1
00:00:00,024 --> 00:00:02,813
I see that no one is 
 thinking,
everyone is doing. 

2
00:00:03,014 --> 00:00:04,954
But you can't build great 
 
stuff without thinking. 

3
00:00:05,244 --> 00:00:08,060
Today's guest is Milan 
Milanovic, 
 author of Laws of 

4
00:00:08,060 --> 00:00:10,665
Software Engineering, 56 laws 
and principles collected 
 from 

5
00:00:10,665 --> 00:00:13,323
real software examples, already 
read in 74 countries. 

6
00:00:13,684 --> 00:00:15,790
Speaking about laws, some people

 think it's all about 

7
00:00:15,790 --> 00:00:17,632
architecture, design, coding 
principles, but 
 actually you 

8
00:00:17,632 --> 00:00:20,358
have something more. 
What we think in software is 

9
00:00:20,358 --> 00:00:21,943
really 
 that everything is 
technical. 

10
00:00:21,973 --> 00:00:25,650
But what I figured out during my

 long career that software 

11
00:00:25,650 --> 00:00:27,234
failures are rarely purely 
technical. 

12
00:00:27,274 --> 00:00:30,132
There's the trap of using AI and
it ended 
 up with complex 

13
00:00:30,132 --> 00:00:32,304
systems or complex code. 
Complex system that works is 
 

14
00:00:32,311 --> 00:00:35,204
always found to have grown from 
a simple one that worked. 

15
00:00:35,313 --> 00:00:38,523
Don't just try to produce 
something 
 because you can now 

16
00:00:38,523 --> 00:00:40,595
easily produce. 
Optimize for judgment, not 

17
00:00:40,595 --> 00:00:42,804
output. 
Writing code is of course 
 our 

18
00:00:42,804 --> 00:00:45,411
main thing how we solve 
problems, but not the only 

19
00:00:45,411 --> 00:00:47,383
thing. 
If you can solve problem without

20
00:00:47,383 --> 00:00:50,054

 coding, that's the best thing.
So no code, no problem. 

21
00:00:50,093 --> 00:00:51,473
Tell us more about Goodhart's 
law. 

22
00:00:51,523 --> 00:00:53,495
This is also something that is 
quite 
 timely and relevant to 

23
00:00:53,495 --> 00:00:55,281
talk about. 
When a measure becomes a target,

24
00:00:55,281 --> 00:00:58,375

 it stops being a good measure.
The moment you measure AI 
 

25
00:00:58,383 --> 00:01:01,663
output, people optimize that 
metric, not the outcome we want.

26
00:01:01,703 --> 00:01:04,870
We also saw this in this 
tokenmaxxing, 
 where some 

27
00:01:04,870 --> 00:01:07,883
companies try to say like more 
tokens is better. 

28
00:01:07,933 --> 00:01:10,710
People talk about, "I can use AI

 and it just runs by itself," 

29
00:01:10,710 --> 00:01:12,183
because people assume things 
will be easy. 

30
00:01:12,264 --> 00:01:13,643
Let's talk about Dunning-Kruger 
effect. 

31
00:01:13,833 --> 00:01:17,262
It says that people with low 
ability 
 have tasks 

32
00:01:17,262 --> 00:01:19,743
overestimated, while experts 
underestimate themselves. 

33
00:01:19,993 --> 00:01:23,613
And when we bound this to AI 
 
code generation, people now ship

34
00:01:23,653 --> 00:01:26,554
convincing looking code they 
 
can't evaluate because the tool 

35
00:01:26,554 --> 00:01:27,853
hides the gap in their 
competence. 

36
00:01:44,336 --> 00:01:46,782
Hey, quick pause. 
My goal with Tech Lead Journal 

37
00:01:46,782 --> 00:01:49,366
is simple. 
Learn from the best in tech 
 so

38
00:01:49,366 --> 00:01:52,736
we can all grow together. 
If this resonates with you, hit 

39
00:01:52,736 --> 00:01:54,386
 subscribe to follow the 
channel. 

40
00:01:54,476 --> 00:01:57,362
It's the biggest way for you to 
support 
 the show and help us 

41
00:01:57,362 --> 00:01:59,396
keep bringing great guests and 
insights to you. 

42
00:01:59,936 --> 00:02:01,856
Thanks for being here, 
 and 
let's get back to it. 

43
00:02:03,140 --> 00:02:05,084
Hello everyone. 
Welcome back to another new 

44
00:02:05,084 --> 00:02:06,820
episode 
 of the Tech Lead 
Journal podcast. 

45
00:02:07,120 --> 00:02:10,580
I have a repeat guest today. 
Very excited to meet 
 Milan 

46
00:02:10,580 --> 00:02:14,272
Milanovic again. 
So I checked earlier today, I 

47
00:02:14,272 --> 00:02:17,620
think we 
 did an episode almost
two years ago. 

48
00:02:18,640 --> 00:02:21,274
So that was about, you know, how
to become 
 a great software 

49
00:02:21,274 --> 00:02:24,360
engineers and all that. 
Milan actually just recently 
 

50
00:02:24,368 --> 00:02:27,780
published another book titled 
Laws of Software Engineering. 

51
00:02:28,500 --> 00:02:31,470
I think it's a very interesting 
 topic, right, talking about 

52
00:02:31,480 --> 00:02:34,575
laws of software engineering. 
For some of you, maybe you may 

53
00:02:34,575 --> 00:02:37,692
have heard 
 a few laws here and
there, but Milan actually came 

54
00:02:37,692 --> 00:02:42,055
up with 56 laws in his 
 books, 
and even published it on a 

55
00:02:42,055 --> 00:02:43,916
website. 
So we'll talk a lot more about 

56
00:02:43,916 --> 00:02:45,877
these laws. 
Are they still applicable in 

57
00:02:45,877 --> 00:02:48,723
this AI era? 
And yeah, welcome, Milan, to the

58
00:02:48,723 --> 00:02:50,150
show. 
Looking forward for this 

59
00:02:50,150 --> 00:02:53,111
exciting topic. 
Thank you, Henry, for the 

60
00:02:53,111 --> 00:02:55,925
invitation. 
So we meet again to talk about 


61
00:02:55,933 --> 00:02:59,010
this, I would say, an 
interesting topic and topic that

62
00:02:59,010 --> 00:03:01,611
will be 
 with us for many years
to come. 

63
00:03:02,404 --> 00:03:05,170
Yeah. 
So Milan, let's go, 
 the first 

64
00:03:05,170 --> 00:03:08,002
place, right? 
How… What made you 
 write this 

65
00:03:08,002 --> 00:03:09,674
book, right? 
What's the background, right? 

66
00:03:09,724 --> 00:03:12,384
Especially when everyone is 
talking 
 about AI, you seem to 

67
00:03:12,384 --> 00:03:15,648
go back to the root of all 
things, which is talking 
 about

68
00:03:15,648 --> 00:03:18,869
law of software engineering. 
Yes, exactly. 

69
00:03:19,169 --> 00:03:25,889
Well, I probably as many others,
in our 
 field of software 

70
00:03:25,889 --> 00:03:31,696
engineering, learned some of 
these laws in their like 

71
00:03:31,696 --> 00:03:35,879
careers. 
When you start as a junior 
 

72
00:03:35,887 --> 00:03:39,681
software engineer and then rise 
up in the ladder, you fail a 

73
00:03:39,681 --> 00:03:45,463
lot. 
And some of these fails, 

74
00:03:45,463 --> 00:03:49,309
failures are, actually 
 
described by some of these laws.

75
00:03:50,829 --> 00:03:54,164
I was always impressed when I 
figured out 
 like, well, okay, 

76
00:03:54,164 --> 00:03:57,909
this thing has a name. 
So someone like twenty years ago

77
00:03:57,909 --> 00:04:03,329

 named this problem I have now.
And I started collecting them. 

78
00:04:03,439 --> 00:04:08,177
For some things I didn't have 
na- 
 like law names, I have 

79
00:04:08,177 --> 00:04:12,649
some, like, my explanation of 
the thing that happened. 

80
00:04:13,399 --> 00:04:18,288
But during the time I slowly 
tried 
 to, you know, put this 

81
00:04:18,288 --> 00:04:23,669
on the paper. 
Then I saw that also many 
 

82
00:04:23,677 --> 00:04:28,258
other people did the same. 
So they also collected these 

83
00:04:28,258 --> 00:04:31,631
things. 
Some things had names, some 

84
00:04:31,631 --> 00:04:36,245
things 
 didn't have name, but 
in the end, when I started to… 

85
00:04:36,245 --> 00:04:40,486
Like when I decided 
 to start 
working on this book, because I 

86
00:04:40,486 --> 00:04:46,981
thought it would be worthwhile, 
I researched like a lot on this 

87
00:04:46,981 --> 00:04:51,933
topic 
 and find, figured out 
that like, there are like a lot 

88
00:04:51,933 --> 00:04:55,514
of these laws scattered 
 all 
around, but not on the one 

89
00:04:55,514 --> 00:04:58,789
place. 
And so no one could sit on one 

90
00:04:58,789 --> 00:05:01,505
place 
 and learn everything 
about them that should be 

91
00:05:01,505 --> 00:05:05,326
learned, you know. 
Yeah, so when you talk about 

92
00:05:05,326 --> 00:05:07,681
failure, 
 right, I also 
remember, you know, myself 

93
00:05:07,681 --> 00:05:11,726
learning some of these laws. 
Definitely no one taught 
 them 

94
00:05:11,726 --> 00:05:14,767
in the university. 
Maybe some lecturer would 
 have

95
00:05:14,767 --> 00:05:17,063
mentioned it, right? 
But it's like nothing in 
 the 

96
00:05:17,063 --> 00:05:20,287
theory book, right? 
Whenever we learn from our 

97
00:05:20,287 --> 00:05:23,800
career or 
 experience, some of 
these laws is actually kind of 

98
00:05:23,800 --> 00:05:25,042
like, wow, very insightful, 
right? 

99
00:05:25,112 --> 00:05:29,232
Someone already figured it out. 
And yeah, it's a, it has a 
 

100
00:05:29,240 --> 00:05:31,732
name, so which is something that
is also interesting, right? 

101
00:05:32,762 --> 00:05:37,221
So I've recently also seen your 
 post that your book has been 

102
00:05:37,221 --> 00:05:42,143
sold thousand-plus copies, 
right, for 
 like 74 countries 

103
00:05:42,143 --> 00:05:45,357
around the world. 
This is something quite 
 unique

104
00:05:45,357 --> 00:05:48,639
in my view, right? 
So there's still a lot of people

105
00:05:48,639 --> 00:05:50,562

 interested in these kind of 
topics. 

106
00:05:50,802 --> 00:05:54,842
So tell me more about your 
 
wide reach of audience here. 

107
00:05:55,622 --> 00:05:58,023
What do you feel about that? 
Yes. 

108
00:05:58,333 --> 00:06:03,338
I analyzed like, book, like how 
it sell 
 in the first two 

109
00:06:03,338 --> 00:06:09,113
months and it was like 1,500 or 
more copies in 74 countries. 

110
00:06:09,113 --> 00:06:13,123
Like that number was not like 
 
interesting, but the number of 

111
00:06:13,313 --> 00:06:19,277
countries was on the one side. 
But on the other side, maybe it 

112
00:06:19,277 --> 00:06:23,293
is 
 not because many of these 
laws are the same in all 

113
00:06:23,293 --> 00:06:26,638
countries, so it 
 doesn't 
change from San Francisco to 

114
00:06:26,638 --> 00:06:30,781
London or Belgrade or Singapore.
So they are, they hold their 

115
00:06:30,781 --> 00:06:33,683
ground 
 on, in all countries 
and in places. 

116
00:06:33,683 --> 00:06:37,943
So from that side, I think 
 it 
could be expected. 

117
00:06:39,324 --> 00:06:41,944
Yeah, nice. 
I think it's very true, right? 

118
00:06:41,944 --> 00:06:44,654
These laws wouldn't change 
 
based on your location, your 

119
00:06:44,654 --> 00:06:47,234
language, software programming 

language, and things like that. 

120
00:06:47,234 --> 00:06:49,164
So I think, yeah, congrats for 
that. 

121
00:06:49,584 --> 00:06:52,964
So one thing that I would like 

to ask before we maybe go into 

122
00:06:52,964 --> 00:06:55,684
some of these laws, right? 
How should we actually 
 use 

123
00:06:55,684 --> 00:06:58,304
your book, right? 
Is it like a dictionary? 

124
00:06:58,304 --> 00:07:00,014
Is it something to read cover to
cover? 

125
00:07:00,044 --> 00:07:02,934
Like how do you advise people 
 
to read and use your book? 

126
00:07:04,793 --> 00:07:07,741
Yes. 
My recommendation is like, 

127
00:07:07,741 --> 00:07:13,421
because 
 many of these laws are
like on a different level of 

128
00:07:13,421 --> 00:07:16,703
abstraction. 
So some are on the like team 

129
00:07:16,703 --> 00:07:21,013
level, 
 some on architecture 
level, some on even on code 

130
00:07:21,013 --> 00:07:24,943
level, on like people 
 level 
and so on different levels. 

131
00:07:25,713 --> 00:07:30,373
So you cannot know all of 
 
these laws in your head. 

132
00:07:30,813 --> 00:07:34,681
That like, there are fifty-six 
laws, 
 but there are more like 

133
00:07:34,681 --> 00:07:38,677
than sixty or even seventy 
principles, different 
 like 

134
00:07:38,677 --> 00:07:42,443
smaller patterns and principles 
which are part of these laws. 

135
00:07:42,983 --> 00:07:48,799
So my recommendation is to read 
the 
 book once, so you have a 

136
00:07:48,799 --> 00:07:53,072
grasp of like, what can happen, 
you know, on 
 these different 

137
00:07:53,072 --> 00:07:56,918
level of abstractions. 
Then when the, like, when the 

138
00:07:56,918 --> 00:08:01,167
problem 
 happens tomorrow or 
some issue you detect, you will 

139
00:08:01,167 --> 00:08:04,206
remember, "Aha, I 
 saw this. 
This is that law." 

140
00:08:04,206 --> 00:08:09,609
Then you go again to the book or
my 
 website and then check the 

141
00:08:09,609 --> 00:08:12,045
details. 
There are of course, 
 more 

142
00:08:12,045 --> 00:08:17,525
details in the book, how to 
better understand this 
 problem

143
00:08:17,525 --> 00:08:21,153
and how to resolve it, you know,
in a proper way. 

144
00:08:22,741 --> 00:08:27,289
Yeah, so I think, for those of 
us 
 who have been around in our

145
00:08:27,289 --> 00:08:29,953
career, I think some of these 
laws are 
 definitely very 

146
00:08:29,953 --> 00:08:32,384
interesting, right? 
But for those of you who are 

147
00:08:32,384 --> 00:08:35,518
new, maybe 
 in the software 
industry, or you may never heard

148
00:08:35,518 --> 00:08:39,751
about these laws before, right? 
I think my advice also would be 

149
00:08:39,751 --> 00:08:42,938
to just 
 at least walk through 
all the books, right, all the 

150
00:08:42,938 --> 00:08:45,717
title, maybe the summary. 
You put actually a good summary,

151
00:08:45,717 --> 00:08:48,343
at 
 the beginning of each of 
the law. 

152
00:08:48,353 --> 00:08:50,563
So at least you kind of like 
 
resonate with some of them. 

153
00:08:51,203 --> 00:08:54,933
So speaking about laws, there 
 
are 56 or maybe more, right, 

154
00:08:54,953 --> 00:08:58,125
like you mentioned, right? 
But what I'm interested is the 


155
00:08:58,133 --> 00:09:01,549
like, the span of categories. 
So it's not just like when we 

156
00:09:01,549 --> 00:09:04,149
talk 
 about laws, some- maybe 
some people think, oh yeah, 

157
00:09:04,149 --> 00:09:06,473
maybe it's all 
 about 
architecture, design, coding 

158
00:09:06,473 --> 00:09:07,843
principles, and things like 
that. 

159
00:09:08,053 --> 00:09:09,753
But actually you have 
 
something more, right? 

160
00:09:09,753 --> 00:09:12,313
You talk about teams. 
You talk about planning. 

161
00:09:12,593 --> 00:09:14,883
You talk about scale and making 
decisions. 

162
00:09:14,893 --> 00:09:18,391
So tell us why there are so many
wide 
 variety of categories 

163
00:09:18,391 --> 00:09:20,443
talking about laws of software 
engineering, right? 

164
00:09:21,282 --> 00:09:25,822
Yes, exactly. 
What we think in software 

165
00:09:25,822 --> 00:09:29,792
usually, 
 especially when we 
are like starting in this field,

166
00:09:29,792 --> 00:09:32,698
that everything 
 is technical, 
so we are only interested in 

167
00:09:32,698 --> 00:09:36,792
technical stuff. 
But what I figured out during my

168
00:09:36,792 --> 00:09:40,622

 like own career, that software
failures are rarely purely 

169
00:09:40,622 --> 00:09:44,612
technical. 
So one production incident can 


170
00:09:44,620 --> 00:09:49,888
lead to organization chart or, 
like to metric or even to some 


171
00:09:49,896 --> 00:09:54,720
kind of a bias, that happened. 
So the book tried to 
 reflect 

172
00:09:54,720 --> 00:09:58,183
that in a way. 
Like architecture decisions 

173
00:09:58,183 --> 00:10:02,122
constrain 
 people decisions, 
and then these people decisions 

174
00:10:02,122 --> 00:10:06,012
move timelines that 
 we have on
projects, and then these 

175
00:10:06,292 --> 00:10:08,982
timelines force quality 
trade-offs. 

176
00:10:09,392 --> 00:10:16,295
So people try to build complex 

systems under limited of, like 

177
00:10:16,295 --> 00:10:19,102
limited time, knowledge, and 
coordination. 

178
00:10:19,112 --> 00:10:22,810
So you need actually all these 
laws 
 about people, math, and 

179
00:10:22,810 --> 00:10:26,162
organization, not only technical
and code stuff, you know. 

180
00:10:27,931 --> 00:10:30,201
Yeah, so many people 
 have 
mentioned about it. 

181
00:10:30,471 --> 00:10:32,781
Software engineering is a 
 
socio-technical thing, right? 

182
00:10:32,781 --> 00:10:36,105
So it's not just the technical 

aspects that you need to think 

183
00:10:36,105 --> 00:10:40,127
about, but it's also the social 
aspect like 
 people, you know, 

184
00:10:40,127 --> 00:10:42,586
project management. 
I think that's always… You 
 

185
00:10:42,594 --> 00:10:45,291
mentioned something that also 
quite insightful to me, right? 

186
00:10:45,291 --> 00:10:48,159
So many of the software 
failures, 
 software project 

187
00:10:48,159 --> 00:10:51,531
failures is actually not because
of just purely technical, right?

188
00:10:51,531 --> 00:10:54,111
So it's like in the people 
aspect, right? 

189
00:10:54,141 --> 00:10:57,109
Maybe in the making the 
decision, putting 
 something 

190
00:10:57,109 --> 00:11:00,606
that is probably not like too 
complex for the stake- state of 

191
00:11:00,606 --> 00:11:02,481
the 
 project at that point in 
time, right? 

192
00:11:02,481 --> 00:11:06,061
So I think some of these 
 
definitely very, you know, very 

193
00:11:06,071 --> 00:11:09,237
exciting to cover some of them. 
So let's go into some of 
 your 

194
00:11:09,237 --> 00:11:11,903
favorites, probably. 
Any one that you would like to 


195
00:11:11,911 --> 00:11:18,554
pick first as the law to cover? 
Yeah, there are laws a few that 

196
00:11:18,554 --> 00:11:22,950
I like 
 the most because I 
think they are, they exist 

197
00:11:22,950 --> 00:11:27,234
almost in all projects or the 
way 
 how we try to work in this

198
00:11:27,234 --> 00:11:30,794
field, at least until we 
understand them properly. 

199
00:11:31,574 --> 00:11:34,894
The first one is Gall's Law that

 say, like, complex system that

200
00:11:34,924 --> 00:11:38,451
works is always found to have 
grown 
 from a simple one that 

201
00:11:38,451 --> 00:11:42,302
worked. 
So I took the example of 

202
00:11:42,302 --> 00:11:47,052
Instagram 
 in my book, where 
Instagram was like first some 

203
00:11:47,052 --> 00:11:50,174
kind of a check-in 
 service, 
with gamification. 

204
00:11:52,544 --> 00:11:57,119
But the thing was that this 
thing 
 didn't work properly for

205
00:11:57,119 --> 00:11:59,896
founders. 
And then they took one small 
 

206
00:11:59,904 --> 00:12:03,742
thing that was like image 
sharing and built Instagram for 

207
00:12:03,742 --> 00:12:06,284
that. 
So, from there. 

208
00:12:06,284 --> 00:12:11,684
So they started from a small 
thing 
 and then build up the 

209
00:12:11,684 --> 00:12:15,496
rest later. 
We… In software engineering 

210
00:12:15,496 --> 00:12:20,014
field, 
 we had similar thing in
previous years with this 

211
00:12:20,014 --> 00:12:23,545
microservices moment. 
Everyone started to put 

212
00:12:23,545 --> 00:12:27,674
microservices 
 in all projects,
even though probably most of 

213
00:12:27,674 --> 00:12:32,504
them didn't need them. 
So Gall's Law say like start 

214
00:12:32,504 --> 00:12:36,694
with 
 the monolith that works, 
start with simple thing and then

215
00:12:36,694 --> 00:12:40,312
split 
 off pieces when you are,
you know, you already understand

216
00:12:40,312 --> 00:12:43,502
them. 
The thing i- the same 
 thing is

217
00:12:43,502 --> 00:12:46,834
with AI now. 
So, you can now generate a lot 

218
00:12:46,834 --> 00:12:49,643
of code, 
 a lot of complex 
systems, you know. 

219
00:12:49,663 --> 00:12:53,303
But this is the trap. 
So generated complexity still 

220
00:12:53,303 --> 00:12:56,453
rests 
 on assumptions nobody 
validated. 

221
00:12:56,513 --> 00:13:03,135
So we saw a lot of generated 
stuff, 
 but still we didn't saw

222
00:13:03,135 --> 00:13:07,363
some great software in the last 
few years 
 that happened to us,

223
00:13:07,363 --> 00:13:10,944
you know. 
Yeah, so, it's very 
 

224
00:13:10,952 --> 00:13:13,150
interesting law, right? 
Gall's law, right? 

225
00:13:14,070 --> 00:13:17,728
So, almost every working system,

 right, a complex system 

226
00:13:17,728 --> 00:13:19,830
actually start from something 
that is simple, right? 

227
00:13:20,080 --> 00:13:23,452
Is this something also quite 
related 
 to the premature 

228
00:13:23,452 --> 00:13:26,074
optimization thing? 
You know, like the root of, you 

229
00:13:26,074 --> 00:13:29,236
 know, of evil is like… 
premature optimization is the 

230
00:13:29,236 --> 00:13:31,402
root of complexity. 
Sorry, it's not the root of 

231
00:13:31,402 --> 00:13:33,590
evil. 
The root of complexity is 
 like

232
00:13:33,590 --> 00:13:35,460
premature optimization. 
Is that also something similar? 

233
00:13:36,611 --> 00:13:41,177
Yes, exactly. 
This is similar, although, 

234
00:13:41,177 --> 00:13:46,431
Donald 
 talked more about the 
code, but this is more on the 

235
00:13:46,431 --> 00:13:49,271
product side, 
 I would say, but
it is the same. 

236
00:13:49,271 --> 00:13:52,571
So don't overcomplicate from the
start. 

237
00:13:52,591 --> 00:13:55,123
Start simple. 
They're more or less talking 
 

238
00:13:55,131 --> 00:13:59,685
about the same thing. 
So if you try to create super 
 

239
00:13:59,693 --> 00:14:03,040
complex thing from the start, 
you will probably fail, you 

240
00:14:03,040 --> 00:14:05,389
know. 
Because there are too much 
 

241
00:14:05,397 --> 00:14:08,591
unknowns in building super 
complex stuff from the start. 

242
00:14:08,591 --> 00:14:12,484
So building the small thing, and
this 
 is actually directly 

243
00:14:12,484 --> 00:14:15,951
bound to MVP approach, which 
goes well with that. 

244
00:14:16,411 --> 00:14:20,033
Start with the simple things. 
Build a small MVP and 
 then 

245
00:14:20,033 --> 00:14:22,331
build up from there. 
Don't build like something 
 

246
00:14:22,339 --> 00:14:28,640
really big from the start. 
Yeah, so I think we all used to 

247
00:14:28,640 --> 00:14:31,426
 hear about all these 
microservices versus monolith 

248
00:14:31,426 --> 00:14:34,040
debate right? 
Although maybe lately it's less 

249
00:14:34,040 --> 00:14:36,845
 prominent on the internet 
simply because everyone just 

250
00:14:36,845 --> 00:14:39,512
talks about AI. 
But I still remember, you know, 

251
00:14:39,512 --> 00:14:42,836
 we all have this passion about,
you know, certain technologies, 

252
00:14:42,836 --> 00:14:45,730
 certain complex systems, right?
And we tend to use that. 

253
00:14:46,080 --> 00:14:49,360
So speaking about AI, you 
mentioned 
 there's the trap of 

254
00:14:49,360 --> 00:14:52,897
using AI and it ended up with, 
you know, complex systems, or 
 

255
00:14:52,905 --> 00:14:54,790
maybe complex code in the first 
place. 

256
00:14:55,100 --> 00:14:58,189
So how would you advise us to 
actually 
 apply this Gall's Law

257
00:14:58,189 --> 00:15:00,795
when using AI? 
Is there some tips maybe 
 that 

258
00:15:00,795 --> 00:15:07,287
you have tried? 
Yes. regarding AI, I wrote 

259
00:15:07,287 --> 00:15:12,771
recently about that. 
I think, what is more important 

260
00:15:12,771 --> 00:15:16,039
is 
 not like big production of 
code. 

261
00:15:16,109 --> 00:15:19,435
That like, I see everyone is 
like, 
 "Okay, I'm using 10 

262
00:15:19,435 --> 00:15:22,365
agents, 20 agents, 30 agents." 
So there are 
 so many agents 

263
00:15:22,365 --> 00:15:26,109
working all the time, but where 
is the judgment? 

264
00:15:26,359 --> 00:15:31,219
And we need to have the right 
 
judgment, you know, what we want

265
00:15:31,259 --> 00:15:36,829
to build, and also to validate 

what was built by those agents. 

266
00:15:36,829 --> 00:15:41,329
Is this like something really 
 
valuable for us or not? 

267
00:15:41,389 --> 00:15:45,227
And like some of these laws are 
actually, 
 including Gall's 

268
00:15:45,227 --> 00:15:49,249
law, talking about that. 
Don't just try to produce 

269
00:15:49,249 --> 00:15:52,715
something 
 because you can now,
you know, easily produce, which 

270
00:15:52,715 --> 00:15:56,149
is like 
 a nice thing for all 
of us. 

271
00:15:56,589 --> 00:16:02,479
But we need to have, like more 

shift now toward the other side.

272
00:16:02,529 --> 00:16:07,499
Like on, from how to 
 why and 
what, you know. 

273
00:16:08,339 --> 00:16:11,149
Just to think more on that 
 
side, on the thinking side. 

274
00:16:11,199 --> 00:16:14,369
I see that no one is 
 thinking,
everyone is doing. 

275
00:16:14,669 --> 00:16:16,939
And there is always problem with
that. 

276
00:16:17,179 --> 00:16:22,255
I heard that from one of my CEOs
once, 
 who said to me like, why

277
00:16:22,255 --> 00:16:25,709
is he don't do it, doing these 
operational things? 

278
00:16:25,909 --> 00:16:28,259
Because when you're doing 
 
operations, you cannot think. 

279
00:16:28,269 --> 00:16:32,875
You don't have time to think. 
And this is somehow, that is 

280
00:16:32,875 --> 00:16:38,950
like, I 
 also find out, that in
this way, no one is thinking, 

281
00:16:38,950 --> 00:16:41,861
everyone is building. 
But you cannot build great 
 

282
00:16:41,869 --> 00:16:44,877
stuff without thinking. 
This is just, you know, my two 


283
00:16:44,885 --> 00:16:47,788
cents on this whole AI movement.
Yeah. 

284
00:16:48,648 --> 00:16:51,928
Especially people who 
 love 
vibe coding, right? 

285
00:16:52,298 --> 00:16:54,688
Or using multiple agents, 
 
right, like what you mentioned. 

286
00:16:54,718 --> 00:16:57,393
Or even like one-shot an 
application, 
 you know, like I 

287
00:16:57,393 --> 00:17:00,498
don't know, build a game 
simulation of blah, blah, blah, 

288
00:17:00,498 --> 00:17:04,038
and the AI 
 just came up with 
the program altogether. 

289
00:17:04,038 --> 00:17:08,136
So I think definitely if you 
just 
 write it once, you know, 

290
00:17:08,136 --> 00:17:10,278
and use it once, probably it's 
okay, right? 

291
00:17:10,278 --> 00:17:13,385
But if you try to evolve it and 
make it a 
 successful product, 

292
00:17:13,385 --> 00:17:17,308
I think it takes more than just,
you know, using AI to do that. 

293
00:17:17,598 --> 00:17:21,016
So I think, yeah, it's very… 
 I
would say it's very timely to 

294
00:17:21,016 --> 00:17:23,468
mention about this, right? 
Because we all focus on 

295
00:17:23,468 --> 00:17:26,324
building, 
 you know, asking AI,
prompting AI, thinking about 

296
00:17:26,324 --> 00:17:28,758
what's the best context 
 
engineering, prompt engineering,

297
00:17:28,978 --> 00:17:32,095
loop engineering and all that. 
But thinking about the what and 

298
00:17:32,095 --> 00:17:33,418
 the why I think is still 
important. 

299
00:17:34,228 --> 00:17:38,338
So let's pick the second law. 
Any law that you want to cover 

300
00:17:38,338 --> 00:17:40,111
as well? 
Yes. 

301
00:17:40,161 --> 00:17:44,747
I mean, we must, mention 
Conway's Law, 
 that say like 

302
00:17:44,747 --> 00:17:48,221
organization shape systems that 
copy their communication 

303
00:17:48,221 --> 00:17:52,101
structure, 
 which means like 
your organization shape your 

304
00:17:52,101 --> 00:17:56,017
software architecture. 
So if you have team, four teams 

305
00:17:56,017 --> 00:17:59,775
 that build a compiler, they 
will build compiler in four 

306
00:17:59,775 --> 00:18:02,596
passes. 
So the organizational chart 
 

307
00:18:02,604 --> 00:18:08,327
shows up in their architecture. 
In modern frame, when we are 
 

308
00:18:08,335 --> 00:18:11,139
talking about microservices 
which we mentioned, so 

309
00:18:11,139 --> 00:18:14,471
microservice 
 boundaries mirror
the team boundaries. 

310
00:18:15,031 --> 00:18:17,901
And there is this thing called 
the 
 reverse Conway maneuver, 

311
00:18:17,901 --> 00:18:21,313
means that you can change the 
team structure first, and 
 then

312
00:18:21,313 --> 00:18:23,221
you get the architecture you 
want. 

313
00:18:23,971 --> 00:18:30,741
And this is not changed in AI, 
also 
 because it can be human 

314
00:18:30,741 --> 00:18:34,875
or AI agent. 
So if you build a multi-agent 

315
00:18:34,875 --> 00:18:38,603
system, 
 with the separate 
agents, with like they have some

316
00:18:38,603 --> 00:18:41,911
handoff contracts, the 
 system 
creates around those handoffs 

317
00:18:42,221 --> 00:18:46,121
and the same like microservices 
 builds around team boundaries. 

318
00:18:46,611 --> 00:18:50,284
So whoever own the planner 
agents 
 versus the, you know, 

319
00:18:50,284 --> 00:18:54,949
executor agents builds 
boundaries that 
 mirror their 

320
00:18:54,949 --> 00:18:58,671
own organization. 
So here Conway runs twice. 

321
00:18:58,681 --> 00:19:02,241
So human organization shape 
 
agent topology, and then agent 

322
00:19:02,371 --> 00:19:08,857
topology shape system output. 
So we should always start with 


323
00:19:08,865 --> 00:19:13,421
the question, what actually, 
what kind of architecture we 

324
00:19:13,421 --> 00:19:17,141
need. 
And then when we define this, we

325
00:19:17,141 --> 00:19:21,873
need 
 to restructure our 
organization, to be in line with

326
00:19:21,873 --> 00:19:25,875
that on our project. 
Because this design ownership 

327
00:19:25,875 --> 00:19:29,782
and 
 communication structure 
first and then our architecture 

328
00:19:29,782 --> 00:19:33,121
will be enabled 
 by that 
organizational structure. 

329
00:19:34,484 --> 00:19:38,648
Yeah, so I think Conway's Law 
 
is, like I think I still 

330
00:19:38,648 --> 00:19:40,004
remember Vaughn Vernon mentioned
it. 

331
00:19:40,004 --> 00:19:42,104
It's like a law of gravity, 
right? 

332
00:19:42,104 --> 00:19:44,821
You cannot run away from it. 
So I think, especially… Exactly,

333
00:19:44,821 --> 00:19:47,571
exactly. 
To me, I think you can only 
 

334
00:19:47,579 --> 00:19:49,974
learn it in the hard way, right?
In your project, right? 

335
00:19:51,064 --> 00:19:53,722
But I think one of the most 
difficult 
 thing about Conway's

336
00:19:53,722 --> 00:19:56,166
Law, although you mentioned 
about reverse you 
 know, 

337
00:19:56,166 --> 00:19:59,009
Conway's Law maneuver, right? 
So some people just started, 
 

338
00:19:59,017 --> 00:20:02,034
you know, with the organization 
structure like it is, right? 

339
00:20:02,034 --> 00:20:05,568
So nobody actually attempt to 
change 
 team structure because,

340
00:20:05,568 --> 00:20:09,049
you know, it takes a lot more 
effort, you know, maybe 
 

341
00:20:09,057 --> 00:20:11,508
changing the manager, maybe 
restructuring the team, and all 

342
00:20:11,508 --> 00:20:14,257
that, right? 
It takes a lot more 
 people to 

343
00:20:14,257 --> 00:20:17,430
be coordinated. 
So maybe what would you have as 

344
00:20:17,430 --> 00:20:21,544
 an advice, you know, for many, 
especially the big corporates, 

345
00:20:21,544 --> 00:20:23,654
right? 
Where they already have a 

346
00:20:23,654 --> 00:20:26,372
structure, they 
 already build 
teams, they have managers, but 

347
00:20:26,372 --> 00:20:28,996
they wanna build a new product. 
So what would be your advice 

348
00:20:28,996 --> 00:20:31,783
here? 
Yeah, there is even a wording 
 

349
00:20:31,791 --> 00:20:34,006
that HR department actually 
create architecture because of 

350
00:20:34,006 --> 00:20:36,131
that. 
You know, because they're 

351
00:20:36,131 --> 00:20:38,363
involved 
 in creation of 
organization. 

352
00:20:39,253 --> 00:20:43,463
What is my proposal is always to

 try to include also architects

353
00:20:43,983 --> 00:20:50,183
in organizational definition. 
So when the manager create, try 

354
00:20:50,183 --> 00:20:54,277
to 
 create its organization 
that will build software, also 

355
00:20:54,277 --> 00:20:58,363
include architects there. 
Because they can also give 

356
00:20:58,363 --> 00:21:02,953
useful 
 insights on this topic 
and help to shape organization 

357
00:21:02,953 --> 00:21:07,053
towards a wanted 
 architecture 
we want to have, you know. 

358
00:21:07,533 --> 00:21:10,913
So this is for sure one 
 good 
step for everyone. 

359
00:21:12,628 --> 00:21:16,546
This is my first time hearing, 
you 
 know, involving architects

360
00:21:16,546 --> 00:21:18,618
in, you know, our organization 
structure. 

361
00:21:18,618 --> 00:21:20,730
But I think, yeah, thinking 
about it, 
 maybe it's not a 

362
00:21:20,730 --> 00:21:23,276
such a bad idea, right? 
So having a software architect 

363
00:21:23,276 --> 00:21:25,178
to 
 actually help structure the
team. 

364
00:21:26,348 --> 00:21:29,908
And speaking about AI, right, 
many people 
 these days assume,

365
00:21:29,908 --> 00:21:33,400
you know, the team size will be 
getting smaller and smaller 
 

366
00:21:33,408 --> 00:21:35,468
simply because now we have AI, 
right? 

367
00:21:35,808 --> 00:21:38,802
And that also means like your 
organization 
 structure will 

368
00:21:38,802 --> 00:21:41,354
probably shrink, right? 
And maybe any kind of a team 

369
00:21:41,354 --> 00:21:44,368
will be 
 just a size of, I 
don't know, like three, maybe to

370
00:21:44,368 --> 00:21:50,108
seven kind of people, right? 
This tends to create like a 

371
00:21:50,108 --> 00:21:53,814
smaller 
 team size, like maybe 
communicating, coordinating with

372
00:21:53,814 --> 00:21:56,955
each other. 
Do you think there is such an 

373
00:21:56,955 --> 00:22:01,098
insight 
 as well about Conway's
Law, like for teams that are, 

374
00:22:01,098 --> 00:22:04,848
you know, like a 
 generalist 
using AI to build products? 

375
00:22:05,088 --> 00:22:07,608
Is there something that is 
 
worth mentioning here as well? 

376
00:22:08,959 --> 00:22:13,979
Yeah, I think it, it's like 
 
Amazon two-pizza team, becomes 

377
00:22:14,139 --> 00:22:18,715
like Amazon one-pizza team. 
And I wrote about this in 
 my 

378
00:22:18,715 --> 00:22:22,513
book in one of the laws. 
I think the thing will be the 

379
00:22:22,513 --> 00:22:24,741
same. 
As I mentioned, with people and 

380
00:22:24,741 --> 00:22:28,787
 also with smaller number of 
people and with agents, whom 

381
00:22:28,787 --> 00:22:33,668
these people 
 are using to 
produce software, it will, 

382
00:22:33,668 --> 00:22:39,526
Conway's Law will still hold. 
So because everything we do and 

383
00:22:39,526 --> 00:22:45,077
 how we do in our organization 
will directly reflect to our 

384
00:22:45,077 --> 00:22:48,136
architecture. 
So people need to have this in 

385
00:22:48,136 --> 00:22:50,309
their 
 minds when they start to
working. 

386
00:22:50,759 --> 00:22:54,701
So this is for anything more 
than 
 one person company, you 

387
00:22:54,701 --> 00:22:57,019
know. 
It will always help. 

388
00:22:58,132 --> 00:23:01,702
Yeah, so I also read somewhere 
when 
 people talking about 

389
00:23:01,702 --> 00:23:05,924
applying AI, right? 
So, think of it as if like you 

390
00:23:05,924 --> 00:23:07,892
have 
 everything from scratch, 
right? 

391
00:23:07,922 --> 00:23:10,362
And you design with AI in place,
right? 

392
00:23:10,632 --> 00:23:12,192
Maybe thinking about agents, 
right? 

393
00:23:12,192 --> 00:23:14,970
Because if you still rely on 
your current 
 organization 

394
00:23:14,970 --> 00:23:18,380
structure, you would end up with
the same number of AI agents 
 

395
00:23:18,388 --> 00:23:20,012
maybe, right, or AI systems, 
right? 

396
00:23:20,122 --> 00:23:23,800
So I think this is where 
Conway's Law, 
 again, is very 

397
00:23:23,800 --> 00:23:27,373
applicable to this. 
So the next I would like 
 to 

398
00:23:27,373 --> 00:23:29,559
pick Goodhart's Law. 
I think this is also something 

399
00:23:29,559 --> 00:23:31,602
that is 
 quite timely and 
relevant to talk about. 

400
00:23:31,602 --> 00:23:33,562
So tell us more about Goodhart's
Law. 

401
00:23:34,343 --> 00:23:35,383
Yes. 
Yes, it is. 

402
00:23:36,083 --> 00:23:39,227
Goodhart's law is always around.
So it say basically that when 
 

403
00:23:39,235 --> 00:23:42,413
a measure becomes a target, it 
stop being a good measure. 

404
00:23:42,923 --> 00:23:48,383
And we also see this now as we 

saw this, like fifty years ago. 

405
00:23:48,893 --> 00:23:55,841
Before, we saw measures like 
lines of code that were like 

406
00:23:55,841 --> 00:23:57,793
really a 
 bad proxy in '80s and
'90s. 

407
00:23:58,673 --> 00:24:01,563
But it is worse now because 
 
you have a model that writes 

408
00:24:01,563 --> 00:24:06,105
five hundred lines in a second. 
So the moment you measure AI 
 

409
00:24:06,113 --> 00:24:09,773
output, people optimize that 
metric, not the outcome we want.

410
00:24:09,773 --> 00:24:17,003
So it is important, when you 
 
create some metrics in your 

411
00:24:17,293 --> 00:24:21,730
organization, to always pair it 
 with some kind of a 

412
00:24:21,730 --> 00:24:23,912
counter-metric. 
So for example, speed against 

413
00:24:23,912 --> 00:24:27,423
stability. 
And then, you know, to always to

414
00:24:27,423 --> 00:24:31,893

 check, can someone game this, 
you know, and would this, like 

415
00:24:31,893 --> 00:24:34,953
this game 
 behavior still 
produce useful work. 

416
00:24:35,313 --> 00:24:38,490
So if that gaming produce 
 
garbage, the metric is 

417
00:24:38,490 --> 00:24:41,910
dangerous. 
The teams, like that win won't 

418
00:24:41,910 --> 00:24:45,732
be the ones 
 that generate the 
most code, I would say. 

419
00:24:46,292 --> 00:24:49,747
They will be the ones whose 
structure 
 isn't fighting their

420
00:24:49,747 --> 00:24:53,489
system, a company, or who 
measure outcome, not output, 
 

421
00:24:53,497 --> 00:24:56,632
which actually Goodhart's law 
said to us. 

422
00:24:57,202 --> 00:25:02,420
And when we talk about AI, we 
also saw 
 this in this 

423
00:25:02,420 --> 00:25:05,924
tokenmaxxing, where some 
companies try to, you know, say,

424
00:25:05,924 --> 00:25:08,852
like, 
 like more tokens the 
better for you. 

425
00:25:09,192 --> 00:25:13,743
And then, of course, people 
always try 
 to align their work

426
00:25:13,743 --> 00:25:16,762
towards that metric. 
They try to produce more and 

427
00:25:16,762 --> 00:25:19,476
more tokens. 
But the real outcome, the real 


428
00:25:19,484 --> 00:25:24,086
quality was not there, so they 
didn't achieve what the 

429
00:25:24,086 --> 00:25:27,692
organization wanted. 
They just optimized on that 
 

430
00:25:27,700 --> 00:25:31,961
particular target they had. 
Yeah. 

431
00:25:32,291 --> 00:25:35,375
So I mean, yeah, talking about 

Goodhart's law, definitely we 

432
00:25:35,375 --> 00:25:38,561
may have seen it from time to 
time, you know, 
 in the 

433
00:25:38,561 --> 00:25:41,167
software development world, 
maybe thinking about velocity, 

434
00:25:41,167 --> 00:25:43,266
right? 
You know, we used to talk about 

435
00:25:43,266 --> 00:25:46,556
 story points, lines of code, 
always, a classic kind of 

436
00:25:46,556 --> 00:25:48,376
example. 
People talk about DORA metrics 

437
00:25:48,376 --> 00:25:50,796
as well. 
And these days it's about 
 

438
00:25:50,804 --> 00:25:52,470
tokenmaxxing, right? 
Yes. 

439
00:25:52,991 --> 00:25:56,575
I think when we had the model 
like 
 cheaper than today, I 

440
00:25:56,575 --> 00:25:59,361
think tokenmaxxing is like a 
very, very applicable thing. 

441
00:25:59,361 --> 00:26:02,810
Many people are like just ask 
developers 
 to spend as much as

442
00:26:02,810 --> 00:26:05,899
possible on tokens, but I think 
there's a little bit of 
 pause 

443
00:26:05,899 --> 00:26:08,821
simply because the AI model cost
is rising at the same time, 

444
00:26:08,821 --> 00:26:11,576
right? 
So I think Goodhart's Law 
 

445
00:26:11,584 --> 00:26:14,411
definitely is very important, 
especially for leaders, right? 

446
00:26:15,171 --> 00:26:18,331
Because leaders sometimes 
 
create this, I would say like, I

447
00:26:18,331 --> 00:26:21,989
don't know, KPIs or part of the 
 performance review, a metrics, 

448
00:26:21,989 --> 00:26:25,351
right? 
You mentioned about having a 
 

449
00:26:25,359 --> 00:26:26,861
counter-metrics or 
counterbalance. 

450
00:26:27,331 --> 00:26:30,461
Any kind of a good examples that
maybe 
 leaders can use these 

451
00:26:30,461 --> 00:26:34,141
days, maybe kind of like a best 
practice in the industry now? 

452
00:26:34,141 --> 00:26:39,554
Like is there a such thing? 
I think we should always start 

453
00:26:39,554 --> 00:26:43,748
from 
 what kind of outcomes we 
want to focus. 

454
00:26:44,198 --> 00:26:48,638
So the focus should always 
 be 
outcome, so not output. 

455
00:26:48,738 --> 00:26:51,178
We are constantly trying to 
 
measure outputs, you know. 

456
00:26:51,328 --> 00:26:53,917
All of these things we're now 
 
talking about are outputs, lines

457
00:26:53,917 --> 00:26:58,233
of code, number of tokens. 
So, and the… what really 

458
00:26:58,233 --> 00:27:05,948
surprised 
 me with this 
tokenmaxxing is that like no one

459
00:27:05,948 --> 00:27:08,248
remembered this 
 Goodhart's 
Law, obviously. 

460
00:27:08,288 --> 00:27:14,638
Because every time when we had 

such thing, it bounced back in 

461
00:27:14,638 --> 00:27:17,964
the history of software. 
And this is one of the reasons 


462
00:27:17,972 --> 00:27:23,620
I, why I wrote the book. 
So we cannot measure things 
 in

463
00:27:23,620 --> 00:27:28,248
software like this, so in number
of things, you know. 

464
00:27:28,798 --> 00:27:33,168
Number of things produced will 

never produce a good value. 

465
00:27:34,738 --> 00:27:38,772
What will produce good value is 
like 
 what outcome we want to 

466
00:27:38,772 --> 00:27:41,988
achieve, and did we meet this 
outcome or not? 

467
00:27:42,238 --> 00:27:46,554
Or we can say also, did we meet 
in 
 timely manner with the 

468
00:27:46,554 --> 00:27:50,973
right resources and, you know, 
in the budget and all 
 of the 

469
00:27:50,973 --> 00:27:54,954
things, but not by measuring 
number of features, tokens, 

470
00:27:54,954 --> 00:27:57,703
whatever. 
So this should always be in mind

471
00:27:57,703 --> 00:28:01,408

 of especially leaders as you 
say, because they define and 

472
00:28:01,408 --> 00:28:05,048
shape this KPI 
 structure and 
the whole organization. 

473
00:28:06,853 --> 00:28:10,981
Yeah, so definitely, right? 
I think focusing on outcome is 

474
00:28:10,981 --> 00:28:12,993
definitely 
 is the most 
important thing, right? 

475
00:28:13,423 --> 00:28:16,248
But talking about it, right? 
Because it is such a common 
 

476
00:28:16,256 --> 00:28:18,949
thing in the industry, right? 
Such anti-pattern that 
 people 

477
00:28:18,949 --> 00:28:21,555
always follow, right? 
I think one of the difficulty, 


478
00:28:21,563 --> 00:28:25,129
right, is just like people find 
it hard to actually measure 

479
00:28:25,129 --> 00:28:27,411
outcome. 
Or maybe sometimes it's 
 also a

480
00:28:27,411 --> 00:28:30,198
lagging thing, right? 
You can only measure it 
 after 

481
00:28:30,198 --> 00:28:33,735
certain times, right? 
And that's why people focus a 

482
00:28:33,735 --> 00:28:37,103
lot on 
 maybe output or proxy 
metrics, right? 

483
00:28:37,323 --> 00:28:41,653
So yeah, this is probably one of

 the danger, because like it's 

484
00:28:41,653 --> 00:28:43,893
easier to just quantify and 
measure it. 

485
00:28:43,893 --> 00:28:46,217
So, and especially some tools 
actually 
 also produce this out

486
00:28:46,217 --> 00:28:48,533
of the box, so you can just use 
that, right? 

487
00:28:48,543 --> 00:28:52,047
So I think focusing on outcomes,
maybe 
 tell us like what kind 

488
00:28:52,047 --> 00:28:55,183
of outcomes that is, you know, 
like we should measure. 

489
00:28:55,193 --> 00:28:58,193
Is it like business outcome? 
Is it like technical outcome? 

490
00:28:59,000 --> 00:29:03,260
It can be either. 
So I would always ba-- like, we 

491
00:29:03,260 --> 00:29:05,360
 should always start from 
impact. 

492
00:29:05,400 --> 00:29:07,640
What kind of impact we want to 
achieve? 

493
00:29:07,770 --> 00:29:12,732
So okay, we want to be like the 
best 
 company in the world in 

494
00:29:12,732 --> 00:29:16,540
this area, or like we want to 
have like five 
 thousand users 

495
00:29:16,540 --> 00:29:19,180
per month, you know? 
Or we want, you know, 
 

496
00:29:19,188 --> 00:29:23,165
whatever, what impact we want. 
Then we translate this to 

497
00:29:23,165 --> 00:29:27,720
outcomes, 
 what kind of outcome
we need to achieve. 

498
00:29:27,740 --> 00:29:32,058
And for example, OKRs are good 
to 
 define and measure that 

499
00:29:32,058 --> 00:29:34,395
outcomes. 
And of course, then comes 
 

500
00:29:34,403 --> 00:29:37,230
outputs, you know, because they 
are connected to these outcomes.

501
00:29:37,260 --> 00:29:40,368
They are fine. 
They-- There could be 
 like 

502
00:29:40,368 --> 00:29:43,524
just indicators. 
So, okay, I listen for these 
 

503
00:29:43,532 --> 00:29:47,962
indicators, like these outputs. 
Okay, we spent like this, we 

504
00:29:47,962 --> 00:29:51,368
created this 
 number of 
features, etc., etc. 

505
00:29:51,368 --> 00:29:54,960
But this should not be the key 
thing, you know? 

506
00:29:54,980 --> 00:29:59,864
The key thing should be like 
what impact 
 we achieved and 

507
00:29:59,864 --> 00:30:03,926
did this impact was actually the
thing we wanted to achieve, 
 

508
00:30:03,934 --> 00:30:06,690
you know, with what we did in 
our project. 

509
00:30:08,133 --> 00:30:10,945
Yeah. 
So for any of leaders who are 
 

510
00:30:10,953 --> 00:30:13,287
listening currently, right? 
Always remember this Goodhart's 

511
00:30:13,287 --> 00:30:15,453
Law. 
Because once you set a certain 


512
00:30:15,461 --> 00:30:19,313
kind of like metrics, right, 
people will just follow. 

513
00:30:19,463 --> 00:30:22,613
I mean, there's a-- this phrase,

 right, "People behave as to 

514
00:30:22,613 --> 00:30:26,263
how they are measured," right? 
So definitely especially metrics

515
00:30:26,263 --> 00:30:29,603

 that can be easily gamified. 
I think in the industry, classic

516
00:30:29,603 --> 00:30:31,423
examples 
 will be like code 
coverage, right? 

517
00:30:31,803 --> 00:30:35,273
Where you can exactly create 
 
100% code coverage but without 

518
00:30:35,293 --> 00:30:38,841
testing anything, right? 
Same thing with, I don't know, 

519
00:30:38,841 --> 00:30:41,159
like 
 with DORA metrics, you 
know, agile, velocity and things

520
00:30:41,159 --> 00:30:44,057
like that, right? 
So always remember the picking 


521
00:30:44,065 --> 00:30:47,289
the counter-metrics, right? 
The kind of things that make 

522
00:30:47,289 --> 00:30:50,313
sure 
 the, there's a check and 
balance between one metric and 

523
00:30:50,313 --> 00:30:52,458
the other. 
So I think definitely very 
 

524
00:30:52,466 --> 00:30:55,553
classic Goodhart's Law. 
I always remember this whenever 

525
00:30:55,553 --> 00:30:59,739
I 
 become a manager, right, so 
that I don't pick the wrong 

526
00:30:59,739 --> 00:31:02,646
metrics to choose. 
So I think looking at the list, 

527
00:31:02,646 --> 00:31:06,815
I 
 would like to cover the 
Hofstadter’s Law next, because I

528
00:31:06,815 --> 00:31:12,043
find it quite 
 interesting to 
discuss with AI. so yeah, maybe 

529
00:31:12,043 --> 00:31:13,443
tell us 
 about Hofstadter’s 
law. 

530
00:31:14,328 --> 00:31:18,868
Yes, Hofstadter's Law was like 

always with us, especially when 

531
00:31:18,878 --> 00:31:23,462
Scrum started, because it 
directly 
 impacted the whole 

532
00:31:23,462 --> 00:31:28,078
thing regarding planning, 
estimations, story points. 

533
00:31:29,158 --> 00:31:32,848
Because it says like it always 
takes 
 longer than you expect, 

534
00:31:32,848 --> 00:31:35,178
even when you account for the 
Hofstadter's Law. 

535
00:31:36,948 --> 00:31:42,244
And why is this? 
Because we-- the software is 
 a

536
00:31:42,244 --> 00:31:47,577
complex thing, you know? 
It's, it is in the area of 
 

537
00:31:47,585 --> 00:31:52,478
very complex things to build. 
And we cannot, you know, know 
 

538
00:31:52,486 --> 00:31:57,933
everything that will happen. 
And it also goes in line with 
 

539
00:31:57,941 --> 00:32:01,648
this rule called the 90-90 rule.
Why? 

540
00:32:01,678 --> 00:32:06,218
Because, you know, let's talk 
about it now. 

541
00:32:06,218 --> 00:32:10,978
You have AI, it like can 
 
generate your 90% maybe of code.

542
00:32:10,978 --> 00:32:16,194
That's the easy part. 
But then that 10% that you 

543
00:32:16,194 --> 00:32:21,286
haven't 
 thought about maybe in
details are like integration, 

544
00:32:21,286 --> 00:32:25,784
edge cases, 
 testing, maybe 
writing docs and similar stuff, 

545
00:32:25,784 --> 00:32:30,348
which could actually 
 took the 
same amount of effort. 

546
00:32:30,388 --> 00:32:33,315
So if you planned like three 
 
days, it could be easily six 

547
00:32:33,315 --> 00:32:36,319
days because you thought, okay, 
the 
 biggest thing is creating 

548
00:32:36,319 --> 00:32:39,159
this. 
I will like, I need like two, 
 

549
00:32:39,167 --> 00:32:42,338
two and a half days, and then 
this half day for the rest. 

550
00:32:42,378 --> 00:32:47,058
But that rest could be the same 
size 
 and same amount as this 

551
00:32:47,058 --> 00:32:50,812
first part. 
And this is exactly this 90-90 

552
00:32:50,812 --> 00:32:53,458
rule, 
 which nicely connect to 
Hofstadter's Law. 

553
00:32:54,258 --> 00:32:59,163
So we are not good at 
estimations 
 and we need to 

554
00:32:59,163 --> 00:33:03,380
take this into account, 
especially when estimating 
 the

555
00:33:03,380 --> 00:33:07,704
complexity of software. 
And of course, there are some 
 

556
00:33:07,712 --> 00:33:12,028
countermeasures that we can do 
here, especially this to try to 

557
00:33:12,028 --> 00:33:18,152
have this in 
 mind when we plan
and to have all of these buffers

558
00:33:18,152 --> 00:33:21,920
and things that could happen and

 that maybe happened in the fu-

559
00:33:21,920 --> 00:33:23,968
in the, past, that could happen 
in the future. 

560
00:33:24,708 --> 00:33:29,112
But we need to have this in mind
when 
 we plan things or when we

561
00:33:29,112 --> 00:33:33,000
say that something will, you 
know, last for 
 that number of 

562
00:33:33,000 --> 00:33:35,821
time and et cetera. 
Yeah. 

563
00:33:36,551 --> 00:33:38,981
So yeah, that I think this 
 is 
also again classic, right? 

564
00:33:38,981 --> 00:33:42,393
So I think it's quite funny. 
It always takes longer than you 

565
00:33:42,393 --> 00:33:44,351
 expect, right, than what you 
estimate. 

566
00:33:44,371 --> 00:33:47,025
It's quite typical, right? 
And you mentioned like people, 

567
00:33:47,025 --> 00:33:50,150
like 
 when you mentioned 
software is a complex thing, but

568
00:33:50,150 --> 00:33:52,891
many people actually 
 assume 
software is easy, right? 

569
00:33:52,891 --> 00:33:55,381
Especially the so called 
 
non-technical people, right? 

570
00:33:55,701 --> 00:33:59,251
And now these days, especially 
when 
 people think about AI, 

571
00:33:59,251 --> 00:34:02,771
it's, they think it's even 
easier for, you 
 know, writing 

572
00:34:02,771 --> 00:34:06,007
or producing software. 
Some people even think like you 

573
00:34:06,007 --> 00:34:08,690
 can just easily, you know, 
reduce number of people, but you

574
00:34:08,690 --> 00:34:10,291
can still 
 produce a lot of 
things, right? 

575
00:34:10,371 --> 00:34:14,101
So tell us, is it still kind of 
like 
 applicable about this law

576
00:34:14,101 --> 00:34:16,623
with AI, right? 
Because now it seems like once 


577
00:34:16,630 --> 00:34:20,275
you can churn out so many agents
working at the same time, maybe 

578
00:34:20,275 --> 00:34:25,130
 you can have things faster. 
You can take f-- you can be 

579
00:34:25,130 --> 00:34:29,317
faster, 
 but I think it's even 
worth it in pause because you 

580
00:34:29,317 --> 00:34:33,831
now cannot know, 
 for example, 
especially with people. 

581
00:34:34,002 --> 00:34:38,420
You talk about vibe coders, they

 cannot anticipate, you know, 

582
00:34:38,420 --> 00:34:42,750
what will happen after they, you
know, vibe code 
 something or 

583
00:34:42,750 --> 00:34:46,270
when they create something. 
How much time the other steps 

584
00:34:46,270 --> 00:34:49,940
will take. 
Because when you hit the wall 

585
00:34:49,940 --> 00:34:54,018
where 
 the, like, maybe AI 
cannot longer help, maybe you 

586
00:34:54,018 --> 00:34:56,902
can need to contact 
 some 
people, need to do something 

587
00:34:57,022 --> 00:35:04,557
manually, then these things will

 hit hard, you know, hard, hard

588
00:35:04,557 --> 00:35:08,377
back. 
So always, for this law in 
 

589
00:35:08,385 --> 00:35:13,527
particular, I always say, like 
we should always have these 
 

590
00:35:13,535 --> 00:35:15,792
contingency buffers in our 
schedules. 

591
00:35:17,242 --> 00:35:19,672
And don't be overly 
 optimistic
in planning. 

592
00:35:19,742 --> 00:35:23,424
So we need to acknowledge that 
delays 
 happens and the factors

593
00:35:23,424 --> 00:35:26,272
that goes into timelines and our
expectations. 

594
00:35:27,302 --> 00:35:30,641
I find out that, you know, 
 
developers are always 

595
00:35:30,641 --> 00:35:32,872
optimistics. 
It always be fast. 

596
00:35:32,912 --> 00:35:37,472
Now with AI even faster. 
But the things will have their 

597
00:35:37,472 --> 00:35:42,244
stops 
 in that timeline, and 
then we need to try to 

598
00:35:42,244 --> 00:35:45,392
anticipate those things 
 and, 
you know, plan accordingly. 

599
00:35:46,655 --> 00:35:49,839
Yeah, developers are always 
optimistic 
 and also try to 

600
00:35:49,839 --> 00:35:52,525
delay everything almost to the 
deadline, right? 

601
00:35:52,525 --> 00:35:54,635
This is talking about 
 maybe 
another law, right? 

602
00:35:55,445 --> 00:35:57,505
When, w- I forgot what's the 
law. 

603
00:35:57,515 --> 00:36:00,229
I think it's like when you have 
a 
 deadline, right, everything 

604
00:36:00,229 --> 00:36:03,015
will just finish closer to the 
deadline. 

605
00:36:03,015 --> 00:36:06,184
I think you have it as well in 
one of, It's ninety-ninety rule,

606
00:36:06,184 --> 00:36:07,435
yeah. 
Yeah. 

607
00:36:07,715 --> 00:36:11,501
So I think definitely this one, 
 you know, when you make 

608
00:36:11,501 --> 00:36:13,715
estimates, right, always have 
buffer, right? 

609
00:36:13,715 --> 00:36:17,888
So things will somehow go wrong,

 especially I think, very 

610
00:36:17,888 --> 00:36:21,440
importantly, right, AI currently
just focus a lot more 
 on the 

611
00:36:21,440 --> 00:36:24,565
producing the code part, right? 
But like you mentioned, right, 

612
00:36:24,565 --> 00:36:27,165
there are 
 other necessary 
things that happen when you want

613
00:36:27,165 --> 00:36:30,265
to, you know, produce software 

that works for, you know, 

614
00:36:30,265 --> 00:36:32,165
external users, right, or 
working in production. 

615
00:36:32,165 --> 00:36:35,463
So it could be integration 
testing, 
 building, deployment,

616
00:36:35,463 --> 00:36:37,065
security stuff, and all that, 
right? 

617
00:36:37,485 --> 00:36:39,999
Always remember there are so 
many 
 aspects that you need to 

618
00:36:39,999 --> 00:36:42,939
think about. 
So maybe if we pick the last 

619
00:36:42,939 --> 00:36:46,330
law, 
 what will be the last law
that you think will be great for

620
00:36:46,330 --> 00:36:49,144
us to cover? 
Well, let's, talk about 
 

621
00:36:49,152 --> 00:36:52,396
Dunning-Kruger effect. 
It is like from the biases 
 or 

622
00:36:52,396 --> 00:36:54,900
behavioral section. 
It's not like about the 
 

623
00:36:54,908 --> 00:36:57,740
project times and estimations. 
Maybe it would be interesting 
 

624
00:36:57,748 --> 00:37:01,990
for people to know about it. 
It says like basically, that 

625
00:37:01,990 --> 00:37:05,523
people with 
 low ability have 
tasks overestimated, while 

626
00:37:05,523 --> 00:37:07,300
experts underestimate 
themselves. 

627
00:37:08,060 --> 00:37:13,400
And when we bound this to the 
thing we 
 talked about, AI code

628
00:37:13,400 --> 00:37:17,076
generation, people now ship like
convincing looking code 
 they 

629
00:37:17,076 --> 00:37:20,556
can't evaluate because the tool 
hides the gap in their 

630
00:37:20,556 --> 00:37:24,828
competence, you know? 
What we can do for sure is 
 

631
00:37:24,836 --> 00:37:28,529
that we produce something. 
So we produce some code, but we 

632
00:37:28,529 --> 00:37:33,344
are 
 not sure is that code good
and how this thing will, you 

633
00:37:33,344 --> 00:37:37,940
know, work when it 
 hits real 
users or real infrastructure and

634
00:37:37,940 --> 00:37:43,174
the real world, you know. 
So we need, always to take this 

635
00:37:43,174 --> 00:37:47,564
into 
 account and to be 
especially when we are like, you

636
00:37:47,564 --> 00:37:50,540
know, we don't have a 
 lot of 
competence about something. 

637
00:37:51,290 --> 00:37:56,714
Don't be ignorant about some 
signals 
 you are getting, you 

638
00:37:56,714 --> 00:38:00,510
know, because our awareness grow
faster than our skill. 

639
00:38:00,520 --> 00:38:04,650
So learning initially reduce 
 
confidence only to rebuild it. 

640
00:38:04,660 --> 00:38:10,427
So we need to, when you saw 
someone who is the 
 expert in 

641
00:38:10,427 --> 00:38:15,948
some field, you will see that 
someone is always talking about,

642
00:38:15,948 --> 00:38:18,360

 you know, trade-offs, 
probabilities. 

643
00:38:18,400 --> 00:38:21,560
Like it depends. 
You heard about this probably. 

644
00:38:22,020 --> 00:38:26,772
Not confident about everything. 
So, you know, and I think we 
 

645
00:38:26,780 --> 00:38:33,503
have the same thing with AI. 
I see a lot of people who say, I

646
00:38:33,503 --> 00:38:38,750
would 
 say pro-- in a proper 
way, we still don't know what is

647
00:38:38,750 --> 00:38:41,440
the best practice 
 now, how to 
work with this thing. 

648
00:38:41,740 --> 00:38:45,504
But on the other side, I see a 
lot 
 of people who are like, 

649
00:38:45,504 --> 00:38:49,152
"Okay, like I have, you know, I 
just started 
 to build this 

650
00:38:49,152 --> 00:38:52,254
thing and I'm the world expert 
in that, and I know 
 everything

651
00:38:52,254 --> 00:38:56,282
about that", you know. 
And this is probably not true, 

652
00:38:56,282 --> 00:39:01,044
you know, 
 because we are still
new in this, and we need some 

653
00:39:01,044 --> 00:39:05,738
time and experience and, you 
 
know, production scars to 

654
00:39:05,738 --> 00:39:09,070
understand these patterns, what 
works, what don't work. 

655
00:39:09,400 --> 00:39:15,390
And then we can confidently say,

 like, confidently, what works 

656
00:39:15,390 --> 00:39:17,010
properly and what don't, you 
know. 

657
00:39:19,273 --> 00:39:21,097
Yeah. 
When you mentioned about that, I

658
00:39:21,097 --> 00:39:23,935
remember 
 a few examples like 
on social media, right, where 

659
00:39:23,935 --> 00:39:27,393
people talk about using 
 AI, 
like especially if the tagline, 

660
00:39:27,423 --> 00:39:31,497
right, see, it's simple, right? 
I can use AI and it just 
 runs 

661
00:39:31,497 --> 00:39:33,588
by itself, right? 
Even they are selling course 
 

662
00:39:33,596 --> 00:39:36,943
for people to make money just by
using AI running autonomously. 

663
00:39:37,563 --> 00:39:39,803
I think I remember about this 
 
example for some reasons. 

664
00:39:40,113 --> 00:39:42,213
Yeah, this is probably 
 quite 
applicable, right? 

665
00:39:42,213 --> 00:39:45,781
So people who probably less 
expert in 
 a certain area, they

666
00:39:45,781 --> 00:39:48,883
think to, like they overestimate
what they can do. 

667
00:39:49,253 --> 00:39:52,142
AI definitely amplifies this a 
lot, right, 
 because people 

668
00:39:52,142 --> 00:39:55,603
assume things will be easy. 
But yeah, speaking about experts

669
00:39:55,603 --> 00:39:57,623

 sometimes, yeah, they also 
kind of like underestimate, 

670
00:39:57,623 --> 00:40:02,023
simply they want to know 
 more 
details and get more confidence 

671
00:40:02,023 --> 00:40:04,203
before, you know, deciding on 
something. 

672
00:40:04,253 --> 00:40:06,685
So I think definitely this is 
kind of 
 like a bias, cognitive

673
00:40:06,685 --> 00:40:09,713
bias thing as a probably that 
happens in our mind, right? 

674
00:40:10,463 --> 00:40:14,135
So speaking about, you know, 
like maybe 
 the software 

675
00:40:14,135 --> 00:40:17,493
engineers who have been doing 
this for a long time, right? 

676
00:40:17,523 --> 00:40:19,343
And AI is kind of like a new 
thing. 

677
00:40:20,203 --> 00:40:22,673
What do you think? 
What will be your advice for 
 

678
00:40:22,681 --> 00:40:25,278
so called these experts, right? 
Like people who are more 

679
00:40:25,278 --> 00:40:27,268
competent about 
 software 
engineering now that facing the 

680
00:40:27,268 --> 00:40:29,513
challenge of, you know, learning
AI. 

681
00:40:29,553 --> 00:40:32,089
Obviously, people will think, 
"Okay, I 
 don't know anything 

682
00:40:32,089 --> 00:40:34,635
about AI." 
Is there any good tips that you 

683
00:40:34,635 --> 00:40:37,173
wanna give with 
 the angle of 
this Dunning-Kruger effect? 

684
00:40:38,082 --> 00:40:42,392
Yeah, I think I see the 
 other 
part of this story. 

685
00:40:42,442 --> 00:40:46,342
It's imposter syndrome. 
I would say a lot of people who 

686
00:40:46,342 --> 00:40:50,340
learned, 
 as I learned and 
probably a lot of people in some

687
00:40:50,340 --> 00:40:53,899
past life, I would say, how to 

do this, we now feel like 

688
00:40:53,899 --> 00:40:56,165
imposters. 
Like, we are like, this is all 

689
00:40:56,165 --> 00:40:59,470
new. 
We don't know how to work, and 

690
00:40:59,470 --> 00:41:04,572
we 
 maybe feel that we don't 
provide value now like we 

691
00:41:04,572 --> 00:41:09,689
provided before. 
But I think, if we learn these 

692
00:41:09,689 --> 00:41:14,194
things 
 properly, like these 
thing we mentioned, the best 

693
00:41:14,194 --> 00:41:18,154
practices, how to work with 
that, 
 and taking into account 

694
00:41:18,154 --> 00:41:22,970
some things that are here to be 
for long time, that, like 
 

695
00:41:22,978 --> 00:41:27,317
Lindy effect talk about this, 
that will be for the next 50 

696
00:41:27,317 --> 00:41:30,750
years probably with us, 
 like 
they were like 50 years behind 

697
00:41:30,750 --> 00:41:34,977
us. 
Your knowledge and value will be

698
00:41:34,977 --> 00:41:39,432
even more 
 important to 
companies than it was before 

699
00:41:39,452 --> 00:41:43,472
because that knowledge don't 
disappear. 

700
00:41:43,812 --> 00:41:49,268
You know, it just changed, and I

 would say even amplified with 

701
00:41:49,268 --> 00:41:52,298
AI. 
Because now we have the same 
 

702
00:41:52,306 --> 00:41:56,802
thing that we had, but now we 
can even produce more, you know.

703
00:41:57,852 --> 00:42:02,782
We can produce more things in 
that same 
 domain, but we still

704
00:42:02,782 --> 00:42:08,402
keep the roots, like on the same
way like we did before. 

705
00:42:09,376 --> 00:42:11,693
Yeah, so for people who don't 
know 
 about Lindy effect, 

706
00:42:11,693 --> 00:42:14,530
right, the, it says like the 
longer something has 
 been in 

707
00:42:14,530 --> 00:42:17,221
use in the maybe software 
industry, right, the more likely

708
00:42:17,221 --> 00:42:19,386
it 
 is continued to be being 
used, right? 

709
00:42:20,046 --> 00:42:22,873
So we're talking about, you 
know, 
 maybe fundamentals, 

710
00:42:22,873 --> 00:42:25,338
technologies where that has been
around for 
 many, many, many, 

711
00:42:25,338 --> 00:42:27,572
years, right? 
I think it will still continue 


712
00:42:27,580 --> 00:42:30,222
to be applicable, right? 
So I think for those who have 

713
00:42:30,222 --> 00:42:32,372
been 
 around in the industry, 
don't get disheartened, you 

714
00:42:32,372 --> 00:42:35,186
know, whenever all 
 this new AI
craze happening, right? 

715
00:42:35,186 --> 00:42:38,546
So, the one, one for- one thing 
 for sure is like we have to 

716
00:42:38,546 --> 00:42:41,823
pick it up and learn, but also 
don't 
 forget all these 

717
00:42:41,823 --> 00:42:45,282
fundamentals, including the laws
that Milan, you 
 know, wrote in

718
00:42:45,282 --> 00:42:48,638
his books, right? 
So, so far all the five laws 

719
00:42:48,638 --> 00:42:51,857
that 
 we have covered seems 
like it is still applicable in 

720
00:42:51,857 --> 00:42:54,146
the AI world. 
How about the others? 

721
00:42:54,226 --> 00:42:56,636
Are there laws that become less 
relevant? 

722
00:42:56,736 --> 00:42:58,826
Is there such thing that you 
have found? 

723
00:43:00,543 --> 00:43:04,269
I wouldn't say. 
So, when I wrote the book, like 

724
00:43:04,269 --> 00:43:09,963
I 
 wrote it du-during the AI 
era, and I tried to reflect in 

725
00:43:09,963 --> 00:43:14,177
almost all of 
 them how they, 
you know, tackle it. 

726
00:43:14,217 --> 00:43:18,485
Of course, there are some laws 
like 
 that are not directly 

727
00:43:18,485 --> 00:43:23,188
connected to producing software,
like these 
 behavioral things 

728
00:43:23,188 --> 00:43:27,743
and some others which are, you 
know, not directly. 

729
00:43:27,763 --> 00:43:31,223
So they are not directly 
connected to 
 AI, but I would 

730
00:43:31,223 --> 00:43:37,233
say AI is present in a lot of 
them and even amplify some 
 of 

731
00:43:37,233 --> 00:43:42,132
them with this because I'm 
afraid that even people who knew

732
00:43:42,132 --> 00:43:45,633
some of 
 this will forget them 
because of AI. 

733
00:43:46,003 --> 00:43:48,389
Everyone thinks like, "Okay, 
this 
 agent will create 

734
00:43:48,389 --> 00:43:51,238
everything for me. 
I don't need to care about 

735
00:43:51,238 --> 00:43:53,831
anything 
 anymore." 
I think they're wrong, and this 

736
00:43:53,831 --> 00:43:56,293
will bite back later if you skip
it. 

737
00:43:58,022 --> 00:44:00,582
Yeah, so I think, yeah, just 
 
what you mentioned, right? 

738
00:44:00,582 --> 00:44:03,382
I think people seem to be 
 in 
this AI mode now, right? 

739
00:44:03,382 --> 00:44:06,552
We seem to forget such things 
 
exist, such laws exist, right? 

740
00:44:06,852 --> 00:44:10,304
And we also probably think this 
is a new 
 disruptor that is 

741
00:44:10,304 --> 00:44:13,550
totally new, that will 
definitely change things for 

742
00:44:13,550 --> 00:44:16,202
good, right? 
That's like different, 
 totally

743
00:44:16,202 --> 00:44:19,004
different altogether. 
But I think just like what we 
 

744
00:44:19,012 --> 00:44:21,562
covered, right, all these laws 
are probably timely, timeless, 

745
00:44:21,562 --> 00:44:23,867
right? 
it's like applicable over 
 a 

746
00:44:23,867 --> 00:44:27,084
long period of time. 
And yeah, maybe looking forward 

747
00:44:27,084 --> 00:44:31,471
 to see new laws being created 
when we talk also about AI, 

748
00:44:31,471 --> 00:44:34,477
right? 
So you mentioned about 
 Lindy 

749
00:44:34,477 --> 00:44:37,008
effect, right? 
So I think this is quite 

750
00:44:37,008 --> 00:44:39,222
important 
 for any software 
engineers because people now 

751
00:44:39,222 --> 00:44:42,172
these days, some of us 
 are 
actually quite disheartened when

752
00:44:42,172 --> 00:44:43,772
we talk about software 
engineering. 

753
00:44:43,972 --> 00:44:46,482
We think the profession will be 
gone. 

754
00:44:46,572 --> 00:44:48,972
You know, like we don't need 
 
so many software engineers. 

755
00:44:49,572 --> 00:44:52,273
So what do you think? 
Is this gonna be true, or is 
 

756
00:44:52,281 --> 00:44:56,552
this gonna be like a fad, right?
So, any thoughts about this? 

757
00:44:57,389 --> 00:45:02,259
I think software engineers will 
be even 
 more valuable than 

758
00:45:02,259 --> 00:45:05,863
before, but people are mostly 
talking about counts, you 
 

759
00:45:05,871 --> 00:45:08,719
know, how mu- how many people we
need. 

760
00:45:09,189 --> 00:45:12,174
And this thing will change. 
For example, you mentioned maybe

761
00:45:12,174 --> 00:45:15,335
on this 
 project, we will not 
need 10 people, maybe we'll 

762
00:45:15,335 --> 00:45:20,601
have, you know, three people. 
But the things, the other things

763
00:45:20,601 --> 00:45:25,735

 will be even more valuable 
because, you know, engineering 

764
00:45:25,735 --> 00:45:28,869
world is not 
 only, you know, 
code production. 

765
00:45:30,819 --> 00:45:34,289
I remember seeing the best 
software 
 engineers I met 

766
00:45:34,289 --> 00:45:37,349
writing one line of code per 
year, you know. 

767
00:45:37,369 --> 00:45:44,793
So their main, you know, thing 
they 
 did for our company was 

768
00:45:44,793 --> 00:45:47,599
not writing code, they are 
solving problems. 

769
00:45:47,659 --> 00:45:51,421
So I basically believe that 
software 
 engineers are problem

770
00:45:51,421 --> 00:45:53,569
solvers. 
Writing code is, of course, 
 

771
00:45:53,577 --> 00:45:56,651
our main thing how we solve 
problems, but not the only 

772
00:45:56,651 --> 00:45:58,177
thing. 
If you can solve problems 

773
00:45:58,177 --> 00:45:59,659
without 
 coding, that's the 
best thing. 

774
00:45:59,659 --> 00:46:03,416
So, you know, no code, no 
problem, 
 because any amount of

775
00:46:03,416 --> 00:46:07,114
code you have, and now with AI 
we will have, it's 
 expensive 

776
00:46:07,114 --> 00:46:11,259
to maintain, to run it. 
You know, you need to pay 
 for 

777
00:46:11,259 --> 00:46:15,639
that in production. 
So it should be always like, 
 

778
00:46:15,647 --> 00:46:20,682
you know, try to be simpler. 
And I think the things will 

779
00:46:20,682 --> 00:46:24,239
restructure. 
Teams will restructure them and 

780
00:46:24,239 --> 00:46:27,707
the 
 companies will become 
maybe different, maybe more 

781
00:46:27,707 --> 00:46:33,485
agile in this regard. 
And, but the judgment code is 

782
00:46:33,485 --> 00:46:36,903
still here. 
As I said, we produce lot of 
 

783
00:46:36,911 --> 00:46:44,131
more code, but I didn't saw the 
new great software that was 

784
00:46:44,131 --> 00:46:47,189
created in the last few years. 
I don't-- cannot remember 

785
00:46:47,189 --> 00:46:50,439
anything new 
 and big we are 
now using, you know. 

786
00:46:51,746 --> 00:46:55,646
Although maybe some of these AI 
frontier 
 models will say 

787
00:46:55,646 --> 00:46:57,668
differently, right? 
Because they will say, okay, we 

788
00:46:57,668 --> 00:47:00,396
have 
 come up with this new 
model that can solve, you know, 

789
00:47:00,396 --> 00:47:03,586
I don't know, even 
 a certain, 
kind of like role, right? 

790
00:47:03,626 --> 00:47:06,326
Maybe accountant, finance, 
 
software engineer, right? 

791
00:47:06,326 --> 00:47:09,110
So I think there's always this 
debate 
 in the industry, I 

792
00:47:09,110 --> 00:47:11,080
always find. 
But I think you, you mentioned 


793
00:47:11,088 --> 00:47:14,151
something that is also, I would 
love to mention it one more 

794
00:47:14,151 --> 00:47:16,516
time, right? 
Lesser code is actually 
 much 

795
00:47:16,516 --> 00:47:19,744
better, right? 
And these days it's very 
 easy 

796
00:47:19,744 --> 00:47:23,404
to produce code. 
And it's also very easy to lose 

797
00:47:23,404 --> 00:47:27,004
 track because it takes a lot of
effort to actually even to 
 

798
00:47:27,012 --> 00:47:30,521
review AI-produced code, right? 
And maybe we can talk about, you

799
00:47:30,521 --> 00:47:33,148
know, 
 building the evals, 
building all these guardrails 

800
00:47:33,148 --> 00:47:36,575
and all that, but I still 
 
think it's probably not humanly 

801
00:47:36,575 --> 00:47:39,231
possible if you continue 
churning code using AI, 
 

802
00:47:39,239 --> 00:47:42,240
especially multiple AI agents at
the same time, right, to 

803
00:47:42,240 --> 00:47:44,496
actually understand 
 what's 
going on with the code, right? 

804
00:47:44,836 --> 00:47:48,836
So I think engineers definitely 
 need to watch out about this. 

805
00:47:49,486 --> 00:47:52,196
How about the juniors, right? 
Juniors who just started. 

806
00:47:52,676 --> 00:47:56,646
I heard so many times that they 
 found it even difficult to get 

807
00:47:56,666 --> 00:48:00,868
the job in the first place. 
Any advice that you know from 
 

808
00:48:00,876 --> 00:48:04,055
your parts of the world there? 
Yeah, missing. 

809
00:48:04,395 --> 00:48:10,290
I mean, we, like, complain a 
lot, but 
 for juniors, it's 

810
00:48:10,290 --> 00:48:14,255
even worse because they cannot 
find jo- new jobs so easily. 

811
00:48:14,765 --> 00:48:20,440
On the other side, what I heard 
also 
 from companies, this 

812
00:48:20,440 --> 00:48:23,192
Dunning-Kruger effect is like 
hitting hard because 
 they 

813
00:48:23,192 --> 00:48:26,485
start to vibe code and they 
think like, "Okay, I knew 

814
00:48:26,485 --> 00:48:29,460
everything now," 
 in a very 
short time, which is not true, 

815
00:48:29,460 --> 00:48:34,295
and then hit the wall later. 
So for them it's even harder 
 

816
00:48:34,303 --> 00:48:38,677
than what it was before. 
But, you know, I think 
 there 

817
00:48:38,677 --> 00:48:42,296
is no magic recipe. 
They need to, you know, 
 to 

818
00:48:42,296 --> 00:48:45,179
learn things properly. 
They need to learn these 

819
00:48:45,179 --> 00:48:47,949
fundamentals, 
 which we 
mentioned with the Lindy effect,

820
00:48:47,949 --> 00:48:51,262
like algorithms, system design, 
 databases, because this 

821
00:48:51,262 --> 00:48:55,095
knowledge will be needed even 
with AI here. 

822
00:48:55,365 --> 00:49:01,171
Then of course AI technologies. 
And then, what I found, find, 

823
00:49:01,171 --> 00:49:07,300
much more 
 important than di- 
just this coding story is like, 

824
00:49:07,300 --> 00:49:11,475
which is what not often 
 
mentioned is like domain 

825
00:49:11,475 --> 00:49:13,923
knowledge. 
I think domain knowledge is 

826
00:49:13,923 --> 00:49:17,424
something 
 much, much more 
important than just knowing how 

827
00:49:17,424 --> 00:49:21,152
to code something. 
Because this is something that 


828
00:49:21,160 --> 00:49:25,508
is hard to gain, you know, and 
expensive to lose people. 

829
00:49:25,578 --> 00:49:28,298
And bus factor is talking 
 
about this, you know. 

830
00:49:28,968 --> 00:49:32,748
Do you have people who have 
 
enough knowledge and how much? 

831
00:49:33,198 --> 00:49:39,248
So we need to think more 
 up in
this direction. 

832
00:49:39,248 --> 00:49:46,252
So these people who want to 
learn and want to work, 
 and 

833
00:49:46,252 --> 00:49:50,787
who has like good skills. 
And when we talk about skills, 

834
00:49:50,787 --> 00:49:54,401
there 
 are also people skills 
which are very much important, 

835
00:49:54,401 --> 00:49:56,578
maybe even more 
 important than
technical skills. 

836
00:49:56,578 --> 00:50:01,894
Then of course technical skills.
To focus on this ownership and 

837
00:50:01,894 --> 00:50:06,508
domain 
 knowledge learning, 
because this part will be needed

838
00:50:06,508 --> 00:50:10,175
for many years to come. 
Yeah. 

839
00:50:10,525 --> 00:50:13,535
I always find it quite difficult

 for them, right, empathize 

840
00:50:13,535 --> 00:50:15,615
them, this kind of situation, 
right? 

841
00:50:15,615 --> 00:50:18,652
Because definitely when you went

 out of school, right, you 

842
00:50:18,652 --> 00:50:20,465
don't have much experience, 
right? 

843
00:50:20,705 --> 00:50:22,885
Probably what you learn 
 also 
quite limited, right? 

844
00:50:23,175 --> 00:50:26,025
My other advice would be just 
 
getting more hands-on, right? 

845
00:50:26,045 --> 00:50:29,445
Build more things, right? 
Maybe, you know, learn about 
 

846
00:50:29,453 --> 00:50:32,265
some technologies and also learn
about fundamentals, right? 

847
00:50:32,265 --> 00:50:33,715
You mentioned about 
fundamentals. 

848
00:50:34,325 --> 00:50:37,373
And I like, like in your recent 
post as 
 well, maybe in your 

849
00:50:37,373 --> 00:50:40,350
newsletter, in your Substack, 
you cover about this must-read 


850
00:50:40,358 --> 00:50:42,245
books for any software 
engineers. 

851
00:50:42,855 --> 00:50:44,605
Maybe talking about also classic
books. 

852
00:50:44,635 --> 00:50:48,888
Any kind of books that would 
want 
 to advise us all to read,

853
00:50:48,888 --> 00:50:50,964
right? 
Maybe if you can pick top five, 

854
00:50:50,964 --> 00:50:53,630
 what would be those books? 
Yes. 

855
00:50:54,200 --> 00:50:58,192
Well, I'm the avid reader, you 
probably 
 know, and I write a 

856
00:50:58,192 --> 00:50:59,990
lot of about books. 
Why? 

857
00:51:00,000 --> 00:51:01,880
Because I think books are 
important. 

858
00:51:01,890 --> 00:51:04,820
They're still important. 
Maybe people think they are not,

859
00:51:04,820 --> 00:51:08,986
but 
 a lot of these books I 
talk about are actually those 

860
00:51:08,986 --> 00:51:14,060
books that held the 
 Lindy 
effect, you know, law there. 

861
00:51:14,500 --> 00:51:18,332
They are like, they are writing 
 about things which are like, 

862
00:51:18,332 --> 00:51:20,810
which will held for many years 
to come. 

863
00:51:20,860 --> 00:51:24,594
And of course, when we talk 
about this, 
 like the best 

864
00:51:24,594 --> 00:51:27,366
books, I always recommend 
Pragmatic Programmer because 

865
00:51:27,366 --> 00:51:31,810
this book 
 actually talking 
about good habits we need to 

866
00:51:31,810 --> 00:51:36,858
have when working in this area. 
How to create, how to become 

867
00:51:36,858 --> 00:51:40,250
real software 
 engineer, and 
this is something, everyone 

868
00:51:40,250 --> 00:51:42,560
should read, like as the first 
book. 

869
00:51:42,930 --> 00:51:46,475
It is old one, but there is new 
 release, like 20th anniversary 

870
00:51:46,475 --> 00:51:49,596
edition. 
Then I would also recommend 

871
00:51:49,596 --> 00:51:52,785
Clean Code. 
Some people don't agree about 

872
00:51:52,785 --> 00:51:56,860
this, but 
 I think Clean Code 
still have a lot of good 

873
00:51:56,860 --> 00:52:00,490
insights on how code should look

 like, how good code should 

874
00:52:00,490 --> 00:52:04,595
look like. 
And, together with Philosophy 
 

875
00:52:04,603 --> 00:52:09,150
of Software Design by Professor 
Ousterhout, they are good pair 

876
00:52:09,150 --> 00:52:12,490
on 
 how the good code should 
look like. 

877
00:52:12,520 --> 00:52:15,688
And when, you know, when AI 
produce 
 some code, you can 

878
00:52:15,688 --> 00:52:19,080
review it, you can check like 
does this make sense or not? 

879
00:52:19,310 --> 00:52:22,240
Maybe sometimes it will not make
sense. 

880
00:52:24,270 --> 00:52:27,690
Then, of course, some of the 
books 
 regarding algorithms 

881
00:52:27,690 --> 00:52:31,795
like Grokking Algorithms, which 
are, you know, basis 
 to learn 

882
00:52:31,795 --> 00:52:36,390
about how things really works. 
And of course, one book I 
 

883
00:52:36,398 --> 00:52:39,990
wouldn't like to skip is 
Designing Data-Intensive 

884
00:52:39,990 --> 00:52:42,320
Applications. 
There is now a second edition. 

885
00:52:42,370 --> 00:52:45,450
I wrote a detailed review 
 like
last year about this. 

886
00:52:46,120 --> 00:52:53,076
And this is, important book for 
anyone 
 working, with like, in 

887
00:52:53,076 --> 00:52:57,692
distributed systems who want to 
know more about advanced 
 data 

888
00:52:57,692 --> 00:53:00,290
concepts, databases, data 
models, transactions, 

889
00:53:00,290 --> 00:53:04,004
replications, consistency. 
All of these things where AI 
 

890
00:53:04,012 --> 00:53:07,725
can help, but you still need to 
understand what's happening 

891
00:53:07,725 --> 00:53:10,911
there. 
So, and of course, as a bonus, 


892
00:53:10,919 --> 00:53:15,140
my book, I think it will go well
along these ones, yeah. 

893
00:53:15,845 --> 00:53:17,915
Yeah. 
I like the bonus part, right? 

894
00:53:17,925 --> 00:53:20,225
So definitely read books still, 
right? 

895
00:53:20,325 --> 00:53:24,543
Some people actually think these
days, 
 I mean, we all kind of 

896
00:53:24,543 --> 00:53:26,775
like getting very little time 
these days, right? 

897
00:53:26,815 --> 00:53:30,301
And, we prefer to use, I don't 
know, maybe 
 last time it was 

898
00:53:30,301 --> 00:53:32,930
like audiobook, right? 
And maybe there are some 

899
00:53:32,930 --> 00:53:34,615
products 
 that also summarize 
book. 

900
00:53:34,625 --> 00:53:37,966
And these days you have AI that 
can 
 summarize books quite 

901
00:53:37,966 --> 00:53:39,901
easily as well. 
Yes. 

902
00:53:39,921 --> 00:53:44,329
I mean, you can use that. 
And I mean, this is the reason 


903
00:53:44,337 --> 00:53:48,839
why I held my book sh- I would 
say short, because every law has

904
00:53:48,839 --> 00:53:51,471
like 
 two to three pages, four 
at most. 

905
00:53:51,851 --> 00:53:54,851
And you also have key 
 
takeaways on the start. 

906
00:53:54,881 --> 00:53:59,471
So it's easy to grasp, every 
 
of these laws, you know. 

907
00:54:00,301 --> 00:54:02,960
There's also audiobook. 
Yeah. 

908
00:54:03,320 --> 00:54:06,970
So one thing that I also learned

 about, even when you use AI 

909
00:54:06,970 --> 00:54:10,300
and book, right, for those 
people who 
 read a lot, you 

910
00:54:10,300 --> 00:54:13,119
will benefit a lot when using AI
simply because first, 
 

911
00:54:13,127 --> 00:54:14,960
obviously you need to craft the 
prompt. 

912
00:54:15,020 --> 00:54:17,791
That's the given thing. 
But also you need to read 
 what

913
00:54:17,791 --> 00:54:20,786
AI is outputting, right? 
So, and if you're an avid 

914
00:54:20,786 --> 00:54:23,046
reader, 
 definitely it gives 
you a good competitive 

915
00:54:23,046 --> 00:54:26,190
advantage, you know, when 
 you 
use AI simply because you love 

916
00:54:26,190 --> 00:54:29,454
to read and you can, you know, 
make 
 sense of stuff faster 

917
00:54:29,454 --> 00:54:31,970
rather than those people who 
don't love to read, right? 

918
00:54:32,373 --> 00:54:36,342
Yes, exactly. 
My way of working with AI, I 

919
00:54:36,342 --> 00:54:40,301
wrote, I 
 think I published 
this today, is like I mostly 

920
00:54:40,301 --> 00:54:44,135
focus on the research and 
 
planning part, because I do this

921
00:54:44,135 --> 00:54:48,063
part, especially when doing some
bigger things. 

922
00:54:49,293 --> 00:54:52,681
I tend to focus more on there, 
not on 
 the implementation 

923
00:54:52,681 --> 00:54:56,849
itself, because if we research 
thing properly and plan 
 thing 

924
00:54:56,849 --> 00:55:01,218
properly, the implementation 
will don't get us surprises, you

925
00:55:01,218 --> 00:55:05,481
know? 
And so try to focus more on this

926
00:55:05,481 --> 00:55:08,261

 thing, creating plans and 
reviewing plans, discussing this

927
00:55:08,261 --> 00:55:10,733
with AI 
 or, you know, with 
other people. 

928
00:55:11,083 --> 00:55:14,167
And then when you have concrete 
 implementation steps, AI is 

929
00:55:14,167 --> 00:55:17,093
good at that part to just spit 
out the code. 

930
00:55:18,355 --> 00:55:20,505
Yeah. 
I think this is a pro tips for 


931
00:55:20,513 --> 00:55:22,909
everyone who use AI, right? 
Always start with plan mode 

932
00:55:22,909 --> 00:55:24,937
first, right? 
Always research first rather 

933
00:55:24,937 --> 00:55:27,795
than jump 
 into the 
implementation, you know, vibe 

934
00:55:27,795 --> 00:55:29,725
code stuff or one-shot thing, 
right? 

935
00:55:29,725 --> 00:55:32,255
So I think definitely you 
 need
to think a lot more. 

936
00:55:32,535 --> 00:55:34,662
And don't forget, I think I like

 what you mentioned earlier, 

937
00:55:34,662 --> 00:55:35,955
right? 
Think about the outcome that 
 

938
00:55:35,963 --> 00:55:38,795
you wanna achieve, right? 
Not the exact output that you 
 

939
00:55:38,803 --> 00:55:41,599
wanna produce with AI, right? 
So think about the outcome, what

940
00:55:41,599 --> 00:55:45,392
you 
 wanna achieve, what you 
wanna do, and then use that, 

941
00:55:45,392 --> 00:55:48,023
with AI, right? 
So Milan, I think we have 
 

942
00:55:48,031 --> 00:55:51,109
covered a lot of things. 
Is there anything else that you 

943
00:55:51,109 --> 00:55:54,295
 think we should cover before we
wrap up to my last question? 

944
00:55:55,714 --> 00:55:58,689
Yeah. 
Maybe just briefly to mention, 


945
00:55:58,697 --> 00:56:03,424
so how I would use these, like 
laws in my daily life. 

946
00:56:04,234 --> 00:56:07,577
So try to name forces, and then 
 think like when you read the 

947
00:56:07,577 --> 00:56:10,124
book, of course, which laws 
apply here? 

948
00:56:10,234 --> 00:56:13,314
Well, there are probably 
 
usually three to five laws. 

949
00:56:13,854 --> 00:56:18,232
Then rank the constraints, 
 
which constraint is the right in

950
00:56:18,232 --> 00:56:21,706
this point of time. 
For example, two-week deadline, 

951
00:56:21,706 --> 00:56:24,694
of course, 
 beats the scaling 
worries that you have. 

952
00:56:25,274 --> 00:56:26,894
And then check also the 
opposite. 

953
00:56:27,104 --> 00:56:30,304
So for each law, you will, you 

know, find where it breaks. 

954
00:56:30,644 --> 00:56:34,989
And this is some kind of, you 
know, 
 algorithm that works 

955
00:56:34,989 --> 00:56:40,164
well when working within 
software. 

956
00:56:40,594 --> 00:56:47,103
And by applying these laws you 
will just 
 go, you know, much 

957
00:56:47,103 --> 00:56:51,455
faster from the, you know, issue
that happens to the right 
 

958
00:56:51,463 --> 00:56:53,321
answer you need for that 
situation. 

959
00:56:55,104 --> 00:56:57,195
Yeah. 
And I would like also to mention

960
00:56:57,195 --> 00:56:58,684
that 
 you also have the 
website, right? 

961
00:56:58,744 --> 00:57:00,213
Maybe a companion website 
 for 
this book, right? 

962
00:57:00,223 --> 00:57:00,923
Yes. 
Yes. 

963
00:57:01,544 --> 00:57:04,332
Where you can easily filter 
based on 
 certain category or 

964
00:57:04,332 --> 00:57:07,404
if you just wanna jump to a 
certain law straight away. 

965
00:57:07,684 --> 00:57:09,704
I think that is kind of 
 like 
also helpful, right? 

966
00:57:10,383 --> 00:57:11,174
Yes. 
So yeah. 

967
00:57:11,884 --> 00:57:14,990
I cncourage people to check out 
this book 
 because this is a 

968
00:57:14,990 --> 00:57:17,488
timely thing, coming back to the
Lindy effect, the fundamentals 


969
00:57:17,496 --> 00:57:18,964
that we talk about software 
engineering. 

970
00:57:19,714 --> 00:57:22,974
And I think many of them, if not
all, 
 are still applicable even

971
00:57:22,974 --> 00:57:25,644
though we are dealing a lot with
AI these days. 

972
00:57:26,224 --> 00:57:28,284
So Dr. Milan, thank you 
 so 
much for your time. 

973
00:57:28,564 --> 00:57:30,094
It's a great pleasure to have 
you again. 

974
00:57:30,114 --> 00:57:32,666
If you still remember last 
 
time, I have one tradition that 

975
00:57:32,666 --> 00:57:35,634
I will ask for my guests. 
I call this the three 
 

976
00:57:35,642 --> 00:57:38,534
technical leadership wisdom. 
Maybe you can share a 
 version 

977
00:57:38,534 --> 00:57:40,506
today, new, right? 
Maybe with some laws, I don't 

978
00:57:40,506 --> 00:57:42,582
know. 
Maybe if you can share, 
 that 

979
00:57:42,582 --> 00:57:46,861
would be great. 
Yes, three leadership wisdoms. 

980
00:57:47,141 --> 00:57:51,246
I will just try to summarize our

 talk, you know, a bit because 

981
00:57:51,246 --> 00:57:53,781
I think we said everything 
important. 

982
00:57:54,691 --> 00:57:57,051
So the first thing is name 
 the
force before you fix it. 

983
00:57:57,231 --> 00:58:00,681
So most technical problems 
 are
actually organizational. 

984
00:58:00,771 --> 00:58:06,181
So the team that can name what's

 pulling it, solve that faster.

985
00:58:06,251 --> 00:58:10,691
So this is the spine of my book.
The second thing is like 

986
00:58:10,691 --> 00:58:12,771
optimize 
 for judgment, not 
output, you know. 

987
00:58:13,241 --> 00:58:16,623
Especially now the cheap part is

 producing code, but the 

988
00:58:16,623 --> 00:58:20,759
variable part is knowing what's 
good, and 
 this will be 

989
00:58:20,759 --> 00:58:24,465
amplified in the future. 
And the third thing, I think 
 

990
00:58:24,473 --> 00:58:27,081
it's important to mention, is 
the write thing down. 

991
00:58:27,641 --> 00:58:32,411
So the knowledge your senior 
 
has is your most fragile assets,

992
00:58:32,411 --> 00:58:37,055
and bus factor talk about this. 
So the whole book exists because

993
00:58:37,055 --> 00:58:40,821
of 
 writing beat this problem 
of constantly rediscovering 

994
00:58:40,821 --> 00:58:46,921
things by new people that 
 are 
here for, you know, for so long.

995
00:58:46,971 --> 00:58:50,031
And it will-- there will 
 be 
for many times to come. 

996
00:58:51,712 --> 00:58:52,452
Yeah. 
Wow. 

997
00:58:52,452 --> 00:58:55,493
I think it's very good. 
I find this kind of like, 
 like

998
00:58:55,493 --> 00:58:58,047
a law also, right? 
It is kind of a bit applicable, 

999
00:58:58,047 --> 00:59:00,287
right? 
I like that optimize judgment 
 

1000
00:59:00,295 --> 00:59:02,272
rather than optimize output, 
right? 

1001
00:59:02,282 --> 00:59:05,648
So I think definitely it's a 
good advice 
 for us, especially

1002
00:59:05,648 --> 00:59:09,362
in this AI era. 
Milan, thank you so much 
 for 

1003
00:59:09,362 --> 00:59:11,430
the conversation. 
Again, if people wanna check out

1004
00:59:11,430 --> 00:59:14,606
more 
 about, you know, your 
book, your stuff, maybe if you 

1005
00:59:14,606 --> 00:59:17,222
can advise some place 
 they can
reach out to you online. 

1006
00:59:18,387 --> 00:59:21,997
Yes. 
They can check the website of 

1007
00:59:21,997 --> 00:59:24,267
the book 
 called the 
lawsofsoftwareengineering.com. 

1008
00:59:25,277 --> 00:59:29,881
And you can-- they can filter 
and 
 check the all the laws, 

1009
00:59:29,881 --> 00:59:32,727
with filters, grouping, and even
by seniority. 

1010
00:59:33,437 --> 00:59:37,457
They can also check the book 
 
there and they can contact me on

1011
00:59:37,457 --> 00:59:41,017
LinkedIn, Twitter, or my 
personal 
 website and to stay 

1012
00:59:41,017 --> 00:59:43,891
in touch. 
I would always-- I'm always open

1013
00:59:43,891 --> 00:59:47,810
to 
 hear what people say, to 
hear for feedback, and anything 

1014
00:59:47,810 --> 00:59:49,668
interesting. 
Yeah. 

1015
00:59:50,028 --> 00:59:53,401
And not to mention you also have

 this popular newsletter, 

1016
00:59:53,401 --> 00:59:55,418
right? 
I think Tech World with, 
 you 

1017
00:59:55,418 --> 00:59:57,354
know, Milan, right? 
So I think please do check it 

1018
00:59:57,354 --> 00:59:59,438
out, right? 
It is quite-- It has been around

1019
00:59:59,438 --> 01:00:02,100

 for quite some time, right? 
So I think it's quite popular 
 

1020
01:00:02,108 --> 01:00:05,950
for people to also follow. 
So again… Yes, tomorrow will be 

1021
01:00:05,950 --> 01:00:08,629
interesting. 
Tomorrow will be interesting 
 

1022
01:00:08,637 --> 01:00:12,781
thing in my newsletter with one 
of Google principal engineers 

1023
01:00:12,781 --> 01:00:15,115
and 
 my recommendation is to 
read it. 

1024
01:00:15,635 --> 01:00:18,652
It is about the future of 
software 
 engineering and it 

1025
01:00:18,652 --> 01:00:21,245
tackles a lot of things we 
talked about today. 

1026
01:00:22,280 --> 01:00:25,000
Nice. 
Yeah, so again, thank you so 

1027
01:00:25,000 --> 01:00:26,310
much. 
So thanks for writing these 

1028
01:00:26,310 --> 01:00:28,226
laws. 
I think, again, it is a timely 

1029
01:00:28,226 --> 01:00:30,350
reminder 
 for all of us in the 
industry, right? 

1030
01:00:30,380 --> 01:00:33,710
About some things that probably 
 don't change often, right? 

1031
01:00:33,750 --> 01:00:37,584
And they are kind of like the 
most, 
 like I would say, one of

1032
01:00:37,584 --> 01:00:40,564
the most important things that 
people should know 
 in the 

1033
01:00:40,564 --> 01:00:43,134
software engineering industry. 
So thanks for writing and 
 

1034
01:00:43,142 --> 01:00:46,984
summarizing these laws. 
Thank you, Henry, for the 

1035
01:00:46,984 --> 01:00:49,359
support and 
 invitation to talk
about this today. 

1036
01:00:49,359 --> 01:00:54,647
I hope this talk and my book 
will be 
 valuable for people in

1037
01:00:54,647 --> 01:00:55,459
our industry.
