1 00:00:00,290 --> 00:00:02,000 These were all the sort of basic primitives. 2 00:00:02,000 --> 00:00:05,750 And then, you know, at some point we said, well, recipes could also 3 00:00:05,770 --> 00:00:09,729 emit structured data in the form of tables, just rows and columns of data. 4 00:00:10,270 --> 00:00:14,080 And we would allow folks to run those over thousands or tens of 5 00:00:14,080 --> 00:00:18,119 thousands of these, loss of semantic tree artifacts and extract data out. 6 00:00:18,949 --> 00:00:25,440 This wound up being the fruitful bed for LLMs eventually arriving is that 7 00:00:25,440 --> 00:00:29,959 we had thousands of these recipes emitting data in various different forms. 8 00:00:30,580 --> 00:00:34,359 And if you could just expose this tools, all of those thousands of recipes to a 9 00:00:34,359 --> 00:00:38,270 model and say, okay, I have a question for you about this business unit. 10 00:00:38,440 --> 00:00:42,760 The model could select the right recipe, deterministically, run it on potentially 11 00:00:42,760 --> 00:00:46,360 hundreds of millions of lines of code, get the data table back, reason about 12 00:00:46,360 --> 00:00:47,799 it, combine it with something else. 13 00:00:48,230 --> 00:00:52,329 And that's the sort of, I think, foundation for large language 14 00:00:52,340 --> 00:00:55,730 models to help with large scale transformation and impact analysis. 15 00:01:01,030 --> 00:01:02,899 Welcome to Screaming in the Cloud. 16 00:01:03,139 --> 00:01:08,679 I'm Corey Quinn, and my guest today has been invited to the show because I've 17 00:01:08,719 --> 00:01:14,830 been experiencing, this may shock you some skepticism around a number of things in 18 00:01:14,830 --> 00:01:17,170 the industry, but we'll get into that. 19 00:01:17,450 --> 00:01:20,740 Jonathan Schneider is the CEO of modern. 20 00:01:20,849 --> 00:01:23,559 Uh, and before that, you've done a lot of things. 21 00:01:23,559 --> 00:01:25,460 Jonathan, first, thank you for joining me. 22 00:01:26,070 --> 00:01:26,369 Yeah. 23 00:01:26,380 --> 00:01:27,580 Thanks for having me here, Corey. 24 00:01:27,580 --> 00:01:28,259 Such a pleasure. 25 00:01:28,740 --> 00:01:31,350 Crying Out Cloud is one of the few cloud security podcasts 26 00:01:31,350 --> 00:01:33,600 that's actually fun to listen to. 27 00:01:33,690 --> 00:01:36,589 Smart Conversations, great guests, and zero fluff. 28 00:01:36,640 --> 00:01:41,340 If you haven't heard of it, it's a cloud and AI security podcast from Wiz run 29 00:01:41,349 --> 00:01:44,000 by CloudSec pros, for CloudSec pros. 30 00:01:44,219 --> 00:01:46,789 I was actually one of the first guests on the show and it's 31 00:01:46,789 --> 00:01:48,279 been amazing to watch it grow. 32 00:01:48,400 --> 00:01:53,410 Make sure to check them out at wiz.io/crying-out-cloud. 33 00:01:53,830 --> 00:01:56,460 Uh, we, we always have to start with a book story because honestly, I'm 34 00:01:56,460 --> 00:01:58,130 envious of those who can write a book. 35 00:01:58,150 --> 00:02:02,610 I just write basically 18 volumes of Twitter jokes over the years, 36 00:02:02,610 --> 00:02:05,560 but never actually sat down and put anything cohesive together. 37 00:02:05,840 --> 00:02:10,210 Uh, you were the author of SRE with Java microservices and the co-author 38 00:02:10,210 --> 00:02:13,980 of Automated Code Remediation, how to Refactor and Secure the 39 00:02:13,990 --> 00:02:15,730 Modern Software Supply Chain. 40 00:02:15,969 --> 00:02:18,459 So you are professionally depressed, I assume. 41 00:02:19,330 --> 00:02:23,979 I mean, I as like most software engineers, I hate writing documentation. 42 00:02:23,980 --> 00:02:28,260 So somehow that translated into, you know, write a, a full scale book instead. 43 00:02:28,260 --> 00:02:30,640 I, I honestly don't remember how that happened. 44 00:02:32,190 --> 00:02:35,270 A series of escalating poor life choices is my experience of it. 45 00:02:35,270 --> 00:02:36,780 I think no one wants to write a book. 46 00:02:36,820 --> 00:02:40,349 Everyone wants to have written a book, and then you went and did it a second time. 47 00:02:41,260 --> 00:02:42,840 Yeah, much more a smaller one. 48 00:02:42,840 --> 00:02:47,340 That second one, you know, just the 35 pager luckily, but, but, you know, still, 49 00:02:47,460 --> 00:02:49,910 um, or it's always, uh, quite the effort. 50 00:02:50,200 --> 00:02:53,879 So one thing that I, I wanted to bring you in to talk about is that the core of 51 00:02:53,880 --> 00:02:57,959 what your company does, which is, I, I, please correct me if I'm wrong on this 52 00:02:58,010 --> 00:03:00,820 software rewrites software modernization. 53 00:03:00,910 --> 00:03:05,850 Effectively, you were doing what Amazon Queue transform purports to 54 00:03:05,850 --> 00:03:08,510 do, uh, before everyone went AI crazy. 55 00:03:08,730 --> 00:03:12,660 Yeah, it started for me almost 10 years ago now at Netflix on the 56 00:03:12,660 --> 00:03:16,290 engineering tools team, where I was responsible for making people move 57 00:03:16,290 --> 00:03:20,910 forward, uh, in part, but they had that freedom and responsibility 58 00:03:20,910 --> 00:03:23,760 culture, so I could tell 'em, you're not where you're supposed to be. 59 00:03:23,760 --> 00:03:25,480 And they would say, great, do it for me. 60 00:03:25,480 --> 00:03:27,130 Otherwise I'm, I got other things to do. 61 00:03:27,340 --> 00:03:32,269 Uh, and so really that forced our team into trying to find ways to 62 00:03:32,269 --> 00:03:34,310 automate that change on their behalf. 63 00:03:35,400 --> 00:03:38,640 I never worked at quite that scale in production. 64 00:03:38,640 --> 00:03:41,880 I mean, I've consulted in there in places like that, but the, that's a 65 00:03:41,880 --> 00:03:45,100 very different experience 'cause you're hyperfocused on a specific problem. 66 00:03:45,500 --> 00:03:49,619 But even at the scales that I've operated at, there was, there was 67 00:03:49,620 --> 00:03:53,369 never a. An intentional decision of someone's gonna start out today, and 68 00:03:53,500 --> 00:03:57,110 we're gonna write this in a language and framework that are 20 years old. 69 00:03:57,260 --> 00:03:59,330 So this stuff has always been extant for a while. 70 00:03:59,370 --> 00:04:04,010 It is, it has grown roots, it has worked its way into business processes and a 71 00:04:04,010 --> 00:04:05,769 bunch of things have weird dependencies. 72 00:04:05,770 --> 00:04:07,109 In some cases on bugs. 73 00:04:07,420 --> 00:04:11,839 People are not, uh, declining to modernize software stacks because they haven't 74 00:04:11,850 --> 00:04:13,359 heard that there's a new version out. 75 00:04:13,770 --> 00:04:17,720 It's because this stuff is painfully hard because people and organizations 76 00:04:17,720 --> 00:04:19,440 that they build are painfully hard. 77 00:04:20,190 --> 00:04:24,510 I I'm curious, in your experience having gone through this at scale 78 00:04:24,510 --> 00:04:28,709 with zeros on the end of it, what, what are the, what are the sticking, 79 00:04:28,740 --> 00:04:29,579 what are the sticking points of this? 80 00:04:29,590 --> 00:04:31,659 Why don't people migrate? 81 00:04:31,700 --> 00:04:34,539 Is it more of a technological problem or is it more of a people problem? 82 00:04:35,370 --> 00:04:39,530 Well, first I would start and hopefully with a, a sympathetic viewpoint for 83 00:04:39,530 --> 00:04:42,700 the developer, which is like, pretend I haven't written any software yet 84 00:04:42,710 --> 00:04:44,200 and I'm actually starting from today. 85 00:04:45,259 --> 00:04:49,570 I look at all the latest available things, and I make the perfectly optimal 86 00:04:49,570 --> 00:04:53,550 choices for every part of my tech stack today, and I, I write this thing 87 00:04:53,550 --> 00:04:58,380 completely clean six months from now, those are no longer the optimal choices. 88 00:04:59,099 --> 00:04:59,680 Oh God, yes. 89 00:04:59,690 --> 00:04:59,950 Me. 90 00:05:00,170 --> 00:05:01,970 Worst developer I ever met is me six weeks ago. 91 00:05:02,160 --> 00:05:02,669 It's awful. 92 00:05:02,679 --> 00:05:04,139 Like, what, what was this idiot thinking? 93 00:05:04,139 --> 00:05:07,960 You do get blame and it's you and wow, we need not talk about that anymore. 94 00:05:08,170 --> 00:05:10,950 But yeah, the past me was terrible at this. 95 00:05:11,300 --> 00:05:11,720 That's right. 96 00:05:11,730 --> 00:05:13,340 And, and always will be future. 97 00:05:13,340 --> 00:05:15,180 You will be the pa the next past you. 98 00:05:15,190 --> 00:05:18,879 So it's, it's, there's never an opportunity where we can say we're making 99 00:05:18,880 --> 00:05:22,490 the optimal choice and that optimal choice will continue to be right going forward. 100 00:05:22,490 --> 00:05:27,650 So, uh, I think that paired with one other fact, which is just that the. 101 00:05:27,930 --> 00:05:32,190 The tools available to us have essentially industrial industrialized software 102 00:05:32,190 --> 00:05:37,159 production to the point where we can write net new software super quickly 103 00:05:37,340 --> 00:05:41,749 using off the shelf and third party open source components we're expected 104 00:05:41,750 --> 00:05:46,270 to because you have to ship ship fast and you know, then what do you do when 105 00:05:46,270 --> 00:05:48,300 that stuff evolves at its own pace? 106 00:05:48,570 --> 00:05:52,820 So nobody's really been good at it, and I think the more. 107 00:05:53,059 --> 00:05:57,600 Authorship, uh, automation that we've, that we've, uh, developed for ourselves 108 00:05:57,600 --> 00:06:02,470 from IDE rule-based intention actions to now, you know, AI authorship. 109 00:06:03,210 --> 00:06:06,890 This, like the time that we spend maintaining what we've previously 110 00:06:06,890 --> 00:06:08,350 written has continued to go up. 111 00:06:08,889 --> 00:06:09,590 I would agree. 112 00:06:10,020 --> 00:06:16,089 I, I think that there has been a. A a, a shift and a proliferation really, of 113 00:06:16,289 --> 00:06:18,299 technical stacks and software choices. 114 00:06:18,350 --> 00:06:23,090 And as you say, even if you make the optimal selection of every piece of 115 00:06:23,090 --> 00:06:26,200 the stack, which incidentally is where, where some people tend to founder, they 116 00:06:26,200 --> 00:06:30,140 spend six months trying to figure out the best approach, pick a direction and go, 117 00:06:30,140 --> 00:06:32,300 even a bad decision can be made to work. 118 00:06:32,740 --> 00:06:36,039 But, but there are so many different paths to go that it's a near certainty 119 00:06:36,039 --> 00:06:37,509 that whatever you have built, you're. 120 00:06:37,999 --> 00:06:41,520 There, you're going to be one of a wide variety of different 121 00:06:41,520 --> 00:06:42,350 paths that you've picked. 122 00:06:42,510 --> 00:06:47,190 You're effectively become a unicorn pretty quickly regardless of how mainstream 123 00:06:47,290 --> 00:06:48,719 each individual choice might be. 124 00:06:48,830 --> 00:06:49,289 That's right. 125 00:06:49,659 --> 00:06:49,919 Yep. 126 00:06:50,139 --> 00:06:52,249 That's just the nature of, of software development. 127 00:06:53,400 --> 00:06:56,770 I I am curious since, uh, you did bring up the, uh, Netflix, uh, 128 00:06:56,790 --> 00:06:58,670 freedom and responsibility culture. 129 00:06:58,940 --> 00:07:02,140 Uh, one thing that has made me skeptical historically of Amazon 130 00:07:02,140 --> 00:07:05,489 queue's, transform abilities and, and many large companies that have taken a 131 00:07:05,490 --> 00:07:10,310 bite at this apple is they, they train these things and build these things. 132 00:07:10,379 --> 00:07:14,710 Inside of a culture that has a very particular point of view that drives 133 00:07:14,710 --> 00:07:16,300 how software development is done. 134 00:07:16,720 --> 00:07:20,560 Uh, I like how many people have we met that have left large tech companies 135 00:07:20,570 --> 00:07:24,289 to go, found a startup, tried to build the exact same culture that they had 136 00:07:24,300 --> 00:07:28,410 at the large company and just founder on the rocks almost immediately. 137 00:07:28,430 --> 00:07:31,890 Because the culture shapes the company and the company shapes the culture. 138 00:07:32,070 --> 00:07:34,700 You, you can't cargo cult it and expect success. 139 00:07:35,290 --> 00:07:37,990 How, how varied do you find that these modernization 140 00:07:38,000 --> 00:07:39,720 efforts are based upon culture? 141 00:07:40,210 --> 00:07:43,560 I'm glad to say that for my own story, I had a degree of indirection here. 142 00:07:43,560 --> 00:07:48,019 I didn't go straight from Netflix to, to founding something, so I was at Netflix. 143 00:07:48,049 --> 00:07:51,640 I think that freedom and responsibility culture meant that Netflix in particular 144 00:07:51,650 --> 00:07:57,249 had far less self similarity or consistency than say, a Google that has a 145 00:07:57,260 --> 00:08:01,189 very prescriptive standard for formatting and everything and the way they do things. 146 00:08:01,799 --> 00:08:03,300 And so I left Netflix. 147 00:08:03,300 --> 00:08:06,329 I went to Pivotal, VMware was working with large. 148 00:08:06,650 --> 00:08:11,250 Enterprise customers like JP Morgan, fidelity, home Depot, et cetera, working 149 00:08:11,250 --> 00:08:15,450 on an unrelated problem in continuous delivery and saw them struggling with the 150 00:08:15,450 --> 00:08:18,730 same kind of problem of like migrations and modern, like everybody does. 151 00:08:19,170 --> 00:08:23,509 And what struck me was that even though they're very different cultures, 152 00:08:24,150 --> 00:08:29,810 uh, JP Morgan much more strongly resembles Netflix than it does Google. 153 00:08:30,300 --> 00:08:36,059 Um, Netflix's uh, lack of consistency was by design or by culture, intentional. 154 00:08:36,380 --> 00:08:39,664 And JP Morgan's is just by the very sheer nature of the fact that they have 155 00:08:39,664 --> 00:08:43,630 60,000 developers and 25 years of this, of history and development on this. 156 00:08:43,630 --> 00:08:49,550 And so a solution that works well for, uh, dissimilar by design 157 00:08:49,600 --> 00:08:53,750 actually works well in the typical enterprise, which is probably closer 158 00:08:53,750 --> 00:08:55,170 to Netflix than it is to Google. 159 00:08:56,160 --> 00:08:57,900 Yeah, a lot of it depends on constraints too. 160 00:08:57,929 --> 00:09:01,820 Uh, JP Morgan is obviously highly reg, sorry, JP Morgan Chase. 161 00:09:01,840 --> 00:09:05,180 They're particular about the naming people are, they're obviously highly 162 00:09:05,180 --> 00:09:09,860 regulated and mistakes matter in a different context than they do when your 163 00:09:09,990 --> 00:09:14,720 basic entire business is streaming movies and also creating original content that 164 00:09:14,720 --> 00:09:16,160 you then cancel just when it gets good. 165 00:09:17,199 --> 00:09:17,426 Right, right, right. 166 00:09:17,479 --> 00:09:17,519 Right. 167 00:09:18,999 --> 00:09:19,809 Yes. 168 00:09:20,350 --> 00:09:23,500 So there's, there is that question, I guess, of how this stuff evolves. 169 00:09:23,500 --> 00:09:27,023 But taking it a bit away from the culture side of it, how do you 170 00:09:27,120 --> 00:09:31,120 find that modernization differs between programming languages? 171 00:09:31,450 --> 00:09:34,140 I mean, I, I dunno if people are watching this on the video or listening to it, 172 00:09:34,190 --> 00:09:38,280 if we may, we take all kinds, but you're wearing a hat right now that says JVM, 173 00:09:38,290 --> 00:09:43,089 so I'm, I'm just gonna speculate wildly that Java might be your first love, given 174 00:09:43,090 --> 00:09:44,760 that you did in fact write a book on it. 175 00:09:45,010 --> 00:09:46,489 It was one of my first loves. 176 00:09:46,490 --> 00:09:46,590 Yeah. 177 00:09:46,880 --> 00:09:49,150 The technically a Java champion right now. 178 00:09:49,150 --> 00:09:52,100 Although, you know, I actually started in c plus plus and I hated Java for 179 00:09:52,100 --> 00:09:53,370 the first few years I worked on it. 180 00:09:53,370 --> 00:09:55,595 But, um, I actually think, uh, 181 00:09:55,900 --> 00:09:57,450 Stockholm Syndrome can work miracles. 182 00:09:57,600 --> 00:09:58,880 It it sure can. 183 00:09:58,940 --> 00:09:59,920 It absolutely can. 184 00:10:00,310 --> 00:10:01,970 I, I don't know that the. 185 00:10:02,260 --> 00:10:04,300 Problems are, are that different? 186 00:10:04,309 --> 00:10:06,520 There's, you know, a lot of different engineering challenges. 187 00:10:06,520 --> 00:10:09,570 How statically typed is something, how dynamically typed is it, how 188 00:10:09,570 --> 00:10:12,780 accurate can a, a transformation be provably made to be? 189 00:10:12,810 --> 00:10:17,640 But in general, I think the problems are, um, the social engineering problems are 190 00:10:17,640 --> 00:10:21,750 harder than the, than the specifics of the transformation that's being made. 191 00:10:22,139 --> 00:10:24,080 And those social engineering problems are like. 192 00:10:24,530 --> 00:10:29,489 Do I build a system that issues mass pull requests from essential team to 193 00:10:29,490 --> 00:10:32,900 all the product teams and expect that everybody's gonna merge them because. 194 00:10:33,300 --> 00:10:38,099 They love it when, you know, random things show up in their, uh, in their PRQ or, 195 00:10:38,540 --> 00:10:43,200 uh, do product teams perceive that, like unwelcome advice coming from an in-law 196 00:10:43,210 --> 00:10:46,100 and they're just looking for a reason to reject it, you know, and then they would 197 00:10:46,100 --> 00:10:50,329 prefer, instead to have an experience where, you know, when they're about to 198 00:10:50,330 --> 00:10:53,970 undergo a large scale transformation that they pull or they initiate the 199 00:10:53,970 --> 00:10:55,420 change and then merge it themselves. 200 00:10:55,430 --> 00:10:59,780 So like those are the things that I think are, are highly similar regardless 201 00:10:59,780 --> 00:11:02,160 of the tech stack or company that's. 202 00:11:02,170 --> 00:11:05,000 Uh, because people are people, uh, kind of everywhere. 203 00:11:05,520 --> 00:11:05,590 Now 204 00:11:05,609 --> 00:11:09,860 you take the suite of Amazon Q transform options and they have a bunch of of 205 00:11:09,900 --> 00:11:13,570 software modernization capabilities, but also getting people off of VMware 206 00:11:13,570 --> 00:11:17,590 due to, you know, extortion as well as getting off of the MI of the mainframe, 207 00:11:17,620 --> 00:11:20,410 which that last one is probably the thing I'm the most skeptical of 208 00:11:20,870 --> 00:11:23,630 companies have been trying to get off of the mainframe for 40 years. 209 00:11:23,639 --> 00:11:26,860 The problem is not that you can't recreate something that does the same 210 00:11:26,860 --> 00:11:30,069 processing, it's that there are thousands of business processes that are critically 211 00:11:30,070 --> 00:11:31,979 dependent on that thing and you can't. 212 00:11:32,139 --> 00:11:34,700 Migrate them one at a time in most cases. 213 00:11:35,969 --> 00:11:40,540 I am highly skeptical that just pour some AI on it is necessarily going to move 214 00:11:40,540 --> 00:11:42,390 that needle in any material fashion. 215 00:11:43,180 --> 00:11:46,530 I think that there's a, a two different kinds of activities here. 216 00:11:46,560 --> 00:11:51,020 One is code authorship, that new authorship, uh, that's what the 217 00:11:51,460 --> 00:11:53,760 copilots are doing, the Amazon queue is doing, et cetera. 218 00:11:53,770 --> 00:11:55,479 It's, it's really assisting in that respect. 219 00:11:55,620 --> 00:11:59,030 And then there's code maintenance, which is, I need to get this thing from 220 00:11:59,030 --> 00:12:00,650 one version of a framework to another. 221 00:12:00,950 --> 00:12:04,520 Maintenance can also include, I'm trying to consolidate one 222 00:12:04,520 --> 00:12:08,200 feature flagging vendor to, or two feature flagging vendors to one. 223 00:12:08,559 --> 00:12:09,500 Um, but. 224 00:12:10,240 --> 00:12:14,969 When I think think of something like a COBOL to a modern stack JV young 225 00:12:15,440 --> 00:12:19,159 or.net or whatever the case might be, I honestly see that less of as, as 226 00:12:19,160 --> 00:12:22,990 a maintenance activity and more as an authorship activity, a new, and 227 00:12:23,260 --> 00:12:26,470 you're, you're writing net new software in a different stack and a different 228 00:12:26,480 --> 00:12:28,430 set of expectations and assumptions. 229 00:12:29,110 --> 00:12:31,000 Um, and so I'm skeptical too. 230 00:12:31,090 --> 00:12:34,494 I don't, I don't think there's a magic wand, but to the extent that our 231 00:12:34,590 --> 00:12:39,010 authorship tools help us accelerate net new development, those problems. 232 00:12:39,610 --> 00:12:41,140 The cost of those problems goes down. 233 00:12:41,210 --> 00:12:41,949 I think over time. 234 00:12:43,290 --> 00:12:47,089 Yeah, that, that does track and makes sense of how I tend 235 00:12:47,090 --> 00:12:48,380 to think about these things. 236 00:12:48,650 --> 00:12:52,259 But at the same time that the cost of these things goes down and the 237 00:12:52,309 --> 00:12:56,640 technology increases, it still feels like these applications that are decades 238 00:12:56,640 --> 00:13:00,960 old in some cases are still exploding geometrically with respect to complexity. 239 00:13:01,850 --> 00:13:02,380 That's right. 240 00:13:02,610 --> 00:13:02,930 Yeah. 241 00:13:03,889 --> 00:13:05,170 Like how do you outrun it all? 242 00:13:05,700 --> 00:13:06,680 Well, um. 243 00:13:08,140 --> 00:13:13,510 Uh, to me there's not just one approach here, but I feel like, um, you know, 244 00:13:14,310 --> 00:13:20,880 for my own sake and my, where my focus is, is really trying to reclaim, uh, 245 00:13:20,900 --> 00:13:25,079 developer time in some area so that it can refocus that effort elsewhere. 246 00:13:25,110 --> 00:13:29,470 And I think one thing I hear pretty consistently is that because 247 00:13:29,470 --> 00:13:32,980 of that explosion in software under management right now. 248 00:13:33,490 --> 00:13:36,280 A developer spending like 30 or 40% of their time just kind 249 00:13:36,280 --> 00:13:39,060 of resiting applications and keeping, keeping the lights on. 250 00:13:39,610 --> 00:13:43,269 And that's something we need to like get rid of a bit or as minimize as 251 00:13:43,270 --> 00:13:47,959 much as possible so that, you know, the next feature they're developing isn't 252 00:13:47,980 --> 00:13:51,810 just a net new feature but is actually, you know, pulling some like old system 253 00:13:51,810 --> 00:13:53,970 into a more modern framework as well. 254 00:13:53,980 --> 00:13:56,210 That's just another activity that can go back onto their, 255 00:13:56,849 --> 00:13:57,880 uh, is something they can do. 256 00:13:58,250 --> 00:14:03,859 But that does track the, I guess the scary part too, is it having lived 257 00:14:03,860 --> 00:14:06,720 through some of these myself where we know that we need to upgrade the 258 00:14:06,730 --> 00:14:11,219 thing to break off the monolith, to master the wolf, et cetera, et cetera, 259 00:14:11,220 --> 00:14:14,960 et cetera, and it feels like there's never time to focus on that because you 260 00:14:14,960 --> 00:14:18,459 still have to ship features, but every feature you're doing feels like it's 261 00:14:18,460 --> 00:14:20,050 digging the technical debt hole deeper. 262 00:14:20,320 --> 00:14:20,770 It is. 263 00:14:21,809 --> 00:14:22,210 It is. 264 00:14:22,250 --> 00:14:22,510 Yeah. 265 00:14:22,510 --> 00:14:27,050 So I mean that's, and this is what I mean is like if we can take the assets that 266 00:14:27,050 --> 00:14:32,970 we have on our management right now and, and like keep them moving forward, um, 267 00:14:33,080 --> 00:14:39,799 then um, we have like less drift and less, you know, um, complexity to deal with. 268 00:14:39,849 --> 00:14:40,749 Overall. 269 00:14:40,940 --> 00:14:43,500 It's an important part of piece of that puzzle I think. 270 00:14:43,990 --> 00:14:45,780 As you said, you've been working on this for 10 years. 271 00:14:45,890 --> 00:14:50,900 Uh, gen AI really took off at the end of 2023, give or take Well, during 2023. 272 00:14:51,559 --> 00:14:55,530 And I'm curious to get your take on how that has evolved. 273 00:14:55,530 --> 00:14:59,290 I mean, yes, we all have to tell a story on some level around that. 274 00:14:59,290 --> 00:15:04,120 Uh, your uur l is modern.ai, so clearly there's, there is some marketing 275 00:15:04,120 --> 00:15:07,240 element to this, but, but you're, but you're a reasonable person on this 276 00:15:07,240 --> 00:15:08,870 stuff and you go deeper than most do. 277 00:15:09,139 --> 00:15:12,390 I think a lot of what, what I've developed over the last several years, 278 00:15:12,390 --> 00:15:16,630 or our team has, has been, you know, accidentally leading towards this moment 279 00:15:16,639 --> 00:15:21,720 where, um, we've got a set of tools that, uh, an LM can take advantage of. 280 00:15:21,730 --> 00:15:25,270 So the first thing was, you know, when I'm looking at a code 281 00:15:25,280 --> 00:15:29,040 base, the text of the code is in. 282 00:15:29,040 --> 00:15:34,090 I think to the abstract syntax tree of the code is insufficient. 283 00:15:34,090 --> 00:15:37,850 So things like tree sitters, you know, that I won't mention all the things 284 00:15:37,850 --> 00:15:40,790 builds on top of tree sitter, but if it's just abstract syntax tree, 285 00:15:41,580 --> 00:15:46,300 there's not enough information often for a model to latch onto to know 286 00:15:46,309 --> 00:15:47,809 how to make the right transformation. 287 00:15:48,550 --> 00:15:51,990 And the reason I started open Rewrite in the very, at the very beginning, 288 00:15:52,029 --> 00:15:55,439 10 years ago was because the very first problem I was trying to to solve 289 00:15:55,440 --> 00:15:59,160 at Netflix was moving from blitz for J and internal logging library to 290 00:15:59,360 --> 00:16:01,980 not blitz for J. We were just trying to kill off something we regretted. 291 00:16:02,470 --> 00:16:08,050 And yet that logging library looked almost identical in syn PAX to S 292 00:16:08,050 --> 00:16:09,870 SL for jj, any of the other ones. 293 00:16:10,490 --> 00:16:12,980 And so just looking at log info. 294 00:16:13,310 --> 00:16:15,890 Well that looks exactly like log on info from another library. 295 00:16:15,890 --> 00:16:19,610 I couldn't, you know, narrowly identify where blitz for day still 296 00:16:19,610 --> 00:16:21,129 was, even in the environment. 297 00:16:21,349 --> 00:16:23,440 So I had to kind of go one level deeper, which is what 298 00:16:23,440 --> 00:16:24,970 does the compiler know about it? 299 00:16:25,780 --> 00:16:30,510 And that is actually a really difficult thing to do, to just take Texta code and 300 00:16:30,510 --> 00:16:32,250 parse it into an abstracts and text tree. 301 00:16:32,710 --> 00:16:36,990 You can use tree sitter to go one step further and actually exercise the 302 00:16:36,990 --> 00:16:38,720 compiler and do all the symbol solving. 303 00:16:38,870 --> 00:16:42,030 Well, that actually means you have to exercise the compiler in some way. 304 00:16:42,059 --> 00:16:43,199 Well, how is that done? 305 00:16:43,200 --> 00:16:44,020 What are the source sets? 306 00:16:44,020 --> 00:16:44,939 What version does it require? 307 00:16:44,940 --> 00:16:49,130 What build tools require, like this winds up being this like Hu, hugely complex 308 00:16:49,130 --> 00:16:53,370 decision matrix to encounter an arbitrary repository and build out that LST. 309 00:16:55,160 --> 00:16:58,039 We built out that that LST or loss of semantic tree, and 310 00:16:58,040 --> 00:17:01,000 then we started building these recipes, which could modify them. 311 00:17:01,000 --> 00:17:04,999 And those recipes stacked on other recipes to the point where like a spring 312 00:17:05,000 --> 00:17:08,109 boot migration has 3,400 steps in it. 313 00:17:08,829 --> 00:17:10,710 And these were all the sort of basic primitives. 314 00:17:10,710 --> 00:17:14,450 And then, you know, at some point we said, well, recipes could also 315 00:17:14,450 --> 00:17:18,430 emit structured data in the form of tables, just rows and columns of data. 316 00:17:18,980 --> 00:17:22,780 And we would allow folks to run those over thousands or tens of 317 00:17:22,780 --> 00:17:26,830 thousands of these loss of semantic tree artifacts and extract data out. 318 00:17:27,660 --> 00:17:34,150 This wound up being the fruitful bed for LLMs eventually arriving, is that 319 00:17:34,150 --> 00:17:38,669 we had thousands of these recipes emitting data in various different forms. 320 00:17:39,290 --> 00:17:42,390 And if you could just expose as tools, all of those thousands 321 00:17:42,390 --> 00:17:43,950 of recipes to a model and say. 322 00:17:44,350 --> 00:17:46,980 Okay, I have a question for you about this business unit. 323 00:17:47,150 --> 00:17:51,469 The model could select the right recipe, deterministically, run it on potentially 324 00:17:51,469 --> 00:17:55,199 hundreds of millions of lines of code, get the data table back, reason about it, 325 00:17:55,219 --> 00:18:00,000 combine it with something else, and that's the sort of, I think, foundation for 326 00:18:00,330 --> 00:18:04,429 large language models to help with large scale transformation and impact analysis. 327 00:18:05,070 --> 00:18:09,780 This episode is sponsored by my own company, the Duck Bill Group, 328 00:18:09,910 --> 00:18:12,250 having trouble with your AWS bill. 329 00:18:12,380 --> 00:18:15,189 Perhaps it's time to renegotiate a contract with them. 330 00:18:15,420 --> 00:18:20,560 Maybe you're just wondering how to predict what's going on in the wide world of AWS. 331 00:18:20,730 --> 00:18:23,649 Well, that's where the Duck Bill group comes in to help. 332 00:18:24,120 --> 00:18:26,350 Remember, you can't duck the duck bill. 333 00:18:26,350 --> 00:18:29,229 Bill, which I am reliably informed by my business partner 334 00:18:29,250 --> 00:18:31,179 is absolutely not our motto. 335 00:18:31,190 --> 00:18:36,090 Uh, to give a a somewhat simplified example, uh, it's easy to envision 'cause 336 00:18:36,119 --> 00:18:40,370 some of us have seen this where we'll have code that winds up cranking on data 337 00:18:40,370 --> 00:18:44,800 and generating an artifact, and then it staes that object into S3 because that is 338 00:18:44,800 --> 00:18:46,669 the defacto storage system of the cloud. 339 00:18:47,080 --> 00:18:50,369 Next, it then picks up that same object and then runs a different 340 00:18:50,370 --> 00:18:52,400 series of transformation objects on it. 341 00:18:52,650 --> 00:18:56,449 Now, from a code perspective, there is zero visibility into whether that. 342 00:18:56,490 --> 00:19:00,389 Artifact being written to S3 is simply an inefficiency that can 343 00:19:00,390 --> 00:19:03,050 be written out and just have it passed directly to that sub-routine. 344 00:19:03,280 --> 00:19:06,800 Or if there's some external process, potentially another business unit 345 00:19:06,980 --> 00:19:09,760 that needs to touch that artifact for something for reporting. 346 00:19:09,990 --> 00:19:13,630 Uh, quarterly earnings are a terrific source where a lot of this stuff sometimes 347 00:19:13,650 --> 00:19:18,889 winds up getting, uh, getting floated up and it's, it is impossible without having. 348 00:19:19,240 --> 00:19:21,590 Conversations in many cases with people in other business 349 00:19:21,590 --> 00:19:23,950 units entirely to, to get there. 350 00:19:24,179 --> 00:19:27,459 That's the stumbling block that I have seen historically. 351 00:19:27,549 --> 00:19:30,229 I Is that the sort of thing that you're, that you wind up having 352 00:19:30,230 --> 00:19:33,530 to think about when you're doing these things or am I contextualizing 353 00:19:33,530 --> 00:19:35,170 this from a very different layer? 354 00:19:35,670 --> 00:19:39,820 I do think of this process of large scale transformation, impact analysis, 355 00:19:39,830 --> 00:19:43,660 very much like what you're describing as like a, a data warehouse, ETL 356 00:19:43,670 --> 00:19:46,669 type thing, which is, you know, I need to take a source of data, which is 357 00:19:46,670 --> 00:19:48,349 the text to the code and enrich it. 358 00:19:48,570 --> 00:19:49,719 Into something that's everything. 359 00:19:49,720 --> 00:19:53,620 To compile our, knows all the dependencies and everything else from that point. 360 00:19:53,820 --> 00:19:57,370 And once I have that data, that's a computationally expensive thing to do. 361 00:19:57,370 --> 00:19:58,530 Once I have that. 362 00:19:58,780 --> 00:20:03,270 There's a lot of different applications of that same data source. 363 00:20:03,920 --> 00:20:08,290 I, I should point out that I have been skeptical of AI in a number of 364 00:20:08,300 --> 00:20:11,680 ways for a while now, and I wanna be clear that when I say skeptical, I 365 00:20:11,680 --> 00:20:13,389 do mean I'm middle of the road on it. 366 00:20:13,410 --> 00:20:14,660 I see its value. 367 00:20:14,660 --> 00:20:17,540 I'm not one of those, it's just a way that kill trees and it's 368 00:20:17,540 --> 00:20:19,259 a dumb markoff chain generator. 369 00:20:19,270 --> 00:20:20,799 No, that is absurd. 370 00:20:21,210 --> 00:20:25,630 I'm also not quite on the fence of this changes everything and every business 371 00:20:25,630 --> 00:20:27,800 application should have AI baked into it. 372 00:20:28,030 --> 00:20:28,450 I am. 373 00:20:28,530 --> 00:20:33,740 I am very middle of the road on it, and the problem that I see as I look 374 00:20:33,750 --> 00:20:37,889 through all of this is it, it feels like it's being used to paper over. 375 00:20:38,000 --> 00:20:40,559 A bunch of these problems where it has to talk to folks. 376 00:20:40,699 --> 00:20:43,600 I've used a lot of AI coding assistance and I see where these things 377 00:20:43,600 --> 00:20:45,520 tend to fall short and fall down. 378 00:20:45,800 --> 00:20:49,699 Uh, a big one is that they are seem incapable of saying, I don't know. 379 00:20:49,710 --> 00:20:51,780 We need to go get additional data. 380 00:20:52,030 --> 00:20:55,630 Instead, they are, uh, they're extraordinarily confident and 381 00:20:55,630 --> 00:20:57,189 authoritative and also wrong. 382 00:20:57,330 --> 00:20:59,699 I say this as a white dude who has two podcasts. 383 00:20:59,730 --> 00:21:02,960 I am conversant with the being authoritatively wrong point of view 384 00:21:02,980 --> 00:21:04,879 here at sort of my people's culture. 385 00:21:05,600 --> 00:21:09,230 So it's, it's one of those, how do you, how do you meet in the middle on that? 386 00:21:09,240 --> 00:21:13,740 How do you get the value without going too far into the realm of absurdity? 387 00:21:14,099 --> 00:21:18,319 Well, I, I do think that these things need to be, they need to collaborate together. 388 00:21:18,320 --> 00:21:22,550 And so, uh, so it is with Amazon Q Code transformer, that's, that's 389 00:21:22,560 --> 00:21:27,020 working to provide migrations for, uh, Java and other things. 390 00:21:27,200 --> 00:21:30,539 You see that, that Amazon Q Code transformer actually uses. 391 00:21:30,690 --> 00:21:33,979 Open rewrite a rule-based or deterministic system behind it to 392 00:21:33,980 --> 00:21:35,660 actually make a lot of those changes. 393 00:21:35,910 --> 00:21:38,210 An an open source tool that incidentally you were the 394 00:21:38,210 --> 00:21:39,410 founder of, if I'm not mistaken. 395 00:21:39,470 --> 00:21:40,210 That's right, yeah. 396 00:21:40,210 --> 00:21:43,690 And that, that our technology is really based on as well, and it's not just 397 00:21:43,690 --> 00:21:45,930 Amazon Q Code transformer as we've seen. 398 00:21:46,420 --> 00:21:49,729 Uh, you know, IBM Watson Migration Assistant built on top of Open rer, 399 00:21:49,849 --> 00:21:53,219 Broadcom application advisor built on top of Open Rewrite Microsoft GitHub, 400 00:21:54,110 --> 00:21:58,929 copilot AI Migration Assistant, I think is the current name also built on that. 401 00:21:59,350 --> 00:22:00,870 And they, they're better together. 402 00:22:00,870 --> 00:22:04,600 I mean, it's, you know, that tool runs, uh, open rewrite to make a bunch of 403 00:22:04,920 --> 00:22:08,180 deterministic changes and then follows that up with further verification steps. 404 00:22:08,750 --> 00:22:13,490 That's, that is the, the golden path, I think, is trying to find ways in which. 405 00:22:13,860 --> 00:22:17,550 Non-determinism is helpful and to stitch together systems that are 406 00:22:17,550 --> 00:22:19,199 deterministic at their core as well. 407 00:22:19,560 --> 00:22:22,529 I hate to sound like an overwhelming cynic on this, but it's one of the 408 00:22:22,530 --> 00:22:27,639 things I'm best at, uh, it the Python two to Python three migration, because 409 00:22:28,100 --> 00:22:33,000 Unicode had no other real discernible reason, uh, took a. Decade in no large 410 00:22:33,040 --> 00:22:36,449 part because the single biggest breaking change was the way that print statements 411 00:22:36,450 --> 00:22:37,690 were then handled as a function. 412 00:22:37,960 --> 00:22:41,650 And you could get around that by importing the future package, uh, which affected a 413 00:22:41,650 --> 00:22:43,260 lot of, uh, two to three migration stuff. 414 00:22:43,590 --> 00:22:48,879 But it still took a decade for the system tools around the Red Hat ecosystem, for 415 00:22:48,880 --> 00:22:52,370 example, just run package management to be written, to take advantage of this. 416 00:22:52,580 --> 00:22:53,530 And that was. 417 00:22:53,830 --> 00:22:56,400 And please correct me if I'm wrong on this, a relatively 418 00:22:56,400 --> 00:23:00,360 trivial, straightforward uplift from Python two to Python three. 419 00:23:00,550 --> 00:23:01,989 There was just a lot of it. 420 00:23:03,530 --> 00:23:07,010 Going, looking at that migration to anything that's even slightly more 421 00:23:07,010 --> 00:23:09,720 complicated than that seal feels like at a past a certain point of 422 00:23:09,720 --> 00:23:12,970 scale, an impossibility, you clearly feel differently given that you've 423 00:23:12,970 --> 00:23:15,850 built a successful company and an open source project around this. 424 00:23:16,020 --> 00:23:16,410 Yeah. 425 00:23:16,450 --> 00:23:18,580 I think actually one of the characteristics that was difficult 426 00:23:18,590 --> 00:23:21,310 about that Python two to three migration are there's things like that that you 427 00:23:21,310 --> 00:23:24,550 described that were fairly simple changes and that were done at the language level. 428 00:23:24,920 --> 00:23:29,240 But alongside that came a host of other library changes that were made. 429 00:23:29,810 --> 00:23:32,740 Not really because of Python two to three, but because there's an 430 00:23:32,740 --> 00:23:34,632 opportunity, they're breaking things, we'll break things and everybody, 431 00:23:34,632 --> 00:23:36,439 everybody, lets just break things, right? 432 00:23:36,480 --> 00:23:40,290 And so a lot of people got stuck on not just the language level changes, 433 00:23:40,290 --> 00:23:43,149 but all those library changes that happened at the same time. 434 00:23:43,720 --> 00:23:47,020 And that's an interesting problem because it's kind of an 435 00:23:47,060 --> 00:23:48,810 unknown scoped problem, right? 436 00:23:49,020 --> 00:23:52,639 Well, how, how much breakage you have in your libraries very much depends 437 00:23:52,639 --> 00:23:54,350 on the libraries that you're using. 438 00:23:54,630 --> 00:23:59,340 Right now, um, so I mentioned earlier like the Spring Boot three 439 00:23:59,359 --> 00:24:01,600 migration, two to three migration open. 440 00:24:01,600 --> 00:24:04,360 Red recipe right now has 3,400 steps. 441 00:24:04,730 --> 00:24:08,000 I promise there's some part of two to three that we don't cover yet. 442 00:24:08,139 --> 00:24:09,230 I don't know what that is. 443 00:24:09,550 --> 00:24:11,049 Uh, but somebody will encounter it. 444 00:24:11,389 --> 00:24:12,710 And for them 445 00:24:12,719 --> 00:24:13,899 in production, most likely 446 00:24:14,420 --> 00:24:16,200 in, yeah, they're gonna be trying to run the recipe. 447 00:24:16,200 --> 00:24:19,750 They're gonna find something that, oh, you don't cover Camel or something great, you 448 00:24:19,750 --> 00:24:23,860 know, like, uh, and so, and that's fine, you know, and we will encounter that. 449 00:24:23,880 --> 00:24:27,030 And probably if they use Camel in one place, they use it a bunch of places. 450 00:24:27,580 --> 00:24:32,239 And so it'll be worth it then to build out that additional recipe that deals 451 00:24:32,250 --> 00:24:35,760 with that camel migration and then, you know, boom, and then you know that, 452 00:24:35,779 --> 00:24:39,020 and then that's sort of contributed back for the benefit of everybody else. 453 00:24:39,820 --> 00:24:45,090 I think what makes this approachable or tractable is really that we're all 454 00:24:45,099 --> 00:24:49,060 sort of building on the same substrate of third party and open source stuff. 455 00:24:50,750 --> 00:24:54,459 From JP Morgan all the way down to tiny like, you know, 15 person 456 00:24:54,470 --> 00:24:55,950 engineering team, modern, like, 457 00:24:56,030 --> 00:24:58,660 oh, oh, we, we can't overemphasize just how much open source 458 00:24:58,660 --> 00:24:59,940 has changed everything. 459 00:24:59,950 --> 00:25:03,960 Back in the Bell Labs days, seventies and eighties, it was everyone 460 00:25:03,960 --> 00:25:06,770 had to basically build their own primitives from the ground up. 461 00:25:06,770 --> 00:25:06,889 It, 462 00:25:07,559 --> 00:25:09,510 yeah, it was all completely bespoke. 463 00:25:10,350 --> 00:25:10,679 Yeah. 464 00:25:10,779 --> 00:25:14,325 Now it's almost become a trope like, so implement quicksort in a whiteboard. 465 00:25:14,390 --> 00:25:16,520 Like, why would I ever need to do that? 466 00:25:17,300 --> 00:25:18,000 Okay. 467 00:25:18,700 --> 00:25:22,299 Uh, I, I guess another, another angle on my skepticism here is 468 00:25:22,330 --> 00:25:27,680 I work with AWS Bills and the AWS billing ecosystem is vast. 469 00:25:27,730 --> 00:25:30,260 Uh, but, but the billing space is a bounded problem. 470 00:25:30,260 --> 00:25:30,650 Space. 471 00:25:30,709 --> 00:25:33,200 Unlike programming languages that are turning complete, you can 472 00:25:33,200 --> 00:25:34,820 build anything your heart desires. 473 00:25:35,240 --> 00:25:38,850 Uh, even in the billing space, I just came back from finops X in 474 00:25:38,850 --> 00:25:44,030 San Diego and none of the vendors are really making a strong AI play. 475 00:25:44,040 --> 00:25:47,820 And I'm not surprised by this because I have done a number of experiments with 476 00:25:47,820 --> 00:25:52,870 LLMs on AWS billing artifacts, and they consistently make the same types of 477 00:25:52,880 --> 00:25:55,210 errors that seem relatively intractable. 478 00:25:55,400 --> 00:25:57,500 Uh, go ahead and make this optimization. 479 00:25:58,160 --> 00:26:01,969 That optimization is dangerous without a little more context fed into it. 480 00:26:02,240 --> 00:26:05,860 So I guess my somewhat sophomoric perspective has been, if you, if you can't 481 00:26:05,860 --> 00:26:09,520 solve these things with AI in a bounded problem space, how can you begin to tackle 482 00:26:09,520 --> 00:26:11,590 them in these open-ended problem spaces? 483 00:26:12,470 --> 00:26:13,919 I, I'm, I'm with you actually. 484 00:26:13,920 --> 00:26:16,019 And there's a counterpoint to this, which is that I think. 485 00:26:16,420 --> 00:26:19,479 That the, all the large foundation models are somewhat undifferentiated. 486 00:26:20,420 --> 00:26:22,930 I mean, they kind of take pull position at any given time, but 487 00:26:23,389 --> 00:26:23,679 Right. 488 00:26:23,690 --> 00:26:25,940 Two weeks later, the whole ecosystem is different. 489 00:26:26,020 --> 00:26:26,500 Yeah. 490 00:26:26,560 --> 00:26:30,410 If they kind of all roughly have the same capabilities and there 491 00:26:30,420 --> 00:26:33,430 are some, like we said, there are very useful things they can do. 492 00:26:33,450 --> 00:26:34,689 There's some utility there. 493 00:26:35,389 --> 00:26:38,420 You know, there are places where non-determinism is useful 494 00:26:38,910 --> 00:26:41,239 and to the extent that you can apply that non-deterministic. 495 00:26:42,930 --> 00:26:43,440 Then great. 496 00:26:43,580 --> 00:26:45,970 You know, like that, that that's, that's fantastic. 497 00:26:46,430 --> 00:26:51,530 But I'm not in a position where I think Spring Boot two to three upgrade or 498 00:26:51,540 --> 00:26:55,790 Python two to three upgrade applied to 5 billion lines of code is going to be 499 00:26:56,440 --> 00:27:00,139 deterministically acceptable either now or six months from now or a year from now. 500 00:27:01,499 --> 00:27:03,830 And maybe I'll be a fool and wrong, but I don't think so. 501 00:27:04,670 --> 00:27:04,939 Oh yeah. 502 00:27:04,940 --> 00:27:08,370 Honestly, this is, this whole AI revolution has sh turned my entire 503 00:27:08,370 --> 00:27:11,389 understanding of how computers work on, its on their heads. 504 00:27:11,389 --> 00:27:14,940 Like short of a rand function, you, you knew what the output of 505 00:27:14,990 --> 00:27:17,590 a given stance of code was going to be, given a certain input. 506 00:27:17,920 --> 00:27:20,050 Now it kind of depends. 507 00:27:20,650 --> 00:27:21,120 It does. 508 00:27:21,410 --> 00:27:22,130 It really does. 509 00:27:23,340 --> 00:27:26,490 Yeah, the problem I run into is no matter how clever I have been able to 510 00:27:26,490 --> 00:27:29,470 be and the people I've worked with we're far smarter than I am, have been able 511 00:27:29,470 --> 00:27:31,930 to pull off, uh, the, these insights. 512 00:27:31,980 --> 00:27:34,989 There's always migration challenges and things breaking in production 513 00:27:34,990 --> 00:27:38,050 just because of edge and corner cases that we simply hadn't. 514 00:27:38,090 --> 00:27:42,080 Considered the difference now is instead because there's a culture 515 00:27:42,080 --> 00:27:45,280 in any healthy workplace about not throwing Steven under the bus. 516 00:27:45,500 --> 00:27:49,140 Well, throwing the robot under the bus is a very different proposition. 517 00:27:49,140 --> 00:27:51,620 I told you AI was craps as the half of your team. 518 00:27:51,620 --> 00:27:54,669 That's AI skeptic, and it's not the ai, it's. 519 00:27:54,670 --> 00:27:57,219 Vault says the, uh, people who are big into I, into I, 520 00:27:57,219 --> 00:27:59,030 ai, ai, business, daddy logic. 521 00:27:59,430 --> 00:28:02,450 And the reality is probably these things are complicated. 522 00:28:02,470 --> 00:28:06,620 Uh, compute, NN neither computer nor me nor beast are going to be able to 523 00:28:06,660 --> 00:28:08,390 catch all of these things in advance. 524 00:28:08,790 --> 00:28:10,019 That is why we have jobs. 525 00:28:10,529 --> 00:28:12,659 I've noticed this just in even managing our team. 526 00:28:12,659 --> 00:28:17,040 You know, I, I catch people when they say, you know, but Junie 527 00:28:17,050 --> 00:28:18,930 said this, or, but, you know. 528 00:28:19,630 --> 00:28:22,139 Uh, don't pass through to me what your assistant said. 529 00:28:22,500 --> 00:28:24,889 You, you're the responsible party when you tell me something. 530 00:28:24,910 --> 00:28:26,769 So you, you, you have a source. 531 00:28:27,049 --> 00:28:28,970 You check that source, you verify the integrity of the 532 00:28:29,020 --> 00:28:30,160 source, then you pass it to me. 533 00:28:30,250 --> 00:28:30,570 Right? 534 00:28:30,630 --> 00:28:32,850 You can outsource the work, but not the responsibility. 535 00:28:32,850 --> 00:28:36,140 A number of lawyers are finding this, uh, uh, uh, to be the case when they're, 536 00:28:36,150 --> 00:28:39,300 they're not checking what paralegals have done or at least blaming the paralegals 537 00:28:39,349 --> 00:28:40,490 for it, I'm sure. 538 00:28:40,840 --> 00:28:41,510 Exactly. 539 00:28:41,550 --> 00:28:42,340 Always has been. 540 00:28:42,809 --> 00:28:43,729 Always has, but 541 00:28:44,090 --> 00:28:47,849 it's, I, I also do worry that a lot of the skepticism around this, uh, even 542 00:28:47,850 --> 00:28:52,680 my own, my own aspect of it comes from a conscious or unconscious level of 543 00:28:52,820 --> 00:28:57,220 defensiveness where I'm worried this thing is going to take my job away. 544 00:28:57,220 --> 00:29:00,390 So the first thing I do, just to rationalize it to myself is point out 545 00:29:00,390 --> 00:29:03,000 that things I'm good at and that this thing isn't good at, at the moment. 546 00:29:03,270 --> 00:29:05,210 Well, that, that's why I'll always have a job. 547 00:29:05,990 --> 00:29:08,770 Conversely, I don't think computers are gonna take jobs away from all 548 00:29:08,770 --> 00:29:10,690 of us in the foreseeable future. 549 00:29:10,990 --> 00:29:14,149 The answer is probably a middle ground in a similar passion fashion. 550 00:29:14,160 --> 00:29:18,149 The way the industrial Revolution sort of did a whole lot of, uh, of a number on 551 00:29:18,170 --> 00:29:20,210 people who are an in independent artisans. 552 00:29:20,620 --> 00:29:25,980 So there's a, it's an evolutional process and I, I just worry that I am 553 00:29:26,140 --> 00:29:28,220 being too defensive, even unconsciously. 554 00:29:29,400 --> 00:29:30,850 I, I think that's sometimes too. 555 00:29:30,850 --> 00:29:35,180 I, I really do feel like this is just a continuum of, of productivity 556 00:29:35,180 --> 00:29:38,899 improvement that's been underfoot for a long time with different technologies. 557 00:29:38,900 --> 00:29:42,960 And I mean, I remember the very first eclipse release and the 558 00:29:42,960 --> 00:29:46,870 very first eclipse release is when they were providing, you know, uh, 559 00:29:46,880 --> 00:29:49,249 rules-based refactorings inside the IDE. 560 00:29:49,489 --> 00:29:51,850 And I remember being super excited every two or three 561 00:29:51,850 --> 00:29:52,980 months when they dropped another. 562 00:29:52,980 --> 00:29:55,180 And just looking at the release notes and seeing all the new. 563 00:29:55,690 --> 00:29:56,980 Things and what did that do? 564 00:29:56,980 --> 00:29:59,650 It made me faster at writing that new code. 565 00:30:00,059 --> 00:30:03,630 And you know, here we've got another thing that has very different characteristics. 566 00:30:03,640 --> 00:30:07,260 It's like, it's almost good at all the things that IDE based refactor 567 00:30:07,320 --> 00:30:10,790 weren't good at, but I still guide it. 568 00:30:11,059 --> 00:30:15,420 And, you know, unlike a, yeah, I think the, the drop CEO said, 569 00:30:15,530 --> 00:30:19,260 uh, our CTO said IDs will be obsolete by the end of the year. 570 00:30:20,060 --> 00:30:21,080 I don't believe this at all. 571 00:30:21,750 --> 00:30:22,520 I don't believe this at all. 572 00:30:22,530 --> 00:30:23,650 I think we're still driving them. 573 00:30:24,400 --> 00:30:26,610 I am skeptical in the extreme on a lot of that. 574 00:30:26,690 --> 00:30:32,389 I, I, I, because again, these, let's be honest here, these people have a thing 575 00:30:32,400 --> 00:30:36,910 they need to sell and they have billions and billions and billions of other 576 00:30:36,910 --> 00:30:38,800 people's money riding on the outcome. 577 00:30:39,250 --> 00:30:40,360 Yeah. 578 00:30:40,360 --> 00:30:42,680 That would shape my thinking in a bunch of ways. 579 00:30:42,680 --> 00:30:44,220 Both subtle and grows too. 580 00:30:44,559 --> 00:30:49,420 I try and take the, a more neutral stance on this, but who knows? 581 00:30:49,440 --> 00:30:53,299 I think it's not just neutral, it's a mature stance and it's 582 00:30:53,309 --> 00:30:57,250 one that's, uh, it's, it's a lot of experience going behind it. 583 00:30:57,310 --> 00:30:58,186 I, I think that you're right. 584 00:30:58,250 --> 00:31:01,080 I don't think we're we're anywhere close to being obsolete. 585 00:31:02,130 --> 00:31:05,989 No, and I, and I also, frankly, I say this coming from an operations background, 586 00:31:05,990 --> 00:31:08,199 CIS had been turned SRE type where. 587 00:31:08,710 --> 00:31:13,899 I have been through enough cycles of seeing today's magical technology become 588 00:31:13,939 --> 00:31:18,860 tomorrow's legacy shit that I have to support that I am, I have a natural 589 00:31:18,860 --> 00:31:23,999 skepticism built into almost every aspect of this just based on history. 590 00:31:24,009 --> 00:31:24,570 If nothing else, 591 00:31:25,000 --> 00:31:26,560 you know what Vibe Coding reminds me of? 592 00:31:26,560 --> 00:31:30,459 It reminds me of model driven architecture about 25 years ago. 593 00:31:30,480 --> 00:31:33,700 The like, you know, just produce a UML diagram and don't 594 00:31:33,700 --> 00:31:36,000 worry like the, the codal. 595 00:31:36,000 --> 00:31:38,630 I'll ship the, I'll just generate the rest of the application. 596 00:31:38,980 --> 00:31:43,010 Or it reminds me of, uh, behavior driven development when we said, oh, we'll 597 00:31:43,010 --> 00:31:44,549 just put in business people's hands. 598 00:31:44,700 --> 00:31:46,709 They write the test and, you know, don't, you know, we don't 599 00:31:46,709 --> 00:31:47,789 want engineers writing the test. 600 00:31:47,789 --> 00:31:48,230 You want business? 601 00:31:48,740 --> 00:31:53,400 Like, I feel like we've seen this play out many, many, many times in various 602 00:31:53,410 --> 00:31:55,230 forms, and maybe this time's different. 603 00:31:55,300 --> 00:31:56,250 I, I don't think so. 604 00:31:57,090 --> 00:31:59,694 And to be honest, I, I like to say that, well, computers used to be 605 00:31:59,830 --> 00:32:03,419 deterministic, but let's, let's be honest with ourselves, we'd long ago crossed 606 00:32:03,420 --> 00:32:07,179 threshold where no individual person can hold the entirety of what even a 607 00:32:07,190 --> 00:32:08,989 simple function is doing in their head. 608 00:32:09,290 --> 00:32:11,649 They, they are putting their trust in the magic think box. 609 00:32:12,920 --> 00:32:13,370 That's right. 610 00:32:13,870 --> 00:32:14,199 Yes. 611 00:32:14,200 --> 00:32:15,010 That's absolutely right. 612 00:32:15,710 --> 00:32:17,849 So I really wanna thank you for taking the time to speak with me. 613 00:32:17,860 --> 00:32:20,379 If people can, people wanna go and learn more, where's the 614 00:32:20,380 --> 00:32:21,939 best place for them to find you? 615 00:32:22,690 --> 00:32:26,929 I think it's, it's easy to find me on LinkedIn these days or, uh, you know, 616 00:32:26,930 --> 00:32:29,370 go find me on Moderna, M-O-D-E-R-N e.ai. 617 00:32:30,130 --> 00:32:31,320 Um, either place. 618 00:32:31,400 --> 00:32:34,230 Happy to always, uh, send me a dm. Happy to answer questions 619 00:32:34,760 --> 00:32:36,689 and we'll of course put that into the show notes. 620 00:32:36,820 --> 00:32:38,840 Thank you so much for your time, I appreciate it. 621 00:32:39,080 --> 00:32:40,040 Okay, thank you Corey. 622 00:32:40,820 --> 00:32:43,610 Jonathan Schneider, CEO at modern. 623 00:32:43,730 --> 00:32:47,070 I'm Cloud Economist Corey Quinn, and this is Screaming In the Cloud. 624 00:32:47,320 --> 00:32:50,480 If you've enjoyed this podcast, please leave a five star review on 625 00:32:50,480 --> 00:32:52,070 your podcast platform of choice. 626 00:32:52,110 --> 00:32:55,820 Whereas if you hated this podcast, please leave a five star review on 627 00:32:55,820 --> 00:33:00,169 your podcast platform of choice along with an insulting comment that maybe 628 00:33:00,170 --> 00:33:04,340 you can find an AI system to transform into something halfway literate.