WEBVTT

NOTE Sentence-level transcript of https://www.youtube.com/watch?v=tUPPVhBBcoM

NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/tUPPVhBBcoM.html. A cue ends where the next begins, or 2 s after its last word.

s1
00:00:01.309 --> 00:00:03.309
[music]

s2
00:00:13.120 --> 00:00:13.679
Okay.

s3
00:00:13.679 --> 00:00:14.400
Hello everyone.

s4
00:00:14.400 --> 00:00:16.000
Uh, I'm excited to be here.

s5
00:00:16.000 --> 00:00:17.840
Uh, my name is Zach Lloyd.

s6
00:00:17.840 --> 00:00:27.760
Um, today I'm going to be talking about self-improving software factories, the new open-source model, and basically what I think is happening to development.

s7
00:00:27.760 --> 00:00:30.000
A little bit about me just to begin.

s8
00:00:30.000 --> 00:00:33.920
So I am a former uh principal engineer from Google.

s9
00:00:33.920 --> 00:00:36.800
Used to lead engineering on the Google Doc Suite.

s10
00:00:36.800 --> 00:00:40.160
I've been an engineer now for over 20 years.

s11
00:00:40.160 --> 00:00:41.200
Long time.

s12
00:00:41.200 --> 00:00:47.600
I am still uh shipping frequently, but I haven't written a line of code in the last six months.

s13
00:00:47.600 --> 00:00:50.160
Uh and I'm the founder of a company called Warp.

s14
00:00:50.160 --> 00:00:55.600
Uh Warp, if you're not familiar, is a open-source agentic development environment.

s15
00:00:55.600 --> 00:00:58.160
Uh you may know us as a terminal.

s16
00:00:58.160 --> 00:00:59.600
That is how the company started.

s17
00:00:59.600 --> 00:01:03.199
We're basically a terminal that has agents built in.

s18
00:01:03.199 --> 00:01:09.600
Uh we open sourced it a couple months ago and I'm going to talk a little bit about that experience and the motivation for it.

s19
00:01:09.600 --> 00:01:12.880
It's popular open source project over 60,000 GitHub stars.

s20
00:01:12.880 --> 00:01:15.520
We've had a couple hundred people contributing.

s21
00:01:15.520 --> 00:01:19.840
We have over 800,000 active developers who are using warp.

s22
00:01:19.840 --> 00:01:30.240
And increasingly we are focused not just on the terminal aspect and the interactive aspect of development but more so on how do you automate development.

s23
00:01:30.240 --> 00:01:33.280
I'm going to talk mostly about that.

s24
00:01:33.280 --> 00:01:44.000
So the thesis that I have uh is that the discipline of software engineering is going to become something more like factory engineering.

s25
00:01:44.000 --> 00:01:47.439
And I'll explain what I mean by this in a minute but just keep that in mind.

s26
00:01:47.439 --> 00:01:49.920
That's that's what I think is going to happen.

s27
00:01:49.920 --> 00:01:57.840
Um, if you look at development over the past couple years, it's I mean it's crazy how it's changed.

s28
00:01:57.840 --> 00:02:09.200
We've gone from a world of chat and AI autocomplete cursor co-pilot to the phase that we're in now, which I consider to be mostly interactive agents.

s29
00:02:09.200 --> 00:02:16.319
So you're sort of sitting at your computer and you are telling cloud code to do something, you're telling warp to do something.

s30
00:02:16.319 --> 00:02:25.920
And I believe what's going to happen over the next six months, a year, hard to predict the pace, is that we're going to move much more towards a world of automation.

s31
00:02:25.920 --> 00:02:28.000
But before I get into that, just quick show of hands.

s32
00:02:28.000 --> 00:02:31.840
How many folks in here are building with agents?

s33
00:02:31.840 --> 00:02:33.920
Every 100% makes sense.

s34
00:02:33.920 --> 00:02:40.720
Uh how many how many folks are building typically with multiple agents at one time?

s35
00:02:40.720 --> 00:02:42.480
So again, almost everyone.

s36
00:02:42.480 --> 00:02:44.239
How many people are running an agent right now?

s37
00:02:44.239 --> 00:02:44.959
I'm not offended.

s38
00:02:44.959 --> 00:02:45.760
I would.

s39
00:02:45.760 --> 00:02:47.120
Okay, that's totally cool.

s40
00:02:47.120 --> 00:02:48.560
I would be doing it too.

s41
00:02:48.560 --> 00:02:53.280
Uh, how many folks are running agents in the cloud out of curiosity?

s42
00:02:53.680 --> 00:02:59.040
So, that looks like less than half, but still significant.

s43
00:02:59.040 --> 00:03:06.159
Uh, and how many folks have have set up a system internally to automate the whole software development life cycle?

s44
00:03:06.159 --> 00:03:11.519
So, everything from like triaging, specking, implementing, reviewing.

s45
00:03:11.519 --> 00:03:14.480
So, I see some hands.

s46
00:03:14.480 --> 00:03:15.760
So some people are doing this.

s47
00:03:15.760 --> 00:03:17.840
So this is what's going to happen.

s48
00:03:17.840 --> 00:03:23.360
Uh every project of significant size I believe is going to have something like this.

s49
00:03:23.360 --> 00:03:27.360
Um and it's going to look kind of like this big loop.

s50
00:03:27.360 --> 00:03:29.200
And everyone is talking about loops.

s51
00:03:29.200 --> 00:03:31.200
There's nothing that complicated about loops.

s52
00:03:31.200 --> 00:03:33.280
This loop says a cloud software factory.

s53
00:03:33.280 --> 00:03:36.239
This loop could literally just say like the software development life cycle.

s54
00:03:36.239 --> 00:03:37.599
It's the same thing.

s55
00:03:37.599 --> 00:03:43.120
Um but just to go through this loop, it's like ideas are going to come in at the top.

s56
00:03:43.120 --> 00:03:45.120
Agents are going to do triage.

s57
00:03:45.120 --> 00:03:47.920
If something is complicated, they will write a spec.

s58
00:03:47.920 --> 00:03:50.239
Uh these little blue boxes are where humans step in.

s59
00:03:50.239 --> 00:03:52.159
Humans will review the spec.

s60
00:03:52.159 --> 00:03:54.239
Uh agents will do the implementation.

s61
00:03:54.239 --> 00:03:56.239
A human and agent will review the code.

s62
00:03:56.239 --> 00:03:57.680
Agents will verify.

s63
00:03:57.680 --> 00:03:59.519
Human will review the product.

s64
00:03:59.519 --> 00:04:03.040
You ship and then you monitor and round you go.

s65
00:04:03.040 --> 00:04:07.920
Uh and this is what software development for better or worse I think is going to end up looking like.

s66
00:04:07.920 --> 00:04:19.382
So I repeat the thesis which is that if this is what software engineering is going to look like um software engineers are going to be the ones who end up building and managing these factories.

s67
00:04:19.382 --> 00:04:20.000
[snorts]

s68
00:04:20.000 --> 00:04:27.040
Now I promised at the beginning and I put in the title of the talk that I was going to talk about open source and so I want to do that for a few minutes.

s69
00:04:27.040 --> 00:04:29.280
I'm going to take a quick digression.

s70
00:04:29.280 --> 00:04:38.400
Um, I bring up open source because one of the main reasons that Warp opens sourced was to build uh build a public factory.

s71
00:04:38.400 --> 00:04:50.160
Um, and so this is a picture of this website we've built called build.warp.dev which shows all of the issues that are flowing through our system and what state they're in,

s72
00:04:50.160 --> 00:04:54.160
what agents are working on them, what contributors are working on them.

s73
00:04:54.160 --> 00:04:58.160
And it's kind of like a protoactory done at scale.

s74
00:04:58.160 --> 00:05:01.280
It's not working perfectly, but it is working.

s75
00:05:01.280 --> 00:05:05.360
And one of the reasons we open sourced was to try to build this.

s76
00:05:05.360 --> 00:05:12.639
Um, just in general, I think it's interesting to talk about open source in the time of agentic development.

s77
00:05:12.639 --> 00:05:16.960
This is a really stupid graph, but it's like it just it you get it.

s78
00:05:16.960 --> 00:05:20.080
It's like it's becoming much cheaper to build software.

s79
00:05:20.080 --> 00:05:27.199
Um, a correlary of that is that it's becoming trivial to clone software.

s80
00:05:27.199 --> 00:05:38.960
And so if you are in the software business, and I don't know how many folks in this room are in the software business per se, but it's very hard to build a software business if it's free to build software.

s81
00:05:38.960 --> 00:05:43.199
It's hard to capture the value, especially if a competitor can clone.

s82
00:05:43.199 --> 00:05:49.840
And so my big takeaway or big tip for everyone in here is that the first thing you should do is patent your code.

s83
00:05:49.840 --> 00:05:50.479
I'm kidding.

s84
00:05:50.479 --> 00:05:52.393
Don't This is a complete joke.

s85
00:05:52.393 --> 00:05:52.560
[laughter]

s86
00:05:52.560 --> 00:05:55.120
Don't don't don't do this.

s87
00:05:55.120 --> 00:06:00.720
Uh my my first tip is obviously you need to have a great product.

s88
00:06:00.720 --> 00:06:02.800
Uh this has always been the case.

s89
00:06:02.800 --> 00:06:06.960
Um but I would say a great product probably was never enough.

s90
00:06:06.960 --> 00:06:15.919
But even now more than ever, if you think that you're going to build a great software business just by building and shipping a great product, you're probably not going to succeed.

s91
00:06:15.919 --> 00:06:18.800
Uh you need advantages beyond the product.

s92
00:06:18.800 --> 00:06:23.440
And so those advantages could look like distribution ecosystem.

s93
00:06:23.440 --> 00:06:26.639
It could be that you have a great brand or a data moat.

s94
00:06:26.639 --> 00:06:28.000
You might have capital.

s95
00:06:28.000 --> 00:06:34.720
Um, but if you're a startup, again, I'm coming from the startup world here, you just don't have these advantages.

s96
00:06:34.720 --> 00:06:38.000
And so, you you still want to break through.

s97
00:06:38.000 --> 00:06:42.639
Uh, and one of the ways that I suggest doing this is by building in the open.

s98
00:06:42.639 --> 00:06:49.919
And so, to be clear, it took Warp five years of building closed to sort of make the leap into building in the open.

s99
00:06:49.919 --> 00:06:51.280
And I'll explain why.

s100
00:06:51.280 --> 00:06:55.360
But if you build in the open, um it helps build your ecosystem.

s101
00:06:55.360 --> 00:06:59.680
It it can take you from being like hated on hacker news to like tolerated.

s102
00:06:59.680 --> 00:07:01.520
Uh it can burnish your brand.

s103
00:07:01.520 --> 00:07:02.479
It creates community.

s104
00:07:02.479 --> 00:07:05.120
And so there's all these advantages to it.

s105
00:07:05.120 --> 00:07:11.680
Uh and some of the things that I think have traditionally been a pain um can now be managed.

s106
00:07:11.680 --> 00:07:25.759
And so like the traditional, you know, pain of open source might be something like you get a lot of noisy issues, you get sloppy PRs, you can end up in code review hell, you can end up having to spend a lot of time

s107
00:07:25.759 --> 00:07:27.120
verifying changes.

s108
00:07:27.120 --> 00:07:35.759
And so the solution, it's a kind of a long-winded way of getting to this for open source or at least for Warp in the case of open source.

s109
00:07:35.759 --> 00:07:46.240
The thing that made us finally decide to do this was that we built a whole set of automations, really a software factory around managing the open-source project.

s110
00:07:46.240 --> 00:07:50.160
And so, like I said, this is what I think the future is going to look like.

s111
00:07:50.160 --> 00:07:58.800
I'm going to drill into it a bit just to get a little bit more technical for folks who want to try to build something like this for their own projects.

s112
00:07:58.800 --> 00:08:01.840
So, what are the components of an effective software factory?

s113
00:08:01.840 --> 00:08:06.080
uh it's really not that complicated to start or at a high level.

s114
00:08:06.080 --> 00:08:08.639
You need a set of automations.

s115
00:08:08.639 --> 00:08:12.560
You need a way of providing context and skills.

s116
00:08:12.560 --> 00:08:16.479
You need a way of bringing humans in at the correct time.

s117
00:08:16.479 --> 00:08:19.840
So sort of like when things get stuck on the factory.

s118
00:08:19.840 --> 00:08:23.919
And then a really important thing is you need some set of self-improvement capabilities.

s119
00:08:23.919 --> 00:08:26.560
So think of this as loops.

s120
00:08:26.560 --> 00:08:33.839
And if you do this right, uh, in the open source world, you can get a, you know, world where agents are helping contributors contribute.

s121
00:08:33.839 --> 00:08:36.959
They're helping maintainers maintain.

s122
00:08:36.959 --> 00:08:40.399
And I want to emphasize there's nothing special about open source here.

s123
00:08:40.399 --> 00:08:43.919
I think every sizable project can benefit from this approach.

s124
00:08:43.919 --> 00:08:52.399
And I I predict that every every company, every open source project will have at its core a software factory.

s125
00:08:52.399 --> 00:08:57.040
kind of like the way that CI/CD became just like, oh, of course you have that.

s126
00:08:57.040 --> 00:09:00.480
Uh maybe I don't know when that happened, 10 years ago.

s127
00:09:00.480 --> 00:09:03.279
Let's tour the t uh the factory floor for a second here.

s128
00:09:03.279 --> 00:09:05.519
So, you're not going to look at this.

s129
00:09:05.519 --> 00:09:06.560
This is too much.

s130
00:09:06.560 --> 00:09:10.320
Uh the the point of this slide is not to have you read the workflow.

s131
00:09:10.320 --> 00:09:21.120
It's that the factory floor is basically a graph um of steps where you are defining like okay, how does software get built for my product?

s132
00:09:21.120 --> 00:09:23.440
And it looks pretty similar for every product.

s133
00:09:23.440 --> 00:09:25.600
Things come in, they flow through.

s134
00:09:25.600 --> 00:09:28.080
Um, they get stuck at certain points.

s135
00:09:28.080 --> 00:09:32.560
Um, and you know, broadly speaking, just to back out a second.

s136
00:09:32.560 --> 00:09:34.160
So there's the inputs.

s137
00:09:34.160 --> 00:09:36.240
The inputs are really ideas.

s138
00:09:36.240 --> 00:09:41.839
Um, the inputs could be coming from your team, they could be coming from your users.

s139
00:09:41.839 --> 00:09:54.000
Um, the inputs themselves tend to come in through certain channels that you should think about as like your task tracker is an obvious one or Slack, your teams, like your communication channels.

s140
00:09:54.000 --> 00:09:57.360
They could come directly from like your terminal or IDE.

s141
00:09:57.360 --> 00:10:03.680
They could come from your monitoring systems, but there's some set of inputs that bring work into the factory.

s142
00:10:03.680 --> 00:10:04.640
There's triage.

s143
00:10:04.640 --> 00:10:06.240
This is a really important step.

s144
00:10:06.240 --> 00:10:18.399
So again, I boil it down to something very simple, but like you want an agent that is looking at issues as they come in and just saying, you know, if this is easy and this is unambiguous,

s145
00:10:18.399 --> 00:10:22.399
just implement it and this is how you can actually get going with a factory.

s146
00:10:22.399 --> 00:10:28.800
If an issue is hard, uh I recommend having uh an agent that produces specs.

s147
00:10:28.800 --> 00:10:32.959
Folks in here using spec driven development, show hands, people follow this.

s148
00:10:32.959 --> 00:10:36.720
Okay, you can do this many different ways, but I think it's very effective.

s149
00:10:36.720 --> 00:10:44.720
The way that uh we do it at Warp that I recommend is having an agent write what we call a product spec and a tech spec.

s150
00:10:44.720 --> 00:10:48.640
Product spec describes the product invariance that you're building towards.

s151
00:10:48.640 --> 00:10:54.000
Techsp spec describes the architecture and the shape of the code.

s152
00:10:54.000 --> 00:10:55.839
Then you have an implementation agent.

s153
00:10:55.839 --> 00:10:59.680
This is basically a coding agent that runs somewhere in the cloud.

s154
00:10:59.680 --> 00:11:00.480
It makes a diff.

s155
00:11:00.480 --> 00:11:03.680
You can use all sorts of coding agents for this.

s156
00:11:03.680 --> 00:11:05.920
You have review.

s157
00:11:05.920 --> 00:11:08.560
This is in many ways the most painful part.

s158
00:11:08.560 --> 00:11:13.600
Like I expect that people are a little bit tired of reviewing a gentic slop.

s159
00:11:13.600 --> 00:11:24.880
Uh I would have an agent do code review first and then it becomes over time like a riskmanagement exercise of like when do you bring in humans to do code review.

s160
00:11:24.880 --> 00:11:28.079
But you want to have a step in here where humans can do it.

s161
00:11:28.079 --> 00:11:31.519
This is a very important step uh for certain types of apps.

s162
00:11:31.519 --> 00:11:33.279
The verification step.

s163
00:11:33.279 --> 00:11:35.920
So this would be things like computer use.

s164
00:11:35.920 --> 00:11:43.600
Uh if you're building a a sort of UI, having the computer actually use the code that the agent produced and producing videos and screenshots.

s165
00:11:43.600 --> 00:11:46.640
CI/CD still use it obviously.

s166
00:11:46.640 --> 00:11:48.240
Uh and then monitoring.

s167
00:11:48.240 --> 00:11:52.320
So agents don't stop in your factory when code is shipped.

s168
00:11:52.320 --> 00:11:54.959
They should observe what's been shipped.

s169
00:11:54.959 --> 00:11:55.920
Is it crashing?

s170
00:11:55.920 --> 00:11:57.680
Is it being used?

s171
00:11:57.680 --> 00:12:04.560
and round and round you go because you take the output of this monitoring step and you feed it back into the top of the factory.

s172
00:12:04.560 --> 00:12:15.200
Now you could try and build this and um I went to a talk earlier that my friend Adam gave where Uber has built an internal version of this um and it's pretty cool.

s173
00:12:15.200 --> 00:12:20.560
I would say for most, you know, most organizations, it really depends where you are.

s174
00:12:20.560 --> 00:12:31.440
You'll be able to build a simple version of this easily, but to build a thing that actually scales is probably like you should probably be focusing on your own product, not building this infrastructure

s175
00:12:31.440 --> 00:12:33.440
because there's a lot of stuff that you end up wanting.

s176
00:12:33.440 --> 00:12:35.040
You're not again not supposed to read this.

s177
00:12:35.040 --> 00:12:36.720
Uh it's just a lot of stuff.

s178
00:12:36.720 --> 00:12:48.320
If you do build it um or if you buy it, you'll end up with something that looks kind of like this, which is um you're going to have a bunch of ways of getting work into your factory.

s179
00:12:48.320 --> 00:12:50.000
That's what's at the top here.

s180
00:12:50.000 --> 00:12:56.560
You're going to have a sort of control plane for figuring out how work gets distributed across your factory floor.

s181
00:12:56.560 --> 00:12:59.360
You're going to have the actual place where the work happens.

s182
00:12:59.360 --> 00:13:02.320
And so that's going to be cloud sandboxes.

s183
00:13:02.320 --> 00:13:05.360
It's going to be figuring out what agent to run.

s184
00:13:05.360 --> 00:13:07.200
So, what's the harness?

s185
00:13:07.200 --> 00:13:08.639
What's the model?

s186
00:13:08.639 --> 00:13:10.959
And then finally, I think this is a really important thing.

s187
00:13:10.959 --> 00:13:15.600
You're going to want to set up some kind of data plane that sits below your factory.

s188
00:13:15.600 --> 00:13:24.160
And so, that's something that lets agents remember what they've done, learn, um, improve over time.

s189
00:13:24.639 --> 00:13:28.560
The factory is not just like a product, it's also a mindset.

s190
00:13:28.560 --> 00:13:33.839
And this brings me back to the thesis I had at the beginning.

s191
00:13:33.839 --> 00:13:36.000
You need to measure and improve.

s192
00:13:36.000 --> 00:13:44.399
So factory, this is where the I don't know, you can stretch this metaphor as far as you want, but like you should be thinking of efficiency.

s193
00:13:44.399 --> 00:13:48.079
And so that means like how much software did you ship?

s194
00:13:48.079 --> 00:13:57.760
How much did it cost in terms of human time and token time and you're going to want to measure this and try to improve it over time.

s195
00:13:57.920 --> 00:14:00.639
A key part of this is creating loops.

s196
00:14:00.639 --> 00:14:04.560
So, uh, loops are, again, they sound complicated.

s197
00:14:04.560 --> 00:14:05.680
They're not that complicated.

s198
00:14:05.680 --> 00:14:14.639
Loops are basically ways of, uh, having agents improve, um, by like observing what they're doing, where they're failing.

s199
00:14:14.639 --> 00:14:19.760
Um, and so, a common kind of loop that you're going to want to put in your factory is like a skill loop.

s200
00:14:19.760 --> 00:14:29.680
That means you're going to have your factory agents that are running skills, and then you'll have observer agents that are seeing how those skills are being applied, looking for issues, and trying to improve the skills.

s201
00:14:29.680 --> 00:14:41.199
So for instance, if you had a code review agent uh and it was leaving comments and a senior engine engineering or team was going and correcting those comments, you'd want an observer agent that would look at that

s202
00:14:41.199 --> 00:14:46.560
and basically uh improve the code review agent for the next run.

s203
00:14:48.079 --> 00:14:52.800
This is one thought just to leave folks with like what is where does this leave engineers?

s204
00:14:52.800 --> 00:14:59.920
Um, I think you're going to have to get into this mindset, and I'm trying very hard to get our team into this mindset.

s205
00:14:59.920 --> 00:15:07.839
It's not always easy that you're not just building the product, but you're building the thing that builds the product.

s206
00:15:07.839 --> 00:15:10.480
And that's like it's just different.

s207
00:15:10.480 --> 00:15:15.040
It's more like process engineering or manufacturing or something like that.

s208
00:15:15.040 --> 00:15:18.800
Um, and you could think like, okay, maybe that's a bummer.

s209
00:15:18.800 --> 00:15:19.839
Like, is that a bummer?

s210
00:15:19.839 --> 00:15:26.560
is that uh you know and it depends like it depends what joy you get out of software engineering.

s211
00:15:26.560 --> 00:15:33.440
If your joy is in writing the code, I think you're everyone in here who is a software engineer is going to be writing less code.

s212
00:15:33.440 --> 00:15:38.320
But if your joy is in shipping product, like it's never been a better time.

s213
00:15:38.320 --> 00:15:40.160
And this is actually where I find my joy.

s214
00:15:40.160 --> 00:15:43.199
It's like I like building and shipping the thing.

s215
00:15:43.199 --> 00:15:46.959
So everyone in here is going to code less, but they're going to ship more.

s216
00:15:46.959 --> 00:15:49.279
And that's going to be a trade-off.

s217
00:15:49.279 --> 00:16:00.240
But if you approach it like you're a a factory engineer, I think you can see like that there's still a really cool set of engineering challenges.

s218
00:16:00.240 --> 00:16:12.480
Um you could almost think of it as like meta-engineering like how do you engineer your system of agents to be the best possible at engineering that I think is a very compelling and interesting set of challenges to solve.

s219
00:16:12.480 --> 00:16:13.279
So that's it.

s220
00:16:13.279 --> 00:16:28.160
Um uh for folks who were interested uh this if you follow the link on this QR code I've set up um a open source GitHub repo where anyone who wants to try building their own factory agents can do it.

s221
00:16:28.160 --> 00:16:32.320
This uses Warps uh agent platform as part of it.

s222
00:16:32.320 --> 00:16:33.839
But you honestly you don't have to use it.

s223
00:16:33.839 --> 00:16:35.759
I'm not trying to like push into our product.

s224
00:16:35.759 --> 00:16:46.480
But this should give you a good sense of like, okay, if you want to set up uh an agent that does uh triage or an agent that does uh specw writing, how do you actually do that?

s225
00:16:46.480 --> 00:16:53.360
How do you get from like the theory of uh working with a factory to actually putting it into practice?

s226
00:16:53.360 --> 00:16:56.800
Um I don't know if we have the capability to do questions in here.

s227
00:16:56.800 --> 00:16:59.519
I saved a few minutes for questions if anyone has questions.

s228
00:16:59.519 --> 00:17:01.839
Otherwise, I will I will wrap up.

s229
00:17:01.839 --> 00:17:03.839
Yes.

s230
00:17:04.880 --> 00:17:10.720
what you said you don't want to build this also build

s231
00:17:10.720 --> 00:17:23.360
it's a great question so I said something that's almost contradictory I think um you should the way you should think of it is like you everyone's going to deploy some sort of factory but then the tuning of the factory

s232
00:17:23.360 --> 00:17:33.440
the like are these the right skills for my domain is this factory building my product in the right way I still think there's a bunch of interesting engineering challenges

s233
00:17:33.440 --> 00:17:45.600
And for some places you can build this, but again I I think I think that the you should probably be focusing on building the core product for your company for the most part, but there's a bunch of like tuning and like

s234
00:17:45.600 --> 00:17:49.120
uh figuring out how to make the factory work for your product that matters.

s235
00:17:49.120 --> 00:17:50.160
That's a great question.

s236
00:17:50.160 --> 00:17:51.360
Yes,

s237
00:17:51.360 --> 00:17:53.520
thank you so much.

s238
00:17:53.520 --> 00:17:55.520
So if you were a college student right now,

s239
00:17:55.520 --> 00:17:56.160
yes,

s240
00:17:56.160 --> 00:18:00.480
graduating and entering the workforce, where would you spend time?

s241
00:18:00.480 --> 00:18:00.960
Yeah.

s242
00:18:00.960 --> 00:18:07.120
So the the question in case people couldn't hear was like if I was a college student graduating and entering the workforce right now.

s243
00:18:07.120 --> 00:18:12.799
So I think that the most important skills in this new world are adaptability.

s244
00:18:12.799 --> 00:18:15.039
I think that that's critical thinking.

s245
00:18:15.039 --> 00:18:18.000
It's like this the speed at which you can learn.

s246
00:18:18.000 --> 00:18:29.679
I do think I don't know if you're a computer science student, but I still think there's a ton of value in understanding like the underlying systems and architecture and being able to reason and understand

s247
00:18:29.679 --> 00:18:34.480
the code, understand the specs that agents are written, sorry, are writing.

s248
00:18:34.480 --> 00:18:37.840
So, I I would focus on on those skills.

s249
00:18:37.840 --> 00:18:42.880
Um, and like I don't know, we're hiring more people than we've ever hired.

s250
00:18:42.880 --> 00:18:48.480
There's a lot of kind of like misdirection around like you know people not being hired because of AI.

s251
00:18:48.480 --> 00:18:50.480
That's not the experience we've had so far.

s252
00:18:50.480 --> 00:19:03.600
And what I'm looking for are like really adaptable product focused thinkers who can like basically be great problem solving problem solvers uh even as the underlying technology changes.

s253
00:19:03.600 --> 00:19:05.440
Yeah, I think I have time for one more question.

s254
00:19:05.440 --> 00:19:07.440
Yes.

s255
00:19:12.480 --> 00:19:22.880
What about the like so the question was what about the the product taste and like the how do you actually build something useful I think is probably the right like maybe the framing and like

s256
00:19:22.880 --> 00:19:36.880
uh where do the ideas come from and so I think that um the problem with the factory metaphor even though I'm like leaning into it because I think that's like that there's something to it is that it can kind of sound like uh mechanizing or dehumanizing

s257
00:19:36.880 --> 00:19:42.799
um and I still think underlying ing all of this, the only thing that matters is like are you building something useful?

s258
00:19:42.799 --> 00:19:49.520
And if you have like a a factory that is like churning out that no one cares about, it's like what's the point?

s259
00:19:49.520 --> 00:20:02.000
And I think that human taste, human input, human product sense, um humans like guiding at those touch points where you can't automate stuff is absolutely like essential.

s260
00:20:02.000 --> 00:20:08.160
And like that's like what I do like I'm trying to figure out what what do customers want, what do people want, what's going to be valuable to them.

s261
00:20:08.160 --> 00:20:11.120
So I think that's an absolutely key point to it.

s262
00:20:11.120 --> 00:20:13.840
I think I'm at time so I I have to go.

s263
00:20:13.840 --> 00:20:16.640
I hope folks uh enjoyed this chat.

s264
00:20:16.640 --> 00:20:18.400
I'm really grateful for being invited to speak.

s265
00:20:18.400 --> 00:20:21.280
So thank you all very much.

s266
00:20:34.689 --> 00:20:36.689
[music]