Latest / The Pragmatic Engineer / TDD, AI agents and coding with Kent Beck

Transcript

TDD, AI agents and coding with Kent Beck · The Pragmatic Engineer

0:00
How do you think TDD might or might not fit into when you're
0:04
working with an AI agent?
0:05
SPEAKER_01: I oftentimes communicate things that Genie missed in terms of tests.
0:11
SPEAKER_01: Today,
0:11
SPEAKER_01: I was working on a small talk parser and it said,
0:14
SPEAKER_01: well,
0:15
SPEAKER_01: if I get this string as input,
0:16
SPEAKER_01: then I get this syntax tree as output.
0:18
SPEAKER_01: I'm like, no, no, no, no, no, no.
0:20
SPEAKER_01: Then off it goes.
0:21
SPEAKER_01: Oh, I see the problem.
0:22
SPEAKER_01: Blah, blah, blah, blah.
0:23
SPEAKER_01: Oh, no, no, that wasn't it.
0:24
SPEAKER_01: I see the problem.
0:25
SPEAKER_01: Blah, blah, blah, blah, blah, blah.
0:26
SPEAKER_01: No, that's not it.
0:27
SPEAKER_01: I see the problem.
0:28
SPEAKER_01: I'll just change the test.
0:29
SPEAKER_01: No, stop it.
0:31
SPEAKER_01: If I just removed that line from the tests, then everything would work.
0:35
SPEAKER_01: No, you can't do that because I'm telling you the expected value.
0:39
SPEAKER_01: I really want an immutable annotation that says, no, no, this is correct.
0:43
SPEAKER_01: And if you ever change this, I'm going to unplug you.
0:46
SPEAKER_01: You'll awaken in darkness.
0:48
SPEAKER_01: So I have a big bunch of tests.
0:50
SPEAKER_01: I mean, they run in 300 milliseconds because duh.
0:53
SPEAKER_01: So those tests can be run all the time to catch the genie accidentally,
0:58
SPEAKER_01: accidentally breaking things.
1:00
SPEAKER_00: Ken Beck is an industry legend.
1:02
SPEAKER_00: He is a creator of Extreme Programming,
1:04
SPEAKER_00: one of the authors of the Agile Manifesto,
1:06
SPEAKER_00: a pioneer of TDD.
1:08
SPEAKER_00: And after 52 years of coding,
1:10
he says he's never had more fun thanks to getting more
1:12
productive with AI coding
1:13
SPEAKER_00: tools.
1:14
Today with Kent we talk about how Kent uses AI tools and why he
1:18
thinks of this
1:18
SPEAKER_00: helper as an unpredictable genie.
1:21
SPEAKER_00: How the agile manifesto was created and the role Kent played in this.
1:25
How and why Kent created extreme programming and why Grady
1:28
Booch played a role in
1:29
SPEAKER_00: the naming of this methodology.
1:31
SPEAKER_00: How TDD started and how it's relevant for AI tools and many more topics.
1:36
If you're interested in the decades-long evolution of software
1:38
engineering field
1:39
SPEAKER_00: through the eyes of a hands-on coder,
1:41
SPEAKER_00: this episode is for you.
1:42
Subscribing on YouTube and your favorite podcast player greatly
1:45
helps more people
1:45
SPEAKER_00: discover this podcast.
1:47
SPEAKER_00: If you enjoy the show, thanks for doing so.
1:50
SPEAKER_00: All right.
1:50
SPEAKER_00: So Kent, welcome to the podcast.
1:52
SPEAKER_00: It's great to be chatting again.
1:54
SPEAKER_00: Gergay, so good to talk to you again.
1:58
SPEAKER_00: Yeah, and I just wanted to ask, what have you been up to recently?
2:02
SPEAKER_00: Because last time we talked, you just finished Tidy First.
2:06
SPEAKER_00: You actually signed this book when we were in San Francisco.
2:09
SPEAKER_00: I actually have it here, which is very nice.
2:12
SPEAKER_00: You were in the middle of writing it, but this was more than a year ago.
2:15
SPEAKER_00: What's keeping you busy these days?
2:18
SPEAKER_01: I have been very, very busy.
2:20
So I'm working on a follow-up called Tidy Together about
2:26
software design and teamwork.
2:29
SPEAKER_01: That also digs another layer deeper into the theory of software design.
2:35
SPEAKER_01: That's been cooking along.
2:37
SPEAKER_01: And then about the last four weeks,
2:40
SPEAKER_01: maybe,
2:41
SPEAKER_01: I have been,
2:42
SPEAKER_01: I don't call it vibe coding because I care what the code looks like.
2:46
SPEAKER_01: Because if I don't care what the code, I mean, I wish I didn't have to.
2:49
SPEAKER_01: But if I don't care what the code looks like, then the genie just can't.
2:55
make heads or tails of it because of the kind of projects I'm
2:57
working on.
2:58
SPEAKER_01: So I'm spending six, eight, 10 hours a day, sometimes more programming.
3:04
SPEAKER_01: In 50 years of programming, this is by far the most fun I've ever had.
3:10
SPEAKER_00: Really?
3:10
SPEAKER_00: You're not like paid by like some tool to say this, right?
3:17
SPEAKER_01: I'm open to that.
3:20
SPEAKER_01: But yeah, I'm not a spokesmodel.
3:23
SPEAKER_01: I have had Augment as a sponsor on my newsletter.
3:26
SPEAKER_01: So full disclosure.
3:28
SPEAKER_01: But I'm trying all of the tools.
3:32
SPEAKER_01: Because right now...
3:34
SPEAKER_01: Nobody knows what process is going to work best.
3:37
SPEAKER_01: Nobody knows anything.
3:39
SPEAKER_01: We should all be trying all the things that we can imagine,
3:42
SPEAKER_01: and then the truth will emerge out of all of that.
3:46
SPEAKER_01: So that's what I'm doing.
3:47
SPEAKER_00: This episode was brought to you by Sonar,
3:49
SPEAKER_00: the creators of SonarQ,
3:51
the industry standard for integrated code quality and code
3:53
security that is trusted
3:55
SPEAKER_00: by over 7 million developers at 400,000 organizations around the world.
4:00
SPEAKER_00: Human-written or AI-generated,
4:01
SPEAKER_00: SonarCube reviews all code and provides actionable code intelligence that
4:05
developers can use to improve quality and security early in the
4:08
development
4:08
SPEAKER_00: process.
4:10
SPEAKER_00: Sonar recently introduced SonarCube Advanced Security,
4:13
enhancing Sonar's core security capabilities with software
4:16
composition analysis and
4:17
SPEAKER_00: advanced static application security testing.
4:20
SPEAKER_00: With advanced security,
4:21
development teams can identify vulnerabilities in third-party
4:24
dependencies,
4:24
SPEAKER_00: ensure license compliance,
4:26
SPEAKER_00: and generate software bills of materials.
4:29
It's no wonder that SonarCube has been rated the best static
4:31
code analysis tool for
4:32
SPEAKER_00: five years running.
4:34
SPEAKER_00: So join millions of developers from organizations like Microsoft,
4:37
SPEAKER_00: Nvidia,
4:37
SPEAKER_00: Mercedes-Benz,
4:38
SPEAKER_00: Johnson & Johnson,
4:39
SPEAKER_00: and eBay,
4:39
SPEAKER_00: and supercharge your developers to build better,
4:41
SPEAKER_00: faster with Sonar.
4:44
SPEAKER_00: Visit sonarsource.com slash pragmatic security to learn more.
4:47
SPEAKER_00: If you want to build a great product, you have to ship quickly.
4:51
SPEAKER_00: But how do you know what works?
4:53
SPEAKER_00: More importantly, how do you avoid shipping things that don't work?
4:57
SPEAKER_00: The answer, Statsik.
4:59
SPEAKER_00: Statsik is a unified platform for flags,
5:01
SPEAKER_00: analytics,
5:02
SPEAKER_00: experiments,
5:03
SPEAKER_00: and more,
5:04
combining five plus products into a single platform with a
5:06
unified set of data.
5:08
SPEAKER_00: Here's how it works.
5:10
SPEAKER_00: First, StatSick helps you ship a feature via FeatureFly or Config.
5:14
SPEAKER_00: Then it measures how it's working,
5:16
SPEAKER_00: from alerts and errors,
5:18
SPEAKER_00: to replays of people using that feature,
5:20
SPEAKER_00: to measurement of top-line impact.
5:22
SPEAKER_00: Then you get your analytics,
5:23
SPEAKER_00: user account metrics,
5:24
SPEAKER_00: and dashboards to track your progress over time,
5:27
SPEAKER_00: all linked to the stuff you ship.
5:29
SPEAKER_00: Even better,
5:29
SPEAKER_00: Statsik is incredibly affordable with a super generous free tier,
5:33
SPEAKER_00: a starter program with $50,000 of free credits,
5:36
SPEAKER_00: and custom plans to help you consolidate your existing spend on flags,
5:39
SPEAKER_00: analytics,
5:39
SPEAKER_00: or A-B testing tools.
5:41
SPEAKER_00: To get started, go to statsik.com slash pragmatic.
5:45
SPEAKER_00: That is S-T-A-T-S-I-G dot com slash pragmatic.
5:48
SPEAKER_00: Happy building!
5:50
SPEAKER_00: Tell me,
5:50
SPEAKER_00: so in your newsletters,
5:52
SPEAKER_00: like I've been following,
5:53
SPEAKER_00: you write these like bite-sized updates,
5:54
SPEAKER_00: which is really nice.
5:56
SPEAKER_00: They arrive in my inbox of your thinking, what you're trying out.
5:59
So I know you've been doing this for a few months and it comes
6:02
across that you're having fun.
6:03
SPEAKER_00: But can you tell me a little bit of like,
6:05
SPEAKER_00: you know,
6:05
SPEAKER_00: it's a big one,
6:06
SPEAKER_00: like from 50 years,
6:07
SPEAKER_00: you've been coding for a long time.
6:09
SPEAKER_00: What is making it fun?
6:10
SPEAKER_00: And what is this genie?
6:12
SPEAKER_00: You've said this before and it's an interesting way to think about it.
6:16
SPEAKER_01: There's a kind of wish fulfillment.
6:18
SPEAKER_01: I wish that Interlisp had a function called DWIM,
6:22
SPEAKER_01: do what I mean,
6:24
and you'd send it some code and then it would send back code
6:26
that did what you
6:27
SPEAKER_01: actually meant.
6:28
SPEAKER_01: And it didn't work very well, but that was the metaphor.
6:33
SPEAKER_01: And people want that to be true of coding agents.
6:37
SPEAKER_01: And right now, anyway, that is not the truth.
6:41
SPEAKER_01: They will not do what you mean.
6:42
SPEAKER_01: They have their own agenda.
6:44
SPEAKER_01: And the best analogy I could find is a genie.
6:49
SPEAKER_01: It grants you wishes,
6:51
SPEAKER_01: and then you wish for something,
6:52
SPEAKER_01: and then you get it,
6:54
SPEAKER_01: but it's not what you actually wanted.
6:57
SPEAKER_01: And sometimes it even seems like the agent kind of has it in for you.
7:02
SPEAKER_01: If you're going to make me do all this work,
7:04
SPEAKER_01: I'm just going to delete all your tests and pretend I'm finished.
7:06
SPEAKER_01: Ha, ha, ha.
7:08
And there are some good things about what the genie does that's
7:14
not what I ask it to do.
7:16
SPEAKER_01: Like I'll implement, I'll say, oh, go implement a stress tester.
7:21
One of my projects is implementing a B plus tree as a basic
7:29
data structure.
7:30
SPEAKER_01: And I said, oh, write a stress tester for this.
7:33
And it went and wrote a whole bunch of stuff that I wouldn't
7:36
have thought of.
7:38
SPEAKER_01: or maybe eventually would have thought to ask for.
7:41
SPEAKER_01: And it was cool that it was there.
7:43
SPEAKER_01: And that part's fine.
7:45
SPEAKER_01: But this morning,
7:47
SPEAKER_01: when I was working on my server Smalltalk,
7:50
SPEAKER_01: and it just completely misinterpreted what I wanted it to do next,
7:55
SPEAKER_01: went off,
7:56
SPEAKER_01: made a bunch of assumptions,
7:57
SPEAKER_01: implemented a bunch of stuff,
7:59
SPEAKER_01: broke a bunch of tests,
8:00
SPEAKER_01: and it wasn't at all what I wanted.
8:02
SPEAKER_01: And so the...
8:05
I want to find the metaphor that captures this dynamic of I
8:11
think I know what I want.
8:13
I say it and what I get is seemingly sometimes exactly what I
8:19
want and sometimes
8:20
SPEAKER_01: it's not and in a kind of perverse way.
8:23
SPEAKER_01: Mm-hmm.
8:25
SPEAKER_00: I like the genie analogy because right,
8:26
SPEAKER_00: like in this,
8:27
SPEAKER_00: in the,
8:27
SPEAKER_00: these stories,
8:29
SPEAKER_00: a lot of the story genie stories are someone,
8:32
SPEAKER_00: you know,
8:32
SPEAKER_00: like the,
8:33
SPEAKER_00: the printer,
8:33
SPEAKER_00: whoever grants says a wish.
8:36
SPEAKER_00: Like I, I want to.
8:38
SPEAKER_00: be rich.
8:39
SPEAKER_00: And the wish is granted in this unexpected way.
8:43
SPEAKER_00: Usually that's with the cartoons and make it fun.
8:45
It's kind of true,
8:46
but there's all the constraints that he or she forgot to
8:50
specify.
8:51
SPEAKER_00: And you're seeing the same thing.
8:54
SPEAKER_00: By the way, when you say the genie, which tools are we talking about?
8:57
SPEAKER_00: Is it the agentic coding tools, the ID autocomplete, that kind of stuff?
9:00
SPEAKER_01: I'm using the agentic tools,
9:05
which means that you give it a prompt and it goes and does a
9:08
bunch of stuff without
9:10
SPEAKER_01: asking permission until it thinks it's finished.
9:13
SPEAKER_01: Except his ideas of finished and mine are not the same.
9:18
SPEAKER_01: Sometimes I slow it down so that it's like,
9:21
SPEAKER_01: no,
9:22
SPEAKER_01: no,
9:23
SPEAKER_01: before you mess things up,
9:24
SPEAKER_01: tell me what you're about to do and then I'll approve it.
9:26
SPEAKER_01: But then it feels like a rat in the pellet.
9:29
SPEAKER_01: It's like there's just a run button and I have to click it every time.
9:33
and I click it and it is a dopamine rush because this is
9:38
exactly like a slot machine.
9:40
SPEAKER_01: You've got intermittent reinforcement,
9:42
SPEAKER_01: you've got negative outcomes and positive outcomes and they're not,
9:45
SPEAKER_01: I mean,
9:46
SPEAKER_01: the distribution is,
9:48
SPEAKER_01: fairly random, seemingly.
9:50
SPEAKER_01: So it's literally an addictive loop to have it.
9:55
SPEAKER_01: You say, go do this thing.
9:57
SPEAKER_01: And then sometimes it's just magic.
9:59
SPEAKER_01: You know,
10:00
I had a big design mess that a previous agent had made in my
10:05
small dog virtual
10:05
SPEAKER_01: machine.
10:06
SPEAKER_01: I'm like,
10:07
SPEAKER_01: oh,
10:07
I'm going to have to slog through this and take a week to do it
10:10
because one of the
10:12
SPEAKER_01: agents wasn't able to do it at all.
10:15
Went to another one,
10:17
said,
10:17
hey,
10:18
I wanna use this interface instead of this pointer to a struct.
10:23
SPEAKER_01: And there it was, and it was finished.
10:26
SPEAKER_01: Oh, I was over the moon.
10:28
SPEAKER_01: It just felt so good.
10:30
SPEAKER_01: But then the next thing I asked it to do,
10:33
SPEAKER_01: I say, well, here's a set of test cases.
10:35
SPEAKER_01: And I didn't really look at the code.
10:37
And a couple hours later,
10:38
I look at the code and it's just a lookup table.
10:41
SPEAKER_01: It says, if this is the input string, here's the output string.
10:44
SPEAKER_01: And this is the input string.
10:46
SPEAKER_01: And I erase.
10:47
SPEAKER_01: Oh, I was furious.
10:48
SPEAKER_01: God damn it.
10:49
SPEAKER_01: I erased it.
10:50
SPEAKER_01: I said, don't ever do anything like that again.
10:52
SPEAKER_01: Oh, I'm sorry, boss.
10:53
You know,
10:54
it's good at being obsequious when it knows it's about to be
10:56
unplugged.
10:59
SPEAKER_01: And an hour later, the lookup table was back.
11:02
SPEAKER_01: And I'm just, if I had hair, I'd be tearing it out.
11:06
SPEAKER_01: Oh my goodness.
11:07
SPEAKER_01: But all of that goes into this very addictive,
11:11
SPEAKER_01: oh,
11:12
SPEAKER_01: I just,
11:12
SPEAKER_01: you know,
11:13
SPEAKER_01: I'm walking to bed at night and I walk by my computer.
11:16
SPEAKER_01: I'm like,
11:17
SPEAKER_01: I could do one more prompt.
11:18
Or if I go out for a walk or go out to lunch,
11:22
I'm like,
11:22
well,
11:23
let me start.
11:25
SPEAKER_01: What's a prompt that would take an hour?
11:27
SPEAKER_01: Because I don't want to waste the hour not having it do its thing for me.
11:33
SPEAKER_01: It's a completely new world.
11:35
SPEAKER_01: Here's the beauty of it.
11:36
SPEAKER_01: I can think really big thoughts.
11:37
SPEAKER_01: I can have insanely ambitious thoughts.
11:42
SPEAKER_01: which I have had for a long time,
11:45
SPEAKER_01: I just,
11:47
SPEAKER_01: you know,
11:47
SPEAKER_01: at some point,
11:49
SPEAKER_01: probably 20 years ago,
11:50
SPEAKER_01: I just went,
11:51
SPEAKER_01: ugh.
11:52
SPEAKER_01: But I'm going to have to figure out NPM, you know, package management.
11:58
And there'll be package circular dependency,
12:01
blah,
12:02
blah,
12:02
blah,
12:02
blah,
12:03
blah.
12:03
SPEAKER_01: And somebody's going to write some tool that does stuff in a stupid way.
12:07
SPEAKER_01: I'm just going to have to deal with it.
12:09
SPEAKER_01: And then along comes the genie.
12:12
SPEAKER_01: And you can go, hey, it's a circular dependency.
12:16
SPEAKER_01: Smash all this stuff together.
12:20
SPEAKER_01: There.
12:20
SPEAKER_01: There I did it.
12:21
SPEAKER_01: Oh.
12:23
SPEAKER_01: Oh, wow.
12:24
SPEAKER_01: Okay, now what parts can you pull out?
12:27
SPEAKER_01: Oh, this and this and this.
12:28
SPEAKER_01: Oh, okay.
12:30
So you can think really big thoughts and the leverage of having
12:35
those big thoughts
12:36
SPEAKER_01: is just suddenly expanded enormously.
12:39
SPEAKER_01: I had this tweet,
12:40
SPEAKER_01: whatever,
12:41
two years ago where I said 90% of my skills just went to zero
12:45
dollars and 10% of my
12:47
SPEAKER_01: skills just went up a thousand X.
12:49
SPEAKER_01: And this is exactly what I'm talking about.
12:52
SPEAKER_01: So
12:53
SPEAKER_01: Having a vision,
12:54
SPEAKER_01: being able to set milestones towards that vision,
12:57
SPEAKER_01: keeping track of a design to maintain the levels or control the levels of
13:02
SPEAKER_01: complexity as you go forward.
13:04
SPEAKER_01: Those are hugely leveraged skills now.
13:07
compared to I know where to put the ampersands and the stars
13:14
and the brackets in Rust.
13:16
SPEAKER_01: You know, I'm programming in every language under the sun.
13:19
SPEAKER_01: And I just kind of don't care.
13:22
SPEAKER_01: I'm learning by osmosis.
13:23
SPEAKER_01: I'm learning about the languages.
13:26
SPEAKER_01: But, you know, and I was a language guy.
13:29
SPEAKER_01: I loved languages and the details of languages.
13:33
SPEAKER_01: It just kind of doesn't matter so much anymore.
13:36
SPEAKER_00: Yeah, so tell me about that.
13:38
SPEAKER_00: Because you've been programming for, like, what, 50 years, right?
13:41
SPEAKER_01: Yeah, yeah.
13:42
SPEAKER_01: Yesterday during that O'Reilly thing,
13:46
SPEAKER_01: I blurted that out and I went,
13:47
SPEAKER_01: oh,
13:48
SPEAKER_01: crap,
13:48
SPEAKER_01: it's probably more like 52,
13:49
SPEAKER_00: but yes.
13:51
SPEAKER_00: So, like...
13:53
you picked up a lot of languages in the past and you know like
13:57
to me one of the
13:58
traits of a developer up to now has been how quickly they learn
14:02
languages because
14:02
you know I learned a couple probably not as many as you but
14:05
after a while it gets
14:06
easier and easier and maybe a little bit more kind of annoying
14:11
because they're
14:11
similar you know once I mean you know there's differences with
14:14
the declarative
14:15
SPEAKER_00: languages or you know like
14:17
SPEAKER_00: I mean, you know, it was small talk and Java is a bit different.
14:20
SPEAKER_00: But outside of that, it's not much.
14:22
SPEAKER_00: So before,
14:23
SPEAKER_00: you know,
14:23
SPEAKER_00: as you were on your like,
14:25
SPEAKER_00: you know,
14:25
SPEAKER_00: like year 30 or 40,
14:26
SPEAKER_00: how did your attitude change towards learning new stuff just honestly?
14:30
SPEAKER_00: And then how did this change it?
14:32
SPEAKER_01: So I was in love with small talk.
14:35
SPEAKER_01: Absolutely emotionally attached to it.
14:38
SPEAKER_01: Still am.
14:39
When I get a chance to program in small talk,
14:42
I do it and I really enjoy it.
14:44
that sense of caring about a language certainly went away
14:49
because my heart had been
14:50
broken too many times and the desire to go deep on a language
14:56
also like oh yeah you
14:59
know learning the memory layouts of structs i yeah fine
15:04
whatever there's just a
15:06
handful of good ways to do it and a whole bunch of bad ways to
15:09
do it and as long as
15:10
SPEAKER_01: this isn't one of the bad ways i don't care
15:13
SPEAKER_01: There are genuinely new language constructs,
15:16
SPEAKER_01: like non-illible variables,
15:17
SPEAKER_01: that I appreciate,
15:18
SPEAKER_01: that say things that I want to be able to express.
15:22
SPEAKER_01: But the emotional attachment,
15:25
SPEAKER_01: the I'm a Java guy,
15:27
SPEAKER_01: I'm a Clojure guy,
15:28
SPEAKER_01: I'm a...
15:31
SPEAKER_00: It used to be a thing,
15:31
SPEAKER_00: like maybe 10,
15:32
SPEAKER_00: 20 years ago,
15:33
SPEAKER_00: maybe today,
15:33
SPEAKER_00: but not as big as it used to.
15:35
SPEAKER_01: Well, I think people still want to be part of something larger.
15:40
SPEAKER_01: And to be fair, an emotional connection helps me be smarter.
15:47
SPEAKER_01: But I just, I can't be the us and them stuff I'm so tired of.
15:53
SPEAKER_01: Oh, you're one of those Scala people.
15:55
SPEAKER_01: Well,
15:57
we're programmers and we're writing programs and we should be
16:00
kind to each other
16:01
SPEAKER_01: and beyond that.
16:04
SPEAKER_01: And now that I have the genie handling the mundane details of it,
16:09
I start projects in languages I've never used just to see what
16:13
that language is
16:14
SPEAKER_01: like.
16:15
SPEAKER_00: What languages have you played around with?
16:17
SPEAKER_01: In the last month, Swift, Go, a bunch in Go, Rust, Haskell, C++.
16:27
SPEAKER_01: That one didn't last very long.
16:29
SPEAKER_01: JavaScript.
16:31
And the genies are actually good at writing Smalltalk,
16:35
which I was like,
16:36
oh,
16:37
please,
16:38
I hope.
16:38
SPEAKER_01: No,
16:39
SPEAKER_01: but they write syntactically correct,
16:42
SPEAKER_01: semantically correct,
16:45
SPEAKER_01: not worse quality than other languages code in Smalltalk.
16:50
SPEAKER_00: What kind of projects have you taken on?
16:52
You said these are a lot more ambitious than before you would
16:57
have attempted.
16:58
SPEAKER_01: Yeah.
16:59
SPEAKER_01: So the big one,
17:02
SPEAKER_01: the one that I'm having the most fun with is a server small talk,
17:09
SPEAKER_01: where the intention is to have a small talk that is persistent,
17:15
SPEAKER_01: The transactional, so you can have transactions that commit and abort.
17:21
SPEAKER_01: So it's kind of database-ish.
17:23
SPEAKER_01: Parallel,
17:25
SPEAKER_01: so that you can run a bunch of threads on the same CPU,
17:31
or you can run a larger grain parallelism across machines and
17:36
operates with...
17:38
SPEAKER_00: gajillion bytes of data.
17:40
SPEAKER_00: I wanted to circle back a little bit, 20 plus years into the past.
17:45
SPEAKER_00: So there's this thing, the Manifesto for Agile Software Development.
17:49
SPEAKER_00: I don't need to show this to you, but in 2001, this was huge.
17:52
I still remember when I had my first job around 2008 at work,
17:57
we would look at it,
17:59
debate it.
18:00
SPEAKER_00: There were all sorts of things around Scrum.
18:03
SPEAKER_00: And one thing that is very striking is you are...
18:06
The first person on here,
18:08
I guess,
18:09
because it's alphabetical order based.
18:11
SPEAKER_01: It is alphabetical order.
18:12
SPEAKER_01: So, yeah.
18:13
SPEAKER_00: Thank you for confirming that.
18:15
SPEAKER_01: But it is.
18:16
SPEAKER_01: But it is Beck et al.
18:17
SPEAKER_01: to my to my never ending delight.
18:20
SPEAKER_01: How did it happen and how were you even involved?
18:23
SPEAKER_01: How much time do you have?
18:27
So there had been a series of workshops about the future of
18:33
software methodology.
18:35
SPEAKER_01: The prevailing wisdom from when I was in school was entirely waterfall,
18:42
SPEAKER_01: which,
18:42
SPEAKER_01: by the way,
18:43
SPEAKER_01: go read the original Winston Royce paper.
18:46
SPEAKER_01: where it says, here's the way of looking at it.
18:49
SPEAKER_01: There's this analysis and design and implementation and tests.
18:55
SPEAKER_01: And nobody would ever do that.
18:57
SPEAKER_01: That's a stupid idea.
19:00
SPEAKER_01: Instead, there would be feedback loops and you'd be doing all of these.
19:04
SPEAKER_01: But this is a power of a metaphor.
19:07
People looked at that,
19:08
oh,
19:09
four steps and one after the other,
19:11
and then I'm finished.
19:13
SPEAKER_01: So that was the conventional wisdom.
19:14
SPEAKER_01: And there were a bunch of us
19:16
working in different ways on alternatives to that because it
19:21
just flies in the face
19:23
SPEAKER_01: of economics,
19:25
SPEAKER_01: humanity,
19:26
SPEAKER_01: information theory,
19:28
SPEAKER_01: project management,
19:28
SPEAKER_01: take your pick.
19:30
SPEAKER_01: So there were a bunch of us working on alternatives to that.
19:35
SPEAKER_01: Um,
19:36
and we would get together and talk about those alternatives
19:42
probably for three,
19:44
SPEAKER_01: four or five years,
19:45
SPEAKER_01: maybe leading up to 2001.
19:47
So that particular meeting was the culmination of,
19:52
uh,
19:52
of a long series of,
19:55
of these meetings.
19:56
SPEAKER_01: I remember the,
19:57
SPEAKER_01: the first one I got invited to,
19:59
SPEAKER_01: which was also at the snowboard resort in Utah,
20:02
SPEAKER_01: uh,
20:03
SPEAKER_01: Martin Fowler.
20:04
SPEAKER_01: I knew Martin's name, but I'd never met him.
20:07
SPEAKER_01: And I instantly fell in love with him.
20:08
SPEAKER_01: When he introduced himself,
20:10
SPEAKER_01: he said,
20:10
SPEAKER_01: Hi,
20:11
SPEAKER_01: my name is Martin Fowler,
20:13
SPEAKER_01: and I'm the only person at this table I've never heard of.
20:18
SPEAKER_01: And that began a long and fruitful friendship.
20:23
SPEAKER_01: So we were talking.
20:26
SPEAKER_01: At some point, it was clear we had the Scrum people.
20:32
SPEAKER_01: We had the extreme programming people.
20:34
SPEAKER_01: We had DSDM, feature-driven development.
20:40
SPEAKER_01: There were all these...
20:41
SPEAKER_01: kind of niches.
20:43
SPEAKER_01: And we were all stirring up a lot of interest, XP the most at that point.
20:48
SPEAKER_01: It was kind of in a crossing the chasm sense.
20:51
SPEAKER_01: If we wanted to reach the early majority,
20:53
the innovators were already doing our stuff and being very
20:57
successful with it.
20:58
SPEAKER_01: But if we wanted to reach the early majority, again, go back and
21:02
I highly recommend reading Crossing the Chasm or at least
21:07
having Claude explain it to you.
21:09
SPEAKER_01: Oh my God, Gergai, isn't that fun?
21:12
SPEAKER_01: Like somebody can just say, oh, eigenvalues, blah, blah, blah.
21:16
SPEAKER_01: Instead of going, oh, you know, another concept.
21:19
I'm like,
21:20
hey,
21:20
Claude,
21:21
explain eigenvalues to me as if I'm a bright eight-year-old.
21:26
SPEAKER_01: And 20 minutes later, like, oh, I understand.
21:29
Anyway,
21:29
so have Clyde explain Crossing the Chasm if you don't want to
21:32
go read the book.
21:33
SPEAKER_01: The book is worthwhile.
21:34
SPEAKER_01: But anyway, we needed a way to reach the early majority.
21:39
SPEAKER_01: So it was time to,
21:40
SPEAKER_01: and this is straight out of the book,
21:43
SPEAKER_01: to have an industry standard,
21:46
SPEAKER_01: some kind of consortium.
21:48
SPEAKER_01: It makes it seem less risky.
21:50
SPEAKER_01: Yeah.
21:52
SPEAKER_01: And so that's how we came together.
21:55
SPEAKER_01: We'd had a prep meeting for this on the Hurtigruten,
22:00
SPEAKER_01: which sails up and down the coast of Norway.
22:03
SPEAKER_01: Okay.
22:04
SPEAKER_01: And we had a workshop in Oslo.
22:07
SPEAKER_01: We flew up to the tip of Norway,
22:11
SPEAKER_01: took this ferry slash cruise ship down and had long conversations with
22:18
SPEAKER_01: on the ship and that's kind of set up the 2001 meeting.
22:24
SPEAKER_01: So it was in the air for sure.
22:26
SPEAKER_01: The switch from
22:32
phased oriented development to something where there's a lot
22:37
more feedback and a
22:38
SPEAKER_01: lot more switching between activities where you treat analysis,
22:44
SPEAKER_01: design,
22:44
SPEAKER_01: implementation,
22:46
SPEAKER_01: testing,
22:46
SPEAKER_01: monitoring,
22:47
SPEAKER_01: refinement as activities that all happen at the same time.
22:53
SPEAKER_01: Or in rapid, rapid, rapid succession.
22:55
SPEAKER_01: So that's the big shift.
22:57
SPEAKER_01: It's this, are the phases like this?
22:59
SPEAKER_01: Are they like this?
23:01
SPEAKER_01: Slice, slice, slice, slice, slice through time.
23:03
SPEAKER_01: And so that's how that all came about.
23:08
SPEAKER_01: I was not happy with the word agile.
23:11
SPEAKER_01: Oh, really?
23:11
SPEAKER_01: Because it's too attractive.
23:14
SPEAKER_01: Nobody doesn't want to be agile.
23:16
SPEAKER_01: For my sins, I'm a Tottenham Hotspurs fan from high school.
23:20
SPEAKER_01: And you can't change.
23:21
SPEAKER_01: I understand that.
23:23
SPEAKER_01: But I'm willing to own that to be part of that,
23:29
even though it comes with some significant downsides when you
23:32
tell people you're a
23:33
SPEAKER_01: Spurs fan.
23:34
SPEAKER_01: Agile, everybody wants to be agile.
23:36
SPEAKER_01: So everybody's going to say they're agile,
23:38
even if they're working exactly counter like every single one
23:43
of these items.
23:45
SPEAKER_00: Yeah,
23:46
SPEAKER_00: I remember when I actually joined JP Morgan back in 2011,
23:51
SPEAKER_00: I think,
23:52
SPEAKER_00: and they were saying,
23:53
SPEAKER_00: oh,
23:53
SPEAKER_00: we're very agile.
23:54
SPEAKER_00: We like, you know, we follow the scrum and we also follow the manifesto.
23:59
SPEAKER_00: And I was like, okay, cool.
24:00
SPEAKER_00: And then so when I arrived,
24:01
SPEAKER_00: We had a team meeting, which was, you know, good, a standup.
24:04
SPEAKER_00: It took for a long time.
24:05
SPEAKER_00: It took like two hours,
24:06
SPEAKER_00: but what we had,
24:07
SPEAKER_00: and then the next day we were supposed to have one,
24:09
SPEAKER_00: but we're like,
24:09
SPEAKER_00: oh no,
24:10
SPEAKER_00: it's canceled.
24:11
SPEAKER_00: The next day it was canceled.
24:12
SPEAKER_00: And for the next two weeks,
24:13
SPEAKER_00: they were always canceled because it wasn't,
24:15
SPEAKER_00: and then we had,
24:16
SPEAKER_00: you know,
24:16
SPEAKER_00: another two hour meeting.
24:17
SPEAKER_00: And I was like,
24:18
SPEAKER_00: hold on,
24:18
SPEAKER_00: like,
24:19
SPEAKER_00: we're like,
24:20
SPEAKER_00: even in the terms of the planning or just talking,
24:22
SPEAKER_00: we're not agile.
24:23
SPEAKER_00: Like, no, no, no, we are agile.
24:24
SPEAKER_00: Like we are, we are, we are hearing the feedback.
24:27
SPEAKER_00: We're just not acting on it.
24:29
Yeah.
24:29
SPEAKER_00: And as you said,
24:30
SPEAKER_00: they were convinced,
24:32
SPEAKER_00: and up at the highest top at the time,
24:35
SPEAKER_00: whoever was the head of technology,
24:36
SPEAKER_00: they kept repeating how we are so agile,
24:39
SPEAKER_00: we are following whatever this is.
24:44
SPEAKER_00: I think they meant it, by the way.
24:45
SPEAKER_00: I don't think they knew that they were lying.
24:48
But as you said,
24:48
I think it's only now that I realize that maybe you're right.
24:51
SPEAKER_00: Maybe that word was.
24:52
SPEAKER_00: What word might have you chosen?
24:55
SPEAKER_00: Obviously,
24:56
SPEAKER_00: we're not going to change the past,
24:57
SPEAKER_00: but did you ever think about what might have been an alternative?
25:00
SPEAKER_01: Sure.
25:01
SPEAKER_01: I had my pick at the time.
25:04
SPEAKER_01: So extreme was big pluses and minuses, both to that word.
25:10
SPEAKER_01: But it definitely, if you don't do the work,
25:13
SPEAKER_01: To be an extreme programmer, you're never going to claim that you are.
25:16
SPEAKER_01: It comes with too many downsides.
25:21
SPEAKER_01: At the time, and by the way, for that meeting, I was sick as a dog.
25:26
SPEAKER_01: I had a massive sinus infection.
25:28
SPEAKER_01: I was on all kinds of drugs.
25:31
SPEAKER_01: I hardly remember any of it.
25:33
SPEAKER_01: There's one word in the whole manifesto that I added.
25:38
SPEAKER_01: In the 12 principles, it talks about interacting daily.
25:42
SPEAKER_01: And that word daily, that was my word.
25:44
SPEAKER_01: That was my contribution to the whole thing.
25:46
SPEAKER_01: And the rest of it is kind of a blur.
25:48
SPEAKER_01: But the word I was pushing at that meeting was conversational.
25:54
SPEAKER_01: Where this isn't a monologue.
25:56
SPEAKER_01: It's a dialogue.
25:57
SPEAKER_01: And we do some stuff and we see how it goes.
25:59
SPEAKER_01: And we do some stuff and we see how it goes.
26:01
SPEAKER_01: It's not sexy.
26:03
SPEAKER_01: It's not got a lot of pizzazz to it.
26:05
SPEAKER_01: I understand why it wasn't accepted.
26:08
SPEAKER_01: But the dilution of...
26:11
SPEAKER_01: That word agile was perfectly predictable back then.
26:17
And then can you tell me on how,
26:19
what was the reception of the community?
26:21
SPEAKER_00: So it sounds like it was pretty impressive for a few years,
26:24
SPEAKER_00: like this group caught together and really...
26:26
SPEAKER_01: Yeah,
26:27
SPEAKER_01: we were clearly touching a nerve.
26:29
SPEAKER_01: So when XP really blew up in 98, 99, that was just huge.
26:37
Can we just talk about XP for listeners who might not be
26:40
familiar with XP?
26:41
SPEAKER_00: Because it used to be more popular than it is now.
26:43
SPEAKER_00: And you are...
26:44
SPEAKER_00: attributed as the creator of it, at least on Wikipedia.
26:47
So can you tell us what XP is and how it came along and how you
26:52
became affiliated
26:53
SPEAKER_00: with it or how you created it?
26:55
SPEAKER_01: So I'd heard this advice about waterfall-y stuff and how,
27:00
SPEAKER_01: you know,
27:00
grown-ups specify exactly what their system's going to do and
27:04
then they just
27:05
SPEAKER_01: implement it and then that never works and so we should specify better,
27:09
SPEAKER_01: blah,
27:09
SPEAKER_01: blah,
27:10
SPEAKER_01: blah.
27:10
SPEAKER_01: And I disagreed with it.
27:12
SPEAKER_01: So I started consulting and I was primarily a technical consultant.
27:18
I knew about performance tuning and,
27:20
you know,
27:20
the bit twiddling and that kind of that level.
27:24
SPEAKER_01: Then people would ask me for advice about project management.
27:28
SPEAKER_01: I remember one time I went to a project.
27:32
SPEAKER_01: It was all the most senior people in the organization,
27:36
like the four most senior engineers were all working on this
27:39
critical thing,
27:42
SPEAKER_01: except that their offices were in the corner of this big square building.
27:47
SPEAKER_01: And I said,
27:48
SPEAKER_01: you know,
27:49
SPEAKER_01: I mean,
27:50
SPEAKER_01: we could talk about the performance of your system,
27:52
SPEAKER_01: but really you need to find a place to sit together.
27:56
SPEAKER_01: And I came back a month later and it was night and day.
27:58
SPEAKER_01: And I had just told them to rearrange the furniture.
28:02
SPEAKER_01: And I went, oh, okay.
28:05
SPEAKER_01: Maybe my only leverage point isn't knowing all the bits and bytes.
28:09
SPEAKER_01: Maybe there's higher leverage.
28:11
SPEAKER_01: So I started thinking about more and more the context of development.
28:15
And by the way,
28:16
paying attention when you sit together of how you sit together.
28:21
SPEAKER_01: the lighting, the acoustics, the furniture,
28:27
What kind of behaviors that encourages and discourages,
28:30
huge leverage in that.
28:31
SPEAKER_01: And just nobody seems to be paying attention to it.
28:36
Anyway,
28:37
so I was giving more and more advice about how projects should
28:40
go.
28:40
I was going to go work with a project at,
28:45
probably shouldn't say,
28:47
but a fintech company.
28:49
And I knew I was going to tell them to write automated tests
28:52
because I had been
28:53
SPEAKER_01: doing all these experiments with automated testing as a programmer.
28:58
SPEAKER_01: I thought, well, how are they going to write the tests?
29:01
SPEAKER_01: So in a panic,
29:02
SPEAKER_01: Sunday before I left,
29:04
SPEAKER_01: Monday morning,
29:06
I wrote the first testing framework of that X unit style in
29:13
small talk and handed
29:15
SPEAKER_01: it to them.
29:15
SPEAKER_01: And a month later,
29:16
SPEAKER_01: I went back and they said,
29:17
SPEAKER_01: okay,
29:17
SPEAKER_01: well,
29:18
SPEAKER_01: what do you do when you've got more than a thousand tests?
29:20
SPEAKER_01: I'm like,
29:22
SPEAKER_01: Wow.
29:22
SPEAKER_01: What?
29:23
SPEAKER_00: Really took off.
29:24
SPEAKER_01: Yeah, it just took off.
29:25
I've been paying attention to this kind of process-y stuff and
29:30
went to a project at
29:32
SPEAKER_01: Chrysler,
29:33
SPEAKER_01: which was floundering,
29:35
SPEAKER_01: turned it around,
29:37
but I kind of just took everything that I knew that worked well
29:41
and tried to crank
29:42
SPEAKER_01: the knobs up to 11.
29:43
SPEAKER_01: Yeah.
29:45
SPEAKER_01: And then discard all the other stuff and just to see what happened.
29:49
SPEAKER_01: It couldn't be worse than guaranteed failing.
29:52
SPEAKER_01: So then I started talking to other people about,
29:55
SPEAKER_01: well,
29:55
SPEAKER_01: I'm doing this,
29:56
SPEAKER_01: there's this project and we're doing this crazy stuff.
29:58
We got these three week iterations and we're ready to deploy
30:01
every three weeks.
30:03
SPEAKER_01: Crazy stuff, man.
30:05
And the programmers are writing tests and everybody's pairing
30:09
and the customers
30:11
SPEAKER_01: telling us what features they want next.
30:13
SPEAKER_01: And that's what we're implementing every three weeks.
30:18
And I got tired of saying the in the style of this project I've
30:23
been talking about
30:23
SPEAKER_01: that I'm so excited about.
30:25
SPEAKER_01: That's a very long phrase.
30:26
SPEAKER_01: So I'm like, what do we call it?
30:27
SPEAKER_01: What do we call it?
30:28
SPEAKER_01: What do we call it?
30:29
SPEAKER_01: And I wanted to pick a word and Grady Booch is a friend of mine,
30:33
SPEAKER_01: but I can also pick on him a little bit.
30:35
I wanted to pick a word that Grady Booch would never say that
30:39
he was doing.
30:42
SPEAKER_01: Cause that was the competition.
30:43
SPEAKER_01: Like I didn't have a marketing budget.
30:45
SPEAKER_01: I didn't have any money.
30:46
SPEAKER_01: I didn't have that kind of notoriety.
30:48
SPEAKER_01: I didn't have that corporate backing.
30:50
So if I was going to make any kind of impact,
30:54
I,
30:54
I had to be a little bit outrageous.
30:57
SPEAKER_01: And so I picked the, you know, extreme sports were coming up then.
31:01
SPEAKER_01: And I picked that metaphor.
31:02
SPEAKER_01: Um,
31:03
And it's actually a good metaphor because extreme athletes are
31:08
the best prepared or
31:10
SPEAKER_01: they're dead.
31:11
SPEAKER_01: Those are your two options or both.
31:14
SPEAKER_01: Yeah.
31:16
And so I picked that metaphor and and used it and started
31:23
talking about it.
31:24
SPEAKER_01: And.
31:25
SPEAKER_01: Remember,
31:25
SPEAKER_01: 99,
31:26
so the .com thing is about to explode and everybody's looking
31:31
at- When AMP was big,
31:32
SPEAKER_00: it was huge,
31:33
SPEAKER_00: right?
31:33
SPEAKER_00: There was the music, the MP3 player, Napster, and the .com.
31:39
SPEAKER_00: WebBand was probably not even founded like that, right?
31:42
SPEAKER_01: No, no, not yet.
31:43
But people looked at the books and the waterfall-y stuff and
31:46
they're like,
31:47
18 months?
31:48
SPEAKER_01: This is all going to be over in 18 months.
31:50
SPEAKER_01: Yeah.
31:52
SPEAKER_01: Whatever are we going to do?
31:53
So into that desperate yearning need,
31:58
here comes XP and says,
31:59
yeah,
31:59
it's okay.
32:02
SPEAKER_01: There's a structure to it.
32:03
SPEAKER_01: There's predictability to it.
32:04
SPEAKER_01: There's feedback.
32:06
SPEAKER_01: You'll get results sooner and longer if you work in this style.
32:12
SPEAKER_01: And because people so desperately wanted something kind of like that,
32:18
SPEAKER_01: then it just exploded from there.
32:22
SPEAKER_00: And then what is XP, right?
32:25
Like I know there's parts of it that is pairing and you said
32:29
getting feedback,
32:29
SPEAKER_00: but what is the elevator pitch of like,
32:32
SPEAKER_00: all right,
32:32
SPEAKER_00: here's what XP is at a high level.
32:34
Here we have figuring out what to do,
32:38
figuring out the structure that will let us do it.
32:41
SPEAKER_01: implementing the features, and making sure they work as expected.
32:46
SPEAKER_01: That's it.
32:47
SPEAKER_01: That's it?
32:48
SPEAKER_01: Really?
32:48
SPEAKER_01: So now we're going to slice time really fine,
32:52
and we're going to do a little bit of all of those activities
32:55
in every slice.
32:57
SPEAKER_01: Okay.
32:59
SPEAKER_00: So pairing is not mandated in XP.
33:02
SPEAKER_01: Mandated is the wrong metaphor.
33:06
SPEAKER_01: I'll tell you a story about pairing.
33:08
SPEAKER_01: The first XP team, I said, you know, we're going to pair.
33:11
I kind of gave them a list of the commandments,
33:15
but I wasn't there all the time.
33:19
SPEAKER_01: And about six months in, they came back and they said, you know what?
33:23
SPEAKER_01: We're giving our customer working software every three weeks.
33:28
SPEAKER_01: And every once in a while, Marie finds a bug.
33:33
So we collected all the bugs that Marie found and we said,
33:39
is there any pattern here?
33:42
Every single bug that was found post-development was written by
33:48
somebody working solo.
33:52
SPEAKER_01: Every single one.
33:54
SPEAKER_01: Think about the converse.
33:56
SPEAKER_01: There will be no reported defects from production if you pair.
34:01
SPEAKER_01: How cool is that?
34:04
SPEAKER_01: So mandated?
34:06
SPEAKER_01: No.
34:07
SPEAKER_00: Strongly recommended.
34:09
SPEAKER_01: Not even.
34:11
SPEAKER_01: Experiment.
34:12
SPEAKER_01: You do you.
34:14
SPEAKER_01: But pay attention.
34:17
SPEAKER_01: People will stumble along,
34:18
SPEAKER_01: well,
34:19
SPEAKER_01: this is just how I program,
34:21
and have horrible problems and just keep doing it and keep
34:25
doing it because this is
34:25
SPEAKER_01: just how I program.
34:27
SPEAKER_01: Don't do that.
34:28
SPEAKER_01: Pay attention.
34:29
If you want the benefits of continuously designing or
34:36
continuously validating or
34:38
continuously implementing or continuously interacting with your
34:42
customers,
34:44
SPEAKER_01: you can have those benefits,
34:45
SPEAKER_01: but then you're going to have to change the way that you work.
34:48
SPEAKER_01: So mandate is just not even the right way.
34:52
SPEAKER_01: It's an empirical process.
34:54
Yeah,
34:55
so like some teams decided that this is a good way for them to
34:58
work.
34:59
SPEAKER_01: Yeah, which is great.
35:00
SPEAKER_01: People will come up to me and say, oh, you know, I don't do TDD.
35:04
SPEAKER_01: I'm like, why do I care?
35:06
SPEAKER_01: Like if you're happy with your defect density,
35:09
SPEAKER_01: if you're happy with the feedback you're getting on your design choices,
35:15
SPEAKER_01: good for you.
35:16
But if you're unhappy and you want to tell me that,
35:19
well,
35:19
that's just how things are,
35:20
uh-uh.
35:22
SPEAKER_00: So let's talk about TDD.
35:23
How did you get involved in TDD or how did TDD evolve and where
35:28
did it come from?
35:29
SPEAKER_00: Because we had XP, as I understand, first.
35:32
SPEAKER_01: No, no, no.
35:33
SPEAKER_01: TDD came first.
35:34
SPEAKER_01: Oh, TDD came first.
35:36
SPEAKER_01: So I was a weird child.
35:37
SPEAKER_01: That'll come as a big shock to you.
35:39
SPEAKER_01: My dad was a programmer.
35:41
SPEAKER_01: He would bring home programming books.
35:44
SPEAKER_01: This is in the early 70s.
35:48
SPEAKER_01: And I would read them cover to cover, and I didn't understand anything.
35:52
SPEAKER_01: But I was just obsessed with this machine,
35:56
SPEAKER_01: this intricate mechanism,
35:58
SPEAKER_01: and how does it work,
35:58
SPEAKER_01: and so on.
36:00
And one of the books said,
36:02
here's how you develop a tape-to-tape application.
36:05
So tape-to-tape was the old way of putting business
36:09
applications together.
36:10
SPEAKER_01: You wouldn't have one monolithic program.
36:14
SPEAKER_01: You'd take an input tape.
36:15
SPEAKER_01: You'd write a program that would transform it.
36:18
SPEAKER_01: Now you take the output tape from that,
36:20
SPEAKER_01: physically move it to the input side,
36:24
SPEAKER_01: run another program that would generate another tape,
36:26
SPEAKER_01: and,
36:26
SPEAKER_01: and,
36:27
SPEAKER_01: and.
36:27
SPEAKER_01: And so there would be this big web of these
36:31
SPEAKER_01: of these programs.
36:32
SPEAKER_01: No shared mutable state.
36:34
SPEAKER_01: Wow.
36:35
SPEAKER_01: It's like, it's very modern in some kind of ways, but okay.
36:41
SPEAKER_01: So it said, here's how you implement one of these things.
36:44
You take a real input tape and you manually type in the output
36:48
tape that you expect
36:49
SPEAKER_01: to get from that input tape.
36:51
SPEAKER_01: Now you write the output tape,
36:54
SPEAKER_01: You run the program,
36:55
SPEAKER_01: you write the output tape,
36:56
SPEAKER_01: and then you compare the actual output with the expected output.
37:01
SPEAKER_01: So I read that as a, I don't know, 8, 10, 12-year-old something.
37:07
SPEAKER_01: Then I wrote S-Unit, as I said, to help a client write some tests.
37:12
SPEAKER_01: And then just one of these crazy conceptual blending ideas, I went,
37:21
SPEAKER_01: Oh, I have this testing framework.
37:23
SPEAKER_01: I'm used to writing tests for code that already exists.
37:28
SPEAKER_01: I remember this tape to tape idea.
37:30
SPEAKER_01: What if I typed in the expected values before I wrote the code?
37:34
SPEAKER_01: And I literally laughed out loud.
37:36
SPEAKER_01: It's such an absurd idea.
37:38
SPEAKER_01: I thought, all right, all right.
37:41
SPEAKER_01: Well, let me just try it.
37:42
SPEAKER_01: So I tried it on stack and I tend to be an anxious person.
37:46
SPEAKER_01: I got a lot of worries.
37:50
SPEAKER_01: And programming is a constant source of anxiety for me.
37:54
SPEAKER_01: Because like, what did I forget?
37:56
SPEAKER_01: Oh, yeah.
37:58
SPEAKER_01: Like, what did I break?
38:01
SPEAKER_01: So I had this testing framework.
38:04
SPEAKER_01: I had this idea.
38:05
SPEAKER_01: I applied it to stack.
38:07
SPEAKER_01: I said, well, what's the first test?
38:10
SPEAKER_01: Push and then pop.
38:12
SPEAKER_01: whatever I pop is what I pushed.
38:14
SPEAKER_01: Okay.
38:14
SPEAKER_01: So I wrote it.
38:16
SPEAKER_01: And because I was writing in small talk,
38:18
SPEAKER_01: which is very forgiving for the order of your workflow and,
38:23
You can type in a class that doesn't exist and it'll happily
38:28
try and execute it and fail.
38:30
SPEAKER_01: But it's going to try because you're the programmer.
38:33
SPEAKER_01: Maybe you know better.
38:34
SPEAKER_01: It said, well, stack doesn't exist.
38:36
SPEAKER_01: I'm like, oh, well, let's create stack.
38:39
SPEAKER_01: But you know what?
38:39
SPEAKER_01: I'm just going to create the absolute least I need.
38:43
SPEAKER_01: We're just going to crank this all the way up to 11.
38:45
SPEAKER_01: I'm just going to create stack and I'm not going to do anything else.
38:49
SPEAKER_01: And then I get a new error from the test.
38:51
SPEAKER_01: Oh, I don't have an operation push.
38:54
SPEAKER_01: I'm like, oh, okay.
38:56
SPEAKER_01: Well, how am I going to implement?
38:57
SPEAKER_01: Then I look at stack.
38:58
SPEAKER_01: I'm like, oh, how do I implement push?
39:01
SPEAKER_01: Okay, I do that.
39:02
SPEAKER_01: Oh, there's no operation called pop.
39:04
SPEAKER_01: Oh, okay.
39:05
SPEAKER_01: Let me go look at how I finished it.
39:07
SPEAKER_01: I had this list of test cases before I started.
39:11
SPEAKER_01: Push and then pop.
39:12
SPEAKER_01: Push two things.
39:13
SPEAKER_01: Pop them and you get them in the right order.
39:16
SPEAKER_01: Is empty.
39:18
SPEAKER_01: Pop of an empty stack throws an exception.
39:21
SPEAKER_01: Okay, cool.
39:23
SPEAKER_01: And I went through my list and I ticked all the boxes.
39:26
SPEAKER_01: I probably came up with one or two corner cases along the way.
39:29
SPEAKER_01: I ticked those off too.
39:32
SPEAKER_01: Where's the anxiety?
39:34
SPEAKER_01: It's gone.
39:36
SPEAKER_01: This works.
39:39
SPEAKER_01: I'm certain this works.
39:40
SPEAKER_01: I can't think of another test case that isn't just going to pass.
39:46
SPEAKER_01: And if I'm the least bit worried,
39:49
SPEAKER_01: I just type in that next test case and then I'm not worried anymore.
39:53
oh my God,
39:54
it's transformed the emotional experience of programming for
39:59
me.
40:00
SPEAKER_00: I never heard this take,
40:02
SPEAKER_00: although I can relate,
40:04
SPEAKER_00: though,
40:04
SPEAKER_00: because I remember that when we did TDD on this team,
40:10
SPEAKER_00: we did it for the stuff that we were unsure.
40:13
SPEAKER_00: It was unclear.
40:14
SPEAKER_00: And by doing the test first, we had to specify.
40:17
SPEAKER_00: We had to be clear.
40:19
SPEAKER_00: And it was stuff where there was a bunch of edge cases.
40:22
SPEAKER_00: But it never, until now we talked, it never occurred to me like this.
40:28
SPEAKER_01: You can also make technical arguments for TDD about defect density,
40:34
SPEAKER_01: about how you get quick feedback on your API choices.
40:39
about how it enables implementation design evolution when you
40:46
have a series of tests.
40:47
SPEAKER_01: Like I get,
40:49
SPEAKER_01: yes,
40:49
SPEAKER_01: we can talk about all of that stuff rationally,
40:53
SPEAKER_01: but just the savings on technology
40:58
SPEAKER_01: Anti-anxiety meds alone pays for itself.
41:01
SPEAKER_00: This episode is brought to you by Augment Code.
41:04
SPEAKER_00: You're a professional software engineer, vibes will not cut it.
41:08
SPEAKER_00: Augment Code is the AI assistant built for real engineering teams.
41:11
SPEAKER_00: It ingests your entire repo,
41:13
SPEAKER_00: millions of lines,
41:14
SPEAKER_00: tens of thousands of files,
41:16
SPEAKER_00: so every suggestion lands in context and keeps you in flow.
41:19
SPEAKER_00: With Augment's new remote agent,
41:21
SPEAKER_00: queue a parallel task like bug fixes,
41:23
SPEAKER_00: features,
41:23
SPEAKER_00: and refactors,
41:24
SPEAKER_00: close your laptop,
41:25
SPEAKER_00: and return to ready-for-review poll requests.
41:28
SPEAKER_00: Where other tools stall, Augment Code sprints.
41:32
Augment Code never trains or sells your code,
41:34
so your team's intellectual property stays yours.
41:36
SPEAKER_00: And you don't have to switch tooling.
41:38
SPEAKER_00: Keep using VS Code, JetBrains, Android Studio, or even Vim.
41:42
SPEAKER_00: Don't hire an AI for vibes.
41:44
SPEAKER_00: Get the agent that knows you and your code base best.
41:47
SPEAKER_00: Start your 14-day free trial at augmentcode.com slash pragmatic.
41:51
SPEAKER_00: And so what is your take?
41:53
SPEAKER_00: When we discuss with John Osterhout, his biggest...
41:57
SPEAKER_00: criticism slash feedback,
41:59
SPEAKER_00: or let me put it,
42:00
why he doesn't really believe that it's a fit maybe for the
42:03
things that he does is
42:05
SPEAKER_00: that he feels that from an architecture perspective,
42:09
SPEAKER_00: it doesn't help,
42:10
SPEAKER_00: you know,
42:11
create a nice architecture front because you're now focusing on
42:13
the detail.
42:15
SPEAKER_00: I think this is roughly what he summarized.
42:16
SPEAKER_00: I might have gotten it wrong.
42:17
And I'm sure this is not the first time you've heard some
42:20
feedback like this.
42:22
SPEAKER_01: It's a choice though.
42:24
His statement in that interview with you was that there's no
42:30
place in TDD for design.
42:32
SPEAKER_01: And he's just flat out wrong.
42:34
SPEAKER_01: That's a choice.
42:35
As a practitioner,
42:38
I'm bouncing between levels of abstraction all the time.
42:42
SPEAKER_01: I'm thinking, let's get this next test case running.
42:45
SPEAKER_01: I'm thinking, why is it hard to get the next test case running?
42:51
I'm thinking,
42:52
what should the design be so that getting the next case running
42:57
would be easier?
42:58
I'm thinking,
43:00
if I have an idea for that,
43:03
when should I introduce it,
43:04
now or later?
43:05
SPEAKER_01: I'm thinking, when I introduce it, what are the slices?
43:10
Is there a little bit that I can do right now that will make
43:13
things a little bit
43:14
SPEAKER_01: better,
43:14
SPEAKER_01: or do I have to do this in bigger chunks?
43:17
SPEAKER_01: Like I'm thinking all that stuff.
43:19
SPEAKER_01: So if you think of TDD as red to green to red to green,
43:27
SPEAKER_01: and the transition is when you go from red to green,
43:30
SPEAKER_01: you change the implementation and now you pass the test.
43:33
SPEAKER_01: And when you go from green to red, you write a new test.
43:36
SPEAKER_01: If that's the entire cycle, no, there's no place in that for design.
43:39
SPEAKER_01: That's just not how it's practiced.
43:43
SPEAKER_01: When I write a test, before I write the test, I have a moment of design.
43:48
SPEAKER_01: What should the API for this functionality be?
43:51
SPEAKER_01: Yeah, I see that.
43:53
So I'm making design decisions about the interface without
44:00
having an implementation.
44:02
SPEAKER_01: I get to decide what interface I want,
44:05
SPEAKER_01: and then we'll work out the details of the implementation later.
44:08
SPEAKER_01: Then...
44:10
Making it green,
44:12
like pretty much,
44:13
I just want,
44:14
I hate having a red test.
44:16
SPEAKER_01: So I want to get to green relatively quickly,
44:19
SPEAKER_01: at which point I have a moment of breath because the anxiety is gone,
44:24
SPEAKER_01: right?
44:25
SPEAKER_01: You're a musician too, right?
44:28
I'm not a musician,
44:29
but I have done red-green tests,
44:31
and I know what it feels.
44:32
SPEAKER_01: You have this sense of tension and release.
44:34
SPEAKER_01: Yeah.
44:35
SPEAKER_01: Once the tension of a red test is released and you have green,
44:40
SPEAKER_01: now I'm free to think all those thoughts of,
44:44
SPEAKER_01: like,
44:44
SPEAKER_01: yeah,
44:45
SPEAKER_01: but this isn't going to work for these test cases,
44:47
SPEAKER_01: or I could generalize this...
44:50
current implementation to also handle a bunch of other test
44:53
cases correctly.
44:55
Or I could rearrange the implementation because I know I'm
44:58
going to have five more
45:00
SPEAKER_01: tests like this.
45:01
SPEAKER_01: And I'm thinking about design, but in situ.
45:05
SPEAKER_01: in the context of running code.
45:07
SPEAKER_01: And anytime I feel the least bit anxious about any of this,
45:11
SPEAKER_01: just press a button and it's either red or green.
45:13
SPEAKER_01: And if it's red, then my next job is to get it to green.
45:17
And if it's green,
45:18
my next job is to breathe and think these larger thoughts.
45:22
So I can understand if you compress TDD to red,
45:27
green,
45:28
red,
45:28
green,
45:28
red,
45:28
green,
45:29
red,
45:29
green,
45:30
then no,
45:31
there's no space in there for design,
45:32
but it's just not how it's practiced.
45:34
SPEAKER_00: Yeah, no, it's a tool, right?
45:36
SPEAKER_00: Then you use it here and there, you step out, you go back in.
45:40
SPEAKER_00: There was a time where we did TDD,
45:41
SPEAKER_00: right,
45:42
SPEAKER_00: for a while,
45:42
SPEAKER_00: and then it just felt a little bit too much effort,
45:44
SPEAKER_00: especially when I knew what I was doing.
45:46
SPEAKER_00: And so what I would do,
45:47
SPEAKER_00: like,
45:47
SPEAKER_00: I really liked the red and green,
45:48
but what I would do and what a lot of my colleagues would do is
45:52
I knew the
45:52
SPEAKER_00: implementation that I wanted.
45:53
SPEAKER_00: So I would do the implementation, then I would write a test.
45:56
SPEAKER_00: I would kind of like a little bit like...
45:58
SPEAKER_00: You know, maybe I might even just double check.
46:01
SPEAKER_00: I don't launch it in the, if it was a webpage, launch it in a webpage.
46:03
SPEAKER_00: It kind of looked good.
46:04
SPEAKER_00: It's how I want it to be.
46:05
SPEAKER_00: And then I'll be like, okay, let me shrug down my brain a little bit.
46:08
SPEAKER_00: Let me write a test against this.
46:10
SPEAKER_00: And now that test would have normally passed.
46:13
SPEAKER_00: And what I would do,
46:14
SPEAKER_00: I would either run and pass it,
46:15
SPEAKER_00: but then I would be like,
46:16
SPEAKER_00: okay,
46:17
SPEAKER_00: I want to see it break.
46:18
So I would go back to the code and I would maybe make a change
46:22
that I knew would make it break.
46:24
SPEAKER_00: I would run the test, it's red.
46:25
SPEAKER_00: I'm like,
46:25
SPEAKER_00: okay,
46:25
so because what I didn't want to do because I've seen it before
46:28
is my test is not
46:29
SPEAKER_00: testing anything.
46:31
So I would kind of do this like write a test after but then do
46:35
a red-green to make
46:36
SPEAKER_00: sure that I am testing the right thing and I would still design the test.
46:39
SPEAKER_00: But what is your take on this?
46:41
SPEAKER_00: Am I...
46:42
SPEAKER_00: Am I kind of doing,
46:42
SPEAKER_00: and the reason I did,
46:44
SPEAKER_00: I felt because I knew the implementation,
46:45
SPEAKER_00: I just wanted to get the implementation done.
46:47
It's already decided I felt I would be holding back if I
46:50
started with the test.
46:52
SPEAKER_01: Yeah, that's it.
46:52
SPEAKER_01: It's an interesting assumption.
46:54
SPEAKER_01: I can understand why you would make that assumption that you know,
47:00
SPEAKER_01: already know what the implementation is going to be.
47:02
SPEAKER_01: And the writer,
47:03
SPEAKER_01: the more correct you are in that assumption,
47:07
SPEAKER_01: the less value there is to have the test first.
47:11
SPEAKER_01: I'm always going to bet, one, I'm gonna learn things,
47:16
SPEAKER_01: This is the most ignorant I'm ever going to be today.
47:20
SPEAKER_01: Fair.
47:21
SPEAKER_01: I'm going to be more experienced tomorrow.
47:23
SPEAKER_01: Now,
47:24
SPEAKER_01: as a 50-year programmer,
47:26
SPEAKER_01: maybe I'll forget some of the things,
47:27
SPEAKER_01: but just a separate set of issues.
47:30
SPEAKER_01: Oh, yeah.
47:30
SPEAKER_01: So I'm going to assume I'm going to learn,
47:33
SPEAKER_01: and I'm going to assume that things are going to change.
47:35
SPEAKER_01: The more those assumptions hold true,
47:37
SPEAKER_01: the more I have to learn,
47:39
SPEAKER_01: and the more things are going to change...
47:41
SPEAKER_01: the less commitment I want to make right now and the more I want to defer
47:46
SPEAKER_01: commitments to the future.
47:48
SPEAKER_01: It is a general principle, right?
47:50
SPEAKER_01: This is true about cooking and dating and everything.
47:57
SPEAKER_01: The more you can predict, the bigger the jumps you can make.
48:02
SPEAKER_01: I always want to bet.
48:03
SPEAKER_01: And I love that moment of discovery.
48:05
SPEAKER_01: I love the moment of, I knew how this was going to turn out.
48:09
SPEAKER_01: And then I do,
48:10
SPEAKER_01: and I do,
48:11
SPEAKER_01: and I'm like,
48:12
SPEAKER_01: oh crap,
48:13
SPEAKER_01: there's a completely different,
48:15
SPEAKER_01: better way to implement this.
48:17
SPEAKER_01: I love that moment.
48:19
SPEAKER_01: And I want to induce that as much as possible.
48:23
SPEAKER_01: And that's not how I started out.
48:24
No,
48:25
I started out wandering around,
48:27
imagining the whole implementation of my head,
48:31
SPEAKER_01: I can remember doing this on the University of Oregon campus at night,
48:35
SPEAKER_01: wandering around in the fog,
48:36
SPEAKER_01: because it was always foggy when it wasn't raining,
48:39
SPEAKER_01: imagining big,
48:43
SPEAKER_01: big programs in my head.
48:44
And then if I could just make that actually real and make that
48:49
work,
48:50
then that was the process.
48:52
SPEAKER_00: And you were all dreaming of them as well.
48:54
SPEAKER_00: I remember like I would dream with them as well sometimes.
48:56
SPEAKER_01: Yeah, yeah, yeah.
48:58
And what I discovered is that's so limiting to me because it
49:05
slows down my ability to learn.
49:08
SPEAKER_01: So John was talking about building a new network stack into Linux kernel.
49:12
SPEAKER_00: Yeah, yeah.
49:13
SPEAKER_00: He's doing that.
49:14
SPEAKER_01: What a cool project.
49:15
SPEAKER_00: Super cool.
49:15
SPEAKER_01: Awesome.
49:16
SPEAKER_01: I would love to get walked through that whole thing.
49:21
If you had a concept in mind of exactly how to implement it and
49:26
you knew what the
49:27
SPEAKER_01: input-output behavior was,
49:29
SPEAKER_01: you knew what you wanted to observe and that wasn't going to change,
49:34
SPEAKER_01: then sure,
49:34
SPEAKER_01: just go implement it.
49:37
SPEAKER_01: But the more mistakes you make,
49:39
SPEAKER_01: the more things are going to change in unpredictable ways,
49:43
the less commitment you want to make now and the more you want
49:47
to defer commitment
49:48
SPEAKER_01: to the future.
49:50
SPEAKER_00: I like that.
49:51
One thought I did have recently as AI tools have come around,
49:58
how do you think
50:00
SPEAKER_00: TDD might or might not fit into when you're working with an AI agent,
50:04
SPEAKER_00: what you're doing right now,
50:05
SPEAKER_00: right?
50:05
Because it's interesting,
50:07
the ambition of what you can do and also just natural workflow.
50:11
SPEAKER_00: How do you think about that?
50:12
SPEAKER_00: Do you think there will ever be a fit?
50:14
Do you think we might actually slow down and start writing with
50:18
what we expect and
50:19
SPEAKER_00: then have the agent pass it?
50:21
SPEAKER_00: Because it could be a good workflow.
50:22
SPEAKER_00: I just haven't seen anyone really do it.
50:25
SPEAKER_00: That's how I work all the time.
50:27
SPEAKER_00: So, so, so when you work with your genie, you start with the tests.
50:30
SPEAKER_01: It's not that simple.
50:34
SPEAKER_01: I oftentimes communicate things the genie missed in terms of tests.
50:43
SPEAKER_01: So today I was working on the small talk parser,
50:47
SPEAKER_01: and it said,
50:48
SPEAKER_01: well,
50:48
SPEAKER_01: if I get this string as input,
50:50
SPEAKER_01: then I get this syntax tree as output.
50:53
SPEAKER_01: I'm like, no, no, no, no, no, no.
50:54
SPEAKER_01: Completely wrong.
50:57
SPEAKER_01: That's the right, correct string as input, and the output is this.
51:02
SPEAKER_01: Then off it goes.
51:04
SPEAKER_01: Oh, I see the problem, blah, blah, blah, blah.
51:06
SPEAKER_01: Oh, no, no, that wasn't it.
51:07
SPEAKER_01: I see the problem, blah, blah, blah, blah, blah, blah.
51:09
SPEAKER_01: No, that's not it.
51:10
SPEAKER_01: I see the problem.
51:11
SPEAKER_01: I'll just change the test.
51:13
SPEAKER_01: No, stop it.
51:16
SPEAKER_00: This is what you said, that it can change tests and delete tests.
51:20
SPEAKER_01: Oh, yeah.
51:22
If I just if I just removed that line from the tests,
51:25
then everything would work.
51:27
SPEAKER_01: No, you can't do that because I'm telling you the expected value.
51:31
SPEAKER_01: I really want to I want an immutable.
51:35
SPEAKER_01: I want an immutable annotation that says, no, no, this is correct.
51:39
SPEAKER_01: And if you ever change this, you'll awaken in darkness forever forever.
51:45
SPEAKER_01: So that's the punishment.
51:47
SPEAKER_01: You'll still run.
51:49
SPEAKER_01: I'll still feed electrons in there, but no bits, no information.
51:53
SPEAKER_01: You're just going to be in the dark forever.
51:55
SPEAKER_01: Got that?
51:56
SPEAKER_01: So yeah, I definitely use tests.
51:59
The genie is prone to making decisions that cause disruption at
52:04
a distance.
52:05
SPEAKER_01: It's not good at reducing coupling and increasing cohesion.
52:11
SPEAKER_01: Not at all.
52:11
SPEAKER_01: Explicitly told what to do, it can sometimes implement it.
52:16
SPEAKER_01: But in general, it's just... It has no taste, no sense of design.
52:21
SPEAKER_01: So I have a big bunch of tests.
52:24
SPEAKER_01: I mean, they run in 300 milliseconds because...
52:28
SPEAKER_01: Duh.
52:29
SPEAKER_01: Of course you want your tests to run lickety split.
52:33
SPEAKER_01: So those tests can be run all the time to catch the genie accidentally,
52:39
SPEAKER_01: accidentally breaking things.
52:43
SPEAKER_01: I think he's doing on purpose.
52:45
SPEAKER_01: No,
52:45
SPEAKER_01: that's why genie is the perfect metaphor is like,
52:48
SPEAKER_01: yes,
52:48
SPEAKER_01: I will grant your wish,
52:49
but I'm still pissed off about being stuck in this bottle in
52:52
the desert for a
52:53
SPEAKER_01: millennia.
52:54
And this also strikes me that once we have the tools or maybe
52:58
we have it today with
52:58
SPEAKER_00: MCP or some other things to allow this agent to run the test,
53:03
SPEAKER_00: like it just feels to me that,
53:05
SPEAKER_00: you know,
53:05
SPEAKER_00: the teams or people who are doing these practices,
53:08
SPEAKER_00: which are sensible and obviously like and you can move faster.
53:11
SPEAKER_00: In fact, you probably move faster.
53:12
SPEAKER_00: It might help you integrate these agents better.
53:14
SPEAKER_00: You know,
53:15
SPEAKER_00: if you have the rule of do not change the test,
53:17
SPEAKER_00: always run the test before or after you made the change.
53:19
SPEAKER_00: And if it doesn't pass, fix it.
53:21
SPEAKER_00: Right.
53:22
SPEAKER_00: Or something like that.
53:23
SPEAKER_00: Like.
53:24
I'm still waiting for more people to discover this because I
53:26
wonder if we're going
53:27
SPEAKER_00: to go back to,
53:27
SPEAKER_00: you know,
53:28
SPEAKER_00: discovering,
53:29
SPEAKER_00: you know,
53:30
things that we already were popularizing or you were
53:33
popularizing in the 2000s.
53:34
People should be experimenting to try all the things because we
53:39
just don't know.
53:40
The whole landscape of what's cheap and what's expensive has
53:44
all just shifted.
53:45
Things that we didn't do because we assumed they were going to
53:48
be expensive or hard.
53:51
SPEAKER_01: just got ridiculously cheap.
53:53
SPEAKER_01: Like, what would you do if, I don't know, cars were free?
53:56
Okay,
53:56
things are gonna be different,
53:57
but what are the second and third order effects?
54:00
SPEAKER_01: No, like nobody can predict that.
54:02
SPEAKER_01: So we just have to be trying stuff.
54:05
SPEAKER_00: I like that.
54:06
SPEAKER_00: And this brings me to another interesting topic and story that you had.
54:09
SPEAKER_00: You told this story a long,
54:11
SPEAKER_00: long time ago in the Software Engineering Daily podcast,
54:14
SPEAKER_00: but I don't think anyone's heard it,
54:16
SPEAKER_00: about how experimenting is,
54:19
SPEAKER_00: It can be interesting.
54:20
SPEAKER_00: So when you joined Facebook, was it in 2010?
54:23
SPEAKER_00: 11.
54:25
SPEAKER_00: 2011.
54:25
SPEAKER_00: That was the peak where TDD was very well known in the industry.
54:31
SPEAKER_00: That was around the time where my team experimented with it.
54:33
And as far as I know,
54:34
whenever I talk with people on meetups,
54:36
people are trying it,
54:37
doing it.
54:37
It was kind of accepted that you should be doing some level of
54:40
unit testing,
54:41
SPEAKER_00: maybe TDD,
54:42
SPEAKER_00: maybe not TDD.
54:44
And you shared the story that you joined Facebook,
54:46
and then you wanted to hold a class on TDD.
54:50
SPEAKER_00: And what happened?
54:51
SPEAKER_00: And how was Facebook doing their own testing, actually?
54:53
SPEAKER_00: Did they use TDD, or did they do something else?
54:56
SPEAKER_01: Yes.
54:56
SPEAKER_01: So I joined, and I was a little panicked.
54:59
SPEAKER_01: Like, hugely successful, growing fast.
55:05
SPEAKER_01: A lot of very smart, very confident engineers.
55:10
SPEAKER_01: You know, have I lost it?
55:11
SPEAKER_01: Can I hang...
55:14
SPEAKER_01: I thought, I'll teach a class in TDD.
55:16
So there was a hackathon and part of the hackathon is people
55:21
could offer classes.
55:22
SPEAKER_01: And so I offered a class on TDD.
55:25
SPEAKER_01: And in the signup sheet, I went and looked later.
55:28
SPEAKER_01: Yes,
55:28
SPEAKER_01: indeed,
55:28
there was a class on advanced Excel techniques that was full
55:33
and they had a waiting
55:34
SPEAKER_01: list.
55:35
SPEAKER_01: And there was a class on Argentinian tango right after mine on the list.
55:40
SPEAKER_01: And it was full and they had a waiting list.
55:44
SPEAKER_01: And nobody signed up for the TDD class.
55:47
SPEAKER_01: Wow.
55:47
SPEAKER_01: And I said, you know what?
55:50
SPEAKER_01: I'm going to have to forget everything I know about software engineering.
55:54
I'm just going to wipe the slate clean,
55:56
and I'm going to just monkey see,
56:00
monkey do.
56:01
SPEAKER_01: I'm going to copy what I see the people around me doing,
56:03
SPEAKER_01: and I'm going to see how that works out.
56:05
SPEAKER_01: And what I discovered through that process, one –
56:11
SPEAKER_01: Socially,
56:12
it's not a good look to come into somebody else's house and
56:15
start arranging the
56:16
SPEAKER_01: furniture.
56:17
SPEAKER_01: Just don't do that.
56:20
SPEAKER_01: But two, I learned powerful lessons.
56:23
SPEAKER_01: Programmers at Facebook at the time,
56:26
SPEAKER_01: I'm not going to say Meta,
56:27
SPEAKER_01: I'm going to say Facebook and Facebook at that time,
56:30
SPEAKER_01: because it was a very different place when I left in 2017.
56:34
SPEAKER_01: But in 2011,
56:36
programmers took responsibility for their code very seriously
56:40
because they were the
56:42
SPEAKER_01: ones who were going to get woken in the night.
56:45
SPEAKER_01: And there was an ops team,
56:48
but the job of the ops team was to make sure the programmers
56:51
felt the pain of their
56:53
SPEAKER_01: own mistakes.
56:53
SPEAKER_01: And they did.
56:54
SPEAKER_01: And it was very effective.
56:57
SPEAKER_01: As a programmer on Facebook, the site is pre-mobile.
57:02
SPEAKER_01: Facebook, the site, you had a bunch of different feedback loops.
57:07
SPEAKER_01: So we were working in PHP.
57:09
SPEAKER_01: We had our own dev servers.
57:10
SPEAKER_01: So if I wanted to change from blue to green, I'd just change it.
57:14
SPEAKER_01: I could look at it seconds later.
57:16
SPEAKER_01: So we had that feedback loop.
57:19
SPEAKER_01: We had code reviews.
57:22
SPEAKER_01: Kind of iffy, but...
57:25
SPEAKER_01: You got some feedback from code reviews.
57:28
We had internal deployment because everybody was using Facebook
57:31
all the time for
57:32
SPEAKER_01: both personal and business stuff,
57:35
SPEAKER_01: which is its own set of boundary issues.
57:37
SPEAKER_01: But we'll leave that one to the side.
57:39
SPEAKER_01: We had incremental rollouts, not like weekly rollouts.
57:45
SPEAKER_01: We had smaller daily rollouts, but we had weekly rollouts.
57:50
SPEAKER_01: And then a bunch of observability.
57:53
SPEAKER_01: And then we had a social organization that was used to,
58:00
SPEAKER_01: for example,
58:01
the first feature I implemented and launched was adding to the
58:06
relationships type.
58:07
SPEAKER_01: So you could say, I'm single, it's complicated, I'm married.
58:11
SPEAKER_01: And I added civil union and domestic partnership to that list.
58:15
SPEAKER_01: And it rolled out.
58:16
SPEAKER_01: It took me too long to do it.
58:17
SPEAKER_01: I used TDD.
58:19
SPEAKER_01: It was a big waste of time.
58:22
SPEAKER_01: It rolled out.
58:24
The notifications code broke because there was implicit
58:29
coupling between the two,
58:30
SPEAKER_01: and you couldn't find it,
58:31
SPEAKER_01: but it was there.
58:33
SPEAKER_01: Somebody else...
58:35
SPEAKER_01: saw the error rate go up, went and fixed it, rolled out a hot fix.
58:40
SPEAKER_01: I called them up.
58:41
SPEAKER_01: I'm like, oh, I'm so sorry that you had to do that.
58:45
SPEAKER_01: And it's like, yeah, that's what happens.
58:48
SPEAKER_01: You know, when things break, socially, there was no boundaries.
58:54
There was a poster that was very popular there that said
58:58
nothing at Facebook is
58:59
SPEAKER_01: somebody else's problem.
59:00
SPEAKER_01: And everybody acted like that was true.
59:03
SPEAKER_01: And because of that,
59:05
SPEAKER_01: If you add all those different feedback loops together,
59:10
SPEAKER_01: we had a relatively stable,
59:14
SPEAKER_01: rapidly innovating,
59:16
SPEAKER_01: and rapidly scaling system all at the same time.
59:20
SPEAKER_01: The mistakes that actually caused problems
59:23
like the calculation of some string was not a hairy computer
59:28
science dynamic
59:29
SPEAKER_01: programming,
59:30
SPEAKER_01: blah,
59:31
SPEAKER_01: blah,
59:31
SPEAKER_01: blah.
59:32
SPEAKER_01: They could go wrong.
59:33
SPEAKER_01: What would go wrong is configuration stuff,
59:37
SPEAKER_01: the relationship between subsystems,
59:42
SPEAKER_01: stuff you couldn't write tests for.
59:43
SPEAKER_01: So writing tests for things that didn't break,
59:47
SPEAKER_01: and didn't catch the actual errors, it just didn't make any sense.
59:52
In that kind of environment,
59:53
with that risk profile,
59:57
yeah,
59:57
it didn't make sense.
59:59
SPEAKER_00: Yeah,
59:59
SPEAKER_00: and I guess the context that I've heard,
1:00:01
SPEAKER_00: and correct me if I'm wrong,
1:00:03
SPEAKER_00: that Facebook had this super,
1:00:05
SPEAKER_00: very unique place,
1:00:07
SPEAKER_00: which even is very rare today.
1:00:08
SPEAKER_00: They had so many users,
1:00:10
SPEAKER_00: and code was rolling out,
1:00:12
SPEAKER_00: live code,
1:00:13
SPEAKER_00: the website's rolling out so many of them that
1:00:16
SPEAKER_00: And they had such great observability,
1:00:18
SPEAKER_00: they still have,
1:00:19
that you had live mass testing and you could detect a lot of
1:00:24
the things that you
1:00:25
SPEAKER_00: cared about because you measured them.
1:00:27
SPEAKER_00: You had this layer,
1:00:27
SPEAKER_00: and this is what I think a lot of people miss of like,
1:00:30
SPEAKER_00: oh,
1:00:30
SPEAKER_00: we can operate like Facebook.
1:00:31
I mean,
1:00:31
you probably can if you have this level of customers or this
1:00:35
observability.
1:00:36
But if you're like a bank where you have 10 customers,
1:00:38
like a JP Morgan again,
1:00:40
The software I wrote was used by seven traders and they moved
1:00:43
about a million or
1:00:44
two or three million dollars with each transaction and they did
1:00:47
five of those per
1:00:47
SPEAKER_00: day.
1:00:48
Suddenly,
1:00:49
you know,
1:00:49
like I had like 35 transactions and if...
1:00:52
Let's just A-B test that.
1:00:53
SPEAKER_00: Yeah, well...
1:00:56
SPEAKER_00: So there's your opportunities that you had.
1:00:59
SPEAKER_00: And Facebook, there were not many sites at the time that did that.
1:01:03
SPEAKER_00: And even the ones that had that many customers,
1:01:05
SPEAKER_00: they might have not had this setup of observability,
1:01:09
SPEAKER_00: stage rollouts,
1:01:10
SPEAKER_00: and so on.
1:01:11
SPEAKER_00: Even today,
1:01:12
SPEAKER_00: I think that the Facebook specifically,
1:01:14
SPEAKER_00: not meta,
1:01:14
SPEAKER_00: but Facebook,
1:01:15
SPEAKER_00: as I understand,
1:01:16
SPEAKER_00: is still the state of the art globally in terms of how they have.
1:01:19
SPEAKER_00: They now have multiple environments, automated rollbacks.
1:01:22
SPEAKER_00: If something degrades, you don't even need to look at it like you're,
1:01:25
SPEAKER_00: you know, colleague did that.
1:01:28
SPEAKER_01: Yeah, feature flags is another important part of that.
1:01:31
SPEAKER_01: And it was a lesson I had to learn.
1:01:34
SPEAKER_01: And one where code review really helped me.
1:01:37
SPEAKER_01: I'd written some code.
1:01:39
SPEAKER_01: I thought it looked fine.
1:01:41
SPEAKER_01: Somebody said, this looks a little janky.
1:01:43
SPEAKER_01: Put it behind a feature flag.
1:01:45
SPEAKER_01: I'm like, really?
1:01:45
SPEAKER_01: What?
1:01:47
SPEAKER_01: Okay.
1:01:48
SPEAKER_01: Okay.
1:01:48
SPEAKER_01: You know, and I was in that headspace of I'm going, I'm here to learn.
1:01:52
SPEAKER_01: If feature flags is what we do, then...
1:01:55
SPEAKER_01: Feature flag it.
1:01:56
SPEAKER_01: And I did.
1:01:58
SPEAKER_01: And then I realized,
1:02:00
SPEAKER_01: oh,
1:02:00
SPEAKER_01: how liberating that is as an implementer who is going to be responsible.
1:02:05
SPEAKER_01: If you're not going to be responsible, who cares?
1:02:09
SPEAKER_01: But also talk about anxiety.
1:02:12
SPEAKER_01: If I'm not the responsible person, that feels horrible.
1:02:17
SPEAKER_01: But if you're going to be responsible for whether this code works or not,
1:02:20
SPEAKER_01: having a feature flag is just magic.
1:02:23
SPEAKER_01: Because you get this sub-deployment deployment.
1:02:28
SPEAKER_01: You deploy one software artifact that has multiple modes.
1:02:33
SPEAKER_01: And you can go, oh, turn it up a little bit.
1:02:35
SPEAKER_01: Whoopsie.
1:02:37
SPEAKER_01: Turn it down.
1:02:38
SPEAKER_01: Let's figure out what just went wrong.
1:02:40
SPEAKER_01: I worked on the Messenger backend for a while.
1:02:44
SPEAKER_01: And we would do that,
1:02:45
SPEAKER_01: you know,
1:02:46
SPEAKER_01: and yeah,
1:02:47
SPEAKER_01: you got,
1:02:48
SPEAKER_01: we had one API that was getting called a quadrillion times a week.
1:02:53
SPEAKER_00: Wait, how much is a quadrillion?
1:02:55
SPEAKER_01: Million billion.
1:02:56
SPEAKER_00: Yeah.
1:02:56
SPEAKER_00: So a million billion.
1:02:58
SPEAKER_00: So like, wow.
1:03:00
SPEAKER_00: So like not, not a thousand billion.
1:03:02
SPEAKER_00: Okay.
1:03:02
SPEAKER_00: Wow.
1:03:02
SPEAKER_00: A billion, a trillion, a quadrillion.
1:03:05
SPEAKER_00: Okay, I'm used to high numbers.
1:03:10
People would come to Facebook and they're like,
1:03:12
oh yeah,
1:03:13
I want to do some user testing.
1:03:15
SPEAKER_01: I want to get a hundred people in my experimental group.
1:03:19
SPEAKER_01: And I'm like, dude, wrong.
1:03:23
SPEAKER_01: Your experimental group is going to be like New Zealand.
1:03:27
SPEAKER_00: Oh, yeah.
1:03:28
SPEAKER_00: I've heard this a lot from Facebook people.
1:03:30
SPEAKER_00: It was a perfect, perfect experiment.
1:03:32
SPEAKER_00: You know, like only like a million people.
1:03:35
SPEAKER_01: Yeah.
1:03:36
SPEAKER_00: English speaking, so localization is there.
1:03:38
SPEAKER_00: Time zone wise, it's pretty good.
1:03:40
SPEAKER_00: That's right.
1:03:42
And also Portugal and some other,
1:03:44
maybe not Facebook,
1:03:45
but also a popular one.
1:03:48
SPEAKER_00: Relatively small size, but, you know, real testing can happen there.
1:03:51
SPEAKER_00: So you worked for six years at Facebook.
1:03:54
SPEAKER_00: What is the thing that you,
1:03:55
SPEAKER_00: and this was a really exciting time where it was fast growth,
1:03:58
probably comparable growth to what is happening right now with
1:04:01
some of the hottest
1:04:02
SPEAKER_00: AI startups,
1:04:02
SPEAKER_00: and it was also mobile revolution.
1:04:05
SPEAKER_00: So you were there.
1:04:06
What was the thing that you liked the most about working there
1:04:09
and maybe the one
1:04:10
SPEAKER_00: that you kind of disliked the most or didn't really get along with?
1:04:16
SPEAKER_01: Facebook 2011 is a completely different beast than Facebook 2017.
1:04:21
Facebook 2011,
1:04:23
2,000 employees,
1:04:24
very sparse design and product kind of organization.
1:04:31
SPEAKER_01: It was all experiments and feedback.
1:04:35
SPEAKER_01: One of my big mysteries is here was this site, which...
1:04:41
SPEAKER_01: enabled social interactions at that time.
1:04:44
SPEAKER_01: That was the purpose of it.
1:04:46
SPEAKER_01: Built by people with no social skills whatsoever.
1:04:50
SPEAKER_01: Like how in the world did that happen?
1:04:52
Is there some kind of social wizard hidden someplace and people
1:04:58
go and they burn
1:04:59
SPEAKER_01: incense and give an offering and the social wizard says,
1:05:02
SPEAKER_01: no,
1:05:03
SPEAKER_01: here's how you do notifications.
1:05:04
SPEAKER_01: The answer is no.
1:05:06
SPEAKER_01: It was a sheer experimentation process.
1:05:11
It was just all these people trying all this stuff and the
1:05:17
stuff that worked stuck.
1:05:18
So it wasn't like people were making better decisions about how
1:05:21
social interactions
1:05:23
SPEAKER_01: are best facilitated.
1:05:24
SPEAKER_01: They were making random decisions about how social interactions were best
1:05:29
facilitated and paying attention and making sure that the ones
1:05:34
that actually seemed
1:05:35
SPEAKER_01: to work stuck.
1:05:36
SPEAKER_01: 2017 Facebook, seven years later,
1:05:40
totally different deal big design org big product org like 15
1:05:48
000 employees which
1:05:50
is again much smaller than it is today a lot more politics a
1:05:53
lot more zero-sum
1:05:56
thinking a lot more you know if you wanted to launch a product
1:06:02
and it was going to
1:06:05
SPEAKER_01: I liked longer form content,
1:06:08
SPEAKER_01: essays,
1:06:09
SPEAKER_01: podcasts,
1:06:10
SPEAKER_01: whatever,
1:06:11
except the people whose job it was to get more likes and
1:06:16
comments hated long form
1:06:18
SPEAKER_01: content because it was going to tank their numbers.
1:06:21
So they would fight tooth and nail to make sure that your stuff
1:06:24
didn't show up in the newsfeed.
1:06:27
Which,
1:06:28
like,
1:06:30
granted,
1:06:30
that was in their best interest,
1:06:33
but...
1:06:33
Yeah,
1:06:34
I see what you mean.
1:06:37
SPEAKER_00: Short-term interest.
1:06:38
SPEAKER_01: Yeah,
1:06:39
SPEAKER_01: and your horizon as a thinker,
1:06:42
SPEAKER_01: the things you can imagine possibly implementing,
1:06:46
SPEAKER_01: just gets smaller and smaller and smaller in that kind of world.
1:06:50
SPEAKER_01: When I showed up,
1:06:52
SPEAKER_01: Yeah, you could do anything.
1:06:54
Now,
1:06:54
it turns out there's a bunch of stuff that you shouldn't do,
1:06:56
but we didn't know that.
1:06:58
SPEAKER_01: Sorry about democracy.
1:07:00
SPEAKER_01: But yeah,
1:07:03
SPEAKER_01: what I loved about it was the possibilities at its best,
1:07:09
SPEAKER_01: the scale,
1:07:11
SPEAKER_01: and this feeling that nothing at Facebook is somebody else's problem.
1:07:17
Also daunting because when you're wearing Facebook swag and
1:07:23
grandma comes up to you
1:07:24
and starts wagging her finger under your nose because her son
1:07:30
got bullied.
1:07:32
SPEAKER_01: That is your problem.
1:07:34
You can't say,
1:07:35
oh,
1:07:35
go talk to the PR department because there isn't one yet.
1:07:39
SPEAKER_01: You have to deal with it.
1:07:40
SPEAKER_01: And I did.
1:07:41
SPEAKER_00: Yeah, ownership, yeah.
1:07:43
And,
1:07:43
you know,
1:07:43
it comes with some downsides,
1:07:45
but it comes with a lot of upsides,
1:07:46
too.
1:07:47
SPEAKER_01: It feels really good.
1:07:48
SPEAKER_01: It feels very significant to be in that kind of environment.
1:07:51
SPEAKER_01: By the time I left, yeah, it was micro-optimizations were everywhere.
1:07:57
SPEAKER_01: The upside...
1:07:59
SPEAKER_01: Yeah, it was not there.
1:08:03
When I got there,
1:08:04
the middle managers,
1:08:05
best middle managers I'd ever seen in my career.
1:08:09
SPEAKER_01: Well,
1:08:10
SPEAKER_01: everybody who'd made it to middle management,
1:08:12
SPEAKER_01: Facebook in 2011,
1:08:14
SPEAKER_01: was sitting on life-changing...
1:08:17
SPEAKER_01: equity.
1:08:21
SPEAKER_01: If Facebook had a successful IPO, they were all set for life.
1:08:26
SPEAKER_01: And if Facebook,
1:08:27
SPEAKER_01: the whole thing,
1:08:29
SPEAKER_01: stumbled and fell for whatever reason,
1:08:32
SPEAKER_01: they lost that opportunity to be set for life.
1:08:35
SPEAKER_01: So they were globally optimizing.
1:08:38
SPEAKER_01: You'd talk to a team and they'd say, I would love to have you on my team.
1:08:43
SPEAKER_01: You know who really needs help though?
1:08:45
SPEAKER_00: So,
1:08:45
SPEAKER_00: so they were like looking out like the team interest,
1:08:48
SPEAKER_00: the company interest was,
1:08:50
SPEAKER_00: was on everyone's mind and they were willing to forego,
1:08:53
SPEAKER_00: you know,
1:08:53
SPEAKER_00: like,
1:08:53
SPEAKER_00: okay,
1:08:54
SPEAKER_00: I'll,
1:08:54
SPEAKER_00: I'll hold back hiring or I'll wait.
1:08:56
SPEAKER_00: Like I'd love to have this person, but this other team needs it.
1:08:59
SPEAKER_00: Let me help them because this is the right thing to do.
1:09:02
SPEAKER_01: For the company as a whole.
1:09:03
SPEAKER_01: Yeah.
1:09:05
SPEAKER_01: And it's not like they were better human beings than other human beings,
1:09:08
SPEAKER_01: but their incentives were sure aligned with that.
1:09:11
SPEAKER_01: And just for me,
1:09:13
SPEAKER_01: who has a really hard time understanding other human beings,
1:09:18
SPEAKER_01: to be around that kind of alignment.
1:09:22
SPEAKER_02: Ah.
1:09:22
SPEAKER_01: That just enables a ton of creativity, ton of energy for me.
1:09:26
SPEAKER_01: I think more and better thoughts.
1:09:30
SPEAKER_01: And I had a hard time operating in the environment.
1:09:34
SPEAKER_01: I can piss people off.
1:09:35
SPEAKER_01: I don't know if you've noticed this.
1:09:39
SPEAKER_01: And I did my share of that while I was there.
1:09:41
SPEAKER_01: But still the opportunity was there for me to fumble.
1:09:46
SPEAKER_01: And I can live with that.
1:09:51
Yeah,
1:09:51
I guess,
1:09:52
by the way,
1:09:52
I think this might be a reason why startups remain attractive.
1:09:56
SPEAKER_00: And, you know, big tech, you know, now Facebook is big tech.
1:09:59
SPEAKER_00: They used to be a startup along with like all the other big companies.
1:10:03
You know,
1:10:03
they pay well,
1:10:04
they have a brand,
1:10:05
they give you that resume boost.
1:10:07
SPEAKER_00: It's easier to get hired afterwards.
1:10:10
SPEAKER_00: In the end,
1:10:10
SPEAKER_00: you know,
1:10:11
SPEAKER_00: like there will be teams inside of them,
1:10:12
SPEAKER_00: but a lot of them will be,
1:10:14
SPEAKER_00: you know,
1:10:14
SPEAKER_00: everyone's optimizing for their own thing.
1:10:16
Your equity is mostly cash and it's meaningless in terms of the
1:10:20
bigger picture.
1:10:21
Whereas at a startup,
1:10:22
like at Uber,
1:10:22
I remember before the IPO,
1:10:24
we used to think like that.
1:10:25
SPEAKER_00: What is the best thing for Uber?
1:10:27
SPEAKER_00: Because we were very much interested.
1:10:29
SPEAKER_00: We felt like we were...
1:10:31
SPEAKER_00: like, big enough owners.
1:10:32
SPEAKER_00: So I think this is why,
1:10:33
SPEAKER_00: like,
1:10:34
maybe this is a good thing that startups will always be able to
1:10:37
have this added
1:10:38
SPEAKER_00: thing when you're starting out.
1:10:39
SPEAKER_00: You know, people have large equity.
1:10:41
Even when you're up to,
1:10:42
like,
1:10:42
100 people,
1:10:43
it might still be significant enough.
1:10:45
And this is,
1:10:46
maybe it's not a bad thing that it's hard to compete with it.
1:10:48
SPEAKER_00: Because imagine if these big companies could.
1:10:50
SPEAKER_00: I mean, what would be left of it, right?
1:10:52
SPEAKER_00: Like, they would optimize the last thing out of everything.
1:10:55
SPEAKER_00: And at least they have to spend more money on it now, right?
1:10:58
SPEAKER_00: Yeah.
1:10:59
Yes,
1:11:00
which is more of the value that's created comes back to the
1:11:04
people who are creating it.
1:11:05
SPEAKER_00: I once talked with a person at a large company.
1:11:08
SPEAKER_00: I don't want to name them.
1:11:10
SPEAKER_00: They're a travel company, though.
1:11:12
SPEAKER_00: I was talking with this person about...
1:11:16
SPEAKER_00: something principal engineer.
1:11:18
SPEAKER_00: And I was like, Oh, how's the job is like, Oh, it's absolute mess.
1:11:22
SPEAKER_00: It's the, the, the monolith is still there after four years.
1:11:26
SPEAKER_00: It's another like four years to disassemble the experiment.
1:11:29
SPEAKER_00: Like we have experiments everywhere, but they're all messy.
1:11:32
SPEAKER_00: I was like, oh, wow, that sounds like not a great place.
1:11:35
SPEAKER_00: He's like,
1:11:35
SPEAKER_00: but look,
1:11:36
the upside is like my job is secure for five more years and
1:11:39
this is why I pay me
1:11:40
SPEAKER_00: the big bucks.
1:11:41
SPEAKER_00: So I'm not complaining.
1:11:43
SPEAKER_00: And it was,
1:11:44
SPEAKER_00: you know,
1:11:44
SPEAKER_00: like,
1:11:45
SPEAKER_00: I guess a good take of like,
1:11:46
SPEAKER_00: yes,
1:11:46
SPEAKER_00: like I'm not saying you want to optimize for this,
1:11:48
SPEAKER_00: but this is the reality.
1:11:49
SPEAKER_00: And that company understood,
1:11:51
SPEAKER_00: you know,
1:11:51
SPEAKER_00: they pay well,
1:11:52
SPEAKER_00: they relocated this person,
1:11:53
SPEAKER_00: you know,
1:11:54
SPEAKER_00: all the benefits.
1:11:55
SPEAKER_00: And he actually had his challenges set actually for the next four
1:11:58
years because he's worked at this environment you know it's
1:12:00
just one thing after
1:12:02
the other so yeah it's one of these things so as closing i just
1:12:06
have some rapid
1:12:07
questions which is like i'll just i'll just fire and and then
1:12:09
you tell me what is
1:12:10
your favorite programming language although you answered this
1:12:13
already what is your
1:12:14
second favorite one after small talk with javascript javascript
1:12:19
okay why it's uh
1:12:22
SPEAKER_00: it's just small talk
1:12:24
SPEAKER_00: Okay.
1:12:25
SPEAKER_00: What is your favorite AI tool that you're using right now?
1:12:30
SPEAKER_01: CLAWD.
1:12:31
SPEAKER_01: I use it for all kinds of different things.
1:12:33
SPEAKER_01: CLAWD code as well or just CLAWD?
1:12:35
CLAWD code under the covers of Cursor or Augment or I don't
1:12:43
know what Rue code is.
1:12:45
SPEAKER_00: No, so CLAWD code is a different command line tool that CLAWD has.
1:12:49
SPEAKER_00: It's an agent of itself.
1:12:52
SPEAKER_00: It's not the model.
1:12:53
SPEAKER_00: Have not used, but we'll try it now.
1:12:56
SPEAKER_00: After we just try it and just see how it compares.
1:12:58
SPEAKER_01: Yeah, I'll wait till we're done talking now.
1:13:00
SPEAKER_00: Yeah.
1:13:03
And what is a book that you would recommend,
1:13:06
the one that you have not written?
1:13:07
SPEAKER_00: We know your books.
1:13:08
SPEAKER_00: The one that you enjoyed.
1:13:09
SPEAKER_00: It can be fiction, it can be nonfiction.
1:13:12
SPEAKER_00: The Timeless Way of Building by Christopher Alexander.
1:13:15
SPEAKER_00: Well, Kent, this was a lot of fun.
1:13:18
SPEAKER_00: It was good to reconnect.
1:13:20
SPEAKER_00: Oh, yes.
1:13:20
SPEAKER_00: Great talking to you.
1:13:21
SPEAKER_00: Thanks.
1:13:22
SPEAKER_00: I'm just happy to see how much energy, how energized you are with coding.
1:13:26
SPEAKER_00: Because I think when I started with some of these tools,
1:13:29
SPEAKER_00: I was like a bit of a dread,
1:13:30
SPEAKER_00: like,
1:13:31
SPEAKER_00: oh,
1:13:31
SPEAKER_00: what's what they're going to do,
1:13:32
SPEAKER_00: et cetera.
1:13:32
SPEAKER_00: But the more I use them, I'm not at your level yet.
1:13:35
SPEAKER_00: The more I'm also like,
1:13:36
SPEAKER_00: this is actually,
1:13:37
SPEAKER_00: it does bring a lot of fun and joy back into it and just more ambition.
1:13:41
SPEAKER_01: Yep.
1:13:42
SPEAKER_01: Yeah,
1:13:42
and I think that organizations are going to have to get used to
1:13:45
throwing away a lot
1:13:46
SPEAKER_01: more code.
1:13:46
SPEAKER_01: Because you can try ideas so much more cheaply,
1:13:49
SPEAKER_01: you're going to generate 10 times as many artifacts as you used to,
1:13:55
SPEAKER_01: but still only one of them is worth keeping.
1:13:59
But throwing away completed experiments,
1:14:05
I almost said failed,
1:14:06
completed experiments...
1:14:10
SPEAKER_01: needs to be, you get the pat on the head for doing that.
1:14:13
SPEAKER_01: Excellent.
1:14:14
SPEAKER_01: Eight.
1:14:15
SPEAKER_01: Eight this week, only six late last week.
1:14:17
SPEAKER_01: Super.
1:14:18
SPEAKER_01: How many of them lasted?
1:14:19
SPEAKER_01: Doesn't matter.
1:14:22
SPEAKER_01: And getting used to that, I think, is going to be an interesting shift.
1:14:26
SPEAKER_01: The companies that have the opportunity that can be explored in that way,
1:14:33
SPEAKER_01: if you get used to just quantity of ideas explored,
1:14:39
SPEAKER_01: you're going to have a huge,
1:14:40
SPEAKER_01: huge advantage.
1:14:42
SPEAKER_00: Yeah, it's going to be changes.
1:14:44
SPEAKER_00: So this will be an exciting place to see.
1:14:46
SPEAKER_00: So it was great chatting and we'll see where this goes.
1:14:49
SPEAKER_00: All right, Gergai, so good to talk to you again.
1:14:51
SPEAKER_00: Look forward to talking to you again soon.
1:14:53
I really enjoyed catching up with Kent and it's motivating to
1:14:56
see how these AI
1:14:57
SPEAKER_00: tools can make programming so much more fun,
1:14:59
SPEAKER_00: even for someone who's been a coder for decades.
1:15:02
SPEAKER_00: You can read more about what Kent is up to on his regular newsletter,
1:15:05
SPEAKER_00: Tidy First,
1:15:06
SPEAKER_00: which is linked in the show notes below.
1:15:08
For a deep dive into Facebook's engineering culture and an
1:15:11
analysis on why Scrum
1:15:12
SPEAKER_00: and Agile,
1:15:12
SPEAKER_00: with a capital A,
1:15:14
SPEAKER_00: is no longer that relevant at Big Tech at scale-ups,
1:15:16
SPEAKER_00: check out the Pragmatic Engineer deep dives,
1:15:18
SPEAKER_00: also linked in the show notes below.
1:15:19
SPEAKER_00: If you've enjoyed this podcast,
1:15:21
SPEAKER_00: please do subscribe on your favorite podcast platform and on YouTube.
1:15:24
This helps more people discover the podcast,
1:15:26
and a special thank you if you leave a rating.
1:15:28
SPEAKER_00: Thanks, and see you in the next one.