1 00:00:00,040 --> 00:00:01,720 Legacy code is code without tests. 2 00:00:01,760 --> 00:00:04,000 If you have code and has lots of tests, and it's relatively easy 3 00:00:04,000 --> 00:00:06,640 to change if you don't have the test, you're really in serious 4 00:00:06,640 --> 00:00:08,600 trouble. It's a strong qualitative 5 00:00:08,600 --> 00:00:10,880 difference between those two different types of code bases. 6 00:00:11,320 --> 00:00:13,760 If you want to teach good design, basically teach people 7 00:00:13,760 --> 00:00:15,960 the things to avoid and everything that's left is good 8 00:00:15,960 --> 00:00:17,760 design. I was kind of shocked in the 9 00:00:17,760 --> 00:00:20,280 beginning to see her vendors basically go and say, hey, we're 10 00:00:20,280 --> 00:00:21,600 going to write all. Your tests for you, yeah. 11 00:00:21,760 --> 00:00:23,800 You know, as the TDD guy, you're looking at this and you're 12 00:00:23,800 --> 00:00:26,160 saying, Oh my God, no, right. But for characterization 13 00:00:26,160 --> 00:00:29,000 testing, it's really kind of powerful because for people 14 00:00:29,000 --> 00:00:30,960 getting much more realistic now, I think in the very beginning 15 00:00:30,960 --> 00:00:33,640 people started going and extrapolating and they just get 16 00:00:33,720 --> 00:00:34,800 completely. Freaked out by the 17 00:00:34,800 --> 00:00:37,080 possibilities, right? But I think many. 18 00:00:37,080 --> 00:00:39,080 Developers now have enough experience what they are and 19 00:00:39,080 --> 00:00:40,920 they realize it's not going to replace many times at. 20 00:00:41,000 --> 00:00:44,440 The end of the day, in the legal system, you can't go and say, 21 00:00:44,600 --> 00:00:47,560 oh, the algorithm did it. It's like, no, a person who 22 00:00:47,560 --> 00:00:50,200 works at a company did this. At the end of the day, we are 23 00:00:50,200 --> 00:01:05,470 responsible for what we do. Hello, guys. 24 00:01:05,470 --> 00:01:07,590 Welcome back to another new episode of the Technical Journal 25 00:01:07,590 --> 00:01:09,750 Podcast. I'm very excited today to meet 26 00:01:09,750 --> 00:01:12,190 Michael Feathers. So we kind of like met nine 27 00:01:12,190 --> 00:01:14,790 years ago in one of his workshop back then in Singapore. 28 00:01:15,150 --> 00:01:18,270 So I think today is the first time we met again after that 29 00:01:18,270 --> 00:01:21,070 such a long time. So Michael, really great to have 30 00:01:21,070 --> 00:01:22,470 you here. Welcome to the show. 31 00:01:23,230 --> 00:01:23,950 Great. Thank you very much. 32 00:01:24,520 --> 00:01:27,160 Right Michael, for people who may not know you, so maybe in 33 00:01:27,160 --> 00:01:29,480 the 1st place, can you help to introduce yourself, maybe 34 00:01:29,480 --> 00:01:31,680 telling us a little bit of highlights or turning points in 35 00:01:31,680 --> 00:01:33,360 your career that we all can learn from? 36 00:01:34,040 --> 00:01:35,720 Sure. So my name is Michael Feathers 37 00:01:35,800 --> 00:01:37,440 and I guess the thing I'm best, whoops. 38 00:01:38,000 --> 00:01:39,680 Here's the thing I'm best known for within the industry. 39 00:01:39,680 --> 00:01:41,560 I was writing a book called Working Effectively with Legacy 40 00:01:41,560 --> 00:01:44,760 Code and I was involved in the early agile movement before it 41 00:01:44,760 --> 00:01:46,160 was even called Agile. Right. 42 00:01:46,800 --> 00:01:49,320 And I spent a lot of time since then just helping people with 43 00:01:49,320 --> 00:01:51,400 like legacy code and refactoring issues. 44 00:01:52,000 --> 00:01:53,640 And yeah, that's just been a thing. 45 00:01:53,640 --> 00:01:57,160 I've been very much interested in fighters that I spent about 9 46 00:01:57,160 --> 00:02:00,680 or 10 years programming, worked in various roles through the 47 00:02:00,680 --> 00:02:04,960 technical hierarchy and. Yeah, just sort of kind of gone 48 00:02:04,960 --> 00:02:06,040 on from there. Yeah. 49 00:02:06,040 --> 00:02:08,639 So I think your book working with the legacy code, right. 50 00:02:09,000 --> 00:02:11,280 I think now it's like the in the 20th anniversary. 51 00:02:11,280 --> 00:02:12,960 So Congrats for that Saturday I think. 52 00:02:13,360 --> 00:02:15,600 Yeah, thanks for reminding me. I forgot what year was in the 53 00:02:15,600 --> 00:02:17,120 20s. Yeah. 54 00:02:17,120 --> 00:02:19,600 So I think it's one of the so-called classic software 55 00:02:19,600 --> 00:02:22,240 programming books, I would say especially people who are 56 00:02:22,240 --> 00:02:25,720 dealing with legacy code. So maybe let's start from that 57 00:02:25,720 --> 00:02:27,920 book first, right. So since you're well known about 58 00:02:28,080 --> 00:02:30,280 that book, so legacy code, right. 59 00:02:30,280 --> 00:02:33,360 So in the 1st place, why did you write something about legacy 60 00:02:33,360 --> 00:02:36,080 code? I guess people back then may 61 00:02:36,080 --> 00:02:38,760 have started working on some legacy code that is difficult to 62 00:02:38,760 --> 00:02:41,280 work on, or that's something that you learn along the way. 63 00:02:42,400 --> 00:02:43,760 Well, it's kind of interesting with that. 64 00:02:43,760 --> 00:02:47,800 So I went to a conference and I met Kent Beck and Ron Jeffries 65 00:02:47,800 --> 00:02:51,880 back in like 1990. 9:00 or so. And I started to go and get some 66 00:02:51,880 --> 00:02:54,040 notice of the people that were elected balls in the very early 67 00:02:54,040 --> 00:02:56,800 agile movement, as I mentioned. And it was interesting because I 68 00:02:56,800 --> 00:02:59,440 learned so much from them. It was interesting because at 69 00:02:59,440 --> 00:03:02,240 that time I was a good programmer and everything, but 70 00:03:02,240 --> 00:03:04,160 they had really thought deeply about how to become better 71 00:03:04,160 --> 00:03:06,560 programmers. So I wanted to refactor. 72 00:03:06,600 --> 00:03:08,840 And I was doing a bit of that back in my workplace. 73 00:03:09,320 --> 00:03:11,680 And they were also saying things like, you should write your test 74 00:03:11,680 --> 00:03:13,400 first, why would you do it any other way? 75 00:03:13,960 --> 00:03:16,080 And I thought this is kind of like a very, you know, bold 76 00:03:16,080 --> 00:03:18,200 thing for them to go and save. Then I understood after I did it 77 00:03:18,200 --> 00:03:20,760 for a while, and it helped me realize I could reduce the 78 00:03:20,760 --> 00:03:22,280 stress of programming a great deal. 79 00:03:23,040 --> 00:03:25,800 And so after that, I was fired to be a consultant. 80 00:03:25,800 --> 00:03:29,520 I lost the tech job that I had, and the job was to go around to 81 00:03:29,520 --> 00:03:32,760 different places and help them acquire the technical practices 82 00:03:32,760 --> 00:03:36,520 that are part of Agile, right? And invariably, it's kind of 83 00:03:36,520 --> 00:03:38,640 like you run into a situation where people have tons and tons 84 00:03:38,640 --> 00:03:41,680 of code and they just don't know quite what to do with it, right? 85 00:03:41,760 --> 00:03:43,960 It's hard to refactor. And you look at Martin Towers 86 00:03:44,320 --> 00:03:47,040 back in that day and they said, OK, and a snippet. 87 00:03:47,040 --> 00:03:49,880 It's kind of like, here's all this refactoring stuff, but you 88 00:03:49,880 --> 00:03:51,240 should have tests. And if you don't have tests, 89 00:03:51,240 --> 00:03:54,320 you're in a bit of trouble. So I just realized that this was 90 00:03:54,320 --> 00:03:56,600 a problem, and I just started collecting techniques for going 91 00:03:56,600 --> 00:03:59,320 and breaking dependencies and getting tests into place. 92 00:03:59,600 --> 00:04:01,880 And yeah, it's been an interesting ride that way. 93 00:04:02,160 --> 00:04:06,560 But it's also been really kind of fascinating for me because I 94 00:04:06,560 --> 00:04:09,960 started like a mathematics major in school and I had this 95 00:04:09,960 --> 00:04:12,120 interesting idea of kind of like being an introvert. 96 00:04:12,120 --> 00:04:15,360 Of course, I read about this mathematician in Pauli Erdish, 97 00:04:15,400 --> 00:04:18,399 and it was like he lived with a suitcase and would go from place 98 00:04:18,399 --> 00:04:20,600 to place and sort of work with other mathematicians. 99 00:04:21,079 --> 00:04:22,960 And the cool thing was he didn't even have to have a piece of 100 00:04:22,960 --> 00:04:24,440 paper or pencil to do this, right? 101 00:04:24,440 --> 00:04:26,640 He just had conversations and do these things. 102 00:04:27,160 --> 00:04:29,160 I thought, wow, that's really an interesting thing to work on 103 00:04:29,160 --> 00:04:31,600 mathematics and just make it purely your life. 104 00:04:32,160 --> 00:04:33,880 But then I kind of realized it's kind of like if you're doing 105 00:04:33,880 --> 00:04:35,800 that, you're not really helping people right directly. 106 00:04:36,160 --> 00:04:39,120 So I've had much more fun going working with people and 107 00:04:39,120 --> 00:04:41,440 realizing that I can make a bit of a difference. 108 00:04:42,000 --> 00:04:44,120 And going back to my book, as much as it's very technical and 109 00:04:44,120 --> 00:04:46,920 has a lot of code in it, I'm only realizing now how much it 110 00:04:46,920 --> 00:04:49,560 was really about going and helping people keep their morale 111 00:04:49,560 --> 00:04:52,360 up and, you know, have a, a positive outlook to what they 112 00:04:52,360 --> 00:04:53,400 can do and know what they can do. 113 00:04:53,400 --> 00:04:56,200 Within code base. Yeah, I think the book itself, 114 00:04:56,200 --> 00:04:59,240 although it has been 20 years, I think it's still pretty relevant 115 00:04:59,240 --> 00:05:01,640 because people now, they still work on legacy code. 116 00:05:01,640 --> 00:05:04,920 In fact, there might be more legacy code than last time 117 00:05:04,920 --> 00:05:06,960 before, right? But in the 1st place, let's go 118 00:05:06,960 --> 00:05:09,680 back to the definition itself about legacy code, because some 119 00:05:09,680 --> 00:05:12,160 people might have different perceptions, different 120 00:05:12,160 --> 00:05:14,400 definition. Some people think it's like old 121 00:05:14,480 --> 00:05:16,760 legacy code. Some people think it's code that 122 00:05:16,760 --> 00:05:19,040 they just don't want to touch or they don't understand. 123 00:05:19,280 --> 00:05:21,880 In your book, actually, you have one good definition that is 124 00:05:21,880 --> 00:05:24,560 probably interesting, like you mentioned something related to 125 00:05:24,560 --> 00:05:27,000 test in the very beginning. So maybe let's go to your 126 00:05:27,000 --> 00:05:28,840 definition. What is legacy code? 127 00:05:28,920 --> 00:05:32,160 Actually, we've got many definitions, but the one I'm 128 00:05:32,160 --> 00:05:34,720 kind of known for is basically the legacy code is code without 129 00:05:34,720 --> 00:05:37,120 tests. And it's funny about this 130 00:05:37,120 --> 00:05:40,600 because this happened when I was working with the team and I had 131 00:05:40,600 --> 00:05:42,360 another consultant with me and I got. 132 00:05:42,360 --> 00:05:44,800 Frustrated at 1 moment and. Somebody asked a question. 133 00:05:44,800 --> 00:05:47,120 I said legacy code is code without tests. 134 00:05:47,120 --> 00:05:48,560 That's just what you need to know, right? 135 00:05:49,200 --> 00:05:51,760 And you know, my friend, the consultant, he basically said, 136 00:05:51,760 --> 00:05:52,720 you know, you got to put that out. 137 00:05:52,720 --> 00:05:54,440 There, that's a really powerful statement to say. 138 00:05:55,280 --> 00:05:59,120 And I knew at the moment that we really was at odds with the 139 00:05:59,120 --> 00:06:01,160 traditional definition. You get code from other people 140 00:06:01,160 --> 00:06:02,600 and that becomes. Legacy code, right? 141 00:06:03,080 --> 00:06:05,400 Or code that you're scared of and don't quite understand. 142 00:06:05,960 --> 00:06:08,280 But I think, you know, with all these definitions, they're 143 00:06:08,280 --> 00:06:11,360 pragmatic in a way because they're used to highlight 144 00:06:11,360 --> 00:06:13,440 something that you basically want to change. 145 00:06:14,080 --> 00:06:15,760 With this one, it was really just like had this. 146 00:06:15,760 --> 00:06:18,040 Intuitive sense that. You know, if you have code and 147 00:06:18,040 --> 00:06:20,040 has lots of tests and it's relatively easy to change. 148 00:06:20,040 --> 00:06:22,240 If you don't have the test, you're really in serious. 149 00:06:22,240 --> 00:06:23,240 Trouble. It's a strong. 150 00:06:23,640 --> 00:06:25,720 Qualitative difference between those two different types of 151 00:06:25,720 --> 00:06:27,880 code bases. And so it's trying to get that 152 00:06:27,880 --> 00:06:29,960 across. But occasionally people will say 153 00:06:29,960 --> 00:06:31,960 who are you to make the definition of legacy? 154 00:06:32,480 --> 00:06:35,200 It's just one or as many. Pick whichever 1 you want. 155 00:06:35,480 --> 00:06:36,320 Pick the one that helps. You. 156 00:06:36,320 --> 00:06:37,920 I think it's the best thing to say like that. 157 00:06:39,040 --> 00:06:40,280 Yeah. So I think in the context of 158 00:06:40,280 --> 00:06:42,840 your definition, you have emphasized a lot about testing, 159 00:06:42,840 --> 00:06:43,760 right. I think in the very beginning 160 00:06:43,760 --> 00:06:46,080 you mentioned you learned from Kendbeck about test first. 161 00:06:46,720 --> 00:06:49,120 Maybe in the refactoring book also it was mentioned that you 162 00:06:49,120 --> 00:06:51,080 refactor also along with the test. 163 00:06:51,280 --> 00:06:54,000 Why the test is so important in your opinion? 164 00:06:54,680 --> 00:06:57,280 Well, I think the main thing is because this is going to be kind 165 00:06:57,280 --> 00:06:58,800 of like getting a philosophical in a way. 166 00:06:58,800 --> 00:07:02,280 It's kind of like most of us walk around with an idea in our 167 00:07:02,280 --> 00:07:03,960 heads about what the system does. 168 00:07:04,520 --> 00:07:06,800 And The thing is, it's kind of like the system has its own idea 169 00:07:06,800 --> 00:07:09,560 about what it does. It's not an idea, it's what it 170 00:07:09,560 --> 00:07:12,160 does, right? And I think that that's an 171 00:07:12,160 --> 00:07:15,280 interesting thing. We quite often get so caught up 172 00:07:15,280 --> 00:07:18,240 with what we think the system is doing and what we think it we 173 00:07:18,240 --> 00:07:20,800 want it to do, and we forget to even think about what it 174 00:07:20,800 --> 00:07:23,840 currently is doing, right? So to me, that's a very 175 00:07:23,840 --> 00:07:26,480 grounding perspective. It's kind of like if you really 176 00:07:26,480 --> 00:07:29,400 know what the system does, you have a great base point to go 177 00:07:29,400 --> 00:07:31,080 and decide what to do next, right? 178 00:07:31,720 --> 00:07:34,000 And over and over again, I've run into situations where 179 00:07:34,000 --> 00:07:36,560 people, once they learn a bit more about what the system 180 00:07:36,560 --> 00:07:39,080 actually does, they have different ideas about what's 181 00:07:39,080 --> 00:07:41,120 possible. And they're kind of like, oh, 182 00:07:41,120 --> 00:07:43,320 you know, it's kind of like people have been asking for this 183 00:07:43,320 --> 00:07:45,760 particular feature and now I see it'll be relatively easy to add 184 00:07:45,760 --> 00:07:46,600 that. Feature, right? 185 00:07:46,600 --> 00:07:47,880 I've always been sort of like saying. 186 00:07:48,360 --> 00:07:50,200 It'll take three weeks to do this sort of thing, but now I 187 00:07:50,200 --> 00:07:51,440 know it's much. Easier to do. 188 00:07:51,840 --> 00:07:54,640 And without that baseline of knowledge, you just really don't 189 00:07:54,920 --> 00:07:57,640 know that. So we talked about tests first a 190 00:07:57,640 --> 00:07:59,160 second ago. One of the things I talk about 191 00:07:59,160 --> 00:08:01,960 in the book is something I call characterization tests. 192 00:08:02,480 --> 00:08:04,440 And quite often people call this like pinning tests now 193 00:08:04,440 --> 00:08:06,640 sometimes. And the idea is that basically 194 00:08:06,640 --> 00:08:09,000 you write tests that you're used to describe the current behavior 195 00:08:09,000 --> 00:08:10,960 of the system. So you're not writing them 196 00:08:10,960 --> 00:08:13,320 first, you're writing them after the code has been written and 197 00:08:13,680 --> 00:08:15,960 you're writing them to go and ask a question of the code base. 198 00:08:16,520 --> 00:08:18,760 And once you get the answer, you basically take that answer and 199 00:08:18,760 --> 00:08:22,080 you put it in as the expectation if the notice. 200 00:08:22,080 --> 00:08:23,800 This is called golden master testing. 201 00:08:23,800 --> 00:08:26,600 In a way it's considered like a poor practice, but I think when 202 00:08:26,600 --> 00:08:29,000 you have existing code without tests, it's just a way of going 203 00:08:29,000 --> 00:08:30,320 and actually understanding what you've. 204 00:08:30,320 --> 00:08:32,840 Got. And you can basically put that 205 00:08:33,280 --> 00:08:36,000 off to the side and distinguish it from the test that you write 206 00:08:36,000 --> 00:08:39,240 1st to go and actually sort of, you know, figure you towards 207 00:08:39,240 --> 00:08:40,840 creating feature understanding when you're done. 208 00:08:41,640 --> 00:08:44,840 So yeah, that's the thing. To me, a test is a way of 209 00:08:44,840 --> 00:08:48,080 grounding our knowledge of a system, right. 210 00:08:48,080 --> 00:08:51,440 When we refer to the tests, are you referring to automated tests 211 00:08:51,440 --> 00:08:54,160 or are you referring also for other types of tests? 212 00:08:54,160 --> 00:08:56,400 Like, I don't know, people have manual tests, UI tests, 213 00:08:56,400 --> 00:08:58,600 regression tests, so many other tests out there. 214 00:08:58,600 --> 00:09:00,320 Right. So are you specifically 215 00:09:00,320 --> 00:09:02,760 referring to just automated tests or something else? 216 00:09:03,720 --> 00:09:06,160 Yeah, generally automated tests. You know, it's kind of like it 217 00:09:06,160 --> 00:09:09,360 seems like I'm finding two or two places that still do much 218 00:09:09,360 --> 00:09:11,960 manual testing and all right. And I think it's you know, 219 00:09:11,960 --> 00:09:14,000 especially with like dev OPS and the accelerate book and 220 00:09:14,000 --> 00:09:16,280 everything like that, you know, it's the idea is going to get 221 00:09:16,280 --> 00:09:19,440 something in place with tests and have that for within an hour 222 00:09:19,440 --> 00:09:21,560 if you can, right. So there doesn't really leave 223 00:09:21,560 --> 00:09:23,760 much time with that. So generally I am talking about 224 00:09:23,760 --> 00:09:25,880 automated tests. Yeah, it's kind of funny. 225 00:09:25,880 --> 00:09:27,960 I think most of the people that are like thought leaders in the 226 00:09:28,720 --> 00:09:31,120 non automatic testing space have really kind of moved into 227 00:09:31,120 --> 00:09:33,720 calling it exploratory testing and basically going and 228 00:09:33,720 --> 00:09:38,640 accenting the fact that as somebody who basically has ideas 229 00:09:38,640 --> 00:09:40,880 though the system. Could be doing, you can just 230 00:09:40,880 --> 00:09:43,480 sort of verify those and look for really terrible things and 231 00:09:43,960 --> 00:09:45,120 make people aware of those, right? 232 00:09:45,120 --> 00:09:46,560 So that's an exploratory approach. 233 00:09:47,720 --> 00:09:50,840 So I think talking about testing, right, although it has 234 00:09:50,840 --> 00:09:53,080 been around for many, many years, right? 235 00:09:53,080 --> 00:09:56,320 I think still the state of the software development practice 236 00:09:56,320 --> 00:09:59,520 these days, testing is still kind of like lacking in terms of 237 00:09:59,520 --> 00:10:02,880 quality, in terms of quantity. And I know that you have written 238 00:10:02,880 --> 00:10:05,680 this book since 20 years ago. Do you think the definition of 239 00:10:05,680 --> 00:10:09,160 legacy code still kind of like relevant or is it something 240 00:10:09,160 --> 00:10:12,640 changing because with so many other tools, techniques and also 241 00:10:12,640 --> 00:10:15,560 maybe now with AI? So is the definition still kind 242 00:10:15,560 --> 00:10:17,760 of like appropriate or do you think something has evolved 243 00:10:17,760 --> 00:10:19,840 along the way? Well, if you don't have any 244 00:10:19,840 --> 00:10:21,000 tests, I think it's appropriate for you. 245 00:10:21,880 --> 00:10:23,280 This is the best way to go and put it, right? 246 00:10:23,800 --> 00:10:26,320 But I think, you know, we might end up in a place where legacy 247 00:10:26,320 --> 00:10:29,280 just reversed and I think it passed to some degree to 248 00:10:29,320 --> 00:10:31,280 basically the thing that we don't understand, right? 249 00:10:31,280 --> 00:10:34,080 And I think that's another definition is like legacy code 250 00:10:34,080 --> 00:10:36,640 is code we don't understand. So yeah, you know, there's a lot 251 00:10:36,640 --> 00:10:38,760 more that we can do now with in order to go and actually 252 00:10:38,760 --> 00:10:41,040 understand. The systems that we have when 253 00:10:41,040 --> 00:10:43,280 using AI through this quite often just like taking a lot of 254 00:10:43,280 --> 00:10:46,040 code and sort of like pulling it into a session and asking 255 00:10:46,040 --> 00:10:48,440 questions about it. And you know, the results have 256 00:10:48,440 --> 00:10:49,480 been mixed. They can, kind of. 257 00:10:49,480 --> 00:10:53,160 See the advantages to this now? And I'm thinking we're kind of 258 00:10:53,160 --> 00:10:54,200 like due. For. 259 00:10:54,760 --> 00:10:56,640 Let's see, it's August 2024 right now, right? 260 00:10:57,320 --> 00:10:59,840 We're due for this kind of backlash and people say, oh, you 261 00:10:59,920 --> 00:11:01,200 know, we're just making things up. 262 00:11:01,280 --> 00:11:04,040 And the second to help me as a. Programmer and stuff like that. 263 00:11:04,480 --> 00:11:07,080 But you know, you can pay attention to the social winds of 264 00:11:07,080 --> 00:11:08,600 things. But I think that. 265 00:11:08,600 --> 00:11:10,720 This technology is only going to get better every time, and it's 266 00:11:10,720 --> 00:11:12,840 really important. For every programmer to learn 267 00:11:12,840 --> 00:11:14,840 how to use it well and understand what limitations are. 268 00:11:14,960 --> 00:11:17,240 So. Yeah, I think speaking about 269 00:11:17,240 --> 00:11:20,840 legacy code, right, I think almost all developers, I believe 270 00:11:20,880 --> 00:11:23,720 that they hate working with legacy code, right, No matter 271 00:11:23,720 --> 00:11:26,040 what, right? So either that they don't know 272 00:11:26,040 --> 00:11:29,680 what the code is doing, either the code is buggy or the code is 273 00:11:29,680 --> 00:11:33,280 just so outdated that you can't even change it to using the 274 00:11:33,280 --> 00:11:34,840 modern practice and things like that, right? 275 00:11:35,280 --> 00:11:38,320 So in your consulting, I'm sure you have seen the worst of 276 00:11:38,600 --> 00:11:41,640 legacy code maybe and some which are probably not so bad. 277 00:11:41,960 --> 00:11:44,240 But if you can tell us a little bit, what are the some of the 278 00:11:44,240 --> 00:11:47,600 typical challenges that people when they deal with legacy code? 279 00:11:47,600 --> 00:11:51,160 Why is it something that is kind of like avoided or hated a lot 280 00:11:51,200 --> 00:11:54,280 in the software development? You know, I think the mindset 281 00:11:54,280 --> 00:11:56,400 has a lot to do with it. And I think also there's the 282 00:11:56,400 --> 00:11:58,320 ambition of creating a new system, right? 283 00:11:59,360 --> 00:12:01,280 And it's kind of like it seems with a lot of modern systems, 284 00:12:01,280 --> 00:12:03,920 you can, you know, you may be working with older code, but you 285 00:12:03,920 --> 00:12:05,840 get to go and sort of like create this little space where 286 00:12:05,840 --> 00:12:08,480 you have, say, new, new service and you can write the code from 287 00:12:08,480 --> 00:12:10,600 scratch and doodles. And you get very happy about 288 00:12:10,600 --> 00:12:12,640 this because you get to sort of display your creativity. 289 00:12:13,360 --> 00:12:15,960 But some people actually enjoy debugging, right? 290 00:12:15,960 --> 00:12:18,000 You know, some people who enjoy debugging, right? 291 00:12:18,520 --> 00:12:19,760 It's the same kind of thing in a way. 292 00:12:19,760 --> 00:12:21,960 It's this thing of trying to understand something. 293 00:12:22,320 --> 00:12:25,720 And I think I might be sort of like more predisposed to this 294 00:12:25,720 --> 00:12:28,320 than most people because from being a kid, I've just always 295 00:12:28,320 --> 00:12:30,960 been curious about everything. Anything I could read, I would 296 00:12:30,960 --> 00:12:33,640 just sort of read, read, read and try things out and that sort 297 00:12:33,640 --> 00:12:36,000 of thing. And if you approach it as an 298 00:12:36,000 --> 00:12:37,920 adventure, I think that's something which is really kind 299 00:12:37,920 --> 00:12:40,520 of good in a way. Reminds me of a story. 300 00:12:40,520 --> 00:12:42,560 And it's not directly about the technical work, but I was 301 00:12:42,560 --> 00:12:44,480 working with a team many years ago. 302 00:12:44,520 --> 00:12:48,120 And it was interesting because I was there for like 4 or five 303 00:12:48,120 --> 00:12:51,080 months on and off. And, you know, at the end of 304 00:12:51,080 --> 00:12:53,600 this, the guy who brought me in, he brought me into his office. 305 00:12:54,080 --> 00:12:55,840 They showed me these graphs. They showed me how the quality 306 00:12:55,840 --> 00:12:58,040 had improved through all the time that I'd been there. 307 00:12:58,040 --> 00:12:59,800 We really, you know, made a big difference. 308 00:13:00,360 --> 00:13:02,480 And I thought, well, this is weird because I feel like a 309 00:13:02,480 --> 00:13:04,720 failure. Why do I feel like a failure? 310 00:13:04,720 --> 00:13:08,240 And what it was, this was like early days, like 2000s, 311 00:13:08,240 --> 00:13:10,000 likethe.com. Bus and all this other stuff. 312 00:13:10,520 --> 00:13:12,280 Was working at a startup and everybody thought they were 313 00:13:12,280 --> 00:13:13,840 going to be a millionaire and then they. 314 00:13:13,840 --> 00:13:15,600 Discovered they weren't going to be a millionaire, right? 315 00:13:16,160 --> 00:13:19,240 And it's just this interesting thing of kind of like, and I 316 00:13:19,240 --> 00:13:20,800 started seeing the friends. It's like, wow, it's, you know, 317 00:13:21,160 --> 00:13:23,520 the thing which really affects whether we feel good today or 318 00:13:23,520 --> 00:13:25,560 not is what our expectations were, right? 319 00:13:26,200 --> 00:13:28,760 And it's, you know, and I talked to a friend who was like in 320 00:13:28,760 --> 00:13:30,640 Eastern religion, he's like, oh, that's his own thing. 321 00:13:30,640 --> 00:13:33,080 It's like, OK, but it's like, it's really an interesting 322 00:13:33,080 --> 00:13:33,880 thing. It's like our. 323 00:13:33,880 --> 00:13:36,200 Expectations set our experience to a strong degree. 324 00:13:36,880 --> 00:13:40,400 And I think if you can actually cultivate the curiosity about 325 00:13:40,400 --> 00:13:44,480 what does the system do right, or the desire to help people in 326 00:13:44,480 --> 00:13:47,520 that context, it can be very, you know, rewarding work. 327 00:13:48,120 --> 00:13:50,000 You may just have to go and put aside your expectations a little 328 00:13:50,000 --> 00:13:52,280 bit. Yeah, you touch on just now 329 00:13:52,280 --> 00:13:55,200 something that a lot of developers probably like to do, 330 00:13:55,200 --> 00:13:58,520 right, which is to create new system or rewrite everything, 331 00:13:58,520 --> 00:14:00,400 right. So since you write the legacy 332 00:14:00,400 --> 00:14:03,520 code book, what is your approach towards fixing the legacy code 333 00:14:03,520 --> 00:14:06,080 or maybe refactoring the legacy code versus rewrite? 334 00:14:06,080 --> 00:14:08,280 Right. So when should we actually 335 00:14:08,440 --> 00:14:11,480 contemplate doing rewrite versus fixing the legacy code? 336 00:14:12,240 --> 00:14:14,080 Yeah, a lot of it comes down to business reason rather than 337 00:14:14,080 --> 00:14:16,200 technical reason, because at the end of the day, the technical 338 00:14:16,200 --> 00:14:19,280 things just kind of like we can deal with anything. 339 00:14:19,280 --> 00:14:21,880 You know, it's a question whether there's payback for this 340 00:14:21,880 --> 00:14:24,080 particular thing. But the business case needs to 341 00:14:24,080 --> 00:14:27,600 be made, needs to be understood that you don't have to rewrite 342 00:14:27,600 --> 00:14:30,840 an entire system most of the time, but you can concentrate on 343 00:14:30,840 --> 00:14:33,360 the things which basically have growth in the future and also, 344 00:14:33,920 --> 00:14:36,360 you know, are presenting difficulty to you in the 345 00:14:36,360 --> 00:14:38,280 meantime. Usually what I'd like to do is I 346 00:14:38,280 --> 00:14:40,320 think I'd like to go and actually like try to do some 347 00:14:40,320 --> 00:14:43,360 refactoring, try to do some planning for some refactoring, 348 00:14:43,360 --> 00:14:46,400 look at perspective features, and then having all that stuff 349 00:14:46,400 --> 00:14:49,280 together thinking about whether a read write is justified or. 350 00:14:49,280 --> 00:14:51,320 Not right as. I mentioned a little earlier 351 00:14:51,320 --> 00:14:54,240 this thing of going and actually doing spot rewrites on 352 00:14:54,240 --> 00:14:56,360 particular things and it's really pretty powerful. 353 00:14:56,360 --> 00:14:58,760 And quite often for people, it's like an all or nothing. 354 00:14:58,760 --> 00:15:00,040 Proposition. And that's really. 355 00:15:00,520 --> 00:15:01,880 A verbal. Position to approach these 356 00:15:01,880 --> 00:15:04,120 things from. But yeah, you know, the thing 357 00:15:04,120 --> 00:15:06,440 with rewrites is that it's great if you have tests. 358 00:15:06,440 --> 00:15:09,000 So if you have tests, you know, really in a golden space. 359 00:15:09,760 --> 00:15:11,440 If you don't, then you're kind of like dealing with the 360 00:15:11,440 --> 00:15:14,040 situation of like, oh, I need to write this thing and I don't 361 00:15:14,040 --> 00:15:16,320 understand what it does. And it's like, OK, well now you 362 00:15:16,320 --> 00:15:18,840 got 2 problems, right? So I think it's really just the 363 00:15:18,840 --> 00:15:21,560 best way to kind of approach it. The thing is that's one that's 364 00:15:21,880 --> 00:15:24,600 rewrite when you're basically dealing with technical debt 365 00:15:24,600 --> 00:15:26,640 directly. But there are these other. 366 00:15:26,640 --> 00:15:28,680 Rewrites, which are really all about architecture that really 367 00:15:28,680 --> 00:15:30,400 have, you know. A different set of 368 00:15:30,400 --> 00:15:31,640 considerations. Leading into them. 369 00:15:32,120 --> 00:15:34,680 And these are the architectural ones are usually like, you know, 370 00:15:34,680 --> 00:15:37,720 much more expensive speaking about architecture, right? 371 00:15:37,720 --> 00:15:41,160 So some people to the extreme right, like doing micro service. 372 00:15:41,160 --> 00:15:43,240 So that one part of the definition of micro service for 373 00:15:43,240 --> 00:15:45,720 some people is like, if you can rewrite the whole service 374 00:15:45,720 --> 00:15:48,400 within, I don't know, X days, right, then that's the micro 375 00:15:48,400 --> 00:15:50,880 service. So with that kind of maybe well 376 00:15:50,880 --> 00:15:52,240 designed micro service, you can do that. 377 00:15:52,240 --> 00:15:54,040 So what's your take about this as well, right? 378 00:15:54,040 --> 00:15:57,560 Like when you refactor the application into a different 379 00:15:57,560 --> 00:16:00,440 service, is that also some strategy that you can do for 380 00:16:00,440 --> 00:16:03,160 tackling legacy code? Yeah, definitely. 381 00:16:03,240 --> 00:16:05,160 You know, I think that's an interesting thing to actually 382 00:16:05,160 --> 00:16:06,800 move things around to different places. 383 00:16:06,800 --> 00:16:09,440 And so it's really, at a macro level, the same thing we do as a 384 00:16:09,440 --> 00:16:12,680 micro level, at a micro level with classes in many ways. 385 00:16:13,040 --> 00:16:14,600 But as far as that thing of like, you should be able to 386 00:16:14,600 --> 00:16:16,920 rewrite in a day or so, I think that's a great goal. 387 00:16:17,280 --> 00:16:18,520 Hard to see it in practice, though. 388 00:16:18,640 --> 00:16:21,560 It's really a weird thing. I think there's this intrinsic 389 00:16:21,560 --> 00:16:23,720 thing about software that basically tends to grow and grow 390 00:16:23,720 --> 00:16:25,880 and grow. And quite often it grows to the 391 00:16:25,880 --> 00:16:28,840 point where you basically have, you know, it's hard to go and do 392 00:16:28,840 --> 00:16:30,600 anything like that. But it snuck up on you, that 393 00:16:30,600 --> 00:16:33,560 kind of thing. Remember like reading through 394 00:16:33,560 --> 00:16:37,080 some research years ago about like how big does a class need 395 00:16:37,080 --> 00:16:38,440 to be? Before you actually go in 396 00:16:38,440 --> 00:16:39,920 refracting, right, that kind of thing. 397 00:16:40,600 --> 00:16:43,360 And looking at actually like quality data, it's kind of like 398 00:16:43,360 --> 00:16:45,680 where do we have the most bugs in the biggest? 399 00:16:45,680 --> 00:16:47,000 Abstraction to the smallest ones. 400 00:16:47,400 --> 00:16:49,280 Turns out the answer is really the one so you touch the most 401 00:16:49,280 --> 00:16:52,040 often, right? And so it's kind of like we all 402 00:16:52,040 --> 00:16:55,000 have a feeling for when things get to be too big, but quite 403 00:16:55,000 --> 00:16:57,360 often it's kind of like there's no set point. 404 00:16:57,360 --> 00:16:59,200 It's kind of like a. Subjective sense in a way. 405 00:16:59,640 --> 00:17:02,040 But yeah, I think there's a lot of stuff that was a dream for 406 00:17:02,040 --> 00:17:05,480 micro services that I think is, you know, it's a shame that we 407 00:17:05,480 --> 00:17:07,520 kind of fell away from some of that stuff. 408 00:17:07,520 --> 00:17:10,359 The thing of being able to use almost any language, be able to 409 00:17:10,359 --> 00:17:13,079 throw things into production almost immediately, be able to 410 00:17:13,079 --> 00:17:14,640 go and rewrite them when we need to. 411 00:17:14,680 --> 00:17:18,040 And yeah, I don't know, I tend to think there was a lot of 412 00:17:18,040 --> 00:17:19,640 these things. It's kind of like we only get 413 00:17:19,640 --> 00:17:21,560 the resilient systems we need when we stress. 414 00:17:21,560 --> 00:17:23,560 Them a bit. If there's a reason to start 415 00:17:23,560 --> 00:17:27,119 rewriting things, then it's good because we actually gain the 416 00:17:27,119 --> 00:17:29,200 skill to do this and don't think we're just being an atypical 417 00:17:29,200 --> 00:17:31,040 thing, right? So it's actually good when 418 00:17:31,040 --> 00:17:32,320 there's a bit. Of stress in the environment, 419 00:17:32,480 --> 00:17:35,520 you know. So speaking about how people 420 00:17:35,520 --> 00:17:38,640 tackle this legacy code, right, maybe from your experience or 421 00:17:38,640 --> 00:17:40,080 maybe some summary of your book, right? 422 00:17:40,080 --> 00:17:43,200 What are the techniques that people can evaluate or maybe 423 00:17:43,200 --> 00:17:46,000 people can try to actually tackle legacy code in a much 424 00:17:46,120 --> 00:17:49,480 elegant way, so to speak? In terms of that, I think the 425 00:17:49,520 --> 00:17:51,760 main thing is to go and figure out exactly where you need to 426 00:17:51,760 --> 00:17:54,520 make changes, right? And then you move outwards from 427 00:17:54,520 --> 00:17:56,400 there. They may be cases where you need 428 00:17:56,400 --> 00:17:59,320 to change two or three lines in the method, for instance, or you 429 00:17:59,320 --> 00:18:01,760 need to go and actually sort of like you add five or six methods 430 00:18:01,760 --> 00:18:04,000 to a class, but maybe you're thinking about creating a new 431 00:18:04,000 --> 00:18:05,920 class. And then what you do is. 432 00:18:05,920 --> 00:18:08,360 So basically you think about what would be involved in 433 00:18:08,360 --> 00:18:11,600 getting tests in place for each one of those places, and you 434 00:18:11,800 --> 00:18:13,920 kind of like crawl the tree upward and see if there's a 435 00:18:13,920 --> 00:18:15,960 common. Place for those changes. 436 00:18:16,600 --> 00:18:18,280 And you try to pragmatically figure out where you want. 437 00:18:18,280 --> 00:18:21,120 To go and have your tests. Now, if you have code that 438 00:18:21,120 --> 00:18:24,000 doesn't really have very many tests in place, chances are the 439 00:18:24,000 --> 00:18:26,080 code isn't really able to accommodate tests. 440 00:18:26,080 --> 00:18:29,000 Yeah, and you have to do quite a bit of dependency breaking work 441 00:18:29,000 --> 00:18:30,800 in order to go and actually put the test in place. 442 00:18:30,800 --> 00:18:34,040 To go and solve these problems. And that's really what my book 443 00:18:34,040 --> 00:18:35,280 is mainly about. Is going and. 444 00:18:35,480 --> 00:18:37,680 Breaking the dependencies and writing the tests going to do 445 00:18:37,680 --> 00:18:40,360 these things, but at a micro level that's what you kind of 446 00:18:40,360 --> 00:18:43,080 need to do, right? I think in the more macro level, 447 00:18:43,080 --> 00:18:45,440 the important thing to go and remember is that essentially 448 00:18:45,960 --> 00:18:49,400 this is the kind of stuff where you're never going to go and 449 00:18:49,400 --> 00:18:51,840 have complete tests on the code base unless you started that. 450 00:18:51,840 --> 00:18:52,480 Way. Right. 451 00:18:53,040 --> 00:18:54,640 And that it's OK to live in this limbo. 452 00:18:54,640 --> 00:18:56,840 Space, and you probably will for a long period of time. 453 00:18:57,120 --> 00:18:59,480 Where you're going to have some areas of great coverage in other 454 00:18:59,480 --> 00:19:03,200 areas, so not so good coverage and that's just normal, right? 455 00:19:03,560 --> 00:19:06,560 But at the end of the day, the coolest thing is that the things 456 00:19:06,560 --> 00:19:08,520 that you're about to change right now, chances are you'll be 457 00:19:08,520 --> 00:19:11,000 changing in the future. So what you do to go and get 458 00:19:11,000 --> 00:19:13,560 tests in place for those things, it's really kind of valuable if 459 00:19:13,560 --> 00:19:15,400 you need to go and actually start to make changes later. 460 00:19:15,400 --> 00:19:17,800 And chances are you will. So you're going to start to get 461 00:19:17,800 --> 00:19:20,360 like very quick returns from doing this sort of thing. 462 00:19:20,400 --> 00:19:23,600 Like Joe be thinking, you know, one day that's like, wow, I need 463 00:19:23,600 --> 00:19:26,280 to go and get in this area of code and there are many tests or 464 00:19:26,280 --> 00:19:29,360 the tests are very good and it's going to be a pain when I do 465 00:19:29,360 --> 00:19:30,640 that tomorrow. And then you get there tomorrow 466 00:19:30,640 --> 00:19:31,840 and you're. Always like oh damn wow, one. 467 00:19:32,320 --> 00:19:34,240 Of my team mates was in there three weeks ago and they put 468 00:19:34,240 --> 00:19:37,240 tests in place and wow, yay, my job is easier than it was 469 00:19:37,240 --> 00:19:39,560 before. So I think there's that, but 470 00:19:39,560 --> 00:19:42,640 there's also I guess the psychological thing of basically 471 00:19:42,640 --> 00:19:44,720 doing that enough to get to the point where you notice the 472 00:19:44,720 --> 00:19:47,720 difference about how it feels. We have an area of code that's 473 00:19:47,720 --> 00:19:49,840 easier to work in versus one that doesn't. 474 00:19:50,400 --> 00:19:53,520 Quite often I run into people that have, you know, even 475 00:19:53,880 --> 00:19:56,640 independent of whether you have tests or not, they're not used 476 00:19:56,640 --> 00:19:58,600 to working a good code. Base at all and so. 477 00:19:58,800 --> 00:20:00,840 They're in pain, but they're in pain, right? 478 00:20:00,840 --> 00:20:02,720 And that's really a tragic situation. 479 00:20:02,720 --> 00:20:04,960 To be in. Yeah, speaking about some of the 480 00:20:04,960 --> 00:20:07,680 techniques you mentioned, like dependency breaking and being in 481 00:20:07,680 --> 00:20:10,440 the limbo situation, you have this one concept called seam, 482 00:20:10,440 --> 00:20:11,600 right? So I think that's really a 483 00:20:11,600 --> 00:20:14,000 powerful thing. Maybe if you can, you know, 484 00:20:14,000 --> 00:20:16,920 define what is seam and how should people use it within 485 00:20:16,920 --> 00:20:18,720 their code base. Sure. 486 00:20:18,840 --> 00:20:21,360 It was kind of funny because I thought of this and I actually 487 00:20:21,360 --> 00:20:23,760 wrote a chapter on my book. They came very close from going 488 00:20:23,760 --> 00:20:26,200 and pulling it out because I thought it was just a peculiar 489 00:20:26,200 --> 00:20:28,600 way that I'm seeing software and it wouldn't be beneficial to 490 00:20:28,600 --> 00:20:30,760 anybody. What it comes down to is this, 491 00:20:30,760 --> 00:20:33,440 and essentially when you look at a class or a set of classes, 492 00:20:34,000 --> 00:20:36,360 there are places where it's easier to go and actually sort 493 00:20:36,360 --> 00:20:40,160 of replace things, replace behavior than other places, 494 00:20:40,240 --> 00:20:42,640 right? So let's say you have a method 495 00:20:42,640 --> 00:20:46,080 and it's basically doing something crazy with IO and you 496 00:20:46,080 --> 00:20:47,840 want to get it out of your way so you can write tests for the 497 00:20:47,840 --> 00:20:49,640 class without actually doing the IO, right? 498 00:20:50,160 --> 00:20:52,800 If it's a virtual method, right, you can basically go ahead and 499 00:20:52,800 --> 00:20:55,400 override it in the subclass and then it's out of your way. 500 00:20:55,960 --> 00:20:58,600 So that part of the method is a scene. 501 00:20:58,600 --> 00:21:01,520 It's kind of like the seam on your shirt is a natural 502 00:21:01,520 --> 00:21:02,320 breaking. Point. 503 00:21:02,600 --> 00:21:05,320 We can actually go and replace one thing with another, replace 504 00:21:05,320 --> 00:21:08,080 one behavior with another. It's interesting about this 505 00:21:08,080 --> 00:21:09,960 because if you start to categorize all the seams, 506 00:21:09,960 --> 00:21:11,440 there's just like lots and. Lots of them. 507 00:21:12,000 --> 00:21:15,040 But it's an opportunistic way of looking at software. 508 00:21:15,040 --> 00:21:18,880 You go and say, what can I do to actually without changing things 509 00:21:18,880 --> 00:21:22,560 drastically, be able to go and get areas where I can get 510 00:21:22,560 --> 00:21:25,560 understanding and coverage? Because quite often it's like 511 00:21:25,560 --> 00:21:27,560 you're, you know, if you're going to make a change and you 512 00:21:27,560 --> 00:21:30,280 don't have all day to go and like refactor everything, you're 513 00:21:30,640 --> 00:21:33,880 going to have to go and get that change in in a way which is not 514 00:21:33,880 --> 00:21:37,200 going to be disruptive has very high quality things along those 515 00:21:37,240 --> 00:21:39,120 lines. And it just might be a spot 516 00:21:39,120 --> 00:21:40,480 change and it's good to know where that's. 517 00:21:40,480 --> 00:21:44,800 Possible so. Is this same concept similar to 518 00:21:44,800 --> 00:21:48,000 the kind of strategy when people doing micro service, you know, 519 00:21:48,160 --> 00:21:51,040 or re architecting certain parts of the API or their service? 520 00:21:51,920 --> 00:21:53,480 Not really. I mean, to me, it's like with 521 00:21:53,480 --> 00:21:56,000 Strangler, it's kind of like you're how what you're doing is 522 00:21:56,000 --> 00:21:57,440 you're kind of like building things in parallel. 523 00:21:57,440 --> 00:21:59,760 You're diverting flow to 1 area of the system so you can go 524 00:21:59,760 --> 00:22:03,520 build out another one. And it's like my favorite before 525 00:22:03,520 --> 00:22:05,560 it was even sort of like written up as a pattern. 526 00:22:05,560 --> 00:22:09,000 It's like I used to live in Miami, FL and there's like this 527 00:22:09,000 --> 00:22:12,240 giant bridge, this seven mile bridge that goes from South. 528 00:22:12,240 --> 00:22:14,600 Florida to Key West. And it was kind of funny because 529 00:22:14,600 --> 00:22:17,360 many years ago what they did is they're kind of like, OK, we've 530 00:22:17,360 --> 00:22:19,920 got the old one, we're going to build a new one next to it, and 531 00:22:19,920 --> 00:22:21,840 then we're going to go and tear down the old one, right? 532 00:22:22,320 --> 00:22:24,200 And it's just, you know, I looked this up one time. 533 00:22:24,200 --> 00:22:26,880 I think it's like parallel replacement is the term that 534 00:22:26,880 --> 00:22:29,720 people use in other engineering disciplines, you know, this kind 535 00:22:29,720 --> 00:22:30,960 of thing. But generally, that's what 536 00:22:31,000 --> 00:22:34,040 strangler fig is. It's just this thing of going 537 00:22:34,040 --> 00:22:37,720 actually sort of selectively rewriting areas by going to 538 00:22:37,720 --> 00:22:40,240 allow yourself to go and have the old path to fall back on. 539 00:22:40,600 --> 00:22:43,400 So in terms of seams, I guess you do have to find seams in 540 00:22:43,400 --> 00:22:45,600 order to actually understand where you can make these 541 00:22:45,600 --> 00:22:47,040 diversion points. So I guess that's the 542 00:22:47,040 --> 00:22:50,400 commonality, right? Thanks for the clarification. 543 00:22:50,840 --> 00:22:54,160 One thing that is always very tightly coupled right with the 544 00:22:54,280 --> 00:22:57,360 legacy code is this practice called refactoring or you know, 545 00:22:57,360 --> 00:22:59,320 writing tests, maybe TDD as well. 546 00:22:59,600 --> 00:23:03,440 So maybe in your view, right, is it very important for us to be 547 00:23:03,440 --> 00:23:06,720 able to do refactoring right or refactoring well? 548 00:23:07,000 --> 00:23:09,840 Should we also read Martin Fowler's book in order to be 549 00:23:09,840 --> 00:23:11,320 able to tackle legacy code better? 550 00:23:12,160 --> 00:23:13,960 Totally. I think it's really, it's a 551 00:23:13,960 --> 00:23:15,440 great book and it's kind of funny. 552 00:23:15,880 --> 00:23:17,880 My daughter many years ago had a friend who was seen by getting 553 00:23:17,880 --> 00:23:20,520 into programming and he's like, oh, what should I do? 554 00:23:20,520 --> 00:23:22,960 What should I do when I give him a copy of the book refactoring. 555 00:23:23,600 --> 00:23:25,760 And the reason I did this was kind of like there was, I think 556 00:23:26,440 --> 00:23:28,880 it was Ralph Johnson who said this many, many years ago when 557 00:23:28,880 --> 00:23:31,520 the book first came out. He said it's kind of interesting 558 00:23:31,520 --> 00:23:34,360 that basically if you want to teach good design, basically 559 00:23:34,360 --> 00:23:37,160 teach people the things to avoid and everything that's left is 560 00:23:37,160 --> 00:23:39,200 good design. And that's really the way the 561 00:23:39,200 --> 00:23:41,160 book is organized. It's kind of like OK, here's all 562 00:23:41,160 --> 00:23:42,040 these code. Smells. 563 00:23:42,640 --> 00:23:44,920 And your job is to go and look at these code smokes and get rid 564 00:23:44,920 --> 00:23:46,840 of them. And if you do, you know you're 565 00:23:46,840 --> 00:23:48,320 done to be better than it was before. 566 00:23:48,320 --> 00:23:51,040 It's a great way of going and conveying design sense, you 567 00:23:51,040 --> 00:23:53,160 know, kind of like what is better than the other thing, 568 00:23:53,200 --> 00:23:55,840 right? So I think it's a very valuable 569 00:23:55,840 --> 00:23:57,760 thing just in terms of understanding systems. 570 00:23:58,360 --> 00:24:01,720 But in general terms of dealing with systems, it's extremely 571 00:24:01,720 --> 00:24:05,200 valuable because refactoring is all about clarifying things to 572 00:24:05,200 --> 00:24:07,800 the point where basically you can understand them and where 573 00:24:07,800 --> 00:24:10,960 change becomes easier. It's not just refactoring 574 00:24:10,960 --> 00:24:14,440 itself, but understanding my predecessors like the code 575 00:24:14,440 --> 00:24:17,480 smells I mentioned, but also design principles so you kind of 576 00:24:17,480 --> 00:24:19,560 understand. What direction to carry things 577 00:24:19,560 --> 00:24:21,520 in. And once you have that, you 578 00:24:21,520 --> 00:24:23,640 know, you're going to just be able to go and sort of like make 579 00:24:23,640 --> 00:24:26,360 things better for yourself in the future and the people around 580 00:24:26,360 --> 00:24:27,760 you. I mean, that's a wonderful 581 00:24:27,760 --> 00:24:30,000 thing. In terms of the structure of 582 00:24:30,040 --> 00:24:34,080 automated tests, do you have any suggestion or should we also 583 00:24:34,080 --> 00:24:36,920 still refer to the like test pyramid or maybe something 584 00:24:36,920 --> 00:24:38,920 around that? Or is there any kind of like 585 00:24:39,120 --> 00:24:41,360 suggestion that you would advocate people when dealing 586 00:24:41,360 --> 00:24:43,720 with legacy code? Well, I think for the most part, 587 00:24:43,720 --> 00:24:46,440 1 of the things that's kind of funny is that there was like BDD 588 00:24:46,440 --> 00:24:48,800 behavior driven development that came out many years ago, right? 589 00:24:48,800 --> 00:24:50,880 And one of the interesting things there is they were 590 00:24:50,880 --> 00:24:53,480 concentrating on like high level behaviors and sort of writing 591 00:24:53,480 --> 00:24:56,040 tests that would deal with collaborations of classes at. 592 00:24:56,040 --> 00:24:58,200 Once and it. Took me a long while to realize 593 00:24:58,200 --> 00:25:00,280 that, you know, I'd run into people all the time. 594 00:25:00,280 --> 00:25:03,520 They're writing like one test case class per test per class, 595 00:25:03,520 --> 00:25:05,640 right? And that you're in this weird 596 00:25:05,640 --> 00:25:08,400 situation with that. But invariably, you know, when 597 00:25:08,400 --> 00:25:10,880 I'm working with legacy systems, I'm trying to find some vantage 598 00:25:10,880 --> 00:25:13,680 point I can write tests at, which isn't too far from the 599 00:25:13,680 --> 00:25:16,560 behavior I'm actually going and covering, but goes and covers, 600 00:25:16,600 --> 00:25:18,480 you know, the behaviors that I care about, right? 601 00:25:19,000 --> 00:25:21,800 And these clusters are really more than just one class quite 602 00:25:21,800 --> 00:25:23,680 often, right? That's just the way things go. 603 00:25:24,080 --> 00:25:27,640 So I think that the conversation we have about whether something 604 00:25:27,640 --> 00:25:31,480 is a unit test or an integration test is not terribly productive 605 00:25:31,480 --> 00:25:32,640 in a way. I think it's really about 606 00:25:32,640 --> 00:25:35,200 testing behaviors. And you know, there's many other 607 00:25:35,200 --> 00:25:36,480 people saying this as well as I am. 608 00:25:37,200 --> 00:25:39,520 It's just a really valuable thing to go and look at is that 609 00:25:39,520 --> 00:25:42,080 you're testing behaviors. So you need to go and find the 610 00:25:42,080 --> 00:25:44,640 places where it's best to do that and kind of move forward 611 00:25:44,640 --> 00:25:46,920 with things. So testing pyramid, you've got a 612 00:25:46,920 --> 00:25:49,040 unit and you go all the way up to like acceptance. 613 00:25:49,040 --> 00:25:52,040 And stuff like that. I guess the thing to say is that 614 00:25:52,400 --> 00:25:54,560 you can. Basically, the lower level tests 615 00:25:54,560 --> 00:25:55,720 that you have are things that you build. 616 00:25:55,720 --> 00:25:57,640 Upon in terms of understanding the system. 617 00:25:58,280 --> 00:26:00,200 And there's no way. Generally to go and have tests 618 00:26:00,200 --> 00:26:02,680 at an extremely high level and cover everything you care about, 619 00:26:02,720 --> 00:26:04,840 right? So it's kind of like I try to 620 00:26:04,840 --> 00:26:07,680 concentrate on the lower levels and then know that's to make the 621 00:26:07,680 --> 00:26:09,400 higher levels. Easier when I need to go 622 00:26:09,400 --> 00:26:11,240 approach to this. Yep. 623 00:26:11,560 --> 00:26:14,480 So what about if you mentioned something very interesting about 624 00:26:14,480 --> 00:26:16,000 behavior, change of behavior, right. 625 00:26:16,000 --> 00:26:18,840 So I think the business most likely will demand a lot of 626 00:26:18,840 --> 00:26:21,800 changes, right? And some program actually are 627 00:26:21,800 --> 00:26:25,280 not touched because some business does not demand it to 628 00:26:25,280 --> 00:26:27,680 change, right? What do you think in terms of 629 00:26:27,680 --> 00:26:31,120 those kind of code, right, for when it happened that we have to 630 00:26:31,120 --> 00:26:33,280 change the code, but there is some limited test. 631 00:26:33,600 --> 00:26:36,560 Is this something that is also common in your consulting 632 00:26:36,560 --> 00:26:38,640 experience, right? How should people think about 633 00:26:38,920 --> 00:26:41,560 code that has been sitting around still producing value, 634 00:26:41,880 --> 00:26:43,160 but one day it has to change, right? 635 00:26:43,160 --> 00:26:45,600 But it's so difficult to change because maybe people have left. 636 00:26:46,440 --> 00:26:48,960 I, I think generally I'm really, you know, in favor of going, 637 00:26:48,960 --> 00:26:51,400 actually just really trying to go and make your changes 638 00:26:51,400 --> 00:26:54,520 supported by tents, OK. Because even though today it 639 00:26:54,520 --> 00:26:56,600 looks like nobody's going to change this and stuff like that, 640 00:26:56,600 --> 00:26:58,440 it's kind of like it's just the safe thing to do. 641 00:26:58,920 --> 00:27:01,120 The fact that you have that code around still means enhanced 642 00:27:01,120 --> 00:27:02,880 value. So you don't want to be in a 643 00:27:02,880 --> 00:27:04,680 situation where you're just making it worse. 644 00:27:04,760 --> 00:27:06,400 Right, but. Of course people are going to be 645 00:27:06,400 --> 00:27:08,240 pragmatic about this. It really depends upon how much 646 00:27:08,240 --> 00:27:10,760 other work they have in front of them and stuff like this. 647 00:27:10,760 --> 00:27:14,520 But you know, I think even going through the exercise of like, 648 00:27:14,840 --> 00:27:16,640 trying to write tests for something which doesn't have 649 00:27:16,640 --> 00:27:19,040 many tests and it's really, really difficult, gives you an 650 00:27:19,040 --> 00:27:22,160 appreciation for, you know, it's like, OK, I'm never going to do 651 00:27:22,160 --> 00:27:23,480 this again. And none of my team mates. 652 00:27:23,480 --> 00:27:25,400 Are because we know how bad this is, right? 653 00:27:26,000 --> 00:27:27,840 So it's really always a valuable exercise. 654 00:27:27,840 --> 00:27:30,040 Try to get the test in place to support the change you're going 655 00:27:30,040 --> 00:27:31,720 to make the. Other thing I want to go and 656 00:27:31,720 --> 00:27:33,720 mention with this is that this is not an uncommon thing. 657 00:27:33,720 --> 00:27:36,480 It's kind of like, you know, people in business all the time 658 00:27:36,480 --> 00:27:38,560 are trying to figure out how much to invest in particular 659 00:27:38,560 --> 00:27:41,160 things in terms of like investing in the, you know, the 660 00:27:41,600 --> 00:27:43,400 current line of business, basically going and creating new 661 00:27:43,400 --> 00:27:44,680 lines of. Business and stuff along those 662 00:27:44,720 --> 00:27:46,640 lines. And we have the same thing as 663 00:27:46,640 --> 00:27:48,720 technical people. There are no easy answers. 664 00:27:48,720 --> 00:27:51,400 To this question, I think it's. Important to know that just 665 00:27:51,440 --> 00:27:53,600 going into it, you know? Yeah. 666 00:27:53,840 --> 00:27:57,120 So one thing that could help us these days is AI, right? 667 00:27:57,120 --> 00:27:59,560 So if people dealing with legacy code, maybe some people will 668 00:27:59,560 --> 00:28:02,920 play around and try to explain this part of the code base, 669 00:28:02,920 --> 00:28:04,760 right, so that you can actually understand it. 670 00:28:04,960 --> 00:28:07,560 So I know that you're also writing this book AI Assisted 671 00:28:07,560 --> 00:28:09,480 Programming, right? So I assume that you have been 672 00:28:09,480 --> 00:28:12,360 playing a lot with AI recently, so maybe let's go there. 673 00:28:12,640 --> 00:28:16,480 How do you think the role of AI in terms of tackling legacy 674 00:28:16,480 --> 00:28:18,000 code? Maybe that's the first question 675 00:28:18,000 --> 00:28:20,320 before we talk about anything in general about AI. 676 00:28:21,160 --> 00:28:23,720 Well, it's really kind of hard to go and actually sort of see 677 00:28:23,720 --> 00:28:25,280 where things are going to go because they're changing all the 678 00:28:25,280 --> 00:28:27,040 time, right? And I think the thing that 679 00:28:27,040 --> 00:28:29,120 people are most concerned with right now is the hallucination 680 00:28:29,120 --> 00:28:31,440 problem where it's kind of like you ask a question, you get an 681 00:28:31,440 --> 00:28:33,000 answer, and you just really don't know whether it's right or 682 00:28:33,000 --> 00:28:35,280 not. And I want to take that as a 683 00:28:35,280 --> 00:28:36,480 given. It's going to get better over 684 00:28:36,480 --> 00:28:37,680 time. But I think the thing that's 685 00:28:37,680 --> 00:28:41,120 important is really to go and understand that if you can't 686 00:28:41,120 --> 00:28:44,080 trust what's being generated, there still is value there. 687 00:28:44,080 --> 00:28:46,120 And I think a lot of this is really about going exploring 688 00:28:46,120 --> 00:28:48,480 alternatives. So it's, you know, if I'm about 689 00:28:48,480 --> 00:28:51,640 to do something, quite often I'm like, OK, well, let me go ahead 690 00:28:51,640 --> 00:28:53,160 and ask for a different way of doing this. 691 00:28:53,760 --> 00:28:55,560 OK, let me ask for three different ways of doing this, 692 00:28:55,680 --> 00:28:58,920 OK, And just compare them side by side and understand what's 693 00:28:58,920 --> 00:29:00,800 possible in this space. And quite often good ideas I 694 00:29:00,800 --> 00:29:03,640 wouldn't have gotten otherwise for things that we can verify 695 00:29:03,640 --> 00:29:06,480 easily, like going and asking a question on a particular. 696 00:29:06,480 --> 00:29:08,120 API and gives you the answer in the course. 697 00:29:08,680 --> 00:29:10,560 You put in place and you find it very quickly whether it works or 698 00:29:10,560 --> 00:29:11,760 not. It's kind of, you know, 699 00:29:11,760 --> 00:29:15,200 brilliant for that. But yeah, I do use it to explain 700 00:29:15,200 --> 00:29:17,600 code to me quite often. And I think, God, one thing I 701 00:29:17,600 --> 00:29:20,760 love is this thing where you can sort of like, I call it lensing 702 00:29:20,760 --> 00:29:24,200 and it's kind of like you drop in a giant area of code and 703 00:29:24,200 --> 00:29:26,840 you're kind of like, OK, this giant class that I have here, 704 00:29:26,840 --> 00:29:28,280 what responsibilities does it have? 705 00:29:28,520 --> 00:29:30,920 Right? And it goes and gives me a list 706 00:29:30,920 --> 00:29:32,320 of them. And usually it's kind of like 707 00:29:32,600 --> 00:29:34,840 so, so and you start to interrogate and you ask other 708 00:29:34,840 --> 00:29:36,600 questions about it. But I love to do this thing 709 00:29:36,600 --> 00:29:40,120 where you're like, OK, you gave me 10 responsibilities, just 710 00:29:40,120 --> 00:29:42,400 give me 6. Basically, if you had to look in 711 00:29:42,400 --> 00:29:44,720 this frost, it's kind of like, what are the six most important 712 00:29:44,720 --> 00:29:47,400 responsibilities? And by scaling back and forth 713 00:29:47,400 --> 00:29:50,200 with this, you start to go and understand, you know, where 714 00:29:50,200 --> 00:29:53,320 perception can change of a class in terms of what the 715 00:29:53,320 --> 00:29:55,200 responsibilities are depending upon the level that you're 716 00:29:55,200 --> 00:29:57,840 looking at, right? And it can give you ideas for a 717 00:29:57,840 --> 00:30:00,120 factor. And so I find that very, very 718 00:30:00,120 --> 00:30:02,080 useful. In a strategy so. 719 00:30:02,560 --> 00:30:04,080 There's a ton of other little things like that. 720 00:30:04,160 --> 00:30:05,200 They're kind of, you know, handy. 721 00:30:05,240 --> 00:30:07,680 Yeah. So I think these days a lot of 722 00:30:07,680 --> 00:30:10,320 programmers have been playing around with AII personally also 723 00:30:10,320 --> 00:30:11,920 have used it. So I think you mentioned about 724 00:30:11,920 --> 00:30:14,800 hallucination problem, right? I think it's also one of the 725 00:30:14,800 --> 00:30:17,280 risks. Would AI also create more legacy 726 00:30:17,280 --> 00:30:18,960 code? Because like some people will 727 00:30:18,960 --> 00:30:21,840 just take it as it is, right? And then just plug it in and 728 00:30:21,880 --> 00:30:24,160 maybe deploy it without even of course has to verify it. 729 00:30:25,080 --> 00:30:27,600 Of course there's something it's the get clear report. 730 00:30:27,600 --> 00:30:29,440 I think one of the things that they were mentioning there is it 731 00:30:29,440 --> 00:30:31,920 seems like there's a lot more churn of code. 732 00:30:31,920 --> 00:30:34,480 Code's being put in public repos much faster. 733 00:30:34,840 --> 00:30:37,240 There's much more churn change after task fund, right? 734 00:30:37,680 --> 00:30:39,320 Yeah. I mean, we're just going to go 735 00:30:39,320 --> 00:30:41,840 and end up in the space for a period of time and everything. 736 00:30:42,320 --> 00:30:44,400 The thing that's tough is that it's kind of like it's shifting 737 00:30:44,400 --> 00:30:45,200 the balance. In a way. 738 00:30:45,200 --> 00:30:48,120 It's kind of like it's easier to write code and it's a bit hard 739 00:30:48,120 --> 00:30:50,320 to understand. And I think for me, I'm always 740 00:30:50,320 --> 00:30:52,600 dealing with this fatigue issue. It's like I'll look for three. 741 00:30:52,600 --> 00:30:53,800 Different ways of doing something. 742 00:30:54,320 --> 00:30:56,840 And then I'm doing code review for all these things and it's 743 00:30:56,840 --> 00:30:58,760 kind of like, at what point is that? 744 00:30:58,760 --> 00:30:59,880 Productive. And what point is that 745 00:30:59,880 --> 00:31:04,400 counterproductive for me, right? But it's kind of like we have to 746 00:31:04,520 --> 00:31:06,200 review code that other people have written or that we've 747 00:31:06,200 --> 00:31:08,960 written a long time ago and have forgotten, and this just goes 748 00:31:08,960 --> 00:31:10,920 and increases the burden in. That space and it's kind of 749 00:31:10,920 --> 00:31:12,800 tough I. Was just saying an area that I 750 00:31:12,800 --> 00:31:14,440 think is really kind of interesting is the entire thing 751 00:31:14,440 --> 00:31:16,120 of going and writing tests. Right, because. 752 00:31:16,760 --> 00:31:19,120 It's kind of shocked at the beginning to see her vendors 753 00:31:19,120 --> 00:31:20,800 basically go and say, hey, we're going to write all your tests 754 00:31:20,800 --> 00:31:21,280 for you. Yeah. 755 00:31:21,440 --> 00:31:23,480 You know, as a TDD guy, you're looking at this and you're 756 00:31:23,480 --> 00:31:25,840 saying, oh, my God, no, Right. But for characterization 757 00:31:25,840 --> 00:31:28,560 testing, it's really kind of powerful because you know if 758 00:31:28,560 --> 00:31:32,560 you're not changing the code right, it is a document of what 759 00:31:32,560 --> 00:31:36,120 the behavior currently is, and you can go and write have. 760 00:31:36,120 --> 00:31:39,320 Tests written for this. And if they pass, then they are 761 00:31:39,320 --> 00:31:41,680 actually showing you, you know, real behavior. 762 00:31:42,200 --> 00:31:44,040 And so that's useful. I think the thing we have to get 763 00:31:44,040 --> 00:31:46,560 used to with this is the idea that tests can be. 764 00:31:46,560 --> 00:31:49,280 Disposable and temporary sometimes and just using the get 765 00:31:49,280 --> 00:31:51,920 insight or to facilitate change. So I think that's a very 766 00:31:51,920 --> 00:31:53,200 valuable. Thing, right. 767 00:31:53,200 --> 00:31:55,480 So you mentioned about generating test cases, right? 768 00:31:55,480 --> 00:31:58,040 And also interrogating code, explaining code, right. 769 00:31:58,280 --> 00:32:01,360 What are some of the other use cases that you typically use 770 00:32:01,360 --> 00:32:03,880 when using AI in your maybe day-to-day work? 771 00:32:04,800 --> 00:32:06,280 Well, all the time. It's just for exploring 772 00:32:06,280 --> 00:32:07,920 alternatives. You know, for the most part. 773 00:32:08,000 --> 00:32:10,080 I mean, it's like even things that have nothing to do with 774 00:32:10,080 --> 00:32:11,800 code. It's kind of like just, you 775 00:32:11,800 --> 00:32:13,560 know, it's an interesting thing to admit. 776 00:32:13,560 --> 00:32:15,120 I'm just so curious. About things that I have a 777 00:32:15,120 --> 00:32:17,840 thought about. You know, something in culture 778 00:32:17,840 --> 00:32:19,200 or something and you know, kind of like. 779 00:32:19,200 --> 00:32:21,640 Science and it's like. Damn, it's like, you know, it's 780 00:32:21,640 --> 00:32:25,800 like, how long is the manta rays, you know, Fin. 781 00:32:25,840 --> 00:32:28,400 In the front, you know, it's like, oh, it's yeah, they're 782 00:32:28,440 --> 00:32:31,120 interesting kind of thing. So it's, yeah, it's an 783 00:32:31,120 --> 00:32:33,600 interesting thing. I just, I think in general, any 784 00:32:33,600 --> 00:32:37,040 domain that you want to dig into, it's easier to do now, you 785 00:32:37,040 --> 00:32:40,600 know, the advent of search browsers gave us a lot more 786 00:32:40,600 --> 00:32:42,400 ability to go and. Discover things and sort of 787 00:32:42,400 --> 00:32:44,520 learn things. And now it's off the charts. 788 00:32:44,520 --> 00:32:47,440 So I think that's just really, you know, incredible so. 789 00:32:48,080 --> 00:32:50,560 You're writing this book AI Assisted Programming, so I think 790 00:32:50,560 --> 00:32:53,080 it's still in progress, right? Very early in the process, 791 00:32:53,080 --> 00:32:55,240 right? So what made you decide to write 792 00:32:55,240 --> 00:32:58,880 a book about AI programming? Well, it seems like a very 793 00:32:58,880 --> 00:33:01,120 strange thing to do considering the how fast things are moving. 794 00:33:01,120 --> 00:33:04,040 But I think what was interesting to me is I felt like it was an 795 00:33:04,040 --> 00:33:06,840 interesting way of going and actually talking more about how 796 00:33:06,840 --> 00:33:10,400 we work in a way. And I felt like there are some 797 00:33:10,400 --> 00:33:12,320 of the things we confront with AI are things we're confronting 798 00:33:12,320 --> 00:33:13,480 with right now, but haven't we talked? 799 00:33:13,480 --> 00:33:14,560 About all that. That's right. 800 00:33:15,240 --> 00:33:16,920 And you know one of those is that. 801 00:33:16,920 --> 00:33:19,280 Fatigue issue that I mentioned a little bit earlier, right? 802 00:33:19,920 --> 00:33:22,080 Let's see, another 1 is really the thing that. 803 00:33:22,080 --> 00:33:24,640 Sometimes you know it's. But often you're working in an 804 00:33:24,640 --> 00:33:26,720 ID and basically gives you code suggestions all the time. 805 00:33:26,720 --> 00:33:29,200 And you're looking at this as kind of like, can I accept this? 806 00:33:29,360 --> 00:33:30,840 How much trust can I place in this? 807 00:33:31,280 --> 00:33:33,240 And the same thing is true with code that you get from somebody 808 00:33:33,240 --> 00:33:34,680 else, right? Where is my trust? 809 00:33:35,000 --> 00:33:36,640 How do I focus on this particular thing? 810 00:33:37,160 --> 00:33:38,960 And then there's a general thing of ideation. 811 00:33:38,960 --> 00:33:41,320 You know, one of the sections I have in the book talks about the 812 00:33:41,320 --> 00:33:44,320 lost potential. Basically, you ask the question 813 00:33:44,320 --> 00:33:46,080 of the AI and you accept the first answer. 814 00:33:46,760 --> 00:33:49,440 Because quite often you just ask this prompt again, you get a 815 00:33:49,440 --> 00:33:50,680 different. Answer and a different answer 816 00:33:50,680 --> 00:33:53,720 and a different answer. And I find that so valuable 817 00:33:53,720 --> 00:33:54,760 because it's sort of open to the. 818 00:33:54,760 --> 00:33:57,440 Possibility space in a way. And gets you thinking about 819 00:33:57,440 --> 00:33:59,480 different things that pokes your mind in a different. 820 00:33:59,480 --> 00:34:01,160 Direction. Which quite often gives you 821 00:34:01,160 --> 00:34:03,120 better ideas. About things and I think that's. 822 00:34:03,640 --> 00:34:05,640 You know, a great thing. The other thing as well is that 823 00:34:05,640 --> 00:34:09,120 basically it makes us better question askers, right? 824 00:34:09,120 --> 00:34:11,040 We get better at it because we have to go and sort of ask 825 00:34:11,040 --> 00:34:15,159 questions to interact with it. So it's really a capability 826 00:34:15,159 --> 00:34:18,040 amplifier. So do you think this prom 827 00:34:18,040 --> 00:34:20,679 engineering thing is one of the skill set that any developers 828 00:34:20,679 --> 00:34:23,719 now need to be able to master? I think so. 829 00:34:23,719 --> 00:34:25,000 I think it's and it's going to change. 830 00:34:25,000 --> 00:34:27,280 So I think with my book, it's less of that prompt engineering 831 00:34:27,280 --> 00:34:30,639 and more bad approach. But yeah, it's it's interesting. 832 00:34:30,639 --> 00:34:33,719 I think 1 section I haven't written yet, but I gave him an 833 00:34:33,840 --> 00:34:35,239 online lecture. About like a while. 834 00:34:35,239 --> 00:34:38,480 Back it's just having a mental model to have these things work 835 00:34:38,960 --> 00:34:41,199 at a behavioral level so they can help you with your 836 00:34:41,199 --> 00:34:44,120 prompting, right? And just, I can very briefly 837 00:34:44,120 --> 00:34:47,400 say, it's almost like you can imagine you have this giant net, 838 00:34:47,400 --> 00:34:48,840 right? And it's kind of like if you ask 839 00:34:48,840 --> 00:34:51,480 a question, it pulls some of those things towards you. 840 00:34:52,040 --> 00:34:53,760 And then basically you ask another question, it pulls other 841 00:34:53,760 --> 00:34:56,199 things towards you. And quite often the way that you 842 00:34:56,199 --> 00:34:58,040 ask questions doesn't lead to in a different path. 843 00:34:58,040 --> 00:35:00,760 So it's kind of like this thing of going in, like what are you 844 00:35:00,760 --> 00:35:02,760 pulling into active memory when you're going in? 845 00:35:02,960 --> 00:35:04,280 Sort of like asking. Questions. 846 00:35:04,800 --> 00:35:06,920 I think you know little things like that quite often. 847 00:35:06,920 --> 00:35:08,800 Make us better at doing these things. 848 00:35:09,720 --> 00:35:11,680 Yeah. So I think what 1, I mean, one 849 00:35:11,680 --> 00:35:14,400 thing that if we always ask question and we got different 850 00:35:14,400 --> 00:35:16,480 answers, right? Obviously reading the code is 851 00:35:16,480 --> 00:35:18,320 like a big cognitive load, right? 852 00:35:18,320 --> 00:35:20,680 We try to understand different ways of doing thing, although it 853 00:35:20,680 --> 00:35:23,640 maybe end up the same and also at the end of the day, still the 854 00:35:23,640 --> 00:35:26,240 same thing, right? So we gotta write the test for 855 00:35:26,240 --> 00:35:28,040 it. So if we don't actually it 856 00:35:28,040 --> 00:35:29,840 creates a legacy code as well, right? 857 00:35:30,120 --> 00:35:33,160 I think reading code is something that I feel is much 858 00:35:33,160 --> 00:35:35,280 more important these days, right, Rather than actually 859 00:35:35,280 --> 00:35:38,200 writing it because the prompt probably could, you know, use a 860 00:35:38,280 --> 00:35:41,400 natural language and they give us a code and you read the code 861 00:35:41,400 --> 00:35:43,760 instead. So in your view, like how can 862 00:35:43,760 --> 00:35:46,400 people master, you know, reading the code much better? 863 00:35:46,400 --> 00:35:49,360 Or maybe sometimes like the design is something that is 864 00:35:49,760 --> 00:35:52,080 sometimes it's not appropriate with the larger code base, 865 00:35:52,080 --> 00:35:54,080 right. So what's your take about 866 00:35:54,080 --> 00:35:57,440 developers taking AI answer and just putting in well, I just, I 867 00:35:58,000 --> 00:35:59,480 feel like everybody. Knows this, but I have to say. 868 00:35:59,520 --> 00:36:02,320 And so I remember like the story of about like somebody is in 869 00:36:02,320 --> 00:36:04,960 like high school and they're taking mathematics course and 870 00:36:04,960 --> 00:36:07,560 their teacher goes and says to them, you know, there's one 871 00:36:07,560 --> 00:36:10,120 secret for getting an A to this course and everybody's like, oh, 872 00:36:10,120 --> 00:36:12,480 what is it? It's like, do your homework, OK, 873 00:36:12,480 --> 00:36:15,120 do the problems, right? And I think the same thing is 874 00:36:15,120 --> 00:36:16,560 true for us in software development. 875 00:36:16,560 --> 00:36:19,560 The secret for a lot of this stuff is take smaller steps, 876 00:36:19,680 --> 00:36:21,520 right? So if you are looking at 877 00:36:21,520 --> 00:36:24,040 generating something, generate something small, right? 878 00:36:24,040 --> 00:36:27,240 If you're trying to understand something, paste in something 879 00:36:27,240 --> 00:36:29,520 that is kind of small and just get its take on things and 880 00:36:29,520 --> 00:36:32,160 you're just trying to understand what you're bringing, what's 881 00:36:32,160 --> 00:36:34,640 going on with something, look at smaller steps and build up from 882 00:36:34,640 --> 00:36:37,080 the foundations. And yeah. 883 00:36:37,080 --> 00:36:38,600 There's a skill to that sort of thing, I think. 884 00:36:39,160 --> 00:36:41,040 There's another thing which hasn't been written up all that 885 00:36:41,040 --> 00:36:43,000 much that I think it's kind of powerful and it has nothing to 886 00:36:43,000 --> 00:36:45,960 do with AI really. And it comes down to basically 887 00:36:45,960 --> 00:36:49,960 seeing enough systems that you develop enough impressions of 888 00:36:49,960 --> 00:36:51,920 different systems, you start to see what's common and what's 889 00:36:51,920 --> 00:36:54,640 different, right? So there's nothing that's a good 890 00:36:54,640 --> 00:36:58,320 substitute for going and having very wide experience of many 891 00:36:58,320 --> 00:37:00,640 different domains. And that experience of even if 892 00:37:00,640 --> 00:37:03,800 you're not hopping from job to job to job is just, you know, 893 00:37:03,800 --> 00:37:05,360 reading. There's books, there's open 894 00:37:05,360 --> 00:37:08,440 source code and stuff like this. And you start to see how people 895 00:37:08,440 --> 00:37:10,320 approach problems differently in different domains. 896 00:37:10,720 --> 00:37:13,040 And then when you see something which is totally atypical to 897 00:37:13,040 --> 00:37:16,320 you, you have, they call it sometimes the coat hangers. 898 00:37:16,320 --> 00:37:18,520 It's kind of like you're given all this information and you're 899 00:37:18,520 --> 00:37:21,640 looking for the things that are like the basis that you can use 900 00:37:21,640 --> 00:37:23,720 to sort of organize your knowledge as you get more 901 00:37:24,080 --> 00:37:25,360 knowledge of the thing you're looking at. 902 00:37:26,040 --> 00:37:28,200 And yeah, I think that's a valuable skill to develop as 903 00:37:28,280 --> 00:37:30,280 well. So, yeah, you mentioned about 904 00:37:30,280 --> 00:37:32,280 understanding domain. Yeah, I still think that AI 905 00:37:32,280 --> 00:37:33,800 still hasn't reached there yet, right? 906 00:37:33,800 --> 00:37:36,840 So maybe like building the whole application end to end, right? 907 00:37:37,160 --> 00:37:40,960 And also maybe building a more distributed complex systems is 908 00:37:40,960 --> 00:37:42,640 something that they haven't reached out yet. 909 00:37:43,040 --> 00:37:47,440 So in your view, what is the ideal, you know, stage where we 910 00:37:47,440 --> 00:37:51,160 should rely on AI more versus something that use human 911 00:37:51,160 --> 00:37:54,480 creativity or brain more? I think it's really for the 912 00:37:54,480 --> 00:37:57,000 areas which are low risk, you know, and this is kind of funny 913 00:37:57,000 --> 00:37:58,680 because there's all different ways of looking at risk. 914 00:37:59,440 --> 00:38:01,040 And one of the things I think is great is actually going to 915 00:38:01,040 --> 00:38:03,960 develop blower plate code. So it's like thinking well back 916 00:38:03,960 --> 00:38:05,480 about creating like a Chrome extension. 917 00:38:05,480 --> 00:38:08,120 And it's kind of like, OK, just ask me to create like a simple 918 00:38:08,120 --> 00:38:10,080 Chrome extension. And it's kind of like, I don't 919 00:38:10,080 --> 00:38:11,280 need to write all that code. It can. 920 00:38:11,320 --> 00:38:13,200 Basically, I can find it very quickly whether it's working or 921 00:38:13,200 --> 00:38:15,120 not, right? But right, we want to 922 00:38:15,120 --> 00:38:17,440 concentrate on the thing that I want the Chrome extension to do, 923 00:38:17,440 --> 00:38:19,880 for instance. There's that, but there's also 924 00:38:19,880 --> 00:38:21,640 things like that that should be like in house. 925 00:38:21,640 --> 00:38:25,400 Tools and stuff like that. Where you have some confidence 926 00:38:25,400 --> 00:38:28,520 that somebody's going to find a problem before it causes loss of 927 00:38:28,520 --> 00:38:31,640 life or money, right? That's a very important thing, 928 00:38:31,680 --> 00:38:33,520 you know, so I think it's finding these low risk 929 00:38:33,520 --> 00:38:36,640 environments and you know, a place where this actually kind 930 00:38:36,640 --> 00:38:40,120 of works very well is if you are creating tools to go and help 931 00:38:40,120 --> 00:38:41,680 you do. Things with your code for 932 00:38:41,680 --> 00:38:43,960 instance. And I guess you know, also 933 00:38:43,960 --> 00:38:46,880 things like quite often I'm feel like I'm rediscovering shell 934 00:38:46,880 --> 00:38:48,400 programming in a way. I've done a lot of shell 935 00:38:48,400 --> 00:38:50,960 programming in my in the past, but when you think about it, 936 00:38:50,960 --> 00:38:55,560 shell is perfect for AI generation in the sense that 937 00:38:56,000 --> 00:38:58,040 you're basically building from these bigger components, the 938 00:38:58,040 --> 00:39:00,200 commands. And it's very easy to find out 939 00:39:00,200 --> 00:39:03,440 exactly what a command does. It has wider knowledge of these 940 00:39:03,440 --> 00:39:05,360 things than we do. And you can see very quickly 941 00:39:05,360 --> 00:39:08,200 whether what it's producing is right or not right. 942 00:39:08,200 --> 00:39:10,000 And it's usually. Very brief and you can see with 943 00:39:10,000 --> 00:39:12,240 others. So going and doing that for my 944 00:39:12,240 --> 00:39:13,440 work. Quite often it's not a very 945 00:39:13,520 --> 00:39:15,800 valuable thing. So yeah, here's that. 946 00:39:16,760 --> 00:39:18,720 I think one topic associated with AI, with software 947 00:39:18,720 --> 00:39:21,480 programming is that a lot of people are worried about their 948 00:39:21,480 --> 00:39:24,040 jobs, right? Or the roles will evolve quite a 949 00:39:24,040 --> 00:39:27,560 lot such that you are becoming obsolete or redundant, right? 950 00:39:27,840 --> 00:39:31,280 We also see some maybe hype technologies like Devin where 951 00:39:31,360 --> 00:39:34,160 they claim they can write code just by itself. 952 00:39:35,160 --> 00:39:37,960 So what is your view about, you know, all these worry about 953 00:39:37,960 --> 00:39:41,240 developers having to maybe upskill themselves or maybe 954 00:39:41,240 --> 00:39:43,040 getting redundant or things like that. 955 00:39:43,040 --> 00:39:45,640 So is that something that we have to fear about? 956 00:39:46,680 --> 00:39:48,520 Well, I think where people are getting much more realistic now. 957 00:39:48,520 --> 00:39:50,240 I think in the very beginning people start doing and 958 00:39:50,240 --> 00:39:52,560 extrapolating and they just get completely. 959 00:39:52,560 --> 00:39:54,400 Freaked out by the possibilities, right? 960 00:39:54,840 --> 00:39:56,560 But I think many. Developers now have enough 961 00:39:56,560 --> 00:39:58,360 experience what they are, they realize it's not going to 962 00:39:58,360 --> 00:40:01,480 replace many times soon, right? It's basically at the end of the 963 00:40:01,480 --> 00:40:05,440 day, you know, it's kind of like in the legal system, you can't 964 00:40:05,440 --> 00:40:07,360 go and say, oh, the algorithm did it. 965 00:40:07,360 --> 00:40:10,560 And it's like, no, a person who works at a company did this. 966 00:40:10,560 --> 00:40:13,240 At the end of the day, we are responsible for what we do. 967 00:40:13,320 --> 00:40:14,400 Right. And I think that's the thing 968 00:40:14,400 --> 00:40:16,160 that's. You know, really important for 969 00:40:16,160 --> 00:40:17,960 us to go and recognize that there has to be somebody in the 970 00:40:17,960 --> 00:40:20,480 seat of responsibility. You have to understand what 971 00:40:20,480 --> 00:40:23,040 you're creating as well. So I think we're going to have a 972 00:40:23,040 --> 00:40:26,240 place for a long time and I can't remember the name of the 973 00:40:26,240 --> 00:40:26,920 book. I wish I. 974 00:40:26,920 --> 00:40:28,360 Could maybe I'll send it to your show notes. 975 00:40:28,360 --> 00:40:30,960 If you do that, but this guy who basically made the case ages 976 00:40:30,960 --> 00:40:33,400 ago, it's kind of like we just need to see ourselves as problem 977 00:40:33,400 --> 00:40:37,080 solvers and programming is one tool to solve a problem, right? 978 00:40:37,640 --> 00:40:39,920 And sometimes we can solve problems without writing code at 979 00:40:39,920 --> 00:40:42,040 all. And that's great, you know, but 980 00:40:42,040 --> 00:40:44,480 we have to basically do that thing of understanding far 981 00:40:44,480 --> 00:40:47,240 beyond the code, you know, everything around it, what's the 982 00:40:47,240 --> 00:40:49,080 business and all these other things because all that. 983 00:40:49,080 --> 00:40:50,720 Information makes us much more effective. 984 00:40:51,520 --> 00:40:53,280 So I think it's this thing of going and expanding. 985 00:40:53,280 --> 00:40:56,200 Out from our roles to that. And I think a lot of people kind 986 00:40:56,200 --> 00:40:58,320 of recognize that that's just. Where the future is, and I do 987 00:40:58,320 --> 00:41:00,920 too, so. Thanks for, you know, giving 988 00:41:00,920 --> 00:41:03,680 your thoughts about this. So one technique that I found in 989 00:41:03,680 --> 00:41:06,280 some of your talk or resources that you shared before this 990 00:41:06,400 --> 00:41:09,080 conversation is that one thing called pseudo Lang, right? 991 00:41:09,080 --> 00:41:10,840 I find it quite interesting, right? 992 00:41:10,840 --> 00:41:12,960 So I think for some people who may not have heard about it. 993 00:41:13,160 --> 00:41:15,800 So tell us what is pseudo Lang and how can we use it in our, 994 00:41:15,800 --> 00:41:17,280 you know, day-to-day work with AI? 995 00:41:18,240 --> 00:41:19,960 Well, it's an interesting thing. This is going to take a minute 996 00:41:19,960 --> 00:41:22,720 or two to explain, right? But I remember I think I found 997 00:41:22,720 --> 00:41:25,640 it through a web search and Slay wrote a couple of media posts 998 00:41:25,640 --> 00:41:27,720 about a thing they called pseudo Lang. 999 00:41:27,720 --> 00:41:31,360 So it's SUDO Lang, right? Just as Apana. 1000 00:41:31,440 --> 00:41:35,160 You know, what the author was saying was that it'd be great to 1001 00:41:35,160 --> 00:41:38,040 go and actually create a language that's kind of like a 1002 00:41:38,040 --> 00:41:41,120 programming language that basically you can feed into an 1003 00:41:41,360 --> 00:41:45,000 LLM that would go and generate code for you and do translations 1004 00:41:45,000 --> 00:41:48,800 and all these other things. And so he described the language 1005 00:41:48,800 --> 00:41:51,040 and had interesting features like constraints and stuff like 1006 00:41:51,040 --> 00:41:54,200 this and a very familiar syntax. And it wasn't quite like any 1007 00:41:54,200 --> 00:41:56,440 other language. It seemed like it was at a bit 1008 00:41:56,440 --> 00:41:59,480 of a higher level because he could replace greater than with 1009 00:41:59,480 --> 00:42:01,680 like the words greater than, for instance, and all this stuff, 1010 00:42:01,680 --> 00:42:03,560 right? But the further you write down 1011 00:42:03,560 --> 00:42:05,880 in the post, the further you realize what had happened. 1012 00:42:06,480 --> 00:42:10,000 What he did is he asked ChatGPT. Hey, I want to have a language 1013 00:42:10,000 --> 00:42:11,880 which makes it easy for me to talk to you about. 1014 00:42:11,880 --> 00:42:13,440 Programming and get the things that I want. 1015 00:42:14,360 --> 00:42:16,720 Give me a language that does this and actually, you know, 1016 00:42:16,720 --> 00:42:19,800 that's what the system did. He created pseudo like, right? 1017 00:42:20,360 --> 00:42:24,280 And then it was even interesting and more tricky because he said, 1018 00:42:24,880 --> 00:42:26,040 you know, I asked for this again. 1019 00:42:26,040 --> 00:42:28,760 And I asked for it. It's kind of like, can you give 1020 00:42:28,760 --> 00:42:31,240 me a syntax to work for this sort of work even if I didn't 1021 00:42:31,240 --> 00:42:32,320 tell you what the speck of pseudo. 1022 00:42:32,320 --> 00:42:35,320 Language was right. And it's interesting because 1023 00:42:35,320 --> 00:42:37,040 it's like it goes full circle at that point. 1024 00:42:37,040 --> 00:42:39,920 So it's really more like pointing in a direction. 1025 00:42:39,920 --> 00:42:41,520 I don't think it's really like the language. 1026 00:42:41,520 --> 00:42:43,480 To learn or anything? And in the book I'm writing 1027 00:42:43,480 --> 00:42:46,280 right now, I basically talk about something I call pigeon 1028 00:42:46,280 --> 00:42:49,600 specification, which is really kind of like a pigeon language 1029 00:42:49,600 --> 00:42:52,480 is a language you know, that people from different languages 1030 00:42:52,760 --> 00:42:54,240 adopt. They can communicate with each 1031 00:42:54,240 --> 00:42:55,400 other. It's usually some subset. 1032 00:42:55,400 --> 00:42:58,880 Of two languages. But I can write like a test case 1033 00:42:58,880 --> 00:43:03,240 using just a very brief test colon dot dot dot dot, you know 1034 00:43:03,400 --> 00:43:06,600 this, assert that, right? I can type that in and basically 1035 00:43:06,600 --> 00:43:08,840 say, give this to me and J unit and it just gives it to me. 1036 00:43:08,840 --> 00:43:10,880 And sometimes it's wrong. But I'm dealing with something 1037 00:43:10,880 --> 00:43:13,000 small enough I can actually see what the results are, right? 1038 00:43:13,480 --> 00:43:16,960 So I think we're going to see an off lot more of this people 1039 00:43:16,960 --> 00:43:19,600 developing like private languages in much the same way 1040 00:43:19,600 --> 00:43:20,920 that you might with people you work with. 1041 00:43:21,400 --> 00:43:23,960 I remember hearing about a musical group many years ago and 1042 00:43:24,440 --> 00:43:26,360 somebody knew joined the band and they had serious trouble 1043 00:43:26,360 --> 00:43:29,000 because the other musicians in the back could talk to each 1044 00:43:29,000 --> 00:43:31,480 other, winking a nod. And this guy was new. 1045 00:43:31,480 --> 00:43:32,720 He had no. Idea what they were where? 1046 00:43:33,040 --> 00:43:35,000 They were going and what they were doing because these people 1047 00:43:35,000 --> 00:43:36,800 knew they, they knew things about each other. 1048 00:43:36,800 --> 00:43:39,280 And I think we're going to go and get into that symbiotic 1049 00:43:39,280 --> 00:43:41,800 state with the coding assistants that we are using. 1050 00:43:41,920 --> 00:43:46,240 So speaking about symbiotic, you know, relationship, right, will 1051 00:43:46,240 --> 00:43:51,080 AI be appropriate for us to do this TDD cycle where maybe we 1052 00:43:51,080 --> 00:43:54,440 generate test 1st and maybe AI help to verify and also create 1053 00:43:54,440 --> 00:43:57,720 the business code or maybe vice versa that kind of stuff like a 1054 00:43:57,720 --> 00:44:01,280 real pair developer or those virtual, right. 1055 00:44:01,440 --> 00:44:04,200 So is it something that you have seen it working before? 1056 00:44:05,080 --> 00:44:07,160 Yeah, I'm done and I basically do and I quite often get 1057 00:44:07,160 --> 00:44:08,400 frustrated. I've seen people. 1058 00:44:08,400 --> 00:44:11,840 Kind of carry a bit further. Loewen Falco is basically some 1059 00:44:11,840 --> 00:44:14,720 videos ongoing and doing this, but I think it's, it's a 1060 00:44:14,720 --> 00:44:16,360 powerful approach. You just need to get the bugs 1061 00:44:16,360 --> 00:44:19,000 out of it to some degree. The thing is that essentially 1062 00:44:19,000 --> 00:44:21,320 it's like you write tests and then you can go and sort of like 1063 00:44:21,320 --> 00:44:23,040 how to generate code to satisfy the test. 1064 00:44:23,520 --> 00:44:26,560 What I've been finding is that essentially if you give it a 1065 00:44:26,560 --> 00:44:28,920 whole bunch of tests at once, quite often it's going to do a 1066 00:44:28,920 --> 00:44:31,760 reasonably good job for you and you can run the test to see 1067 00:44:31,760 --> 00:44:34,720 whether it has or not, right? But quite often I've done this 1068 00:44:34,720 --> 00:44:37,080 thing where I basically start to go and say, OK, here's a test, 1069 00:44:37,400 --> 00:44:39,160 here's another test, here's another test, here's another 1070 00:44:39,160 --> 00:44:42,920 test and have it generate code. And I'm asking it to go and 1071 00:44:42,920 --> 00:44:47,280 always make sure that when it generates code that it's going 1072 00:44:47,280 --> 00:44:49,440 to satisfy all the tests I've shown so far. 1073 00:44:49,800 --> 00:44:52,800 And it's kind of like a little bit problematic with that right 1074 00:44:52,840 --> 00:44:53,920 now. I'm asked to do with I guess 1075 00:44:54,240 --> 00:44:55,880 context window problems and stuff on your. 1076 00:44:55,880 --> 00:44:58,520 So I think, you know, it's moving forward in a good way. 1077 00:44:58,520 --> 00:45:01,280 I think there's so much you can learn just by going and taking a 1078 00:45:01,280 --> 00:45:03,760 project and going and deleting all the source code and just 1079 00:45:03,760 --> 00:45:06,600 giving all the tests to an AI assistant saying write the code 1080 00:45:06,600 --> 00:45:08,640 for me, right? The things that it's missing are 1081 00:45:08,640 --> 00:45:11,800 things that you were missing in. Coverage for that kind of thing. 1082 00:45:12,320 --> 00:45:14,600 It's, yeah, there's a lot of interesting stuff we can do in 1083 00:45:14,600 --> 00:45:16,440 that space. It's going to get better and I 1084 00:45:16,440 --> 00:45:18,320 think TDD. Is still going to be around so. 1085 00:45:19,280 --> 00:45:21,360 So you mentioned about context window, maybe some people are 1086 00:45:21,360 --> 00:45:24,800 not familiar with the concept and I know that some of the LLM 1087 00:45:24,800 --> 00:45:27,600 newer models always mentioned about this context thing which 1088 00:45:27,600 --> 00:45:29,240 gets larger and larger and larger. 1089 00:45:29,240 --> 00:45:32,000 Tell us what's the difference with having larger context 1090 00:45:32,000 --> 00:45:34,560 window? Well, it's like when you look at 1091 00:45:34,560 --> 00:45:37,120 how an LLM works, it's kind of like it's rather deep and 1092 00:45:37,120 --> 00:45:39,080 involved many layers of neural networks. 1093 00:45:39,560 --> 00:45:42,720 But it seems that at least for a little while ago, when it 1094 00:45:42,720 --> 00:45:44,920 basically goes and does inference like of your prompt 1095 00:45:44,920 --> 00:45:47,920 and processing and things, it's doing almost like a quadratic 1096 00:45:47,920 --> 00:45:50,200 operation. It's just like the more the 1097 00:45:50,200 --> 00:45:52,280 needs in memory, the longer it's going to go and take. 1098 00:45:52,280 --> 00:45:54,760 To go and do things. And so for that reason, it's 1099 00:45:54,760 --> 00:45:57,120 kind of like you have this thing that happens quite often where 1100 00:45:57,720 --> 00:46:00,160 you're working in a session, and then after a period of time, 1101 00:46:00,160 --> 00:46:03,080 you're asking it to do something and then it goes and gives you 1102 00:46:03,080 --> 00:46:05,360 an answer, which makes it obvious that it doesn't know 1103 00:46:05,840 --> 00:46:07,920 what you gave it like 10 minutes ago, right? 1104 00:46:08,480 --> 00:46:10,680 And so it's kind of like, I called this to like performance 1105 00:46:10,680 --> 00:46:13,400 degradation, right? That just has to do with the 1106 00:46:13,400 --> 00:46:15,200 physics pretty much. You know, like how 1107 00:46:15,560 --> 00:46:19,000 computationally complex. These operations happen to be so 1108 00:46:19,000 --> 00:46:21,360 it puts limitations on how much we can go and basically do it 1109 00:46:21,360 --> 00:46:23,120 once. The thing is, I've heard 1110 00:46:23,120 --> 00:46:25,400 recently that some of the newer models, I think Gemini. 1111 00:46:26,080 --> 00:46:28,360 Claims that it has like, you know, an infinite context 1112 00:46:28,360 --> 00:46:30,360 window. And I haven't really kept up in 1113 00:46:30,400 --> 00:46:32,520 that space to go and see whether those things still have that 1114 00:46:32,520 --> 00:46:34,000 issue. I know some of the tools I'm 1115 00:46:34,000 --> 00:46:35,520 using today still have that. Issue. 1116 00:46:36,080 --> 00:46:39,120 But I think it's going to be with us regardless because it's 1117 00:46:39,120 --> 00:46:43,200 very much like, I took a lot of cognitive science in college and 1118 00:46:43,200 --> 00:46:45,200 it's just like the way humans interact with things. 1119 00:46:45,200 --> 00:46:47,760 It's kind of like if I divert your attention to over there, 1120 00:46:48,320 --> 00:46:50,520 it's not going to be immediately top of mind the thing that we 1121 00:46:50,520 --> 00:46:52,600 were just talking about and vice versa, right? 1122 00:46:53,040 --> 00:46:56,600 We have attention and it's relatively fixed because nothing 1123 00:46:56,600 --> 00:46:59,280 can have any attention. It's just it violates physics. 1124 00:46:59,280 --> 00:47:02,160 So we're always going to have that thing of kind of like 1125 00:47:02,560 --> 00:47:04,640 what's in our intent to focus and what is not. 1126 00:47:05,120 --> 00:47:07,720 And I think that basically understanding that's very useful 1127 00:47:08,320 --> 00:47:10,320 to being effective with AI right now. 1128 00:47:11,120 --> 00:47:13,480 So I guess I hope that's, you know, a decent summary. 1129 00:47:14,200 --> 00:47:17,160 Yep, thanks for explaining that. So we have spoken a lot of 1130 00:47:17,160 --> 00:47:19,600 things about AI. Any things that we haven't 1131 00:47:19,600 --> 00:47:23,280 spoken that you figure it out recently that people need to 1132 00:47:23,280 --> 00:47:27,200 know about AI programming. It's funny because in writing 1133 00:47:27,200 --> 00:47:29,640 this, it seems like I've had a lot of topics and it's been like 1134 00:47:29,760 --> 00:47:33,080 digging down deep in them. So recent isn't really, you 1135 00:47:33,080 --> 00:47:34,640 know, the best way of going and putting. 1136 00:47:34,640 --> 00:47:35,520 Here's one. OK. 1137 00:47:35,520 --> 00:47:39,000 And I just released on LinkedIn a couple of days ago a chapter 1138 00:47:39,000 --> 00:47:41,640 when it's writing and it's something I call way words. 1139 00:47:41,640 --> 00:47:43,160 And it's kind of a powerful idea. 1140 00:47:43,920 --> 00:47:45,960 And what it comes down to is this. 1141 00:47:45,960 --> 00:47:49,080 It's kind of like supposing you're asking, you know, the 1142 00:47:49,080 --> 00:47:50,760 tool to go to retract or something, right? 1143 00:47:51,200 --> 00:47:52,720 And you write a prompt to go and do this. 1144 00:47:53,280 --> 00:47:55,720 And then it's kind of like it does send you give it more test 1145 00:47:55,720 --> 00:47:58,560 cases and you discover you need to change the prompt a bit and 1146 00:47:58,560 --> 00:48:01,000 change it a little bit more. At that point, you probably want 1147 00:48:01,000 --> 00:48:02,240 to name this. Thing right? 1148 00:48:02,360 --> 00:48:05,320 You really want to go and say, OK, you know, this is a, an 1149 00:48:05,320 --> 00:48:06,800 operation called da, da, da, da da. 1150 00:48:07,280 --> 00:48:09,920 And then basically you can just in English, you know, or your 1151 00:48:09,920 --> 00:48:13,560 language, just go ahead and just sort of like use that as a piece 1152 00:48:13,560 --> 00:48:15,280 of new nomenclature in terminology. 1153 00:48:15,880 --> 00:48:17,960 And so it's just like introducing a, a method name or 1154 00:48:17,960 --> 00:48:20,080 a variable. One thing you can do sometimes 1155 00:48:20,080 --> 00:48:23,480 is go and say, OK, please, you know, now that I defined this 1156 00:48:23,480 --> 00:48:26,960 thing, give me a summary prompt, give me a prompt that I can use 1157 00:48:26,960 --> 00:48:29,360 to put into another session so that we can go and actually do 1158 00:48:29,360 --> 00:48:32,120 the same operations doesn't always work, but it's getting 1159 00:48:32,120 --> 00:48:34,600 better. But I really find a lot of 1160 00:48:35,120 --> 00:48:37,800 utility in this thing of naming things as you go along when 1161 00:48:37,800 --> 00:48:38,680 you're. Working with things. 1162 00:48:39,320 --> 00:48:41,640 It can just be OK. Well this example I just created 1163 00:48:41,640 --> 00:48:45,120 call this response ABC and then you can go back and say well 1164 00:48:45,480 --> 00:48:47,200 when we did. ABC that's it, right? 1165 00:48:47,200 --> 00:48:49,080 It's going and creating these handles for going and deal with 1166 00:48:49,080 --> 00:48:51,760 things. And yeah, very useful thing to 1167 00:48:51,760 --> 00:48:54,440 go and use. So speaking about this 1168 00:48:54,440 --> 00:48:57,320 nomenclature, right, we just spoken about the context window 1169 00:48:57,320 --> 00:49:00,760 in the different sessions. Do you have to sometimes copy 1170 00:49:00,760 --> 00:49:03,200 and paste the prompts that you had previously in some of the 1171 00:49:03,200 --> 00:49:06,240 notes and then when you work on a different sessions, maybe some 1172 00:49:06,400 --> 00:49:09,680 couple of days later that you have to paste it back, look sort 1173 00:49:09,680 --> 00:49:12,920 of the memory or the cognitive thing to the AI brain and then 1174 00:49:12,920 --> 00:49:14,960 try again. Is that something that is also 1175 00:49:14,960 --> 00:49:17,560 commonly done by people who are using AI? 1176 00:49:18,640 --> 00:49:21,080 I think for the most part I kind of cheat by going basically 1177 00:49:21,080 --> 00:49:23,000 making the source code the thing, right? 1178 00:49:23,000 --> 00:49:25,440 It's kind of large. If I knows I can transport 1179 00:49:25,440 --> 00:49:28,120 source code from one session to another, that's kind of good. 1180 00:49:28,120 --> 00:49:30,160 But what I have durable things I'm working on. 1181 00:49:30,640 --> 00:49:33,440 Yeah, it gets down to copying. Things back and forth between 1182 00:49:33,440 --> 00:49:36,240 sessions or just sort of re describing something and quite 1183 00:49:36,240 --> 00:49:38,080 else I can get it back to the point where I feel it's useful. 1184 00:49:38,720 --> 00:49:40,480 So yeah, it's kind of problematic with. 1185 00:49:41,200 --> 00:49:43,280 So I'm looking forward to things getting a bit better in that 1186 00:49:43,280 --> 00:49:45,400 space. I think there's somebody who's 1187 00:49:45,400 --> 00:49:47,480 talking about actually doing this and calling them skills 1188 00:49:47,480 --> 00:49:50,240 that you can actually sort of like make almost like plug 1189 00:49:50,240 --> 00:49:52,360 insurance for particular things. And there's many different ways 1190 00:49:52,360 --> 00:49:55,560 to going and tuning LMS to go and sort of like facilitate 1191 00:49:55,560 --> 00:49:58,040 this, but to the point where we can actually create user 1192 00:49:58,040 --> 00:49:59,440 defined. Skills to move them across 1193 00:49:59,440 --> 00:50:01,800 sessions if it's valuable. The other thing I'll say is that 1194 00:50:01,800 --> 00:50:04,800 even if we have an Internet context window, sessions by 1195 00:50:04,800 --> 00:50:07,560 themselves are so valuable because quite often I'll keep 1196 00:50:07,560 --> 00:50:09,760 one main session and I'll basically just create small ones 1197 00:50:09,760 --> 00:50:12,720 I dispose of, right? And it's almost like managing 1198 00:50:12,720 --> 00:50:15,200 tabs in a browser, right? It's kind of like, here's the 1199 00:50:15,200 --> 00:50:17,760 main thing, I'm working and I've got 567 eights and you have to 1200 00:50:17,760 --> 00:50:19,280 kind of know when to get rid of them and when. 1201 00:50:19,280 --> 00:50:21,280 To keep them. But it's nice to go and have a 1202 00:50:21,280 --> 00:50:23,400 good separated space. We can, actually. 1203 00:50:23,400 --> 00:50:25,320 Go and do some working things and experiments. 1204 00:50:25,800 --> 00:50:27,560 Without actually sort of corrupting or degrading 1205 00:50:27,560 --> 00:50:28,960 performance. And if you have the. 1206 00:50:28,960 --> 00:50:32,240 Original session you have. I know it's a fast changing 1207 00:50:32,240 --> 00:50:35,200 world these days, right, But any tools, favorite tools that you 1208 00:50:35,200 --> 00:50:38,600 use these days, you know, related to AI encoding or is it 1209 00:50:38,600 --> 00:50:40,120 Co piloted? It's going to be so weird just 1210 00:50:40,240 --> 00:50:43,280 there are so you know, Co piloted is great, but it's one 1211 00:50:43,280 --> 00:50:45,600 of the things I kind of am drawing the distinction between 1212 00:50:45,600 --> 00:50:49,880 is that let's see, we're in a financial in the world right 1213 00:50:49,880 --> 00:50:52,200 now. The market is in this place 1214 00:50:52,200 --> 00:50:56,040 where every time that you make a general LM better, it's kind of 1215 00:50:56,040 --> 00:50:58,360 like you're kind of eating into the space where people do 1216 00:50:58,360 --> 00:50:59,760 specialty things. Right. 1217 00:51:00,280 --> 00:51:03,280 So it's kind of like you can try to specialize the tool, but if 1218 00:51:03,320 --> 00:51:05,720 people can just wait for the next general LM to do things, 1219 00:51:05,720 --> 00:51:08,320 then what's it do right? But I think it's starting to 1220 00:51:08,320 --> 00:51:11,200 separate now and you're starting to see decent coding tools to 1221 00:51:11,200 --> 00:51:14,480 have their own utility. But what I like to do is even if 1222 00:51:14,480 --> 00:51:17,320 I'm using copilot or I'm using code scene or all these other 1223 00:51:17,720 --> 00:51:22,040 tools is to have my own window open in something like cloud, 1224 00:51:22,040 --> 00:51:25,040 for instance, whatever I consider to be the best LM at 1225 00:51:25,040 --> 00:51:27,160 the moment. And basically do my more 1226 00:51:27,160 --> 00:51:30,640 ideation work in that space because they aren't clamped 1227 00:51:30,640 --> 00:51:32,360 down. And by what I mean by that is 1228 00:51:32,880 --> 00:51:34,960 sometimes you'll be using a tool and an IDE. 1229 00:51:34,960 --> 00:51:38,000 And so you'll say, well, give me three other approaches for going 1230 00:51:38,000 --> 00:51:40,680 and doing this or tell me more about this thing in the domain 1231 00:51:40,680 --> 00:51:43,120 and they'll say, no, I'm sorry, I'm here to help you with your 1232 00:51:43,120 --> 00:51:44,800 code and refactoring. And that's that. 1233 00:51:45,320 --> 00:51:48,040 And you don't, you know, know what the tools are good for, but 1234 00:51:48,040 --> 00:51:51,840 always keep your most capable general tool next to you in 1235 00:51:51,840 --> 00:51:54,600 order to go and actually do the things that you are going to 1236 00:51:54,600 --> 00:51:56,440 help you have high creative value. 1237 00:51:56,480 --> 00:52:00,840 So yeah, personally also subscribe to many AI tools, 1238 00:52:00,840 --> 00:52:02,240 right? So I think sometimes you just, 1239 00:52:02,440 --> 00:52:05,360 you know, pick the best from whatever AI that you use, right? 1240 00:52:05,360 --> 00:52:08,040 And sometimes, yeah, one works best for certain contexts and 1241 00:52:08,040 --> 00:52:11,000 the other works best for others. So I think that this landscape 1242 00:52:11,000 --> 00:52:13,760 is changing really, really fast. So I think there will be more 1243 00:52:13,760 --> 00:52:15,840 exciting stuff happening within this space. 1244 00:52:16,400 --> 00:52:18,520 But one thing for sure, like what you said, I think probably 1245 00:52:18,520 --> 00:52:21,080 developers role won't be like gone, right? 1246 00:52:21,080 --> 00:52:23,840 People are still having to problem solve something, right? 1247 00:52:24,080 --> 00:52:26,080 And especially also the understanding about domain, 1248 00:52:26,080 --> 00:52:27,720 right? I think that's still one key. 1249 00:52:28,160 --> 00:52:30,840 So thank you so much for this exciting talk for Michael. 1250 00:52:30,840 --> 00:52:33,200 Unfortunately, we reached the end of our conversation. 1251 00:52:33,440 --> 00:52:35,920 But before I let you go, I have one last question that I always 1252 00:52:35,920 --> 00:52:38,040 ask my guest. I call this 3 technical 1253 00:52:38,040 --> 00:52:39,920 leadership wisdom. You can think of it just like 1254 00:52:39,920 --> 00:52:42,160 advice that you want to give to us, the listeners here. 1255 00:52:42,320 --> 00:52:44,440 So maybe you can share your version of the wisdom. 1256 00:52:45,640 --> 00:52:47,600 Yeah, It's kind of funny because I was looking back through, you 1257 00:52:47,600 --> 00:52:50,320 know, the other, you know, some of your earlier episodes. 1258 00:52:50,320 --> 00:52:54,000 And, you know, there's a lot of things about basically managing 1259 00:52:54,040 --> 00:52:56,280 organizations and basically dealing with people and stuff 1260 00:52:56,280 --> 00:52:58,800 like that purely from a technical point of view. 1261 00:52:58,800 --> 00:53:00,840 The thing I want to go and basically mention is that 1262 00:53:01,640 --> 00:53:03,760 basically looking for things that aren't there is very 1263 00:53:03,760 --> 00:53:04,760 valuable. Right. 1264 00:53:05,240 --> 00:53:07,760 And I think just generally in life, it's kind of like if, if 1265 00:53:07,760 --> 00:53:09,920 you can detect the absence of something that should be there, 1266 00:53:09,920 --> 00:53:12,960 right, then you're in a good position. 1267 00:53:13,440 --> 00:53:15,640 The legacy code book came about because I realized it's like, 1268 00:53:16,320 --> 00:53:17,960 this is a tough problem and nobody's going to touch it. 1269 00:53:17,960 --> 00:53:19,600 Nobody wants to touch this problem. 1270 00:53:19,600 --> 00:53:21,680 So it's like, OK, might as well do this, right? 1271 00:53:22,280 --> 00:53:24,240 So developing a sense for the absence. 1272 00:53:24,240 --> 00:53:25,560 Of things is extremely. Valuable. 1273 00:53:26,040 --> 00:53:29,160 The other is like, you know #2 I guess would be basically 1274 00:53:29,160 --> 00:53:31,080 learning as many domains as you can, right? 1275 00:53:31,080 --> 00:53:34,760 Really digging into things because it's like the more 1276 00:53:34,760 --> 00:53:37,920 impressions you get of different things, the more you have to 1277 00:53:37,920 --> 00:53:39,560 draw upon when you're formulating solutions. 1278 00:53:39,560 --> 00:53:43,200 Yourself, right? And this is almost like how an 1279 00:53:43,200 --> 00:53:46,400 Lol works in a way, right? It basically is read almost all 1280 00:53:46,400 --> 00:53:49,560 the online content in the world. And one thing that makes it very 1281 00:53:49,560 --> 00:53:51,840 effective is because it has that breadth, right? 1282 00:53:52,240 --> 00:53:55,560 So you'll quite often see yourself, I'm reading about this 1283 00:53:55,560 --> 00:53:57,920 and you'll say, wow, that's very close to this thing over there. 1284 00:53:57,920 --> 00:54:00,080 What's is there a commonality here? 1285 00:54:00,080 --> 00:54:02,480 And then you start asking questions and going deep. 1286 00:54:02,680 --> 00:54:04,640 Right. And I guess the third thing is 1287 00:54:04,640 --> 00:54:07,240 really just maintain your curiosity more than anything 1288 00:54:07,280 --> 00:54:09,360 else. You know, everybody who's in 1289 00:54:09,360 --> 00:54:12,280 software development is in a position to be able to learn 1290 00:54:12,280 --> 00:54:15,200 more about many adjacent things. You know, you're able to learn 1291 00:54:15,200 --> 00:54:17,400 more about business, you're able to learn more about management 1292 00:54:17,800 --> 00:54:19,680 and the resources available to all of us. 1293 00:54:19,680 --> 00:54:22,360 It's just a question of actually going there and dealing with 1294 00:54:22,360 --> 00:54:23,480 those things. So I guess those are the main 1295 00:54:23,480 --> 00:54:25,880 things and those have served me well through my career. 1296 00:54:25,920 --> 00:54:28,440 You know, and I, I don't know to what degree people can trick 1297 00:54:28,440 --> 00:54:30,760 themselves in the curiosity if they're not terribly curious. 1298 00:54:30,760 --> 00:54:34,160 But yeah, I hope they can because I, you know, it's, it's 1299 00:54:34,160 --> 00:54:35,080 fun to me. It's fun. 1300 00:54:35,240 --> 00:54:38,160 So yeah. I'm sure when people opt for the 1301 00:54:38,160 --> 00:54:41,120 programmer's job, I think there is a sense of curiosity because 1302 00:54:41,120 --> 00:54:43,600 otherwise, right, I mean, they they can't be a good programmer 1303 00:54:43,600 --> 00:54:46,360 anyway. So sure, just some of the most 1304 00:54:46,360 --> 00:54:48,400 interesting conversations ever had at conferences. 1305 00:54:48,760 --> 00:54:50,560 They have nothing to do with software, just people are just 1306 00:54:50,560 --> 00:54:52,880 very into things and. Curious. 1307 00:54:53,120 --> 00:54:54,760 Sorry, Yeah. No problem. 1308 00:54:55,200 --> 00:54:56,680 So thanks again for this conversation. 1309 00:54:56,680 --> 00:54:58,680 So if people like this conversation or they want to 1310 00:54:58,680 --> 00:55:01,520 reach out or check out with you certain stuff, is there a place 1311 00:55:01,520 --> 00:55:03,680 where they can reach out online? Sure. 1312 00:55:03,680 --> 00:55:06,640 There's my sub stack Michael Chathers at substack.com. 1313 00:55:06,640 --> 00:55:09,320 There's also on X or Twitters M Chathers. 1314 00:55:09,720 --> 00:55:11,840 So just either one of those places is perfectly fine to go 1315 00:55:11,840 --> 00:55:14,160 and reach me. You can also search for the book 1316 00:55:14,280 --> 00:55:17,680 AI assisted programming on Reenpub and it's early days. 1317 00:55:17,680 --> 00:55:19,320 I'm about 15% done. But I'm. 1318 00:55:19,720 --> 00:55:21,160 Releasing new chapters every week or so. 1319 00:55:21,160 --> 00:55:25,200 So, all right, thanks. I don't know how far it can be a 1320 00:55:25,280 --> 00:55:29,000 legacy code book, right, which is very classic, but I think I 1321 00:55:29,000 --> 00:55:32,560 hope I can see the book in a full form, right, and learn how 1322 00:55:32,560 --> 00:55:35,280 to use the AI much better. So thanks for writing the book 1323 00:55:35,280 --> 00:55:36,920 as well. Sure. 1324 00:55:37,040 --> 00:55:37,320 Thank you.