1
00:00:00,151 --> 00:00:00,871
Hi everyone.

2
00:00:00,871 --> 00:00:06,295
Welcome to the Bright Signal Podcast, where we cut through the noise and bring you the
latest tech news and interviews.

3
00:00:06,295 --> 00:00:08,626
My name is Marillo and I'm joined by my friend Bart.

4
00:00:08,626 --> 00:00:09,636
Hey Bart.

5
00:00:10,021 --> 00:00:14,587
And we just had a chat with Alexandre Pereira from twenty five one.

6
00:00:14,587 --> 00:00:17,308
Yeah, it was a very, very interesting chat.

7
00:00:17,308 --> 00:00:17,908
2501.

8
00:00:17,908 --> 00:00:25,413
What they do is that they build autonomous SRE agents and bring them to uh large corporate
enterprises.

9
00:00:25,413 --> 00:00:36,218
Why this is very interesting is that they, what their value proposition is, is that they
basically do root cause analysis on anything that goes wrong on their infrastructure, but

10
00:00:36,218 --> 00:00:38,580
also try to automate the remediation.

11
00:00:38,580 --> 00:00:47,722
which in these type of enterprises is typically not easy because they have a complex
infrastructure, a large part on-prem, part in the cloud, part of it modern, part of it

12
00:00:47,722 --> 00:00:48,102
legacy.

13
00:00:48,102 --> 00:00:49,713
So it's a very complex environment.

14
00:00:49,713 --> 00:00:55,535
And if you can execute something like this correctly, then it becomes a very, very
interesting file proposition.

15
00:00:55,535 --> 00:00:55,866
Indeed.

16
00:00:55,866 --> 00:00:57,431
Yeah, was a very nice chat.

17
00:00:57,431 --> 00:00:59,647
So yeah, let's go to the interview.

18
00:00:59,763 --> 00:01:00,174
to the interview.

19
00:01:00,155 --> 00:01:00,877
welcome everyone.

20
00:01:00,877 --> 00:01:01,919
Hey Bart.

21
00:01:02,168 --> 00:01:03,171
And hey Alexandre.

22
00:01:03,171 --> 00:01:04,245
Alexandre Pereira.

23
00:01:04,245 --> 00:01:05,661
Is that that that sounds good?

24
00:01:05,661 --> 00:01:06,535
How are we doing?

25
00:01:06,535 --> 00:01:07,830
Warm outside.

26
00:01:07,830 --> 00:01:09,214
Or yeah yeah yeah.

27
00:01:09,214 --> 00:01:12,720
thanks for joining us, Alexander, and welcome to the Bright Signal podcast.

28
00:01:12,720 --> 00:01:18,319
maybe you can dive in already and like for the people that don't know you, like would you
like to introduce yourself, share a bit of your background and

29
00:01:18,319 --> 00:01:18,980
Yeah, sure guys.

30
00:01:18,980 --> 00:01:21,561
So Alex, I'm obviously French.

31
00:01:21,561 --> 00:01:23,082
You will notice with the accent.

32
00:01:23,082 --> 00:01:28,074
I'm like, let's say an entrepreneur since 10, 12 years now.

33
00:01:28,074 --> 00:01:30,286
Working essentially on tech for enterprise.

34
00:01:30,286 --> 00:01:39,063
Also spent the large majority of my starting career into corporate world, launching
e-commerce websites, working with like these big brands in Europe, in France especially.

35
00:01:39,063 --> 00:01:42,455
Doing that like for the last 20 years now and...

36
00:01:42,455 --> 00:01:44,017
It's very cool spot to be.

37
00:01:44,017 --> 00:01:49,440
And since two years now, we work on AI, ah on agent-y AI especially for corporate as well.

38
00:01:49,440 --> 00:01:52,162
So that's gonna be one of the topic today.

39
00:01:52,162 --> 00:01:52,803
yeah, that's it.

40
00:01:52,803 --> 00:01:56,796
So I would consider myself as a, I don't know, a mix between the tech guy.

41
00:01:56,796 --> 00:02:02,521
I'm not an engineering by training, but I'm an engineering by love and trying to code when
I can.

42
00:02:02,521 --> 00:02:05,903
But now, yeah, I'm not coding anymore so that much.

43
00:02:05,903 --> 00:02:08,808
I'm just trying to make this thing move forward.

44
00:02:08,808 --> 00:02:10,647
and do the best product we can.

45
00:02:10,647 --> 00:02:14,712
So you don't you didn't study any like uh computer science or anything like that.

46
00:02:14,712 --> 00:02:19,139
You just like on the side, you're just tinkering with it and and and learn by that by
doing.

47
00:02:19,139 --> 00:02:22,384
Yeah, I was kind of a cliché nerd when I was a kid.

48
00:02:22,384 --> 00:02:27,963
ah My parents and my family were like super nerdy as well.

49
00:02:27,963 --> 00:02:31,428
I was playing video game like on Sega Master System with my grandparents.

50
00:02:31,428 --> 00:02:33,121
So that was kind of fun.

51
00:02:33,121 --> 00:02:43,376
I started to have my own computer quite early, coding, coding to pay, you know, I was
having like this 125cc moped and I needed to pay everything for it, so at some point I was

52
00:02:43,376 --> 00:02:50,626
like, should you work in McDonald's or should you work in tech and do websites for the
groceries or for the bakery?

53
00:02:50,698 --> 00:02:56,334
I started to do that and ultimately I landed in this company that is named Decathlon that
you may know.

54
00:02:56,334 --> 00:02:57,646
It's a sportswear group.

55
00:02:57,646 --> 00:03:05,723
The end of marketing of this brand there was completely crazy and it was like, okay,
you're 18 years old but you look fun and you know how to code websites so come to work

56
00:03:05,723 --> 00:03:06,114
with us.

57
00:03:06,114 --> 00:03:11,266
I was not having any other experience than that and that launched the thing, so yeah.

58
00:03:11,266 --> 00:03:12,287
Super cool.

59
00:03:13,399 --> 00:03:15,322
You also have a sim something similar, right Bart?

60
00:03:15,322 --> 00:03:21,504
Also tinkering with the stuff and you you you went to officially to tech later, right, in
life as well.

61
00:03:21,504 --> 00:03:23,999
There are some parallels indeed, yeah.

62
00:03:23,999 --> 00:03:24,420
Okay, cool.

63
00:03:24,420 --> 00:03:30,271
And also how did you so we're gonna talk about uh twenty five one or two thousand two
thousand five hundred and one?

64
00:03:30,271 --> 00:03:33,639
I don't know how twenty five one, yeah.

65
00:03:33,639 --> 00:03:38,331
Everyone is asking about it and I think we figured it out only two or three months ago.

66
00:03:38,331 --> 00:03:40,971
We're like, everyone is asking, how do we pronounce it?

67
00:03:40,971 --> 00:03:42,282
And we're like, we don't know.

68
00:03:42,282 --> 00:03:44,262
So we have to choose.

69
00:03:44,702 --> 00:03:46,373
So 2501 in English.

70
00:03:46,373 --> 00:03:47,745
Okay, there we go.

71
00:03:47,745 --> 00:03:50,387
And uh it's it's yeah, we mentioned agentic, right?

72
00:03:50,387 --> 00:03:54,161
How did you how did you get in touch with uh agents and all these different things?

73
00:03:54,161 --> 00:03:56,384
Did you just jump on it early on or?

74
00:03:56,384 --> 00:03:59,542
So, yeah, if you want the genesis of the project...

75
00:03:59,542 --> 00:04:06,005
Basically, we were trying to create LLM chaining technologies at the very, very start of
it.

76
00:04:06,005 --> 00:04:11,127
I don't know, we were on GPT 3.0 or something like that, 3.5 maybe.

77
00:04:11,127 --> 00:04:16,400
You started to have these guys online on GitHub doing what we call agent orchestration.

78
00:04:16,400 --> 00:04:18,280
It was not named agent at this time.

79
00:04:18,280 --> 00:04:22,973
It was name, I don't know, LLM orchestration or chaining LLMs.

80
00:04:22,973 --> 00:04:28,715
And you had, I don't know if you heard about this kind of GitHub repositories like
BabyAGI, AutoGP.

81
00:04:28,715 --> 00:04:30,295
Auto GPT went crazy.

82
00:04:30,295 --> 00:04:34,166
I think was over to GitHub's fastest repository to get on.

83
00:04:34,166 --> 00:04:36,026
And they were just doing something simple.

84
00:04:36,026 --> 00:04:38,577
They were just chaining and looping on LLM response.

85
00:04:38,577 --> 00:04:40,788
And so it was just very, very basic at the beginning.

86
00:04:40,788 --> 00:04:43,579
But it was like the start of what could be Agentic.

87
00:04:43,579 --> 00:04:44,950
I'll call it Agentic further.

88
00:04:44,950 --> 00:04:51,832
So yeah, it started all of this kind of crowd, started to innovate a bit like in startup
garage mode, GitHub.

89
00:04:51,832 --> 00:04:54,343
So we started to do the same with one of my colleagues.

90
00:04:54,343 --> 00:04:56,266
just like trying to see if we can do something.

91
00:04:56,266 --> 00:05:00,502
And I was not at all in this mood of, I don't know, creating a new company or whatever.

92
00:05:00,502 --> 00:05:02,595
We're like just playing with the tech.

93
00:05:02,595 --> 00:05:09,732
And I don't know, a few months after, early 2024, we started to see fundraising happening
in there.

94
00:05:09,732 --> 00:05:10,377
Mm-hmm.

95
00:05:10,377 --> 00:05:21,040
So I think the one that was significant and that was kind of the compelling event starting
everything on Agentic was Cognition Lab DevIn product launch um that raised, I think, 100

96
00:05:21,040 --> 00:05:24,002
million from scratch at the beginning by...

97
00:05:24,002 --> 00:05:26,160
um

98
00:05:26,160 --> 00:05:26,381
US.

99
00:05:26,381 --> 00:05:34,337
So everyone after that was like, okay, there is something interesting because he found
this fund, he's interested in this kind of thing that could be very good maybe in the

100
00:05:34,337 --> 00:05:34,627
future.

101
00:05:34,627 --> 00:05:45,042
So it started to get credibility to these GitHub repositories and all these innovators
that were trying to do something and trying to make sure that, I don't know, the VCs

102
00:05:45,042 --> 00:05:48,165
getting there put a lot of legitimacy in the end.

103
00:05:48,477 --> 00:05:49,928
And so I started to do the same.

104
00:05:49,928 --> 00:05:51,530
was like, okay, maybe there's something to do.

105
00:05:51,530 --> 00:05:56,485
And at this point I was a bit at the end of my previous entrepreneurial story, which was
in Japan.

106
00:05:56,485 --> 00:05:58,638
And I was like, okay, let's look at what we can do.

107
00:05:58,638 --> 00:06:05,314
Ultimately, like in a few months, we raised our first million euro that became like a two
million round precede.

108
00:06:05,314 --> 00:06:07,229
And then we went into that.

109
00:06:07,229 --> 00:06:16,594
Then I can explain later how the construction of the product, but the very origin of it
was just like pure innovation research with people trying stuff with this technology.

110
00:06:16,594 --> 00:06:26,800
And I like that because we are entering a phase again with AI where we innovate a lot
based on pure, how to say, pure random ideas sometime.

111
00:06:26,800 --> 00:06:30,702
And this chaining of LLM was not at all anticipated by guys.

112
00:06:30,762 --> 00:06:33,898
OpenAI didn't anticipated it, Anthropeak didn't anticipated it.

113
00:06:33,898 --> 00:06:34,995
It's just like two years ago.

114
00:06:34,995 --> 00:06:40,059
that Sam Altman says, no, no, the chat is not what will be the end product, but there is
something else cooking.

115
00:06:40,059 --> 00:06:43,882
And there is just one guy stupidly saying, but maybe it can generate code.

116
00:06:43,882 --> 00:06:46,684
And if it can generate code, it can generate action itself.

117
00:06:46,684 --> 00:06:48,345
And so here we go.

118
00:06:48,345 --> 00:06:48,988
So.

119
00:06:48,988 --> 00:06:54,892
So that means that when you created the company, you didn't really have this exact vision
that you have today, like what you want to do.

120
00:06:54,892 --> 00:07:02,667
It really started from you see that there is a value in this and we're gonna see an
experiment and innovate and see what direction we can take this.

121
00:07:02,667 --> 00:07:05,069
Yeah, that was basically like this kind of thing.

122
00:07:05,069 --> 00:07:12,619
We were like, okay, how can we create something that is autonomous in a computer terminal
system, in a shell script system?

123
00:07:12,619 --> 00:07:18,945
And how we can make something that will be, don't know, the next AIOS or the next
developer experience?

124
00:07:18,945 --> 00:07:21,112
So everyone was kind of touching everything everywhere.

125
00:07:21,112 --> 00:07:23,694
was like two years ago, so Cloud Code was nothing.

126
00:07:23,694 --> 00:07:29,494
You had like bit of projects trying to do this with, I don't know, Copilot was trying to
do some stuff.

127
00:07:29,494 --> 00:07:37,998
You had like some guys at Y Combinator starting to create companies similarly, like I
don't know, Pythagora, Marble, that has pivoted or died since.

128
00:07:37,998 --> 00:07:44,349
And what I liked is that, yeah, it was very, very R &D driven and...

129
00:07:44,349 --> 00:07:46,702
the need started to emerge after that.

130
00:07:46,702 --> 00:07:49,335
And it's not a need that is a new need.

131
00:07:49,335 --> 00:07:57,472
You don't create a market, and I will come back to our market later, but the technology is
so huge, so powerful, that you can apply it to a lot of stuff.

132
00:07:57,472 --> 00:08:05,249
And this was what was the, know, kind of the guess, the gut feeling about it with the
first investors and with...

133
00:08:05,249 --> 00:08:14,944
colleagues were like we're gonna figure it out it's just like so big so huge it's a
once-of-a-lifetime maybe once a system to every revolution right now that we're that if we

134
00:08:14,944 --> 00:08:16,838
play well this thing it might work

135
00:08:16,838 --> 00:08:26,953
And you said like we like your co-founders as well, like or was it like back then when you
s were being curious and like trying things out and tinkering, was it already the the

136
00:08:26,953 --> 00:08:27,614
co-founders?

137
00:08:27,614 --> 00:08:28,515
Was it like colleagues?

138
00:08:28,515 --> 00:08:29,567
Was it just friends?

139
00:08:29,567 --> 00:08:35,745
How like what what are the when you were sharing these ideas, who what was the team or who
are the people that you were discussing these things with?

140
00:08:35,745 --> 00:08:40,021
So at beginning we were like kind of racing with one of the guys that is my friend.

141
00:08:40,021 --> 00:08:41,953
He was having his project, I was doing my project.

142
00:08:41,953 --> 00:08:46,529
We were like a bit competing every night trying to code what could be the next thing.

143
00:08:46,529 --> 00:08:52,889
Ultimately, he joined me on the project because I was like, okay, now I think we have
something interesting.

144
00:08:52,889 --> 00:08:55,220
I have a bit of pre-seed money arriving.

145
00:08:55,220 --> 00:08:55,960
Join me.

146
00:08:55,960 --> 00:08:56,600
We don't compete.

147
00:08:56,600 --> 00:08:57,771
We continue together.

148
00:08:57,771 --> 00:09:00,382
And he was not the co-founder, but the first employee of the company.

149
00:09:00,382 --> 00:09:06,173
And then my co-founder Alex joined me a few months after from his current thing.

150
00:09:06,173 --> 00:09:09,924
So yeah, it was a bit of a story of just...

151
00:09:09,924 --> 00:09:11,927
I don't know, coding and seeing how it goes.

152
00:09:11,927 --> 00:09:13,720
And Investor, we are like in the same way.

153
00:09:13,720 --> 00:09:20,171
We are super lucky to have these guys that joined on as well because we got like, think,
what is the best configuration of Proceed Investors.

154
00:09:20,171 --> 00:09:23,147
And we're just super supportive on, yeah, you're gonna figure it out.

155
00:09:23,147 --> 00:09:29,208
And that also means that you got your pre-seed round, also really before there was a very
strong ID to sell.

156
00:09:29,208 --> 00:09:32,360
Yeah, and that's what it should be, in my opinion.

157
00:09:32,360 --> 00:09:37,954
This is the thing, know, the VC world is a bit like paradoxes because there is always a
reason to not invest.

158
00:09:37,954 --> 00:09:40,786
There is always one million reasons to not invest in a company.

159
00:09:40,786 --> 00:09:44,277
But in the end, if you look at pre-seed rounds, what do you look for?

160
00:09:44,277 --> 00:09:52,133
You're looking like at co-founders, team, a bit the vision, but the vision can be a bit
blurry and gross still.

161
00:09:52,133 --> 00:09:54,295
But you're not looking at any market trends.

162
00:09:54,295 --> 00:10:00,342
You're not looking at, okay, market research can happen, but not if you are purely
tech-driven, R &D-driven.

163
00:10:00,342 --> 00:10:02,935
It's different than just if you launch a service company.

164
00:10:02,935 --> 00:10:05,358
Yeah.

165
00:10:05,358 --> 00:10:12,464
Maybe to make a step to today, like what is the product that you're offering today as
2501?

166
00:10:12,464 --> 00:10:12,915
Yeah, sure.

167
00:10:12,915 --> 00:10:16,149
So if you want to summarize what we do in a nutshell.

168
00:10:16,149 --> 00:10:21,642
we automate what is IT maintenance and cloud maintenance of infrastructure for large
corporates.

169
00:10:21,642 --> 00:10:31,317
So our job is to make easier the way to manage this large hybrid cloud on-premise data
centers, IT corporation infrastructures.

170
00:10:31,317 --> 00:10:36,371
So in a way that this thing has not changed for the last 30 years.

171
00:10:36,371 --> 00:10:38,072
It has been a bit modernized.

172
00:10:38,072 --> 00:10:40,193
They did a bit of automation in the job.

173
00:10:40,193 --> 00:10:44,636
The arrival of the cloud changed a bit on stuff, but it's still very complex stuff that's
happening.

174
00:10:44,636 --> 00:10:50,509
If you look at S &P 500, CAC 40 in France, these guys have large infrastructure, super
difficult to manage.

175
00:10:50,509 --> 00:10:54,109
By design, they are not state of the art because they their product.

176
00:10:54,109 --> 00:10:55,449
They are not a tech company.

177
00:10:55,449 --> 00:10:57,269
We are not speaking like the NASDAQ companies.

178
00:10:57,269 --> 00:11:02,320
We are speaking like the whole guys that are managing the main part of the economy.

179
00:11:02,320 --> 00:11:09,811
so it has been like for the last decades a very kind of simple job because it was human
led.

180
00:11:09,811 --> 00:11:12,811
A lot of shoring was done as well.

181
00:11:13,145 --> 00:11:15,407
And right now this world needs to change.

182
00:11:15,407 --> 00:11:26,987
what we've seen is that technology like agent I can kind of help this transition and help
these guys to modernize their park to modernize their processes to optimize what is like

183
00:11:26,987 --> 00:11:37,898
their velocity of production release their maintenance responsiveness and Also reply to a
lot of topics that are like quite crunchy right now like sovereignty of data.

184
00:11:37,898 --> 00:11:39,519
So at all so there is a

185
00:11:39,519 --> 00:11:50,547
of trends also right now on trying to avoid offshoring, trying to get back on EU soil, on
the US soil, make the data private again, back to the data center, back to the on-prem.

186
00:11:50,547 --> 00:11:52,668
So this is all this world that we're serving.

187
00:11:52,668 --> 00:12:03,506
So we have a big agent-tki solution that is now able to manage all of these services that
are like not always AWS, GCP, Azure, but could be like some VMware server, some Fortinet

188
00:12:03,506 --> 00:12:05,408
firewalls, stuff like that.

189
00:12:05,408 --> 00:12:08,301
We do kind of every technology that we can connect to.

190
00:12:08,301 --> 00:12:17,832
and serve the spark that could be sometimes complex because sometimes it's fully on-prem,
fully air-gap, fully privatized, no access to the internet at all and stuff like that.

191
00:12:17,832 --> 00:12:27,032
So you need to create a product that is adapted to this kind of large variety of
technologies that is covering and also to the privacy concerns and the privacy

192
00:12:27,032 --> 00:12:31,303
restrictions there because they are often like government-related, vital asset-related.

193
00:12:31,303 --> 00:12:35,865
So when you work for banks, energy companies, government-related assets, et cetera, et
cetera.

194
00:12:35,865 --> 00:12:36,756
So, $25.

195
00:12:36,756 --> 00:12:38,178
one is all of that.

196
00:12:38,178 --> 00:12:42,294
We serve our customers by pushing the most optimized solution for them.

197
00:12:42,294 --> 00:12:44,636
That is completely agnostic in terms of technology.

198
00:12:44,636 --> 00:12:47,290
That is completely agnostic also in terms of LLM usage.

199
00:12:47,290 --> 00:12:51,427
We can plug ourselves to kind of any LLM that is, uh let's say, agency credit.

200
00:12:51,427 --> 00:12:52,227
And...

201
00:12:52,279 --> 00:13:00,602
We deploy that in a few hours to our customers so they can automate everything that is
going down or that they need to maintain in their infrastructure.

202
00:13:00,602 --> 00:13:05,014
if you are like, don't know, BNP Paribas, you have a server going down at 4 a.m.

203
00:13:05,014 --> 00:13:10,626
in the morning, some customers will feel this impact because they don't know some part of
the application doesn't work, stuff like that.

204
00:13:10,626 --> 00:13:19,140
Instead of waiting one hour that someone would wake up and take the ticket in the ITSM
systems, it will be automatically done or at least automatically started and investigated

205
00:13:19,140 --> 00:13:19,721
by the AI.

206
00:13:19,721 --> 00:13:25,589
maybe you made like last point how how far is the LLM or the agent autonomy goes?

207
00:13:25,589 --> 00:13:29,876
Do they also rem like make fixes or do they just do the initial investigation?

208
00:13:29,876 --> 00:13:34,663
what to which extent gets the does the agent have autonomy to interact with the system?

209
00:13:34,663 --> 00:13:37,625
Yeah, so our philosophy is that agents should be autonomous.

210
00:13:37,625 --> 00:13:40,557
This is what we are kind of obsessed with that.

211
00:13:40,557 --> 00:13:43,278
And our design of products is quite unique because of that.

212
00:13:43,278 --> 00:13:45,940
And this is maybe what's pushing us a bit ahead of the curve.

213
00:13:45,940 --> 00:13:47,300
We don't build workflows.

214
00:13:47,300 --> 00:13:53,444
You will not see in our interface some point and click complex things where you need to do
if else do that.

215
00:13:53,444 --> 00:13:54,744
It's quite very simple.

216
00:13:54,744 --> 00:13:59,517
It's a lot of prompt and securities that we build in the agentic system, but then
everything is autonomous.

217
00:13:59,517 --> 00:14:01,017
You can put like restrictions.

218
00:14:01,017 --> 00:14:02,398
So of course we have a human industry.

219
00:14:02,398 --> 00:14:11,138
loop mode where the agent is restricted from doing action, but still all the commands are
generated by the agent and it's just an appreciation of command that will use, I don't

220
00:14:11,138 --> 00:14:15,249
know, LLM as a judge technology or stuff like that to ensure safety of the product.

221
00:14:15,249 --> 00:14:18,448
So yeah, we want the agent to be fully autonomous.

222
00:14:18,448 --> 00:14:27,520
That's a bit our goal and we want to share that with our customers because otherwise ROI
will be lower and also technology will be less kind of innovative.

223
00:14:27,520 --> 00:14:31,193
So we push forward this with our customers and this is quite important for us.

224
00:14:31,193 --> 00:14:35,910
But at the same time, we build 50 % of our work is about safety.

225
00:14:35,910 --> 00:14:46,660
50 % of what we do is human in the loop situations, how to be sure to inject the right
business rules, the right restriction rules, blacklist, how to measure performance of the

226
00:14:46,660 --> 00:14:48,051
agents, et cetera, et cetera, cetera.

227
00:14:48,051 --> 00:14:51,444
So it's not just about like the AI.

228
00:14:51,444 --> 00:14:52,061
Mm-hmm.

229
00:14:52,061 --> 00:15:02,201
products are just AI and there is nothing around it but it's 50 % or so of tool set that
we give around the product that is allowing the customer to feel safe to measure and to

230
00:15:02,201 --> 00:15:04,092
not do things that will go wrong.

231
00:15:04,092 --> 00:15:05,683
And it's an interesting approach.

232
00:15:05,683 --> 00:15:12,974
I perhaps like crack me if I'm wrong, like classify this a bit as AI as RE.

233
00:15:12,974 --> 00:15:22,636
so site reliability engineering, where there are a number of competitors, but I think a
lot of your competitors, their, their, their value proposition is more root cause

234
00:15:22,636 --> 00:15:23,527
analysis.

235
00:15:23,527 --> 00:15:25,738
It's like pointing at this is going wrong.

236
00:15:25,738 --> 00:15:29,876
but not necessarily this remediation phase that you're discussing as well, right?

237
00:15:29,876 --> 00:15:32,119
And I think what you're focusing on is the two.

238
00:15:32,191 --> 00:15:37,036
And this is why also we kind of are feeling a bit six months ahead of the curve.

239
00:15:37,036 --> 00:15:40,478
Or maybe we're just a crazy guy in turn, I don't know.

240
00:15:40,478 --> 00:15:43,739
Because yeah, from the beginning, well, no, we should not just observe.

241
00:15:43,739 --> 00:15:45,821
Of course, it's obvious that everyone can observe.

242
00:15:45,821 --> 00:15:49,242
But the value of agentic is about doing things.

243
00:15:49,242 --> 00:15:51,012
This is why we build this agentic.

244
00:15:51,012 --> 00:15:55,386
My job is to just watch the agent work and not touching my fucking keyboard.

245
00:15:55,386 --> 00:15:56,947
that's what it should do.

246
00:15:56,947 --> 00:16:03,283
So because we are onto that since the beginning, we have built kind of a framework that
works super well on that.

247
00:16:03,283 --> 00:16:04,726
Everything is in house built.

248
00:16:04,726 --> 00:16:09,462
And this choice is a bit risky because in sales, you will have like...

249
00:16:09,462 --> 00:16:15,682
more, let's say, it's taking more time to convince customers and to show that your product
is safe, et cetera, et cetera.

250
00:16:15,682 --> 00:16:18,373
So it could be like a bit more difficult to sell.

251
00:16:18,373 --> 00:16:21,673
But when you are in, the value proposition is insane.

252
00:16:22,573 --> 00:16:27,797
And the ROI, when we send ROI matrix to our customers or leads, it's like, it's obvious.

253
00:16:27,797 --> 00:16:31,279
This is more like, okay, in three years, course it will be the standard.

254
00:16:31,279 --> 00:16:36,030
Because you put so much value back to the customer that they have no choice in doing that.

255
00:16:36,030 --> 00:16:37,412
This is where it's interesting.

256
00:16:37,412 --> 00:16:39,524
It's very early days of agentic.

257
00:16:39,524 --> 00:16:41,156
Adoption is still ongoing.

258
00:16:41,156 --> 00:16:46,241
But at the same time, it's unanimous that everyone says it will be done like that.

259
00:16:46,770 --> 00:16:55,072
Maybe to double click a bit, like you I understand what I I think I can definitely agree
that it may be hard to convince people there's more scrutiny, right?

260
00:16:55,072 --> 00:16:57,764
When they are when they have like agency, basically.

261
00:16:57,764 --> 00:17:00,375
What are how do you how do you get people over the hump?

262
00:17:00,375 --> 00:17:07,468
Like what are the what are the things that you can show them to say like okay, it's not
gonna it's not gonna mess up your infrastructure, it's not gonna make a bigger mess out of

263
00:17:07,468 --> 00:17:08,489
the the incident.

264
00:17:08,489 --> 00:17:13,322
What are the things that you can that you can share that you can say to them to get people
to get in, basically?

265
00:17:13,322 --> 00:17:20,170
Yeah, so first, I'm not saying it's never gonna fuck up because one day I will have a very
bad call and it's gonna be a bad day.

266
00:17:20,170 --> 00:17:25,444
But it's the same with if you hire some engineers offshore or even near shore, even in
shore, it's the same.

267
00:17:25,444 --> 00:17:28,617
Someone's gonna fuck up someday and your infrastructure is gonna be down.

268
00:17:28,617 --> 00:17:29,488
I leave that.

269
00:17:29,488 --> 00:17:37,073
thousands of times in my life when I was working in big corporates or even when I was in
the startup, the bad call on Saturday night will happen.

270
00:17:37,073 --> 00:17:38,395
But it will happen maybe less.

271
00:17:38,395 --> 00:17:40,257
It will happen maybe in different stage.

272
00:17:40,257 --> 00:17:41,609
so we are like...

273
00:17:41,609 --> 00:17:44,640
saying that AI is not less intelligent than human.

274
00:17:44,640 --> 00:17:49,551
It's actually on this kind of task more precise often, depending on the model of course.

275
00:17:49,551 --> 00:17:52,062
We can debate on the model stuff and the model choice.

276
00:17:52,062 --> 00:17:59,104
But let's say we are convinced with our customers that in the end, a human being will make
more mistakes than an AI that is well done.

277
00:17:59,104 --> 00:18:02,055
And if we look at new AI technology like cloud.

278
00:18:02,055 --> 00:18:09,531
these new models and even smaller models like if we speak about like the latest Quen
models, Zai models, GML, it's like these guys are crazy.

279
00:18:09,531 --> 00:18:12,663
It's the quality of work that is done by the AI is very, very good.

280
00:18:12,663 --> 00:18:23,438
So it's quite an optimistic view of what's gonna happen but I'm pretty sure that on the
statistic level once everything is on production, we have like, I don't return of

281
00:18:23,438 --> 00:18:28,592
experience on thousands of hours of work, et cetera, the statistics would be in favor of
the AI.

282
00:18:28,592 --> 00:18:33,036
I'm 100 % convinced of it and almost of our customers say so as well.

283
00:18:33,036 --> 00:18:35,858
The second part is that you need to bring the tool set as I was saying.

284
00:18:35,858 --> 00:18:46,156
You need to bring not only the AI, but you need the tools to control the AI, to measure
the AI, to make sure that we are not having something that is wrong going on.

285
00:18:46,156 --> 00:18:56,573
so for that, for example, we built one of the first benchmark engine that is able not to
monitor only if the AI does the job as it succeeded the task or not,

286
00:18:56,573 --> 00:19:00,046
also measure the compliance of the tasks that have been done inside the job.

287
00:19:00,046 --> 00:19:11,335
So we measure every part of the job done against scenarios that the customer can customize
and say, okay, am I seeing this task and this part of the task and this part of the task?

288
00:19:11,335 --> 00:19:15,748
Is this forbidden things never happening because we don't want to see them, et cetera, et
cetera.

289
00:19:15,748 --> 00:19:25,156
So the agent is able to work in full autonomy and the customer can monitor them and say,
okay, this type of task that I want to see in production now is stable.

290
00:19:25,156 --> 00:19:28,287
is going well and we can continue to push it in production.

291
00:19:28,287 --> 00:19:35,788
And if there is an anomaly, you go to this qualification environment, to the sandbox
environment, and you can see the anomaly happening because I don't know, you change the

292
00:19:35,788 --> 00:19:39,620
model, you change some points here and there, and there is maybe a bad impact.

293
00:19:39,620 --> 00:19:46,041
As soon as you will see that, you can stop your agent in production, find the fix in the
sandbox, and then replicate.

294
00:19:46,041 --> 00:19:49,282
And so we bring the process of managing these things.

295
00:19:49,282 --> 00:19:50,392
It's still early days.

296
00:19:50,392 --> 00:19:51,733
We are building as we fly.

297
00:19:51,733 --> 00:19:53,313
But at the same time,

298
00:19:53,313 --> 00:19:54,625
It's the new way to work.

299
00:19:54,625 --> 00:19:58,490
You were speaking about these SREs or C-SOPs guys in the corporate world.

300
00:19:58,490 --> 00:19:59,730
Their job is going to change.

301
00:19:59,730 --> 00:20:01,973
They are kind of changing.

302
00:20:01,973 --> 00:20:08,591
Before they were piloting human work or they were doing themselves like production
engineering, for the run engineering.

303
00:20:08,591 --> 00:20:11,636
Now they like evolving to be AI pilots.

304
00:20:11,636 --> 00:20:12,737
And this is a new world.

305
00:20:12,737 --> 00:20:13,678
This is a new game.

306
00:20:13,678 --> 00:20:15,519
We are discussing with our customers every day.

307
00:20:15,519 --> 00:20:24,675
And it's so cool to see these guys that were like before just doing installation by hands,
map notes by hands, that use this knowledge that they have accumulated to guide AI and to

308
00:20:24,675 --> 00:20:31,980
build the best AI systems with us and that will do the job for like a fraction of the cost
and that's a fraction of the time necessary.

309
00:20:31,980 --> 00:20:33,187
So, yeah.

310
00:20:33,187 --> 00:20:33,657
Okay.

311
00:20:33,657 --> 00:20:37,931
So maybe maybe to to rephrase what you said, put in my words to make sure I'm following.

312
00:20:37,931 --> 00:20:45,980
Like you're saying, you're not promising that the the LEI will never make mistakes because
but just the same as people do, people have done for a long time.

313
00:20:45,980 --> 00:20:51,005
And you really double down on also the traceability to make sure you can check what
happened.

314
00:20:51,005 --> 00:20:56,028
Let's let's let's give the visibility and give that feedback loops to make sure that these
things don't happen again and all these different things.

315
00:20:56,028 --> 00:20:59,682
It's more auditable as well, it's more maybe even predictable, I guess, 'cause uh

316
00:20:59,682 --> 00:21:06,615
agents if especially if you have open models, it's uh it's easier to to achieve
reproducibility with people, maybe it's more blurry, right?

317
00:21:06,615 --> 00:21:08,969
Does that does that sound about right?

318
00:21:08,969 --> 00:21:09,861
Yeah, totally.

319
00:21:09,861 --> 00:21:13,584
We cannot lie saying it's perfect, it's never going to do something wrong.

320
00:21:13,584 --> 00:21:14,485
It's a lie.

321
00:21:14,485 --> 00:21:17,277
Don't do that and no one will trust you if you say that.

322
00:21:17,277 --> 00:21:27,176
But yeah, everything is in how to make sure that the technology will continue to evolve
and be more and more qualitative and now we are, think, in a kind of a murlot of model

323
00:21:27,176 --> 00:21:33,737
accuracy and we cannot even release models anymore without thinking they are too strong
now with the mythos.

324
00:21:33,737 --> 00:21:35,088
events are arriving.

325
00:21:35,088 --> 00:21:48,349
So I think the model technology is kind of still evolving quite fast, even if it's maybe
not anymore the exponential curve that we've seen three years ago, but it's still very,

326
00:21:48,349 --> 00:21:49,410
very impressive.

327
00:21:49,410 --> 00:21:54,015
And there is a new generation of models arriving, a lot of research in world models, et
cetera, et cetera.

328
00:21:54,015 --> 00:21:57,288
So it's going to be better and better every day we speak.

329
00:21:57,288 --> 00:22:02,830
But yeah, the need of having products is that a lot of people think that AI is just AI,
end of the story.

330
00:22:02,830 --> 00:22:12,455
And you can see a lot of companies that are just having PhD guys doing incredible job on
models, but then on the application side, there is something missing because you need to a

331
00:22:12,455 --> 00:22:15,556
good harness, you need to build a good tool set, et cetera, et cetera.

332
00:22:16,093 --> 00:22:16,357
So.

333
00:22:16,357 --> 00:22:18,040
you, you, touched a bit on sales.

334
00:22:18,040 --> 00:22:28,569
like, uh, it's, and I can imagine that you're a bit of this dichotomy because you're like,
assume that your typical company that you sell to, like, they are very focused on having

335
00:22:28,569 --> 00:22:30,340
robustness, having reliability.

336
00:22:30,340 --> 00:22:36,868
And because AI agents and especially autonomous agents are still so new, like the Chris,
this question, like, will this not

337
00:22:36,868 --> 00:22:44,225
take our reliability down, even though your value proposition is actually to improve it or
detect issues more quickly or resolve them more quickly, right?

338
00:22:44,225 --> 00:22:53,637
Like you have this, you're in this time period where we're still gathering data on to
prove, to basically make this argument it's definitely better than how we used to do it.

339
00:22:53,637 --> 00:23:01,233
Yeah, so on our side, is a way that we have, this is something that we have implemented
directly in the release process with our customer.

340
00:23:01,233 --> 00:23:07,009
So instead of saying, we are going to show you proofs, we build the proof together with
the customer.

341
00:23:07,009 --> 00:23:15,855
So when we onboard the customer, the customer is entering what we call a sandbox phase on
which we work with the customer, like next to them, they have access to the tool, they

342
00:23:15,855 --> 00:23:16,906
build things with us.

343
00:23:16,906 --> 00:23:19,797
And this is important because this is when they see that by

344
00:23:19,797 --> 00:23:20,989
to prove that it's working.

345
00:23:20,989 --> 00:23:22,390
And they continue the product themselves.

346
00:23:22,390 --> 00:23:24,532
We train them, we help them, et cetera.

347
00:23:24,532 --> 00:23:26,653
But in the end, they become this AI operator.

348
00:23:26,653 --> 00:23:33,669
So this is very interesting because as I was saying, you see a lot of top managers guys
that are like reading Gartner's or Forrester every day.

349
00:23:33,669 --> 00:23:37,601
And these guys will, of course, be convinced that in the next few years, it's going to be
the trend.

350
00:23:37,601 --> 00:23:40,023
It's going to be the default solution.

351
00:23:40,023 --> 00:23:41,684
And everyone agrees on that.

352
00:23:41,684 --> 00:23:47,702
But then you need to make sure also the entire teams, especially large corporate, agrees
with that vision.

353
00:23:47,702 --> 00:23:51,381
And if you lose the acceptance of the vision under the...

354
00:23:51,381 --> 00:23:54,392
the big C level people, it's a bit a mess also to implement.

355
00:23:54,392 --> 00:23:57,472
You will have adoption issues and we've seen that also in the company.

356
00:23:57,472 --> 00:24:05,343
We've seen in some customers real adoption issues because it's a change management for
them that is quite critical to happen, especially when you are an IT company.

357
00:24:05,343 --> 00:24:08,805
Your job is completely upside down since a few months, now.

358
00:24:08,805 --> 00:24:18,559
So the only way is not to have the best product on them but also to onboard them, part of
your team kind of, and use the product as quickly as possible to big

359
00:24:18,559 --> 00:24:23,114
convince themselves that what they do is providing value and this is going to be the
default for them.

360
00:24:23,114 --> 00:24:29,465
Yeah, and I think that's what you're also doing is you're uniquely positioned in the sense
that the...

361
00:24:29,465 --> 00:24:36,196
the skill that you're productionizing, like you're at the frontier of what is possible
with autonomous agents.

362
00:24:36,196 --> 00:24:45,096
And at the same time, you have a strong expertise in how these large corporates, how their
infrastructure looks like, which is typically not the most modern cloud setup, right?

363
00:24:45,096 --> 00:24:48,867
Like they have a lot on-premise, they have a lot of old school hardware that they're
running on.

364
00:24:48,867 --> 00:24:56,098
And I think that combination of skills and combining them in a product is quite a unique
position,

365
00:24:56,098 --> 00:25:00,843
It's even harder to find just people, individual people that know about these two things.

366
00:25:00,843 --> 00:25:03,414
So yeah, that was actually a fun thing.

367
00:25:03,414 --> 00:25:12,496
yeah, you would see on the competition mapping that this enterprise, AA agent equal is
quite still a blue ocean, which is surprising.

368
00:25:12,496 --> 00:25:15,416
Everyone is SaaS cloud based.

369
00:25:16,076 --> 00:25:17,667
This part is overcrowded.

370
00:25:17,667 --> 00:25:21,598
On this you have like half of white communities two years ago was there.

371
00:25:21,598 --> 00:25:25,809
You still have like a lot of companies that raise actually a lot of money because there is
still a huge market there.

372
00:25:25,809 --> 00:25:30,728
Still have all these retail brands that are completely cloud-based, AWS still.

373
00:25:30,728 --> 00:25:34,779
These guys are doing an extremely good job at doing this on this perimeter.

374
00:25:34,779 --> 00:25:43,310
But on our side, it's very interesting because you are on this whole kind of world that is
still the default for the majority of the companies.

375
00:25:43,310 --> 00:25:53,370
I was super surprised when we are in New York often we speak with guys and a lot of people
were saying, okay, how much on-prem is JP Morgan Chase?

376
00:25:53,501 --> 00:25:55,341
60, 70%.

377
00:25:55,341 --> 00:25:57,892
Goldman Sachs, 90 % on-prem.

378
00:25:57,892 --> 00:25:58,523
All of them.

379
00:25:58,523 --> 00:25:59,283
All of them.

380
00:25:59,283 --> 00:26:01,585
I was in the Office of Societe Generale a few weeks ago.

381
00:26:01,585 --> 00:26:03,206
It's still majority on-prem as well.

382
00:26:03,206 --> 00:26:15,888
oh And this is what no one has understood is that there is still a large spectrum of IT
infrastructure that is not based on the cloud, that will never move to the cloud.

383
00:26:15,888 --> 00:26:17,671
It's not designed to be in the cloud.

384
00:26:17,671 --> 00:26:20,185
How do you move ABM, AS400?

385
00:26:20,185 --> 00:26:22,566
machines, yeah, no, it's not possible.

386
00:26:22,566 --> 00:26:23,246
It's not possible.

387
00:26:23,246 --> 00:26:25,918
And it's not the work of AWS, of GCP, of Azure to do that.

388
00:26:25,918 --> 00:26:27,699
They are not able to host these things.

389
00:26:27,699 --> 00:26:29,441
So how do you serve this market?

390
00:26:29,441 --> 00:26:31,433
And this market is well underserved.

391
00:26:31,433 --> 00:26:39,218
It's still like all these actors that are still here forever, SAP, IBM, Fortinet, yeah,
ServiceNow.

392
00:26:39,218 --> 00:26:40,222
These guys are here.

393
00:26:40,222 --> 00:26:41,219
They are not moving.

394
00:26:41,219 --> 00:26:44,281
And this is like a huge chunk of the IT services in the world.

395
00:26:44,281 --> 00:26:52,740
So yeah, when you are speaking about the unique skill set and the unique understanding,
this is quite true because we were lucky to be surrounded by people from this world from

396
00:26:52,740 --> 00:26:58,204
time ago and I was at the Bayer site a few years ago, so I could have understood both
sides of the game.

397
00:26:58,204 --> 00:27:00,487
But I was so shocked that it's still the case.

398
00:27:00,487 --> 00:27:04,914
The on-prem majority component is still like a huge thing everywhere.

399
00:27:04,914 --> 00:27:09,991
And no one is like kind of interested in that part because it's super painful to interact
with.

400
00:27:09,991 --> 00:27:12,265
Connectivity is not super easy.

401
00:27:12,265 --> 00:27:14,369
Sales cycles are super long.

402
00:27:14,369 --> 00:27:19,976
We are speaking about like six to 12 months in sales to sign a deal.

403
00:27:19,976 --> 00:27:25,275
This is why we kind of raised our second round because we needed to be strong enough to
close our deals.

404
00:27:25,275 --> 00:27:29,638
So yeah, and you need to speak to some people that have kind of a unique culture as well.

405
00:27:29,638 --> 00:27:35,143
When you speak to these IT services persons, it's not like the usual startup client that
you're gonna have.

406
00:27:35,143 --> 00:27:37,784
These guys are super, super critical.

407
00:27:37,784 --> 00:27:39,425
They know their job super well.

408
00:27:39,425 --> 00:27:40,396
They are here forever.

409
00:27:40,396 --> 00:27:42,852
You speak to guys that are here since.

410
00:27:42,852 --> 00:27:45,136
30 years more sometime, they have seen everything.

411
00:27:45,136 --> 00:27:48,232
They have seen like the dot-com revolution.

412
00:27:48,232 --> 00:27:50,195
They have seen like the mobile revolution.

413
00:27:50,195 --> 00:27:53,700
They have seen everything and they're like still here doing data center stuff.

414
00:27:53,902 --> 00:27:55,752
okay, so it's very.

415
00:27:55,752 --> 00:27:58,842
all the time, the mainframe just stayed the same mainframe.

416
00:27:59,880 --> 00:28:08,529
These guys are like, yeah, I've been here when we did the first automation for Sokira's
code with Terraform and I'm still here doing now the agent-y AIOps transition.

417
00:28:08,529 --> 00:28:19,390
ah And so that's very cool because you speak to some people that have like a large
understanding of what is tech, but not shiny tech always, but very useful tech for the

418
00:28:19,390 --> 00:28:19,811
world.

419
00:28:19,811 --> 00:28:20,842
Maybe I have a question also.

420
00:28:20,842 --> 00:28:26,213
Like uh I imagine that for for LLMs it's much, much harder to to debug the scenarios.

421
00:28:26,213 --> 00:28:31,781
And I mean, 'cause just thinking of documentation, like the the the the data that it was
trained on, right?

422
00:28:31,781 --> 00:28:33,482
I'm sure there's it's very disproportionate.

423
00:28:33,482 --> 00:28:36,245
There's way more talking about AWS, G C P, all these things.

424
00:28:36,245 --> 00:28:39,247
Have you do you have any comparisons, views or anything?

425
00:28:39,247 --> 00:28:42,844
Like, yeah, like like I don't know, um let's say minimax.

426
00:28:42,844 --> 00:28:47,961
without much fine-tuning all these things does very well on debugging AWS and on-prem we
see that it drops a lot.

427
00:28:47,961 --> 00:28:50,706
Do you have any any more numbers, any more details on that?

428
00:28:50,706 --> 00:28:54,954
Yeah, I have a good example on this, which is, so.

429
00:28:54,954 --> 00:29:05,391
For the last, let's say, 12 months, and especially until very recently, our best player of
recommendation was Quen, because Quen models are small, light, efficient.

430
00:29:05,391 --> 00:29:11,344
They have kind of the right balance between size, accuracy, cost, running on custom GPUs.

431
00:29:11,344 --> 00:29:13,645
So we were all in Quen.

432
00:29:13,645 --> 00:29:14,726
We are completely agnostic.

433
00:29:14,726 --> 00:29:18,468
We can work on any model, but this was our top one favorite model to push on.

434
00:29:18,468 --> 00:29:22,360
So our customers were like, which model we should choose if we have the choice?

435
00:29:22,360 --> 00:29:24,521
We said, start with Quen.

436
00:29:24,601 --> 00:29:26,802
Up to 135B, 80B, next.

437
00:29:26,802 --> 00:29:29,084
Now we have new ones arriving since a few months.

438
00:29:29,084 --> 00:29:31,025
So these guys are the best.

439
00:29:31,025 --> 00:29:37,340
But when we tested something, which was like managing Windows server, we noticed that Quen
was not that good.

440
00:29:37,340 --> 00:29:41,373
It was hallucinating a lot in, I don't know, versioning of PowerShell and stuff like that.

441
00:29:41,373 --> 00:29:53,430
And we tested some alternatives and we found that GPT-OSS, the open source, open weight
LLM of OpenAI was performing well better on PowerShell commands, Windows Server commands

442
00:29:53,430 --> 00:29:55,301
than Quinn for our use case.

443
00:29:55,301 --> 00:29:57,943
Maybe it's not absolutely, but our use case was the case.

444
00:29:57,943 --> 00:29:59,574
And even if it was a smaller model.

445
00:29:59,574 --> 00:30:05,858
So we were like, okay, maybe there is like some interesting conclusion to have that maybe,
I don't know.

446
00:30:05,858 --> 00:30:08,299
It was from that gut guessing because we are not in this

447
00:30:08,299 --> 00:30:19,461
companies doing the training with them but potentially GPT and OpenAI being closer to
Microsoft are having more data set on Windows than the rest of the world.

448
00:30:19,962 --> 00:30:30,668
Maybe Chinese companies that are training mostly on public resources or on own inside
resources have less training on Windows server documentation and so that was some

449
00:30:30,668 --> 00:30:32,800
assumption we're having and now we see that

450
00:30:32,800 --> 00:30:35,961
GPT models are good for some certain type of task, et cetera.

451
00:30:35,961 --> 00:30:38,863
I don't know if you heard about the skateboard benchmark.

452
00:30:38,863 --> 00:30:43,005
um It's quite a niche thing online.

453
00:30:43,005 --> 00:30:44,445
You should search for it.

454
00:30:44,445 --> 00:30:46,196
So there is something named the skateboard benchmark.

455
00:30:46,196 --> 00:30:53,748
And if you compare a Chinese model to an American model, the American model will know all
the tricks of skateboard, where the Chinese doesn't really know it.

456
00:30:54,629 --> 00:30:55,268
Yeah.

457
00:30:55,268 --> 00:30:57,888
because the training corpus is not the same.

458
00:30:58,052 --> 00:31:06,512
And that's cool because I think the English models and the American models especially are
like training on everything that's online about skateboarding and stuff where the Chinese

459
00:31:06,512 --> 00:31:10,372
maybe social media is not that much on that part.

460
00:31:11,492 --> 00:31:12,412
Exactly.

461
00:31:12,412 --> 00:31:15,472
And what we've seen is kind of equivalent of this skateboard benchmark.

462
00:31:15,472 --> 00:31:18,601
We've seen that Chinese model will perform maybe less on Microsoft products.

463
00:31:18,601 --> 00:31:19,404
Mm-hmm.

464
00:31:19,404 --> 00:31:27,016
But aside of it, if you go to the extreme position of this, as long as the model has,
let's say, public corpus to train on...

465
00:31:27,016 --> 00:31:34,342
it's quite okay to work on and you can work without that much fine tuning, just like a
good prompting engineering that we do inside the products is enough.

466
00:31:34,342 --> 00:31:43,853
But when we see that some products is exotic, and for example, there is a very famous
firewall in Europe named Storm Shield that is used mostly by government in France and in

467
00:31:43,853 --> 00:31:44,324
Europe.

468
00:31:44,324 --> 00:31:46,707
This one has almost no documents online.

469
00:31:46,707 --> 00:31:49,540
And engineers working on these products are super niche.

470
00:31:49,540 --> 00:31:54,836
There is only like a 600 page product on their website that is uh the documentation of the
product.

471
00:31:54,836 --> 00:31:57,549
And so on this, you can try every model you want.

472
00:31:57,549 --> 00:31:59,652
Even Claude cannot work on that.

473
00:31:59,652 --> 00:32:03,716
So on this, we start to feel that, okay, we will need some fine tuning.

474
00:32:03,716 --> 00:32:06,750
We need to put more adjustment, reinforcement.

475
00:32:06,750 --> 00:32:07,783
That sense.

476
00:32:07,874 --> 00:32:10,176
Maybe one last question about model comparison.

477
00:32:10,176 --> 00:32:11,448
You mentioned Claude as well.

478
00:32:11,448 --> 00:32:17,385
How from your view, because I imagine you have a very good view on this, what is the gap
between open and closed models?

479
00:32:17,385 --> 00:32:22,171
Like I'm thinking uh GDPT and and Claude versus Chinese models.

480
00:32:22,171 --> 00:32:23,594
given demo accuracy.

481
00:32:23,594 --> 00:32:24,094
Yeah, yeah.

482
00:32:24,094 --> 00:32:31,030
So for example, I don't know, uh you have this client and then you say, Okay, you have
this type of problems and you say, Okay, Cloud I'm pretty sure is gonna do well, but maybe

483
00:32:31,030 --> 00:32:32,190
the open models not sure.

484
00:32:32,190 --> 00:32:34,382
I imagine the open models are still behind, right?

485
00:32:34,382 --> 00:32:36,663
But like is the gap still big?

486
00:32:36,663 --> 00:32:43,215
Is it is it is it noticeable or like you think it's it's the we're closing the gap, like
open source and open models.

487
00:32:43,215 --> 00:32:47,789
So I would say the gap gets thinner and thinner, but still super visible.

488
00:32:47,789 --> 00:32:57,131
You still have these flagship models, foundations, rock stars that are above the crowd
because their model size is like, we don't know them.

489
00:32:57,785 --> 00:33:02,436
And they have this kind of weird architectures that are super, super complex to make this
happen.

490
00:33:02,436 --> 00:33:08,478
When you speak to Opus, when you speak to Fable, these kind of models, it's not
comparable.

491
00:33:08,478 --> 00:33:14,581
You're not speaking to a 200, 400 billion parameters model that can sit on a few GPUs in
the data.

492
00:33:14,581 --> 00:33:24,925
center you're speaking to some stuff I think people don't realize the level of complexity
that this flagship model has just running on something compared to what is done on the

493
00:33:24,925 --> 00:33:35,979
small open source open wide models and this is where it's impressive because if you you
compare I don't know the Concorde plane with the small touring plane that your grandfather

494
00:33:35,979 --> 00:33:37,730
is driving on the weekend you know

495
00:33:38,308 --> 00:33:38,782
Yeah.

496
00:33:38,782 --> 00:33:47,367
It's still flying, both of them, but there is one that is doing Mach 3, 3 hours to New
York, and the other one can just fly you 200 kilometers to the lake.

497
00:33:48,928 --> 00:33:50,249
I think this is the kind of thing.

498
00:33:50,249 --> 00:33:51,090
It's still a plane.

499
00:33:51,090 --> 00:33:53,222
It has an engine and two wings.

500
00:33:53,222 --> 00:33:57,075
But we are not speaking about the same level of complexity at all.

501
00:33:57,102 --> 00:34:02,777
But this is where I think it's super impressive to see these kind of models going shrinker
and shrinker.

502
00:34:02,777 --> 00:34:06,013
And there is now two different ways to do it.

503
00:34:06,013 --> 00:34:06,646
That is...

504
00:34:06,646 --> 00:34:08,228
Do you need a flagship usage?

505
00:34:08,228 --> 00:34:12,600
And on this case, you pay per token, you pay per intelligence, and this is a use case you
want on some case.

506
00:34:12,600 --> 00:34:14,711
For example, code development.

507
00:34:15,012 --> 00:34:18,953
All the guys now doing coding are using Cloud by default almost.

508
00:34:18,953 --> 00:34:20,414
And you don't want to pay less.

509
00:34:20,414 --> 00:34:21,775
You're very happy to use Cloud.

510
00:34:21,775 --> 00:34:23,916
It's super intelligent, does the job well.

511
00:34:23,916 --> 00:34:28,268
When you use your Open Cloud virtual assistant or digital twin, same.

512
00:34:28,268 --> 00:34:31,610
You want to use Cloud because it does the job super well end end.

513
00:34:31,610 --> 00:34:35,144
When you are in more guided space and when you're looking for efficient

514
00:34:35,144 --> 00:34:41,820
when you're looking for cost performance, when you're looking for privacy, which is a
huge, huge topic also.

515
00:34:41,820 --> 00:34:45,644
Main reason why we don't use Cloud is privacy first and then cost after.

516
00:34:45,644 --> 00:34:50,895
And this is where you can start to build on open-weight open-source model that are
smaller.

517
00:34:50,895 --> 00:34:53,800
On this game, people are trying to make the most compact models.

518
00:34:53,800 --> 00:34:55,613
for me, it's two different games.

519
00:34:55,613 --> 00:34:57,866
It's not anymore competing together.

520
00:34:57,866 --> 00:34:59,510
I don't think it will ever compete anymore.

521
00:34:59,510 --> 00:35:00,816
Interesting.

522
00:35:01,772 --> 00:35:12,280
I can imagine that you're maybe you can shed a bit of light on that like your typical
customer like do they prefer open weight models that they can host locally on their

523
00:35:12,280 --> 00:35:12,882
premises?

524
00:35:12,882 --> 00:35:15,195
do you see a preference at these large corporates?

525
00:35:15,195 --> 00:35:25,778
Okay, so there is first rule that is universal nothing goes outside the corporate world We
don't use any public LLM for none of our customers We do it in demo we do it in some test,

526
00:35:25,778 --> 00:35:32,919
but that's it at the moment We are installing the corporate we use either their inference
partnership with Amazon Bedrock

527
00:35:32,919 --> 00:35:39,386
because they have a private contract with them and there is like a deal or we use on-prem
GPUs on the customer side.

528
00:35:39,687 --> 00:35:41,029
So that's the first rule.

529
00:35:41,029 --> 00:35:44,694
Nothing goes outside because by design it's not secure for their data.

530
00:35:44,694 --> 00:35:48,238
And if you want to stay like socked to and stuff, you need to respect that.

531
00:35:48,238 --> 00:35:51,208
And especially if you work like in, don't know.

532
00:35:51,208 --> 00:35:57,044
health situation, defense, corporate related or critical assets and all of them are almost
critical assets.

533
00:35:57,044 --> 00:35:58,325
this is mandatory.

534
00:35:58,325 --> 00:36:06,774
And then you see the planning now that we have with majority of our customers that year
one, we use an inference partner because it's easier to deploy.

535
00:36:06,774 --> 00:36:07,525
don't lose time.

536
00:36:07,525 --> 00:36:14,102
Year two, we move from OPEX to Capex and we move like investing in GPU.

537
00:36:14,102 --> 00:36:17,690
So in the end, the vision is to go into open-source models.

538
00:36:17,690 --> 00:36:21,818
And also what we do when we use the inference partner is to use this type of model as
well.

539
00:36:21,818 --> 00:36:23,860
We don't use Claude, almost never.

540
00:36:23,860 --> 00:36:24,545
Yeah.

541
00:36:24,545 --> 00:36:29,718
use Bedrock, we will use a bit of mix of everything, GPT-OSS, et cetera.

542
00:36:29,718 --> 00:36:31,220
On Azure, we use Quenolot.

543
00:36:31,220 --> 00:36:35,284
So it's more like we start paying the token on this market that we target.

544
00:36:35,284 --> 00:36:39,907
It allows us to show to the customer that it's working well, that it's not that difficult.

545
00:36:39,907 --> 00:36:42,660
We don't need to invest day one in GPU, et cetera.

546
00:36:42,660 --> 00:36:46,008
And then we move into more long-term situation from here too.

547
00:36:46,008 --> 00:36:53,073
so you're also saying that long term for, you have this also this vision to build your own
data center to host your models.

548
00:36:53,073 --> 00:36:53,987
How do you see this?

549
00:36:53,987 --> 00:37:01,975
So yes and no, because it will not serve the customer needs that much, because we will
become a liability for them.

550
00:37:01,975 --> 00:37:02,948
Yeah, yeah, see what you mean.

551
00:37:02,948 --> 00:37:03,506
oh

552
00:37:03,506 --> 00:37:12,937
No, our vision is more that customer environments are quite complex and sometimes even if
you are in a big company that has announced a partnership with Mistral, these business

553
00:37:12,937 --> 00:37:14,880
units are very split and separated.

554
00:37:14,880 --> 00:37:18,920
So if you work with the business case, they have already Mistral, GPU and stuff.

555
00:37:18,920 --> 00:37:23,771
But if you work, for example, with the IT infrastructure, this guy maybe doesn't have a
contract with Mistral as yet.

556
00:37:23,771 --> 00:37:30,162
And so sometimes we kind of fail in this situation and we try to fill the gap.

557
00:37:30,162 --> 00:37:30,868
Yeah, okay.

558
00:37:30,868 --> 00:37:35,068
So our job is to provide first the agentic solution and to make it work well.

559
00:37:35,068 --> 00:37:36,419
This is 90 % of our job.

560
00:37:36,419 --> 00:37:44,530
But we have also this capacity to bring hardware if necessary and to deliver to the
customer the capacity in GPU to host our models that we need.

561
00:37:44,530 --> 00:37:45,571
Interesting.

562
00:37:45,571 --> 00:37:58,043
And maybe to briefly touch upon the open weight models of today, because like you and I
think a lot of people are leaning on those very heavily, especially for these repetitive

563
00:37:58,043 --> 00:38:01,126
tasks, guided tasks, where they are very cost efficient basically.

564
00:38:01,126 --> 00:38:02,548
How do you see the future of these things?

565
00:38:02,548 --> 00:38:08,814
When I hear this, always a bit worried in the sense that it costs a lot of capital to
train these open weight models and they're...

566
00:38:08,814 --> 00:38:19,056
more like most of them are run by for-profit companies that release this today for
marketing purposes because we don't trust DeepSeek for whatever reason to BMP will not

567
00:38:19,056 --> 00:38:21,808
open a contract with DeepSeek tomorrow, but they might with Mistel.

568
00:38:21,808 --> 00:38:24,081
So there is some marketing that needs to be done.

569
00:38:24,081 --> 00:38:29,666
Maybe there are geopolitical reasons to be competitive towards the States or something.

570
00:38:29,666 --> 00:38:33,210
But like there's, I don't see a very long-term future for these.

571
00:38:33,210 --> 00:38:39,646
these models that require so much capital to stay quote unquote open source.

572
00:38:39,646 --> 00:38:44,748
So you're right, there is like something still to dig on that part on how the...

573
00:38:44,748 --> 00:38:48,350
small, middle-sized model business will continue to exist?

574
00:38:48,350 --> 00:38:51,722
Is it something that will continue to stay sponsored for the sake of it?

575
00:38:51,722 --> 00:38:58,907
You have also a lot of sponsorship happening when NVIDIA sponsors a lot of people with
free GPU usage, etc.

576
00:38:58,907 --> 00:39:03,256
It's actually marketing that is done also for these inference partners.

577
00:39:03,256 --> 00:39:10,470
If you're like Base 10, for example, you want to sponsor some guys on a game face to
deploy the models.

578
00:39:10,470 --> 00:39:13,422
Because if one of them is going to be a hit,

579
00:39:13,422 --> 00:39:16,846
it's going to be good marketing for you because you're going to train these models as
well.

580
00:39:16,846 --> 00:39:20,227
So I would say that there is always an interest somewhere to do it.

581
00:39:20,227 --> 00:39:25,482
ah And I would say that it's how good research should be done.

582
00:39:25,482 --> 00:39:26,434
It should be independent.

583
00:39:26,434 --> 00:39:28,186
Look at what has been done with DeepSeq.

584
00:39:28,186 --> 00:39:32,711
DeepSeq is a very good example of a model that has been a bit oh cheating.

585
00:39:32,711 --> 00:39:33,983
But at the same time,

586
00:39:33,983 --> 00:39:35,054
of a distraction of the ghost.

587
00:39:35,054 --> 00:39:39,747
Yeah, it was a proof of concept that distillation is sometimes better than training.

588
00:39:40,309 --> 00:39:45,505
And that's actually something super interesting because ultimately it was not the right
way to do it.

589
00:39:45,505 --> 00:39:54,358
But in the end, I don't know what they did is not, they showed that we should be able to
train models more efficiently without kind of spending that much money.

590
00:39:54,358 --> 00:39:55,619
And so now,

591
00:39:55,690 --> 00:40:02,567
People don't do that this way, they did it with Anthropic, but they tried to find a way to
replicate this philosophy of distillation.

592
00:40:02,567 --> 00:40:03,579
So I would say.

593
00:40:03,579 --> 00:40:04,379
It's a balance.

594
00:40:04,379 --> 00:40:06,239
We should have people innovating.

595
00:40:06,239 --> 00:40:07,279
We will have newcomers.

596
00:40:07,279 --> 00:40:09,799
We will have guys stopping to do it.

597
00:40:09,799 --> 00:40:11,579
And probably maybe at some point you're right.

598
00:40:11,579 --> 00:40:20,612
The new wave will be Mistral, OpenAI, Anthropic, just releasing small models that are not
very time expensive to run.

599
00:40:20,612 --> 00:40:24,663
And this is just part of their product lineup.

600
00:40:24,663 --> 00:40:27,723
That would be maybe the new standard, but I don't see that for the moment opening.

601
00:40:27,723 --> 00:40:35,612
If you look at the only one that has succeeded to do that is GPT OSS with OpenAI that is
how to say super small market share still.

602
00:40:35,612 --> 00:40:38,512
If it was not Windows Server, we would not use it on our side.

603
00:40:38,986 --> 00:40:51,438
Mistral is, I think, trying to find their business model because a lot of the revenue of
Mistral is services and engineering model tokenization.

604
00:40:51,438 --> 00:40:55,462
Let's say selling API is not the majority of the revenue of this file.

605
00:40:55,462 --> 00:40:57,153
So yeah, let's see how it goes.

606
00:40:57,153 --> 00:40:59,616
It's still, think, building as we fly.

607
00:41:00,176 --> 00:41:01,749
Sure, yeah, there as well, yeah.

608
00:41:01,749 --> 00:41:03,260
Talking building as we fly.

609
00:41:03,260 --> 00:41:09,891
Can you give us a bit of a view on, say, where do you see 2501 going in coming two years?

610
00:41:09,891 --> 00:41:11,642
What are your big focuses?

611
00:41:11,642 --> 00:41:12,673
That's a good question.

612
00:41:12,673 --> 00:41:14,724
Now we see things a bit everywhere.

613
00:41:14,724 --> 00:41:15,884
when we...

614
00:41:15,884 --> 00:41:20,075
So our value proposition and the core thing we're doing right now is remediation.

615
00:41:20,075 --> 00:41:24,235
It is to do maintenance and fix of incidents automatically.

616
00:41:24,235 --> 00:41:27,511
So we focus 100 % almost of our time on this.

617
00:41:27,511 --> 00:41:31,944
our job is to do remediations to the incidents and maintenance request management.

618
00:41:31,944 --> 00:41:34,446
Observe, remediate, report.

619
00:41:34,446 --> 00:41:36,699
This is what we want to do in a nutshell.

620
00:41:36,699 --> 00:41:40,660
and we try to be the best at doing this because there is a lot of ROI for the customer.

621
00:41:40,660 --> 00:41:44,411
Then you can extend in a lot of different spaces.

622
00:41:44,411 --> 00:41:53,082
There is naturally the observability world that is super, super linked to us on which we
have some part of the product that start to compete with some people there.

623
00:41:53,082 --> 00:42:00,582
For example, we start to have a releasing quite soon CMDB auto exploration, which is like
a huge pain point for our customers.

624
00:42:00,582 --> 00:42:04,402
And there is no natural lineup of product that is solving this issue.

625
00:42:04,402 --> 00:42:08,873
So which we know that we're going to like go a bit upper layer on observability.

626
00:42:08,873 --> 00:42:12,757
if we're gonna go full scale there, or we're just gonna just touch it a bit.

627
00:42:12,757 --> 00:42:19,574
But there is a lot of values there, because you have companies like Dynatrace, have
companies like Big Panda, Datadog, all these guys are here.

628
00:42:19,574 --> 00:42:25,374
So we don't want to compete with them, but there is like maybe a sub layer of their parts
that is interesting maybe to absorb.

629
00:42:25,374 --> 00:42:27,885
You have like the CyberSec world.

630
00:42:27,885 --> 00:42:39,071
Enormous and with these new models arriving the cybersec is something super super super
big to consider budget in cyber security is increasing like I don't know we never seen

631
00:42:39,071 --> 00:42:50,428
that the new mythos release has put in the highlights some big problems and now cyber
security team have a problem that there is Every day too many threats arriving that they

632
00:42:50,428 --> 00:42:55,307
don't have like any more capacity to follow up and patch everything oh

633
00:42:55,307 --> 00:43:02,937
It's gonna be something interesting and our technology is perfect to do that as well, like
cybersecurity audits, cybersecurity compliance check.

634
00:43:02,937 --> 00:43:13,963
So see sock sock remediation everything that you try to do on cyber so We don't go there
yet because we feel we don't have the credentials to do it It's not our focus and we

635
00:43:13,963 --> 00:43:21,977
should show some battle at some point So we want to stay on the remediation of IT
infrastructure, but we are for example some customers already I was discussing with one

636
00:43:21,977 --> 00:43:24,108
yesterday that are like okay inside the use cases.

637
00:43:24,108 --> 00:43:30,851
I'm gonna use two five zero one I'm gonna put some vulnerability vulnerability tests we
feel that cyber is gonna be a big topic for us as well in

638
00:43:30,851 --> 00:43:41,263
coming years and and on the below I was saying before the hardware also there is a lot of
difficulties around this world how to build the hardware that is in capacity to host your

639
00:43:41,263 --> 00:43:46,782
models for the usage of the models what is the hardware looking like there is today a lot
of

640
00:43:46,782 --> 00:43:51,064
I don't know, under-optimized infrastructure on AI usage.

641
00:43:51,064 --> 00:43:56,515
You will buy too many GPUs, you will not use that capacity, you will not optimize your
model correctly.

642
00:43:56,515 --> 00:43:57,695
So there is also a lot of stuff.

643
00:43:57,695 --> 00:44:01,495
On this, it's a very small part of our job because we don't do it for every customer.

644
00:44:01,495 --> 00:44:05,015
We do it for very few customers that are using it.

645
00:44:05,015 --> 00:44:06,835
But we feel that there is a small need there.

646
00:44:06,835 --> 00:44:11,746
But I would say our big topic is observability and cyber next to what we're doing right
now.

647
00:44:11,746 --> 00:44:23,787
But it makes me think, because again, you have a very interesting position there where you
can traverse a bit into different fields because you start from this, you're at the

648
00:44:23,787 --> 00:44:24,999
infrastructure level.

649
00:44:24,999 --> 00:44:29,092
You're doing observance there, you're doing root cause analysis.

650
00:44:29,092 --> 00:44:34,327
It's not a big step to say from there we're also going to do more in the cyber sec, we're
going to do firewall monitoring.

651
00:44:34,327 --> 00:44:36,868
It's not a big step to make.

652
00:44:36,868 --> 00:44:45,856
on the other side, if you would start with a firewall anomaly detection and then say, I'm
going to go the other way, I'm going to do now, do infrastructure failure detection,

653
00:44:45,856 --> 00:44:48,258
that's a way bigger leap.

654
00:44:48,258 --> 00:44:52,731
It's an interesting point to start from and to further build up.

655
00:44:52,790 --> 00:44:53,401
Exactly.

656
00:44:53,401 --> 00:45:00,125
It's way more natural to start from infrastructure and then to go on these things that are
observability, cyber and audit.

657
00:45:00,125 --> 00:45:04,809
Because doing the opposite is more like you need to go back to the central core problem
that is infrastructure itself.

658
00:45:04,809 --> 00:45:13,915
And in terms of sales, it's better as well because you're already working with the key
decision makers that are like the CIO or the CTO of these companies.

659
00:45:13,915 --> 00:45:20,312
And if you're doing well your job on the core infrastructure part, naturally you're going
to have opportunities elsewhere.

660
00:45:20,312 --> 00:45:22,357
Maybe a last topic, fundraising.

661
00:45:22,357 --> 00:45:24,429
You've been quite successful in fundraising.

662
00:45:24,429 --> 00:45:30,221
Can you share a bit your approach there and also the potential for the needs in the
future?

663
00:45:30,221 --> 00:45:38,111
Yeah, so we raised 10 million in two years, especially it was like last year almost
everything because our first round was a bit in two stages.

664
00:45:38,111 --> 00:45:41,155
So yeah, last year we confirmed to 10 million.

665
00:45:41,155 --> 00:45:47,705
So now the fundraising topic is about how we can accelerate into different markets, into
different needs.

666
00:45:47,705 --> 00:45:53,420
So, yeah, the fundraising strategy is something that is always also very evolving day by
day.

667
00:45:53,420 --> 00:45:59,866
So, try to also measure what is the capacity in terms of capital you need to deploy to
make your mission a success.

668
00:45:59,866 --> 00:46:06,424
Because if you raise too much, you're then stuck in certain way if you don't achieve your
purpose.

669
00:46:06,424 --> 00:46:14,114
We see a lot of startups, I will not name them, that have raised first round 50, 100
million and then are stuck in this position.

670
00:46:14,114 --> 00:46:17,243
are so high that uh...

671
00:46:17,243 --> 00:46:24,474
how do you evaluate a company that is like already evaluated at 500 million, half a
billion dollars when there is no revenue?

672
00:46:24,474 --> 00:46:27,245
And it's not foundation, it's applied AI.

673
00:46:27,245 --> 00:46:28,265
we need to do stuff.

674
00:46:28,265 --> 00:46:29,385
Foundation, we can understand.

675
00:46:29,385 --> 00:46:31,304
Yeah, your model performs well, it does good benchmark.

676
00:46:31,304 --> 00:46:34,556
Maybe you have a few rounds before it starts to be mainstream.

677
00:46:34,556 --> 00:46:39,078
But on applied, you need to have like results on revenue, signing customers, et cetera.

678
00:46:39,078 --> 00:46:41,429
So right now our focus is...

679
00:46:41,429 --> 00:46:43,240
commercial deployment.

680
00:46:43,240 --> 00:46:44,920
We are 200 % on sales.

681
00:46:44,920 --> 00:46:45,840
It's going well for us.

682
00:46:45,840 --> 00:46:49,220
We're welcoming a lot of new customers this summer.

683
00:46:49,224 --> 00:46:53,386
It's very seasonal because it's a long sales cycle, but we are quite happy.

684
00:46:53,386 --> 00:47:02,179
yeah, for the moment, the big focus is on sales, but we are expecting probably to
fundraise again by the end of the year or mid next year maximum.

685
00:47:02,179 --> 00:47:02,525
So.

686
00:47:02,525 --> 00:47:06,958
not out of need, but out of creating this legitimacy.

687
00:47:06,958 --> 00:47:08,429
And I will be transparent with that.

688
00:47:08,429 --> 00:47:20,607
is, you know, in fundraising in VC world, you have at some points a kind of tacit election
that's happening where you raise Series A or Series B and the investors and the community

689
00:47:20,607 --> 00:47:24,250
and the customers are like, okay, these guys are gonna be the default for a few years now.

690
00:47:24,250 --> 00:47:27,522
And this is what we try to be in the next coming round.

691
00:47:27,522 --> 00:47:28,633
Yes, see.

692
00:47:28,633 --> 00:47:34,251
How did you experience the ecosystem for fundraising in what you're doing?

693
00:47:34,251 --> 00:47:42,413
Because if I'm correct, the first round you did in Paris, It's a Galleon, I think it's
Persian.

694
00:47:42,413 --> 00:47:47,198
we started in Paris, in everyone there's been like a majority of European actors.

695
00:47:47,198 --> 00:47:50,702
Actually, right now everyone is European based investors.

696
00:47:50,702 --> 00:47:52,394
So yeah, we started with Galleon.

697
00:47:52,394 --> 00:47:54,937
That was our first crazy follower.

698
00:47:54,937 --> 00:48:00,042
And last year what happened is also Cast Capital from Germany joined and...

699
00:48:00,042 --> 00:48:09,025
the same vision with us and Galleon continued to follow us and we also made a very
interesting fund, a follower named Axelio and we have also K-Fund in Spain, which is part

700
00:48:09,025 --> 00:48:09,927
of our captable.

701
00:48:09,927 --> 00:48:18,919
So yeah, it's a very pan-European captable that we have right now, but our vision is to be
global, so we are completely open-minded to raise elsewhere.

702
00:48:18,919 --> 00:48:21,142
Right now, I'm a lot in New York.

703
00:48:21,142 --> 00:48:23,773
for customers but also for potential investors.

704
00:48:23,773 --> 00:48:25,313
So it's not a thing.

705
00:48:25,313 --> 00:48:34,826
But the feedback is, if you want my honest feedback on pre-seed seed stage, and this is
something that is, I would say, yeah, everyone says it but no one really confirms it is

706
00:48:34,826 --> 00:48:40,417
that there is only a few actors that are pre-seed, that are really, really invested in
pre-seed.

707
00:48:40,607 --> 00:48:46,969
I would say right now, yeah, only two, maybe four VCs in Paris.

708
00:48:46,969 --> 00:48:49,907
our pure press player that will play the game.

709
00:48:49,907 --> 00:48:53,432
though there are a lot of funds that claim to be pre-seed, right?

710
00:48:53,432 --> 00:48:59,809
But they are still expecting you to show revenue and early market fit and that's what
you're saying, right?

711
00:48:59,809 --> 00:49:01,460
Yeah, it's a very formal world.

712
00:49:01,460 --> 00:49:06,049
So sometimes these funds that are expecting more metrics will just follow the crazy guy.

713
00:49:06,049 --> 00:49:14,497
But yeah, the very precede logic, which is like purely founders led, vision led, I would
say there is less than five in Paris that are doing it properly.

714
00:49:14,497 --> 00:49:18,970
I will not name people, of course, Gaglion is part of them.

715
00:49:18,970 --> 00:49:20,758
Kima is part of them.

716
00:49:20,758 --> 00:49:22,630
But this philosophy is still...

717
00:49:22,630 --> 00:49:25,332
Not, would say, very French and very European.

718
00:49:25,332 --> 00:49:27,183
We have seen that more in Silicon Valley.

719
00:49:27,183 --> 00:49:31,774
It's more like, okay, I like what you do, you look fun, you look like a crazy founder.

720
00:49:31,774 --> 00:49:33,805
Here is one million dollar and figure it out.

721
00:49:33,805 --> 00:49:36,985
This is not something that we are used to do in Europe.

722
00:49:37,330 --> 00:49:43,550
And in seed, it's a bit kind of the same because you're still on pre-market fits, kind of.

723
00:49:43,550 --> 00:49:47,661
You start to have some spark and something is starting, but...

724
00:49:47,661 --> 00:49:49,923
the conviction is still what's driving the investment.

725
00:49:49,923 --> 00:50:02,645
yeah, having funds like Cusp helping us on this and having the same conviction with us,
putting kind of a good amount because oh it's still a good fundraising that we did, is

726
00:50:02,645 --> 00:50:10,604
very, yeah, I would say it's very motivating because you're like, okay, people are also
thinking about it and people that does like good experience in the market are thinking

727
00:50:10,604 --> 00:50:11,156
about it.

728
00:50:11,156 --> 00:50:17,763
So yeah, but you know, as long as VCs follow the power law, this is what I say to people
fundraising.

729
00:50:17,763 --> 00:50:21,949
Find VCs that are like extremely following the power law.

730
00:50:21,949 --> 00:50:24,762
They need one investment to return the fund.

731
00:50:24,762 --> 00:50:26,824
The rest is more or less just running.

732
00:50:26,824 --> 00:50:34,324
And if they follow this logic, they will be able to also invest in kind of gut feeling and
conviction.

733
00:50:34,324 --> 00:50:34,815
Yeah.

734
00:50:34,815 --> 00:50:36,151
Interesting point of view.

735
00:50:36,151 --> 00:50:36,484
Yeah.

736
00:50:36,484 --> 00:50:36,859
point.

737
00:50:36,859 --> 00:50:37,320
Okay.

738
00:50:37,320 --> 00:50:39,676
Very impressive story, Alexander.

739
00:50:39,676 --> 00:50:40,337
Exactly.

740
00:50:40,337 --> 00:50:40,837
Yeah.

741
00:50:40,837 --> 00:50:46,175
Is there maybe if people I wanna hear can follow up this story, how can they reach you?

742
00:50:46,175 --> 00:50:47,777
How can they stay up to date?

743
00:50:47,777 --> 00:50:51,423
The website is twenty five one so two five zero one dot AI.

744
00:50:51,423 --> 00:50:56,092
But uh how can people find you, reach you, yeah, stay up to date and

745
00:50:56,092 --> 00:51:07,868
My LinkedIn is the best way to say hi to me and I reply to every message so that's
something that I continue to do every day and I'm doing enterprise so LinkedIn is my first

746
00:51:07,868 --> 00:51:09,149
social network now.

747
00:51:09,184 --> 00:51:09,775
Yeah.

748
00:51:09,775 --> 00:51:13,318
We'll put it also on the show notes for people that wanna follow you.

749
00:51:14,545 --> 00:51:15,686
Yes, thanks a lot.

750
00:51:15,686 --> 00:51:17,368
Is there anything else you wanna you wanna say?

751
00:51:17,368 --> 00:51:20,082
Any any less less words before we call it?

752
00:51:20,082 --> 00:51:22,753
Now I would say, yeah, people continue to grind.

753
00:51:22,753 --> 00:51:26,504
It's been like 30 years that we didn't see something like that happening.

754
00:51:26,504 --> 00:51:34,995
So AI revolution is the moment to create, it's the moment to grind, it's the moment to do
some code.

755
00:51:34,995 --> 00:51:38,906
I think I'm very optimistic person.

756
00:51:38,906 --> 00:51:40,697
I'm more like pro.

757
00:51:40,697 --> 00:51:44,701
innovation than afraid of it, it's gonna change a lot of things.

758
00:51:44,701 --> 00:51:46,322
Let's not be realistic.

759
00:51:46,322 --> 00:51:55,440
But at the same point, we are potentially seeing what is one of the first step to, I don't
know, a new kind of economy, a new kind of humanity.

760
00:51:55,460 --> 00:52:00,105
Yeah, our life is gonna change, and I think in quite interesting way.

761
00:52:01,207 --> 00:52:01,827
Yeah.

762
00:52:01,827 --> 00:52:02,459
Cool.

763
00:52:02,459 --> 00:52:03,372
Thanks a lot.

764
00:52:03,372 --> 00:52:05,428
And yeah, thanks for sitting with us and chatting with us.

765
00:52:05,428 --> 00:52:06,203
Pleasure having you.

766
00:52:06,203 --> 00:52:08,247
you