1
00:00:07,832 --> 00:00:17,312
Welcome back to Adventures in DevOps, podcast host, repeat founder, technical advisor,
conference speaker, and started out as a Microsoft intern, proves that Microsoft even has

2
00:00:17,312 --> 00:00:18,453
its good parts.

3
00:00:18,453 --> 00:00:26,431
It's by no stretch of the imagination to call her tech celebrity, senior director of
developer advocacy, Cassidy Williams.

4
00:00:26,431 --> 00:00:27,702
Welcome to the show.

5
00:00:28,034 --> 00:00:29,778
Hello, wow, thank you for the intro.

6
00:00:29,778 --> 00:00:31,295
Thank you for having me.

7
00:00:31,310 --> 00:00:37,098
You know, I almost didn't say like you apparently were on a documentary as well as the
starring member.

8
00:00:37,454 --> 00:00:42,294
I forget about that so often because that was so many years ago.

9
00:00:42,754 --> 00:00:45,754
I was just like the clown of that documentary, feel like.

10
00:00:45,754 --> 00:00:48,270
But was good times back in the day.

11
00:00:48,270 --> 00:00:55,550
Well, I think part of the problem is if LinkedIn experiences are a proxy for success, then
you have achieved quite a lot there.

12
00:00:55,550 --> 00:00:58,222
And it makes me question what we should even talk about first.

13
00:00:58,222 --> 00:00:59,045
I don't even know.

14
00:00:59,045 --> 00:01:00,689
What would you like to talk about first?

15
00:01:00,689 --> 00:01:02,508
I could talk about most things.

16
00:01:02,508 --> 00:01:10,710
Well, you know, uh last week we did a breakdown of CI, CD systems and their challenges and
found that they all suffer one way or another.

17
00:01:10,710 --> 00:01:18,892
And given your current work is at GitHub, I can only imagine the sorts of constructive
feedback you must be getting from your community about the recent outages.

18
00:01:19,218 --> 00:01:20,008
yes.

19
00:01:20,008 --> 00:01:20,458
yes.

20
00:01:20,458 --> 00:01:27,079
And this is where I'll like put on my work hat and be like, performance is a priority at
GitHub.

21
00:01:27,079 --> 00:01:31,721
We are making sure that it is number one in everything that we ship.

22
00:01:31,721 --> 00:01:34,042
But uh yeah, no, it's been pretty tough.

23
00:01:34,042 --> 00:01:45,745
oh It's one of those things where my work in particular is very much focused on the open
source community making maintainers and individual developers happy with all the features

24
00:01:45,745 --> 00:01:46,575
that we ship.

25
00:01:46,575 --> 00:01:48,846
And it's been exciting to get those shipped.

26
00:01:48,846 --> 00:02:00,046
But as the company is growing so tremendously in terms of just usage, it's wild to be
like, we want to ship this for maintainers.

27
00:02:00,046 --> 00:02:06,126
But also in 2025, we crossed like the 1 billion commit mark, which is amazing.

28
00:02:06,126 --> 00:02:09,166
Like 2025, we had a billion commits on the platform.

29
00:02:09,166 --> 00:02:14,706
Now we're hitting like 275 million a week, which is absurd numbers.

30
00:02:14,706 --> 00:02:17,734
And that scale is so...

31
00:02:17,844 --> 00:02:26,864
astronomical that all of the tech debt that we've seen in the past that normally would be
like, okay, we're going to fix it incrementally, these things here and there.

32
00:02:26,864 --> 00:02:29,006
Now it's just really showing its age.

33
00:02:29,006 --> 00:02:39,667
And so that's something where we have to navigate that while also shipping releases and
hope that the releases can shine at least a little bit when folks are angry about

34
00:02:39,667 --> 00:02:40,908
availability.

35
00:02:41,034 --> 00:02:43,095
See, I think a lot of people are unfair in this regard.

36
00:02:43,095 --> 00:02:52,329
And it's one thing to have a root cause analysis or a public release on problems that
happen, saying like, oops, we had some engineer that changed the policy and that got

37
00:02:52,329 --> 00:02:57,641
deployed and deleted a whole S3 bucket from everyone's AWS account.

38
00:02:57,641 --> 00:03:07,255
It's another thing when I think officially the public story is the lack of capacity and
dealing with how to deal with those sorts of challenges that you get when you have to

39
00:03:07,255 --> 00:03:08,598
scale significantly.

40
00:03:08,598 --> 00:03:18,478
And I feel like it comes at a time when there's a lot more, for better or worse, code and
pull requests being entered into every source of record that exists.

41
00:03:18,478 --> 00:03:27,238
The usage is, the numbers that I cannot say are so astronomical that it's hard for me even
to comprehend.

42
00:03:27,238 --> 00:03:38,638
And it's cool because that means that everything we work on has that kind of impact for a
lot of people, people who haven't been devs before, people who are using agents to run

43
00:03:38,638 --> 00:03:45,038
like fleets of different things to commit and make pull requests and stuff and all that
jazz.

44
00:03:45,078 --> 00:03:48,286
But yeah, it's a lot.

45
00:03:48,286 --> 00:03:59,016
And it's something where it's really nice for me to be able to go speak at conferences and
be in person with developers where they're less angry in person than they are on the

46
00:03:59,016 --> 00:03:59,736
internet.

47
00:03:59,736 --> 00:04:06,062
And so I can kind of hear more reasonable level concerns and feature requests and stuff
and navigate through those.

48
00:04:06,062 --> 00:04:15,570
Whereas if I were to go to the average Twitter feed or something, there will be people
rage baiting for clicks and it's much harder to navigate.

49
00:04:16,126 --> 00:04:25,410
I want to ask you about that because I feel like you've put yourself in a not such a great
position when your customers are super technical and you've chosen to be almost a liaison

50
00:04:25,410 --> 00:04:35,384
for those technical users who historically have their own personal, I want to say
vendettas, but maybe philosophies is the more appropriate term when it comes to how

51
00:04:35,384 --> 00:04:36,715
systems should work.

52
00:04:36,715 --> 00:04:44,138
And I feel like this may be, I mean, you've been in the developer advocacy or bridging
that gap for a long time, feel like, but...

53
00:04:44,138 --> 00:04:49,733
Now you're like in one of the most technical systems out there, which is not like first
class production.

54
00:04:49,733 --> 00:04:58,189
Like we're not talking about cloud providers specifically, but I think everyone knows when
there's something happening with their build system in some regard.

55
00:04:58,189 --> 00:05:01,592
Has that transitioned into the space in GitHub?

56
00:05:01,592 --> 00:05:09,568
Have you seen like a specific kind of, I don't know, eye-opening experience or new sort of
challenges to actually communicate with the users?

57
00:05:09,742 --> 00:05:20,846
I think the main thing that I've been trying to communicate to people is like, if it were
easy, it would be done already for so many of the things that they want.

58
00:05:22,247 --> 00:05:26,269
I even went in literally day one when I joined GitHub.

59
00:05:26,269 --> 00:05:30,571
I went in and made a pull request on something where I was just like, I've been wanting
this for so long.

60
00:05:30,571 --> 00:05:34,132
Shipped it, broke production, had to roll it back really quick.

61
00:05:34,132 --> 00:05:35,857
And it wasn't even that big of a change.

62
00:05:35,857 --> 00:05:38,034
It was just like a small CSS fix.

63
00:05:38,034 --> 00:05:39,146
But I just

64
00:05:39,146 --> 00:05:40,167
I wanted it.

65
00:05:40,167 --> 00:05:45,320
And after breaking production, figuring out what I did wrong, all that, it was day one.

66
00:05:45,320 --> 00:05:46,661
Luckily, people were forgiving.

67
00:05:46,661 --> 00:06:00,762
That was like the first mini lesson in something much bigger in that there's so many
things that seem like it's easy to do, but there's a lot of historical context why things

68
00:06:00,762 --> 00:06:06,830
have been done certain ways that you have to bridge certain gaps where people will come in
saying like,

69
00:06:06,830 --> 00:06:10,123
Well, this feature is in pull requests, but it's not in issues.

70
00:06:10,123 --> 00:06:11,874
But they look the same.

71
00:06:12,034 --> 00:06:14,076
Why can't it just be in both?

72
00:06:14,116 --> 00:06:23,824
When historically, yes, a lot of the back end of both of them are the same, but two
separate front end teams worked on both of those and they have completely different state

73
00:06:23,824 --> 00:06:32,622
management systems and UI libraries that they used for some reason because history and
they're slowly building that bridge between them.

74
00:06:32,622 --> 00:06:35,726
So that way there can be more feature parity between the two.

75
00:06:35,726 --> 00:06:36,826
categories of things.

76
00:06:36,826 --> 00:06:38,606
And that's just an example.

77
00:06:38,746 --> 00:06:43,566
But yeah, I think that's really the main thing where there's a lot of history.

78
00:06:43,566 --> 00:06:52,506
This is a platform that has been serving developers for so many years that changes have to
be incremental and small.

79
00:06:52,506 --> 00:07:00,046
And one of the teams that is really growing a lot internally, they started just by calling
themselves the Tiny Winds team.

80
00:07:00,046 --> 00:07:05,654
Or it was a few people saying, what if we just went through the maintainer feature request
backlog?

81
00:07:05,654 --> 00:07:08,755
And anything that takes like less than two weeks to ship, let's just ship it.

82
00:07:08,755 --> 00:07:15,617
And that started to make such a big difference for so many maintainers where it was little
things here and there.

83
00:07:15,617 --> 00:07:22,339
Like one of them that was shipped recently was sorting notifications between oldest and
newest, instead of from newest to oldest.

84
00:07:22,439 --> 00:07:27,281
Small thing, but big thing, depending on who is asking for it.

85
00:07:27,661 --> 00:07:29,952
That team is truly just taking all of those.

86
00:07:29,952 --> 00:07:31,622
And it's been making such...

87
00:07:31,622 --> 00:07:42,365
strides in small things on the platform that ends up being much bigger overall and
impacting so many people overall that like we're trying to build that love with developers

88
00:07:42,385 --> 00:07:52,628
from that angle while everything else is just trying to keep everything afloat while the
platform is truly just being hammered.

89
00:07:52,628 --> 00:07:58,004
And honestly, I think that the platform is doing as well as it could.

90
00:07:58,004 --> 00:08:09,178
Given that absolutely astronomical, once again, growth that we're seeing, I challenge so
many developers to try to solve some of the problems that our teams are solving, but it's

91
00:08:09,178 --> 00:08:09,789
a lot.

92
00:08:09,789 --> 00:08:10,830
It's a lot.

93
00:08:10,830 --> 00:08:15,170
I think that is worth repeating realistically.

94
00:08:15,170 --> 00:08:21,450
There's this question of, as a user of any service, you're complaining about it in some
regard.

95
00:08:21,450 --> 00:08:24,850
And my question is, do you want that thing to go away?

96
00:08:24,850 --> 00:08:27,510
And for someone else to have to deal with that?

97
00:08:27,510 --> 00:08:29,330
I was like, no, no, I want it to stay.

98
00:08:29,330 --> 00:08:31,966
It's like, well, great, then maybe you should be happy about it.

99
00:08:31,966 --> 00:08:34,589
To just be a little bit nicer to the humans.

100
00:08:34,589 --> 00:08:39,753
Like I understand communicating frustration and we hear the frustrations for sure.

101
00:08:39,753 --> 00:08:43,836
We hear them personally as individuals all the time.

102
00:08:43,836 --> 00:08:55,465
But at the same time when people are bashing, they kind of feel like they're bashing this
faceless organization when there's so many humans being like, I'm working the best that I

103
00:08:55,465 --> 00:08:57,007
can on this specific thing.

104
00:08:57,007 --> 00:09:01,430
And there could be a lot more empathy in socials.

105
00:09:01,452 --> 00:09:05,265
Honestly, we've had quite a few GitHub members.

106
00:09:05,265 --> 00:09:11,270
uh Obviously, they weren't always GitHub people in their whole lives on this podcast and
one day before when I've met quite a few.

107
00:09:11,270 --> 00:09:13,152
And they always seem like great people.

108
00:09:13,152 --> 00:09:15,414
I mean, we're only like 10 minutes in here.

109
00:09:15,414 --> 00:09:17,396
So I don't know you that well yet.

110
00:09:17,396 --> 00:09:19,917
This whole thing could just still turn around.

111
00:09:20,142 --> 00:09:21,665
That could be a real jerk.

112
00:09:22,446 --> 00:09:30,307
uh No, and I don't want to make this whole episode, you know, it's not all about the
struggles you have to deal with, but one of the things that did come up for me earlier on

113
00:09:30,307 --> 00:09:36,666
in my career is I was forced to be also that sort of point person, except my customers
were less technical.

114
00:09:36,666 --> 00:09:42,352
I was working in the healthcare space and I had to explain to them why the system couldn't
do what they want.

115
00:09:42,352 --> 00:09:50,854
And often while I'm sitting there trying to like rehearse what my retort is for why this
incredibly critical piece of functionality

116
00:09:50,930 --> 00:10:00,874
Needs to be three pixels over on the screen and can't be moved I just feel like a total
clown like I know it's better for them and after doing that for almost a year I'm just

117
00:10:00,874 --> 00:10:04,076
like that's it no more like I Lying doesn't make sense.

118
00:10:04,076 --> 00:10:13,710
I'm just gonna be honest with them like I have no idea why this is so can't do exactly
that you want it's the right thing it's the right thing from my perspective and I just

119
00:10:13,710 --> 00:10:18,454
know that's right here's are way less forgiving that uh healthcare IT admin

120
00:10:18,454 --> 00:10:20,275
Because they're just like, but I could do it.

121
00:10:20,275 --> 00:10:21,676
And I'm like, sure.

122
00:10:21,676 --> 00:10:22,517
Yes.

123
00:10:22,517 --> 00:10:22,997
Great.

124
00:10:22,997 --> 00:10:25,159
There's a lot at play.

125
00:10:25,159 --> 00:10:28,041
Nothing is ever simple.

126
00:10:28,041 --> 00:10:30,843
yeah, it's been an interesting challenge.

127
00:10:30,843 --> 00:10:38,718
We definitely kind of rotate on the team who's the punching bag of the day, depending on
which thing uh went viral or something.

128
00:10:38,718 --> 00:10:40,929
But it's not all bad.

129
00:10:40,950 --> 00:10:47,138
The fact that people approach us so regularly being like, I got a job because of this or

130
00:10:47,138 --> 00:10:54,963
my project exploded because of this or I'm able to change the trajectory of my career
because of this, that stuff keeps us going.

131
00:10:54,963 --> 00:11:05,759
And it's been really, really exciting to see people be thrilled with different releases
and with different things that they're able to be enabled to do because of our work.

132
00:11:05,759 --> 00:11:09,661
And so that fully makes it worth it, where there's highs and lows.

133
00:11:09,661 --> 00:11:15,474
Sometimes the internet is more angry than other times, but it's not all bad.

134
00:11:15,530 --> 00:11:24,934
So one thing I did notice is recently on the GitHub blog of which I assume there's just
thousands and thousands of articles in there.

135
00:11:24,934 --> 00:11:30,556
I did stumble across something that you wrote which was almost like a throwaway blog post
on typed languages.

136
00:11:30,616 --> 00:11:35,498
And uh I say throwaway because it was a little bit like uh from Ma's last theorem.

137
00:11:35,498 --> 00:11:44,984
If you're familiar, it's like this uh in one of his notebooks in the margin, it said like
A to the X plus B to the X uh equals C to the X.

138
00:11:44,984 --> 00:11:48,556
there are no values greater than two that it is valid for.

139
00:11:48,616 --> 00:11:57,341
And it's like, it took us like, I think almost a hundred years to actually prove that the
only answers are like Pythagorean theorem, uh where it's a square value or one.

140
00:11:57,481 --> 00:12:06,286
And when I was coming across your article on type languages, it did feel like this secret
genius that was written in the margin of a notebook somewhere that only seems obvious

141
00:12:06,286 --> 00:12:10,268
after you like actually have accumulated lots of significant experience on the subject
matter.

142
00:12:10,268 --> 00:12:13,132
And I think there's a lot of maybe interesting things to be said about.

143
00:12:13,132 --> 00:12:21,749
like where type languages are going and whether or not we're accelerating towards that or
whether or not there's a whole branch of computer science or programming that is left

144
00:12:21,749 --> 00:12:23,286
untouched that we're going to get into.

145
00:12:23,286 --> 00:12:32,250
Yeah, I feel like I love that you brought up this blog post because it was kind of one
that I wrote and I was like, that was quiet because it was over the holidays and so fun.

146
00:12:32,250 --> 00:12:38,604
Yeah, so for those who didn't read said blog post, we could link it in the show notes.

147
00:12:38,604 --> 00:12:39,534
Yeah, for sure.

148
00:12:39,534 --> 00:12:40,954
All links will be in the description.

149
00:12:40,954 --> 00:12:42,478
So someone will see it for sure.

150
00:12:42,478 --> 00:12:53,657
uh Very long story short, looking at the data, typed languages are on the rise and in a
significant way, like by a fairly large margin.

151
00:12:53,657 --> 00:13:06,878
And I feel like the main reason behind that is because so much code is being written by
non-humans now, we need more languages and tools that have guardrails already built in.

152
00:13:06,878 --> 00:13:11,712
And types are a very simple guardrail that you can have where me personally as a human,

153
00:13:11,712 --> 00:13:12,962
I prefer JavaScript.

154
00:13:12,962 --> 00:13:17,144
just want like sometimes the code to get out of my way so I can make something.

155
00:13:17,524 --> 00:13:22,606
But I understand the value of types because then linting is a little bit stronger.

156
00:13:22,606 --> 00:13:25,587
You get some more explicit error messages and stuff.

157
00:13:25,587 --> 00:13:31,419
And you need that when you don't necessarily know what code is being shipped line by line.

158
00:13:31,419 --> 00:13:36,040
And I think that's a very high level summary of what that post is about.

159
00:13:36,258 --> 00:13:36,909
Yeah, for sure.

160
00:13:36,909 --> 00:13:39,462
And I think the interesting thing here is I'm with you.

161
00:13:39,462 --> 00:13:43,217
When I discovered JavaScript, was just a little bit after the jQuery period.

162
00:13:43,217 --> 00:13:47,213
So I didn't have to go through that horrific existence.

163
00:13:47,213 --> 00:13:47,863
started.

164
00:13:47,863 --> 00:13:49,788
I kind of miss jQuery.

165
00:13:50,131 --> 00:13:51,634
Things are so simple.

166
00:13:52,086 --> 00:13:58,198
I don't think there's ever a time where I'm like, I'm so glad this existed, but I was not
like a UI developer in any regard.

167
00:13:58,198 --> 00:14:02,589
I liked my C++, anything programming on physical hardware.

168
00:14:02,589 --> 00:14:04,190
was hardware descripting language.

169
00:14:04,190 --> 00:14:09,131
Like I would program FPGAs uh for uh CPUs, et cetera.

170
00:14:09,131 --> 00:14:09,851
Yeah.

171
00:14:09,851 --> 00:14:19,774
So the interesting thing is that I think I totally agree with you that because we're
entering a world where more software is written, whether or not more software is kept,

172
00:14:19,774 --> 00:14:21,744
or more software products are created?

173
00:14:21,744 --> 00:14:27,156
I think it's a separate question, but definitely for sure more software is being written
for better or for worse.

174
00:14:27,156 --> 00:14:34,959
And there is a reliability problem, not with what's actually being produced, but the code
quality or safety that's baked into it.

175
00:14:34,999 --> 00:14:44,264
And while I much preferred JavaScript early on in my life, in the last 10 years, I started
an experiment where I went through and found every single bug that...

176
00:14:44,264 --> 00:14:53,072
I caused to enter production and found like a majority of them would have been solved
with, well, not just type languages, but type languages that had certain extra guardrails

177
00:14:53,072 --> 00:14:54,042
in place.

178
00:14:54,903 --> 00:15:03,554
And that sort of like completely eliminated whatever preference I had for any sort of
scripting language that didn't make types a required aspect.

179
00:15:03,554 --> 00:15:04,195
Right.

180
00:15:04,195 --> 00:15:13,761
It's interesting how that has also changed my tune, where so many of my projects that I've
made for myself just side projects in the past.

181
00:15:13,761 --> 00:15:15,573
It's me saying like, get out of my way.

182
00:15:15,573 --> 00:15:17,564
I just want to make this app exist.

183
00:15:17,564 --> 00:15:18,345
Yay.

184
00:15:18,345 --> 00:15:22,698
And because it's just me working on it, I know how to test my own code.

185
00:15:22,698 --> 00:15:23,688
It'll be fine.

186
00:15:23,688 --> 00:15:27,901
And now that, yeah, I'm starting to use AI more.

187
00:15:27,901 --> 00:15:30,253
playing with agents more and things like that.

188
00:15:30,493 --> 00:15:31,894
I'm realizing, mm.

189
00:15:31,894 --> 00:15:37,406
It would be nice to have some of these errors handled when I don't know all of the ins and
outs of things.

190
00:15:37,406 --> 00:15:42,887
yeah, I'm realizing now like, wow, types actually are good, aren't they?

191
00:15:42,887 --> 00:15:47,519
They aren't just annoying, which is a joke for everybody listening.

192
00:15:47,519 --> 00:15:49,219
That is called humor.

193
00:15:49,219 --> 00:15:59,582
But yeah, it's been interesting to see that shift where like my opinions in different
regards have been changed because

194
00:15:59,818 --> 00:16:02,247
we need to be able to be flexible with the times.

195
00:16:02,247 --> 00:16:03,974
Things are changing constantly.

196
00:16:03,974 --> 00:16:06,561
cannot be loyal to a language or to a tool.

197
00:16:06,985 --> 00:16:09,513
I think that's what's known as personal growth.

198
00:16:09,958 --> 00:16:11,027
Yes.

199
00:16:11,027 --> 00:16:11,969
Gross.

200
00:16:13,282 --> 00:16:22,481
You know when Russ came out I was not like the biggest fan like I mean the complexity
wasn't was just something that would stop me personally but over time I definitely learned

201
00:16:22,481 --> 00:16:30,528
to lean on it more because there's just like a whole class of problems that goes away and
The current idea that I have floating around in my head is this one of formal

202
00:16:30,528 --> 00:16:31,738
verification.

203
00:16:31,799 --> 00:16:35,566
I don't know if this is Terminology you've you've run into before

204
00:16:35,566 --> 00:16:38,706
I, maybe, but I'd like you to continue.

205
00:16:38,706 --> 00:16:40,926
Why don't you tell me what you think it means?

206
00:16:41,426 --> 00:16:42,446
If you insist.

207
00:16:42,446 --> 00:16:52,379
uh So the idea is that we can't be sure whether the semantics of the language or the
function that we're writing is absolutely correct and types help us do that.

208
00:16:52,379 --> 00:16:52,789
Right.

209
00:16:52,789 --> 00:16:59,411
But at some point, if everything becomes typed that we have, there's still a class of bugs
that we can run into.

210
00:16:59,411 --> 00:17:00,702
And one of them is simple.

211
00:17:00,702 --> 00:17:08,584
Like I write a function, how can I be sure without writing every unit test possible that
the results of the function are actually accurate?

212
00:17:08,830 --> 00:17:17,427
And a good example is like if you have a difference function like a minus b where a is
greater than b, you should be able to conclude that for all values of a greater than b,

213
00:17:17,427 --> 00:17:19,238
the result is always positive.

214
00:17:19,318 --> 00:17:24,683
And if you can't imagine there being a set of unit tests, you can write to prove that.

215
00:17:24,683 --> 00:17:34,156
The challenge is that realistically, you want uh a language that you can actually validate
that the difference in this case could always be positive.

216
00:17:34,156 --> 00:17:36,727
Some function, no matter what the inputs are, actually work.

217
00:17:36,727 --> 00:17:38,548
And this is what formal verification actually does.

218
00:17:38,548 --> 00:17:40,969
It allows you to write guarantees about it.

219
00:17:40,970 --> 00:17:49,314
And in the world of LLMs, I think the current thought is we'll shift everything to type
languages because more software is being written and we can have guarantees about it.

220
00:17:49,314 --> 00:17:56,178
But over time, I think we're going to realize there's going to be another optimization on
top of that, which is how do we make sure that the output is actually semantically

221
00:17:56,178 --> 00:17:56,962
accurate?

222
00:17:56,962 --> 00:18:01,867
wonder if we will eventually see a shift to functional languages as a result.

223
00:18:02,702 --> 00:18:14,122
I want to say, so it was interesting because I think it was a couple episodes ago, we were
actually talking about, we had one of the foremost Haskellers on and he was talking about

224
00:18:14,122 --> 00:18:18,122
writing the library for Othel and some other fancy frameworks.

225
00:18:18,122 --> 00:18:24,122
And basically there is this aspect where you can get quite into the complexity of how you
define your types.

226
00:18:24,122 --> 00:18:28,622
Basically, you can have a language that has so complicated or complex

227
00:18:28,622 --> 00:18:35,006
uh type semantics that you can write like the length of the string must be between 10
characters and 12 characters.

228
00:18:35,006 --> 00:18:42,390
And this is great because you can make a whole bunch of assumptions about the amount of
data that you can save in a particular database column if it's limited.

229
00:18:42,390 --> 00:18:48,283
But on the flip side, it becomes incredibly challenging to actually reason about in
practice what you're looking at.

230
00:18:48,283 --> 00:18:56,694
And I think anyone that's written any sort of TypeScript can sort of see from their
experience, like you look at the types in TypeScript and you're like, I have no idea.

231
00:18:56,694 --> 00:19:01,310
what invariant this actually captures because I have no idea what this actually is saying.

232
00:19:01,568 --> 00:19:07,942
Yeah, yeah, it's a different kind of set of problems.

233
00:19:07,942 --> 00:19:20,559
I've been thinking about like, kind of what you're saying, but not with this naming in
terms of like, if so much software is going to be written by machines and read by humans

234
00:19:20,559 --> 00:19:29,174
less and less, possibly, probably, who knows where we'll be in five years, let's circle
back then, is a more like...

235
00:19:29,774 --> 00:19:38,241
machine-y level language where it's either super low level or super, super functional
where like we look at it, we're like, what does this even mean?

236
00:19:38,241 --> 00:19:39,713
But a machine is like, perfect.

237
00:19:39,713 --> 00:19:41,724
It's just math.

238
00:19:42,966 --> 00:19:45,447
where will that balance be?

239
00:19:45,753 --> 00:19:48,610
I don't know the answer to that question.

240
00:19:48,664 --> 00:19:57,932
So do you feel yourself reaching for a particular language right now and do you think that
your choice or preference is going to change as the years go on?

241
00:19:57,932 --> 00:20:01,004
What's interesting is no, not at all.

242
00:20:01,004 --> 00:20:07,899
Like there have been a couple things that I've experimented with now that I haven't
experimented with before.

243
00:20:07,899 --> 00:20:12,302
Like I've written Rust for fun pre-AI.

244
00:20:12,302 --> 00:20:17,776
I've written various Python scripts and JavaScript scripts and all of the different
things.

245
00:20:17,776 --> 00:20:27,052
But recently, for example, I was just like, you know, I've never made a Go CLI and I'd
like to play with these libraries, but I don't know Go.

246
00:20:27,052 --> 00:20:28,743
But the machines do.

247
00:20:28,743 --> 00:20:44,036
And so I just kind of told some, I was using the GitHub Copilot CLI, hashtag shill, said I
want to use this CLI, or I want to use these libraries in the CLI to make this CLI tool.

248
00:20:44,036 --> 00:20:46,938
And it made something that worked.

249
00:20:46,938 --> 00:20:51,382
And I admit, I was looking at the code, was like, I generally know how this works.

250
00:20:51,382 --> 00:20:53,463
Could I write this from scratch?

251
00:20:53,524 --> 00:20:56,450
Maybe if I had a lot of docs and experience.

252
00:20:56,450 --> 00:20:57,730
Does that matter?

253
00:20:57,730 --> 00:21:00,061
It's a very existential question.

254
00:21:01,134 --> 00:21:09,632
Do you feel like having written that, that you actually created it or that it was just
something that you then owned?

255
00:21:09,632 --> 00:21:14,456
You know that meme where it's someone on the internet saying like, look what I made.

256
00:21:14,456 --> 00:21:19,169
And then like they give the object to someone and the person holding the object is like, I
made this.

257
00:21:19,169 --> 00:21:21,171
Like that meme comic.

258
00:21:21,171 --> 00:21:24,874
I feel like there's an element of that with AI tools where people are just like, look what
I made.

259
00:21:24,874 --> 00:21:26,615
I'm like, but did you?

260
00:21:26,615 --> 00:21:28,096
But also kind of.

261
00:21:28,096 --> 00:21:35,141
I feel like with some of the tools that I've worked on, note how I'm saying worked on
rather than made.

262
00:21:35,141 --> 00:21:36,334
uh

263
00:21:36,334 --> 00:21:44,781
Some of the tools that I've worked on, I say that I did make it because I put in so much
significant effort in addition to the AI helping me along.

264
00:21:44,781 --> 00:21:50,900
Some tools, one that I'm editing right now is like an electron app that I started, think
literally in 2016.

265
00:21:50,900 --> 00:21:56,752
So about 10 years ago was when I first started this app and I handcrafted every line of
code.

266
00:21:56,752 --> 00:22:02,176
And there's been a handful of issues in my repo that I haven't touched because it's
painful.

267
00:22:02,236 --> 00:22:04,310
And literally yesterday I was like, what if I just

268
00:22:04,310 --> 00:22:08,231
what if I just let go of control a little bit and let the AI do some of it?

269
00:22:08,531 --> 00:22:09,692
It wasn't perfect.

270
00:22:09,692 --> 00:22:11,132
I still had to tweak it.

271
00:22:11,132 --> 00:22:16,554
And I did sit there thinking like, did I make this feature or did I just let a bot do it?

272
00:22:16,554 --> 00:22:18,914
But also at the same time, I did refine it.

273
00:22:18,914 --> 00:22:21,515
It's a very weird balance.

274
00:22:21,515 --> 00:22:26,166
And once again, this is where like a lot of people might say, but doesn't matter.

275
00:22:26,166 --> 00:22:29,813
And that's a good question.

276
00:22:29,813 --> 00:22:31,198
I don't know if it does matter.

277
00:22:31,198 --> 00:22:32,518
feel like we're...

278
00:22:32,984 --> 00:22:37,636
collectively as an industry having a bit of an identity crisis in that regard.

279
00:22:38,086 --> 00:22:40,838
I mean, I think you're definitely onto something there in that regard.

280
00:22:40,838 --> 00:22:51,607
So one of the things is that we still have opinions about the technology that we can
create like individually with our own brains and that technology that we make or someone

281
00:22:51,607 --> 00:22:57,332
else makes independently using the tools that we think they, how they work and have
assumptions about what actually comes out there.

282
00:22:57,332 --> 00:23:00,604
So when I'm using an LLM, maybe I have assumptions about what it can make.

283
00:23:00,925 --> 00:23:07,682
And I think right now fundamentally those things are still different enough to the point
where I feel like I can

284
00:23:07,682 --> 00:23:11,170
have those opinions, but at some point, I think that's gonna shift.

285
00:23:11,170 --> 00:23:18,556
So much so that I will say, I'm gonna go on a limb here, that in a couple years, we are
going to get artisanal software engineering.

286
00:23:19,598 --> 00:23:24,059
Yeah, handcrafted websites made just for you.

287
00:23:24,059 --> 00:23:29,761
Yeah, I feel like I'm seeing some of that right now, honestly.

288
00:23:29,841 --> 00:23:40,264
it's a really interesting thing, uh phenomenon, I guess you could say, where I have a big
list of side projects on my computer where I pull from it sometimes where I'm I'm

289
00:23:40,264 --> 00:23:42,124
interested in making something.

290
00:23:42,124 --> 00:23:48,256
there's some of them that I really care about existing, and I have not started touching
them yet because I'm like, I

291
00:23:48,256 --> 00:23:49,456
I want to make it myself.

292
00:23:49,456 --> 00:23:50,907
It's my project.

293
00:23:50,907 --> 00:23:57,970
But also, if I just use these AI tools, it would be alive right now and I could actually
be using the project.

294
00:23:58,090 --> 00:24:05,473
And it's a really weird thing where I want to make it, but also maybe I should just use
the tools so that way I can make it exist.

295
00:24:05,473 --> 00:24:13,136
And some of the projects I have made exist, some of them still feel precious to me where
I'm squatting on the domain name and I should just ship it.

296
00:24:13,136 --> 00:24:17,958
And it's been interesting to talking to people who aren't.

297
00:24:17,976 --> 00:24:22,567
technical or don't have a technical background and how they're just starting to ship
things a lot.

298
00:24:22,567 --> 00:24:29,509
Where we're seeing that a ton internally, but even just a neighbor of mine, um I'm going
to just tell this story of her.

299
00:24:29,509 --> 00:24:32,330
I introduced her to agents in general.

300
00:24:32,330 --> 00:24:37,551
She had used ChatGPT a little bit before, but that was kind of the extent of her usage.

301
00:24:37,551 --> 00:24:39,652
She does not work in tech.

302
00:24:39,652 --> 00:24:43,693
She is very artistic, fully, fully separate from our industry.

303
00:24:43,693 --> 00:24:45,833
And I introduced her to

304
00:24:47,790 --> 00:24:51,442
uh Zo Computer, which is kind of like a hosted open clause sort of thing.

305
00:24:51,442 --> 00:24:56,115
It's this online computer, agented computer thing.

306
00:24:56,575 --> 00:25:01,078
She got so into it so quickly and was like, I can make an app for this, for this, for
this.

307
00:25:01,078 --> 00:25:02,859
it was amazing to see her ship.

308
00:25:02,859 --> 00:25:06,171
But then at the same time, she would ask me for feedback on her apps.

309
00:25:06,171 --> 00:25:09,443
And I'd be like, it's amazing that you got this far.

310
00:25:09,443 --> 00:25:11,164
It is, it does exist.

311
00:25:11,164 --> 00:25:14,136
But there's so many holes, but it does exist.

312
00:25:14,136 --> 00:25:22,693
But at the same time, she asked me for like specific details and I was like, okay, I can
see that your password is admin123 by just inspecting the console.

313
00:25:22,693 --> 00:25:24,575
You got to have these guardrails in place.

314
00:25:24,575 --> 00:25:33,142
She was talking to me about like bun runtime, which is wild for someone who knows nothing
about tech, but at the same time, she doesn't know what an array is.

315
00:25:33,142 --> 00:25:42,912
Like, it's crazy to see all of this usage of tools and people shipping from all these
different backgrounds where I feel like...

316
00:25:42,912 --> 00:25:48,305
A lot of engineers are accelerated and shipping so many different things and able to push
stuff.

317
00:25:48,305 --> 00:25:55,299
A lot of them are paralyzed, kind of like me in a lot of my own projects where it's like,
can do so many things.

318
00:25:55,299 --> 00:25:56,340
I want to do some things.

319
00:25:56,340 --> 00:25:59,592
I'm nervous about doing other things, all of that.

320
00:25:59,592 --> 00:26:04,675
And then there's some people where they're just taking it and going and being super
creative with it.

321
00:26:04,922 --> 00:26:08,187
yeah, it's wild to see just the differences of everything.

322
00:26:08,187 --> 00:26:10,038
And I feel like we're going to...

323
00:26:10,378 --> 00:26:13,611
look back on this time right now and be like, man, you knew nothing.

324
00:26:13,611 --> 00:26:19,446
And in a few years, our roles are going to be looking incredibly different.

325
00:26:19,722 --> 00:26:22,503
I don't know if that's optimism or pessimism.

326
00:26:22,638 --> 00:26:23,721
That's little bit of both.

327
00:26:23,721 --> 00:26:27,508
feel like I'm like both nervous and excited.

328
00:26:27,658 --> 00:26:33,702
I like coding, but also I like building and I feel like those are sometimes the same and
sometimes different.

329
00:26:33,720 --> 00:26:34,360
More so now.

330
00:26:34,360 --> 00:26:42,733
And I feel like you're also in a particular role where you're connecting over the
collaboration, over a communication mechanism with other humans.

331
00:26:42,733 --> 00:26:51,795
Is that role, like, you fearful for that role that I assume you enjoy doing going away in
the near future because you're, like, the people who are building aren't going to

332
00:26:51,795 --> 00:27:01,440
necessarily be technical or have a clear understanding of uh technical aspects, which
you're the expert or working for a company where...

333
00:27:01,440 --> 00:27:07,114
you need to communicate along that sort of product architecture or tool set.

334
00:27:07,114 --> 00:27:16,420
In a weird way, and I don't mean to be like, my job's amazing, but in a weird way, I feel
like more communication heavy roles are going to be more valuable than ever.

335
00:27:16,420 --> 00:27:28,867
Because now if anybody can quote unquote code and ship things, first of all, I feel like
mentorship is more important than ever in general, because if everybody is hiring senior

336
00:27:28,867 --> 00:27:34,614
engineers and people who can use AI and stuff, who will the future senior engineers of
tomorrow be?

337
00:27:34,614 --> 00:27:42,077
And we need to be able to kind of lift people up, but then also learn via teaching on the
technical side.

338
00:27:42,077 --> 00:27:51,461
But at the same time, these junior folks entering the industry are going to be moving
faster than ever and learning faster than ever.

339
00:27:51,461 --> 00:27:55,963
So that level of communication from like a purely engineering perspective is important.

340
00:27:56,103 --> 00:28:04,236
The growth that I've seen for developer advocacy and like technical marketing and that
type of external communication.

341
00:28:04,436 --> 00:28:17,695
It has been growing tremendously from my perspective, just from what I've seen around
hiring boards and forums and Discord groups and just trends that I've seen, where I feel

342
00:28:17,695 --> 00:28:27,262
like a few years ago when all of the massive layoffs were happening and stuff, a lot of
people were leaving those kinds of roles and leaning just into the engineering side and

343
00:28:27,262 --> 00:28:29,083
doing some career pivots.

344
00:28:29,083 --> 00:28:33,710
And now I feel like the pendulum is swinging in the other direction where now that

345
00:28:33,710 --> 00:28:44,118
quote unquote, everyone can code, we need people who can communicate both to humans and to
the agents of the world and be able to communicate concepts and teach concepts more than

346
00:28:44,118 --> 00:28:44,938
ever.

347
00:28:45,419 --> 00:28:59,060
And the LinkedIn inboxes for everyone I know in this type of role are on fire right now
because the AI companies are realizing it too, where you can't necessarily use AI for all

348
00:28:59,060 --> 00:28:59,522
of your...

349
00:28:59,522 --> 00:29:04,610
communication marketing and communication teaching and education and all of that.

350
00:29:04,610 --> 00:29:10,338
You need a person to be able to explain to other people how to use these tools.

351
00:29:10,402 --> 00:29:19,729
Well, that's like directly opposite of the marketing documentation on all of the AI tools
out there that suggest I can just fire my whole team, technical or not, and replace it

352
00:29:19,729 --> 00:29:20,940
with agents.

353
00:29:21,164 --> 00:29:25,217
Yeah, everyone thinks that other roles don't matter anymore.

354
00:29:25,818 --> 00:29:27,259
But that's not true.

355
00:29:27,259 --> 00:29:35,205
I feel like it's now we can abstract away some of the boring stuff and repetitive stuff
and do more of our specialized skills.

356
00:29:35,205 --> 00:29:36,646
That's what it feels like.

357
00:29:37,134 --> 00:29:37,994
I like your framing.

358
00:29:37,994 --> 00:29:47,859
It's something that almost, I think it was over 20 years ago now, where one of my
professors had said that the industry and corporations, especially enterprises, want

359
00:29:47,859 --> 00:29:50,920
engineers who can do at least one other thing.

360
00:29:50,920 --> 00:29:54,321
And I didn't really understand what that meant at the time.

361
00:29:54,321 --> 00:29:57,743
And it's sort of something that has stuck with me even after so long.

362
00:29:57,743 --> 00:29:59,863
I don't have that good of a memory, I guess.

363
00:29:59,904 --> 00:30:01,905
And it does, feel like, come to light more.

364
00:30:01,905 --> 00:30:05,304
Like, who am I outside of being able to be a

365
00:30:05,304 --> 00:30:08,195
technologist in any way, you know, where are my skills?

366
00:30:08,195 --> 00:30:17,538
I do think I got a little bit of advantage given my non-historical path, like I didn't
major in computer science, I didn't start out as a software developer or build websites,

367
00:30:17,538 --> 00:30:27,341
so I feel like I avoided some of the canonical, I'll say failure modes, where you spend
your whole life coding in some regard, maybe backends, dealing with support or tickets,

368
00:30:27,341 --> 00:30:31,422
not even support tickets, and now you may be questioning what you're doing.

369
00:30:31,434 --> 00:30:36,227
Yeah, it's very interesting to see that in general.

370
00:30:36,227 --> 00:30:47,803
Like all of the best engineers that I know and people who are not even just engineers, but
technical in some way using AI tools, grand majority of them don't have a traditional

371
00:30:47,803 --> 00:30:48,853
computer science background.

372
00:30:48,853 --> 00:30:56,477
Like I do have a traditional background and I feel like I'm in the minority these days
with all of the people who are really accelerating their work.

373
00:30:56,477 --> 00:30:59,599
And that's been interesting to see and very cool, honestly.

374
00:30:59,599 --> 00:31:01,510
And I feel like that is

375
00:31:01,546 --> 00:31:06,608
an important thing and feels like a move in the right direction.

376
00:31:06,608 --> 00:31:09,150
I feel like people should do more things.

377
00:31:09,150 --> 00:31:18,074
um And my very, very, very, very idealist take is if we're more productive than ever,
should we have four-day work weeks?

378
00:31:18,114 --> 00:31:19,965
We should be working less days, right?

379
00:31:19,965 --> 00:31:21,535
Because we're shipping more.

380
00:31:21,535 --> 00:31:27,298
But I know that that is not one that the capitalist machine would like to hear.

381
00:31:27,348 --> 00:31:29,351
You're opening the can of worms there.

382
00:31:29,351 --> 00:31:33,218
is this book, uh I think it's a four hour work week.

383
00:31:33,218 --> 00:31:34,246
ah

384
00:31:34,311 --> 00:31:35,634
I will say.

385
00:31:35,990 --> 00:31:36,951
Yeah.

386
00:31:36,951 --> 00:31:39,022
So not a new concept, realistically.

387
00:31:39,022 --> 00:31:40,613
And it doesn't matter what the research says.

388
00:31:40,613 --> 00:31:48,637
So there is a dedicated episode that we made, ah I think, at the beginning of the year
called, what is productivity even?

389
00:31:48,637 --> 00:31:56,352
Because I feel like every time someone brings up term productivity, they're hard pressed
to actually identify, like, what is that thing that's happening?

390
00:31:56,352 --> 00:32:03,852
And I don't want to spoil the episode, what we chatted about there, but it definitely felt
very subjective as far as determining what that thing is.

391
00:32:03,852 --> 00:32:08,476
What I do want to ask you about is maybe to just change the topic a little bit.

392
00:32:08,476 --> 00:32:14,712
Uh, you did mention basically the role potentially changing as it's going forward and
something that is in huge need.

393
00:32:14,712 --> 00:32:24,892
so my question will be, if you see the role changing and that it's in a huge need, are
there things that you're already doing to prepare for the future and how you see the role

394
00:32:24,892 --> 00:32:25,832
evolving?

395
00:32:26,058 --> 00:32:31,620
I feel like the biggest skill anyone can have right now is flexibility.

396
00:32:31,740 --> 00:32:36,582
like, don't be too attached to all the things because everything is changing.

397
00:32:36,582 --> 00:32:38,763
Just everything's changing so much.

398
00:32:38,763 --> 00:32:43,035
I feel like every talk I've given recently is like, hey, so things are changing.

399
00:32:43,035 --> 00:32:43,705
Are you okay?

400
00:32:43,705 --> 00:32:55,808
uh And having that flexibility and willingness to learn things, I feel like we used to say
that was a big thing in tech and we got comfortable and now it's more true than ever.

401
00:32:55,808 --> 00:33:03,583
And yeah, in general, I feel like we just need to be willing to learn and be flexible with
the way the waves go.

402
00:33:03,583 --> 00:33:08,406
And like you said earlier, I'm sure there's going to be a lot of like, make our to snow
websites here.

403
00:33:08,406 --> 00:33:10,367
We do not use these tools.

404
00:33:10,488 --> 00:33:11,829
I think that's okay.

405
00:33:11,829 --> 00:33:18,393
And I don't think necessarily people will be left behind if they don't use certain tools.

406
00:33:18,393 --> 00:33:25,888
But at the same time, you shouldn't be completely, fully AI-pilled or like an AI skeptic.

407
00:33:26,011 --> 00:33:30,547
or anything like that for anything because we don't know what the future is going to look
like.

408
00:33:30,547 --> 00:33:40,822
I think we have power in shaping it, but being super strongly in one way or another could
only be detrimental when everything is changing so much and we don't have a lot of control

409
00:33:40,822 --> 00:33:42,360
over what the future will be.

410
00:33:42,360 --> 00:33:52,297
Do you feel like there's a particular area that you're focused on personally, like as far
as like areas where, you know, you're currently weak and you see that given the overall

411
00:33:52,297 --> 00:34:00,362
topology of the industry changing over time that you want to adjust yourself to, I don't
know, meet it where you expect it to be going.

412
00:34:00,482 --> 00:34:02,363
That's a good question.

413
00:34:02,363 --> 00:34:17,144
think honestly right now, what I've been personally dealing with is seeing how I'm
frustrated that rage sells in terms of how products are shipped and gone viral and so many

414
00:34:17,144 --> 00:34:17,614
things.

415
00:34:17,614 --> 00:34:27,801
everybody posts rage bait and makes enemies and gets loud and then they raise tons of
money or they get tons of views or anything like that.

416
00:34:28,525 --> 00:34:38,823
Personally for me, I'm trying to figure out how do I work against that current and still
learn and still be able to make things that are interesting for people both at work at

417
00:34:38,823 --> 00:34:44,898
GitHub but also on my personal projects and stuff and be less mean, guess.

418
00:34:44,898 --> 00:34:51,964
It's kind of a wishy-washy answer to that, but it's something that I'm really struggling
with seeing.

419
00:34:52,745 --> 00:34:54,612
Every single person that...

420
00:34:54,612 --> 00:35:02,430
Once again, we're listening to constantly, we're listening on socials and we're looking at
how things are getting shaped.

421
00:35:02,430 --> 00:35:05,573
It's literally all just people being upset at something.

422
00:35:05,573 --> 00:35:07,555
Too much AI or too little AI.

423
00:35:07,555 --> 00:35:09,797
Good AI, not good enough AI.

424
00:35:09,797 --> 00:35:14,302
Like, this model is dead, this one's the best one now.

425
00:35:14,302 --> 00:35:18,956
There's so much of those upset definitive statements.

426
00:35:18,956 --> 00:35:27,648
like software engineering is dead, software engineering is alive, but marketing is dead,
things like that where I'm like, it's not any of these things.

427
00:35:27,870 --> 00:35:36,573
Is there any way to make the non-black and white, the gray, peek through and be a little
colorful and fun?

428
00:35:36,573 --> 00:35:48,378
I'm genuinely hopeful that in these new eras, we're going to see more personal tools and
more like artistic things coming out of it as a result because

429
00:35:48,494 --> 00:35:59,581
humans who were once stopped by not being able to do certain things because of technical
ability or because of like certain reach, they suddenly have the freedom of making what

430
00:35:59,581 --> 00:36:01,392
they want in an artistic way.

431
00:36:01,392 --> 00:36:06,034
And I use artistic to be like in a creative human expressive way.

432
00:36:06,034 --> 00:36:13,254
I want to see more of that because people are unburdened by the way things have already
been.

433
00:36:13,254 --> 00:36:14,912
And I don't know how

434
00:36:14,912 --> 00:36:22,498
I'm going to turn that into what I'm learning and in my career, but that's what I'm
struggling with and thinking about a whole lot right now.

435
00:36:23,062 --> 00:36:34,032
So I do think that there is this challenge of at least the question I've been asking
myself is am I part of an industry where a majority of my colleagues or Network, etc.

436
00:36:34,032 --> 00:36:42,419
People who are in the same area of tech as I am are all online Or is it just a small part
of it?

437
00:36:42,419 --> 00:36:51,266
And I'm just seeing those people who scream on rooftops and of course they pick the most
click-baity thing to to say or trigger to get views

438
00:36:51,330 --> 00:36:58,092
And I worry that it's the former and I'm seeing the entire industry and I'm not happy with
that.

439
00:36:58,212 --> 00:37:06,444
And it's not really the latter, which would be sort of both the cynical perspective where
you only listening to those people who should not be listened to because there's nothing

440
00:37:06,444 --> 00:37:07,694
else to be there.

441
00:37:07,795 --> 00:37:16,977
But I guess I'm optimistic in that way that there is a secret part of humanity that our
software engineers, they have no desire to be on LinkedIn or blue sky or wherever people

442
00:37:16,977 --> 00:37:17,777
go today.

443
00:37:17,777 --> 00:37:19,852
Maybe they'll show up on hacker news one time.

444
00:37:19,852 --> 00:37:22,023
if some random company they happen to like mention.

445
00:37:22,023 --> 00:37:24,504
But I don't know.

446
00:37:24,504 --> 00:37:33,807
It's like there are so many people using chat bots today, LLMs in some regard, that I fear
that a majority of our tech industry is also online in some regard.

447
00:37:33,807 --> 00:37:40,149
And there's no secret rock out there in the forest that you can uncover.

448
00:37:40,149 --> 00:37:42,670
There's engineers doing an amazing thing there.

449
00:37:42,670 --> 00:37:44,650
Hasn't that been totally abused?

450
00:37:44,746 --> 00:37:47,028
Yeah, that's so real.

451
00:37:47,028 --> 00:37:57,234
feel like because I have to post so much online on the various networks for work, at least
half of the replies are always bots.

452
00:37:57,234 --> 00:37:59,616
And I can just, you can smell it a mile away.

453
00:37:59,616 --> 00:38:06,761
And I do have like a nice Discord community of people where I'm like, look, the humans are
here and it's great.

454
00:38:06,761 --> 00:38:09,022
And that's honestly...

455
00:38:09,194 --> 00:38:18,141
Those kinds of group chats, whether it's a Discord group or a Signal group or various
group chats where I have some humans that I know know how to code.

456
00:38:18,141 --> 00:38:21,653
I love being able to talk to them about projects and different things.

457
00:38:21,653 --> 00:38:31,710
And uh it's something that I feel like I have been limited now, just given the state of
social media and how I can talk about it online where, for example, one of my projects, I

458
00:38:31,710 --> 00:38:37,994
was deciding if I wanted to sync data using SQLite or uh JSON.

459
00:38:38,262 --> 00:38:43,824
A simple thing, there's pros and cons to both depending on how the app will ultimately be
used.

460
00:38:43,824 --> 00:38:47,646
And unfortunately, using AI, they're just like, you're absolutely right.

461
00:38:47,646 --> 00:38:49,266
This is a tough decision.

462
00:38:49,266 --> 00:38:52,128
And it just doesn't actually give me opinions.

463
00:38:52,128 --> 00:38:59,571
And I was like, I need some engineers who have seen things to just tell me stuff and not
use a chat bot to be like, wow, what a quandary.

464
00:38:59,571 --> 00:39:06,834
And so that is where I posted the Discord chats and in the Signal chats and stuff and be
like, I want you to roast this concept.

465
00:39:06,834 --> 00:39:08,258
Tell me what you think.

466
00:39:08,258 --> 00:39:15,689
Some people had really strong opinions in both ways, but being able to see that with that
experience was so nice and is so refreshing.

467
00:39:15,689 --> 00:39:17,732
And I feel like we need more of that.

468
00:39:17,732 --> 00:39:20,735
But people are doing the things for the clicks.

469
00:39:21,056 --> 00:39:25,292
so I feel like we're losing a lot of that uh vulnerability.

470
00:39:25,292 --> 00:39:26,243
Yeah, for sure.

471
00:39:26,243 --> 00:39:27,443
Genuineness.

472
00:39:27,443 --> 00:39:29,925
There is some genius in what you've shared there.

473
00:39:29,925 --> 00:39:43,833
I remember early on in one of my previous companies that I had worked at, an engineer had
told me that she paid a uh used car salesman to go to a uh new lot for a competitor to

474
00:39:43,833 --> 00:39:45,894
negotiate buying her car for her.

475
00:39:45,894 --> 00:39:50,546
uh And I'm like, genius, honestly.

476
00:39:50,546 --> 00:39:53,602
And I think the same thing applies to the LLMs where you just get...

477
00:39:53,602 --> 00:40:00,017
they'll allow the first one to give you the plan and then you bring it to another one and
be like, hey, this other LLM said this is the best thing ever.

478
00:40:00,017 --> 00:40:01,949
Tell me why it's atrocious.

479
00:40:01,949 --> 00:40:03,510
And the stats are back on that.

480
00:40:03,510 --> 00:40:08,544
Actually, it's pretty effective if you, instead of saying that I did this thing, why is it
so great?

481
00:40:08,544 --> 00:40:09,994
No leading questions.

482
00:40:09,994 --> 00:40:11,897
ah I didn't do this thing.

483
00:40:11,897 --> 00:40:12,908
I think it's terrible.

484
00:40:12,908 --> 00:40:14,119
Do you agree?

485
00:40:14,119 --> 00:40:16,981
Is a great way for it to expose all the problems there.

486
00:40:16,981 --> 00:40:23,456
I will say that there is a little bit of an issue where, especially when it comes to
reasoning, that there's like this

487
00:40:23,456 --> 00:40:32,323
It can come up with like 10 ideas about how to solve a different problem a particular way
and it'll be missing the real like the one that I believe is the objectively correct

488
00:40:32,323 --> 00:40:36,977
answer number 11 and only if you go back to your desk and bang your head against the
keyboard for a while.

489
00:40:36,977 --> 00:40:43,276
uh I guess not in your case because you have probably have like really heavy keyboard keys
and that would probably be damaging.

490
00:40:43,276 --> 00:40:44,351
But still.

491
00:40:46,422 --> 00:40:48,068
You actually can get out a real answer.

492
00:40:48,068 --> 00:40:49,164
Yeah, for sure.

493
00:40:49,164 --> 00:40:51,155
Yeah, okay, so really quick shill hat.

494
00:40:51,155 --> 00:40:57,916
Something that I am actually excited that GitHub released recently is they have a rubber
duck feature now that's built into the CLI.

495
00:40:57,916 --> 00:41:05,558
So when you do make plans, it automatically consults another model to be like, hey, vet
this plan to make sure it's good and there aren't holes.

496
00:41:05,558 --> 00:41:07,519
And I used it yesterday and it was fun.

497
00:41:07,519 --> 00:41:12,480
Taking the shill hat off, yeah.

498
00:41:12,800 --> 00:41:18,880
like you said, it's cool to consult the different tools on this and I feel like that's

499
00:41:18,880 --> 00:41:22,894
was they're so agreeable and it's a nice thing.

500
00:41:22,894 --> 00:41:28,970
But I kind of want a grumpy engineer who's seen something to vet certain plans for me.

501
00:41:28,970 --> 00:41:38,109
And I feel like that's something we kind of have lost with Stack Overflow kind of going
away more and more and why it's nice to have these chats where I just want to see these

502
00:41:38,109 --> 00:41:43,444
kinds of debates because we learn so much from failure.

503
00:41:43,478 --> 00:41:50,086
and things going poorly and things you've had to debug for hours and hours and you're
like, was a one line change after all of that.

504
00:41:50,167 --> 00:41:59,758
Those sorts of things you learn so much from and we just aren't having the same levels of
failures and debugging sessions as we once did.

505
00:41:59,839 --> 00:42:01,860
And so we're less haggard.

506
00:42:02,824 --> 00:42:09,542
Sorry, I'm laughing because I know you have a special place in your heart for the Stack
Overflow community.

507
00:42:10,364 --> 00:42:21,260
You uh chaired the authoring of the podcast and got it started originally, But no one, I
don't think anyone in quite a long time has said, how great Stack Overflow is.

508
00:42:21,260 --> 00:42:22,931
I know, I know.

509
00:42:22,971 --> 00:42:23,731
That's the thing.

510
00:42:23,731 --> 00:42:35,418
I don't think I would have said I liked it when people were mean to me on Stack Overflow
until recently because I just, need something to roast my projects.

511
00:42:35,418 --> 00:42:44,742
And you can like train some kind of custom LLM fine tune or something and say like, hey,
everything I do roast me, it's not the same.

512
00:42:44,843 --> 00:42:46,594
It's not.

513
00:42:46,594 --> 00:42:51,148
And I feel like those kinds of perspectives that

514
00:42:51,148 --> 00:42:54,640
They come from community and they come from experience.

515
00:42:54,640 --> 00:42:58,282
And finding that community is particularly hard right now.

516
00:42:58,282 --> 00:43:00,643
And when you do have it, hold onto it tight.

517
00:43:00,643 --> 00:43:02,635
again, I love my Discord server.

518
00:43:02,635 --> 00:43:05,596
I love my group chats for that kind of experience.

519
00:43:05,596 --> 00:43:12,460
And I hope those types of communities can continue to exist as we all move towards this
future.

520
00:43:12,460 --> 00:43:20,092
I mean, you've given me some optimism because I think one of my personal failure modes has
been I just like tearing things apart and so much so that when I come up with an idea to

521
00:43:20,092 --> 00:43:24,643
do a project, I'm like, but there's an open source version already and it's already there
on the internet.

522
00:43:24,643 --> 00:43:25,854
Like I can just use that.

523
00:43:25,854 --> 00:43:28,234
I have no justification to do this myself.

524
00:43:28,234 --> 00:43:31,965
And I take that same optimism and I bring it to other people's projects too.

525
00:43:31,965 --> 00:43:33,736
But you you saying I missed this.

526
00:43:33,736 --> 00:43:36,276
I missed when tearing apart my stuff.

527
00:43:36,457 --> 00:43:37,677
Why the idea is so bad.

528
00:43:37,677 --> 00:43:42,010
You know, maybe there is a whole in this future world where

529
00:43:42,010 --> 00:43:45,994
The artisanal software engineer review comes into play.

530
00:43:46,695 --> 00:43:50,760
We hand pull requests review all of your code that you write.

531
00:43:50,760 --> 00:43:53,041
There's no bots involved here.

532
00:43:53,198 --> 00:43:56,560
uh I honestly like that.

533
00:43:56,560 --> 00:44:02,342
Who knows if I should be saying that as a GitHub employee right now.

534
00:44:02,342 --> 00:44:12,047
like, I think there's so much value in the human experience there and it'll be interesting
once again to see where our roles go.

535
00:44:12,047 --> 00:44:16,950
One of the things that I genuinely really like doing is like untangling legacy code.

536
00:44:16,950 --> 00:44:18,208
It's so fun.

537
00:44:18,208 --> 00:44:23,483
I love going into some kind of hairy 3000 line file and be like, why is it like this?

538
00:44:23,483 --> 00:44:24,824
I'm going to make it perfect.

539
00:44:24,824 --> 00:44:32,568
It's like untangling headphones or earbuds or a pile of shoelaces or something.

540
00:44:32,568 --> 00:44:33,840
I know what you're talking about.

541
00:44:33,840 --> 00:44:42,484
just, fear that there's just like very small part of audience, human population that
understands like, oh yes, there's a knot here.

542
00:44:42,484 --> 00:44:43,743
I get to untangle.

543
00:44:43,743 --> 00:44:44,782
can't wait to.

544
00:44:44,782 --> 00:44:47,164
It's like solving a puzzle like a sicko.

545
00:44:47,164 --> 00:44:49,656
uh it's really fun.

546
00:44:49,656 --> 00:44:59,513
Something that I've honestly been doing is having AI make some feature on an app, and then
I just rip it to shreds, and I treat it like legacy code and fix it up and clean it up

547
00:44:59,513 --> 00:45:00,454
myself.

548
00:45:00,454 --> 00:45:02,746
Some models, I have to do that more or less.

549
00:45:02,746 --> 00:45:03,796
Certain features, I do it.

550
00:45:03,796 --> 00:45:11,532
But that's honestly how I've been kind of enjoying certain side projects where I make the
feature exist, and then I make it better myself.

551
00:45:11,920 --> 00:45:13,736
That just sounds horrifyingly bad.

552
00:45:13,736 --> 00:45:16,776
Like I just can't imagine what to review a-a-a-a-

553
00:45:17,062 --> 00:45:17,863
I know.

554
00:45:17,863 --> 00:45:27,627
Yeah, believe me, it's like not always a great experience, like that was something I did
yesterday where this feature that I've been wanting in a side project, I was like, fine,

555
00:45:27,627 --> 00:45:29,287
we'll have an AI model do it.

556
00:45:29,287 --> 00:45:30,428
I hated it.

557
00:45:30,428 --> 00:45:32,569
The code was not what I wanted.

558
00:45:32,569 --> 00:45:42,313
But because the code existed, I was like, I can make it exactly what I want because it
almost gave me clarity from a rubber duck type perspective being like, oh, it makes sense

559
00:45:42,313 --> 00:45:44,984
why it was implemented this way, but it should be implemented this way.

560
00:45:44,984 --> 00:45:46,278
And I was able to like...

561
00:45:46,278 --> 00:45:49,301
use my opinions to edit it more.

562
00:45:49,362 --> 00:45:59,582
So I don't know, maybe I just craved a thought exercise, but uh it'll be interesting once
again to see, is this always how we're going to be shipping things?

563
00:45:59,582 --> 00:46:00,393
Probably not.

564
00:46:00,393 --> 00:46:01,334
don't know.

565
00:46:01,582 --> 00:46:10,062
I mean, I almost fear for you because you're like, spend a lot of time online with
AI-generated bots, with AI-generated comments for my messages.

566
00:46:10,102 --> 00:46:17,622
I get to work and I have these angry developers complaining about the platform that I'm
helping to make a success.

567
00:46:17,622 --> 00:46:18,262
You know what?

568
00:46:18,262 --> 00:46:19,902
I just crave.

569
00:46:20,366 --> 00:46:22,606
I breaking something.

570
00:46:22,606 --> 00:46:24,586
Yeah, no, believe me.

571
00:46:24,586 --> 00:46:28,405
I feel like I'm not super AI-pilled.

572
00:46:28,405 --> 00:46:30,846
I still consider myself somewhat of a skeptic.

573
00:46:30,846 --> 00:46:34,906
And I'm at this point where like I just smell it a mile away.

574
00:46:35,086 --> 00:46:40,266
Every single time I hear someone reading a script, I'm just like, I know this is an
AI-generated script.

575
00:46:40,266 --> 00:46:45,566
Within seconds, I can see it in blog posts, I can see it in comments, I can see it in
code.

576
00:46:45,666 --> 00:46:48,346
It's just what I see now.

577
00:46:48,374 --> 00:47:00,190
And it's so nice to see human writing and stuff and human code and fixing things with
human changes that, I don't know, I feel like I'm trying to add that more and more

578
00:47:00,190 --> 00:47:00,910
wherever I can.

579
00:47:00,910 --> 00:47:04,142
Like I refuse to use AI on my own personal blog.

580
00:47:04,142 --> 00:47:07,543
I refuse to use AI to write social media posts for me.

581
00:47:08,004 --> 00:47:12,716
And at the same time, I see people using it and I understand why they use it.

582
00:47:12,716 --> 00:47:18,970
And I want to figure out how do I get them to a place where they don't have it right for
them, but they might have it edit for them or something.

583
00:47:19,351 --> 00:47:22,893
There's balances everywhere that we're trying to meet.

584
00:47:23,406 --> 00:47:32,366
I feel like this is you trying to teach someone how to be enlightened and you can't drag,
like they have to want to get there and see their own path for them.

585
00:47:32,666 --> 00:47:38,886
It's just, so we had a vibe coding episode of this podcast a bunch of months ago.

586
00:47:39,246 --> 00:47:47,006
And the reason I bring it up is because the guest referenced a post about that AI art has
no soul.

587
00:47:47,006 --> 00:47:50,686
And I think this is true for a lot of those things why at least,

588
00:47:50,786 --> 00:47:52,747
you can reference, like identify immediately.

589
00:47:52,747 --> 00:47:54,728
can usually identify what's going on.

590
00:47:54,728 --> 00:47:59,850
And the thing that stabs me a lot is some of the things that I noticed that it does, like
I do intentionally.

591
00:47:59,850 --> 00:48:06,182
Like the thing about dashes broke me because right with dashes, I love dashes.

592
00:48:06,182 --> 00:48:07,553
dashes are so great.

593
00:48:07,553 --> 00:48:08,543
And now I'm just like, what?

594
00:48:08,543 --> 00:48:15,176
People are going to mistake my work for LLM generated garbage because I use this prose
like naturally.

595
00:48:15,176 --> 00:48:17,677
I have a shortcut key on my keyboard for dash.

596
00:48:17,677 --> 00:48:19,378
uh I know.

597
00:48:19,378 --> 00:48:20,302
Yeah.

598
00:48:20,302 --> 00:48:24,646
uh But now I'm like second-guessing, like, do I have to take it out of my writing?

599
00:48:24,646 --> 00:48:29,731
Because people will start to assume that it's also generated that way.

600
00:48:29,731 --> 00:48:35,196
It's so different, though, and I don't know what it is, like, why I can recognize it in
what I'm reading.

601
00:48:35,262 --> 00:48:40,822
I mean, the joke out there right now is like, the person who filed the support ticket
thanked me.

602
00:48:40,822 --> 00:48:43,674
ah That's the clincher.

603
00:48:43,674 --> 00:48:45,706
Like, a human would never be that nice.

604
00:48:46,859 --> 00:48:53,886
I did see something gross, just like people think they're AI bots more than support
agents, which was so sad.

605
00:48:54,038 --> 00:49:07,521
yeah, no, I feel like I've been trying to, it's been a good exercise in me like practicing
my voice on the internet more where I'm not gonna care if I use an emdash or anything.

606
00:49:07,521 --> 00:49:09,002
I'm just gonna write.

607
00:49:09,034 --> 00:49:14,217
And I have like an AI manifesto on my website where I say like, I'm not going to use AI
for my writing.

608
00:49:14,217 --> 00:49:19,200
I very much strongly think if you can't be bothered to write it, I can't be bothered to
read it.

609
00:49:19,200 --> 00:49:21,751
And so I want to make something that's like worth reading.

610
00:49:21,751 --> 00:49:27,344
honestly, yeah, something about AI writing is so sterile, I guess.

611
00:49:27,344 --> 00:49:32,297
Like, like whenever I read it, it's, it's almost unreadable because I'm like, I know this
is correct.

612
00:49:32,297 --> 00:49:38,560
The words are in the correct order, but it's so hard to read because there's no like depth
to it.

613
00:49:38,678 --> 00:49:40,363
And yeah.

614
00:49:40,654 --> 00:49:41,914
I know what it is.

615
00:49:42,295 --> 00:49:47,197
You're picking up on the uncanny valley of that writing.

616
00:49:47,197 --> 00:49:51,299
You see that there's no soul there and it's deeply disturbing.

617
00:49:51,299 --> 00:49:53,380
And I think for a lot of people, they just don't see it.

618
00:49:53,380 --> 00:49:57,522
They don't necessarily understand or don't perceive it.

619
00:49:57,522 --> 00:49:58,793
They don't observe it in that way.

620
00:49:58,793 --> 00:50:02,024
So it may not actually bother them as much.

621
00:50:02,165 --> 00:50:04,586
I guess I'll go on record in saying this.

622
00:50:04,756 --> 00:50:12,788
just recently wrote a couple of blog posts where I now can distinguish between having an
LLM help you write and it doing the writing for you.

623
00:50:12,788 --> 00:50:19,250
I think there are things that like I really wanted to get written like a long time, like
those tickets in your backlog or that tech debt you want to take care of and I wouldn't

624
00:50:19,250 --> 00:50:20,430
have done it.

625
00:50:20,510 --> 00:50:27,872
And because of the scope creep that LLMs keep throwing into everything that I do, every,
you here's some extra features.

626
00:50:27,872 --> 00:50:29,473
Here's something like writing command line prompt.

627
00:50:29,473 --> 00:50:32,640
Here's five input parameters that you can optionally add in.

628
00:50:32,640 --> 00:50:33,474
I'm like, hadn't.

629
00:50:33,474 --> 00:50:34,392
This is a one-time script.

630
00:50:34,392 --> 00:50:35,632
I don't need any of those.

631
00:50:35,632 --> 00:50:36,714
Stop that.

632
00:50:36,838 --> 00:50:38,290
Don't make a wizard.

633
00:50:38,926 --> 00:50:45,174
uh Here's a whole UI that I know you didn't want that but you know it's now there's a file
called UI.

634
00:50:45,174 --> 00:50:48,574
A little Unicode loading m animation and everything.

635
00:50:48,574 --> 00:50:49,455
yeah, for sure.

636
00:50:49,455 --> 00:50:51,196
I think the same thing applies to the writing though.

637
00:50:51,196 --> 00:50:59,321
uh It is really helpful to have it build out all the scaffolding, not the outline, it's
terrible at that, not the paragraphs, it's terrible at that.

638
00:50:59,321 --> 00:51:06,335
But if you tell it to write a bunch of things, then I can go back through and be like, you
know what, I do want to say something here, but what it wrote is so atrocious, I need to

639
00:51:06,335 --> 00:51:07,754
rewrite the whole thing.

640
00:51:07,754 --> 00:51:15,456
And so going back through and rewriting an article that I didn't just find on the internet
that's the appropriate length is really motivating for me.

641
00:51:15,456 --> 00:51:23,050
Again, it's the, found this knot or puzzle that I have to uns- like fix in some way and I
know that's not everyone's particular preference.

642
00:51:23,340 --> 00:51:32,378
Yeah, no, there was a point where my team started taking on certain newsletters that
GitHub puts out and my team writes those newsletters.

643
00:51:32,378 --> 00:51:39,324
And we had to tell the folks who are assigning us these newsletters, hey, don't write it
with AI for us to edit.

644
00:51:39,324 --> 00:51:40,965
That is actually harder.

645
00:51:40,965 --> 00:51:44,408
If you just give us the headlines that you want us to hit, we can write it.

646
00:51:44,408 --> 00:51:45,429
It will read better.

647
00:51:45,429 --> 00:51:47,910
It will make more sense.

648
00:51:48,351 --> 00:51:52,054
so now if we ever use AI for any sort of writing at work,

649
00:51:52,054 --> 00:51:56,076
all we do is get like, these are the bullet points we want to hit, then that's it.

650
00:51:56,076 --> 00:51:58,617
And we fill it out that way.

651
00:51:58,617 --> 00:52:07,541
yeah, I feel like I'm constantly fighting with myself on what it's useful for, what it's
not useful for, what we should use it for, and what we shouldn't.

652
00:52:07,541 --> 00:52:13,634
really, I just want humans to do well at being humans and use their brains.

653
00:52:13,634 --> 00:52:18,476
Because if the internet goes out one day, what will you be able to do?

654
00:52:18,638 --> 00:52:28,827
Well, I think there was something that I was dealing with recently, which is I sort of
want LLMs to go away because of the environmental impact or data issues or whatever.

655
00:52:28,827 --> 00:52:31,278
I keep thinking to myself, but will it really?

656
00:52:31,278 --> 00:52:32,831
I don't think so.

657
00:52:32,831 --> 00:52:42,560
And so maybe it is really more of a bet of, since it's going to be, even if it never gets
better, we're just at this level, it still means that things are fundamentally different

658
00:52:42,560 --> 00:52:43,290
than they were before.

659
00:52:43,290 --> 00:52:45,560
And I don't think there's any way of getting away from that.

660
00:52:45,560 --> 00:52:58,409
I think there are waves of tech trends, a very significant one, for example, that like
grew and then fell with crypto a few years ago, where I think we're deep in an AI wave and

661
00:52:58,409 --> 00:53:01,521
it's less like it's going to come and then go away.

662
00:53:01,521 --> 00:53:10,077
It's more going to be like everyone is using it and then the tools that are good will
probably rise to the top and the ones that end up not actually being useful because they

663
00:53:10,077 --> 00:53:13,689
just threw AI at you will probably go away.

664
00:53:13,729 --> 00:53:15,640
And so I feel like

665
00:53:15,640 --> 00:53:21,914
Hopefully, we'll get to a point where like right now everyone's like, get on the AI train
or you're going to be left behind.

666
00:53:21,914 --> 00:53:28,877
Like I'm sure there's elements of that, but I think there's no such thing as getting left
behind in this regard.

667
00:53:28,877 --> 00:53:37,622
There might be money to be made, but we're going to see the tools that end up being good
productivity tools, however you define productivity.

668
00:53:37,622 --> 00:53:44,490
Those ones will rise to the top, be useful for us, and the ones that end up being
superfluous extra.

669
00:53:44,490 --> 00:53:46,272
unnecessary will go away.

670
00:53:46,272 --> 00:53:51,035
And hopefully by then things will even out a little bit.

671
00:53:51,035 --> 00:53:56,980
But right now everyone's just like, I got to use these tokens because they're for free now
and they're going to be more expensive later.

672
00:53:56,980 --> 00:53:59,522
we're just seeing that everywhere right now.

673
00:53:59,662 --> 00:54:11,656
Okay, my spicy take is that I think that the companies that create LLMs uh are taking a
page from the book of all those social media platforms that have corrupted the teenagers

674
00:54:11,656 --> 00:54:23,709
for generations and are using those same strategies to convince or trap uh software
engineers in vicious loops of continual usage and basically abusive relationships.

675
00:54:23,709 --> 00:54:26,446
And it's incredibly hard to get out of one of those.

676
00:54:26,446 --> 00:54:42,146
Yes, I am aware of who I work for, but I think that's a very real thing where like a few
years ago, people would be enraged being just like, have to spend $20 a month to use AI.

677
00:54:42,326 --> 00:54:48,106
And now they're just like, I will spend $1,000 a month if it means that I will get as many
tokens as I want.

678
00:54:48,366 --> 00:54:53,346
Like that type of attitude shift, it's like an addiction.

679
00:54:54,530 --> 00:55:02,295
fascinating to see and I think that that's by design and once again kind of above our pay
grade to stop it.

680
00:55:02,295 --> 00:55:13,233
And so, yeah, it's a very weird thing where I do see usage gains and I do see productivity
gains where some of the domain names that I've had sitting for years, I actually have

681
00:55:13,233 --> 00:55:14,484
shipped the projects.

682
00:55:14,484 --> 00:55:17,726
Like that is an actual value to me that I'm excited about.

683
00:55:17,746 --> 00:55:23,750
And then at the same time, I'm nervous for like AI psychosis brain.

684
00:55:23,874 --> 00:55:24,534
Hmm.

685
00:55:24,534 --> 00:55:28,886
It's a cybernetic rot, like brain rot.

686
00:55:28,886 --> 00:55:31,225
Like you use AI and it's coming.

687
00:55:31,225 --> 00:55:40,942
mean, Microsoft backed the paper at Carnegie Mellon that suggested using LLMs removes your
usage of critical thinking.

688
00:55:40,942 --> 00:55:42,663
you, it's gone.

689
00:55:42,663 --> 00:55:44,394
Like you just don't use it.

690
00:55:44,394 --> 00:55:49,576
And it, hate when people say the brain is like a muscle that has to be trained.

691
00:55:49,778 --> 00:55:58,224
But they use that analogy in the paper and I mean in some regard it's probably true like
if you don't continually think about this we know that patients with with any sort of

692
00:55:58,224 --> 00:56:11,493
dementia that do like word puzzles or like strategic thinking games are much better at
Keeping their their wits about them rather than ones that watch uh non interactive content

693
00:56:11,493 --> 00:56:13,554
uh I don't know

694
00:56:14,199 --> 00:56:15,068
Yeah.

695
00:56:15,692 --> 00:56:18,972
Okay, so maybe it's a good moment to move on to picks.

696
00:56:18,972 --> 00:56:21,610
So Cassidy, what did you bring for us today?

697
00:56:22,070 --> 00:56:32,458
My pick is the note-taking app Obsidian, um where if you haven't used Obsidian before, if
you look at it, it's like, this is your second brain with infinite plugins.

698
00:56:32,559 --> 00:56:36,062
It is a glorified markdown editor, but it has plugins.

699
00:56:36,062 --> 00:56:47,722
And I think the open source plugin ecosystem is where its superpower is, where I honestly,
I could, to some circles, be considered an Obsidian power user, and to others be like a

700
00:56:47,722 --> 00:56:50,560
total noob, because I really do just use it.

701
00:56:50,560 --> 00:56:55,291
as a markdown editor for my blog posts and for taking notes on projects and stuff.

702
00:56:55,291 --> 00:57:02,233
But the fact that you can backlink your notes and tag between them is where it's really,
really powerful.

703
00:57:02,233 --> 00:57:11,936
I've been able to like make wikis for myself on my projects, on my ideas, on any talks
that I listen to or things that I see.

704
00:57:11,936 --> 00:57:15,627
And I love that it's such an open ecosystem.

705
00:57:15,627 --> 00:57:17,957
I have like a theme that I made for it.

706
00:57:17,957 --> 00:57:20,514
There's so many different plugins that you can use for

707
00:57:20,514 --> 00:57:27,232
formatting markdown or turning it into a Kanban board or really just a whole personal
dashboard, but I'm a big fan.

708
00:57:27,234 --> 00:57:33,527
I think the Zettelkasten aspect of it to do the deep linking is where it really shined
initially.

709
00:57:33,527 --> 00:57:44,501
Yeah, I think it can be very intimidating at first because of all of its capabilities, but
that's where I say to anybody who first opens it, start with just like writing stuff and

710
00:57:44,501 --> 00:57:48,510
get into that habit and slowly build up workflows that make it make sense for you.

711
00:57:48,510 --> 00:57:53,453
You know, I started with it early on and I was a huge obsidian user.

712
00:57:53,453 --> 00:57:55,595
I didn't go down the whole plugin route though.

713
00:57:55,595 --> 00:58:01,096
And I've been going, like making my way back to VS Code as my note-taking application.

714
00:58:01,096 --> 00:58:05,385
And honestly, there's so many similarities.

715
00:58:05,385 --> 00:58:07,032
Just, it makes sense.

716
00:58:07,032 --> 00:58:08,548
but it's the marketplace.

717
00:58:09,152 --> 00:58:10,163
Yeah.

718
00:58:10,324 --> 00:58:10,825
Yeah.

719
00:58:10,825 --> 00:58:12,337
Well, now they have a CLI.

720
00:58:12,337 --> 00:58:19,538
I haven't actually played with it yet, but I'm like, that does unlock so many things to be
able to use Obsidian in a headless way.

721
00:58:19,538 --> 00:58:23,883
But I'm going to stick with it as just like a quick place for me to take notes.

722
00:58:23,924 --> 00:58:25,550
And we'll leave it there.

723
00:58:25,550 --> 00:58:31,118
How do you do the syncing between your, I assume, laptop or desktop and like your phone if
you're on the go?

724
00:58:31,118 --> 00:58:37,326
So they have like a sync platform that I pay for, but as a backup, I keep my vault in
Dropbox.

725
00:58:37,326 --> 00:58:43,250
And so I have both of those and I occasionally make a copy in Dropbox in case I lose it.

726
00:58:43,250 --> 00:58:47,357
No, but like you ever like you're out somewhere and you just have your phone and you want
to take a note?

727
00:58:47,357 --> 00:58:50,501
How do get those notes into your obsidian vault?

728
00:58:52,325 --> 00:58:53,142
I see.

729
00:58:53,142 --> 00:58:54,022
Yeah.

730
00:58:54,664 --> 00:59:00,310
If it is really, really quick and I don't feel like using it, I'll text myself, but that's
rare.

731
00:59:01,279 --> 00:59:02,260
No, I think it's fair.

732
00:59:02,260 --> 00:59:11,236
know, after I've done a lot of the editors tried a bunch of different things and I feel
like when I'm outside of a coding experience, it is for some reason obsidian.

733
00:59:11,236 --> 00:59:19,303
And I don't know if that's just some I'm on Linux and maybe there's some magical set of
solutions on Mac OS because everyone's like, yeah, I just install whatever I think bear is

734
00:59:19,303 --> 00:59:20,986
the one people use for writing.

735
00:59:20,986 --> 00:59:28,110
I used to use bare, but I switched so much between a PC and a Mac and any given day
because I stream from my PC.

736
00:59:28,110 --> 00:59:30,312
I use my Mac for some things and other things.

737
00:59:30,312 --> 00:59:34,474
But yeah, I needed something more neutral that could work across platforms.

738
00:59:34,574 --> 00:59:35,254
No, got it.

739
00:59:35,254 --> 00:59:36,074
No, it's a good pick.

740
00:59:36,074 --> 00:59:36,634
That's a good pick.

741
00:59:36,634 --> 00:59:38,034
And no one's done it before.

742
00:59:38,274 --> 00:59:40,834
So you got the unique holder of Obsidian.

743
00:59:40,934 --> 00:59:44,674
We did have one guest that went back through and was like, oh yeah, and you're like 300
episodes.

744
00:59:44,674 --> 00:59:45,554
No one picked this.

745
00:59:45,554 --> 00:59:50,793
like, I don't even think there's a canonical list of where, what the pick's worth.

746
00:59:50,793 --> 00:59:51,814
don't know.

747
00:59:53,000 --> 00:59:56,853
where an agent can go through and find all the pics and make a list.

748
00:59:57,450 --> 01:00:06,931
Yeah, an agent is coming to make a tool to crawl the blog and the website and have a page
dedicated to it.

749
01:00:06,931 --> 01:00:09,023
But no one asked for that.

750
01:00:09,965 --> 01:00:11,294
That's in the backlog.

751
01:00:11,294 --> 01:00:19,161
That's something where I also want to do some kind of thing like that for my newsletter,
because my newsletter has been going for nine years now.

752
01:00:19,161 --> 01:00:24,926
And I want to just make some kind of visualization on the emojis that I've had in the
subject line.

753
01:00:24,926 --> 01:00:26,767
Nobody cares, but I do.

754
01:00:26,767 --> 01:00:30,530
And so that's a project in the backlog for another time.

755
01:00:30,754 --> 01:00:32,178
Do things for yourself.

756
01:00:32,178 --> 01:00:34,844
That's the most important modus

757
01:00:38,382 --> 01:00:39,802
Yeah, so what did I bring?

758
01:00:39,802 --> 01:00:41,302
I struggled today.

759
01:00:41,482 --> 01:00:43,642
I'm running out of things to pick, honestly.

760
01:00:43,642 --> 01:00:48,462
I think it's almost going on three years of just hosting the podcast myself.

761
01:00:48,462 --> 01:00:54,362
That's a lot of to At that point, I have to have something every single week that's like
150 picks already.

762
01:00:54,362 --> 01:01:04,682
So I've started this new thing where I have to have an idea about something that I could
do a pick on, then find that thing in the universe so that I could actually pick it.

763
01:01:05,152 --> 01:01:08,055
A guest, uh I think last year actually suggested this.

764
01:01:08,055 --> 01:01:09,868
It's a book called The Light Eaters.

765
01:01:09,868 --> 01:01:14,023
It dives into the scientific understanding of how sentient plants are.

766
01:01:14,023 --> 01:01:16,515
How sentient are they?

767
01:01:17,457 --> 01:01:18,588
Yeah, the light eaters.

768
01:01:18,588 --> 01:01:19,679
Yeah.

769
01:01:19,679 --> 01:01:21,401
Like they eat light, photosynthesis.

770
01:01:21,401 --> 01:01:22,502
ah

771
01:01:22,515 --> 01:01:25,558
I at first thought it was like people who just don't eat that much

772
01:01:27,982 --> 01:01:38,531
uh The author like details her almost philosophical journey to like full plant
understanding and there's like some interesting things and I keep a list of every time

773
01:01:38,531 --> 01:01:45,817
someone tells me like, I'm a vegetarian because I don't like eating animals and I'm like,
well, you can be a vegetarian for lots of reasons though.

774
01:01:45,817 --> 01:01:53,644
The one to not like eating sentient beings is not really a good argument because, you
know, I feel like you're putting plants into disservice realistically.

775
01:01:53,644 --> 01:01:56,526
There's like a couple of really good research papers.

776
01:01:56,526 --> 01:02:05,626
that on my list that suggest like plants that are from the same species who have similar
genealogy will try to avoid stealing each other's sunlight.

777
01:02:05,626 --> 01:02:06,506
Wow.

778
01:02:06,506 --> 01:02:11,946
And I just find like, how do you justify that plants aren't intelligent?

779
01:02:12,046 --> 01:02:13,246
I mean, at this point you can't.

780
01:02:13,246 --> 01:02:17,354
If you read the book, you'll be convinced too.

781
01:02:17,354 --> 01:02:18,280
Yeah, okay.

782
01:02:18,280 --> 01:02:19,155
I'll look that one up.

783
01:02:19,155 --> 01:02:20,283
That sounds cool.

784
01:02:20,283 --> 01:02:27,097
I don't know if it's like a personal vendetta just to argue against people when they say
things, but ah I mean I appreciate a lot of things, but

785
01:02:27,362 --> 01:02:29,154
Bike-driven developments fun.

786
01:02:29,166 --> 01:02:36,204
ah I mean it used to be a thing like this can't be done and then you'd be like I'm gonna
do it and

787
01:02:36,935 --> 01:02:41,942
I played sousaphone and marching band purely because my band director said that I
couldn't.

788
01:02:41,942 --> 01:02:45,467
You're too Literally that.

789
01:02:45,467 --> 01:02:46,068
Yeah.

790
01:02:46,068 --> 01:02:49,533
I had a bruise on my shoulder for a year, but I played it.

791
01:02:49,533 --> 01:02:50,978
So there you go.

792
01:02:50,978 --> 01:02:55,690
So yeah, so that's the one that's like the marching band version of 2-BUTT, right?

793
01:02:55,690 --> 01:02:58,768
Yeah, yeah, it's like the big circle around your body and stuff.

794
01:02:58,768 --> 01:03:00,412
My sister and I both ended up playing it.

795
01:03:00,412 --> 01:03:03,892
It was a blast, literally, but it was also incredibly heavy.

796
01:03:03,892 --> 01:03:05,292
But we had to them wrong.

797
01:03:05,292 --> 01:03:08,427
I never understood how anyone in marching band played one of those honestly.

798
01:03:08,427 --> 01:03:17,581
I was a trumpet player and that was much easier to carry around for nine and a half
minutes than it was to have a whole harness holding up your instrument so that it didn't

799
01:03:17,581 --> 01:03:18,572
fall over.

800
01:03:18,572 --> 01:03:27,104
I started as trumpet and then I kind of just progressed to larger and larger horns where I
went from like trumpet to mellophone to baritone to sousaphone.

801
01:03:27,566 --> 01:03:28,606
Do you still play?

802
01:03:28,896 --> 01:03:35,471
Yeah, I granted I have two toddlers now, so I don't have as much time to these days.

803
01:03:35,471 --> 01:03:39,344
before I had them, my husband and I, were in the Seattle Video Game Orchestra together.

804
01:03:39,344 --> 01:03:41,665
And so he was on violin and I was on trumpet and that.

805
01:03:41,665 --> 01:03:44,527
we would play like Zelda and Skyrim music around the city.

806
01:03:44,527 --> 01:03:45,707
It was really fun.

807
01:03:46,151 --> 01:03:47,958
I don't think that's on your LinkedIn profile.

808
01:03:47,958 --> 01:03:48,619
It's not.

809
01:03:48,619 --> 01:03:50,940
oh But it was a good time.

810
01:03:51,201 --> 01:03:58,398
I miss it and I'm looking forward to teaching my kids how to play these instruments
eventually.

811
01:03:58,398 --> 01:03:59,158
No, not LLMs.

812
01:03:59,158 --> 01:04:01,146
No, this is artistry.

813
01:04:01,146 --> 01:04:07,358
I think if LLMs completely take over our profession and there are no software engineers
left, least you'll still have something to fall back to.

814
01:04:07,358 --> 01:04:08,063
Exactly.

815
01:04:08,063 --> 01:04:08,577
There we go.

816
01:04:08,577 --> 01:04:09,010
Okay.

817
01:04:09,010 --> 01:04:11,118
Yeah, I'll take that angle.

818
01:04:12,276 --> 01:04:15,345
Well, thank you, Cassidy, for this episode of Adventures in DevOps.

819
01:04:15,345 --> 01:04:17,378
I've totally enjoyed having you on.

820
01:04:17,378 --> 01:04:18,356
Thank you so much for having me.

821
01:04:18,356 --> 01:04:19,593
It was a fun time.

822
01:04:19,630 --> 01:04:21,072
It's been great.

823
01:04:21,072 --> 01:04:26,918
Thanks all listeners for coming and tuning into this episode and hopefully we'll see you
all back again next week.