
hardware abstraction layer?
Especially with me. That's the

1400
01:19:08,418 --> 01:19:12,595
drive by wire. So we've looked
at, hey, how can we use, AI to

1401
01:19:12,595 --> 01:19:16,772
get there? And we've seen some
like early cool stuff. But my,

1402
01:19:16,772 --> 01:19:21,087
my hunch is, I think, I think in
general, much of the work that

1403
01:19:21,087 --> 01:19:25,332
we have to do is good software,
best practices more than it is

1404
01:19:25,332 --> 01:19:29,373
like build AI to recognize a
novel situation and categorize

1405
01:19:29,373 --> 01:19:33,276
it into some, archetypical
situation that we have a known

1406
01:19:33,276 --> 01:19:37,316
behavior for because, like the
types of vehicles that we're

1407
01:19:37,316 --> 01:19:41,562
focusing on shouldn't be making
judgment calls like that. Even

1408
01:19:41,562 --> 01:19:43,000
another human driven.

1409
01:19:43,000 --> 01:19:44,000
Audrow Nash: Is very procedural.

1410
01:19:44,000 --> 01:19:47,977
Stefan Seltz-Axmacher: Yeah.
Like I've seen procedure docs

1411
01:19:47,977 --> 01:19:51,817
for sites where, you know,
they're driving a large wheel

1412
01:19:51,817 --> 01:19:55,931
loader that is, say, you know,
50 to 100,000 pounds. And the

1413
01:19:55,931 --> 01:20:00,183
rules for operating that wheel
loader are, you know, if you're

1414
01:20:00,183 --> 01:20:04,503
in reverse and you see someone
standing within 100m behind you,

1415
01:20:04,503 --> 01:20:08,891
turn the engine call, get on the
radio, call for help. If you're

1416
01:20:08,891 --> 01:20:12,937
driving from here to there and
you notice that this gate is

1417
01:20:12,937 --> 01:20:17,189
open, turn off the engine. Get
on the radio. Call for help. If

1418
01:20:17,189 --> 01:20:21,303
this beeper goes off, turn off
the engine. Get on the radio.

1419
01:20:21,303 --> 01:20:25,554
Call for help. These large, kind
of terrifying machines should

1420
01:20:25,554 --> 01:20:28,777
not be judging pedestrian
intents. It's kind of

1421
01:20:28,777 --> 01:20:32,000
deterministic. If pedestrian,
then stop moving.

1422
01:20:32,000 --> 01:20:35,309
Audrow Nash: So that makes
sense. Yeah, yeah. What a funny

1423
01:20:35,309 --> 01:20:38,846
thing that it I mean yeah. Makes
sense. They follow the simple

1424
01:20:38,846 --> 01:20:42,040
and it probably increases how
fast they can train people

1425
01:20:42,040 --> 01:20:45,520
rather than handling all the
appropriate conditions. And they

1426
01:20:45,520 --> 01:20:49,000
need one well-trained person
instead of every truck operator.

1427
01:20:49,000 --> 01:20:52,449
Stefan Seltz-Axmacher: And I
mean, I think those kind of

1428
01:20:52,449 --> 01:20:56,390
things like that, a lot of these
sites, things can go wrong with

1429
01:20:56,390 --> 01:21:00,331
like very once very simple rules
are broken. So it's a rehashing

1430
01:21:00,331 --> 01:21:03,472
constantly of simple,
straightforward rules. And in

1431
01:21:03,472 --> 01:21:07,290
part they're simple because it's
an easier for other people to

1432
01:21:07,290 --> 01:21:10,000
notice if someone's not
following the rules.

1433
01:21:10,000 --> 01:21:14,136
Audrow Nash: Oh, interesting.
Well let's see. So one thing

1434
01:21:14,136 --> 01:21:18,557
that I wanted to make sure we
talk about today, so I'm having,

1435
01:21:18,557 --> 01:21:22,906
a baby girl in bed about six
weeks from when we're recording.

1436
01:21:22,906 --> 01:21:27,398
You guys both have kids? Yep. I
don't often hear this discussed

1437
01:21:27,398 --> 01:21:31,605
with founders and this kind of
thing. I'd love to hear your

1438
01:21:31,605 --> 01:21:35,670
perspectives on how it is to
have a family. But also, you

1439
01:21:35,670 --> 01:21:39,805
guys are doing your startup.
Startup life is notorious for

1440
01:21:39,805 --> 01:21:44,297
being chaotic and busy all the
time and everything. I'd love to

1441
01:21:44,297 --> 01:21:48,718
hear what you think about it and
how your experience has been.

1442
01:21:48,718 --> 01:21:51,000
Yeah. You want to start, Stefan?

1443
01:21:51,000 --> 01:21:53,945
Stefan Seltz-Axmacher: Yeah.
I've done a startup with no kids

1444
01:21:53,945 --> 01:21:56,986
I was at a startup with. I get.
So if it's, It's good. It's got

1445
01:21:56,986 --> 01:21:58,000
a good middle ground.

1446
01:21:58,000 --> 01:21:59,000
Audrow Nash: Well informed.

1447
01:21:59,000 --> 01:22:03,343
Stefan Seltz-Axmacher: Yeah. It
is significantly harder. It is.

1448
01:22:03,343 --> 01:22:07,203
That is not a joke. I mean, I
think, there's an old, Sam

1449
01:22:07,203 --> 01:22:11,339
Altman quote from back when he
was just, a measly VC partner

1450
01:22:11,339 --> 01:22:15,751
and, you know, the CEO of an AI.
That was, you know, when you're

1451
01:22:15,751 --> 01:22:20,025
a startup founder, you get to
have, like, one hobby. Like that

1452
01:22:20,025 --> 01:22:24,023
hobby might be that you have
friends, it might be that you

1453
01:22:24,023 --> 01:22:27,814
like kite surfing. It might be.
It might be that you go

1454
01:22:27,814 --> 01:22:31,744
backpacking, but you get like
one. And when I didn't have

1455
01:22:31,744 --> 01:22:35,190
kids, my hobby was I had
friends. And now my hobby

1456
01:22:35,190 --> 01:22:39,326
largely is I have a kid. Yeah.
And that is, that is where my

1457
01:22:39,326 --> 01:22:43,462
free, not work nonproductive
time goes. Which is. Yeah, that

1458
01:22:43,462 --> 01:22:47,736
is a trade off. That is a thing.
But, you know, that's, that's

1459
01:22:47,736 --> 01:22:51,941
kind of that's kind of part of
the deal helps that, you know,

1460
01:22:51,941 --> 01:22:56,077
in my 30s and I'm a I did a
bunch of fun social things in my

1461
01:22:56,077 --> 01:22:59,938
20s. I'm okay with like the.
Yeah. Now the change. Yeah.

1462
01:22:59,938 --> 01:23:04,005
Yeah. The there's, a lot of
parallels, I think, my, so I, I

1463
01:23:04,005 --> 01:23:08,141
don't know if you guys have
signed up for this. You and your

1464
01:23:08,141 --> 01:23:11,794
you and your partner. There's a
great, Montessori toy

1465
01:23:11,794 --> 01:23:14,000
subscription called, love every.

1466
01:23:14,000 --> 01:23:15,000
Audrow Nash: Okay.

1467
01:23:15,000 --> 01:23:18,642
Stefan Seltz-Axmacher: Which I
highly, highly recommend. My

1468
01:23:18,642 --> 01:23:20,000
kids, my kids, you've.

1469
01:23:20,000 --> 01:23:22,384
Ilia Baranov: Got, like, you've
got, like, five months until

1470
01:23:22,384 --> 01:23:24,728
this starts becoming relevant.
To be clear, the first three

1471
01:23:24,728 --> 01:23:26,000
months, they're not going to be.

1472
01:23:26,000 --> 01:23:29,271
Stefan Seltz-Axmacher: Oh, no,
no, but Fritz loves his, like,

1473
01:23:29,271 --> 01:23:32,489
first three month toys. He like,
he like. Wow. Yeah. Like, I

1474
01:23:32,489 --> 01:23:35,868
mean, there were a lot of them
were like, black and white cards

1475
01:23:35,868 --> 01:23:39,032
with patterns on them, that you
like to look at. Like, we'd

1476
01:23:39,032 --> 01:23:42,410
probably plop them in front of
them anyway. Yeah. Like, what is

1477
01:23:42,410 --> 01:23:43,000
that? Yeah.

1478
01:23:43,000 --> 01:23:45,000
Audrow Nash: Oh, contrast. Very
interesting. Yeah.

1479
01:23:45,000 --> 01:23:49,580
Stefan Seltz-Axmacher: But, with
that toy cat in the first box,

1480
01:23:49,580 --> 01:23:54,088
there was a coffee cup that had
the phrase, the days are long,

1481
01:23:54,088 --> 01:23:58,159
but the years are short. And
that is exactly, a parallel

1482
01:23:58,159 --> 01:24:02,594
experience having a startup,
because actually, it's the thing

1483
01:24:02,594 --> 01:24:06,738
that Ilia was joking about
earlier on the podcast of a, a

1484
01:24:06,738 --> 01:24:11,100
beer getting humanoid. Actually
just being a vending machine

1485
01:24:11,100 --> 01:24:15,753
with, autonomous cart. That was
not a year ago. Earlier that was

1486
01:24:15,753 --> 01:24:20,188
like two and a half years ago.
We had that. Oh, there you go.

1487
01:24:20,188 --> 01:24:24,405
The, startup time is like a
black hole. Where I started my

1488
01:24:24,405 --> 01:24:29,058
first startup when I was, like,
26 or so. And now I'm 35, ish, I

1489
01:24:29,058 --> 01:24:33,638
think. I don't know, and I'm not
really sure. And just like, in

1490
01:24:33,638 --> 01:24:38,000
my mental model of myself, I'm
still, like, maybe 26.5. And.

1491
01:24:38,000 --> 01:24:41,000
Audrow Nash: Only six months is
advanced in some sense. Yeah.

1492
01:24:41,000 --> 01:24:42,000
Stefan Seltz-Axmacher: And.

1493
01:24:41,000 --> 01:24:43,000
Audrow Nash: I think years has
gone by otherwise.

1494
01:24:43,000 --> 01:24:46,257
Stefan Seltz-Axmacher: Like the
fact that having a kid is

1495
01:24:46,257 --> 01:24:49,686
similar is useful in that, like,
I can, I can grok the joint

1496
01:24:49,686 --> 01:24:53,000
mental model of it all where
there's so much happening all

1497
01:24:53,000 --> 01:24:55,000
the time, but also not a lot.
It's,

1498
01:24:55,000 --> 01:24:57,000
Audrow Nash: Yeah, it's like a
paradox. Yeah. For everything.

1499
01:24:57,000 --> 01:25:00,549
Stefan Seltz-Axmacher: Like,
we've like polymath has achieved

1500
01:25:00,549 --> 01:25:03,924
much less than I would have
wanted to achieve in the years

1501
01:25:03,924 --> 01:25:07,240
that we have. But like, it's
impossible to achieve what I

1502
01:25:07,240 --> 01:25:10,964
want to achieve in any, anytime.
But we've also done incredible,

1503
01:25:10,964 --> 01:25:14,513
amazing amounts of things. We've
been, we've, we've gotten to

1504
01:25:14,513 --> 01:25:17,771
work on some really cool
programs. We've been on lots of

1505
01:25:17,771 --> 01:25:21,262
different robots. We've seen
really neat things. And I think

1506
01:25:21,262 --> 01:25:24,869
the same thing is true with like
kids where like, if Fritz, my

1507
01:25:24,869 --> 01:25:28,360
son is now, we just moved from
saying, months to he's a year

1508
01:25:28,360 --> 01:25:29,000
and a half.

1509
01:25:29,000 --> 01:25:32,000
Audrow Nash: And wow, he's a big
transition.

1510
01:25:32,000 --> 01:25:35,871
Stefan Seltz-Axmacher: Yeah,
yeah, he can do lots of things,

1511
01:25:35,871 --> 01:25:39,806
but, like, he still can't talk
in sentences, so I don't know.

1512
01:25:39,806 --> 01:25:43,871
It must be dumb. But yeah, it's
cool. I know it. At least being

1513
01:25:43,871 --> 01:25:46,000
a dad for way longer than I
have.

1514
01:25:46,000 --> 01:25:49,471
Ilia Baranov: Yeah, I have, I
have a four year old daughter

1515
01:25:49,471 --> 01:25:53,000
and, one in the three quarter
year old son. More than three.

1516
01:25:53,000 --> 01:25:54,000
Audrow Nash: Quarter. That's
great.

1517
01:25:54,000 --> 01:25:56,000
Stefan Seltz-Axmacher: Yeah.

1518
01:25:56,000 --> 01:25:59,397
Ilia Baranov: I keep saying one
and a half, and then I have to

1519
01:25:59,397 --> 01:26:02,740
keep mentally pumping a lot of
things. Yeah, far beyond that.

1520
01:26:02,740 --> 01:26:06,247
Yeah. But the thing, the thing I
want to highlight a lot because

1521
01:26:06,247 --> 01:26:09,425
my experience is a little bit
different than stuff. And my

1522
01:26:09,425 --> 01:26:12,877
daughter was already born when,
you know, Stephan and I met and

1523
01:26:12,877 --> 01:26:16,329
started discussing, co-founding
and doing this. And the thing I

1524
01:26:16,329 --> 01:26:19,342
really want to highlight is the
enormous role that your

1525
01:26:19,342 --> 01:26:22,466
significant other spouse or
whoever plays in this. Right?

1526
01:26:22,466 --> 01:26:25,534
Like, I, I at the time was
working at Amazon, relatively

1527
01:26:25,534 --> 01:26:28,493
well-paid, you know, decent
relative stability, decent

1528
01:26:28,493 --> 01:26:31,507
program, was still working on
robots. I was still doing

1529
01:26:31,507 --> 01:26:34,740
something I like. And I brought
up this crazy idea of like,

1530
01:26:34,740 --> 01:26:36,000
let's drop all of that.

1531
01:26:36,000 --> 01:26:37,000
Stefan Seltz-Axmacher: Move.

1532
01:26:37,000 --> 01:26:40,988
Ilia Baranov: To a startup. But,
hopefully it works. And, what do

1533
01:26:40,988 --> 01:26:44,546
you think? And, and, you know,
all credit to my wife. She,

1534
01:26:44,546 --> 01:26:47,982
incredibly, was on board with
this, she really wanted to

1535
01:26:47,982 --> 01:26:51,663
understand the scope of the idea
and need my co-founder, you

1536
01:26:51,663 --> 01:26:55,528
know, try to understand if we're
building something reasonable.

1537
01:26:55,528 --> 01:26:59,454
And what is it? You know, is it
sane? And, and. Yeah. So I think

1538
01:26:59,454 --> 01:27:03,196
that's a huge chunk, right? Like
that, that makes all of this

1539
01:27:03,196 --> 01:27:06,571
possible. And her infinite
patience to have me suddenly

1540
01:27:06,571 --> 01:27:10,252
have a trip planned and needing
to have her take care of two

1541
01:27:10,252 --> 01:27:13,994
children and by herself. And
then a day before that, the trip

1542
01:27:13,994 --> 01:27:17,552
is no longer planned. But now
there's a new trip, but some

1543
01:27:17,552 --> 01:27:21,356
other point. And now I have to
go to this event and have to go

1544
01:27:21,356 --> 01:27:24,975
to that event. And that that
introduces a lot of chaos. And

1545
01:27:24,975 --> 01:27:28,779
so that that is, you know, the
only thing that keeps the sane,

1546
01:27:28,779 --> 01:27:32,583
I would say the only way that
this would be harder would be to

1547
01:27:32,583 --> 01:27:36,202
be a single parent and also run
a startup. And I can't even

1548
01:27:36,202 --> 01:27:37,000
imagine that.

1549
01:27:37,000 --> 01:27:40,606
Stefan Seltz-Axmacher: I, I feel
totally I don't know how anyone

1550
01:27:40,606 --> 01:27:43,817
is a single parent. That seems
to me like not even that's

1551
01:27:43,817 --> 01:27:46,915
inside of a startup. I mean,
like and just like, take a

1552
01:27:46,915 --> 01:27:49,000
shower, like, yeah, yeah, yeah,
yeah.

1553
01:27:49,000 --> 01:27:51,000
Audrow Nash: Yeah. Well, we
super, super.

1554
01:27:51,000 --> 01:27:52,000
Ilia Baranov: Human levels of
ability.

1555
01:27:52,000 --> 01:27:55,913
Audrow Nash: We have a puppy now
and the puppies too. You might

1556
01:27:55,913 --> 01:27:59,640
have heard Embarek a little
earlier, but he, is so much work

1557
01:27:59,640 --> 01:28:03,553
and I could not even imagine. I
imagine, like, single parenting

1558
01:28:03,553 --> 01:28:07,342
a puppy. Like there's no way in
hell I could do it. And I can

1559
01:28:07,342 --> 01:28:11,255
only imagine a kid will be more
for. Yeah, one or more kids. Oh

1560
01:28:11,255 --> 01:28:12,000
my goodness.

1561
01:28:12,000 --> 01:28:15,675
Stefan Seltz-Axmacher: But it's
also really cool. It's great.

1562
01:28:15,675 --> 01:28:19,410
It's really cool. It's really
amazing. It's, you know, ten out

1563
01:28:19,410 --> 01:28:22,000
of ten would do it again. Highly
recommend.

1564
01:28:22,000 --> 01:28:23,000
Audrow Nash: Two thumbs up.
Yeah.

1565
01:28:23,000 --> 01:28:26,888
Stefan Seltz-Axmacher: I don't
know, five stars. Like, there's

1566
01:28:26,888 --> 01:28:30,839
annoying stuff. Like he wakes up
screaming bloody murder in the

1567
01:28:30,839 --> 01:28:33,975
middle of the night, and,
there's disagreements on

1568
01:28:33,975 --> 01:28:37,738
parenting strategies in the
middle of the night because kids

1569
01:28:37,738 --> 01:28:41,564
screaming in the middle of the
night. Yeah, and everyone's at

1570
01:28:41,564 --> 01:28:45,326
their worst. And then there's,
like, you know, when, when he

1571
01:28:45,326 --> 01:28:49,340
went out for the day with, with
his, nanny, you know, he kept on

1572
01:28:49,340 --> 01:28:53,166
hugging me and my wife on his
way out the door, and it's just

1573
01:28:53,166 --> 01:28:56,929
like, oh, my God. Or like last
night, I'm given a bath. And,

1574
01:28:56,929 --> 01:29:00,880
like, I figure this, like, kind
of rubbery bottle. If I held it

1575
01:29:00,880 --> 01:29:04,831
underwater, I could squeeze it
and, like, a burst of water. And

1576
01:29:04,831 --> 01:29:08,656
he, like, laughed incredibly
loud. It's like it's the coolest

1577
01:29:08,656 --> 01:29:12,607
thing ever. It's pretty amazing.
And yeah, I mean, I think it's

1578
01:29:12,607 --> 01:29:16,621
just a work and kids, it's a all
the time always like if, if, if

1579
01:29:16,621 --> 01:29:20,509
instead we had done, you know
how to not for cats. Probably we

1580
01:29:20,509 --> 01:29:24,398
could just be hanging out right
now either rich or unemployed.

1581
01:29:24,398 --> 01:29:28,349
Yeah. But it's just it's a high
level of intensity. And I mean,

1582
01:29:28,349 --> 01:29:32,049
I think, you know, honestly,
before, before alien, I teamed

1583
01:29:32,049 --> 01:29:35,875
up when I was thinking about,
like, service and what I wanted

1584
01:29:35,875 --> 01:29:39,575
to do, I desperately did not
want to start another robotics

1585
01:29:39,575 --> 01:29:43,275
company. I, like, desperately
like I wanted to do some easy

1586
01:29:43,275 --> 01:29:47,226
mode SaaS company. And the thing
that I realized really quickly

1587
01:29:47,226 --> 01:29:50,926
is that none of those things
were exciting enough for me to

1588
01:29:50,926 --> 01:29:54,438
want to work on a Saturday
morning, instead of literally

1589
01:29:54,438 --> 01:29:58,201
anything I could do on Saturday.
On a Saturday morning, even

1590
01:29:58,201 --> 01:30:01,838
during like peak Covid, it was
like, I'd rather watch like

1591
01:30:01,838 --> 01:30:05,915
Netflix reruns than work on like
a trucking software company. And

1592
01:30:05,915 --> 01:30:09,364
like, robotics has this
incredible intensity that makes

1593
01:30:09,364 --> 01:30:13,190
it incredibly compelling and
rewarding. And I think life with

1594
01:30:13,190 --> 01:30:16,827
a kid is a similar sort of
uplift where, yes, it is really

1595
01:30:16,827 --> 01:30:20,653
hard that they scream in the
night. Yes, it is really hectic.

1596
01:30:20,653 --> 01:30:24,353
Yes, it is like a puppy on
steroids, however, but it's this

1597
01:30:24,353 --> 01:30:28,304
like extra cool thing in day to
day life, so I don't know, I'll

1598
01:30:28,304 --> 01:30:31,377
just be tired until he's
probably like 8 or 10 or

1599
01:30:31,377 --> 01:30:35,203
something. And yeah, and until
polymath is in the billions of

1600
01:30:35,203 --> 01:30:39,216
dollars a year revenue and I can
mostly be the founder who shows

1601
01:30:39,216 --> 01:30:42,917
up and waves his hand, hit his
hands, in a black turtleneck

1602
01:30:42,917 --> 01:30:46,491
every six months. But like,
yeah, that's I feel like it's

1603
01:30:46,491 --> 01:30:49,000
better living than the other way
around.

1604
01:30:49,000 --> 01:30:52,000
Audrow Nash: I suspect that's
true. Yeah. What do you think?

1605
01:30:52,000 --> 01:30:55,945
Ilia Baranov: Well, I was just
going to give a hint to the, to

1606
01:30:55,945 --> 01:30:59,953
yourself and to whoever might be
listening on, on the waking up

1607
01:30:59,953 --> 01:31:03,707
screaming and disagreements on
parenting styles. The golden

1608
01:31:03,707 --> 01:31:07,779
rule I read was that you have to
make up your mind on how you're

1609
01:31:07,779 --> 01:31:11,596
going to handle that crying
during the day. And the rule is,

1610
01:31:11,596 --> 01:31:14,905
you cannot change your mind at
night like there's no

1611
01:31:14,905 --> 01:31:18,786
discussion. You cannot change
it. Whatever you do in it. That

1612
01:31:18,786 --> 01:31:22,221
is the plan. Period. Like that
doesn't even enter your

1613
01:31:22,221 --> 01:31:26,039
thoughts, and you have to be
ready to reevaluate that during

1614
01:31:26,039 --> 01:31:29,983
the day when you have a little
bit more sleep every few weeks,

1615
01:31:29,983 --> 01:31:33,737
because it will change pretty
rapidly. But like the plan is

1616
01:31:33,737 --> 01:31:37,618
the plan, you don't discuss it
like that. That saved a lot of

1617
01:31:37,618 --> 01:31:38,000
chaos.

1618
01:31:38,000 --> 01:31:39,000
Audrow Nash: That is great.

1619
01:31:39,000 --> 01:31:42,282
Stefan Seltz-Axmacher: So have
you and, you and your, partner,

1620
01:31:42,282 --> 01:31:45,512
my wife, Michelle. Your wife?
Yeah. Have you and Michelle had

1621
01:31:45,512 --> 01:31:48,000
your parents try to chime in,
not raising your.

1622
01:31:48,000 --> 01:31:50,947
Audrow Nash: But they haven't.
So, I mean, we're already

1623
01:31:50,947 --> 01:31:54,105
getting parents talking, like,
giving us advice all the time

1624
01:31:54,105 --> 01:31:57,263
for things for the baby now,
puppy. They didn't really care.

1625
01:31:57,263 --> 01:32:00,526
Our dogs pretty well trained.
Okay, so, like, I don't know, we

1626
01:32:00,526 --> 01:32:02,000
generally impress with them.

1627
01:32:02,000 --> 01:32:06,173
Stefan Seltz-Axmacher: We I, we
had a really useful experience,

1628
01:32:06,173 --> 01:32:10,081
when we just had a puppy, pre
having a kid. Yeah. Where, we

1629
01:32:10,081 --> 01:32:13,857
went to my mother in law's
house, and my dog is great and

1630
01:32:13,857 --> 01:32:17,699
lovely and charming, but, you
know, neurotic and crazy and

1631
01:32:17,699 --> 01:32:21,739
scared of some of the other.
And, Cassie, my wife and I would

1632
01:32:21,739 --> 01:32:25,846
be arguing about some course of
action for the dog in front of

1633
01:32:25,846 --> 01:32:29,622
my mother in law, who then
would, like, take my side. And

1634
01:32:29,622 --> 01:32:33,662
then, like, Cassie felt teamed
up on. And then, like, I tried

1635
01:32:33,662 --> 01:32:37,306
to switch to her side, and
there's this musical chairs.

1636
01:32:37,306 --> 01:32:41,479
Who's our brother in law? And it
was just incredibly stressful.

1637
01:32:41,479 --> 01:32:45,652
And no one likes it. And yeah,
another piece of advice I'd give

1638
01:32:45,652 --> 01:32:49,626
you, similar to Eliza's on,
Night Cry is, in front of anyone

1639
01:32:49,626 --> 01:32:53,600
else. You guys agree? Don't
argue or just you're on the set.

1640
01:32:53,600 --> 01:32:57,840
If you're arguing about what to
do about Nate crying in front of

1641
01:32:57,840 --> 01:33:02,013
someone else and they chime in
supporting one of you, you know,

1642
01:33:02,013 --> 01:33:04,000
support whoever is being, who.

1643
01:33:04,000 --> 01:33:06,000
Audrow Nash: Yeah. Or ganged up
on. Yeah.

1644
01:33:06,000 --> 01:33:08,000
Ilia Baranov: Better just not to
discuss it.

1645
01:33:08,000 --> 01:33:11,088
Audrow Nash: Yeah I think yeah.
Don't discuss it in front of

1646
01:33:11,088 --> 01:33:14,125
anyone else or this kind of
thing. Not a, like an impromptu

1647
01:33:14,125 --> 01:33:15,000
passion argument.

1648
01:33:15,000 --> 01:33:17,474
Ilia Baranov: Yeah. Impassioned
argument. There's many wrong

1649
01:33:17,474 --> 01:33:19,000
answers, but there's no right
answer.

1650
01:33:19,000 --> 01:33:21,000
Stefan Seltz-Axmacher:
Realistically.

1651
01:33:21,000 --> 01:33:24,785
Audrow Nash: So that's
hilarious. Yeah. One thing that

1652
01:33:24,785 --> 01:33:29,061
you may laugh at with this is we
moved to San Antonio, Texas.

1653
01:33:29,061 --> 01:33:33,336
And, when we ask people or when
people ask, about like how we

1654
01:33:33,336 --> 01:33:36,000
chose it, we say it was the
least bad.

1655
01:33:36,000 --> 01:33:39,000
Stefan Seltz-Axmacher: Move, all
that.

1656
01:33:39,000 --> 01:33:40,000
Audrow Nash: In a similar vein.

1657
01:33:40,000 --> 01:33:43,000
Stefan Seltz-Axmacher: Yeah.
Yes. Yeah.

1658
01:33:43,000 --> 01:33:48,178
Audrow Nash: Any other advice of
kids, that you can think of or

1659
01:33:48,178 --> 01:33:52,945
any other things to share? Maybe
we'll go around one more.

1660
01:33:52,945 --> 01:33:55,000
Stefan. Anything for you?

1661
01:33:55,000 --> 01:33:59,069
Stefan Seltz-Axmacher: I think
on the kids side, I said all my

1662
01:33:59,069 --> 01:34:02,942
stuff earlier. Probably got
more. Yeah. I mean, you'd think

1663
01:34:02,942 --> 01:34:07,142
so. If you're working on robots,
we'd love to. We'd love to talk

1664
01:34:07,142 --> 01:34:11,015
to you in general. Where. So
this, this, this approach that

1665
01:34:11,015 --> 01:34:14,690
we're talking about, about
mostly focusing on these big,

1666
01:34:14,690 --> 01:34:18,562
chunky, industrial things that
approval that are, you know,

1667
01:34:18,562 --> 01:34:22,500
obviously vulnerable, have put
us in an interesting position

1668
01:34:22,500 --> 01:34:26,373
where we're starting to have
cash flow, positive bumps who,

1669
01:34:26,373 --> 01:34:29,917
and that that's so exciting.
Yeah. Like, there's a big

1670
01:34:29,917 --> 01:34:33,789
fundamental difference from the
groups that I know, who are

1671
01:34:33,789 --> 01:34:37,924
saying, like, hey, we need raise
$100 million to get a bunch of

1672
01:34:37,924 --> 01:34:42,125
synthetic data to then maybe get
$500 million and then maybe get

1673
01:34:42,125 --> 01:34:46,260
3 million. And I think, I think
we're in a time where there's a

1674
01:34:46,260 --> 01:34:50,460
lot of noise going on elsewhere
about those sorts of approaches,

1675
01:34:50,460 --> 01:34:54,595
and 99% of those companies won't
exist in three years. But like

1676
01:34:54,595 --> 01:34:58,468
the folks who can, can actually
build a robot that actually

1677
01:34:58,468 --> 01:35:02,406
creates real value are, you
know, half of us will be around,

1678
01:35:02,406 --> 01:35:06,409
you know, like, you still screw
up a pretty generous product.

1679
01:35:06,409 --> 01:35:07,000
But like.

1680
01:35:07,000 --> 01:35:09,000
Ilia Baranov: Me and us will be
around.

1681
01:35:09,000 --> 01:35:12,231
Stefan Seltz-Axmacher: That most
of the companies will die, not

1682
01:35:12,231 --> 01:35:13,000
100%, but most.

1683
01:35:13,000 --> 01:35:15,175
Audrow Nash: Maybe not in three
years. Yeah, or they'll be

1684
01:35:15,175 --> 01:35:16,000
acquired or something.

1685
01:35:16,000 --> 01:35:20,005
Stefan Seltz-Axmacher: Yeah, but
like, many of them are too big

1686
01:35:20,005 --> 01:35:23,438
to be acquired. You know,
there's not a lot of natural

1687
01:35:23,438 --> 01:35:26,808
acquirers for $10 billion
products. Humanoid robotics

1688
01:35:26,808 --> 01:35:30,750
company fair point. Like there
was when that company was worth

1689
01:35:30,750 --> 01:35:34,564
$200 million and it was 2016 and
interest rates were low and

1690
01:35:34,564 --> 01:35:38,315
like, cool, like Google or
whoever, cash was expensive now,

1691
01:35:38,315 --> 01:35:42,003
for sure. Like they did. Fine.
But now at like the current

1692
01:35:42,003 --> 01:35:45,372
valuations, that's just not
feasible. So I think this

1693
01:35:45,372 --> 01:35:49,250
there's, there's this incredible
like, slope of enlightenment

1694
01:35:49,250 --> 01:35:53,001
that we're currently on for
robotics around building things

1695
01:35:53,001 --> 01:35:57,007
that are actually buildable and
actually sellable for, but, but

1696
01:35:57,007 --> 01:36:00,630
can be built today and I
encourage more people to look at

1697
01:36:00,630 --> 01:36:04,000
getting on that slope instead of
the next hype cycle.

1698
01:36:04,000 --> 01:36:07,891
Audrow Nash: Oh yeah. Let's see.
Yeah. I guess, wrapping up,

1699
01:36:07,891 --> 01:36:11,717
because I see that we are coming
close to our end time, so,

1700
01:36:11,717 --> 01:36:15,478
Stefan, that sounds like what
you want people to take away

1701
01:36:15,478 --> 01:36:19,369
from this, which makes good
sense. Yeah. What would you hope

1702
01:36:19,369 --> 01:36:23,000
sticks out in people's heads?
From this whole interview?

1703
01:36:23,000 --> 01:36:26,919
Ilia Baranov: Yeah. You know,
definitely, definitely. Stefan

1704
01:36:26,919 --> 01:36:31,035
making the points on the on the
business and robotics side, I'd

1705
01:36:31,035 --> 01:36:34,758
say I gave kind of a talk, last
year at Robot Business on

1706
01:36:34,758 --> 01:36:38,743
running a startup and running a
robotics started, and I think

1707
01:36:38,743 --> 01:36:42,858
it's on our YouTube channel if
people are interested. And, that

1708
01:36:42,858 --> 01:36:46,908
what I ended up, that one was
kind of two critical things. One

1709
01:36:46,908 --> 01:36:50,958
or let's add one more, but two
to start. One is even if you're

1710
01:36:50,958 --> 01:36:54,943
building robots, you have to be
humans. And I think that, you

1711
01:36:54,943 --> 01:36:59,058
know, like having a family, both
myself and Stephan and many of

1712
01:36:59,058 --> 01:37:03,108
our team members are younger and
starting families. There's no

1713
01:37:03,108 --> 01:37:06,832
blood from a stone that works
well with complex technical

1714
01:37:06,832 --> 01:37:10,947
things like kind of crunch time
that comes from gaming industry

1715
01:37:10,947 --> 01:37:14,801
or heavy software dev does not
work well and is, in my mind

1716
01:37:14,801 --> 01:37:18,917
pretty clearly the less capital
efficient way to do things over

1717
01:37:18,917 --> 01:37:23,097
time, like even for shareholder
value, like it's almost always a

1718
01:37:23,097 --> 01:37:27,213
negative and so be humans. While
building robots is number one.

1719
01:37:27,213 --> 01:37:31,263
And then the other thing is the
lesson we keep having to learn

1720
01:37:31,263 --> 01:37:35,117
over and over, and as core to
robots is you have to do less

1721
01:37:35,117 --> 01:37:38,971
work, love, do less, but do a
better love. Focused approach

1722
01:37:38,971 --> 01:37:42,760
like do that as much as you can.
And I think both of those

1723
01:37:42,760 --> 01:37:46,810
things, are applicable almost
equally to child rearing as they

1724
01:37:46,810 --> 01:37:50,990
are building robots. Very, very
common skill sets of like, don't

1725
01:37:50,990 --> 01:37:54,844
try to bring your kid to 30
different events like, yeah, do

1726
01:37:54,844 --> 01:37:57,000
three, but do them really
nicely.

1727
01:37:57,000 --> 01:38:00,483
Stefan Seltz-Axmacher: What is
child rearing if not building a

1728
01:38:00,483 --> 01:38:02,000
really slow dev time robot.

1729
01:38:02,000 --> 01:38:05,053
Ilia Baranov: Yeah, yeah. Oh
yeah, I think of that all the

1730
01:38:05,053 --> 01:38:08,316
time, especially with a with a
newborn. You're like, oh like I

1731
01:38:08,316 --> 01:38:11,316
keep, I keep giving this
example. When my son was born, I

1732
01:38:11,316 --> 01:38:14,474
noticed he would only smile if
my face was oriented the same

1733
01:38:14,474 --> 01:38:17,579
way. His face was like, he can
only recognize faces in this

1734
01:38:17,579 --> 01:38:20,474
orientation. Wow. And then,
yeah, and that's apparently

1735
01:38:20,474 --> 01:38:23,684
pretty common. And then like, a
some amount of time, I forget

1736
01:38:23,684 --> 01:38:27,000
exactly when suddenly it snapped
that like, even if my face was

1737
01:38:27,000 --> 01:38:30,263
rotated this way, it was still a
human face and it was like an

1738
01:38:30,263 --> 01:38:33,263
overnight change of like,
rotation works now for a vision

1739
01:38:33,263 --> 01:38:34,000
like that job.

1740
01:38:34,000 --> 01:38:35,000
Stefan Seltz-Axmacher: Yeah, it
works.

1741
01:38:35,000 --> 01:38:37,375
Ilia Baranov: Right. And like,
you could see these things

1742
01:38:37,375 --> 01:38:38,000
turning on and.

1743
01:38:38,000 --> 01:38:41,000
Stefan Seltz-Axmacher: I don't
think even had it enabled.

1744
01:38:41,000 --> 01:38:42,000
Audrow Nash: Yeah. Yeah.

1745
01:38:42,000 --> 01:38:44,000
Ilia Baranov: No, no, nothing
real impressive, you know.

1746
01:38:44,000 --> 01:38:45,000
Stefan Seltz-Axmacher: Yeah.

1747
01:38:45,000 --> 01:38:50,243
Ilia Baranov: Reinforcement
learning I don't know. But yeah.

1748
01:38:50,243 --> 01:38:54,000
So, be humans and, do less but
do a better.

1749
01:38:54,000 --> 01:38:55,915
Stefan Seltz-Axmacher: Oh, yeah.
Well, thank you so much for

1750
01:38:55,915 --> 01:38:57,000
having us on today. So lot of
fun.

1751
01:38:57,000 --> 01:38:59,682
Audrow Nash: Oh, yeah. Well,
thank you both. Do you want to

1752
01:38:59,682 --> 01:39:01,000
plug your podcast real quick?

1753
01:39:01,000 --> 01:39:04,015
Stefan Seltz-Axmacher: So you
can learn more about us at

1754
01:39:04,015 --> 01:39:07,138
Polymath robotics.com. And our
podcast is automated.com or

1755
01:39:07,138 --> 01:39:08,000
automated parts.

1756
01:39:08,000 --> 01:39:10,000
Audrow Nash: I'll include the
link.

1757
01:39:10,000 --> 01:39:12,000
Ilia Baranov: It's at the bottom
of the webpage.

1758
01:39:12,000 --> 01:39:13,880
Stefan Seltz-Axmacher: It's
called Automate It. It's

1759
01:39:13,880 --> 01:39:15,000
wherever you get your podcasts.

1760
01:39:15,000 --> 01:39:17,000
Audrow Nash: Okay. Hell, yeah.
All right. Thank you both.

1761
01:39:17,000 --> 01:39:18,000
Stefan Seltz-Axmacher: Thank
you.

1762
01:39:18,000 --> 01:39:19,000
Ilia Baranov: See you next time.

1763
01:39:19,000 --> 01:39:20,000
Audrow Nash: Bye, everyone.
