Whisper.cpp 生成的字幕 https://emacstalk.codeberg.page/podcast/001/ https://rustacean-station.org/episode/andrew-kelley/

EmacsTalk 第一期.srt

1
00:00:00,000 --> 00:00:06,800
大家好,今天是2021年6月27号,是InMessTalk的第一期。

2
00:00:06,800 --> 00:00:11,700
这一期主要由我来讲一下这个节目的背景。

3
00:00:11,700 --> 00:00:17,200
在介绍背景之前,首先进行一下主播的个人介绍。

4
00:00:17,200 --> 00:00:25,100
我是14年大学毕业,现在是在杭州的一家互联网公司工作。

5
00:00:25,300 --> 00:00:31,300
我之前用过的语言也比较多,Java,Python,Curl都用过。

6
00:00:31,300 --> 00:00:34,400
现在主要是用Go和Rust。

7
00:00:34,400 --> 00:00:39,900
我接触Emacs的原因最早是由于SRCP。

8
00:00:39,900 --> 00:00:44,700
由于SRCP是用Scheme来作为讲解,

9
00:00:44,700 --> 00:00:49,400
而Scheme作为一门Lisp方言,

10
00:00:49,400 --> 00:00:52,400
用Emacs来编辑是。

11
00:00:52,470 --> 00:00:53,710
再恰当不过的了

12
00:00:53,710 --> 00:00:58,590
但是由于emacs的上手难度比较高

13
00:00:58,590 --> 00:01:01,550
我前前后后折腾了好多次

14
00:01:01,550 --> 00:01:06,510
大概是在16年才正式把emacs真正用了起来

15
00:01:06,510 --> 00:01:10,950
到现在算起来的话大概有五六年的样子

16
00:01:10,950 --> 00:01:15,390
我最早使用的配置是一个clurl的

17
00:01:15,390 --> 00:01:19,310
名字大概是叫emacs-for-clurl

18
00:01:19,670 --> 00:01:24,670
在那基础之上我陆陆续续加了很多自己的配置

19
00:01:24,670 --> 00:01:28,790
现在我的配置相当于把它已经抛弃了

20
00:01:28,790 --> 00:01:31,750
完全是自己来做管理

21
00:01:31,750 --> 00:01:38,390
我现在主要是用user package来进行包的安装与配置

22
00:01:38,390 --> 00:01:41,510
对于一些比较重要的包

23
00:01:41,510 --> 00:01:44,110
比如说像magate, evo

24
00:01:44,950 --> 00:01:46,890
submodule把它管理起来,

25
00:01:46,890 --> 00:01:51,190
防止由于升级导致软件的一些崩溃。

26
00:01:51,190 --> 00:01:57,190
除此之外,我主要是用在evol下使用emacs,

27
00:01:57,190 --> 00:02:02,510
由于在使用emacs大概在两到三年以后,

28
00:02:02,510 --> 00:02:05,110
我的小指隐隐作痛,

29
00:02:05,110 --> 00:02:09,150
后来把大小写件改成control,

30
00:02:09,150 --> 00:02:11,230
情况有所缓解,

31
00:02:11,230 --> 00:02:14,190
但是也不能根本上解决问题。

32
00:02:14,590 --> 00:02:17,070
然后大概在一年多以前,

33
00:02:17,070 --> 00:02:22,430
我把我的所有按键放到了evol之上,

34
00:02:22,430 --> 00:02:25,870
现在我的小指基本上已经彻底解放了。

35
00:02:25,870 --> 00:02:29,750
除了这两个最重要的以外,

36
00:02:29,750 --> 00:02:35,790
其他的package还有像company用来做代码的补全,

37
00:02:35,790 --> 00:02:37,430
然后是listview。

38
00:02:37,430 --> 00:02:40,430
用来做代码的跳转

39
00:02:40,430 --> 00:02:46,030
基本上这几个是所有编程语言都需要用到的功能

40
00:02:46,030 --> 00:02:48,590
除了编程以外

41
00:02:48,590 --> 00:02:54,830
我现在用MU-FoE来进行邮件的收发与读取

42
00:02:54,830 --> 00:02:59,870
用EL-Feed来进行我RSS管理

43
00:02:59,870 --> 00:03:01,350
来管理我RSS的订阅

44
00:03:01,350 --> 00:03:07,110
基本上我所有的工作流都不用离开Emacs

45
00:03:07,990 --> 00:03:13,270
熟悉我的读者应该还知道我之前还有写一个博客

46
00:03:13,270 --> 00:03:15,790
刘家财.net

47
00:03:15,790 --> 00:03:19,550
感兴趣的读者也可以去参考

48
00:03:19,550 --> 00:03:24,670
由于在之前的公司有个花名叫西瓜

49
00:03:24,670 --> 00:03:29,870
所以说在现在互联网上我的笔名一般是

50
00:03:29,900 --> 00:03:33,900
如果你看见了这个名字一般就是我没错了

51
00:03:33,900 --> 00:03:40,100
好下面介绍一下为什么要做这一档节目

52
00:03:40,100 --> 00:03:44,700
首先我个人是podcast资深受益者

53
00:03:44,700 --> 00:03:51,300
我在很多闲暇时间都会把podcast开着

54
00:03:51,300 --> 00:03:54,100
听的东西也比较杂

55
00:03:54,100 --> 00:03:58,300
与编程相关的无关的都会去听一听

56
00:03:58,300 --> 00:04:00,300
作为一种陪伴音

57
00:04:00,300 --> 00:04:03,900
就像之前小时候家里面没人看电视

58
00:04:03,900 --> 00:04:05,900
电视还要开着一样

59
00:04:05,900 --> 00:04:09,900
另一方面工作了大概有六七年

60
00:04:09,900 --> 00:04:14,300
我发现我们不仅仅需要干巴巴的干货

61
00:04:14,300 --> 00:04:18,900
也需要一些湿货来丰富我们的生活

62
00:04:18,900 --> 00:04:20,900
毕竟人不是机器人

63
00:04:20,900 --> 00:04:22,900
我们的生活是一种生活

64
00:04:22,900 --> 00:04:26,280
就像我们有时候解决一个bug,

65
00:04:26,280 --> 00:04:29,780
飞机了飞了老大进也没有去解决,

66
00:04:29,780 --> 00:04:32,380
但是有可能睡觉就解决了。

67
00:04:32,380 --> 00:04:35,980
所以在做这档节目之前,

68
00:04:35,980 --> 00:04:40,380
我就给他的定位是进行失货知识的分享。

69
00:04:40,380 --> 00:04:45,380
读者不用花费太大的精力来收听,

70
00:04:45,380 --> 00:04:48,780
就把它当做一种陪伴音就好了。

71
00:04:48,780 --> 00:04:53,180
最后一点,作为e-mess的资深用户,

72
00:04:53,180 --> 00:04:57,580
发现这个社区有很多有意思的东西,

73
00:04:57,580 --> 00:05:02,380
然后这些有意思的东西是由谁创造的呢?

74
00:05:02,380 --> 00:05:05,980
我猜想肯定是一些比较有趣的人,

75
00:05:05,980 --> 00:05:09,980
有趣的人肯定会有一些有趣的事情。

76
00:05:09,980 --> 00:05:15,980
希望通过这档节目能把它的内容带给更多的观众。

77
00:05:15,980 --> 00:05:18,700
他们挖掘出来呈现在读者面前

78
00:05:18,700 --> 00:05:23,300
用一句话总结这档节目的目标

79
00:05:23,300 --> 00:05:26,380
就是深入专注e-Market社区

80
00:05:26,380 --> 00:05:29,460
探寻e-Market社区背后的人与事

81
00:05:29,460 --> 00:05:34,820
在开始这档节目之前我也问过我这个问题

82
00:05:34,820 --> 00:05:36,620
为什么是我

83
00:05:36,620 --> 00:05:40,100
为什么要由我来做这件事情

84
00:05:40,100 --> 00:05:41,540
我能做好吗

85
00:05:41,540 --> 00:05:43,020
我能做多久

86
00:05:43,780 --> 00:05:47,380
要回答这些问题其实很难

87
00:05:47,380 --> 00:05:51,660
我也不敢说自己有什么雄心壮志

88
00:05:51,660 --> 00:05:56,700
只希望尽自己的一些力量发挥自己的光和热

89
00:05:56,700 --> 00:05:59,300
私心当然也是有的

90
00:05:59,300 --> 00:06:04,740
就是希望能结识更多的更多志同道合的朋友

91
00:06:04,740 --> 00:06:07,500
毕竟程序员的交际圈

92
00:06:07,500 --> 00:06:12,330
如果说能认识喜欢相同事物的人

93
00:06:12,330 --> 00:06:16,730
我感觉这也是一件对我个人比较有帮助的事情

94
00:06:16,730 --> 00:06:20,130
毕竟多条朋友多条路嘛

95
00:06:20,130 --> 00:06:28,410
最后我在这里向emacs社区内所有乐于分享的朋友发出一个邀请

96
00:06:28,410 --> 00:06:31,410
希望大家能来联系我们

97
00:06:31,410 --> 00:06:36,010
把你自己的故事分享给社区内的所有人

98
00:06:36,410 --> 00:06:40,410
希望大家不要在意自己的emacs水平

99
00:06:40,410 --> 00:06:45,010
毕竟每个人的学习背景知识都不一样

100
00:06:45,010 --> 00:06:49,010
文道有先后 术业有专攻

101
00:06:49,010 --> 00:06:52,010
只要你认为你的故事够精彩

102
00:06:52,010 --> 00:06:54,210
我就希望你来联系我

103
00:06:54,210 --> 00:06:57,410
最后祝大家吃好喝好玩好

104
00:06:57,410 --> 00:06:58,810
下期节目再见

Zig with Andrew Kelley.srt

1
00:00:00,000 --> 00:00:10,720
 [Music]

2
00:00:10,720 --> 00:00:14,880
 Hello and welcome to another episode of "Restation Station" on your host Alan Weimar.

3
00:00:14,880 --> 00:00:18,480
 Today I am with Andrew Kelly, the creator of Zig.

4
00:00:18,480 --> 00:00:21,120
 It's really a great honor to have a language creator here.

5
00:00:21,120 --> 00:00:23,840
 I like I told him before the show, I have great respect for language creators,

6
00:00:23,840 --> 00:00:27,040
 especially ones that can get the language very popular.

7
00:00:27,040 --> 00:00:29,920
 Also, as you guys have probably heard in different episodes,

8
00:00:29,920 --> 00:00:32,240
 we've been asking people what their experience with Zig is,

9
00:00:32,240 --> 00:00:33,920
 and we haven't heard much about it.

10
00:00:33,920 --> 00:00:36,160
 Myself have heard a lot of good things about Zig,

11
00:00:36,160 --> 00:00:39,120
 but the other thing on the other side is I heard people say that Zig just doesn't

12
00:00:39,120 --> 00:00:40,960
 quite click for them as it does for Rust.

13
00:00:40,960 --> 00:00:43,760
 So I think this is a great opportunity for people to understand, you know,

14
00:00:43,760 --> 00:00:48,640
 what is Zig and is it actually better for them than maybe Rust or maybe something else.

15
00:00:48,640 --> 00:00:51,840
 So Andrew, why don't you go ahead and introduce yourself and let's get started.

16
00:00:51,840 --> 00:00:54,560
 Hey, Alan, nice to meet you and thanks for having me on the show.

17
00:00:54,560 --> 00:01:00,480
 So I'm Andrew Kelly, I originally made Zig and have now transitioned into the role of

18
00:01:00,480 --> 00:01:04,080
 President of the Board of the Zig Software Foundation,

19
00:01:04,080 --> 00:01:09,200
 as well as still the lead developer, the Zig project.

20
00:01:09,200 --> 00:01:11,840
 So we're full steam ahead.

21
00:01:11,840 --> 00:01:13,440
 Right now we're working towards the...

22
00:01:13,440 --> 00:01:16,720
 It's hard for me to talk about anything else because it's...

23
00:01:16,720 --> 00:01:20,480
 We're right, super close to finishing the self-hosted compiler,

24
00:01:20,480 --> 00:01:22,960
 which is major milestone for any programming language.

25
00:01:23,520 --> 00:01:26,240
 Oh, I feel like maybe we should have waited for that to happen now because,

26
00:01:26,240 --> 00:01:27,840
 I mean, this is going to happen, right?

27
00:01:27,840 --> 00:01:30,160
 I feel like now just too soon to record.

28
00:01:30,160 --> 00:01:32,000
 Yeah, it's imminent. We're right on the edge right now,

29
00:01:32,000 --> 00:01:34,960
 so it'd be kind of funny to review, to listen to this.

30
00:01:34,960 --> 00:01:37,760
 You know, if you're hearing the podcast after it's been released,

31
00:01:37,760 --> 00:01:41,840
 it might be kind of funny to hear the difference in attitude and my attitude before and after.

32
00:01:41,840 --> 00:01:44,560
 Or could it also be difference in, you know, implementation?

33
00:01:44,560 --> 00:01:48,000
 Maybe things change before it gets released and it happens a lot, I believe, right?

34
00:01:48,000 --> 00:01:48,320
 Oh, yeah.

35
00:01:48,320 --> 00:01:48,960
 What do they say?

36
00:01:48,960 --> 00:01:51,840
 The last 10% takes 90% of the time.

37
00:01:53,280 --> 00:01:57,520
 This one I haven't heard by, I just remember the 80/20 rule that applies to so many situations,

38
00:01:57,520 --> 00:01:59,040
 but that could also be...

39
00:01:59,040 --> 00:02:00,720
 Is it really 90/10? I haven't heard this one.

40
00:02:00,720 --> 00:02:02,160
 Oh, wait, I got it. I got it. Okay.

41
00:02:02,160 --> 00:02:09,600
 The first 90% takes 90% of the time, obviously, and then the last 10% also takes 90% of the time.

42
00:02:09,600 --> 00:02:10,960
 That's the joke.

43
00:02:10,960 --> 00:02:12,880
 Okay. Got it.

44
00:02:12,880 --> 00:02:15,920
 Yeah, I mean, as we're talking before the show for a long time,

45
00:02:15,920 --> 00:02:18,960
 I did hear... I did some research about you and Ziglang.

46
00:02:19,680 --> 00:02:23,520
 The... My understanding, maybe you can let me know if I misunderstand, but my understanding

47
00:02:23,520 --> 00:02:29,520
 about Zig is that Zig is meant to correct the problems of C, right?

48
00:02:29,520 --> 00:02:31,680
 Russ is a little bit different in their mindset.

49
00:02:31,680 --> 00:02:35,440
 I think it's more about like, they want to make sure things are safe when you program.

50
00:02:35,440 --> 00:02:39,680
 And of course, by that, it's like, well, let's compare that to, you know, unsafe things such as C.

51
00:02:39,680 --> 00:02:44,640
 But I think Zig, particularly the way that I see it is, they really are trying to replace

52
00:02:45,680 --> 00:02:50,320
 C or correct the problems of C in particular. Is that a proper way that I could say the kind

53
00:02:50,320 --> 00:02:51,920
 of like the mission statement for Zig?

54
00:02:51,920 --> 00:02:55,200
 Yeah. I mean, I think that's fair. If you'll allow me to say it in my own words,

55
00:02:55,200 --> 00:03:03,680
 I would say that Russ successfully manages to eliminate a certain category of problems,

56
00:03:03,680 --> 00:03:07,760
 which is a really nice benefit and Russ is succeeding in that arena.

57
00:03:07,760 --> 00:03:14,560
 But it does that at the cost of complexity and sometimes at the cost of performance.

58
00:03:14,560 --> 00:03:20,800
 And so this is a tradeoff. And if you're willing to pay the cost for eliminating entire class of

59
00:03:20,800 --> 00:03:26,320
 bugs, then that's great. And Russ is excellent for that tradeoff. But with Zig's approach,

60
00:03:26,320 --> 00:03:32,480
 we're looking at the situation and saying, there's a lot we can do to improve on the C or C++

61
00:03:32,480 --> 00:03:36,960
 approach to things without giving up simplicity or performance.

62
00:03:36,960 --> 00:03:42,800
 So I'd say that's kind of like the Russ Zig tradeoff in their respective domains.

63
00:03:43,360 --> 00:03:47,280
 Well, here's the interesting thing. People always talk about performance, right? So I hear a lot of

64
00:03:47,280 --> 00:03:51,120
 people say, you know, I don't want to use Russ because performance will suffer. Like you said,

65
00:03:51,120 --> 00:03:56,640
 there's some cost to it, right? But you make quite some claims about Zig that even Zig is faster than

66
00:03:56,640 --> 00:04:01,120
 C, right? Do you actually have some benchmarks to prove that? Or is it based off of kind of like

67
00:04:01,120 --> 00:04:06,720
 wet finger testing? Yeah. So we can look at micro benchmarks or we can look at macro benchmarks,

68
00:04:06,720 --> 00:04:12,640
 but at the micro level, the semantics of the language are just more optimizable. So just as an example,

69
00:04:12,640 --> 00:04:19,520
 in C, you have a difference between signed and unsigned integers. Even though optimizers would

70
00:04:19,520 --> 00:04:23,520
 be able to make more assumptions about unsigned integers because they have a different range,

71
00:04:23,520 --> 00:04:30,000
 they're hamstrained by the fact that arithmetic unsigned integers in C is well defined to

72
00:04:30,000 --> 00:04:36,160
 wrap. And for some reason, they do this split on signed because if you use a signed integer in C,

73
00:04:36,160 --> 00:04:42,160
 then now it's undefined behavior to wrap the integer. So now the optimization of a compiler

74
00:04:42,160 --> 00:04:47,840
 is able to do math on signed integers better. But this is kind of a weird arbitrary situation.

75
00:04:47,840 --> 00:04:54,480
 So as just a simple example, and Zig arithmetic is there's just different operators for wrapping

76
00:04:54,480 --> 00:05:00,960
 or assertion arithmetic. So if you just use the plus operator, then you're asserting that that's

77
00:05:00,960 --> 00:05:05,840
 not going to overflow the integer. And that's a difference with C. And you can just look at

78
00:05:05,840 --> 00:05:11,280
 like a Godbolt example and just notice that machine code is better for unsigned integers in Zig.

79
00:05:11,280 --> 00:05:16,560
 That's like a micro one. And then at the macro level, just take a look at the Zig self-hosted

80
00:05:16,560 --> 00:05:22,240
 compiler. It's way faster than them clang at for an equivalent source code size. And that's just

81
00:05:22,240 --> 00:05:29,040
 because Zig is a language that guides you towards writing faster code. Okay, yeah, that makes sense.

82
00:05:29,040 --> 00:05:32,080
 It's kind of a little bit weird, right? Because like to me, I think about it, it's like, they're

83
00:05:32,080 --> 00:05:35,680
 just bits, right? I guess you could just slightly change how you interact with them.

84
00:05:35,680 --> 00:05:39,680
 Why would you have undefined behavior that this doesn't make sense for me? Is this like a

85
00:05:39,680 --> 00:05:44,960
 committee choice or something? Or is it like initial guys who wrote C kind of did it one way,

86
00:05:44,960 --> 00:05:48,000
 and then the committee said, okay, let's do it some other way. Do you have any idea?

87
00:05:48,000 --> 00:05:51,520
 Yeah, it's funny that you say they're just bits, right? You're kind of going at this at an

88
00:05:51,520 --> 00:05:56,640
 undefined behavior angle. But if I could just side tangent real quick, the idea is you put into your

89
00:05:56,640 --> 00:06:01,680
 head, yeah, they are just bits, aren't they? So what if we just treat all of our entire input space

90
00:06:01,680 --> 00:06:06,400
 as bits and our output space as bits and just do the best possible thing given the hardware that

91
00:06:06,400 --> 00:06:11,760
 we have to make the input go to the output. And that kind of modeling of programs is what makes

92
00:06:11,760 --> 00:06:16,320
 your performance the best. But that's not what you're talking about. You're asking me about

93
00:06:16,320 --> 00:06:20,240
 undefined behavior. Like, why can't we just define all the behavior is kind of what you're asking?

94
00:06:20,240 --> 00:06:23,760
 Is that right? Well, that's a second question, right? My first question is like, wait a minute,

95
00:06:23,760 --> 00:06:26,720
 there are just bits, right? You just have to kind of treat them slightly differently, right? If

96
00:06:26,720 --> 00:06:30,400
 like you have sign, right, I think it's always the first bit is the one that's, you know,

97
00:06:30,400 --> 00:06:36,720
 signed and unsigned. Now, if it's sorry, plus or minus, right? If it's unsigned, then obviously you have,

98
00:06:36,720 --> 00:06:41,280
 I don't know what the calculation is up top of my head, but you have more data you can store,

99
00:06:41,280 --> 00:06:47,440
 but all positive. So it's just like trying to understand like, why is it such an issue to do

100
00:06:47,440 --> 00:06:51,600
 this stuff? And then the second question, like you said, is why is it that we have undefined behavior,

101
00:06:51,600 --> 00:06:56,080
 right? Why can't we just define it eventually? And then I think things would work out. But I

102
00:06:56,080 --> 00:07:01,200
 understand that there is, you know, a code out there that is using this kind of undefined behavior.

103
00:07:01,200 --> 00:07:05,840
 And maybe there is, I mean, the thing is undefined behavior, like, really, nobody writes it. And then

104
00:07:05,840 --> 00:07:10,960
 like something somebody actually implements something. And then people just rely on that. Or like,

105
00:07:10,960 --> 00:07:15,040
 how can we say something about this kind of situation? I think I can shed some light on this.

106
00:07:15,040 --> 00:07:19,600
 So I think that part of what's confusing about undefined behavior for people is that

107
00:07:19,600 --> 00:07:25,040
 it's actually a broad term, and we need to kind of narrow in on different kinds of undefined behavior.

108
00:07:25,040 --> 00:07:30,560
 So as an example, let's just say that we took Rust and everything's the same, but we're going to make

109
00:07:30,560 --> 00:07:37,520
 one change to the language specification. And the change is that if you ever multiply by exactly

110
00:07:37,520 --> 00:07:44,160
 1000, that's undefined behavior. Okay, this would be ridiculous. This would be a really bad idea

111
00:07:44,160 --> 00:07:49,040
 that would only be a problem. And you would be right to ask, that doesn't make any sense. Why

112
00:07:49,040 --> 00:07:54,160
 don't we just define multiplication by 1000 to do what you would expect it to do? And then no one's

113
00:07:54,160 --> 00:08:01,440
 going to run into this problem. Right. That's one angle. And the weird thing about C, specifically,

114
00:08:01,440 --> 00:08:08,080
 and to some degree, C++ kind of by inheritance, upon intended, is that there are some things in

115
00:08:08,080 --> 00:08:12,000
 the specification like this, be skeptical about what I'm about to say, because I can't remember

116
00:08:12,000 --> 00:08:16,400
 very clearly. But I think there's something like, if your source file doesn't end in a new line,

117
00:08:16,400 --> 00:08:20,960
 it's undefined behavior, or something crazy like that. That's kind of like the multiplying by a

118
00:08:20,960 --> 00:08:25,520
 1000 thing. Right. Like, what is that doing in there? Just get that out of there, simplify it

119
00:08:25,520 --> 00:08:31,760
 by taking that out. Right. That's like one angle to look at undefined behavior. But there's another

120
00:08:31,760 --> 00:08:37,520
 really important different angle to look at. And that is that the compiler needs to make some

121
00:08:37,520 --> 00:08:43,840
 assumptions about things that will never happen so that it can do optimizations. And this is true

122
00:08:43,840 --> 00:08:49,040
 in Rust too. Like Rust, the Rust compiler needs to assume that certain things will never happen.

123
00:08:49,040 --> 00:08:54,240
 And it is doing these things in order to make the output reasonable. And an example of this would be

124
00:08:54,240 --> 00:09:01,200
 if you allocate memory from, and then you grab some memory, and I'm holding up my left hand here,

125
00:09:01,200 --> 00:09:06,320
 and then I allocate some more memory, and I'm holding up my right hand. And the premise is that

126
00:09:06,320 --> 00:09:12,800
 you can never access these two allocations through each other. This is called pointer provenance.

127
00:09:12,800 --> 00:09:18,160
 And this is a really, really important assumption for compilers to make. Otherwise,

128
00:09:18,160 --> 00:09:21,760
 they're going to out garbage code. So even Rust wants to have this assumption.

129
00:09:21,760 --> 00:09:26,720
 And the language is built specifically so that you can never violate this premise.

130
00:09:26,720 --> 00:09:31,280
 And that way in the back end, when Rust emits LLVM code, it can promise LLVM,

131
00:09:31,280 --> 00:09:36,400
 don't worry, this will never happen. You can safely assume that if one of these pointers

132
00:09:36,400 --> 00:09:41,120
 accesses the other pointer's data, that's undefined behavior. And that assumption,

133
00:09:41,120 --> 00:09:46,640
 that existence of undefined behavior is what gives the optimizer the capability to do anything

134
00:09:46,640 --> 00:09:51,200
 useful at all with the program. So that's the other angle. And I feel like the sometimes the

135
00:09:51,200 --> 00:09:55,920
 discourse on undefined behavior gets a little confused because people are thinking about one,

136
00:09:55,920 --> 00:10:00,080
 but they're talking about the other. So can we actually talk about undefined behavior? So

137
00:10:00,080 --> 00:10:04,160
 I hear this term a lot. And sometimes I'm in meetings with people, with clients, and they

138
00:10:04,160 --> 00:10:07,520
 keep using acronyms and I ask them, what does this actually mean? And sometimes they're not even

139
00:10:07,520 --> 00:10:12,560
 able to actually tell me or even explain what terms mean. But like, we keep hearing this a lot

140
00:10:12,560 --> 00:10:17,600
 in the community, undefined behavior. So undefined behavior means that anything can happen, right?

141
00:10:17,600 --> 00:10:22,400
 That there's no true definition. So you can just not expect something. Is that a way to actually

142
00:10:22,400 --> 00:10:27,360
 kind of define undefined behavior? Yeah, I mean, what you just said would be a consequence of the

143
00:10:27,360 --> 00:10:32,880
 definition, a pithy way that I would try to say it would be undefined behavior is a promise that

144
00:10:32,880 --> 00:10:38,160
 the optimizer can assume will never be broken. It's a premise that it's going to operate, that's

145
00:10:38,160 --> 00:10:44,000
 going to do its optimizations around. It's going to assume something so that it can know that it's

146
00:10:44,000 --> 00:10:47,680
 safe to do certain changes to your code. That sounds like more like when it's operating on

147
00:10:47,680 --> 00:10:52,080
 defined behavior. No, that's the way I'm understanding what you're saying. Well, specifically, though,

148
00:10:52,080 --> 00:10:58,400
 the undefined behavior is what happens if the premise is violated. So here's an example, I can

149
00:10:58,400 --> 00:11:04,000
 make an analogy. Let's say that you buy a computer and the motherboard has the hardware. This is a

150
00:11:04,000 --> 00:11:09,440
 hardware analogy. It has the text specs and it says, okay, this thing needs to be powered by five volts.

151
00:11:09,440 --> 00:11:16,240
 If you give it one volts, it's not going to behave correctly. It needs five volts, right? So powering

152
00:11:16,240 --> 00:11:21,120
 your motherboard with one volt, it's going to have undefined behavior. Who knows what it will do?

153
00:11:21,120 --> 00:11:25,360
 Will it be if correctly, will it have wrong results undefined? The only way to get defined

154
00:11:25,360 --> 00:11:29,360
 behavior is give it the five volts that the hardware needs. That's like a hardware example.

155
00:11:29,360 --> 00:11:33,040
 But we're talking about software. So just take that analogy and apply it to software.

156
00:11:33,040 --> 00:11:37,200
 Yeah, how about Zig, though? Does it actually have some undefined behavior?

157
00:11:37,200 --> 00:11:43,200
 Yeah, yeah. One of the differences between Zig and Rust is the existence of more or different

158
00:11:43,200 --> 00:11:48,320
 categories of release modes. So Rust has a debug, which is the default, and a release,

159
00:11:48,320 --> 00:11:54,480
 which you opt into. Zig has a debug, which is the default, and three release modes.

160
00:11:55,200 --> 00:12:02,000
 The three release modes are small, fast, and safe. You have to choose one of these three

161
00:12:02,000 --> 00:12:07,920
 things. You don't get all of them. So that's a difference with Rust. So in Zig, if you choose

162
00:12:07,920 --> 00:12:15,360
 safe, then we have safety checks that check for undefined behavior and we'll make it defined to

163
00:12:15,360 --> 00:12:21,680
 call panic instead of invoking undefined behavior. But if you choose small or fast,

164
00:12:21,680 --> 00:12:27,600
 then you are opting into the optimizer assuming that undefined behavior will never happen.

165
00:12:27,600 --> 00:12:34,080
 And then also, there are some kinds of undefined behavior that even the safe modes cannot detect.

166
00:12:34,080 --> 00:12:39,680
 And that's an important fact for the when use cases to use Rust versus to use Zig.

167
00:12:39,680 --> 00:12:45,120
 You know, if your goal is 100% in memory safety encapsulation, that's what Rust is going to give

168
00:12:45,120 --> 00:12:49,360
 you, because there are some instances in Zig where even our safe mode is not going to catch

169
00:12:49,360 --> 00:12:53,040
 certain kinds of undefined behavior. I want to ask a question, but I feel like the way I wanted

170
00:12:53,040 --> 00:12:59,280
 to ask it was not not proper. But do you feel that this is not something that you want to actually

171
00:12:59,280 --> 00:13:02,800
 fix that you want to make it really safe? I mean, because I can't imagine people saying,

172
00:13:02,800 --> 00:13:07,760
 oh, I just want a really, really fast program. I don't really care if the memory is not safe.

173
00:13:07,760 --> 00:13:14,640
 Well, there's some cases where that is correct. So as an example, if you're targeting WebAssembly

174
00:13:14,640 --> 00:13:19,040
 and you're already going to run the code in the sandbox, then there's no reason to pay the cost

175
00:13:19,040 --> 00:13:23,440
 of safety since you're already sandboxing it in the runtime. So that'd be one example.

176
00:13:23,440 --> 00:13:30,000
 Another example would be you're shipping a video game like Ocarina of Time to the console or

177
00:13:30,000 --> 00:13:34,480
 client's computer, and the player's just going to run the game on their computer. So if they

178
00:13:34,480 --> 00:13:40,400
 manage to exploit their own system, then it just kind of doesn't matter. So that'd be another

179
00:13:40,400 --> 00:13:47,520
 fair example. But and also answer to your question, I do have a personal goal changing this list

180
00:13:47,520 --> 00:13:53,360
 to be roughly equivalent to like, go or rust in terms of safety. So as an example, one of the

181
00:13:53,360 --> 00:13:59,840
 problem areas right now would be escaped pointers to local variables. So you have a function,

182
00:13:59,840 --> 00:14:05,280
 take the address of a local variable, and then you store that somewhere, and then the function

183
00:14:05,280 --> 00:14:10,000
 returns. That pointer now becomes a dangling pointer. And if you use it, that's called a use

184
00:14:10,000 --> 00:14:15,200
 after free. Now there is an important difference between Zig and Rust here, which is that in Rust,

185
00:14:15,200 --> 00:14:20,480
 the existence of that pointer is already undefined behavior. I hope someone can correct me if I'm

186
00:14:20,480 --> 00:14:24,480
 wrong. I might be wrong about this exact example, but I know that Rust does have an important

187
00:14:24,480 --> 00:14:30,640
 distinction about even creating like bad types, even if they're never like loaded or stored,

188
00:14:30,640 --> 00:14:35,200
 that's already undefined behavior. Whereas in Zig, only an access would cause the problem. So you

189
00:14:35,200 --> 00:14:40,000
 can actually store it fine as long as you never load or save it later. Anyway, my point is that

190
00:14:40,000 --> 00:14:45,120
 I do actually have a plan to address this, which is to do escape analysis on local variables,

191
00:14:45,120 --> 00:14:51,440
 keep allocate them with an allocator that does not reuse virtual memory addresses,

192
00:14:51,440 --> 00:14:56,960
 free them at the end of the function, and then these no longer will cause use after free. They'll

193
00:14:56,960 --> 00:15:04,000
 cause either a segfault or undefined value will be caught. So in summary, there's two

194
00:15:04,000 --> 00:15:10,400
 answers to this question. One is it's actually okay to take certain risks for certain targets.

195
00:15:10,400 --> 00:15:17,600
 And number two is I actually do want to bring this list down roughly to go or Rust safety level,

196
00:15:17,600 --> 00:15:23,680
 but we're not at 1.0 yet. So check back then and you can evaluate the safety of Zig at that point

197
00:15:23,680 --> 00:15:28,800
 in time. Yeah, that was actually what I wanted to ask next was like, you know, Rust has this

198
00:15:28,800 --> 00:15:33,520
 thing that since once point, oh, they always want to remain backwards compatible. Is it going to

199
00:15:33,520 --> 00:15:38,960
 have the same kind of guarantees? 1.0. Yeah, yeah, our goal is to pretty much just never touch the

200
00:15:38,960 --> 00:15:45,520
 language again after 1.0. Now that could be a catalyst to have a never-reducing 1.0 kind of

201
00:15:45,520 --> 00:15:49,920
 stage is that the way you see it happening or you actually see a 1.0 happening sometime relatively

202
00:15:49,920 --> 00:15:55,600
 soon. Yeah, I mean, I see what you're saying, but it's offset by the fact that the language is just

203
00:15:55,600 --> 00:16:02,080
 so much simpler than other languages. So while yes, it's still a monumental amount of work, it is

204
00:16:02,080 --> 00:16:07,680
 lessened by the fact that the language doesn't require it's a much more finite set of language

205
00:16:07,680 --> 00:16:14,640
 features than other projects. So I do know that people are using Zig in production, right? And so

206
00:16:14,640 --> 00:16:21,280
 how are they handling like if there is changes to the language, right? Is there an upgrade guide or

207
00:16:21,280 --> 00:16:25,520
 is there like, I think in some languages, they have like a tool that which can scan the code and

208
00:16:25,520 --> 00:16:30,000
 change it for you. Do you have something like this? Yeah, we have both of these things. Yeah, so if

209
00:16:30,000 --> 00:16:35,040
 any listeners this podcast right now, I would encourage you to go to the Zig website and just

210
00:16:35,040 --> 00:16:42,160
 look at the latest release notes for, I don't know, 0.9.0 and then just scroll down in awe of two weeks

211
00:16:42,160 --> 00:16:51,600
 that I just spent typing away these notes very laboriously, but I just but also, yeah, we sometimes

212
00:16:51,600 --> 00:16:56,080
 are able to have the Zig Formatter tool, auto upgrade things. It doesn't have type awareness,

213
00:16:56,080 --> 00:17:00,160
 so it's limited in what it can do, but when it can, it's pretty ideal. Now, what kind of things does

214
00:17:00,160 --> 00:17:04,400
 it actually do for you? Does it rename variables? Does it just change the syntax or what kind of

215
00:17:04,400 --> 00:17:08,400
 things does it actually do for you? Yeah, I was thinking about the formatter. Yeah, the formatter

216
00:17:08,400 --> 00:17:12,800
 and if it does actually change and upgrade your code for the new version. Yeah, I mean, it's a

217
00:17:12,800 --> 00:17:18,640
 pretty standard formatter. If you've ever used Go format, it's pretty much the same. I don't have

218
00:17:18,640 --> 00:17:22,400
 any experience with Rust format. I don't know how that works, but yeah, there's no configuration

219
00:17:22,400 --> 00:17:27,280
 options. It's opinionated, so everyone has to use the same style. And then what we do is just,

220
00:17:27,280 --> 00:17:31,200
 I don't know, let's say that one of the built-in functions gets renamed, it's able to just do that

221
00:17:31,200 --> 00:17:35,520
 rename for you and forget what I was going to say, but it doesn't have access to type information.

222
00:17:35,520 --> 00:17:39,920
 So the only thing you can do is send tactical changes. So there must be like the AST or something

223
00:17:39,920 --> 00:17:45,600
 like that, that kind of level. The main purpose of the formatter is just reformat the code. So

224
00:17:45,600 --> 00:17:50,320
 double-purposing it, do language upgrades sometimes works and sometimes does not work.

225
00:17:50,320 --> 00:17:54,800
 Please don't tell me that you also add underscores between the thousands place

226
00:17:54,800 --> 00:18:02,160
 and numbers, do you? Automatically? Yeah. No. I actually rejected that language proposal for a

227
00:18:02,160 --> 00:18:07,360
 while, but then I caved. Okay. To even support underscores and integer literals at all.

228
00:18:07,360 --> 00:18:12,240
 I totally am okay with that, but what I really hate is like a formatter that I use for Elixir.

229
00:18:12,240 --> 00:18:15,440
 They only, like you said, it's also opinionated, right? It follows very much the same ideas,

230
00:18:15,440 --> 00:18:20,720
 but for some reason they decided to automatically add the thousands separated underscores in between

231
00:18:20,720 --> 00:18:25,600
 and it drives me nuts because sometimes I have an integer that actually is a reference for something

232
00:18:25,600 --> 00:18:30,400
 from a client and it's like, okay, you know, I format all the time because I want my code to look

233
00:18:30,400 --> 00:18:35,280
 okay, but then I do command F or command shift F try to find that integer and I cannot find it

234
00:18:35,280 --> 00:18:40,720
 because it's going to be a formatter. Yeah. That's annoying. Also, that's a regional thing. So if you

235
00:18:40,720 --> 00:18:44,880
 put something in the thousands place, then that might actually just not be what a different culture

236
00:18:44,880 --> 00:18:50,400
 does. So it's just kind of weird. You know, it's opinionated. So what can I do? But that's my biggest

237
00:18:50,400 --> 00:18:53,280
 gripe. Otherwise, everything else is fine. I don't really care too much about other stuff.

238
00:18:53,280 --> 00:18:56,880
 It's kind of nice having a formatter opinion because you don't need to fight about code style.

239
00:18:56,880 --> 00:19:01,280
 Just run the formatter and go. Yeah, that's the idea. Yeah. We do have a kind of a nice thing

240
00:19:01,280 --> 00:19:06,160
 where if you want to put, so Zig just supports commas at the end of lists and stuff. I think

241
00:19:06,160 --> 00:19:11,680
 Rust does too, if I remember correctly. So like enums or structs or even function calls. Anyway,

242
00:19:11,680 --> 00:19:18,800
 point being if you just trail your an extra comma, the formatter will actually notice that and then

243
00:19:18,800 --> 00:19:22,000
 say, I see that you want it to put everything on a separate line. I'll just go ahead and do that

244
00:19:22,000 --> 00:19:26,000
 for you. And then likewise, if you delete the trailing comma, it'll say, Oh, okay, I guess you

245
00:19:26,000 --> 00:19:29,600
 want me to put it on one line. Okay, I'll go ahead and do that. You kind of communicate to the

246
00:19:29,600 --> 00:19:34,000
 formatter based on the existence of the trailing comma or not. I kind of really liked that because

247
00:19:34,000 --> 00:19:38,320
 like it does affect me because I don't want to like add in a bunch of artificial changes. It's

248
00:19:38,320 --> 00:19:41,920
 like, okay, I just want to add an extra item to the end of this list. Well, that means I have to

249
00:19:41,920 --> 00:19:46,720
 actually change two lines, which is just yeah, yeah, that's a good reason for things on different

250
00:19:46,720 --> 00:19:50,720
 lines. Yeah, kind of reminds me of when I worked in the bank before and like the most productive

251
00:19:50,720 --> 00:19:55,840
 person according to get changes was actually the configuration guy who would just manage the ammo

252
00:19:55,840 --> 00:20:00,400
 files all day, because you just be adding and removing stuff. And then eventually decided to

253
00:20:00,400 --> 00:20:04,480
 start gamifying his work, because you know, oh, you just add a space remove a space,

254
00:20:04,480 --> 00:20:08,480
 commit it. And then at the end of the day, look how many lines of code you changed this month,

255
00:20:08,480 --> 00:20:11,920
 you tripled your output, what's going on, you're doing a good job, here's a race.

256
00:20:11,920 --> 00:20:16,240
 Oh, wow, they wait. So their performance was actually being measured by lines of code.

257
00:20:16,240 --> 00:20:20,800
 So when I was working in the bank, we were working on, you know, cost allocations, right?

258
00:20:20,800 --> 00:20:24,880
 Because here's the thing is that IT people, how do you judge the value of a person?

259
00:20:24,880 --> 00:20:29,760
 If you sit down with somebody, you're assigned to help them optimize their work day, right? Usually,

260
00:20:29,760 --> 00:20:34,320
 it means writing a piece of software. But let's say that you, you know, did all your checking,

261
00:20:34,320 --> 00:20:37,440
 and all of a sudden at the end of the day, you say, actually, if you just remove these pieces and

262
00:20:37,440 --> 00:20:42,400
 then connect this stuff and whatever, you can actually save yourself three hours or whatever,

263
00:20:42,400 --> 00:20:46,080
 right? Just give me an idea. There's times when you're a developer, I'm sure you know this,

264
00:20:46,080 --> 00:20:50,880
 where you can find ways to improve without actually writing a piece of code.

265
00:20:50,880 --> 00:20:56,240
 To improve your own skill or to improve some of the company's infrastructure or what?

266
00:20:56,880 --> 00:21:01,760
 You can improve the company's way of doing things, right? So a lot of times people say,

267
00:21:01,760 --> 00:21:05,600
 okay, I want to make my process better. Let me just get the developer to write

268
00:21:05,600 --> 00:21:11,360
 software for me, so I can just drag and drop and go all my day. It's like, well, wait a minute,

269
00:21:11,360 --> 00:21:16,000
 this car is coming from Dropbox, and you want to just email to this person. So why not just have

270
00:21:16,000 --> 00:21:20,480
 Dropbox automatically send the invite and share the link, and then you're ready to go. Oh, I didn't

271
00:21:20,480 --> 00:21:24,640
 know we could do that. Okay. Honestly, like you can find simple things without writing a piece of

272
00:21:24,640 --> 00:21:30,640
 code or a lot of code. And then how do you measure that? Yeah, you need a human actually using

273
00:21:30,640 --> 00:21:35,040
 critical thinking in order to measure that. Exactly. And also like, I mean, everything too is,

274
00:21:35,040 --> 00:21:39,120
 if you're pair programming, which is what we did a lot also, how do you measure that? Because if

275
00:21:39,120 --> 00:21:43,360
 I'm telling you what to write and you're just typing away, technically, I think the guy who's got

276
00:21:43,360 --> 00:21:47,680
 the idea and it works very well is probably contributing more, right? I mean, you're just typing out the

277
00:21:47,680 --> 00:21:52,160
 stuff. We can discuss that one, but that's my thought on it. So now it's the guy who wrote

278
00:21:52,160 --> 00:21:56,800
 everything who gets all the credit. So like, yeah, it's interesting. Because the thing is,

279
00:21:56,800 --> 00:22:00,960
 IT salaries are quite high in relative comparison to like a regular clerk. So how do you judge the

280
00:22:00,960 --> 00:22:06,400
 value? Yeah, well, I will say one thing. I think that how you judge the value is different,

281
00:22:06,400 --> 00:22:11,280
 depending on if you're a large company or a small one. In any case, they had basically had four

282
00:22:11,280 --> 00:22:16,880
 buckets, right? Equipment, people's salaries, IT equipment, because you know, servers and stuff

283
00:22:16,880 --> 00:22:22,960
 are not cheap. And IT salaries, because IT salaries are, you know, rather high in comparison for the

284
00:22:22,960 --> 00:22:26,720
 same people. And over here, they just keep trying to get more developed. Anyways, I'm talking too

285
00:22:26,720 --> 00:22:30,480
 much about my own stuff, right? But in general, right? I forgot why we got onto this topic. But

286
00:22:30,480 --> 00:22:36,080
 yeah, I mean, judging value is really difficult. I completely agree that judging value is of human

287
00:22:36,080 --> 00:22:41,600
 contributions is very difficult. Okay, so where are we going with this? Yeah, so maybe this is a

288
00:22:41,600 --> 00:22:45,440
 good time since we both kind of, at least for me, I lost my way. What else going with everything?

289
00:22:45,440 --> 00:22:48,880
 Why don't we talk more about your past, right? Because I think it gives us a good idea about,

290
00:22:48,880 --> 00:22:53,120
 you know, where you came from and why you decided to get onto this path. Now, you started off,

291
00:22:53,120 --> 00:22:57,200
 you're definitely doing higher level languages, right? At the beginning. Yeah, earlier on, yeah,

292
00:22:57,200 --> 00:23:07,360
 I started with Visual Basic 6, I did Pearl, Python, JavaScript, C++, Java, C, then Rust,

293
00:23:07,360 --> 00:23:12,560
 then I made Zig. Now, VB is kind of known. They're not a very popular language, right? At the time,

294
00:23:12,560 --> 00:23:16,720
 though, I guess I worked out for you now. I mean, as for a 12 year old, it was awesome.

295
00:23:16,720 --> 00:23:19,280
 Okay, fine. 12 years old. Okay, I can't judge a 12 year old, I guess.

296
00:23:19,280 --> 00:23:24,640
 Yeah, I just know that in the programming world that people would say, oh, VB is not good,

297
00:23:24,640 --> 00:23:29,040
 et cetera. I never really got into VB. I never really did basic. I think I did basic once

298
00:23:29,040 --> 00:23:34,880
 when I was playing around with robots in university. We had a class called Build Your Own Robot Army,

299
00:23:34,880 --> 00:23:39,840
 and we're actually able to write basic for something. Wait, you're doing some kind of like software,

300
00:23:39,840 --> 00:23:43,840
 like bots fighting each other? Well, the name of the course was How to Build Your Own

301
00:23:43,840 --> 00:23:48,960
 Robot, Evil Robot Army, was actually the name of the course. So, oh, wow. So for my school, we have,

302
00:23:48,960 --> 00:23:53,520
 it's a trimester school. So you got your spring, you got your fall, but in the middle of January,

303
00:23:53,520 --> 00:23:57,440
 it's something called J-term. So you can take off, which doesn't really make sense because in

304
00:23:57,440 --> 00:24:02,240
 general, you pay for that, right? You can take a trip, which I did also one time I went to China,

305
00:24:02,240 --> 00:24:06,320
 and that's actually part of a course, right? And then the other one is that you could take a course.

306
00:24:06,320 --> 00:24:10,320
 You sit in class about four hours a day, every day for like the whole month,

307
00:24:10,320 --> 00:24:13,920
 and you can get the course done. So it could be maybe you get a missing credit that you need.

308
00:24:13,920 --> 00:24:18,240
 Some courses are only taught in J-term, and sometimes there's courses like that, which are kind of like

309
00:24:18,240 --> 00:24:22,960
 fun courses. You get the credit and everything, but they're more fun, right? Building robots

310
00:24:22,960 --> 00:24:27,440
 every day for four hours is kind of fun. That sounds great. Was there hardware involved?

311
00:24:27,440 --> 00:24:30,320
 It was all hardware in the beginning, and then at the very end, we had a special chip that we

312
00:24:30,320 --> 00:24:35,920
 could actually program. Wow. Wait, you said a robot army? So were you making them like swarm or

313
00:24:35,920 --> 00:24:41,680
 something? For some names may be confusing, right? So the name of the course was How to Build Your Own

314
00:24:41,680 --> 00:24:46,640
 Evil Robot Army, but we were not making swarms of armies of robots. I wish we were. That would be

315
00:24:46,640 --> 00:24:50,960
 fun, but maybe not enough time and four hours a day for only the month of January. Okay, so you

316
00:24:50,960 --> 00:24:55,760
 made one robot. I made several different robots, and I think they would basically build a robot,

317
00:24:55,760 --> 00:25:00,400
 and then disassemble it, and then make a new one. That sounds really cool. It's like over 10 years

318
00:25:00,400 --> 00:25:04,640
 ago, right? I'm getting old over here. I can't remember exactly what we did. Yeah, it was something

319
00:25:04,640 --> 00:25:09,760
 like that, right? So it sounds cool, but the details of them have done anything to take that class when

320
00:25:09,760 --> 00:25:13,680
 I was young. Yeah, I'm not saying it wasn't fun, but why are we talking about this again? I totally

321
00:25:13,680 --> 00:25:18,320
 forgot. Oh, you said that you heard the Visual Basic was a bad language. Oh, yeah, yeah, sorry.

322
00:25:18,320 --> 00:25:22,320
 Yeah, so yeah, visual. And then I remember that for programming, that was actually basic, right?

323
00:25:22,320 --> 00:25:27,200
 And that was my first and only time I've actually using basic. So quite basic in terms, right? But

324
00:25:27,200 --> 00:25:31,280
 I do know that basic was a huge thing, right? I mean, Microsoft wrote basic, I think for the first

325
00:25:31,280 --> 00:25:36,800
 Apple or the second Apple II, I think it was. They sold basic to one of those original switches

326
00:25:36,800 --> 00:25:42,000
 with that computer. What is that? The altar, I think it was. That's before my time. You're teaching

327
00:25:42,000 --> 00:25:47,280
 me history at this point. I would thought you would know this stuff. No, I'm only 33. We're about

328
00:25:47,280 --> 00:25:52,160
 the same age. I'm just the promise that I love you too much. That's the issue. But any case, right?

329
00:25:52,160 --> 00:25:55,760
 So you did VB at 12 years old. What kind of stuff are you doing with VB? And by the way,

330
00:25:55,760 --> 00:26:00,800
 you know, as much as I could, but also not that much. My parents only let me use the computer,

331
00:26:00,800 --> 00:26:07,200
 actually any electronics for one hour a day, which as you know, programming is a pretty time-consuming

332
00:26:07,200 --> 00:26:12,560
 hobby. So it was pretty challenging. I read a lot of books and then like planned out everything I

333
00:26:12,560 --> 00:26:16,720
 would do on paper so that when I like got my hour, I would just like put it in the computer as fast

334
00:26:16,720 --> 00:26:21,600
 as possible. But I forget what your actual question was. No, I was just curious about what you were

335
00:26:21,600 --> 00:26:25,440
 actually kind of programming with, but I can understand like, so was it like actually kind of like you

336
00:26:25,440 --> 00:26:30,000
 were acting like programmers were a long time ago where it's like you had to wait in line to drop

337
00:26:30,000 --> 00:26:33,520
 your card into the machine? Yeah, kind of. Yeah. But even though it was sitting right there,

338
00:26:33,520 --> 00:26:38,720
 I just wasn't allowed to use it most of the time. Yeah, I would just try anything. So I,

339
00:26:38,720 --> 00:26:43,440
 what did I do? I tried to make like a WYSIWYG website editor, which did not work very well.

340
00:26:43,440 --> 00:26:48,240
 I made a lot of games without learning anything about game design yet. So I have a lot of like

341
00:26:48,240 --> 00:26:53,120
 bad game ideas that I did. I did make one kind of fun thing, which was like a macro recorder.

342
00:26:53,120 --> 00:26:58,880
 So record your keyboard and mouse movements and then replay them. And I actually managed to prank

343
00:26:58,880 --> 00:27:02,240
 my high school teacher with that one years later. So that was pretty funny.

344
00:27:02,240 --> 00:27:08,320
 Yeah, I mean, to me just experimenting with like the Windows APIs and just learning how to code was

345
00:27:08,320 --> 00:27:13,840
 really fun. So from there on, you're like, wow, I really want to continue working on with computers

346
00:27:13,840 --> 00:27:17,680
 or continue programming or just like this is just fun to play around with at the beginning.

347
00:27:17,680 --> 00:27:21,760
 Well, I had a lot of hobbies. I also got into animation. I was really involved in albino

348
00:27:21,760 --> 00:27:26,080
 black sheep back in the day. I got on day using FL Studio to try to make music,

349
00:27:26,640 --> 00:27:31,680
 I tried learning piano. What else? I wanted to like build hardware, but I just didn't

350
00:27:31,680 --> 00:27:35,600
 wasn't smart enough to figure out how to get started and none of the adults around me figured

351
00:27:35,600 --> 00:27:39,280
 out that they needed to just buy me a book on hardware. And I didn't know that that's what I needed.

352
00:27:39,280 --> 00:27:44,720
 So that never happened. But anyway, there's alternate universes in which I have many different

353
00:27:44,720 --> 00:27:48,640
 specialties. But yeah, in this one, I landed on software. Yeah, I think I hear that in some

354
00:27:48,640 --> 00:27:52,160
 podcasts. Like if you want to develop, or what would you be? And you hear some interesting answers,

355
00:27:52,160 --> 00:27:57,280
 some of them actually like a farmer or artist sounds like you have many different possibilities.

356
00:27:57,280 --> 00:28:02,640
 Yeah, I like just love to create, you know. So what happened after VB, right? So after some time

357
00:28:02,640 --> 00:28:07,760
 you're playing with VB and then like when do you think you really started to realize that, hey,

358
00:28:07,760 --> 00:28:12,080
 this is actually what I want to do. I want to program. I love this. I think it's just the hobby

359
00:28:12,080 --> 00:28:18,560
 that I kept coming back to. So, you know, when it came time to pick a major in college, it was

360
00:28:18,560 --> 00:28:22,800
 an obvious choice to go with computer science. That's basically it. Not much more to say there.

361
00:28:22,800 --> 00:28:30,400
 What was your first job actually out of university? Amazon.com. It was the worst job I ever had. I

362
00:28:30,400 --> 00:28:35,520
 started looking for a different job after just three months after moving across the country

363
00:28:35,520 --> 00:28:40,400
 for this dream job out of college. Wow, I'm sure a lot of people at home are thinking like,

364
00:28:40,400 --> 00:28:43,600
 wow, that doesn't make any sense because we all want to work for Amazon because they're one of

365
00:28:43,600 --> 00:28:49,760
 the biggest rust hires or kind of rust places out there, I would say. Amazon is really good at treating

366
00:28:49,760 --> 00:28:55,520
 programmers like cogs in a machine. So you're not going to have a good experience at Amazon,

367
00:28:55,520 --> 00:29:02,240
 but you will be exploited for labor effectively. And Jeff Bezos will make a lot of money and you'll

368
00:29:02,240 --> 00:29:06,640
 quit soon after like a year because of the high attrition rate. But it doesn't matter because

369
00:29:06,640 --> 00:29:10,880
 someone will replace you and they've designed a system where everyone is replaceable. It sucks.

370
00:29:10,880 --> 00:29:15,200
 Don't work there. But I mean, as a business owner, right? For me, I can give the idea that, hey,

371
00:29:15,200 --> 00:29:19,040
 I do want people to be replaceable because you don't want to rely your whole business on somebody.

372
00:29:19,040 --> 00:29:22,560
 Yeah, I mean, if your only goal is to make money. Yeah. Well, I mean, not about that, right? You

373
00:29:22,560 --> 00:29:26,800
 need to be able to handle people leaving, especially for tech, right? People always change jobs. That's

374
00:29:26,800 --> 00:29:31,360
 kind of, oh, you're still at the same job for what? Two years? That doesn't make sense. Why don't you

375
00:29:31,360 --> 00:29:36,960
 change jobs? I mean, okay. I see what you're saying, right? Your point is that the business needs to

376
00:29:36,960 --> 00:29:42,640
 survive, you know, like regular, healthy employee turnover, which is a fair point. But you also need

377
00:29:42,640 --> 00:29:48,160
 to, from another angle, people need to have a place where they can fit in and they can feel

378
00:29:48,160 --> 00:29:53,360
 like they're doing important work. And if you are just completely replaceable, you're not doing

379
00:29:53,360 --> 00:29:59,680
 important work. You don't have a place to fit in. And it's not a happy, fulfilling job. Like,

380
00:29:59,680 --> 00:30:05,200
 at some point, in order to just have a healthy working environment, you have to give people

381
00:30:05,200 --> 00:30:10,960
 some autonomy and some responsibility and some meaning. So what you think would be a healthy

382
00:30:10,960 --> 00:30:17,120
 situation between an employer and employee where an employer, their employee is replaceable, but

383
00:30:17,120 --> 00:30:21,680
 still not quite replaceable. I'm not sure how you'd even define that in a way that

384
00:30:21,680 --> 00:30:26,080
 is easily understandable from the employer side. I mean, I think this is a pretty natural

385
00:30:26,080 --> 00:30:32,160
 phenomenon. So if you take any average company who's hiring, they're spending company resources

386
00:30:32,160 --> 00:30:37,920
 trying to find a new candidate, right? And when that person, if they find someone qualified and

387
00:30:37,920 --> 00:30:42,000
 they join, they have then have to train them, have to catch them to speed and that person has

388
00:30:42,000 --> 00:30:47,840
 to take time to get familiar with not only like the code, but also the people and how to work well

389
00:30:47,840 --> 00:30:51,840
 with other people and all the nuances of everyone's personalities and that sort of thing. What is

390
00:30:51,840 --> 00:30:56,400
 that called? Institutional knowledge? I forget what it's called. But I mean, this, we already have

391
00:30:56,400 --> 00:31:01,120
 words for these things, right? This is already just an observable phenomenon in the real world.

392
00:31:01,120 --> 00:31:06,640
 And companies generally don't want people to quit too much or too fast. They don't want

393
00:31:06,640 --> 00:31:11,840
 to hire attrition rate because it's inefficient to be hiring all the time. It's more efficient for

394
00:31:11,840 --> 00:31:16,640
 the existing people to invest in your existing employees and have them do good work and better

395
00:31:16,640 --> 00:31:20,560
 and better work as they get more and more familiar and power and ability within the

396
00:31:20,560 --> 00:31:25,600
 organization to accomplish the goals that help the organization. So I don't think I'm saying

397
00:31:25,600 --> 00:31:29,840
 something outrageous. I mean, I think I'm just describing an existing phenomenon and I'm just

398
00:31:29,840 --> 00:31:36,720
 pointing out that Amazon has optimized for a high attrition rate. They said, fine, if you all are

399
00:31:36,720 --> 00:31:41,920
 going to quit because it sucks to work here, we're just optimizing our business around that being a

400
00:31:41,920 --> 00:31:48,000
 premise. So screw you, we don't care if you're happy or not. Yeah, I mean, I understand from both

401
00:31:48,000 --> 00:31:52,400
 sides, right? It's like you want to feel valued. You want to feel like you fit in. You want to have

402
00:31:52,400 --> 00:31:57,840
 all this kind of stuff at the same time. It's like, well, people naturally will go wherever they want

403
00:31:57,840 --> 00:32:02,160
 to go. I mean, it depends on the motivation, right? Some people are motivated by money. There's not

404
00:32:02,160 --> 00:32:07,680
 much you can do about that. Some people are motivated by whatever, right? Whatever the motivation is.

405
00:32:07,680 --> 00:32:12,160
 I mean, there's been some people who, like I recently hired some guy, he said, you know, I really want

406
00:32:12,160 --> 00:32:15,840
 to work here. I just didn't have anything for him to work at the time. So I just say, hey, you have

407
00:32:15,840 --> 00:32:20,000
 to wait for a moment. And he was basically like, listen, I really want to work with you. I got some

408
00:32:20,000 --> 00:32:24,560
 other offers, but you know, I'm willing to wait, but I can't wait for so long, right? Because he's got

409
00:32:24,560 --> 00:32:29,200
 things to handle, right? So I mean, there's this kind of thing, right? But there's other people who

410
00:32:29,200 --> 00:32:33,520
 just like, oh, I got this, you know, I got this offer today. You know, what can you do about it?

411
00:32:33,520 --> 00:32:37,600
 Can you match it? It's like, well, you know, how do I handle that? Right? Like, why you actually

412
00:32:37,600 --> 00:32:41,280
 hear? Do you actually want to be here? So that's, I mean, there's so many ways to look at it,

413
00:32:41,280 --> 00:32:44,800
 right? But it's good to hear your side. And I think you have some points, obviously.

414
00:32:44,800 --> 00:32:48,480
 Yeah. I mean, you're talking to someone who's running a nonprofit. So we obviously have different

415
00:32:48,480 --> 00:32:53,840
 motives than a company whose explicit goals are to give a return on an investment.

416
00:32:54,320 --> 00:32:58,160
 Well, yeah, I mean, but you're also for a nonprofit, you also have to explain, okay, I paid this money

417
00:32:58,160 --> 00:33:02,960
 for this person, then, you know, here's the output, right? As a nonprofit, I think you have to have

418
00:33:02,960 --> 00:33:06,400
 something like that. Every time you spend money, I think you have to kind of justify it.

419
00:33:06,400 --> 00:33:09,920
 Yeah, how it works for nonprofits is you have a mission statement, which is public. You can go

420
00:33:09,920 --> 00:33:15,760
 look at ours on our homepage. And your legal responsibility is to spend all of your funds,

421
00:33:15,760 --> 00:33:20,080
 all of them. So nobody gets any, you know, there's no leftover funds in a nonprofit. You have to spend

422
00:33:20,080 --> 00:33:25,520
 all your funds serving the mission statement. So when people donate money to us, then presumably

423
00:33:25,520 --> 00:33:30,080
 what they're thinking is that submission statement, I can stand behind. I trust the people involved

424
00:33:30,080 --> 00:33:33,760
 that they are going to use this money to work towards that mission statement.

425
00:33:33,760 --> 00:33:38,000
 You know what? I think this is a perfect segue. I do want to go back to your career,

426
00:33:38,000 --> 00:33:41,760
 but I think this is a perfect segue since you're already thinking about the nonprofit. Rust has a

427
00:33:41,760 --> 00:33:47,040
 nonprofit, right? And I did interview Rebecca Rumbel, the episode's not out yet at the time of

428
00:33:47,040 --> 00:33:51,120
 this recording. But she did quite explain about what their mission statement is and what they're

429
00:33:51,120 --> 00:33:55,840
 doing. Can you talk a little bit more about what Zig mission statement is and what you guys are

430
00:33:55,840 --> 00:34:02,400
 doing as a nonprofit? Yeah, sure. Yeah. So we're 501c3, same as the Python software foundation.

431
00:34:02,400 --> 00:34:06,880
 And we actually published all of our financial transactions recently. So you can just go look

432
00:34:06,880 --> 00:34:12,800
 them up if you want every single one. And that's all open. And if you do that, what you'll find is

433
00:34:12,800 --> 00:34:18,880
 that almost all the money goes directly towards paying contributors who are working on the Zig

434
00:34:18,880 --> 00:34:24,320
 project. And that's something I'm really proud of. It's a really efficient organization. And it's

435
00:34:24,320 --> 00:34:29,200
 using the donations that we get to send that money directly to people who deserve it for working

436
00:34:29,200 --> 00:34:33,760
 on Zig. Now, to be clear, there are people, a lot of people who are doing really good work,

437
00:34:33,760 --> 00:34:38,400
 and they're not getting paid. They're just volunteering their time for the open source project. I would

438
00:34:38,400 --> 00:34:43,840
 love to pay them if we had more budget. But obviously we have to spend wisely and not too much.

439
00:34:43,840 --> 00:34:47,920
 Well, I mean, I guess you can kind of disclose like, what are some of the developers actually

440
00:34:47,920 --> 00:34:53,360
 getting paid? Are they can pay per patch or some kind of weekly thing or a stipend or what kind of

441
00:34:53,360 --> 00:34:59,040
 thing are usually they get paid with? Everyone has a pretty similar contract. It's 50 US dollars

442
00:34:59,040 --> 00:35:07,200
 per hour. They set whatever schedule they want. And the like what they're supposed to work on is

443
00:35:07,200 --> 00:35:12,640
 intentionally loose so that they can decide what they think is best for the Zig project.

444
00:35:12,640 --> 00:35:17,840
 And they get paid for their time. No questions asked. Because the thing that we were able to do

445
00:35:17,840 --> 00:35:23,680
 is pick people who already demonstrated intrinsic motivation to help the project. So for example,

446
00:35:23,680 --> 00:35:29,840
 one of our full-time hires is Jakub Kanka. He does awesome work. And the thing is he was already

447
00:35:29,840 --> 00:35:35,680
 volunteering his free time before he became a full-time hire. So it's kind of an advantage that we have

448
00:35:35,680 --> 00:35:41,840
 over startup because he demonstrated before he even saw any money that he cared about the project.

449
00:35:41,840 --> 00:35:48,000
 He wanted to succeed and that he could self-direct. So he doesn't require management. He just knows

450
00:35:48,000 --> 00:35:52,560
 what he wants to do and he has a good sense for it already. I don't need to tell him what to do.

451
00:35:52,560 --> 00:35:58,960
 And that contract is the same kind that everyone has. So everyone has a high level of trust and

452
00:35:58,960 --> 00:36:04,480
 they're paid by the hour. Their schedule is whatever they want it to be. And their tasks are self-directed.

453
00:36:04,480 --> 00:36:08,560
 That's not a huge amount of money. I mean, 50 bucks an hour, I think is not a lot in comparison to

454
00:36:08,560 --> 00:36:13,280
 other companies, especially like you said Amazon. So I mean, their motivation is probably not

455
00:36:13,280 --> 00:36:18,400
 money, obviously, right? My understanding is that, well, actually, I'm going to try and protect the

456
00:36:18,400 --> 00:36:23,840
 privacy of people. So I don't want to say too much. But my understanding that it's a comfortable

457
00:36:23,840 --> 00:36:30,320
 rate for those who are getting it and not the United States. And but also I do want to see that

458
00:36:30,320 --> 00:36:36,480
 higher. I do want to keep the rate competitive. And I think that as we move forward, I think that

459
00:36:36,480 --> 00:36:40,400
 we'll be able to take in more donations. And I think that we will be able to do that. I'm hoping

460
00:36:40,400 --> 00:36:44,960
 that that is the case within the next couple of years. I mean, the way I looked at it, right?

461
00:36:44,960 --> 00:36:48,960
 I think, yeah, you're right overseas. That could be quite a bit of money. But let's just say the

462
00:36:48,960 --> 00:36:54,640
 person is in like US or when he's kind of higher level European countries. Now, in that case, right,

463
00:36:54,640 --> 00:36:59,040
 you could say, okay, this is not really meant for them to necessarily live on, but it is meant for

464
00:36:59,040 --> 00:37:03,200
 them to say like, thank you. And then they could use that to say maybe get a new laptop that they

465
00:37:03,200 --> 00:37:08,160
 could use for contributing more. Like the new M1 ships are coming out. Those ones could help them

466
00:37:08,160 --> 00:37:12,400
 compile faster just to give an idea. These kinds of things, right? Well, like a thank you. And

467
00:37:12,400 --> 00:37:17,120
 here's some compensation. And then you could do with it, however you want. That's kind of the way

468
00:37:17,120 --> 00:37:22,640
 I look at it least. Yeah, we do have some part time contractors as well. And I think that I think

469
00:37:22,640 --> 00:37:28,160
 that it functions more like that for the part time contractors where they for whatever reason are

470
00:37:28,160 --> 00:37:32,720
 not able or willing to go full time on zig or we don't have the funds to pay them. But they're

471
00:37:32,720 --> 00:37:37,120
 able to do some side work, you know, part time and have a little bit of extra spending money at

472
00:37:37,120 --> 00:37:40,800
 the same rate. I mean, how do you guys decide what's going to be the next release? I mean,

473
00:37:40,800 --> 00:37:44,640
 if you said it sounds like people are really autonomous, but then there must be like a milestone

474
00:37:44,640 --> 00:37:48,080
 listen on GitHub and then all the issues attached to it and people just kind of pop it off the stack

475
00:37:48,080 --> 00:37:52,480
 whenever they feel like they want to work towards a new release. No. Yeah, so there's issues on there

476
00:37:52,480 --> 00:37:56,480
 and that's to help the issues are to help coordinate between people. So that's what those are there

477
00:37:56,480 --> 00:38:02,960
 for. I have a vision for what I think's important and I do my best to share and inspire, but I don't

478
00:38:02,960 --> 00:38:08,960
 dictate. I mean, I do review code and I do veto technical decisions and I do guide the technical

479
00:38:08,960 --> 00:38:15,840
 aspects of what is in scope of the project and what we're working towards. But I don't dictate the

480
00:38:15,840 --> 00:38:21,440
 tasks of what people decide they want to work on. So as an example, we know that we want to have

481
00:38:22,160 --> 00:38:28,480
 zig support for the major operating systems, Linux, Windows, Mac OS. We know that we want

482
00:38:28,480 --> 00:38:33,440
 the self hosted compiler to have its own linker for these three operating systems for various

483
00:38:33,440 --> 00:38:42,400
 reasons. Now, personally, my priorities would be like Windows x86, Linux x86, my priorities.

484
00:38:42,400 --> 00:38:47,440
 Eventually, I'd want to get Mac OS to, but for Jakub, Mac OS is really important to him.

485
00:38:47,440 --> 00:38:51,920
 And so that's what he's been focusing a lot of time on. And that's great. I'm happy about that.

486
00:38:51,920 --> 00:39:01,600
 And thanks to that, we've had zig has a cross compiling across linker for Mac OS that other

487
00:39:01,600 --> 00:39:06,880
 projects don't have access to. And if it were me, that's not what I would have prioritized my time

488
00:39:06,880 --> 00:39:12,160
 on. But that's great. That is something that we want that's in scope. And I'm thrilled that

489
00:39:12,160 --> 00:39:16,000
 that's something that Jakub is enthusiastic about. And then he's managed to push to this level of

490
00:39:16,000 --> 00:39:20,880
 quality. That's great. As somebody who's an avid Mac user, let me make it clear the main reason I

491
00:39:20,880 --> 00:39:25,600
 really want. Well, I like using Mac, the fits well for me. And it's like a kind of a major OS. But

492
00:39:25,600 --> 00:39:31,280
 everything too, that I have to say, like in general, like I build iOS apps for some clients. And I

493
00:39:31,280 --> 00:39:35,440
 just cannot do that. And otherwise, I mean, there are services I could use, but in general,

494
00:39:35,440 --> 00:39:39,920
 like run the app and develop on my machine directly. So I'm kind of limited to that.

495
00:39:39,920 --> 00:39:44,080
 So this makes sense for me to buy Mac. So for that, Mac OS support makes it more attractive for me,

496
00:39:44,080 --> 00:39:48,800
 at least. Yeah, that makes sense. The new M1s are pretty nice. It's pretty nice machines.

497
00:39:48,800 --> 00:39:52,400
 Yeah, we can go into M1 Ultra versus the other one. I've heard weird benchmarks for

498
00:39:52,400 --> 00:39:56,000
 their basically about the same. So you may be paying extra for something, which is not

499
00:39:56,000 --> 00:40:00,240
 great in terms of most tasks, maybe long running ones that might win out. But for some short-term

500
00:40:00,240 --> 00:40:05,680
 ones, it doesn't really make sense. But yeah, okay, I think that's a good kind of overview,

501
00:40:05,680 --> 00:40:09,760
 kind of going back. Okay, so you worked at Amazon. Obviously, I don't think you enjoyed your time

502
00:40:09,760 --> 00:40:15,840
 there as you made it indirectly clear. But then what you're doing, and then of course, I think

503
00:40:15,840 --> 00:40:20,400
 everybody knows that you were working at, what was that dating website? Okay, Cupid, right?

504
00:40:20,400 --> 00:40:24,080
 Before it was actually called, okay, Cupid, but before, okay, Cupid, you think you also had

505
00:40:24,080 --> 00:40:28,240
 some work you're doing too, right? Yeah, did a bunch of work at startups, worked at a couple

506
00:40:28,240 --> 00:40:34,320
 music startups, worked at Backtrace. I think that's it. And where was it that, because I know

507
00:40:34,320 --> 00:40:38,640
 before, okay, Cupid, you're working on Zig, right? Like, what was the motivating factor where you're

508
00:40:38,640 --> 00:40:42,640
 like, okay, I want to build this thing? Were you working with Sea of the Time and you're just so

509
00:40:42,640 --> 00:40:47,120
 frustrated by all these issues? Or was it just like you wanted to have fun? Or what was the

510
00:40:47,120 --> 00:40:53,920
 motivating factor? Yeah, I was working on a digital audio workstation in C++. And yeah,

511
00:40:53,920 --> 00:40:58,240
 running into a lot of issues. Actually, you know what, I tried a bunch of different languages

512
00:40:58,240 --> 00:41:02,960
 for that project. I tried Go and I realized that it was not powerful enough, not low level enough.

513
00:41:02,960 --> 00:41:08,800
 I tried, actually gave Rust a really solid shot for the digital audio workstation. And this was

514
00:41:08,800 --> 00:41:13,280
 pretty 1.0. So with the caveat that I'm sure the experience is a lot smoother these days,

515
00:41:13,280 --> 00:41:19,920
 back then I really struggled to be productive. The learning curve was very steep for me.

516
00:41:19,920 --> 00:41:25,360
 I spent a month just trying to do some basic kind of like font rendering stuff for a GUI.

517
00:41:25,360 --> 00:41:31,840
 And I mean, I got it working, but then I found that the way I built it didn't abstract properly.

518
00:41:31,840 --> 00:41:36,000
 So it felt like even though I'd finally got it working, I couldn't even use my code.

519
00:41:36,000 --> 00:41:40,720
 Didn't know how to build on it. Anyway, long story short, I had to be honest with myself.

520
00:41:40,720 --> 00:41:47,520
 And I realized I am not productive with Rust. I am not making enough progress at a speed that

521
00:41:47,520 --> 00:41:53,440
 is satisfactory to myself. So at that point, I switched to C++. And despite all of its flaws,

522
00:41:53,440 --> 00:41:58,480
 I did experience a lot more productivity. And I was able to get a lot further in a shorter

523
00:41:58,480 --> 00:42:04,080
 amount of time on the project with C++. But still, I mean, none of these things are

524
00:42:04,720 --> 00:42:10,480
 were ideal for me. So at that point, I'd ping-ponged back for a bunch of different languages.

525
00:42:10,480 --> 00:42:17,600
 And having had just compared all of them for this one use case, I had a pretty clear idea of

526
00:42:17,600 --> 00:42:22,480
 what I wanted. And that's when I started Zig. How did you kind of prep yourself for writing a

527
00:42:22,480 --> 00:42:28,880
 language? And that's a huge endeavor now. Yep. Step one is just get in it, get touch, read me,

528
00:42:28,880 --> 00:42:34,480
 commit, then make a source directory. I mean, I'm just being facetious, but point being

529
00:42:34,800 --> 00:42:40,960
 you just start, just go. The first thing I did was I tried to write a parser generator.

530
00:42:40,960 --> 00:42:45,840
 Instead of using a parser generator, tried to make one. And a few weeks, I just gave up because

531
00:42:45,840 --> 00:42:50,080
 that's way too hard. I was not able to do that. And I just started doing recursive descent. And

532
00:42:50,080 --> 00:42:56,080
 that was way easier. It's just the first problem I ran into. So I don't know, you just start and

533
00:42:56,080 --> 00:42:59,760
 then you run into problem after problem after problem. And if you don't give up, then eventually

534
00:42:59,760 --> 00:43:03,760
 you have something. I think my biggest question is like, if you're coming from higher level

535
00:43:03,760 --> 00:43:09,040
 languages, even if you did see in C++, like designing the language, I mean, there's a lot of things

536
00:43:09,040 --> 00:43:11,840
 you need to know, a lot of things you need to consider, like, how does this stuff actually work?

537
00:43:11,840 --> 00:43:15,520
 Like, if you want to know how to get to the bytecode, then LLVM, you could just kick out

538
00:43:15,520 --> 00:43:19,120
 finally machine code at the end, right? Or binary code, whatever you want to call it,

539
00:43:19,120 --> 00:43:23,440
 the final end product, right? But Zig is not using LLVM, right? Do you think something else or

540
00:43:23,440 --> 00:43:27,280
 how does it actually work? You're characterizing me as coming from higher level languages, but I

541
00:43:27,280 --> 00:43:32,640
 think that might not be accurate because, I mean, I had been using C++ even professionally for a

542
00:43:32,640 --> 00:43:39,120
 while, and I had done a project where I recompile, aesthetically recompiled Nintendo games

543
00:43:39,120 --> 00:43:45,440
 to native executables using LLVM. So, I mean, I was pretty familiar with a bunch of low level

544
00:43:45,440 --> 00:43:50,160
 stuff. And for that project, I wrote a disassembler and assembler. I had to learn about JIT, I had to

545
00:43:50,160 --> 00:43:54,960
 learn about a lot of stuff. So, I mean, I definitely was familiar with a lot of these topics. It wasn't

546
00:43:54,960 --> 00:43:59,840
 just me going from high level languages and then making a low level language. Also, the

547
00:43:59,840 --> 00:44:04,800
 to answer your question, the Zid compiler that everyone is using today is using LLVM.

548
00:44:04,800 --> 00:44:12,240
 And the self-hosted compiler that is imminently shipping, it has a LLVM backend. It also has a

549
00:44:12,240 --> 00:44:17,680
 C backend, a native x86 backend, a native ARM backend, a native WebAssembly backend,

550
00:44:17,680 --> 00:44:22,960
 even a SpearV backend. These are all various levels of completion. The LLVM backend is the most

551
00:44:22,960 --> 00:44:28,080
 complete. Oh, now I feel I've been embarrassed I didn't know so much about these details, but

552
00:44:28,080 --> 00:44:32,000
 yeah, okay. Like I said, no worries. Yeah, no shade on you. I just wanted to make sure the

553
00:44:32,000 --> 00:44:36,800
 listeners didn't think that I was a green language user when I started making Zig.

554
00:44:36,800 --> 00:44:40,960
 No, and I think that's fine, right? I mean, nobody, like, when you started those pieces,

555
00:44:40,960 --> 00:44:43,920
 right, you have pretty much no idea like, how do you actually learn them? Did you actually crack

556
00:44:43,920 --> 00:44:48,800
 open a book and start reading or did you reading stuff online or I mean, I have no idea what like

557
00:44:48,800 --> 00:44:51,360
 where you even would start to learn these kind of pieces and why?

558
00:44:51,360 --> 00:44:54,160
 I just spent a lot of time practicing.

559
00:44:54,160 --> 00:44:59,280
 And when I were chunks, right? Yeah. Well, at some point I moved out of my parents' house and I spent

560
00:44:59,280 --> 00:45:03,440
 a lot more than one hour a day on the computer. Wait, wait, wait, was it similar to like, once you

561
00:45:03,440 --> 00:45:06,400
 move out of your parents' house, you start eating ice cream all the time because there was nobody

562
00:45:06,400 --> 00:45:11,280
 telling you not to? Yep, that's exactly right. I'm doing the kid who lives under our rock and

563
00:45:11,280 --> 00:45:17,280
 then drinks too much alcohol in college except for computers. Karamazoo, that Amash TV show,

564
00:45:17,280 --> 00:45:20,880
 I think I've seen the bits and pieces on TLC. Have you seen that one? No, I don't know what you're

565
00:45:20,880 --> 00:45:25,360
 talking about. There's a show, I really haven't watched the whole episode but I've seen bits and

566
00:45:25,360 --> 00:45:29,280
 pieces where these people who are Amash, I think they have like, when you're getting to just about

567
00:45:29,280 --> 00:45:33,520
 being an adult in their community, you're allowed to kind of go out and have like a crazy wild weekend

568
00:45:33,520 --> 00:45:41,280
 or day and they get to experience, you know, electricity, drinking, bikinis, you know, stuff

569
00:45:41,280 --> 00:45:44,720
 that that's just not in the community and then they kind of come back and say, yeah, I think I

570
00:45:44,720 --> 00:45:47,920
 want to stay, you know, here or they say, you know what, the world outside a little bit more

571
00:45:47,920 --> 00:45:53,200
 interesting. That's what I understand, right? So don't take as my word. It's like the Amash purge.

572
00:45:53,200 --> 00:45:59,280
 Yeah, something like that. So they actually videotape it on TLC and you're like, wow, these guys are

573
00:45:59,280 --> 00:46:03,200
 wild but you know, like we've talked about before, hey, you know, if you never had a chance to go out

574
00:46:03,200 --> 00:46:06,880
 and do things when you do get the opportunity, you may not be able to control yourself.

575
00:46:06,880 --> 00:46:13,200
 Yeah, I don't know if that's real or not. Maybe it is but here we are. Yeah, so okay, it's good to

576
00:46:13,200 --> 00:46:17,760
 hear it because I didn't know if you actually use LLVM or not because it wasn't clear for me.

577
00:46:17,760 --> 00:46:21,600
 I know Russ definitely makes it clear and also I'm more in the community of Russ, I am with

578
00:46:21,600 --> 00:46:24,960
 Zig. So it wasn't super clear to me that you guys are using that because I was like, hmm,

579
00:46:24,960 --> 00:46:28,640
 how can you guys optimize and be faster than C but that's must be

580
00:46:28,640 --> 00:46:33,120
 but we're talking about runtime performance, right? So that's related to the fact that it's

581
00:46:33,120 --> 00:46:40,800
 using LLVM. Here's what I'll say. It's hard to beat LLVM's performance at runtime and it's hard

582
00:46:40,800 --> 00:46:46,880
 not to beat LLVM's performance at compile time. Oh, okay, I see. Yeah, yeah. Well, that's one of

583
00:46:46,880 --> 00:46:49,920
 the things that Russ is famous for now. So because you guys are using LLVM,

584
00:46:49,920 --> 00:46:55,040
 is there also a slow compilation problem too because that's kind of also what Russ is known for

585
00:46:55,040 --> 00:46:59,840
 but that may also be his borrow checker. If you look at the timings of a Zig compilation today,

586
00:46:59,840 --> 00:47:07,040
 it's 90% waiting for LLVM but that's the thing though, so this new self-hosted compiler that we're

587
00:47:07,040 --> 00:47:12,560
 shipping soon, it's going to ship with the LLVM backend. That's the first step but we have the

588
00:47:12,560 --> 00:47:19,920
 x86 backend, what is it? 60% done or something like that. So when this one gets done, that's no LLVM.

589
00:47:19,920 --> 00:47:26,640
 So what we're going to have is we're basically following in these footsteps, which is you have a

590
00:47:26,640 --> 00:47:33,360
 pure self-hosted compiler for debug builds and the purpose of those is to go fast, compile fast,

591
00:47:33,360 --> 00:47:38,640
 so you can have that quick iteration time, you know, edit compile test cycle. And then when you

592
00:47:38,640 --> 00:47:41,920
 want to create your production build, then you're going to do a release build and you're going to

593
00:47:41,920 --> 00:47:46,640
 use the LLVM backend for that. Take a little bit longer, get better results.

594
00:47:46,640 --> 00:47:52,480
 Now I'm kind of curious, so why did you take LLVM over, say GCC? I mean, I always heard that GCC

595
00:47:52,480 --> 00:47:57,280
 is really good but I think there's some issues with it. When I hear the difference between LLVM

596
00:47:57,280 --> 00:48:02,240
 and GCC is that GCC is like all optimizations built into it because of history and LLVM is

597
00:48:02,240 --> 00:48:06,800
 kind of catching up. And they're kind of neck and neck, there are some cases, one of them wins,

598
00:48:06,800 --> 00:48:13,280
 some cases, the other one wins. The two reasons are license, LLVM has a liberal license, GCC has

599
00:48:13,280 --> 00:48:20,480
 GPL, Zeg is MIT, so GCC license is just not compatible. And the other one is just older and

600
00:48:20,480 --> 00:48:26,960
 craftier, so I mean, LLVM is already a behemoth and it already has its own craft, but GCC has more.

601
00:48:26,960 --> 00:48:31,040
 Okay, yeah, because I know in Rust, like there's some people working on GCC support and there's a

602
00:48:31,040 --> 00:48:35,040
 lot of people really interested in that. And like I said, what I told you is what I understood about

603
00:48:35,040 --> 00:48:40,480
 GCC versus LLVM, but I don't know how true that is and what exactly why people are interested

604
00:48:40,480 --> 00:48:46,080
 because you make it sound like LLVM is really much better in terms of licensing, of course,

605
00:48:46,080 --> 00:48:50,400
 but performance you're saying is really similar. I'm saying it's similar, yeah, I don't know where

606
00:48:50,400 --> 00:48:55,440
 we're at with one versus the other. I think Foronix has a bunch of comparison benchmarks. If you want

607
00:48:55,440 --> 00:49:00,080
 to know the latest, you can probably find it there. Okay, the other thing I thought was very

608
00:49:00,080 --> 00:49:03,440
 interesting, I want to kind of get back more in your career is that like you basically said, hey,

609
00:49:04,720 --> 00:49:07,920
 I really don't want to work on Zig. I think right after you left, okay, Cupid, right,

610
00:49:07,920 --> 00:49:11,120
 I think that's when you really want focusing full time or do you have one more thing before that,

611
00:49:11,120 --> 00:49:15,920
 one more gig? I did quit. Okay, Cupid directly to go full time on Zig. Yeah.

612
00:49:15,920 --> 00:49:20,240
 Okay, yeah, and you went full time on Zig and you actually did it really mathematically. I thought

613
00:49:20,240 --> 00:49:24,800
 it was kind of interesting and makes a lot of sense. It's like, okay, now I have enough cash and the

614
00:49:24,800 --> 00:49:29,600
 way things are going with donations, like I'm going to like get very close to being at rock bottom

615
00:49:29,600 --> 00:49:32,800
 and then be able to start going up from there. Right. I thought that was super interesting and

616
00:49:32,800 --> 00:49:36,720
 I'm happy. Yeah. You're very much a programmer and you're like very analytical about this,

617
00:49:36,720 --> 00:49:42,640
 looking at like a graph, I guess, and everything. It was very stressful. My fiance was very supportive

618
00:49:42,640 --> 00:49:47,200
 of me during that time. You know, you look at the data, but you kind of almost can't believe your

619
00:49:47,200 --> 00:49:51,920
 eyes and helps to have that other person saying, yeah, the number is check out, you can quit your

620
00:49:51,920 --> 00:49:56,880
 job and saying, okay, yeah, okay, I'm going to do it. Yeah, I mean, having a supportive essay was

621
00:49:56,880 --> 00:50:02,560
 really helpful, right? I can imagine too many people that would be like standing by you, but

622
00:50:02,560 --> 00:50:07,040
 yeah, I understand. I guess you must have really talked about your passion for this and then that

623
00:50:07,040 --> 00:50:10,800
 was like the kind of the thing. It's okay, let's just do it. Let's have a try, right?

624
00:50:10,800 --> 00:50:14,880
 Oh, you mean by way of kind of like convincing her to go along with it?

625
00:50:14,880 --> 00:50:18,720
 I mean, it's not about convincing, right? It's like, hey, I really want to do this. And then I

626
00:50:18,720 --> 00:50:22,880
 think that your passion alone should probably already convince them, hey, let's just give it a shot.

627
00:50:22,880 --> 00:50:26,720
 Right? Yeah, that sounds about right. I don't think it's really much of convincing. It's just like,

628
00:50:26,720 --> 00:50:31,200
 okay, I trust you and let's give it a try. Yeah. Did you also have to bring your data to and

629
00:50:31,200 --> 00:50:34,400
 really justify all your use cases? Do you ever give her an elevator pitch and everything?

630
00:50:34,400 --> 00:50:40,240
 Yeah, I think I'm the one who is a little more focused on making sure that crunching the numbers

631
00:50:40,240 --> 00:50:44,640
 and make sure the bottom line is black. So I think ultimately I had to convince myself, I think.

632
00:50:44,640 --> 00:50:50,080
 Okay, but I guess seeing that she's online, like, okay, this is really working out. I'm happy we did

633
00:50:50,080 --> 00:50:56,320
 it. Oh, yeah. She said that I was just obviously having some sort of a crisis of every day,

634
00:50:56,320 --> 00:51:00,800
 having to go to work when I wanted to not be going to work and working on this other thing.

635
00:51:00,800 --> 00:51:04,800
 And it was just a, I don't know, there's a difference in someone in someone's aura,

636
00:51:04,800 --> 00:51:08,880
 you know, when they're working on what they want to do with their life rather than, you know,

637
00:51:08,880 --> 00:51:13,040
 being a wage slave. So obviously it's been nice. Yeah, I mean, when you work on something

638
00:51:13,040 --> 00:51:17,680
 as you really want to work on, I can also justify this. I'm working longer hours, but I feel happier

639
00:51:17,680 --> 00:51:23,280
 overall because I'm doing things that I actually want to do. So I totally understand. Yeah, yeah.

640
00:51:23,280 --> 00:51:27,760
 Okay. Yeah, I mean, maybe we can kind of start to wrap up with some final points, right? So

641
00:51:28,400 --> 00:51:33,280
 Zigg is going towards a hosted compiler. Is there any other kind of short term things that

642
00:51:33,280 --> 00:51:37,360
 are going to be happening sometime soon? Short term. Yeah, well, I'll tell you what's happening

643
00:51:37,360 --> 00:51:44,960
 in a couple of weeks, we're having the first ever zig meetup in Milan, Italy. So yeah, about,

644
00:51:44,960 --> 00:51:49,840
 I don't know, 40 people, it looks like should be pretty fun. Yeah, I mean, people may

645
00:51:49,840 --> 00:51:55,040
 laugh about 40 people, but we have a saying in Asia, at least in China, it's like, every,

646
00:51:55,040 --> 00:52:00,320
 was a 1000 kilometer March starts with one step, right? So 40 people now, it could be

647
00:52:00,320 --> 00:52:06,160
 triple that the next time I have no idea. Like I said, I do see there's a community for zig. And

648
00:52:06,160 --> 00:52:09,440
 yeah, because people don't make it doesn't mean that people don't care. Well, this is the first

649
00:52:09,440 --> 00:52:13,440
 one. So I mean, as far as I'm concerned, the smaller the better, right? Because I like more

650
00:52:13,440 --> 00:52:19,120
 intimate gathering. So the way I see it is, I'm going to enjoy the smaller numbers while they last,

651
00:52:19,120 --> 00:52:22,880
 because in the future, they're going to be much bigger and it's going to have to be a different

652
00:52:22,880 --> 00:52:27,600
 affair, you know, everything too is that I'm sure you're finding out doing a conference is a whole

653
00:52:27,600 --> 00:52:31,840
 different thing. I've never done it before, but I've talked to people about them. And it's a lot of

654
00:52:31,840 --> 00:52:37,440
 work. Thankfully, I can thank Lawrence for that. Laurie Crow is doing all the setup. He lives in

655
00:52:37,440 --> 00:52:42,560
 Milan and he's set up the room and organizing and doing all the stuff that I would find it very

656
00:52:42,560 --> 00:52:47,600
 difficult to stay on top of. So all I have to do is show up, which is super nice for me.

657
00:52:47,600 --> 00:52:54,240
 So please don't mess this one up, please. Right. Yeah. Exactly. I hope you don't feel too much

658
00:52:54,240 --> 00:52:57,840
 pressure on that part. Right. It's always like that. It's like, it was Occam's razor is at the

659
00:52:57,840 --> 00:53:01,920
 one where it's like something they can go wrong or something like that. You're thinking of Murphy's law.

660
00:53:01,920 --> 00:53:05,920
 Sorry, Murphy's law. Yeah. Occam's razor is the one about like the most obvious answer is the one

661
00:53:05,920 --> 00:53:12,960
 that's true or something I forgot now. Occam's razor is if your explanation includes unnecessary

662
00:53:12,960 --> 00:53:16,800
 pieces of explanation, then just remove the unnecessary pieces because you don't need them.

663
00:53:17,680 --> 00:53:22,880
 Yeah. There it is. Again, I'm still waking up as I told you this morning. It's 10 30 almost.

664
00:53:22,880 --> 00:53:28,160
 So yeah, that's fair. You have some coffee there. Yeah. If I didn't have coffee, I think I would,

665
00:53:28,160 --> 00:53:34,000
 I don't know if I were to be able to even live. It's like my lifeblood. Nice. Ask me what kind of tea

666
00:53:34,000 --> 00:53:39,440
 I'm drinking. What kind of tea you're drinking? This tea is Ulong tea. It's thematic. Are you

667
00:53:39,440 --> 00:53:45,600
 ready? It's called Iron Goddess of Mercy. There's a tea called Teep is Cantonese word for steel or

668
00:53:45,600 --> 00:53:50,080
 no iron. So I'm wondering if that's the same one. I don't know. I just picked it because we're doing

669
00:53:50,080 --> 00:53:55,920
 the Rust podcast. But Guinyam is like a famous like mythological person in the Chinese community.

670
00:53:55,920 --> 00:54:00,160
 And also I think it's the same person who was like, there's also a woman in like the Indian

671
00:54:00,160 --> 00:54:03,760
 stuff that changes shape. It's actually similar, right? The India and China, they have a lot of

672
00:54:03,760 --> 00:54:09,040
 similarities with their mythology. That makes sense. They're next to each other. Also the British,

673
00:54:09,040 --> 00:54:13,680
 right? They brought British, you know, Indians over to Hong Kong and then you know, Hong Kong,

674
00:54:13,680 --> 00:54:17,360
 you know, they mix together, etc. Right. So a lot of this kind of stuff trading on that.

675
00:54:17,360 --> 00:54:21,440
 That makes sense. But yeah, I mean, I would love to chat some more, but I do have

676
00:54:21,440 --> 00:54:26,160
 something I need to do relatively soon. I appreciate your time. Maybe we can kind of wrap up with kind

677
00:54:26,160 --> 00:54:30,320
 of some, maybe if you have some last minute kind of calls, you want to call people to calls to action.

678
00:54:30,320 --> 00:54:35,440
 Maybe people come check out the language, contribute, go to the conference, anything.

679
00:54:35,440 --> 00:54:39,840
 Let's see. We have under Rust audience members, right? They're all probably pretty happy using

680
00:54:39,840 --> 00:54:43,600
 Rust. They're great. They don't need any calls to action. You know, if they feel inspired and they

681
00:54:43,600 --> 00:54:48,240
 want to throw some donation money our way, that's always welcome. You can find a link on the homepage,

682
00:54:48,240 --> 00:54:53,200
 ziggling.org. But it's all good, you know, we can all use the tools that work best for us.

683
00:54:53,200 --> 00:54:56,880
 No problem. Yeah, I think my call to action is just give it a try, especially if you're like,

684
00:54:56,880 --> 00:55:00,400
 you're struggling with SIG or you think that maybe Rust doesn't work for everything,

685
00:55:00,400 --> 00:55:04,560
 I'm sorry, struggling with Rust, not struggling with SIG. Give it a try. Maybe you find a new love,

686
00:55:04,560 --> 00:55:09,120
 right? You never know until you try. I never would have predicted I would be really into Rust as

687
00:55:09,120 --> 00:55:14,720
 I am right now. So maybe if I do try Zig, maybe I have a Zig station station after this. I don't know.

688
00:55:14,720 --> 00:55:18,960
 Zig station station. Yeah, you might have to come up with a new pun.

689
00:55:18,960 --> 00:55:22,080
 I don't know. If you have any ideas, iguana, sauna.

690
00:55:22,080 --> 00:55:26,160
 Why iguana? Is that the mask or iguana? Sorry. Yeah. It's like the Rust station, right? So

691
00:55:26,160 --> 00:55:30,480
 ziguana. So is an iguana actually the mask out? I didn't actually see that.

692
00:55:30,480 --> 00:55:36,480
 Yeah, there's a ziguana. So you can call it the ziguana sauna podcast. There you go. Done. It's a fork.

693
00:55:36,480 --> 00:55:42,000
 Okay. All right. Take a look at that. All right. Again, thank you for your time.

694
00:55:42,000 --> 00:55:45,440
 Maybe we'll have you back later on to talk more about what's going on with zig. I think

695
00:55:45,440 --> 00:55:49,360
 I'm happy that can give us an explanation finally because I've been asking people and

696
00:55:49,360 --> 00:55:53,680
 I haven't really gotten that. And you definitely give a very good outlook on it. And so hopefully

697
00:55:53,680 --> 00:55:58,880
 this inspires people to take a look at it. And maybe you'll have 41 or 43 people come into Milan.

698
00:55:58,880 --> 00:56:05,200
 Well, thank you for your time. Nice to meet you, Alan, and I hope you enjoy your coffee and the

699
00:56:05,200 --> 00:56:08,560
 rest of your day. And enjoy your tea. All right. Hopefully it doesn't make you stay up too late

700
00:56:08,560 --> 00:56:12,800
 tonight. It's nighttime for you, right? Yeah, no problem. Ulong tea is low caffeine.

701
00:56:12,800 --> 00:56:16,800
 Okay. I'm gonna look you in that one. All right. Thanks again for your help.

702
00:56:16,800 --> 00:56:25,520
 And I'll catch you again next time. Take care.