UDSKRIFT 🔗
WEBVTT
1
00:00:00.111 --> 00:00:10.441
I think the idea that, oh, well, there's like product market fit, like we do need to pay attention to the market, did lead over then to the second company, which we started to ideate and think about, okay, what do we really want to build?
2
00:00:11.535 --> 00:00:15.477
On today's episode of Inside the Silicon Mind, we're excited to welcome Clement Pang.
3
00:00:15.577 --> 00:00:27.844
Clement is a serial entrepreneur and software engineer with over 15 years in tech, known for co-founding Wavefront, which VMware acquired for $400 million, and Lovana Labs, where he built AI learning companions to transform children's literacy.
4
00:00:27.944 --> 00:00:33.928
He began his career at Google, leading fraud and abuse infrastructure to protect billions of users and later helped build
5
00:00:34.128 --> 00:00:39.291
GLMX, a money market trading platform now processing over $1 trillion daily.
6
00:00:39.351 --> 00:00:46.174
Clement is an entrepreneur in residence at Sutton Hill Ventures, exploring new ventures at the intersection of AI, data and human impact.
7
00:00:46.374 --> 00:00:53.458
In this episode, we dive into Clement's journey through startups, AI and the future of building impactful technology at scale.
8
00:00:53.658 --> 00:01:01.963
At what point did you know, whether it's the university, whether it's being at Google, at what point did you know, I'm going to be a founder, I'm going to be, or at least start thinking about being part of a startup?
9
00:01:03.124 --> 00:01:15.144
I think when I started in my first company... Clement, I'm curious.
10
00:01:16.085 --> 00:01:19.587
How has AI changed what it means to be a founder today?
11
00:01:19.887 --> 00:01:21.068
Well, it's changed so many things, right?
12
00:01:21.088 --> 00:01:27.672
Like there is the how you build stuff, how you ideate about, you know, product, how you even decide what to build these days, right?
13
00:01:27.712 --> 00:01:37.878
Because I think there's always this joke, like, you know, one of the foundation model companies like OpenAI, Anthropic, like they're just so close to like the adjacent GPT wrappers.
14
00:01:37.898 --> 00:01:42.801
So you want to be making sure that you're not building something that they could just release like the next day, right?
15
00:01:42.861 --> 00:01:44.202
And just replace your company.
16
00:01:44.938 --> 00:01:47.539
But in terms of ideation, I'm in the car.
17
00:01:48.239 --> 00:01:55.142
I'm chatting with ChatGPT and thinking about, OK, I want to learn about thorium reactors, because I want to understand how they work.
18
00:01:55.182 --> 00:01:57.943
And they would just explain to me I'm five.
19
00:01:58.143 --> 00:02:02.124
And that's helped me with so many things, because I just want to understand, hey, why is this a thing?
20
00:02:02.144 --> 00:02:03.085
Why is this company a thing?
21
00:02:03.105 --> 00:02:04.285
Why do people use it?
22
00:02:04.705 --> 00:02:06.166
How this file format works?
23
00:02:06.226 --> 00:02:07.586
And why are companies buying this?
24
00:02:07.687 --> 00:02:08.667
And how are they selling?
25
00:02:09.227 --> 00:02:10.228
You just learn so much about that.
26
00:02:10.388 --> 00:02:13.869
And then finally, of course, as somebody who builds, writes code, right?
27
00:02:13.889 --> 00:02:18.091
Like cloud code has just changed kind of just everything.
28
00:02:18.111 --> 00:02:26.315
You feel like you have so many junior devs, or even potentially senior devs as they are becoming more intelligent.
29
00:02:27.295 --> 00:02:28.596
that are helping you constantly, right?
30
00:02:28.616 --> 00:02:32.318
So just how can you run multiple agents in parallel doing things?
31
00:02:32.378 --> 00:02:38.382
And then, you know, in the past, like maybe you don't write all the unit tests and now there's no excuse that you have to write all the unit tests.
32
00:02:38.402 --> 00:02:39.582
Like it's all covered, right?
33
00:02:39.622 --> 00:02:46.987
And so, yeah, it's just changed, I think, how you would build, what you would build and, yeah, everything in between.
34
00:02:47.238 --> 00:02:52.799
I'm guessing as far as like a shift in the industry, is this the first big change since when?
35
00:02:52.839 --> 00:02:54.119
Since cloud, since mobile?
36
00:02:54.139 --> 00:02:55.219
I think since cloud, yeah.
37
00:02:55.239 --> 00:02:58.200
Started my first company, we're building, we're racking stuff in Equinix, right?
38
00:02:58.240 --> 00:03:00.620
And the second company is already in AWS.
39
00:03:01.141 --> 00:03:10.042
I think that was a big change because you're not, well, it used to be fun because I guess you could go to the data center and you could see their giant flywheel like power backups.
40
00:03:10.602 --> 00:03:13.283
But now it's just something in the, you don't even know where it is.
41
00:03:13.443 --> 00:03:14.683
It's probably in Oregon somewhere, right?
42
00:03:15.717 --> 00:03:18.058
But now it's like, yeah, it's another fundamental shift.
43
00:03:18.078 --> 00:03:21.780
I think the first shift to cloud allowed us to just play with stuff, right?
44
00:03:21.820 --> 00:03:25.181
Nowadays, even like it's impacting our own lives, right?
45
00:03:25.461 --> 00:03:31.264
I think about whether I should buy a GPU or like, oh, maybe now I could just, you know, go to one of the SaaS services and do that.
46
00:03:31.284 --> 00:03:34.005
So that shift, of course, like I think has come and stayed.
47
00:03:34.505 --> 00:03:43.129
But I think AI is just another step function where everything is like, you know, like careers are being altered as we speak, right?
48
00:03:43.803 --> 00:03:52.329
It's interesting that the way I look at it from an industry standpoint, is it similar to the internet age as far as like the big shift that's happening?
49
00:03:53.370 --> 00:03:53.650
Yeah.
50
00:03:53.810 --> 00:03:57.713
From, you know, more from like a, you need to think about AI.
51
00:03:57.753 -->
Load full transcript
00:04:02.196
Like when the internet came around, it was everyone needs to become an internet company.
52
00:04:02.236 --> 00:04:02.436
Yeah.
53
00:04:02.556 --> 00:04:03.877
I think I would think of it as.
54
00:04:04.667 --> 00:04:13.994
is it a disruptor that is so big that a new player could completely unseat a traditional entrenched player?
55
00:04:14.354 --> 00:04:16.356
And I think the advent of the computer, right?
56
00:04:16.416 --> 00:04:28.545
So if you were to get on the bandwagon early enough, you could build something better than companies that did not have computers, companies that did not operate on the internet as an e-commerce store, you could build something Amazon to do that.
57
00:04:28.585 --> 00:04:30.706
And I think AI now has the same effect.
58
00:04:30.746 --> 00:04:31.307
I think a lot of...
59
00:04:32.122 --> 00:04:44.388
existing legacy companies, they would have so much processes, so much code base, so much, you know, cruft to deal with that, yeah, another company could come in and, you know, rethink the whole process of, I don't know, like booking airline tickets, right?
60
00:04:44.408 --> 00:04:45.629
That's currently cited.
61
00:04:46.489 --> 00:04:48.230
And you could build something brand new and it could be better.
62
00:04:48.790 --> 00:04:52.352
And suddenly, like, you know, they have a real major problem.
63
00:04:52.754 --> 00:04:54.354
And you're in the ideation phase right now.
64
00:04:54.394 --> 00:04:56.895
You're thinking about what's next and you're building.
65
00:04:58.336 --> 00:05:06.258
This is an interesting time because when you think about where you are today, you're a serial entrepreneur, you've done this many times over and we'll get to that.
66
00:05:06.898 --> 00:05:16.081
But as far as the funding piece and starting a company, the size of the company, how many engineers do you need today versus what you needed before?
67
00:05:16.881 --> 00:05:19.642
What is the cost to building a startup today versus before?
68
00:05:20.782 --> 00:05:23.083
What are the most significant changes you're seeing right now?
69
00:05:23.363 --> 00:05:30.947
Well, certainly I think there is this, as I just said, like there's this feeling that you could build so much, right, with agents helping you to do that, right?
70
00:05:30.967 --> 00:05:33.688
So are you going to hire that many junior devs and senior devs?
71
00:05:34.009 --> 00:05:38.651
But then you also think like, oh, but then, you know, what happens to the industry or even your company, right?
72
00:05:38.951 --> 00:05:40.872
Because you play it a couple of years down, right?
73
00:05:40.892 --> 00:05:42.773
You really do want people... Because...
74
00:05:43.244 --> 00:05:47.607
it's kind of cool to be running agents for a while, but then you realize, okay, so I'll be running agents for like customer support.
75
00:05:47.627 --> 00:05:50.770
I'm going to run agents for like my DevOps, my ops, my, you know, everything.
76
00:05:51.170 --> 00:05:54.353
And at some point, right, well, that also is, you know, takes on a toll, right?
77
00:05:54.373 --> 00:05:57.335
So maybe you have some meta agent that's managing all of that, right?
78
00:05:58.156 --> 00:06:12.807
But it becomes this question of like, okay, if I really truly built this and it is just a one man army, like I always have this quote from all the companies I've been at, which is like people tell me like, you know, there's the Clement run by bus problem, like in many of the places I've been at.
79
00:06:13.641 --> 00:06:16.642
And now you're like concentrating that risk even further, right?
80
00:06:16.662 --> 00:06:18.862
Because nobody really knows what's going on.
81
00:06:18.982 --> 00:06:25.543
Like, again, maybe at some point there's this meta agentic things that even if you're like stepping away completely, like the company still runs.
82
00:06:26.064 --> 00:06:28.184
But we're not there even with coding.
83
00:06:28.204 --> 00:06:31.665
So let alone like you could just run marketing, you know, with an agentic thing.
84
00:06:31.725 --> 00:06:32.105
Right.
85
00:06:32.165 --> 00:06:37.846
So as far as like the role of a software engineer, how is that going to look in the coming years?
86
00:06:38.324 --> 00:06:40.645
Yeah, I think recently someone told me a joke.
87
00:06:40.985 --> 00:06:43.446
It was like, oh, we don't need developers anymore.
88
00:06:43.466 --> 00:06:50.390
We just need people who knows exactly how to tell the computer to do the things that we want it to do, right?
89
00:06:50.470 --> 00:06:56.552
Which, yeah, so I think developers probably will still be around, but it's kind of like, okay, we don't program an assembly.
90
00:06:56.612 --> 00:07:00.714
We don't program, most people don't program that low level these days, right?
91
00:07:01.175 --> 00:07:03.456
But I think there's another level of abstraction.
92
00:07:03.476 --> 00:07:04.316
Of course, it opens up.
93
00:07:04.919 --> 00:07:07.662
development to like product managers and to designers.
94
00:07:08.083 --> 00:07:15.973
But I think for even for engineers, like it's the, at least the agents of these days, like they, they get very eager to like, just solve your problem, right.
95
00:07:16.013 --> 00:07:19.838
Without thinking too much about like, how is it going to interoperate with existing systems?
96
00:07:19.978 --> 00:07:21.580
Are you going to build it in a way that.
97
00:07:22.483 --> 00:07:28.489
you could do the next 20 things you want to do with it, because you just give it a single prompt, like, I need just this little piece of software.
98
00:07:28.529 --> 00:07:31.612
It's going to just maybe take the path of least resistance and build it for you.
99
00:07:32.313 --> 00:07:38.298
So developers are still there just trying to think a little bit ahead of like, hey, can I ask more questions?
100
00:07:38.318 --> 00:07:42.302
Because probably you'd have a little bit more of the business sense of what are you going to do?
101
00:07:42.402 --> 00:07:43.624
What is the next 20 things you want to do?
102
00:07:43.884 --> 00:07:46.066
And so you may prompt it better, or when it's building,
103
00:07:46.566 --> 00:07:52.349
I think right now what I often have to do is just go into a planning phase with the AI and be like, I don't write any code.
104
00:07:52.369 --> 00:07:57.331
Just let me ideate with you for a little bit, write a design docs, and then go attempt the work.
105
00:07:57.471 --> 00:08:02.733
And I think that might still – that might be just software developers or ideation.
106
00:08:03.435 --> 00:08:09.317
kind of partners and then the person that's certainly writing the unit tests and the commit messages like that's the AI.
107
00:08:09.397 --> 00:08:17.499
But then you're monitoring the actual code that is being written and stopping it and pausing it at times and then giving additional suggestions.
108
00:08:17.519 --> 00:08:21.460
So I think it's an elevated plane of work, I think.
109
00:08:21.640 --> 00:08:22.661
Yeah.
110
00:08:22.741 --> 00:08:29.843
And Clement, when you think about the scope of AI today as a new piece of technology that everyone is playing with and building new companies.
111
00:08:31.170 --> 00:08:34.719
Is it now harder or easier to start companies?
112
00:08:35.621 --> 00:08:38.347
I think it's easier to start companies, but...
113
00:08:39.261 --> 00:08:40.542
well, it's harder to get funding.
114
00:08:40.702 --> 00:08:44.043
So it's like, is it easy to start a company that stands a test of time?
115
00:08:44.063 --> 00:08:44.584
Right.
116
00:08:44.624 --> 00:08:55.389
I think now it's hard because you really could have somebody who looks, I mean, you could probably just go into any of the, you know, coding help and just say like, write me this, right.
117
00:08:55.409 --> 00:08:57.050
Write me this thing that I see on the app store.
118
00:08:57.370 --> 00:09:00.171
And it'll probably get you like, you know, maybe 80% of the way.
119
00:09:00.231 --> 00:09:05.114
Like, so, so you always have somebody who could, who could be competing with you and, um,
120
00:09:06.419 --> 00:09:23.272
Yeah, so, but easy because, you know, you have things like Vercel, you have things like, you know, like serverless, like databases, clouds, you have, yeah, of course, Lovable and Bolt if you want to feel like completely, you know, too agentic and just prompt based.
121
00:09:24.352 --> 00:09:24.933
product work.
122
00:09:25.033 --> 00:09:27.375
But yeah, but I think it is getting easier.
123
00:09:27.435 --> 00:09:34.721
But I think the problem is once you get to that stage, then what is really good software like starts to stand out, right?
124
00:09:34.761 --> 00:09:40.706
Because, you know, I've gone through my, you know, companies where, you know, it's infrastructure companies, right?
125
00:09:40.726 --> 00:09:44.109
A lot of it is you need to have really good algorithms, right?
126
00:09:44.129 --> 00:09:49.513
You need to have very good architecture design, scalability, resiliency, and all that kind of stuff, right?
127
00:09:49.653 --> 00:09:51.435
And so those are still...
128
00:09:52.173 --> 00:09:53.554
you know, requirements, right?
129
00:09:53.614 --> 00:09:58.937
And to the degree that, you know, you could, you're just vibing all of this, right?
130
00:09:58.977 --> 00:10:02.520
Where you actually know what the best practices are and whatnot.
131
00:10:02.540 --> 00:10:06.142
I think those companies will probably stand out a bit more, I feel.
132
00:10:06.162 --> 00:10:18.510
Because I, yeah, it's, again, like, this like zero shot situation where you're just typing something and it goes off and does something, like oftentimes means it will just, you know,
133
00:10:19.743 --> 00:10:23.265
build your company maybe on shaky foundation, it allows you to build it quickly.
134
00:10:23.785 --> 00:10:29.088
But then, yeah, maybe you then ask the agent to come back and be like, okay, let me build it like it would be Google scale.
135
00:10:29.388 --> 00:10:34.931
And then does it know how to come back and do all of that work and refactor and do all the migration and all that kind of stuff?
136
00:10:35.051 --> 00:10:37.993
So is it better then at that point to build it right the first time?
137
00:10:39.215 --> 00:10:43.438
So it's easy, but how do you build something that's enduring that really could scale?
138
00:10:43.578 --> 00:10:45.059
And also differentiable, right?
139
00:10:45.079 --> 00:10:50.582
So is there actual IP, actual algorithms, actual new insights that you're bringing to a particular problem?
140
00:10:50.602 --> 00:10:51.683
I think that's still going to be there.
141
00:10:52.404 --> 00:10:58.088
One phrase I hear a lot nowadays is this whole concept of it, as you said earlier on, a wraparound chat GPT.
142
00:10:58.268 --> 00:11:04.232
And the risk is, you know, at what point does chat GPT completely take over and see what's working?
143
00:11:04.493 --> 00:11:08.615
And do you see that as like a real risk right now in the market?
144
00:11:08.635 --> 00:11:13.859
Are there a lot of companies doing this and are they playing it quite a... Yeah, yeah, I think so.
145
00:11:14.640 --> 00:11:18.062
You know, some companies I am familiar with even, right, like that they,
146
00:11:18.577 --> 00:11:20.200
started like, let's use Claude, right?
147
00:11:20.220 --> 00:11:23.686
Like the, it was more like a chat GPT clone in the very beginning and
148
00:11:24.231 --> 00:11:25.972
people noticed that it was really good at coding.
149
00:11:26.332 --> 00:11:33.474
So I think maybe two years ago, I was pasting actual problems into ChatGPT and then doing cut and paste and all that kind of stuff.
150
00:11:33.494 --> 00:11:36.035
And then Cloud came along and be like, ooh, MCP is really cool.
151
00:11:36.155 --> 00:11:38.616
So I used MCP to just give a terminal access and do that.
152
00:11:39.257 --> 00:11:45.439
And then of course, Anthropic probably saw
that, hey, why don't we just even own that space and just make Cloud Code a thing?
153
00:11:45.979 --> 00:11:51.141
But I'm sure enough companies, whether it's the Cursor or the Windsurf and other companies who,
154
00:11:52.176 --> 00:12:07.780
who saw that the models are really good at coding, built a product out of it, now realizing that actually the models that they're using or that they may be fine tuning or at least calling millions of times a second, is actually going to compete with them and you're actually giving them the training data.
155
00:12:08.300 --> 00:12:11.461
So whether they use it or not, but yeah.
156
00:12:11.601 --> 00:12:15.742
But that's a real issue, because these are like billion dollar businesses.
157
00:12:15.982 --> 00:12:18.023
And so if you are so adjacent to
158
00:12:19.043 --> 00:12:23.011
the textual completion problem, then you might be at risk.
159
00:12:23.901 --> 00:12:29.642
Are we at a point now in the industry where we know which companies own the application land and really own the market from an AI standpoint?
160
00:12:29.902 --> 00:12:31.203
I don't think so.
161
00:12:31.623 --> 00:12:32.283
Yeah.
162
00:12:32.863 --> 00:12:36.964
So I use like ChatGPT for it to plan travel, right?
163
00:12:37.164 --> 00:12:43.225
But it's, you understand like text chat is only gets you so far because if you're traveling at an airport, right?
164
00:12:43.245 --> 00:12:48.286
Like you're not going to scroll your text and be like, when did I talk to, well, you could ask it again, like, when's my flight, blah, blah, blah.
165
00:12:48.306 --> 00:12:51.807
But it's probably easier if they have a UX that actually tells you like, hey, this is,
166
00:12:52.227 --> 00:13:03.756
you know, flight information in some sort of a, like Flighty, right, which is a great app, right, that actually, you know, shows you that information, surface it in a nice way at a glance on your phone, on your watch and all that kind of stuff, right?
167
00:13:03.776 --> 00:13:19.349
So I think the app layer, I don't know, maybe it's IO, you know, like whatever it is, right, that will come where it incorporates IO into the more daily motions of life that's beyond just, you know, textual chat and voice, right?
168
00:13:19.696 --> 00:13:20.236
chats, right?
169
00:13:20.476 --> 00:13:22.077
And also there's the memory question too, right?
170
00:13:22.097 --> 00:13:24.777
Because a lot of these chats are just one shot, right?
171
00:13:24.797 --> 00:13:26.418
Like you could have memory and chat GPT.
172
00:13:26.458 --> 00:13:29.819
It's not, it doesn't have your entire life in it, right?
173
00:13:29.859 --> 00:13:33.500
And so how do you actually, you know, do you want that, right?
174
00:13:33.520 --> 00:13:36.040
Do you want like an AI that actually knows everything about your life?
175
00:13:36.080 --> 00:13:41.041
It probably will help you, you know, respond to emails and texts and all that, you know, a little bit more intelligently.
176
00:13:41.922 --> 00:13:47.103
Or do you not want that and just want everything to have a fresh, you know, state when it tries to attempt a problem?
177
00:13:48.026 --> 00:13:48.326
I don't know.
178
00:13:48.426 --> 00:13:57.533
Like I recently just heard someone told me like the friends were saying like your chat GPT just always sounds like you because there's enough memory in it that it just colors its response.
179
00:13:57.733 --> 00:13:58.574
Yeah.
180
00:13:58.774 --> 00:14:00.615
And so sometimes maybe that's not a good thing.
181
00:14:00.635 --> 00:14:05.979
Like you actually want the real or the unvarnished chat GPT.
182
00:14:06.239 --> 00:14:07.300
And yeah.
183
00:14:07.400 --> 00:14:09.141
So I think there's all these questions out there.
184
00:14:09.181 --> 00:14:17.227
Like are people looking for that universal, you know, full memory, you know, full reasoning AI to help with their lives and work for a person or a family?
185
00:14:17.789 --> 00:14:19.710
Or are they looking at more siloed, right?
186
00:14:19.750 --> 00:14:22.872
Like something that actually understands, let's say, all of my travel, right?
187
00:14:22.892 --> 00:14:29.537
And it's like the app is engineered only for travel, and it's all the bells and whistles that it requires.
188
00:14:30.417 --> 00:14:34.840
Completely out of the box question, but are we training these models by actually using the models?
189
00:14:35.825 --> 00:14:43.286
If you use the non-enterprise API versions, you're giving them consent to do that.
190
00:14:43.627 --> 00:14:57.029
And surely, as far as how much we're paying to use ChatGPT, whether it's $20 or $60, ultimately the cost of running these queries that we're putting in, it costs more.
191
00:14:58.029 --> 00:14:59.330
Yeah, for sure.
192
00:14:59.550 --> 00:15:01.130
So I think it's certainly a land grab.
193
00:15:01.612 --> 00:15:05.834
There are certainly, you're helping them, you know, build the models doing some of the reinforcement learning.
194
00:15:06.935 --> 00:15:21.782
But, yeah, but I think people are generally, I mean, in the history of different models of how it works, right, the Ubers and the Lyfts, right, like the, maybe even the early days, the Amazon, right, like, so it's okay to be losing money for a while as long as you're grabbing Mindshare and whatnot.
195
00:15:21.822 --> 00:15:27.745
And I would always say, like, kids these days, like, ChatGP is such a thing that...
196
00:15:29.158 --> 00:15:31.580
Like it's actually very hard to displace that.
197
00:15:31.700 --> 00:15:36.944
Even if you say like Gemini is there like Google search, like Google to them is the search engine, right?
198
00:15:36.964 --> 00:15:39.866
And ChatGPT is, you know, how I do my homework, let's say.
199
00:15:40.687 --> 00:15:47.152
And even if you could go to Gemini and do the same thing, I think that, you know, ChatGPT, which is not a very –
200
00:15:47.903 --> 00:15:50.985
you know, well engineered product name, but, but
it just really stuck with people.
201
00:15:51.385 --> 00:15:52.646
Everybody is going to chat GPT.
202
00:15:52.686 --> 00:15:56.328
If you're not using chat GPT, people think like you should be using chat GPT.
203
00:15:56.348 --> 00:15:58.089
Like why are you on cloud or on Gemini or something?
204
00:15:58.109 --> 00:15:58.770
That's true.
205
00:15:58.790 --> 00:16:05.954
Like again, benchmarks are one thing, but you know, I think the land grab, the perception, all of that is, is certainly, you know, what they got.
206
00:16:06.441 --> 00:16:12.586
Yeah, I mean, I hear my nephews and nieces saying, ChatGBT as opposed to Google it nowadays.
207
00:16:12.606 --> 00:16:13.287
Yeah, yeah, yeah.
208
00:16:13.307 --> 00:16:17.630
I'm sure the Oxford Dictionary will incorporate it in some years.
209
00:16:18.091 --> 00:16:19.832
Clement, let's go to the very beginning, right?
210
00:16:19.852 --> 00:16:21.514
So you had multiple startups.
211
00:16:21.594 --> 00:16:24.496
So I'd love to get a glimpse into that.
212
00:16:24.536 --> 00:16:28.639
Let's unpack the beginning, you know, that founder journey.
213
00:16:28.680 --> 00:16:30.421
Before the founder journey, the engineer journey.
214
00:16:30.441 --> 00:16:35.025
So what's that transition from the engineer to the founder?
215
00:16:35.544 --> 00:16:41.407
Yeah, I think there are, of course, engineers turned founders in the Valley.
216
00:16:41.947 --> 00:16:44.668
I feel maybe there is not enough of it.
217
00:16:45.289 --> 00:17:04.238
And I think that's probably because Google does a very good job at just, you know, having people, you know, I think, again, I think this was something that people told me a while back was like one of the best things that Google did was how to figure out, like, to get engineers to work in their, you know, in their problem space and spend like, you know, 15, 20 years doing it.
218
00:17:05.399 --> 00:17:09.881
Because, like, that's how they make sure there's not a competitor competing with them, right?
219
00:17:10.041 --> 00:17:14.864
So I think, if anything, like, I think my journey started at Google, right?
220
00:17:15.284 --> 00:17:16.824
Like, yeah, unlike a lot of people.
221
00:17:17.785 --> 00:17:21.867
2006 joined Google as an intern, then turned full-time.
222
00:17:23.007 --> 00:17:26.969
But I never actually had, like, entrepreneurial aspirations, right?
223
00:17:26.989 --> 00:17:27.730
So I think...
224
00:17:28.370 --> 00:17:29.210
It's not in the family?
225
00:17:29.310 --> 00:17:30.911
It's not in the family, yep.
226
00:17:31.331 --> 00:17:33.372
I grew up in an Asian family.
227
00:17:33.412 --> 00:17:34.833
Like, they were...
228
00:17:35.400 --> 00:17:36.801
They were actually against me joining Google.
229
00:17:36.841 --> 00:17:39.062
They were actually saying, like, you better off join Microsoft.
230
00:17:39.342 --> 00:17:40.062
I got offers from both.
231
00:17:40.603 --> 00:17:42.844
Because at the time, Microsoft was.
232
00:17:42.924 --> 00:17:48.507
Yeah, Microsoft is a bullshit company, and Google might disappear tomorrow.
233
00:17:49.047 --> 00:17:54.070
But I interned at Google, and again, I wasn't considering the risk and all that.
234
00:17:54.130 --> 00:17:55.871
I was just saying, oh, California is great.
235
00:17:55.911 --> 00:17:56.951
Like, weather is great.
236
00:17:57.052 --> 00:17:57.892
Like, the team is great.
237
00:17:58.443 --> 00:17:59.844
solving really wonderful problems.
238
00:17:59.884 --> 00:18:01.605
I remember Peter Norvik was in the team.
239
00:18:01.625 --> 00:18:02.305
I think he's still there.
240
00:18:03.025 --> 00:18:06.007
And he's the author of my artificial intelligence text book.
241
00:18:06.747 --> 00:18:07.187
So yeah.
242
00:18:08.228 --> 00:18:11.429
And many years later, VMware acquired my second company.
243
00:18:11.489 --> 00:18:14.210
So I think VMware and Google started in the same year.
244
00:18:14.230 --> 00:18:18.232
So Google is this cool information retrieval search company.
245
00:18:18.292 --> 00:18:19.673
VMware is an infrastructure company.
246
00:18:19.793 --> 00:18:23.695
MARK MANDELSKI- So when you were at Google, there was no idea that one day you were going to start this company.
247
00:18:24.759 --> 00:18:26.020
Like be a founder one day.
248
00:18:26.200 --> 00:18:27.240
That was not a concept for you.
249
00:18:27.260 --> 00:18:27.681
Yeah.
250
00:18:28.401 --> 00:18:31.783
I think I was just a really nerdy person.
251
00:18:31.823 --> 00:18:36.385
Like I just like, you know, I mean at Google it was like, it's like Disneyland, right?
252
00:18:36.405 --> 00:18:36.885
It's the buffet.
253
00:18:36.905 --> 00:18:37.726
Like it's everything.
254
00:18:37.746 --> 00:18:37.986
Sure.
255
00:18:39.045 --> 00:18:44.746
everything at Google is a single repo, like all the documentation was open, at least I think it's still true back today.
256
00:18:45.246 --> 00:18:48.287
And so you could just be like talking to really interesting people.
257
00:18:48.387 --> 00:18:54.008
And I think back then, like even production services would not be blocked off from teams.
258
00:18:54.368 --> 00:18:56.948
So you could suddenly call some service that you don't know.
259
00:18:57.088 --> 00:19:03.109
And it gives you this amazing, like let's say nearest neighbor service, or it gives you this amazing like browser emulation.
260
00:19:03.129 --> 00:19:05.550
I guess now it's back in Vogue.
261
00:19:05.630 --> 00:19:08.210
Like all that kind of stuff, like it used to be like, oh, I just call this and then,
262
00:19:08.728 --> 00:19:11.629
you call it and then you're using it and you call the other team and be like, oh, I'm using it.
263
00:19:11.669 --> 00:19:13.589
Like, this is in production.
264
00:19:14.370 --> 00:19:14.770
I'm sorry.
265
00:19:14.850 --> 00:19:16.930
So it was so fun, I think, at that point.
266
00:19:16.950 --> 00:19:19.311
But it was just, you know, really nerdy stuff.
267
00:19:20.051 -->
00:19:27.454
And I think I got to go a little bit further back because I started, of course, doing programming and computer science at Rice in Houston.
268
00:19:28.014 --> 00:19:28.774
And I remember, like...
269
00:19:29.981 --> 00:19:36.811
To give you the contrast of how unlikely I would be an entrepreneur, I remember back then I would be like at my room.
270
00:19:36.972 --> 00:19:40.537
I'm doing CS and double E, double major, and I would be coding and stuff.
271
00:19:40.617 --> 00:19:44.143
And my next-door neighbors would be having parties and like there were people coming along.
272
00:19:44.583 --> 00:19:45.084
And I would...
273
00:19:45.968 --> 00:19:47.189
I would just be staring at my screen.
274
00:19:48.049 --> 00:19:52.452
I just completely disconnected from what's happening in the next room.
275
00:19:52.992 --> 00:19:55.953
And I would just be typing and coding and stuff.
276
00:19:56.354 --> 00:19:59.455
I'm that nerdy of a guy.
277
00:20:00.396 --> 00:20:02.557
Before we sort of dive into the family, that's interesting.
278
00:20:02.597 --> 00:20:04.378
I didn't quite expect us to dive into that.
279
00:20:04.398 --> 00:20:07.720
But that level of focus and attention to what you're doing.
280
00:20:08.340 --> 00:20:09.200
You're at university.
281
00:20:09.281 --> 00:20:10.921
Everyone's partying and having fun.
282
00:20:10.941 --> 00:20:13.243
But you're there focused.
283
00:20:14.883 --> 00:20:15.684
Where does that come from?
284
00:20:16.183 --> 00:20:17.204
I think it's an obsession.
285
00:20:17.224 --> 00:20:19.707
I think all of us have some OCD going on.
286
00:20:20.228 --> 00:20:23.371
And my parents certainly told me I did have that as a kid.
287
00:20:23.431 --> 00:20:35.965
So I think just computers, like, you know, going even further backwards, like the first time you installed for my generation, like Visual Basic and putting it on Windows and be like, in file, there's a command called make.exe.
288
00:20:37.552 --> 00:20:40.214
And I'm like, oh, so that's how you make computer software.
289
00:20:40.794 --> 00:20:46.918
Because until that time, you're just saying, okay, well, you just, you know, you buy something or you got something and then it's .exe, right?
290
00:20:46.938 --> 00:20:49.700
And that's an app or that's a program.
291
00:20:49.760 --> 00:20:54.043
And then when you got that, then you'd be like, okay, this is how you make a program.
292
00:20:54.423 --> 00:20:59.246
And so you started on that journey and then you're trying to learn, you know, Visual Basic and all that kind of stuff, right?
293
00:20:59.426 --> 00:21:05.330
And so I think that just, the act of making, right, I think this is something that is true universally.
294
00:21:05.370 --> 00:21:05.470
Like.
295
00:21:06.179 --> 00:21:08.443
you know, whether you're making art, whether you're making music, right?
296
00:21:08.463 --> 00:21:14.815
But making like something where it looks and feels and, you know, people can touch it and
be like, well, this is really cool.
297
00:21:15.196 --> 00:21:16.879
Like that reverberates, right?
298
00:21:16.939 --> 00:21:17.560
And so I think...
299
00:21:18.538 --> 00:21:22.061
And then there's this, I think this is, there was that obsession about like trying to solve something.
300
00:21:22.201 --> 00:21:23.342
Like you want to make something faster.
301
00:21:23.362 --> 00:21:25.183
You want to make something, you know, scale better.
302
00:21:25.203 --> 00:21:29.346
You want to make something just a little bit more performance sometimes.
303
00:21:29.466 --> 00:21:29.606
Right.
304
00:21:29.706 --> 00:21:31.768
And, and it just, you know, sucks.
305
00:21:32.428 --> 00:21:37.472
Something, something is in your soul that you feel like, okay, I want to, you know, just act out a little bit of performance.
306
00:21:37.592 --> 00:21:39.093
I don't like it failing like this.
307
00:21:39.133 --> 00:21:41.435
I don't like it, you know, spewing a message.
308
00:21:42.196 --> 00:21:42.396
Yeah.
309
00:21:42.436 --> 00:21:43.957
So I think it's a, it is an obsession.
310
00:21:43.977 --> 00:21:44.077
Yeah.
311
00:21:44.637 --> 00:21:53.341
But that obsession and that, as you said, making something has AI and I'm slightly digressing here, but has AI somewhat taken that away from us?
312
00:21:53.981 --> 00:21:56.202
Like, especially for engineers and founders, they're possible.
313
00:21:56.222 --> 00:22:05.647
We live in a world where our ability to create something and as you said, make something it's going to be through prompting an AI to do it for you.
314
00:22:05.827 --> 00:22:06.067
Yeah.
315
00:22:06.287 --> 00:22:10.089
Well, I think it's certainly leveled the playing field because a lot of people who don't
316
00:22:10.613 --> 00:22:13.516
necessarily have all the training to write code, right?
317
00:22:13.556 --> 00:22:14.917
They could do that with prompting.
318
00:22:15.577 --> 00:22:20.922
I think for a lot of engineers, right, I'm a big fan of the book Building Right It, right?
319
00:22:21.202 --> 00:22:26.106
And I think engineers tend to build it right, but not oftentimes build the right it.
320
00:22:27.067 --> 00:22:36.876
And I think the ability to have AI allows us to be like, okay, why don't we really figure out what is the right it?
321
00:22:37.216 --> 00:22:37.516
Sure.
322
00:22:37.556 --> 00:22:37.857
Because
323
00:22:38.760 --> 00:22:40.781
I have this problem still today, right?
324
00:22:40.821 --> 00:22:43.962
Where I would just sit down and be like, oh, I really want to use this package.
325
00:22:44.222 --> 00:22:45.403
I really want to use this.
326
00:22:45.823 --> 00:22:48.284
I really want it to be refactored and look like this.
327
00:22:48.364 --> 00:22:48.644
Right.
328
00:22:48.664 --> 00:22:50.625
Because I, this is an obsession, right?
329
00:22:51.065 --> 00:22:56.267
And I would get, you know, an afternoon, you know, this time, not wasted, but, you know, it's spent on it.
330
00:22:56.767 --> 00:23:03.450
But now if you like have the prompt layer on top of it, then you'd be like, okay, I don't, I don't really care at this point because I want to put something out there.
331
00:23:03.510 --> 00:23:05.171
I want to see like what people are thinking about it.
332
00:23:05.491 --> 00:23:05.691
Right.
333
00:23:05.711 --> 00:23:06.131
And so I think,
334
00:23:06.642 --> 00:23:12.924
Currently in my current role, I get a lot of chances to be like, okay, let me do a week-long kind of spike and be like, okay, I'll build this.
335
00:23:13.264 --> 00:23:15.264
I don't really care if there are unit tests.
336
00:23:15.724 --> 00:23:19.625
Right now it does, because nowadays you can't really have an excuse for that.
337
00:23:19.645 --> 00:23:28.627
But yeah, but if it doesn't have all the right vulnerability checks when you're checking in code, that's okay.
338
00:23:28.928 --> 00:23:30.688
Even though it feels wrong, it feels like the
339
00:23:31.071 --> 00:23:34.532
You know, the coding style needs to have, you know, spotless or check style running.
340
00:23:34.552 --> 00:23:36.152
Like you don't feel that you need that.
341
00:23:36.212 --> 00:23:39.393
But yeah, so just code away and do all of that.
342
00:23:39.453 --> 00:23:42.394
And then when you really figure out that there's some there there, right?
343
00:23:42.634 --> 00:23:43.754
This is onto the right it.
344
00:23:44.314 --> 00:23:50.516
Then you may come back and be like, okay, I really need to come back and redo some of the core stuff.
345
00:23:50.576 --> 00:23:51.756
Or you even just rewrite it.
346
00:23:52.296 --> 00:23:56.039
And just now really put, you know, the mind into that motion.
347
00:23:56.059 --> 00:24:03.383
So there's still an element of, or at least there is still a world of creativity where you could still make something in this new age.
348
00:24:03.423 --> 00:24:03.884
Oh, I think so.
349
00:24:03.924 --> 00:24:04.444
For sure.
350
00:24:04.464 --> 00:24:05.044
For sure.
351
00:24:05.064 --> 00:24:13.070
I think, you know, to the degree that we have good APIs and good documentation because we want people to integrate with you.
352
00:24:13.090 --> 00:24:15.271
Like, I think all the new stuff still needs to have.
353
00:24:15.948 --> 00:24:17.329
all of those pieces in place.
354
00:24:17.369 --> 00:24:19.550
Meaning people like to integrate with things that work well.
355
00:24:20.370 --> 00:24:24.832
And so I think there is a creative component to that because there's no right or wrong.
356
00:24:25.353 --> 00:24:31.936
It's how you even architect some of these APIs or documentation or UX or the flow and all that kind of stuff.
357
00:24:31.956 --> 00:24:39.600
So there's still that piece where, yeah, it's as low level as the API or as high level as how UX works.
358
00:24:39.680 --> 00:24:42.441
I think there's certainly creativity involved.
359
00:24:43.558 --> 00:24:46.380
Let's go back to the university story.
360
00:24:46.440 --> 00:24:49.741
So
at what point was that inflection point?
361
00:24:49.801 --> 00:24:57.846
At what point did you know, whether it's the university journey or whether it's being at Google, at what point did you know, I'm going to be a founder?
362
00:24:57.866 --> 00:25:00.928
I'm going to be at least start thinking about being part of a startup.
363
00:25:01.208 --> 00:25:06.010
I think when I started in my first company, so I will rewind to my Google days.
364
00:25:06.050 --> 00:25:07.111
So at that point.
365
00:25:07.863 --> 00:25:12.447
You know, a colleague of mine said, okay, there's, you know, Sutter Hill is starting a new company.
366
00:25:12.567 --> 00:25:19.614
It's in the financial market space and it's going to revolutionize, you know, Fed funds, time deposits, repo.
367
00:25:20.715 --> 00:25:26.140
And I don't know anything about financial markets, let alone Fed funds or time deposits or all that.
368
00:25:26.280 --> 00:25:26.500
Right.
369
00:25:26.881 --> 00:25:29.123
But I just feel like, oh, but you could build a new exchange.
370
00:25:29.787 --> 00:25:36.512
And then the things just start to go in your mind and be like, okay, so exchange all the big exchanges I know are like mainframes, right?
371
00:25:36.652 --> 00:25:39.133
Okay, but you don't probably want to be voting on tandem.
372
00:25:39.474 --> 00:25:42.315
So you'd be like, okay, so are we going to do this all in memory?
373
00:25:42.375 --> 00:25:44.057
Like how are you going to do your transaction logs and stuff?
374
00:25:44.097 --> 00:25:45.157
Are you going to do it on database?
375
00:25:45.397 --> 00:25:48.219
So that was more of the obsession, I think.
376
00:25:48.820 --> 00:25:52.983
And then, okay, so let me quit my Google job and do all that.
377
00:25:53.023 --> 00:25:54.444
And of course, there's a lot of thinking when I do that.
378
00:25:54.484 --> 00:25:57.446
And I always go back to like two stories.
379
00:25:57.526 --> 00:25:58.006
One story is,
380
00:25:58.602 --> 00:26:02.485
I told my SRE at that point at Google, I'm quitting.
381
00:26:02.985 --> 00:26:03.866
And he was like, oh.
382
00:26:04.206 --> 00:26:05.748
And I was like, oh, I'm joining a startup.
383
00:26:05.768 --> 00:26:09.250
And he'll be like, oh, so you'll be back soon.
384
00:26:09.430 --> 00:26:11.592
And I was like, okay.
385
00:26:12.093 --> 00:26:14.274
And then I think I remember telling my director that I'm leaving.
386
00:26:14.514 --> 00:26:18.838
And she told me like, okay, if you want to come back, you got to give me two days.
387
00:26:19.952 --> 00:26:23.597
Because one day you can't come back the next day because HR doesn't work that way.
388
00:26:23.757 --> 00:26:24.758
So I'm like, okay.
389
00:26:24.778 --> 00:26:26.601
So there's like, to me, like there's no risk.
390
00:26:27.001 --> 00:26:27.502
Right.
391
00:26:27.902 --> 00:26:29.785
Like the job is guaranteed to come back to.
392
00:26:30.105 --> 00:26:30.325
Yeah.
393
00:26:30.345 --> 00:26:35.051
Of course you lose, you know, some of the upside of being working, continuing to work at Google.
394
00:26:35.071 --> 00:26:35.192
Yeah.
395
00:26:35.812 --> 00:26:39.936
But it just feels like, oh, there's this really cool problem to solve.
396
00:26:40.176 --> 00:26:43.479
And I'm not even thinking, is there a market that TAM brought market?
397
00:26:43.559 --> 00:26:44.619
I'm not even thinking of any of that.
398
00:26:44.639 --> 00:26:46.341
Were these even ideas in your mind?
399
00:26:46.361 --> 00:26:48.423
Did you think about the business end of startups?
400
00:26:49.003 --> 00:26:49.724
No.
401
00:26:50.044 --> 00:26:52.406
I think it just seemed like it was a really cool problem.
402
00:26:52.646 --> 00:26:55.307
So it still goes back to the idea of obsession and making something.
403
00:26:55.407 --> 00:26:56.107
Yeah.
404
00:26:56.127 --> 00:26:59.028
Because, and not that I want to get personal, but you're building a house.
405
00:26:59.408 --> 00:26:59.588
Yeah.
406
00:26:59.608 --> 00:26:59.748
Right.
407
00:26:59.788 --> 00:27:00.949
It's again, making something.
408
00:27:00.989 --> 00:27:04.770
Like everything is about that obsession to see something turn into something.
409
00:27:04.950 --> 00:27:05.290
Yeah.
410
00:27:05.430 --> 00:27:05.650
Yeah.
411
00:27:05.830 --> 00:27:07.751
And I think it is something universal.
412
00:27:07.951 --> 00:27:08.211
Sure.
413
00:27:08.911 --> 00:27:18.614
I mean, not to disparage like some professions, but like there are obviously some professions that are there that perhaps are more about supporting, right?
414
00:27:18.834 --> 00:27:21.895
Or, you know, the currency in modern society.
415
00:27:22.306 --> 00:27:26.171
But there are many professions that are, you know, they're making like making food, making art, making music.
416
00:27:26.211 --> 00:27:32.640
You know, I think I always think programming is very much close to more of an art form than to like a science.
417
00:27:32.660 --> 00:27:37.366
Because you ask like 20 different people to do the same thing and they all have like a different.
418
00:27:37.750 --> 00:27:39.271
Well, you ask the same AI to do the same thing.
419
00:27:39.291 --> 00:27:43.415
Like, you know, today and tomorrow, like it will probably give you a whole different architecture.
420
00:27:43.895 --> 00:27:45.817
So it's like, so what do you decide?
421
00:27:45.877 --> 00:27:46.838
Like, do you want to put it this way?
422
00:27:46.858 --> 00:27:49.180
Do you want to, you know, is it going to be running on this architecture?
423
00:27:49.200 --> 00:27:50.481
Does it have this API?
424
00:27:50.721 --> 00:27:51.542
Does it have this flow?
425
00:27:51.562 --> 00:27:53.603
So it's like a very creative process.
426
00:27:53.623 --> 00:28:05.233
So I think it is a universal thing where I think some people, certainly I think, or even most people, like if I give kind of a one to my wife and she's not a coder, right?
427
00:28:05.670 --> 00:28:07.531
and says, okay, well, you
could build anything you want.
428
00:28:08.251 --> 00:28:09.491
She has a lot of ideas, right?
429
00:28:09.511 --> 00:28:15.913
Like, you know, whether it's for travel, whether it's for cooking, whether it's for like, whatever it is, like she has a lot of ideas.
430
00:28:16.133 --> 00:28:19.474
And I feel, yeah, I think this is universal.
431
00:28:19.494 --> 00:28:25.115
So for me, I think when I was at Google, like I, and even at school, like I really wanted to build something.
432
00:28:25.155 --> 00:28:35.218
And so the allure of building exchange just seemed, like I wanted to understand, like, I think that's one of the draws to leaving like a big company is,
433
00:28:36.406 --> 00:28:39.188
you tend to specialize, right, like in big companies.
434
00:28:39.328 --> 00:28:45.892
So you get visibility into just a very small, like your surface is just very limited in terms of like your inputs and outputs.
435
00:28:46.312 --> 00:29:00.021
But in a startup, like you're basically, you know, deciding the database, the front end, building the client, building the back end, building the office, right, how do you want the process to work, you know, the code reviews and all that kind of stuff, right?
436
00:29:00.041 --> 00:29:00.902
And so I think it just...
437
00:29:01.928 --> 00:29:05.112
Yeah, that I think then piqued my interest.
438
00:29:05.272 --> 00:29:14.924
I remember the first time going to the startup, like meaning starting, I left Google on a Friday, joined the startup on Monday, and then we went and be like, oh, we need to go to Apple Store, right?
439
00:29:14.944 --> 00:29:16.846
Because we all, you need to buy a laptop, right?
440
00:29:16.866 --> 00:29:17.507
So nobody is...
441
00:29:17.940 --> 00:29:18.561
no IT, right?
442
00:29:18.581 --> 00:29:20.883
So nobody's giving you a laptop and setting up your desk.
443
00:29:20.923 --> 00:29:30.152
You go to Apple Store and then we went back to the office to be like, oh, we now have three people and it's not kind of easy to all be like looking at the same monitor.
444
00:29:30.172 --> 00:29:33.575
I guess we're not, you know, we have big bodies.
445
00:29:33.935 --> 00:29:36.758
So we'd be like, okay, let's go to Staples and buy a projector, right?
446
00:29:37.278 --> 00:29:37.659
So then it,
447
00:29:39.242 --> 00:29:42.803
But imagine that mindset, but it's like with coding and building stuff, right?
448
00:29:42.823 --> 00:29:43.063
Sure.
449
00:29:43.083 --> 00:29:48.204
Like you're like thinking like what's the most scrappy way to build something, to test it?
450
00:29:48.724 --> 00:29:49.945
And everything is just a time.
451
00:29:49.965 --> 00:29:50.965
But it's so much fun, right?
452
00:29:51.025 --> 00:29:51.885
It's so much fun, right?
453
00:29:52.085 --> 00:30:02.148
And so – but of course the reality of whether your things – the thing that you built actually is – you know, has probably market fit, right?
454
00:30:02.168 --> 00:30:03.608
The market is big enough.
455
00:30:03.908 --> 00:30:05.409
eventually we'll set in, right?
456
00:30:05.869 --> 00:30:14.074
So for us, I think we built the exchange and three years later, I remember like we were going to launch the exchanges for big banks, right?
457
00:30:14.094 --> 00:30:21.918
So like the Bank of America, Chase, I think like all the big banks had like letter of intent to say like they would trade on our platform.
458
00:30:22.658 --> 00:30:25.860
And we said, okay, this is the go live date.
459
00:30:26.180 --> 00:30:28.021
So make sure it runs on your browser, right?
460
00:30:28.081 --> 00:30:28.962
All that kind of stuff, right?
461
00:30:29.022 --> 00:30:32.063
We were running on Flash, the front end, running in Internet Explorer on Flash.
462
00:30:32.283 --> 00:30:33.524
That's how long ago this was.
463
00:30:34.140 --> 00:30:40.001
And so they all got there, they got their logins, they set up all their centers, their currencies, all the credit lines, all kinds of stuff.
464
00:30:40.301 --> 00:30:49.644
And I remember my CEO was like, you, Clement, probably need to stay in Palo Alto in a hotel for the whole week because trading starts around, I guess, 5 over here.
465
00:30:49.984 --> 00:30:54.365
Just make sure you're ready in case you support stuff, support the trades that happen.
466
00:30:54.625 --> 00:30:54.885
Right.
467
00:30:55.362 --> 00:30:57.723
And so I still have photos from those days, right?
468
00:30:57.743 --> 00:30:59.524
Because like Palo Alto downtown is nice at night.
469
00:30:59.604 --> 00:31:03.106
So I remember like waking up and be like, okay, now's the go live date.
470
00:31:03.326 --> 00:31:04.567
We're going to start have trades.
471
00:31:05.407 --> 00:31:06.048
And there were no trades.
472
00:31:06.748 --> 00:31:08.049
Like none at all.
473
00:31:08.229 --> 00:31:09.730
Like on the first day.
474
00:31:09.890 --> 00:31:11.591
And I think for the first week, we had one trade.
475
00:31:12.471 --> 00:31:18.114
And for a product that was designed to replace voice brokers, the voice brokers would
476
00:31:18.835 --> 00:31:26.159
would do a better job than us, certainly for one trade, and certainly less money to have that infrastructure.
477
00:31:26.579 --> 00:31:29.080
So we actually thought it was the wrong thing to build.
478
00:31:29.860 --> 00:31:31.641
So I think that is the moment then dawned on me.
479
00:31:31.661 --> 00:31:32.362
It was like, okay.
480
00:31:32.442 --> 00:31:35.103
PMF sells in and you start thinking about the business side.
481
00:31:35.443 --> 00:31:35.863
Right.
482
00:31:35.943 --> 00:31:37.804
And then be like, okay, so now we want to figure out this.
483
00:31:37.824 --> 00:31:39.845
So we have all these letter of intent, right?
484
00:31:39.865 --> 00:31:41.526
Like it has a J.B. Morgan and stuff like that, right?
485
00:31:41.546 --> 00:31:43.207
So isn't that going to be...
486
00:31:43.812 --> 00:32:08.952
it right like they it's not right so just start like asking more
questions i know like uh business or or execs like you know i mean i'm a founding co i'm a founding engineer but like don't let don't want to ask just like i don't want somebody to ask me like why are you using like my sequel like why are you using like xa transactions uh with jboss or really i don't want people to ask me a question and having to defend but now i'm starting to ask more questions and be like okay like we built all this like why isn't it happening and all that
487
00:32:09.712 --> 00:32:13.254
And we decided at that time, like, maybe this is the wrong thing to build.
488
00:32:13.294 --> 00:32:14.814
We thought, like, that's maybe the wrong market.
489
00:32:14.834 --> 00:32:17.215
Maybe the market doesn't exist and all that.
490
00:32:18.075 --> 00:32:26.539
And so I thought, going back to the story, like, I could pull my two-day card back to Google and we could all go back to where it came from.
491
00:32:27.359 --> 00:32:31.781
Sutter Hill at that time said, no, no, let's just, you folks could build things.
492
00:32:32.181 --> 00:32:36.663
And so let's just write you a check and you could just continue to build whatever you want.
493
00:32:37.163 --> 00:32:38.704
I didn't realize it was an SHP company.
494
00:32:39.107 --> 00:32:40.408
The first one and second one, yeah.
495
00:32:40.428 --> 00:32:41.109
Both of them, yeah.
496
00:32:41.249 --> 00:32:48.675
So, of course, there are other rounds, but the seed funding for both my first and second company come from Sudhiral.
497
00:32:49.055 --> 00:32:51.237
And you were working out of their office, I assume?
498
00:32:51.597 --> 00:32:54.920
So the first company, we did work out of their office a little bit.
499
00:32:55.461 --> 00:32:56.181
Only very briefly.
500
00:32:56.201 --> 00:32:59.124
I think we got a Regis office very soon.
501
00:32:59.164 --> 00:33:02.967
Like, this is before we worked, so the Regis office have no windows.
502
00:33:02.987 --> 00:33:04.588
So no bar or anything.
503
00:33:04.909 --> 00:33:07.711
So the second company worked out of the first company,
504
00:33:08.216 --> 00:33:09.859
because both other companies.
505
00:33:09.879 --> 00:33:14.667
So initially we worked out there, eventually we moved again to another Regis location.
506
00:33:15.468 --> 00:33:18.173
But yeah, they said, okay, don't go back to Google.
507
00:33:18.193 --> 00:33:21.298
We'll just give you some money and you can work out an idea together.
508
00:33:22.282 --> 00:33:31.888
But the first company, the CEO, Glenn, bless his soul, like he just kept at it and it kept saying, like, eventually we'll figure this out.
509
00:33:31.988 --> 00:33:35.930
Eventually interest rates will pick up and these markets will heat up.
510
00:33:36.310 --> 00:33:38.311
And he continued to do that.
511
00:33:38.352 --> 00:33:48.417
We started the company in 2010, left in 2013, thinking again, with my puny engineering mind, that it's not going to fly or we built the wrong thing.
512
00:33:48.537 -->
00:33:49.038
Build it right.
513
00:33:49.818 --> 00:33:50.418
Built the wrong thing.
514
00:33:51.079 --> 00:33:51.299
Yeah.
515
00:33:52.547 --> 00:33:55.108
But now it's a trillion dollar a day exchange.
516
00:33:55.409 --> 00:33:55.629
Wow.
517
00:33:55.869 --> 00:34:04.654
So, like, we're getting checks from GLMX and like I hold it and I'm like, this is crazy.
518
00:34:04.674 --> 00:34:04.754
Wow.
519
00:34:04.774 --> 00:34:07.355
Like 17 years, no, 15 years later or 14 years later.
520
00:34:09.597 --> 00:34:16.799
There's a timing component to product market fit, of course, that maybe it's more obvious in certain markets.
521
00:34:17.360 --> 00:34:22.101
But yeah, you just persevered, you know, hopefully replaced the flash front end.
522
00:34:22.321 --> 00:34:25.342
And now they're very successful as a company.
523
00:34:25.783 --> 00:34:28.183
And I think that speaks to equity, of course, right?
524
00:34:28.243 --> 00:34:32.925
Meaning you own something that will potentially pay out.
525
00:34:33.432 --> 00:34:36.414
This is almost like a wild bet at that point of what it is.
526
00:34:36.514 --> 00:34:36.694
Right.
527
00:34:37.075 --> 00:34:48.243
But I think the idea that, oh, well, there's like product market fit, like we do need to pay attention to the market did bleed over then to the second company, which like we started to ideate and think about, okay, what do we really want to build?
528
00:34:48.895 --> 00:34:55.357
And we just, I remember like every week we would just sit around and be like, okay, let's all like, you know, come up with crazy ideas to try.
529
00:34:55.377 --> 00:34:56.718
And we did try them, right?
530
00:34:56.738 --> 00:35:01.259
So we were like, okay, maybe everybody could share their searches on Google.
531
00:35:01.279 --> 00:35:06.741
So we built like a Chrome plugin and be like, you see your social friends, like what they're searching for and what links they clicked on.
532
00:35:06.981 --> 00:35:09.262
Maybe there's a little privacy.
533
00:35:10.303 --> 00:35:13.684
But we did that and then we send it out to people.
534
00:35:13.724 --> 00:35:17.345
You know, we build like kind of FanDuel and that kind of like real-time business.
535
00:35:17.998 --> 00:35:20.220
betting market for sports, if you will.
536
00:35:20.420 --> 00:35:24.803
And we built it and we actually had it run in like sports bars and whatnot.
537
00:35:24.823 --> 00:35:25.984
And so people could win stuff.
538
00:35:26.404 --> 00:35:27.165
So we tried that.
539
00:35:27.205 --> 00:35:30.888
And I am not a big sports fan.
540
00:35:30.988 --> 00:35:34.850
So after building it and we were like, oh, let's go to like AT&T Park back then.
541
00:35:34.870 --> 00:35:35.611
AT&T Park.
542
00:35:35.811 --> 00:35:37.372
Like back then AT&T Park.
543
00:35:37.693 --> 00:35:39.254
And just went to AT&T Park.
544
00:35:39.734 --> 00:35:41.776
And I remember like using our app to play with it.
545
00:35:41.936 --> 00:35:44.798
And like I don't understand baseball that much.
546
00:35:45.218 --> 00:35:46.299
And the sun was really bright.
547
00:35:46.861 --> 00:35:51.646
And I'm always missing the questions because I'm not like, you know, I'm not paying attention to the phone, the phone keeps on chatting off.
548
00:35:52.146 --> 00:35:55.129
And I went back and I was like, no, this is not, not for me.
549
00:35:55.169 --> 00:36:01.615
Like, I'm sure there might be something and I'm sure like, we discovered like, DraftKings and FanDuals are things.
550
00:36:02.496 --> 00:36:03.357
But it was not for us.
551
00:36:03.417 --> 00:36:04.778
And, and then,
552
00:36:05.373 --> 00:36:17.628
Wavefront, which is my second company, kind of came to be because when I left Google and I went to my first company, at Google, I had a lot of obsession about how Google manages their production infrastructure.
553
00:36:17.648 --> 00:36:21.653
So there's Borg and there's Borgmon back then at Google.
554
00:36:22.053 --> 00:36:23.714
And when I left Google, there's no Borgmon.
555
00:36:23.935 --> 00:36:24.795
There's no Borg, right?
556
00:36:24.815 --> 00:36:27.477
Because we're racking stuff in Equinix.
557
00:36:27.978 --> 00:36:30.900
So I said, okay, well, there must be like a better thing.
558
00:36:30.920 --> 00:36:32.862
And people say, okay, but there's Splunk.
559
00:36:32.942 --> 00:36:35.304
Splunk nowadays is a lot more advanced than Splunk back then.
560
00:36:35.344 --> 00:36:36.945
So, but Splunk is this log thing.
561
00:36:37.065 --> 00:36:39.607
So I'd be like, okay, so, okay, I could do logs everywhere.
562
00:36:39.627 --> 00:36:41.328
And it's very slow.
563
00:36:41.388 --> 00:36:42.229
It's very expensive.
564
00:36:42.565 --> 00:36:48.933
it's on-prem, so you're like getting a license, but not really, like you need a license to use your thing, to use your disk space.
565
00:36:49.834 --> 00:36:56.662
So we're like, no, like there must be a better thing, and Google has that better thing, and Google is not going to open source Borg and Borgmon, not even today.
566
00:36:56.683 --> 00:36:58.505
I guess Kubernetes came out of that, but
567
00:36:59.376 --> 00:37:05.780
So we said, okay, let's look at whether Borgmon or any other things in Google back then would be interesting things to build.
568
00:37:06.300 --> 00:37:10.202
And at the same time, I think Sam Pilara came out of Twitter as their chief architect.
569
00:37:10.242 --> 00:37:11.463
And he also joined Sutter Hill.
570
00:37:11.903 --> 00:37:12.784
And he kind of met with us.
571
00:37:12.844 --> 00:37:20.408
And he said, you need to build Viz, which was kind of the people who left Google, went to Twitter and said, you need Borgmon to actually slay the fail whale.
572
00:37:20.848 --> 00:37:23.890
And he said, look, that's the thing that's slayed the fail whale.
573
00:37:24.682 --> 00:37:29.223
And so if you build this and then become a SaaS platform, then a lot of people will probably want that.
574
00:37:29.863
--> 00:37:35.024
And so I think both of those things came together and we decided to build that, yeah.
575
00:37:35.724 --> 00:37:36.805
Oh, there's a side note as well.
576
00:37:37.185 --> 00:37:42.826
We also met Ben and Thierry from Snowflake like when they were just two founders.
577
00:37:43.406 --> 00:37:52.008
And I remember sitting together at the GLMX office because we were still thinking about what we would build, but we were working out of the GLMX office.
578
00:37:52.914 --> 00:37:57.637
And they came in and said, okay, we're going to build this company, Snowflake, and it's going to be a data warehouse in the cloud and all that kind of stuff.
579
00:37:57.657 --> 00:38:00.519
And we may join forces together and go do that.
580
00:38:00.659 --> 00:38:03.081
And well, of course, we didn't.
581
00:38:03.501 --> 00:38:16.770
And at that time, we sat down, the GLMX engineering crew sat down and be like, I don't think PNG or Tide or all these companies are going to move their data warehouse to the cloud.
582
00:38:16.850 --> 00:38:20.853
At that time, we were still just in that early journey to the cloud, and so we...
583
00:38:21.534 --> 00:38:24.996
Thought that wouldn't quite work, but obviously they proved us very wrong.
584
00:38:26.257 --> 00:38:29.379
But hey, we then started to do Wayfront.
585
00:38:29.459 --> 00:38:30.920
They started to do Snowflake.
586
00:38:30.940 --> 00:38:32.921
Snowflake became a Wayfront customer, right?
587
00:38:32.961 --> 00:38:34.682
And we do a little bit of Snowflake as well.
588
00:38:34.722 --> 00:38:36.983
But yeah, so I think it kind of turned out well.
589
00:38:37.163 --> 00:38:42.026
But yeah, but there was a small anecdote story there.
590
00:38:42.086 --> 00:38:46.449
Like if we've decided another way, then we would have a very different conversation as to...
591
00:38:47.309 --> 00:38:49.432
what transpired afterwards.
592
00:38:49.673 --> 00:38:51.776
That's so fascinating.
593
00:38:51.796 --> 00:38:57.504
So when you think about the Wayfront story, at that point, you had...
594
00:38:58.530 --> 00:38:59.750
it wasn't just an obsession anymore.
595
00:38:59.890 --> 00:39:01.691
You were thinking about the business side of things.
596
00:39:01.711 --> 00:39:05.852
You were thinking about figuring out whether what you're building has legs.
597
00:39:06.852 --> 00:39:08.532
What does that journey look like in the beginning?
598
00:39:08.852 --> 00:39:10.313
Are you speaking to customers?
599
00:39:10.353 --> 00:39:12.733
Are you, are you doing a lot of fact finding?
600
00:39:12.753 --> 00:39:19.275
What do you just go in and build and let, you know, be more of a product-led growth and see if it works out?
601
00:39:19.295 --> 00:39:21.055
Like what does that, how did it look away from?
602
00:39:22.028 --> 00:39:26.732
I think we're still a very much engineer-focused kind of company.
603
00:39:26.812 --> 00:39:30.975
So I don't think we had our first product manager until much later.
604
00:39:31.355 --> 00:39:43.164
So, of course, Dev, our CTO, my co-founder, he is amazing in terms of bridging the technical and the product and the use case and all that.
605
00:39:43.204 --> 00:39:45.806
So he's kind of the face of the company as we go,
606
00:39:46.283 --> 00:39:53.209
visit Yammer, right, visit Workday, visit Box, right, he would be able to kind of sell the vision and Borgmon and this and all that.
607
00:39:54.050 --> 00:40:04.619
So I actually, for a very long period of time, even at Wayfront, was kind of just I get the field reports and be like, okay, like, okay, this is like, this will, they're looking for this.
608
00:40:04.679 --> 00:40:07.101
If we have this, then they will buy this and they will sign this and blah, blah, blah.
609
00:40:07.121 --> 00:40:08.162
And so I'll be like, okay, cool.
610
00:40:08.182 --> 00:40:08.462
Yeah.
611
00:40:08.730 --> 00:40:11.231
And be like, oh, the next renewal comes and they want it to be cheaper.
612
00:40:11.251 --> 00:40:14.152
I'd be like, okay, so we need to get the margins to be like this and blah, blah, blah.
613
00:40:14.172 --> 00:40:16.312
And be like, okay, so let's actually work on like the margins.
614
00:40:17.132 --> 00:40:26.155
So I would still think I was more, because that's still not the exact thing that, I mean, they're not giving me the spec of how to make it better.
615
00:40:26.195 --> 00:40:28.515
It's more like, this is the customer's problem, right?
616
00:40:28.575 --> 00:40:30.216
Is it either too expensive sometimes, right?
617
00:40:30.276 --> 00:40:34.017
Or it's not fast enough or they want to, you know, scale 10 times the traffic.
618
00:40:34.077 --> 00:40:37.058
And, you know, Clement, like, do you want 10 times the traffic?
619
00:40:37.078 --> 00:40:37.858
Didn't you take it, right?
620
00:40:38.269 --> 00:40:43.775
And so all of that is just kind of that kind of product or almost business requirements that lands on my lap.
621
00:40:43.795 --> 00:40:46.798
And then I just think about like, okay, so how is this going to work?
622
00:40:46.818 --> 00:40:48.539
How is this going to be translated into code?
623
00:40:49.460 --> 00:40:56.688
But then over time, I think as the company grew, right, we have Dev and it goes to all of the companies, right?
624
00:40:56.928 --> 00:41:00.792
But there are going to be cases where, you know, he can't cover all of them and,
625
00:41:02.005 --> 00:41:09.988
And I think at that point, I think also Sam kind of like, okay, we got to give you a title and then you could just like, you could go to talk to the companies.
626
00:41:10.048 --> 00:41:12.729
And I was like, oh, I like, if you think that's a good idea.
627
00:41:12.749 --> 00:41:23.514
And I think that's where, because again, going back to my university days where I'm just this person like that stares at that computer screen, like don't even look at, don't even turn my head.
628
00:41:23.814 --> 00:41:26.975
And then I think entrepreneurship like really,
629
00:41:27.530 --> 00:41:32.575
you know, kind of force you to become a different person because like getting out of your comfort zone.
630
00:41:33.096 --> 00:41:36.600
And then you'll realize, actually, you have rapport, right?
631
00:41:36.620 --> 00:41:38.522
Because a lot of your customers are engineers.
632
00:41:39.223 --> 00:41:41.265
So they rather just talk to somebody.
633
00:41:41.345 --> 00:41:44.288
I remember like when I was visiting many places, right?
634
00:41:44.328 --> 00:41:45.109
Including like the
635
00:41:45.496 --> 00:41:48.217
The workdays, like the PayPals, the Reddit, the Lyft, right?
636
00:41:48.357 --> 00:41:49.537
You're talking to engineers, right?
637
00:41:49.557 --> 00:41:54.719
So they'd rather, you know, you tell them like, hey, this is actually really hard to do when you say you want this.
638
00:41:55.219 --> 00:41:58.600
And I could explain to them like why this is really hard because we sort of do it this way.
639
00:41:59.360 --> 00:42:04.062
And so if you want to do it this way, then if you want this particular feature to work, like this is actually really hard.
640
00:42:04.082 --> 00:42:10.744
But we could, you know, we could, I'm almost like ideating with them like live over there and be like, yeah, we'll be able to do that.
641
00:42:10.784 --> 00:42:13.905
And so people like get a sense like you're not just –
642
00:42:14.831 --> 00:42:15.771
you know, BSing, right?
643
00:42:15.811 --> 00:42:20.213
When you say, oh, we'll do this in the next release if you sign, usually when you sign, is for the next year.
644
00:42:21.414 --> 00:42:24.855
So I think that kind of feed into itself.
645
00:42:24.935 --> 00:42:34.359
I had many bad, not bad, but like not ideal customer meetings where I just go like way into the weeds on like a very technical issue.
646
00:42:35.359 --> 00:42:36.820
But I think, yeah, I think there...
647
00:42:37.493 --> 00:42:47.537
Because so many of these companies are engineer-led and we're building a DevOps product, if you will, or engineer R&D product, I think just having myself out there was helpful.
648
00:42:47.597 --> 00:42:57.980
And I think that helped me kind of realize my gaps, like how I could be a better communicator, more compassionate usually to people's problems.
649
00:42:58.060 --> 00:43:05.063
And yeah, so I think at Wavefront is where I slowly kind of, I think, grew out of my shell, if you will.
650
00:43:05.344 --> 00:43:15.789
As far as Wavefront is concerned and the role you were playing there, speaking to customers, I'm assuming they were design partners and they were giving you the feedback and you were iterating.
651
00:43:16.349 --> 00:43:27.295
But how do you know which feedback is great feedback that you need to do something about and when it's quite siloed to that one customer and perhaps might not be the right feedback?
652
00:43:28.635 --> 00:43:29.276
Yeah, I think...
653
00:43:30.515 --> 00:43:37.959
Well, I wouldn't say I'm also very
good at it either because oftentimes, like, as an engineer, like, I'm at a customer site and they show me, like, a very slow query.
654
00:43:38.099 --> 00:43:39.199
And I'm like, oh, crap.
655
00:43:39.219 --> 00:43:40.380
Like, I don't like this.
656
00:43:40.460 --> 00:43:41.260
I don't like it slow.
657
00:43:41.720 --> 00:43:42.241
Right.
658
00:43:42.581 --> 00:43:44.101
And honestly, like, maybe it's only for them.
659
00:43:44.121 --> 00:43:45.002
But that's your obsession side, right?
660
00:43:45.022 --> 00:43:45.822
Yeah, but that's, yeah.
661
00:43:46.062 --> 00:43:47.643
Yeah, I'm like, no, I'm going to make it fast.
662
00:43:47.683 --> 00:43:49.584
I'm going to make the, I'm going to make the query fast.
663
00:43:49.684 --> 00:43:51.345
And I'm just going to go back and I'm going to think about it.
664
00:43:51.985 --> 00:44:00.149
And I think, but I think that, you know, having said that, there are many of these cases where you just made something fast, um,
665
00:44:01.110 --> 00:44:08.654
There are, if it is slow enough that somebody's complaining to you, maybe only one out of ten, you know, of your users is actually complaining to you.
666
00:44:08.934 --> 00:44:09.154
Sure.
667
00:44:09.254 --> 00:44:10.475
Because you happen to visit them, maybe.
668
00:44:10.615 --> 00:44:15.798
Or because you happen to, like, serve the one to ten, the score, the NPS score.
669
00:44:16.618 --> 00:44:18.599
So you kind of know that because you gather that information.
670
00:44:19.219 --> 00:44:23.782
But if you actually make it better, like, people actually notice these things, right?
671
00:44:23.902 --> 00:44:25.863
And engineers are really bad at giving feedback, too.
672
00:44:25.903 --> 00:44:27.404
Like, they expect the thing to work.
673
00:44:27.444 --> 00:44:29.705
And if it's better, like, you'll be like, cool, yeah.
674
00:44:30.281 --> 00:44:30.522
Nice.
675
00:44:31.042 --> 00:44:33.666
I'm not going to come back and congratulate and be like, oh, nice.
676
00:44:33.706 --> 00:44:34.988
Like, this is like 10 seconds now.
677
00:44:35.028 --> 00:44:36.290
It's like they're never going to do that.
678
00:44:36.690 --> 00:44:39.634
So I think it just fed into that flywheel.
679
00:44:39.795 --> 00:44:42.298
So I wouldn't say I'm very good at that.
680
00:44:42.478 --> 00:44:46.985
But I think ultimately, like, you – of course, like, you're –
681
00:44:47.596 --> 00:44:57.962
If you're doing kind of either customer collected feedback and working from that or maybe from the other angle is like you're just deep diving into your own infrastructure and seeing like where things are spending time.
682
00:44:58.102 --> 00:45:07.348
So both of those like probably are informing your understanding of your infrastructure and where things are spending a lot of CPU cycles or disk IO.
683
00:45:07.928 --> 00:45:09.169
And you're just optimizing that.
684
00:45:09.249 --> 00:45:14.592
And maybe both of those come together and you realize, okay, this is what I want to do today and just make this better.
685
00:45:14.672 --> 00:45:17.854
And lo and behold, some customers get really happy.
686
00:45:19.855 --> 00:45:36.304
Honestly, I remember back after we got acquired by VMware, Wavefront got acquired by VMware, I did some of the automated schema management stuff to basically reshape the data adaptively based on the incoming data source and to accelerate essentially query execution speeds.
687
00:45:36.364 --> 00:45:36.424
And
688
00:45:37.104 --> 00:45:46.713
And the PM actually went out to, I believe one of the customers, I forgot who they were, and they came back and be like, whatever you did in the last quarter, right, like, all the features don't matter.
689
00:45:46.913 --> 00:45:50.296
Like, it was, it was like this thing that just made everything fast.
690
00:45:50.336 --> 00:45:50.636
Wow.
691
00:45:51.056 --> 00:45:54.299
And it just, like, that's the thing, like, they didn't ask for that, right?
692
00:45:54.359 --> 00:45:55.981
But, but it was like, that's the thing.
693
00:45:56.501 --> 00:45:57.242
And my PM was like,
694
00:45:57.730 --> 00:45:59.191
Yeah, you do whatever you want.
695
00:46:00.231 --> 00:46:03.473
Yeah, because of course, like, of course, there are other features that matter.
696
00:46:03.513 --> 00:46:09.917
But he's just like, but the customer, like, really noticed, like, because once something is like, I think Paul Graham said that, right?
697
00:46:09.937 --> 00:46:11.798
When something is like 10 times better, like people notice, right?
698
00:46:11.818 --> 00:46:12.438
So if your latency...
699
00:46:13.403 --> 00:46:17.467
Yeah, all the tail latencies, like, you could, like, bring that up by 10 times, and people will notice that.
700
00:46:18.168 --> 00:46:32.020
And at the end of the day, I think we spend, well, depending on the product, but I think we don't spend enough time, like, probably industry-wide, to think about those frustration points and just, yeah, accelerate latency or accelerate, like, cost or, like, reduce cost, meaning.
701
00:46:32.441 --> 00:46:37.465
And just do, and because customers will notice, they don't, they won't ask you oftentimes.
702
00:46:37.725 --> 00:46:39.447
But you do that, and people will be like, wow.
703
00:46:40.565 --> 00:46:46.530
It kind of makes me think, right, so when you think about the customers and the ones you worked with at Vaifront, what did they care about the most?
704
00:46:46.550 --> 00:46:48.251
Was it the UI, how it all looked?
705
00:46:48.311 --> 00:46:49.913
Was it the speed, the latency?
706
00:46:49.973 --> 00:46:52.815
Was it, you know, another feature?
707
00:46:53.736 --> 00:46:56.378
Is less more, you know, less features, but they do better?
708
00:46:56.558 --> 00:46:58.340
What were the key things that you could think of?
709
00:46:58.920 --> 00:47:02.043
Yeah, I think, interestingly, maybe similar to Google.
710
00:47:02.083 --> 00:47:06.446
Google has a search bar that has
nothing else, right, when all the other things have a portal, right?
711
00:47:06.686 --> 00:47:06.987
Mm-hmm.
712
00:47:07.781 --> 00:47:11.805
like the request that we get the most from Wavefront was, give me the blank page.
713
00:47:12.085 --> 00:47:13.847
Like, don't even show me graphs or anything.
714
00:47:13.867 --> 00:47:17.912
Like, just let me type, like just let me type the query and then I just want to do these deep dive explorations.
715
00:47:18.032 --> 00:47:18.512
Right.
716
00:47:18.552 --> 00:47:28.903
So, yeah, so I think there are some similar themes to that where, you know, if you're working with, well, I guess Google is not with technical people, but like technical audience, but like just in general,
717
00:47:29.415 --> 00:47:32.056
you give people the feeling that this is in control.
718
00:47:32.296 --> 00:47:33.177
Yeah, they're in control.
719
00:47:33.617 --> 00:47:36.438
There's like just this vast engine behind you.
720
00:47:36.478 --> 00:47:37.339
It's crazy.
721
00:47:37.719 --> 00:47:39.660
Maybe even if that feels like ChatGPT too, right?
722
00:47:39.680 --> 00:47:40.500
But I was thinking about that.
723
00:47:40.540 --> 00:47:47.003
I was thinking like when I look up Google or ChatGPT, it just, so what do you want to do today?
724
00:47:47.243 --> 00:47:47.924
Kind of question.
725
00:47:48.104 --> 00:47:48.684
And you just type.
726
00:47:49.144 --> 00:47:49.284
Yep.
727
00:47:49.604 --> 00:47:49.945
Simple.
728
00:47:51.079 --> 00:47:58.823
Yeah, it's not forcing you to like fit your question or your problem into like some drag and drop or some like pipelining thing, right?
729
00:47:58.843 --> 00:48:00.644
You just like sit there and be like, okay, let me start typing.
730
00:48:01.184 --> 00:48:05.166
And you start typing and this iteration process, like chat, of course, conversational process.
731
00:48:05.186 --> 00:48:09.388
But I think there's like a, it's like a, it's a smell, right?
732
00:48:09.408 --> 00:48:18.473
I think with product design where generally like you want people to have this feeling of this immense power, this immense possibility.
733
00:48:19.145 --> 00:48:19.385
Right.
734
00:48:19.645 --> 00:48:21.186
And I think for Wayfram, we have that.
735
00:48:21.926 --> 00:48:24.427
And people just...
736
00:48:24.767 --> 00:48:27.969
I don't think all of them used that power before they bought it.
737
00:48:28.369 --> 00:48:35.952
But the fact that not me, but our sales folks and Dev mostly demonstrated that power, they'd be like, oh, this is cool.
738
00:48:35.972 --> 00:48:42.635
So that means if someday I need this to do this and that and this and that and all that, this platform will be able to do that.
739
00:48:42.695 --> 00:48:43.055
I didn't just
740
00:48:43.479 --> 00:48:47.342
told you everything that is possible and it just checked all your boxes.
741
00:48:47.542 --> 00:48:51.765
There were like many boxes that potentially in the future you could be using the platform for.
742
00:48:51.785 --> 00:48:54.927
And I think that was the thing that I think unlocked...
743
00:48:55.385 --> 00:48:56.206
A lot of customers.
744
00:48:56.846 --> 00:48:57.687
And of course the scale, right?
745
00:48:57.807 --> 00:48:58.468
Of course the cost.
746
00:48:58.508 --> 00:49:00.009
Yeah.
747
00:49:00.049 --> 00:49:08.635
One last question I want to ask you about Wavefront and then I'd like to dive into your journey since Wavefront, since the acquisition, being an investor, being an advisor.
748
00:49:08.656 --> 00:49:10.177
But the last question I want to ask you about Wavefront.
749
00:49:10.737 --> 00:49:16.782
So Clement, if you were to rebuild Wavefront today with AI native tools, what would you do differently?
750
00:49:16.802 --> 00:49:16.902
Yeah.
751
00:49:17.565 --> 00:49:21.848
I think these days, like there are a lot of open source tooling and observability.
752
00:49:22.249 --> 00:49:25.731
So I would definitely build it open source or I would say open core.
753
00:49:26.452 --> 00:49:44.466
So you would want, you want people to understand because beyond what I just said about maybe that, you know, the, the, the sense of power, the, you know, scalability, the cost, all that people do want these days to understand, like you don't own all of their data and there's no way for you, for you.
754
00:49:45.187 --> 00:49:45.567
There's not,
755
00:49:46.058 --> 00:49:51.202
Not that you can't just send all the data back, but they would need, you know, a ginormous amount of infrastructure to catch it.
756
00:49:51.683 --> 00:49:52.083
Right.
757
00:49:52.544 --> 00:49:56.647
Or you just egress all of your kind of bits and help them how to interpret them.
758
00:49:56.687 --> 00:50:07.136
But people actually want to understand, like, hey, even though you're a SaaS platform and an observatory platform, like, you actually don't own the data in a particular form that they can't really get back.
759
00:50:08.140 --> 00:50:13.224
So if I were to build it today, I would probably build it open, open core, run it as a SaaS service, right?
760
00:50:13.284 --> 00:50:15.265
So we could be the best person to run it, right?
761
00:50:15.325 --> 00:50:16.406
At the lowest cost possible.
762
00:50:16.426 --> 00:50:18.467
You don't have to think about HA, high availability.
763
00:50:18.608 --> 00:50:23.871
You have somebody to call if your queries are slow, you have integration problems, you could call someone to support or all that.
764
00:50:24.292 --> 00:50:30.136
But also understanding that at some point, if you were to run it in an air gap situation or you want to run it locally, you want to run it in a...
765
00:50:30.756 --> 00:50:32.297
If you want to try it, right?
766
00:50:32.337 --> 00:50:33.578
You could run it locally.
767
00:50:33.938 --> 00:50:36.400
And if the SaaS platform is any good, then...
768
00:50:36.727 --> 00:50:42.588
at least we could demonstrate to you that, hey, we could in fact run it better than you because we probably know the code base better than you do.
769
00:50:43.169 --> 00:50:52.831
But I would, that feeling, I think it is starting to shift where people feel like, hey, it's better if we own the data, right?
770
00:50:52.871 --> 00:50:54.151
In observability especially.
771
00:50:54.171 --> 00:51:00.253
And you're this, maybe you're indexed in some way and you're able to still have the same power
772
00:51:00.906 --> 00:51:05.892
But if I were able to look at the data myself, I have the ability to do that.
773
00:51:05.952 --> 00:51:08.976
I actually have the data in my possession.
774
00:51:09.777 --> 00:51:13.461
I think that would be my technical answer to how I would build it differently.
775
00:51:13.561 --> 00:51:14.382
I don't know if that's it.
776
00:51:15.103 --> 00:51:15.244
Yeah.
777
00:51:16.605 --> 00:51:22.393
It is definitely, I guess, the concern in the market right now is the sharing of this data.
778
00:51:22.413 --> 00:51:22.573
Right.
779
00:51:23.394 --> 00:51:28.461
And there are companies that are finding new ways to solve that problem.
780
00:51:28.761 --> 00:51:28.901
Yep.
781
00:51:29.081 --> 00:51:31.064
With that in consideration, shifting to...
782
00:51:32.100 --> 00:51:45.705
you being an investor, you being an advisor, since the acquisition of Wayfront, which is a phenomenal success, you've been, you know, you had so much successes, especially, you know, starting off at a company like Google and having multiple startups.
783
00:51:46.125 --> 00:51:58.649
Now you're an SHV, sorry, a ER at SHV, again, SHV powerhouse, like you're keeping that, you know, staying with the same family, which is incredible, because I know how amazing they are.
784
00:52:00.269 --> 00:52:06.591
But before you start this journey again, thinking about starting your next company, you've been investing, you've been advising.
785
00:52:08.091 --> 00:52:11.812
As an advisor and as an investor, what are you seeing?
786
00:52:12.592 --> 00:52:13.753
What gets you most excited?
787
00:52:14.633 --> 00:52:21.334
What are the sort of ideas and problem statements that you're seeing in the market today that gets you really excited to write a check?
788
00:52:21.575 --> 00:52:23.295
That's a great question because I think...
789
00:52:24.172 --> 00:52:32.839
as an angel investor, after the Wayfront story, I didn't sit down and... Because I think, I guess, most funds would have, like, a mandate or a thesis or something, right?
790
00:52:32.879 --> 00:52:33.700
I didn't have that.
791
00:52:33.880 --> 00:52:37.683
So I sat down and, like, people started coming, right?
792
00:52:37.703 --> 00:52:40.205
Like, because they need a seat check and all that.
793
00:52:41.066 --> 00:52:43.368
And so it was, like, more of, like, friends of friends, right?
794
00:52:43.388 --> 00:52:46.450
Like, so you'd meet with them, and then you'd be like, okay, this is a good idea.
795
00:52:46.470 --> 00:52:47.611
Does it feel right?
796
00:52:47.651 --> 00:52:50.253
Like,
you know kind of what their backgrounds are and whatnot.
797
00:52:50.613 --> 00:52:51.234
And, um...
798
00:52:53.690 --> 00:52:57.851
I would go, and I did that for a couple months, hopefully, maybe years, maybe too long.
799
00:52:58.372 --> 00:53:04.933
And Sam kind of one day told me, like, he just told me point blank, Sam being from Sutter Hill, like he said.
800
00:53:05.113 --> 00:53:05.674
Sam Pilaro.
801
00:53:05.754 --> 00:53:06.074
Pilaro.
802
00:53:06.354 --> 00:53:11.595
And he said, Clement, you'd be like a really bad VC because you're really good at coding.
803
00:53:12.155 --> 00:53:16.256
So you, if somebody tells you there's a problem, you'd be like, okay, well, this is like how you would solve it, right?
804
00:53:16.276 --> 00:53:17.497
Like this is like blah, blah, blah, blah, blah.
805
00:53:17.517 --> 00:53:17.697
Okay.
806
00:53:17.797 --> 00:53:18.597
Then you would do that.
807
00:53:19.120 --> 00:53:20.081
maybe that's the wrong answer.
808
00:53:20.101 --> 00:53:24.245
So, but like you would, in your mind think, okay, this is the steps you would take to solve the problem.
809
00:53:24.586 --> 00:53:26.187
And so it'd be like, okay, so you need a hundred K to do it.
810
00:53:26.207 --> 00:53:26.508
Oh, cool.
811
00:53:26.548 --> 00:53:26.768
Nice.
812
00:53:26.808 --> 00:53:27.929
Like you, that's a real problem.
813
00:53:27.969 --> 00:53:29.511
I think you could solve it and then you could write a check.
814
00:53:30.051 --> 00:53:34.776
And then Sam would, Sam would told me, but then a year later you would go visit the company and be like, they did nothing like that.
815
00:53:34.896 --> 00:53:36.878
They, they did, they may have pivoted already.
816
00:53:37.118 --> 00:53:37.439
Sure.
817
00:53:37.459 --> 00:53:39.581
Or if they're still in the same problem, you would, you would discover that,
818
00:53:40.280 --> 00:53:42.021
oh, you did that.
819
00:53:42.101 --> 00:53:45.122
And that happens even at big companies, right?
820
00:53:45.142 --> 00:53:47.342
Like I'm at VMware for- So you're making a great advisor.
821
00:53:47.922 --> 00:53:50.903
I probably could be a better advisor than investor, yes.
822
00:53:51.003 --> 00:53:56.625
Because I think I would misjudge the ability, because you have no control.
823
00:53:56.645 --> 00:54:02.687
Like you're not going to be there and just saying like, hey, you know, do this or look into this piece of technology.
824
00:54:02.707 --> 00:54:05.508
Go to this cloud and oh no, like we've tried this.
825
00:54:05.568 --> 00:54:08.149
Like maybe personally, but or other.
826
00:54:08.565 --> 00:54:10.606
folks have tried this, like that doesn't smell right.
827
00:54:10.646 --> 00:54:11.146
And all that.
828
00:54:11.166 --> 00:54:21.030
So I think I, and then I realized, okay, maybe I should, I should like step back a little bit from the directly investing at companies route.
829
00:54:21.070 --> 00:54:22.871
And so I'm an LP in a number of funds.
830
00:54:22.891 --> 00:54:27.173
So at least I could, and those GPs are kind enough to bring me in and also like,
831
00:54:27.456 --> 00:54:29.618
here and there it'd be like, Clevin, can you like vet this?
832
00:54:29.658 --> 00:54:31.479
And I could actually go vet this, right?
833
00:54:32.620 --> 00:54:38.244
Instead of just, you know, believing in, you know, usually a deck, right?
834
00:54:38.304 --> 00:54:42.247
Or a coffee meeting and then you wire the money and then you don't hear back from them ever.
835
00:54:42.708 --> 00:54:44.009
So, yeah.
836
00:54:44.029 --> 00:54:46.991
So advising, I think it has been a lot better.
837
00:54:47.151 --> 00:54:49.433
I think, of course, you take equity as well.
838
00:54:49.493 --> 00:54:54.477
So you get something, you get something in the game and you're not, you know, out a bunch of,
839
00:54:55.426 --> 00:54:57.670
of cash flow maybe that's a little bit more risk adjusted.
840
00:54:58.932 --> 00:55:05.564
But yeah, but then people actually want you because they thought you were of some help to them.
841
00:55:05.965 --> 00:55:06.205
Right?
842
00:55:07.070 --> 00:55:07.870
in some way, right?
843
00:55:07.890 --> 00:55:10.691
Maybe it's technical, maybe it's hiring, maybe it's like strategy.
844
00:55:11.431 --> 00:55:13.392
I mean, the name itself holds weight, right?
845
00:55:13.632 --> 00:55:21.494
If I'm a founder and I have, you know, you've invested in my company, it's having your name on the cap table looks good in front of other potential investors, even customers maybe.
846
00:55:21.554 --> 00:55:22.634
Yeah, yeah, yeah.
847
00:55:22.694 --> 00:55:27.216
So, I mean, there are certainly many advisor deals that it's just to get my face on their webpage.
848
00:55:27.536 --> 00:55:27.796
Sure.
849
00:55:29.937 --> 00:55:50.592
beyond but a lot of it was just to get my face on the webpage but the rest of it was yeah to talk about potential customers or helping them think through is that the right architecture and whatnot but yeah but I think advising then people feel like okay well I'm giving you equity so let me actually get you on a call like every month or so
850
00:55:51.317 --> 00:55:56.602
And we could talk about something, maybe something that I saw, maybe a competitor and I would be like, what do you think about that?
851
00:55:56.642 --> 00:55:57.964
Have you considered that?
852
00:55:57.984 --> 00:56:04.110
And if I need to make an intro to a VC that I could, I could be of help.
853
00:56:04.130 --> 00:56:07.033
So I think that it seems a little bit more aligned.
854
00:56:07.293 --> 00:56:11.177
So investing through LPs and advising directly, you're getting equity.
855
00:56:11.757 --> 00:56:14.100
And then actually, I enjoy that better too, right?
856
00:56:14.120 --> 00:56:15.101
Because I get to
857
00:56:15.738 --> 00:56:16.799
see so many things, right?
858
00:56:16.819 --> 00:56:20.783
Like, oh, you're building insurance products or you're building observability, right?
859
00:56:20.844 --> 00:56:23.987
Verticalized observability for different silos, right?
860
00:56:24.067 --> 00:56:27.611
Or, you know, I've invested, like, in restaurants and stuff as well.
861
00:56:27.631 --> 00:56:32.576
So, like, but it helps, like, I've invested in advice for those.
862
00:56:32.816 --> 00:56:34.698
So I have, like, oh, okay, cool.
863
00:56:34.718 --> 00:56:35.219
Like, because...
864
00:56:36.252 --> 00:56:39.173
I mean, I've been in the enterprise SaaS business for a long time, right?
865
00:56:39.194 --> 00:56:43.896
So I don't get to be exposed to so many other industries.
866
00:56:43.956 --> 00:56:48.738
And so just helping me understand like, oh, this is the cost lever there.
867
00:56:48.758 --> 00:56:50.259
This is what people are looking for.
868
00:56:50.279 --> 00:56:53.100
If nothing else, just for my intellectual curiosity.
869
00:56:53.381 --> 00:56:56.542
But it's, you know, that's the same thing as like making things you want to be.
870
00:56:57.142 --> 00:56:57.923
That's interesting.
871
00:56:59.119 --> 00:57:02.122
In many ways, you got me thinking, right?
872
00:57:03.123 --> 00:57:11.692
As a founder, being a serial entrepreneur, you've had exits, you've had successes, and you are thinking, right, what's next?
873
00:57:11.972 --> 00:57:14.174
But I have an opportunity to invest now.
874
00:57:14.274 --> 00:57:16.517
I have this newfound wealth I can do something with.
875
00:57:17.317 --> 00:57:32.622
Do you find that perhaps from your experience, being an LP first before going out on your own and deciding where you put the money might be a good way to see how that particular fund is investing and supporting them with advising them on where to, you know, doing the due diligence basically.
876
00:57:34.323 --> 00:57:35.003
I don't know.
877
00:57:35.263 --> 00:57:42.506
Well, I, for the funds that I are investing in as LPs, like I, it's back to the, I'm trusting the person situation again.
878
00:57:42.546 --> 00:57:44.346
Cause of course I could look at their thesis.
879
00:57:44.366 --> 00:57:46.487
I could look at their past funds and performance and all that, but,
880
00:57:47.020 --> 00:57:50.542
I think at the end of the day, like past performance are not indicative of future returns, right?
881
00:57:50.562 --> 00:57:56.146
And so it is also in some sense like a venture investment, right?
882
00:57:56.166 --> 00:58:01.750
So I think you got to make sure like in your portfolio, like it's not like 50% of your net worth is in that, right?
883
00:58:01.770 --> 00:58:12.017
Like you want to limit your exposure there because you don't want to be like just woken up, you know, all the time when interest rate changes and it affects your portfolio in very wild swings.
884
00:58:12.037 --> 00:58:12.877
So...
885
00:58:14.298 --> 00:58:26.554
No, so I think I – but I'm interested to know what the GP is seeing because they give me a lot of insights, right, as – me as an LP investing and they'd be like, okay,
these are the deals that I'm seeing.
886
00:58:26.634 --> 00:58:29.697
So most of the funds that I'm investing in, they're in –
887
00:58:30.465 --> 00:58:32.486
in a space that I am interested in, right?
888
00:58:32.666 --> 00:58:34.006
So like probably software.
889
00:58:34.247 --> 00:58:36.147
You believe in their thesis, you like what they're doing.
890
00:58:36.487 --> 00:58:37.208
Right, right.
891
00:58:37.348 --> 00:58:41.209
But then I could also say, okay, well, what are some interesting things that you recently saw?
892
00:58:41.229 --> 00:58:44.611
Right, and then they'll tell me, and be like, oh, that's cool, right?
893
00:58:44.631 --> 00:58:45.771
So are they open source?
894
00:58:45.791 --> 00:58:47.072
So I could go tinker with that.
895
00:58:47.752 --> 00:58:48.893
Or I could talk to the founders.
896
00:58:49.253 --> 00:58:51.894
And at the very least, I could be like, okay, you're investing in it.
897
00:58:51.914 --> 00:58:53.135
Then I could use it.
898
00:58:53.195 --> 00:58:58.198
I could use Cal.com because we're investing through a fund.
899
00:58:58.618 --> 00:59:02.100
So let's go use it and maybe give some feedback.
900
00:59:02.160 --> 00:59:03.141
But otherwise, just at least.
901
00:59:03.822 --> 00:59:07.544
don't support their competitor in a weird way.
902
00:59:08.305 --> 00:59:09.166
It's interesting.
903
00:59:09.206 --> 00:59:15.570
For me, as an LP in a couple of funds, I get to see the statements and I get to see how they're performing and where their money's being spent.
904
00:59:16.710 --> 00:59:24.155
I guess that's where I was thinking of because it gives you some sort of perspective, even though you are trusting this fund manager to do the investment.
905
00:59:24.375 --> 00:59:26.737
One last question I have regarding investing and advising.
906
00:59:26.777 --> 00:59:26.977
Right.
907
00:59:27.617 --> 00:59:34.823
When you look at what's happening right now in AI, with this AI era that we're in, is it more about engineering or is it more storytelling?
908
00:59:35.023 --> 00:59:38.386
Well, I think it really depends on the company, I would say, right?
909
00:59:38.446 --> 00:59:51.457
Because if you're building perhaps, let's say, in the, if you're solving like a pain point, like for a customer, let's say a B2C or B2B customer, then you have to be able to sell the story, right?
910
00:59:51.477 --> 00:59:55.300
Just the fact that you have AI and you could chat with something like doesn't, like,
911
00:59:56.130 --> 01:00:01.032
some companies just slapped AI on that so you could chat against all of your, let's say your tickets, all your bugs and stuff like that.
912
01:00:01.332 --> 01:00:01.792
Okay.
913
01:00:01.812 --> 01:00:05.394
That's kind of nice, but that's like not what I'm, what's the story?
914
01:00:05.434 --> 01:00:07.535
Like, cause every day I can't find my tickets.
915
01:00:07.555 --> 01:00:10.416
So I need an AI to chat with so I could find my tickets.
916
01:00:10.856 --> 01:00:11.156
I don't know.
917
01:00:11.276 --> 01:00:11.416
Right.
918
01:00:11.436 --> 01:00:16.658
But in, but in many cases, like, I think, you know, if you could tell a story like, oh, we could give this AI, it helps you.
919
01:00:17.079 --> 01:00:18.479
Again, you may use ticket example.
920
01:00:18.979 --> 01:00:20.480
Oh, maybe you could free up all your, your,
921
01:00:21.012 --> 01:00:21.552
support staff.
922
01:00:21.572 --> 01:00:26.494
So they are actually working on like the most important tickets or like the most VIP customers and the rest of it.
923
01:00:26.534 --> 01:00:28.954
Like they could do like first layer triaging and all that.
924
01:00:28.994 --> 01:00:32.135
And all of that means like your customer, like it's happier.
925
01:00:32.155 --> 01:00:38.677
They get instant, like we all, I like at least Amazon's like Amazon had this chatbot a long time ago before even LMS came in.
926
01:00:38.737 --> 01:00:38.877
Right.
927
01:00:38.917 --> 01:00:41.398
And I just, I just feel like, Oh, I don't have to talk to a person.
928
01:00:41.438 --> 01:00:44.099
Like that person can click, click, click, click, click and be like, okay, it's done.
929
01:00:44.159 --> 01:00:44.399
Right.
930
01:00:44.499 --> 01:00:44.779
It's either
931
01:00:45.152 --> 01:00:47.973
they figure out it's a refund or they need to do something about the order.
932
01:00:48.334 --> 01:00:49.414
Like it's nice, it's nice.
933
01:00:49.434 --> 01:00:49.694
Really?
934
01:00:49.754 --> 01:00:53.756
So, so if you could sell that feelings to people and be like, okay, this could be your product.
935
01:00:53.816 --> 01:00:55.717
Like this could be the support experience.
936
01:00:55.737 --> 01:01:02.621
You, people don't have to feel like they're, you know, they, they have to, you know, talk to someone that doesn't get it or, or to be transferred like 20 times.
937
01:01:02.661 --> 01:01:04.241
And, and every time they lose context.
938
01:01:04.902 --> 01:01:06.182
Um, I think that's, that's helpful.
939
01:01:06.743 --> 01:01:07.103
But I think,
940
01:01:07.707 --> 01:01:15.433
on the engineering front, right, then that's, the question becomes like, how do you use AI to actually build IP?
941
01:01:15.813 --> 01:01:16.054
Right.
942
01:01:16.194 --> 01:01:16.474
Right.
943
01:01:16.914 --> 01:01:33.447
And I think that's still, and I find that to be very true today where you can have a good solid chat with AI or do a deep research or something, right, or clawed and like pull in all of the project knowledge and come up with it with an understanding of something that you did not know before.
944
01:01:33.927 --> 01:01:34.708
Oh yeah, there was some,
945
01:01:35.281 --> 01:01:39.245
like there was some research that is in this space that I am not aware of at all.
946
01:01:39.445 --> 01:01:40.607
And it's able to surface that.
947
01:01:40.887 --> 01:01:42.729
And I've chatted and be like, okay, so how does that work?
948
01:01:42.929 --> 01:01:45.011
Like, what is the core idea there?
949
01:01:45.031 --> 01:01:47.333
How does that compare to what I know already?
950
01:01:47.694 --> 01:01:51.057
And I'd be like, okay, so if I want to know more about this, like, can you tell me?
951
01:01:51.357 --> 01:01:57.824
So I think there is this, you know, kind of research partner that is already there, right?
952
01:01:57.964 --> 01:01:59.045
And it may not write,
953
01:01:59.338 --> 01:02:01.639
that code properly in some of these cases.
954
01:02:02.039 --> 01:02:07.581
But at least it will give you, you know, cause we had talked about hallucinations and wrong stuff getting spewed all the time.
955
01:02:07.601 --> 01:02:11.202
But I feel the researchy side of it is actually quite good.
956
01:02:11.362 --> 01:02:15.204
Cause I probably, because all these LMs are trained on all the papers and all that, you know, right.
957
01:02:15.224 --> 01:02:18.165
And they could search and these things are readily available.
958
01:02:18.586 --> 01:02:22.029
But I find like my hallucinations is more about asking like, where is the World Expo?
959
01:02:22.109 --> 01:02:29.237
And it's somehow like, still today could hallucinate and just say, oh, it's in like Tokyo this year and be like, no, that's not.
960
01:02:29.597 --> 01:02:33.381
But like with research stuff, like it oftentimes will get it and I'll have a conversation.
961
01:02:33.421 --> 01:02:38.366
It will not just tell me about an algorithm that did not exist or tell me something completely gibberish.
962
01:02:38.386 --> 01:02:39.747
So why the algorithm...
963
01:02:40.077 --> 01:02:44.620
works and the reason behind it and the proofs and all that, like it will be able to do that.
964
01:02:44.760 --> 01:02:53.266
So it's, well, it's also probably because most of the models are tuned to like all these math Olympiads and research oriented questions, right?
965
01:02:53.286 --> 01:02:53.486
So yeah.
966
01:02:53.866 --> 01:02:54.647
Yeah, absolutely right.
967
01:02:54.667 --> 01:02:56.268
I think the research side is definitely strong.
968
01:02:56.308 --> 01:03:01.792
I mean, I, as you said at the start, I found myself talking to it in the cloud, just having a conversation.
969
01:03:03.313 --> 01:03:13.900
But the reason I asked that question around the engineering and the storytelling, I just think about Amazon's Alexa and Apple's Siri and then opening eyes chat GPT.
970
01:03:14.600 --> 01:03:20.804
Was the engineering just not there with Alexa and Siri and they were probably first in the market in a way, but they failed?
971
01:03:21.424 --> 01:03:23.486
Or was it the timing wasn't right?
972
01:03:23.806 --> 01:03:27.288
You know, it's like the timing of chat GPT was just perfect.
973
01:03:28.160 --> 01:03:28.400
Yeah.
974
01:03:29.321 --> 01:03:29.741
I don't know.
975
01:03:29.861 --> 01:03:33.263
I think chat interfaces have been around for quite some time.
976
01:03:33.343 --> 01:03:44.430
So if you tell me if chat GPD were 10 years earlier, if it came in 2012 or something, I think it would still blow up in the same way, I think.
977
01:03:44.610 --> 01:03:45.390
Right.
978
01:03:45.470 --> 01:03:48.732
It would certainly basically mean that nobody would use Siri or even invent it.
979
01:03:49.072 --> 01:03:50.793
Everybody would be like, okay, why would you do it that way?
980
01:03:51.414 --> 01:03:53.515
I'm just surprised, though, that,
981
01:03:54.145 --> 01:04:20.384
having chat gpt around for so long like even the entrenched players is still catching up right we were talking about three years in right and i'm using siri and i think i was asking it just a very simple translate question just so you get it like on my watch and be like i can't do this and i'm like wow like you crazy right and and i think google is a little better like google systems i think it is a little better alexa i think that's still i i haven't used it in a while so i shouldn't comment um but
982
01:04:21.341 --> 01:04:28.023
But it still goes to show like just the – how like a new player like in this age, just like in the internet age, right?
983
01:04:28.043 --> 01:04:28.703
Maybe in the cloud age.
984
01:04:28.763 --> 01:04:32.405
Like you have this ability to like just display something, right?
985
01:04:32.425 --> 01:04:34.565
Google obviously is very concerned like people are not doing searches.
986
01:04:34.905 --> 01:04:39.207
Because like, you know, me and my wife, like we – like everything is like ChatGPT now, right?
987
01:04:39.227 --> 01:04:42.908
Like it's like Google like caught up by having the search –
988
01:04:43.332 --> 01:04:45.573
AI previews and stuff like that today.
989
01:04:45.593 --> 01:04:47.393
But like, it lost the mindset.
990
01:04:47.853 --> 01:04:50.394
It lost that, you know, the zeitgeist.
991
01:04:50.454 --> 01:04:51.755
It's like, it's not in there anymore.
992
01:04:51.795 --> 01:04:58.377
And so people just be like, okay, I'm just going to go to chat GPT and I'm going to have a chat with it and I could have follow-up questions and all that.
993
01:04:58.557 --> 01:05:06.459
And so, yeah, I really don't know, honestly, like why, what is holding these assistants back?
994
01:05:06.820 --> 01:05:08.520
Like, is it, there's a legal problem?
995
01:05:08.580 --> 01:05:10.361
Is there, like, there's a correctness problem?
996
01:05:10.381 --> 01:05:11.401
Like, of course you could just say,
997
01:05:12.245 --> 01:05:14.547
I almost feel like, yeah, you could just say like it makes mistakes.
998
01:05:14.668 --> 01:05:16.009
It's not like it doesn't make mistakes today.
999
01:05:16.049 --> 01:05:26.119
So why don't you just like open it up, open up the floodgates to like 1% of people and just be like, let's see like if people really get super angry or do they just, oh, okay.
1000
01:05:26.279 --> 01:05:33.286
Like maybe I won't need chat GPT today because I just asked, I could ask my watch or I could ask my Nest Hub or something.
1001
01:05:33.447 -->
01:05:33.627
Right.
1002
01:05:34.177 --> 01:05:38.160
I mean, just like you, I mean, you may have, but I have never used Alexa.
1003
01:05:38.180 --> 01:05:40.261
But at Siri, I just always switched it off.
1004
01:05:40.661 --> 01:05:44.904
Like every time I got a new phone, I just switched it off because I just found it so annoying.
1005
01:05:45.304 --> 01:05:46.625
I'm having a conversation.
1006
01:05:46.705 --> 01:05:48.126
I'll just accidentally say something.
1007
01:05:48.166 --> 01:05:50.368
It was just, it'll think I'm talking to it.
1008
01:05:51.108 --> 01:05:57.412
And I think based on that, I kind of see it as, is it a possibility that the engineering just wasn't there, but the storytelling was great?
1009
01:05:57.472 --> 01:05:59.734
Like they sold the idea with Siri.
1010
01:06:00.154 --> 01:06:00.894
Yeah.
1011
01:06:01.115 --> 01:06:01.695
I think the,
1012
01:06:02.630 --> 01:06:03.170
I use Siri.
1013
01:06:03.350 --> 01:06:06.031
So the reason is because I have a newborn child.
1014
01:06:06.191 --> 01:06:11.672
So I realized for the very first time, like I have no hands to control any of my devices.
1015
01:06:12.192 --> 01:06:22.754
And so I actually went ahead and learned how to use Siri properly, because there's like the shortcuts app and you could actually build like really weird like way to build like procedural logic.
1016
01:06:23.774 --> 01:06:27.835
And sometimes because I just don't want to say so many things when I need to log a diaper or something.
1017
01:06:27.855 --> 01:06:29.735
So I'd be like, okay, let me just program shortcuts.
1018
01:06:30.165 --> 01:06:31.065
So I learned how to use it.
1019
01:06:31.145 --> 01:06:32.706
It's very clunky, but it works.
1020
01:06:32.906 --> 01:06:38.388
And I, of course, they're still asking like, why can't, because there's a natural language solution to all of this, but they don't have it.
1021
01:06:38.848 --> 01:06:41.169
So, so I think, yeah, it does.
1022
01:06:42.450 --> 01:06:42.590
Yeah.
1023
01:06:42.610 --> 01:06:43.610
There is a story there.
1024
01:06:43.670 --> 01:06:49.973
Like there, it does help with, with quality of life, but I feel like they, I don't, I don't know.
1025
01:06:49.993 --> 01:06:50.213
Right.
1026
01:06:50.253 --> 01:06:53.734
And I guess that's why Apple is trying to figure out like what that means.
1027
01:06:54.147 --> 01:07:00.858
I think Apple is uniquely difficult because they have the privacy mandate or the privacy claim.
1028
01:07:01.660 --> 01:07:03.503
So how are you actually going to do that, right?
1029
01:07:03.543 --> 01:07:07.730
Because the most powerful models are in the cloud and the on-device models are not going to be as good.
1030
01:07:09.215 --> 01:07:10.576
And this is the legacy problem, right?
1031
01:07:10.596 --> 01:07:13.458
Like they probably have APIs into all these apps, right, in this weird way.
1032
01:07:13.498 --> 01:07:16.420
And, you know, are we going to just like expose all of it to the model?
1033
01:07:16.740 --> 01:07:17.761
And does the model know what to do?
1034
01:07:17.821 --> 01:07:24.405
And if it does something very wrong, like order stuff from Uber or DoorDash, like how are you going to like unwind all of that, right?
1035
01:07:24.425 --> 01:07:28.888
Maybe some of them are not as easy as just like because you said something.
1036
01:07:28.928 --> 01:07:30.589
So obviously we'll order your…
1037
01:07:30.889 --> 01:07:48.716
your pizza but but now if it's just thinking you want a pizza and then ordered a pizza then you know do who is responsible right and how do you unwind all right so maybe there's there's some of it there going on but again i i just feel like at this stage like because you know you're either in the zeitgeist or you're not like and they're not right so like why not
1038
01:07:49.767 --> 01:07:53.069
Why not bet the farm on it, right?
1039
01:07:53.129 --> 01:07:54.809
I mean, you're turning it off already.
1040
01:07:54.830 --> 01:07:57.071
So, you know, it's that.
1041
01:07:57.191 --> 01:08:00.552
Or you have maybe some users that will not turn it off.
1042
01:08:00.612 --> 01:08:02.153
And the users that got annoyed get turned it off.
1043
01:08:02.593 --> 01:08:04.534
I think now you could talk to it through.
1044
01:08:04.654 --> 01:08:07.055
So you talked to Siri, but you can use ChatGPT.
1045
01:08:07.075 --> 01:08:10.397
So it would say, I would tell it to play a song or something.
1046
01:08:10.417 --> 01:08:14.239
And it would say, should I use ChatGPT to, like, I'm asking it to do something.
1047
01:08:14.259 --> 01:08:15.780
And it would say, should I use ChatGPT?
1048
01:08:15.920 --> 01:08:17.841
But I think that's like doing a search thing.
1049
01:08:18.541 --> 01:08:20.242
It's not like if you're controlling your phone.
1050
01:08:20.662 --> 01:08:21.002
Right.
1051
01:08:21.142 --> 01:08:21.422
Yeah.
1052
01:08:21.522 --> 01:08:29.444
So I think it's not like if it's integrated in your calendar, like you want to like, okay, I would love it to have a time and be like, okay, can you clear my day for tomorrow?
1053
01:08:29.704 --> 01:08:29.944
Yes.
1054
01:08:30.064 --> 01:08:30.684
Like I'm sick today.
1055
01:08:30.704 --> 01:08:31.485
Like clear my day.
1056
01:08:31.785 --> 01:08:31.945
Yeah.
1057
01:08:31.965 --> 01:08:32.345
That'd be nice.
1058
01:08:32.385 --> 01:08:33.525
Email to everyone.
1059
01:08:33.545 --> 01:08:34.365
Email, decline.
1060
01:08:34.446 --> 01:08:34.766
Right.
1061
01:08:34.786 --> 01:08:35.366
That'd be nice.
1062
01:08:35.466 --> 01:08:37.066
But it doesn't do that.
1063
01:08:37.106 --> 01:08:37.586
That's pretty cool.
1064
01:08:37.606 --> 01:08:38.507
That's a great business idea.
1065
01:08:39.087 --> 01:08:41.207
I think Claude is pretty close to that.
1066
01:08:41.227 --> 01:08:42.708
You can connect everything.
1067
01:08:42.848 --> 01:08:43.048
Yeah.
1068
01:08:43.508 --> 01:08:45.329
Clem, it's been really awesome having you on the show.
1069
01:08:45.609 --> 01:08:50.652
I'd like to always finish off these episodes by asking a question.
1070
01:08:50.672 --> 01:08:53.834
I mean, this whole show is about learning.
1071
01:08:53.854 --> 01:09:05.041
Inside Silicon Mind is about getting in the mind of individuals who are in Silicon Valley, or even if they're not in Silicon Valley, but they've built companies, built technology companies, software companies, and now AI companies.
1072
01:09:06.101 --> 01:09:11.648
As far as like learning and developing and expanding our minds, one thing I'm quite avid about is reading.
1073
01:09:11.868 --> 01:09:14.531
And I always love a good book in my hands.
1074
01:09:14.972 --> 01:09:16.494
I'm not a fan of audibles.
1075
01:09:16.514 --> 01:09:18.436
I like to buy the book and read it.
1076
01:09:19.818 --> 01:09:23.562
But to our viewers and listeners, is there a book that you've read?
1077
01:09:24.462 --> 01:09:38.636
one that's truly resonated with you that you feel has had a impact on your career, on, it could be your career, it could be on how you think about technology, AI, one that has truly been quite profound for you.
1078
01:09:41.078 --> 01:09:41.319
Okay.
1079
01:09:41.379 --> 01:09:47.785
Well, so I think if you go back to my career, I would, I would go back to the book, The Right It, I better support his book.
1080
01:09:48.305 --> 01:09:48.726
I think that
1081
01:09:49.900 --> 01:09:53.143
As an engineer, I wanted to understand why things fail.
1082
01:09:53.163 --> 01:09:54.184
Right.
1083
01:09:54.244 --> 01:09:56.446
Like, fail to take off and all that.
1084
01:09:56.746 --> 01:10:03.051
And because I think just as an engineer, you spend so much time thinking about, like, yeah, everything.
1085
01:10:03.111 --> 01:10:04.192
Like, line spaces, tap spaces.
1086
01:10:04.212 --> 01:10:04.993
How do you align?
1087
01:10:05.073 --> 01:10:07.175
Is it passive voices?
1088
01:10:07.215 --> 01:10:08.115
All that kind of stuff like you do.
1089
01:10:08.436 --> 01:10:14.881
But if it doesn't work, right, if nobody uses it, then it's just a theoretical exercise that you just did and wasted all your time.
1090
01:10:15.682 --> 01:10:16.523
So I think that book...
1091
01:10:17.660 --> 01:10:24.546
really gave me the framework to think about like, yes, if I'm an engineer, I want to build it right.
1092
01:10:25.226 --> 01:10:26.788
But like, what is the right it to build?
1093
01:10:27.168 --> 01:10:31.371
And I think that, you know, I think Aborah Savory was a VP of engineering at Google.
1094
01:10:31.391 --> 01:10:35.635
And so like he came for, he has that kind of engineering mindset.
1095
01:10:36.736 --> 01:10:41.520
So he understood like how frustrating it is to spend so much of your life building something.
1096
01:10:41.740 --> 01:10:44.122
And Google has a lot of those things going on, building so much,
1097
01:10:44.542 --> 01:10:47.725
And then not getting traction and all that.
1098
01:10:47.745 --> 01:10:55.792
So asking the
right questions and, you know, doing experiments and iterating and getting data quickly, cheaply, quickly, easily, all that.
1099
01:10:55.812 --> 01:11:09.565
So I think that book has shaped me immensely, I think, to the degree that, you know, after Wayfront, I actually built an ed tech company, Generative AI, with my sister.
1100
01:11:09.985 --> 01:11:11.006
And I think that helped me, like,
1101
01:11:11.622 --> 01:11:15.506
really try to figure out like, hey, what is the it in that space?
1102
01:11:15.526 --> 01:11:16.306
Because I have no idea.
1103
01:11:16.326 --> 01:11:18.508
Like I have not been in the education space, right?
1104
01:11:18.608 --> 01:11:20.130
I'm not in early childhood education for sure.
1105
01:11:20.770 --> 01:11:33.721
And so I think that helped me to kind of, you know, ask the questions, like design the experiments or at least design, you know, how do we want the first, you know, versions of the app to look like and feel like what are the in-person testing that we could do?
1106
01:11:33.861 --> 01:11:37.385
And I think that helped inform me a lot on that.
1107
01:11:38.249 --> 01:11:42.252
When you think about that book, what was it specifically about it?
1108
01:11:42.292 --> 01:11:43.933
What was it really teaching you?
1109
01:11:45.273 --> 01:11:51.797
I think it's really saying that at the end of the day, nothing really speaks success to it.
1110
01:11:53.333 --> 01:11:56.037
apart from somebody actually using it and paying for it.
1111
01:11:56.237 --> 01:12:04.369
It seems like very cliche, but I think there's a lot of cases where, okay, you do a study, you could do a market study, or you could do like a research.
1112
01:12:04.389 --> 01:12:09.556
You could ask, I'm sure like all the consulting companies can give you like, okay, if you do this, like that, you know, that you get a lot of people, right?
1113
01:12:10.477 --> 01:12:12.318
Well, I love users.
1114
01:12:12.338 --> 01:12:18.603
Or you'd be like, okay, we could, like, spend all this time and effort, right, to build, you know, the final shiny thing.
1115
01:12:18.683 --> 01:12:20.164
Because people will say that, right?
1116
01:12:20.184 --> 01:12:24.287
Like, oh, it needs to be perfect before, like, you know, before people will use it.
1117
01:12:25.027 --> 01:12:27.389
But that book is arguing, like, for the complete opposite.
1118
01:12:27.429 --> 01:12:35.515
It's like, oh, no, like, you, you know, use the minimal amount of effort, minimal amount of code or, you know, whatever.
1119
01:12:35.535 --> 01:12:35.655
Right.
1120
01:12:36.256 --> 01:12:43.934
and then you design these experiments there's one story in there that always like resonates with me which is when IBM built speech recognition back in I guess in the 70s
1121
01:12:45.493 --> 01:12:46.794
which became IBM via voice.
1122
01:12:46.834 --> 01:12:52.597
It's before all these like text-to-speech, speech-to-text stuff that we have on the phone, which now works perfectly.
1123
01:12:53.058 --> 01:12:54.458
They thought about building it.
1124
01:12:54.899 --> 01:12:59.221
And before they built it, they wanted to understand like whether people would pay for it and how people would use it.
1125
01:12:59.621 --> 01:13:05.745
So they just set it up in a room where there's the computer there and they just invite people there and be like, okay, there's now a computer that could just transcribe your voice.
1126
01:13:06.225 --> 01:13:09.727
And people would just sit there and be like, oh, nice, like I could talk and go, what's up here on the screen?
1127
01:13:10.347 --> 01:13:13.449
And of course, there's just a typist in the next room, right, typing...
1128
01:13:14.122 --> 01:13:16.825
And so people would get this feeling that they have this technology.
1129
01:13:16.845 --> 01:13:17.165
Oh, wow.
1130
01:13:17.205 --> 01:13:18.066
So they haven't built it yet.
1131
01:13:18.086 --> 01:13:18.867
They haven't built it yet.
1132
01:13:18.947 --> 01:13:19.267
Yeah.
1133
01:13:19.828 --> 01:13:23.071
And then they very quickly realized people would not dictate their emails because...
1134
01:13:23.413 --> 01:13:27.035
then your entire office would know what you're doing.
1135
01:13:27.095 --> 01:13:27.615
That's true.
1136
01:13:27.755 --> 01:13:28.416
That's very true.
1137
01:13:28.496 --> 01:13:30.036
Still nowadays, right?
1138
01:13:30.056 --> 01:13:32.638
Like we don't get emails in our offices, right?
1139
01:13:32.858 --> 01:13:34.139
And then they'd be like, okay, this is nice.
1140
01:13:34.199 --> 01:13:35.079
Okay, so what would happen?
1141
01:13:35.119 --> 01:13:37.600
Like, oh, the secretaries may not have a work.
1142
01:13:38.141 --> 01:13:39.181
But it starts to ideate.
1143
01:13:39.201 --> 01:13:42.023
Like it starts to like people would start to think like, okay, this is how much I would pay for it.
1144
01:13:42.063 --> 01:13:44.484
Like, oh, if I have this, like, would you pay for it?
1145
01:13:44.544 --> 01:13:45.324
Would you pay for it now?
1146
01:13:45.844 --> 01:13:49.606
And then you could test like, okay, this is like why people would want something or not want something.
1147
01:13:49.646 --> 01:13:52.608
So I think that really stuck with me, which is like, okay, I could,
1148
01:13:52.848 --> 01:14:18.713
because as an engineer like you know you totally could get me to be like let's build that thing right that's so cool that's like speech recognition right um but i think now like you you need to figure out like how people would use it what people would pay for it who would pay for it and all that and then you really could understand like okay the the product should be built in a particular way so i think that that really that's a great story yeah the ibm story that's interesting yep um thank you for the recommendation that's a great book um and
1149
01:14:20.677 --> 01:14:26.282
As far as books are concerned, do you read a lot or is it?
1150
01:14:26.523 --> 01:14:27.323
I do.
1151
01:14:27.504 --> 01:14:30.106
You said no Audible, but I am a big Audible fan.
1152
01:14:30.126 --> 01:14:30.546
Oh, you are?
1153
01:14:30.687 --> 01:14:31.027
Okay.
1154
01:14:31.067 --> 01:14:34.650
In the car, it's either ChatGPT or Audible or a podcast, I guess.
1155
01:14:35.411 --> 01:14:37.413
What sort of Audible do you tend to listen to in the car?
1156
01:14:37.773 --> 01:14:38.374
It's nonfiction.
1157
01:14:38.634 --> 01:14:39.074
Nonfiction?
1158
01:14:39.094 --> 01:14:39.355
For sure.
1159
01:14:39.375 --> 01:14:39.495
Yeah.
1160
01:14:39.775 --> 01:14:40.075
Awesome.
1161
01:14:40.495 --> 01:14:40.755
Cool.
1162
01:14:41.476 --> 01:14:43.276
Well, Clement, it was a pleasure, honestly.
1163
01:14:43.296 --> 01:14:48.918
Thank you so much for making the time to see me here today and record this episode with me.
1164
01:14:49.538 --> 01:14:51.399
I had a lot of fun, a lot of learns.
1165
01:14:52.139 --> 01:14:56.001
And obviously, I wish you all the best of luck on your next big journey.
1166
01:14:56.301 --> 01:14:56.621
Thank you.
1167
01:14:57.041 --> 01:14:57.841
Yeah, it was really fun.
1168
01:14:58.101 --> 01:14:58.561
Appreciate it.
1169
01:14:59.202 --> 01:14:59.602
Thanks, Clement.
1170
01:14:59.902 --> 01:15:00.122
Thanks.
1171
01:15:04.258 --> 01:15:08.622
This podcast was brought to you by Sarmana Labs and powered by Harrison Clark.
1172
01:15:09.082 --> 01:15:19.851
At Harrison Clark, we're at the heart of Silicon Valley's evolution and at the intersection of search and venture, funding the company shaping the future and connecting them to the most elite cloud data and AI talent.
1173
01:15:20.272 --> 01:15:22.514
To learn more, visit HarrisonClark.com.
1174
01:15:22.894 --> 01:15:27.298
If today's episode challenged your thinking, then share it with someone who's driving innovation.
1175
01:15:27.718 --> 01:15:36.702
For even more insights on achieving product market fit, subscribe to our newsletter, The PMF Playbook, where we break down the journey from zero to product market fit.
1176
01:15:37.122 --> 01:15:39.523
Visit thepmfplaybook.com.
1177
01:15:39.963 --> 01:15:45.045
For more episodes, follow us on social media or visit insidethesiliconmind.com.
1178
01:15:45.465 --> 01:15:50.767
Until next time, stay curious, stay consistent, and stay inside the Silicon Mind.
1179
01:15:51.228 --> 01:15:51.748
See you soon.