1 00:00:00,120 --> 00:00:02,680 Secure your Elixir apps with paraxial IO. 2 00:00:02,880 --> 00:00:05,920 From code analysis and bot defense to dependency checks, 3 00:00:05,920 --> 00:00:09,160 they've got you covered. Elixir first design with GitHub 4 00:00:09,160 --> 00:00:11,600 integration. Elevate your security at 5 00:00:11,600 --> 00:00:16,520 paraxial dot IO. Hey, I'm Kimberly with Elixir 6 00:00:16,520 --> 00:00:20,040 mentor Jacob recently added recruitment services to help 7 00:00:20,040 --> 00:00:22,240 companies hire pre vetted Elixir devs. 8 00:00:22,760 --> 00:00:26,000 I help with company outreach, so if you're hiring, I may be 9 00:00:26,000 --> 00:00:29,000 reaching out, or you can connect with us anytime at 10 00:00:29,000 --> 00:00:39,560 elixirmentor.com. Testing what's up? 11 00:00:39,560 --> 00:00:42,320 I'm Jacob Blitzo, your Elixir mentor, and welcome to another 12 00:00:42,320 --> 00:00:44,840 exciting episode of the Elixir Mentor Podcast. 13 00:00:45,040 --> 00:00:47,240 This is where we discuss everything related to Elixir, 14 00:00:47,240 --> 00:00:49,320 from interviews with enthusiasts, pioneers in the 15 00:00:49,320 --> 00:00:52,240 community, to innovative projects and libraries shaping 16 00:00:52,240 --> 00:00:55,640 Elixir's future. Today I'm joined by Josh Price, 17 00:00:55,640 --> 00:00:59,440 founder and CTO of Olympic. You've probably heard of him 18 00:00:59,440 --> 00:01:03,560 just being involved with ASH, and we're going to talk a little 19 00:01:03,560 --> 00:01:06,640 bit about a new project he's been working on called Clarity, 20 00:01:06,680 --> 00:01:09,320 and probably talk about just building a consultancy in 21 00:01:09,320 --> 00:01:13,560 general and who knows, maybe what AI means for developers. 22 00:01:13,840 --> 00:01:17,080 I think we have a whole assortment of conversations, so 23 00:01:17,080 --> 00:01:18,760 thank you so much for joining me today. 24 00:01:20,120 --> 00:01:22,960 No worries, Jacob. Great to be here. 25 00:01:23,600 --> 00:01:25,880 Yeah, I think it's going to be a fun conversation. 26 00:01:25,880 --> 00:01:29,640 And for everyone joining us, I know this isn't our normal time, 27 00:01:29,640 --> 00:01:33,840 so thanks for tuning in. It's difficult to connect with 28 00:01:33,840 --> 00:01:37,520 people from from across the the world, right? 29 00:01:37,520 --> 00:01:40,880 I think you're the furthest you can be as far as time. 30 00:01:41,040 --> 00:01:42,960 Zones go. The New Zealanders will be mad 31 00:01:42,960 --> 00:01:46,640 that you said that, Jacob. Yeah, but you don't care about 32 00:01:46,640 --> 00:01:52,720 them, do you? We care greatly, but we like we 33 00:01:52,720 --> 00:01:57,080 like making fun of each other. Something about sheep and stuff, 34 00:01:57,080 --> 00:01:59,760 right? I don't know if you can go there 35 00:01:59,760 --> 00:02:03,480 man. I have one of my cousins is 36 00:02:03,480 --> 00:02:06,520 married to a Kiwi so I think I'm allowed to make fun of them a 37 00:02:06,520 --> 00:02:07,920 little. Bit so you've heard the jokes. 38 00:02:08,160 --> 00:02:10,440 Yeah, I don't. Know if we do those anymore? 39 00:02:11,160 --> 00:02:14,720 You don't do them anymore. I don't think so, but they got a 40 00:02:14,720 --> 00:02:16,840 lot of mileage. That's funny. 41 00:02:19,040 --> 00:02:22,160 Yeah. So to kick things off, you wanna 42 00:02:22,160 --> 00:02:26,360 kind of for those of us that might not know you kind of 43 00:02:26,360 --> 00:02:29,320 explain who you are and your background a little bit. 44 00:02:30,400 --> 00:02:33,200 Yeah, sure. So I'm Josh, if you haven't 45 00:02:33,200 --> 00:02:37,960 figured that out already. I've been doing Elixir for 46 00:02:37,960 --> 00:02:44,920 probably over 10 years, which is, you know, a fairly long time 47 00:02:45,480 --> 00:02:47,360 I've had. OG status. 48 00:02:48,160 --> 00:02:54,360 Probably 30 years of programming plus, no, probably 40 if you 49 00:02:54,360 --> 00:02:58,160 count, you know, where I started, you know, poking 50 00:02:58,160 --> 00:03:03,640 Commodore 60 Fours and Vic 20s and very early computers, Apple 51 00:03:03,640 --> 00:03:08,800 2ES and CS. But yeah, I probably had three 52 00:03:09,440 --> 00:03:12,320 decades of my career. Like the first one was Java, 53 00:03:12,760 --> 00:03:16,360 second one was Ruby, and then the third one was definitely 54 00:03:16,400 --> 00:03:19,560 focused on the Lixa. And yeah, I don't see that 55 00:03:19,560 --> 00:03:22,840 changing. And it's funny that we've 56 00:03:22,840 --> 00:03:24,840 arrived at and we can talk about this later. 57 00:03:25,560 --> 00:03:28,640 You know, the, the the Twittersphere, if we call it 58 00:03:28,640 --> 00:03:32,480 that, I'm not calling it the new thing is abuzz with the fact 59 00:03:32,480 --> 00:03:35,640 that, you know, people realizing that Elixir and Erlang and the 60 00:03:35,640 --> 00:03:40,120 BEAM are perfect, you know, infrastructure for running 61 00:03:40,120 --> 00:03:43,680 modern identic systems. I'm surprised that's too long to 62 00:03:43,720 --> 00:03:45,520 be honest. Right. 63 00:03:46,240 --> 00:03:50,160 And then not long after I discovered Elixir, I was like, I 64 00:03:50,160 --> 00:03:51,800 need to start a company doing this. 65 00:03:51,800 --> 00:03:55,880 And me and a buddy, James Sadler, who is off doing some 66 00:03:55,880 --> 00:04:01,000 very, very deep tech these days doing like, what's it called? 67 00:04:01,000 --> 00:04:03,680 He works for a company called Cipherstash where it's super 68 00:04:03,680 --> 00:04:07,080 deep encryption where you can do lots of stuff while still 69 00:04:07,080 --> 00:04:10,520 keeping the data encrypted. Encryption and use is sort of 70 00:04:10,520 --> 00:04:11,960 the thing. OK, nice. 71 00:04:11,960 --> 00:04:16,680 I hear they're probably making some waves in the space. 72 00:04:16,680 --> 00:04:20,120 It's pretty cool, but yeah, we've been doing that for like 9 73 00:04:20,120 --> 00:04:23,280 years now. I think it was our birthday in 74 00:04:23,280 --> 00:04:25,240 February. We are very bad at celebrating 75 00:04:25,240 --> 00:04:26,560 these things. Maybe we'll wait for the 10th 76 00:04:26,560 --> 00:04:27,800 one. Plan that better. 77 00:04:28,600 --> 00:04:29,640 That's awesome. Yeah, that's very cool. 78 00:04:31,240 --> 00:04:33,560 It's a short history of what I've been doing. 79 00:04:33,920 --> 00:04:38,040 So you started your consultancy after finding Elixir? 80 00:04:39,120 --> 00:04:42,640 Right. So I was working at a gaming and 81 00:04:42,640 --> 00:04:49,000 betting company for my sins like the the domain was not really my 82 00:04:49,160 --> 00:04:53,080 jam. But yeah, it was super 83 00:04:53,080 --> 00:04:56,360 interesting work. And what it required was because 84 00:04:56,360 --> 00:04:58,640 it was very fast moving market data. 85 00:04:59,120 --> 00:05:03,920 And the faster, the closer you got to an event, the faster and 86 00:05:03,920 --> 00:05:06,600 the more traffic and the more that the data would move. 87 00:05:07,560 --> 00:05:13,200 And so we solved that pretty inelegantly with basically 88 00:05:13,200 --> 00:05:15,920 caching and refreshing every 10 seconds. 89 00:05:17,160 --> 00:05:19,360 It did the trick. It's a plus on that how you'll 90 00:05:19,480 --> 00:05:20,000 do. Things. 91 00:05:21,000 --> 00:05:24,800 I don't know. It's, look, it's, I think we can 92 00:05:24,800 --> 00:05:27,360 talk about this later as well as like sometimes the dump solution 93 00:05:27,360 --> 00:05:29,920 is really brutally effective and that worked great. 94 00:05:29,920 --> 00:05:34,080 But you had sort of an average time of stale data at 5 seconds, 95 00:05:34,160 --> 00:05:37,680 which when you have really, really fast moving data may not 96 00:05:37,920 --> 00:05:40,840 cut it. If anyone's done any trading of 97 00:05:40,840 --> 00:05:42,880 any sort, then you know what I mean. 98 00:05:43,320 --> 00:05:44,600 Yeah, for. Sure. 99 00:05:44,600 --> 00:05:47,840 And so, yeah, I saw a couple of things in that project. 100 00:05:49,360 --> 00:05:53,640 Not only was Node really expensive and, you know, AWS can 101 00:05:53,920 --> 00:05:56,520 charge you an arm and a leg and you need a lot of Node servers 102 00:05:57,040 --> 00:06:01,160 to to do that kind of strategy. And I was like, and we also had 103 00:06:01,160 --> 00:06:07,040 like a REST API that was I guess artisanally handcrafted. 104 00:06:07,680 --> 00:06:12,640 And so, you know, we saw drift across all kinds of different 105 00:06:12,840 --> 00:06:14,960 parts of the system that people were building out. 106 00:06:14,960 --> 00:06:18,360 Like everywhere you looked, every bit was different to every 107 00:06:18,360 --> 00:06:21,160 other bit, just in subtle, nuanced ways. 108 00:06:22,200 --> 00:06:24,160 You just didn't have that consistency. 109 00:06:24,480 --> 00:06:27,920 And so, you know, with an artisanally handcrafted REST 110 00:06:28,040 --> 00:06:31,200 API, which was honestly the winning strategy here because we 111 00:06:31,200 --> 00:06:35,680 were wrapping old, you know, legacy technology that they'd 112 00:06:35,680 --> 00:06:39,560 spent a lot of money and a lot of time failing to, to migrate 113 00:06:39,560 --> 00:06:43,560 to something more modern. And so we kind of threw the, the 114 00:06:43,560 --> 00:06:46,560 API rug over it. And we were able to replace an 115 00:06:46,560 --> 00:06:50,800 existing Flash website that a BIG4 consultancy had put in 116 00:06:50,800 --> 00:06:55,360 there after Steve Jobs had said, hey, don't, don't no flash on my 117 00:06:55,480 --> 00:06:59,000 Apple devices. And so they had a Flash website 118 00:06:59,000 --> 00:07:02,560 and they had no mobile, you know, and that was not 119 00:07:02,600 --> 00:07:08,000 acceptable at the time. So basically the API surface 120 00:07:08,000 --> 00:07:11,240 area sort of saved us from, you know, the monsters underneath. 121 00:07:11,320 --> 00:07:15,360 And we built all of this modern stuff that was pretty great, but 122 00:07:15,360 --> 00:07:18,640 I thought it could be better. And Elixir and Graph QL was sort 123 00:07:18,640 --> 00:07:21,200 of the two dimensions on which I thought it could be better. 124 00:07:23,720 --> 00:07:27,480 First of all, like REST APIs, it's HTTP One, it's request 125 00:07:27,480 --> 00:07:30,480 response. So there was no way to push fast 126 00:07:30,480 --> 00:07:34,000 changing market data to the front end and to the user's 127 00:07:35,040 --> 00:07:37,680 devices. And so you have to request. 128 00:07:37,680 --> 00:07:40,680 And so that's why the caching strategy every 10 seconds worked 129 00:07:41,720 --> 00:07:44,120 so brutally well, brutally effective. 130 00:07:44,120 --> 00:07:47,160 Sometimes you know you just have to make your hacky solution and 131 00:07:47,280 --> 00:07:48,960 no one really needs to know how you did it. 132 00:07:49,720 --> 00:07:52,200 Well, exactly. But what I was dreaming of was 133 00:07:52,200 --> 00:07:54,480 like, well, why aren't we pushing the data only when it's 134 00:07:54,480 --> 00:07:58,680 changed? And so that led me down a path 135 00:07:58,680 --> 00:08:00,120 of like, well, what do we need here? 136 00:08:00,120 --> 00:08:03,240 Like web sockets would be great and open connections so you 137 00:08:03,240 --> 00:08:04,800 don't have to like how do you do push? 138 00:08:05,720 --> 00:08:08,640 And APIs, certainly REST APIs didn't support it. 139 00:08:08,640 --> 00:08:12,560 So I was looking at Graph QL, which at the time was pretty new 140 00:08:13,120 --> 00:08:20,440 and also had a subscription idea and some of the spec was also 141 00:08:20,440 --> 00:08:25,040 looking at defer and stream and sort of all all that kind of 142 00:08:25,040 --> 00:08:27,640 stuff. How do you do you do you still 143 00:08:27,640 --> 00:08:31,440 like graph QL today I. Know we use it a lot. 144 00:08:31,640 --> 00:08:34,919 Yeah, but I'm a I'm a big fan of it personally, but a lot of 145 00:08:34,919 --> 00:08:39,280 people have very harsh opinions of it, but I think it's. 146 00:08:40,559 --> 00:08:46,720 I was super bullish. I was super bullish about it and 147 00:08:46,720 --> 00:08:53,640 I think possibly over bullish. I think it has some problems and 148 00:08:53,640 --> 00:08:57,960 I think, you know, it's really picked a tool for the job and in 149 00:08:58,120 --> 00:09:01,360 it's kind of if I need an API, it's the first tool I would 150 00:09:01,360 --> 00:09:03,800 reach for. But it's not the only tool in 151 00:09:03,800 --> 00:09:06,000 the tool belt. So knowing when to use it is 152 00:09:06,000 --> 00:09:08,320 important. Yeah, but I think it doesn't get 153 00:09:08,320 --> 00:09:13,320 enough airtime and credit for the subscription stuff which is 154 00:09:13,320 --> 00:09:17,000 allowing push through the API, which changes the game in terms 155 00:09:17,000 --> 00:09:19,680 of APIs. Yeah, to be able to subscribe to 156 00:09:19,680 --> 00:09:21,960 events and things is pretty awesome. 157 00:09:22,920 --> 00:09:26,160 Or you know, just changes whatever you're subscribing to, 158 00:09:26,160 --> 00:09:28,080 right? I also like just the 159 00:09:28,080 --> 00:09:33,280 flexibility, like I feel like as long as you have some 160 00:09:33,280 --> 00:09:37,680 restrictions around your like mutations and queries, it's nice 161 00:09:37,680 --> 00:09:40,320 the flexibility you can still give the front end for 162 00:09:40,440 --> 00:09:43,720 requesting data without just getting a bloated response, you 163 00:09:43,720 --> 00:09:45,640 know? Yeah. 164 00:09:45,640 --> 00:09:49,080 And look, there's a ton of ways that graph QL is just super 165 00:09:49,080 --> 00:09:53,000 elegant. And, you know, the way I liked 166 00:09:53,000 --> 00:09:55,920 to think about it and I still like to think about it is it's, 167 00:09:57,200 --> 00:10:00,920 although, you know, Ash probably changes that graph QL by itself 168 00:10:00,920 --> 00:10:03,320 is probably a super flexible API toolkit. 169 00:10:03,800 --> 00:10:09,000 And you can do REST with, you know, a graph QL, you know, 170 00:10:09,160 --> 00:10:12,960 graph because all you're doing is sort of projecting a tree and 171 00:10:12,960 --> 00:10:16,480 starting at a root node and projecting a static tree over 172 00:10:16,480 --> 00:10:19,040 that. And think of just mounting a 173 00:10:19,040 --> 00:10:24,960 graph QL query at an endpoint like a URL and you've got REST 174 00:10:25,200 --> 00:10:28,200 and you can just define, you know what all the queries for 175 00:10:28,200 --> 00:10:32,640 the things you can get. So you can sort of get REST 176 00:10:33,200 --> 00:10:35,480 through graph QL, which is very elegant. 177 00:10:35,520 --> 00:10:38,840 And you can have graph QL, surf graph QL, surf graph QL, it 178 00:10:38,840 --> 00:10:41,840 doesn't matter where it is. It was, you know, a standardised 179 00:10:41,840 --> 00:10:44,720 way of, you know, describing the world. 180 00:10:45,160 --> 00:10:51,120 So that was 1 dimension of what I saw at this this company. 181 00:10:51,680 --> 00:10:54,360 And then the other dimension was like, OK, well, if you've got 182 00:10:54,440 --> 00:10:59,440 open sockets, there's no way that node would reasonably and 183 00:10:59,680 --> 00:11:04,480 and sort of economically efficiently serve that many open 184 00:11:04,480 --> 00:11:07,040 connections. Look, it probably could, but it 185 00:11:07,040 --> 00:11:10,680 would cost you a lot. And I do recall the AWS fuels 186 00:11:10,680 --> 00:11:15,440 being quite large. It's no surprise or irony that 187 00:11:16,160 --> 00:11:21,200 my manager and buddy now runs AWS in Australia. 188 00:11:22,040 --> 00:11:26,440 Oh, that's awesome. And so that gives you a sense of 189 00:11:26,440 --> 00:11:30,560 like, yeah, OK, you can turn the taps on and, and certainly 190 00:11:30,560 --> 00:11:32,760 charge a lot of people a lot of money if you don't choose the 191 00:11:32,760 --> 00:11:37,440 right technical solution. So yeah, like that was where I 192 00:11:37,440 --> 00:11:39,400 landed on Elixir. And I was like, Oh my God. 193 00:11:39,400 --> 00:11:42,160 Like Elixir and Erlang, this is perfect for this like open 194 00:11:42,160 --> 00:11:47,320 connections, pushing fast moving data up to client devices. 195 00:11:47,320 --> 00:11:49,760 And that was the sort of 2 dimensions on which we founded 196 00:11:49,760 --> 00:11:53,240 Olympic. So that was, that was pretty 197 00:11:53,240 --> 00:11:56,200 much it. And you know, graph QL look, we 198 00:11:56,200 --> 00:11:59,320 use it a ton. Like we don't sort of think it's 199 00:11:59,520 --> 00:12:03,560 the most important thing like we used to look, use it where you 200 00:12:03,560 --> 00:12:05,080 need it. Yeah, I think that's a great. 201 00:12:05,080 --> 00:12:09,080 Deal for we usually, I mean we usually still use it for mobile 202 00:12:09,080 --> 00:12:14,640 apps, so and then Ash came along a bit later, basically allowing 203 00:12:14,640 --> 00:12:17,560 us to spin up whatever API we want. 204 00:12:17,640 --> 00:12:21,360 I'm curious what brought, what brought ash on your radar and 205 00:12:21,360 --> 00:12:24,360 like what made you go so bullish with it and like bring Zach 206 00:12:24,360 --> 00:12:29,320 Daniel on board and all that? So this is a good question, and 207 00:12:29,320 --> 00:12:32,360 I'd actually forgotten the origin story until I just 208 00:12:32,360 --> 00:12:33,840 started telling the previous story. 209 00:12:35,280 --> 00:12:41,640 So it came through graph QL and so I thought graph QL was super 210 00:12:41,640 --> 00:12:46,400 simple, like it's only got a couple of primitives and in 211 00:12:46,400 --> 00:12:49,840 terms of primitive types, that is it's kind of problem these 212 00:12:49,840 --> 00:12:51,440 days. We can talk about that later if 213 00:12:51,440 --> 00:12:54,000 you want. So I had a couple of primitive 214 00:12:54,000 --> 00:12:57,120 types and then a way to model like a graph of your domain. 215 00:12:57,560 --> 00:13:02,840 And so I started thinking, could we using graph QLSDL, the schema 216 00:13:03,240 --> 00:13:08,000 definition language actually modeled domains in this and you 217 00:13:08,000 --> 00:13:12,280 basically could. What it lacked was a few things 218 00:13:12,280 --> 00:13:16,760 like union types and, you know, I would model domains and go, 219 00:13:16,760 --> 00:13:23,720 yeah, like API first modelling. It wasn't terrible except there 220 00:13:23,720 --> 00:13:27,560 wasn't anywhere to put the business logic or describe what 221 00:13:27,560 --> 00:13:29,280 happened when you mutated a thing. 222 00:13:29,400 --> 00:13:36,920 And so, you know, fun fact, Elixir Graph QL was my first 223 00:13:37,840 --> 00:13:42,000 hello world Elixir project. And so I spent like months 224 00:13:42,000 --> 00:13:46,200 writing the parser for that, which actually got borrowed and 225 00:13:46,200 --> 00:13:50,640 used into absinthe. And I wrote it in lex and yek, 226 00:13:51,120 --> 00:13:54,800 which is, if you don't know what those are, those are Erlang's 227 00:13:54,880 --> 00:14:02,360 lex and yek, pasa and lexa. And so yeah, like I'd spent a 228 00:14:02,360 --> 00:14:04,480 ton of work in that. And then Absinthe used that to 229 00:14:04,480 --> 00:14:09,120 sort of leapfrog into a much more elegant solution that we 230 00:14:09,120 --> 00:14:12,400 see today. And yeah, so I was trying to 231 00:14:12,400 --> 00:14:14,720 build graph QL and Elixir for exactly that reason. 232 00:14:14,720 --> 00:14:17,240 And I was like, we can model stuff in this, but there was 233 00:14:17,240 --> 00:14:22,600 nowhere in graph QL to actually put the domain logic and the 234 00:14:22,600 --> 00:14:25,560 things that sort of aren't just the nouns and relationships and 235 00:14:25,560 --> 00:14:28,440 the types. So I was like, this is nearly 236 00:14:28,440 --> 00:14:34,240 there, but it's not quite. And I did meet, actually my 237 00:14:34,240 --> 00:14:36,960 first Elixir conference was in Berlin in 2016. 238 00:14:37,640 --> 00:14:42,120 And me and James Sadler went over there and we met up with, 239 00:14:42,600 --> 00:14:46,560 so this was Alexicomp EU and I think it was one of the first, 240 00:14:47,400 --> 00:14:51,080 one of the first ones. But we also met up with a whole 241 00:14:51,080 --> 00:14:54,720 bunch of Graf QL folks. So Bruce Williams was there. 242 00:14:57,120 --> 00:14:59,720 Who else was there? The guy that wrote Sangria, I 243 00:14:59,720 --> 00:15:03,120 forget his name, who's no longer with us, which is a shame. 244 00:15:03,120 --> 00:15:06,120 He was amazing. He built built the scalar graph 245 00:15:06,120 --> 00:15:08,880 QL implementation. Andy Marek built the Java one, 246 00:15:09,560 --> 00:15:14,080 and then Johannes Schickling, who started graphical and then 247 00:15:14,080 --> 00:15:17,040 ended up doing Prisma. We're also there. 248 00:15:17,040 --> 00:15:21,000 So we met up in a pub in Berlin, had some beers and chatted. 249 00:15:21,040 --> 00:15:23,440 Graph. So that was. 250 00:15:23,600 --> 00:15:25,800 That's awesome. Berlin's a fun city. 251 00:15:25,800 --> 00:15:28,160 I I wanna go back. I need an excuse. 252 00:15:28,160 --> 00:15:30,720 I was there for like exactly 8 hours once. 253 00:15:31,160 --> 00:15:35,640 I'm like a long layover and I just. 8 hours is not enough time 254 00:15:35,680 --> 00:15:38,040 in. No, I know I but I took 255 00:15:38,040 --> 00:15:40,520 advantage of being there and I just like paid a taxi cab 256 00:15:40,520 --> 00:15:43,320 driver. I was like, take me to all like 257 00:15:43,760 --> 00:15:46,840 the German food, like all the good restaurants that I should 258 00:15:46,840 --> 00:15:50,440 go to and we just like drove around for 8 hours and I just 259 00:15:50,440 --> 00:15:52,040 ate a ton of food and drank beer. 260 00:15:53,280 --> 00:15:56,800 Did you have Curry versed? I don't think so. 261 00:15:56,840 --> 00:15:59,120 No street food Classic. Nice. 262 00:15:59,400 --> 00:16:02,800 I've got to say I I'm I'm sorry to all the Germans, but I don't 263 00:16:02,800 --> 00:16:05,040 rate it. It's a street food, like it's 264 00:16:05,040 --> 00:16:08,160 good, but it's not amazing. Yeah, yeah, I hear you. 265 00:16:08,800 --> 00:16:11,240 And I travel a lot and I'm it's just heavy, but. 266 00:16:11,240 --> 00:16:16,040 Just all meat, you know? But it's, you know, it's like a 267 00:16:16,040 --> 00:16:17,520 process to me. So it's like. 268 00:16:18,240 --> 00:16:20,920 Yeah, it's. A very particular taste and I I 269 00:16:21,960 --> 00:16:24,960 definitely had a long conversation last time I was in 270 00:16:25,520 --> 00:16:27,480 Berlin and. Yeah. 271 00:16:28,600 --> 00:16:31,520 I offended someone so apologies to all the Germans. 272 00:16:31,800 --> 00:16:33,680 German Curry. Offender way. 273 00:16:33,680 --> 00:16:35,880 No, I'm just kidding. Yeah. 274 00:16:35,880 --> 00:16:41,560 So was when you met Zach Daniel, was he just already working on 275 00:16:41,560 --> 00:16:45,120 Ash or was it like an idea that you kind of? 276 00:16:45,120 --> 00:16:46,920 Both. Came up with. 277 00:16:48,160 --> 00:16:52,240 Well, remember how I was like trying to, you know, model 278 00:16:52,240 --> 00:16:57,440 domains in graph QL and you know, people thought I was crazy 279 00:16:57,480 --> 00:17:02,160 and it didn't quite work because we were missing whole chunks and 280 00:17:02,160 --> 00:17:04,000 the type system wasn't quite there. 281 00:17:04,680 --> 00:17:07,280 And I'm like, this is close. What is it? 282 00:17:07,440 --> 00:17:13,119 And at the same time, I saw like an early incarnation of Ash and 283 00:17:13,119 --> 00:17:16,400 it was it sort of, I don't even remember what what it said. 284 00:17:16,920 --> 00:17:18,520 I think we looked at the Wayback machine. 285 00:17:18,520 --> 00:17:21,160 If you look at the early ash websites, you can find it. 286 00:17:22,280 --> 00:17:25,480 We did that the Ash Summit. And it's kind of funny, like it 287 00:17:25,480 --> 00:17:28,920 was on my To Do List, but I didn't realize really what it 288 00:17:28,920 --> 00:17:30,520 did. And so I looked at it and 289 00:17:30,520 --> 00:17:33,120 thought, oh, that's a way of sort of auto spinning up some 290 00:17:33,120 --> 00:17:37,840 graph QL AP is using Alexa. And I, I sort of had it on my To 291 00:17:37,840 --> 00:17:39,480 Do List for like a couple of years. 292 00:17:40,240 --> 00:17:41,760 I'm like, yeah, I should check that out one day. 293 00:17:41,880 --> 00:17:46,640 I didn't get around to it. Then we had a client project 294 00:17:46,920 --> 00:17:52,920 where it was like, oh, we need to build, you know, basically 295 00:17:52,920 --> 00:17:58,920 Elixa, we needed to build out a mobile like API because they 296 00:17:58,920 --> 00:18:01,360 wanted to have, you know, mobile devices. 297 00:18:02,120 --> 00:18:06,880 And this was our sort of, you know, first really big Ash 298 00:18:06,880 --> 00:18:08,840 project. And I was like, Ash would be 299 00:18:08,840 --> 00:18:11,320 perfect for this. So I hit up Zack and I tried it 300 00:18:11,320 --> 00:18:12,920 out and I'm like, oh, this is quite good. 301 00:18:13,840 --> 00:18:17,360 And then I realized once I drove into Ash that it actually solved 302 00:18:17,360 --> 00:18:19,560 a lot of the problems that I was trying to solve. 303 00:18:20,280 --> 00:18:25,280 And so I sort of saw the power immediately because it just 304 00:18:25,280 --> 00:18:27,840 filled in the blanks what I was trying to do, modelling domains 305 00:18:27,840 --> 00:18:31,000 in Graph QL. And it modelled it, you know, in 306 00:18:31,120 --> 00:18:33,040 elegant DSLS at the Elixir layer. 307 00:18:33,040 --> 00:18:36,360 And then from that you could derive Graph QL or REST or 308 00:18:36,360 --> 00:18:38,120 whatever. And then I realized it was more 309 00:18:38,120 --> 00:18:41,600 than just an API generator. And I think a lot of people fell 310 00:18:41,600 --> 00:18:45,480 into the, it's just an API generator trap for a long time. 311 00:18:46,200 --> 00:18:48,640 And, you know, I was like, oh, we could use this for 312 00:18:48,640 --> 00:18:52,520 everything. And, you know, the more I dug, 313 00:18:52,520 --> 00:18:57,800 the more things I could see. And, you know, the client was a, 314 00:18:58,240 --> 00:19:01,840 you know, deeply technical and appreciated what that brought. 315 00:19:02,520 --> 00:19:05,000 And he basically asked a whole bunch of enterprise architects 316 00:19:05,000 --> 00:19:08,120 like that. He knew because we were like, we 317 00:19:08,120 --> 00:19:10,800 should try Ash on this. We're also trying to try live 318 00:19:10,800 --> 00:19:13,640 view. So, you know, he was already 319 00:19:13,640 --> 00:19:16,040 sort of bought into live view, but we didn't know about the 320 00:19:16,040 --> 00:19:18,880 component system. But we looked at Ash as well, 321 00:19:18,880 --> 00:19:22,680 trying to go, well, if we model everything in this, we can 322 00:19:22,680 --> 00:19:26,120 generate so much of the, you know, the enterprise stuff that 323 00:19:26,120 --> 00:19:29,480 we know we're going to need for this, you know, eventually very 324 00:19:29,480 --> 00:19:33,600 enterprisey app. And so he asked a bunch of his 325 00:19:33,600 --> 00:19:36,080 buddies that were enterprise architects and had worked on 326 00:19:36,080 --> 00:19:38,840 really, really big enterprise systems. 327 00:19:39,400 --> 00:19:44,400 And they all basically, you know, came back with the idea 328 00:19:44,400 --> 00:19:47,640 that, well, you're eventually going to have to hand roll and 329 00:19:47,640 --> 00:19:51,160 hand build something like Ash eventually yourself. 330 00:19:51,920 --> 00:19:55,240 And so I think in the forward of the Ash book, I make a joke 331 00:19:55,240 --> 00:19:58,840 about, I can't remember what Laura I called it, but it's a a 332 00:19:58,840 --> 00:20:03,560 riff on Greenspun's Law that any sufficiently old and advanced, 333 00:20:04,520 --> 00:20:08,720 you know, enterprise system contains, you know, a buggy, 334 00:20:08,720 --> 00:20:11,880 slow I'll fitting version of Ash Well. 335 00:20:13,440 --> 00:20:17,200 That's funny. And so, yeah, I mean, I think 336 00:20:17,200 --> 00:20:20,160 that's true. Like the bigger an app gets, the 337 00:20:20,160 --> 00:20:23,440 more ash matters. And I think that's where people 338 00:20:23,520 --> 00:20:26,080 struggle to realise its power because if you try it out on a 339 00:20:26,080 --> 00:20:27,440 toy app, it doesn't give you anything. 340 00:20:28,040 --> 00:20:31,160 And it's not until you have a really big domain and lots of 341 00:20:31,160 --> 00:20:34,880 things to generate from that. I feel like it gives you a lot 342 00:20:34,880 --> 00:20:37,920 with a toy app because it gets you up and running really 343 00:20:37,920 --> 00:20:39,960 quickly. I'm not saying it doesn't give 344 00:20:39,960 --> 00:20:41,640 you anything with a toy app. Yeah, OK. 345 00:20:41,840 --> 00:20:45,840 Like the value isn't there because people go like this is 346 00:20:45,840 --> 00:20:48,080 not doing that much more than what I could do, yeah. 347 00:20:48,160 --> 00:20:50,320 Fair, especially with. LLMS it's like. 348 00:20:52,520 --> 00:20:56,280 Yeah, I actually struggled. I liked the idea of Ash. 349 00:20:56,280 --> 00:21:01,040 I liked, you know, I liked Zach. I had Zach on my podcast before 350 00:21:01,040 --> 00:21:03,120 I started diving into Ash as well. 351 00:21:03,520 --> 00:21:06,520 And I struggled. I gave up. 352 00:21:06,520 --> 00:21:09,160 Like I ripped it out of a couple projects where I tried to bring 353 00:21:09,160 --> 00:21:12,680 it in and I was like, I'm just, I can move faster. 354 00:21:12,680 --> 00:21:15,560 Just, you know, use using Elixir in Phoenix. 355 00:21:15,560 --> 00:21:17,240 I'm just going to stick with what I'm comfortable with. 356 00:21:17,840 --> 00:21:20,280 And then when the book came out, I was like, all right, I'm going 357 00:21:20,280 --> 00:21:22,400 to force myself. I'm going to actually learn this 358 00:21:22,400 --> 00:21:24,480 thing. And now all my personal 359 00:21:24,480 --> 00:21:27,920 projects, like if I'm in control of them, I'm using hash on all 360 00:21:27,920 --> 00:21:30,960 of them now because you just like get up and running so 361 00:21:30,960 --> 00:21:33,440 quickly. Exactly. 362 00:21:33,440 --> 00:21:36,600 I mean, it's not just up and running and like Igniter changes 363 00:21:36,600 --> 00:21:39,920 the game on that. One, just like so much less 364 00:21:40,200 --> 00:21:43,480 template code that you're writing as well, like you just 365 00:21:43,480 --> 00:21:46,520 create your resource, your policies and like you don't have 366 00:21:46,520 --> 00:21:50,400 to think about your migrations or like anything like you so 367 00:21:50,400 --> 00:21:54,800 much comes out-of-the-box. I think having spent like a 368 00:21:54,800 --> 00:21:58,000 decade of writing migrations in Ruby and. 369 00:21:58,000 --> 00:22:02,520 Migrations and just suck. Like, honestly, it was the worst 370 00:22:02,520 --> 00:22:04,960 part of my day. And it's like, what are we, What 371 00:22:04,960 --> 00:22:07,960 are we doing here? And not having that, like 372 00:22:07,960 --> 00:22:12,520 people, you know, undervalue, like just how great not having 373 00:22:12,520 --> 00:22:15,440 to think about migrations is. I mean, you still have to read 374 00:22:15,440 --> 00:22:18,400 and review them, but unless they're destructive you don't 375 00:22:18,400 --> 00:22:19,800 have to look at them. Right. 376 00:22:20,040 --> 00:22:21,360 Just a projection of your domain. 377 00:22:22,120 --> 00:22:24,960 I have a horrible habit of always reading all the and maybe 378 00:22:24,960 --> 00:22:27,000 it's not a horrible habit. It's probably a a safe. 379 00:22:27,000 --> 00:22:27,720 Habit. It's a good habit. 380 00:22:28,120 --> 00:22:30,680 Yeah, I read. All your kids out there, read 381 00:22:30,680 --> 00:22:33,040 your migrations. Don't be like Uncle Josh. 382 00:22:33,560 --> 00:22:37,680 Yeah. And but so far I haven't done 383 00:22:37,680 --> 00:22:42,480 anything, you know, complicated like basic web apps, you know, 384 00:22:42,480 --> 00:22:44,680 they're just CRUD operations essentially, right? 385 00:22:44,920 --> 00:22:48,160 But I have not had to adjust anything like my indexes get 386 00:22:48,160 --> 00:22:50,800 created properly. I I have not had to edit 387 00:22:50,800 --> 00:22:52,440 anything. And that's nice. 388 00:22:52,560 --> 00:22:53,640 Right. Yeah. 389 00:22:53,640 --> 00:22:56,360 And so I've been doing that long enough that I think, you know, 390 00:22:56,400 --> 00:23:00,880 unless there's a problem later, which is very rare, like unless 391 00:23:00,880 --> 00:23:04,160 you have very specific needs, if you're just doing, you know, 392 00:23:04,160 --> 00:23:07,000 normal stuff, ash, ash has your back on that. 393 00:23:07,080 --> 00:23:09,000 Well, it's actually technically ash Postgres. 394 00:23:09,040 --> 00:23:13,360 And that's the other complexity about ash is like ash core is 395 00:23:13,360 --> 00:23:16,520 really, you know, like the starting point. 396 00:23:16,560 --> 00:23:20,120 And then ash Postgres means you can derive your data layer and 397 00:23:20,120 --> 00:23:23,120 there's all kinds of data layers popping up. 398 00:23:24,160 --> 00:23:27,120 And I think, you know, it may be controversial, but I think we're 399 00:23:27,120 --> 00:23:33,440 seeing that 2026 may be the end of like the road for relational 400 00:23:33,720 --> 00:23:38,600 data because it just doesn't work really in an agentic world 401 00:23:38,600 --> 00:23:43,160 where you need, you know, graph, sparse graph relationships, you 402 00:23:43,160 --> 00:23:47,120 need vector, you need time series and temporal 403 00:23:47,280 --> 00:23:50,120 understanding as well. Like what facts of the world 404 00:23:50,120 --> 00:23:55,800 were replaced by new facts. And yeah, the multi model DB is 405 00:23:55,800 --> 00:23:57,200 something we're looking into at the moment. 406 00:23:58,280 --> 00:24:01,760 Sounds complicated and now another thing I have to learn. 407 00:24:03,440 --> 00:24:08,400 Well, look, the relational DB has done well, I think it was 408 00:24:08,480 --> 00:24:12,800 sort of invented in the 60s and 70s, I believe, you know, and 409 00:24:12,800 --> 00:24:17,760 the relational model has has had a good run, but it just taps out 410 00:24:17,760 --> 00:24:24,000 at the the things that you need for a Genting memory and being 411 00:24:24,000 --> 00:24:26,800 able to query, you know, context about the world extremely 412 00:24:26,800 --> 00:24:30,720 quickly and thoroughly. And so anyone who's you know, 413 00:24:31,160 --> 00:24:35,040 I've been doing stuff with agents, it's like intelligence 414 00:24:35,040 --> 00:24:37,320 without context really isn't that useful. 415 00:24:38,120 --> 00:24:39,400 Right. Well, that's. 416 00:24:39,400 --> 00:24:40,720 Been the biggest problem. Right. 417 00:24:41,880 --> 00:24:43,640 Right. And the the more context you can 418 00:24:43,640 --> 00:24:47,120 give it, the better, which is, you know, open floor, maybe you 419 00:24:47,120 --> 00:24:51,640 know, effectively A weaponized virus in disguise, but you know, 420 00:24:51,640 --> 00:24:54,680 it's very, it's when you give it, when you give it all the 421 00:24:54,680 --> 00:24:57,200 context and all the tools, it can be extremely helpful. 422 00:24:58,080 --> 00:24:59,600 I have not played around with it. 423 00:24:59,600 --> 00:25:02,720 I have like kept my eye on it but I haven't used it. 424 00:25:02,840 --> 00:25:06,560 Maybe I need to? Look, it's definitely worth 425 00:25:06,720 --> 00:25:10,560 testing it out in a very, you know, sandbox. 426 00:25:10,600 --> 00:25:14,360 Don't give it any tools that yeah, yeah, I might regret and 427 00:25:14,360 --> 00:25:18,720 definitely do it in the sandbox. You know, the more sandboxes the 428 00:25:18,720 --> 00:25:21,840 better. Yeah, sandbox inside of 429 00:25:21,840 --> 00:25:23,240 sandboxes. But yeah, I haven't played with 430 00:25:23,320 --> 00:25:26,720 it either. But what I've done is largely 431 00:25:26,720 --> 00:25:30,040 just put a whole bunch of context in files on disk, which 432 00:25:30,040 --> 00:25:32,120 is largely A clawed native knowledge graph. 433 00:25:33,040 --> 00:25:38,120 And so I've been running sort of my own very Macgyvid open clause 434 00:25:38,120 --> 00:25:41,720 style thing just with clawed code with files on disk, and it 435 00:25:41,720 --> 00:25:43,400 works great. That's awesome. 436 00:25:44,280 --> 00:25:47,520 It definitely falls down though, because you know where it puts 437 00:25:47,520 --> 00:25:51,640 stuff it's using like file system search primitives like 438 00:25:51,640 --> 00:25:55,640 graph and you know, find and that kind of thing. 439 00:25:55,840 --> 00:26:00,480 And so that starts tapping out pretty quickly. 440 00:26:00,720 --> 00:26:05,160 And I think it's interesting because I've also been, I don't 441 00:26:05,160 --> 00:26:07,680 know if you've ever tried this, but just opening up Claude 442 00:26:07,680 --> 00:26:11,720 inside your doc Claude directory with all of the sessions because 443 00:26:11,720 --> 00:26:14,320 it saves everything. I have not tried that. 444 00:26:14,320 --> 00:26:18,400 That's a good idea though. Yeah, but only recently. 445 00:26:18,400 --> 00:26:20,240 And I don't know what tipped it off. 446 00:26:20,600 --> 00:26:25,560 And I think did this by default because everything, all of the 447 00:26:25,560 --> 00:26:28,520 sessions are stored as Jason L, which is like, you know, Jason 448 00:26:28,560 --> 00:26:31,880 strings his lines on in a, in a text file. 449 00:26:32,240 --> 00:26:35,880 And yeah, grip starts falling down in that sort of format. 450 00:26:36,640 --> 00:26:40,040 And so, yeah, what if you do that? 451 00:26:40,040 --> 00:26:43,720 You can not only introspect, you know, what your clawed sessions 452 00:26:43,720 --> 00:26:47,160 look like. And it took me a while to figure 453 00:26:47,160 --> 00:26:48,880 this out. And this is probably a pro tip 454 00:26:48,920 --> 00:26:51,920 for everyone. Clode does keep things for 30 455 00:26:51,920 --> 00:26:54,040 days, but it has a 30 day clean up window. 456 00:26:54,120 --> 00:26:56,680 So if you want to save sessions older than 30 days, there is a 457 00:26:56,680 --> 00:27:01,920 setting that you can use to go. And I think it's like clean up 458 00:27:03,040 --> 00:27:05,360 frequency or whatever and you can just say how many days it's 459 00:27:05,360 --> 00:27:06,880 taken. Set it to a really high number. 460 00:27:07,400 --> 00:27:12,200 So you keep all your sessions and sort of 1 easy way into this 461 00:27:12,200 --> 00:27:14,640 is you. There is now a command on the 462 00:27:14,640 --> 00:27:15,880 latest Clode. I don't know when they 463 00:27:15,880 --> 00:27:18,200 introduced it. I just found it the other day, 464 00:27:18,480 --> 00:27:25,000 slash insights and so maybe run it on yours if you've got a 465 00:27:25,000 --> 00:27:27,520 cloud open. If you type slash insights it 466 00:27:27,520 --> 00:27:31,840 will basically go in introspect all of your sessions and give 467 00:27:31,840 --> 00:27:33,400 you some. I can sign out right now so I 468 00:27:33,440 --> 00:27:36,320 don't forget. I mean, it says some interesting 469 00:27:36,320 --> 00:27:39,440 stuff. It's really, really kind of 470 00:27:39,440 --> 00:27:43,080 personal and it's kind of helpful because it, it actually 471 00:27:43,080 --> 00:27:45,880 goes and analysis like where did things go wrong? 472 00:27:46,240 --> 00:27:49,200 And it gives you prompts that you can put into your Claude MD 473 00:27:49,200 --> 00:27:50,720 like so you can just copy paste them. 474 00:27:50,760 --> 00:27:54,680 So it gives you a big HTML report and it gives you like 475 00:27:54,960 --> 00:27:57,360 ways to fix up where it went wrong. 476 00:27:58,760 --> 00:28:01,960 Someone had a friend of mine had posted on LinkedIn this morning 477 00:28:01,960 --> 00:28:06,160 that he was like literally just searching his Claude sessions 478 00:28:06,160 --> 00:28:08,760 for swear words. And that's obviously a place 479 00:28:08,760 --> 00:28:11,600 where you got frustrated and the botch is doing the wrong thing. 480 00:28:11,600 --> 00:28:15,120 And it's like we'll find all those times and figure out what 481 00:28:15,120 --> 00:28:16,600 went wrong. Don't do that thing. 482 00:28:17,560 --> 00:28:24,880 'Cause I have been like, are you an F an idiot like after you? 483 00:28:24,880 --> 00:28:28,200 So I was thinking that go in depth explaining what you want 484 00:28:28,200 --> 00:28:30,600 and it still does the wrong thing and you're like what are 485 00:28:30,600 --> 00:28:33,200 you doing? Right. 486 00:28:33,440 --> 00:28:38,360 And I've been doing, you know, all kinds of tasks in just my, 487 00:28:39,120 --> 00:28:43,040 you know, my folder of Claude, yeah, like state files. 488 00:28:43,040 --> 00:28:45,080 And Claude sort of constructed it and. 489 00:28:45,240 --> 00:28:48,440 Have you played around with beads at all for context stuff? 490 00:28:49,400 --> 00:28:51,320 Well beads is just like A to do list. 491 00:28:51,360 --> 00:28:54,000 I did try it out probably last year, but. 492 00:28:54,040 --> 00:28:56,520 OK, it kind of fell on the floor. 493 00:28:56,520 --> 00:28:59,040 It kind of fell on the floor. I think they started trying to 494 00:28:59,040 --> 00:29:01,760 get too fancy with it. And actually a buddy of mine 495 00:29:02,840 --> 00:29:07,200 rewrote beads and it's called beadwork and he's using like get 496 00:29:07,200 --> 00:29:10,280 branches and stuff now for keeping track of tasks and like 497 00:29:10,280 --> 00:29:12,680 epics, It's actually, it works really well. 498 00:29:13,720 --> 00:29:15,880 So Oh. Very cool. 499 00:29:16,440 --> 00:29:19,720 Yeah, I think. For context and like just like I 500 00:29:19,720 --> 00:29:23,760 always make like markdown plans and things, but it's kind of 501 00:29:23,760 --> 00:29:28,440 nice to keep like the plan split into issues and like a nice To 502 00:29:28,440 --> 00:29:29,680 Do List, right? But. 503 00:29:31,040 --> 00:29:34,240 Yeah, for sure. The challenge is when, you know, 504 00:29:34,280 --> 00:29:37,760 you may change direction and like things get invalidated. 505 00:29:38,360 --> 00:29:42,240 So, you know, but certainly there's lots of ways to do that 506 00:29:42,240 --> 00:29:45,720 and I think people are exploring different tools and. 507 00:29:46,120 --> 00:29:49,520 Yeah, there's a new cool ways to do everything I feel like daily 508 00:29:50,080 --> 00:29:54,120 I can't keep up. Well, here's what I've realized, 509 00:29:55,440 --> 00:29:59,040 and this may be sort of explain what's happening with the 510 00:29:59,640 --> 00:30:02,560 explosion of software is it's actually easier to write your 511 00:30:02,560 --> 00:30:05,280 own now than understand someone else's thing. 512 00:30:05,280 --> 00:30:08,000 And I don't know if you played with Gastown, which was Steve 513 00:30:08,000 --> 00:30:11,360 Yegi's sort of advancement beyond beads. 514 00:30:11,560 --> 00:30:13,520 And I think he uses beads in Gastown. 515 00:30:14,120 --> 00:30:17,320 But honestly, that was just, you know, someone's fever dream. 516 00:30:17,320 --> 00:30:20,120 And I'm like, I tried it. And I'm like, yeah, yeah, I'm 517 00:30:20,320 --> 00:30:24,520 the mayor of Gastown. And then I ended up falling back 518 00:30:24,520 --> 00:30:30,480 into like just raw clawed planning, like I was just in 519 00:30:30,480 --> 00:30:33,520 called plan mode. And raw plan mode is what a lot 520 00:30:33,520 --> 00:30:35,880 of people are discovering is actually pretty good. 521 00:30:35,880 --> 00:30:38,640 It keeps getting better. And so you don't actually need 522 00:30:38,640 --> 00:30:41,080 to live inside someone else's fever drink tap. 523 00:30:41,400 --> 00:30:45,600 You can create your own. And that's honestly better than, 524 00:30:45,760 --> 00:30:49,320 you know, writing your own tools, hyper personalizing them 525 00:30:49,320 --> 00:30:51,640 for yourselves. And software engineers are are 526 00:30:51,640 --> 00:30:55,920 the most capable of doing that. It's like wild too for me is 527 00:30:55,920 --> 00:31:02,400 like I've been just using the the app, the Claude app a lot 528 00:31:02,400 --> 00:31:06,200 and like Co work you can. I've been just writing my own 529 00:31:06,200 --> 00:31:12,080 like Claude MD's for directories and like I have like a candidate 530 00:31:13,760 --> 00:31:19,880 evaluation skill now and it will go through candidate projects 531 00:31:19,880 --> 00:31:22,720 and give me like PDF write ups. So I know if I want to go 532 00:31:22,720 --> 00:31:25,080 physically look at their code or just trash it. 533 00:31:25,360 --> 00:31:27,760 Like it will tell me if it's like good or not. 534 00:31:27,760 --> 00:31:30,400 And then I'm like, OK, now I'll manually review this code. 535 00:31:30,640 --> 00:31:32,520 It saves me so much time. It's amazing. 536 00:31:33,880 --> 00:31:37,480 And look, that's that's it. Like we're producing output at 537 00:31:37,480 --> 00:31:41,000 such a volume that actually human cognitive limits of reach 538 00:31:41,000 --> 00:31:45,480 very quickly and you can't read and reason about the world at 539 00:31:45,480 --> 00:31:47,320 that speed. And so you actually. 540 00:31:47,320 --> 00:31:49,400 Have to. You actually have to have the 541 00:31:49,440 --> 00:31:53,560 box like do it for you. Like as we see, like you just 542 00:31:53,560 --> 00:31:57,200 made a tweet about, you know, AGI being like right around the 543 00:31:57,200 --> 00:32:00,400 corner, right? And like, what do you see it 544 00:32:00,400 --> 00:32:06,080 doing to like our worlds? Like you and I both get paid to 545 00:32:06,080 --> 00:32:07,960 write code. Like you're a consultancy. 546 00:32:07,960 --> 00:32:13,240 So you're, you have employees and all of this and like, are 547 00:32:13,240 --> 00:32:16,400 you worried or do you think it's just going to kind of like shift 548 00:32:16,400 --> 00:32:19,920 how we do it and we're still going to be the technical 549 00:32:19,920 --> 00:32:22,040 architects that have to build things? 550 00:32:24,000 --> 00:32:29,280 Well, I like to think of it this way, you know, and maybe zooming 551 00:32:29,280 --> 00:32:32,040 out to like not just software developers, but humanity is 552 00:32:32,040 --> 00:32:35,840 about to get a massive uplift. Now we may be speed running the 553 00:32:35,840 --> 00:32:39,520 industrial revolution in sort of a tenth of the time, you know, 554 00:32:39,520 --> 00:32:44,520 like a year instead of 10 of that nature, so. 555 00:32:44,520 --> 00:32:48,400 Can we, can we keep up with our own advancements? 556 00:32:48,400 --> 00:32:51,200 Like that's kind of like what kind of scares me. 557 00:32:51,200 --> 00:32:52,800 I don't know. I don't know about you, but I'm 558 00:32:52,800 --> 00:32:55,280 reaching my own personal cognitive and. 559 00:32:55,280 --> 00:32:58,080 Physics. Yeah, I've hit that a while ago. 560 00:32:59,000 --> 00:33:03,000 And and so, right. And so that is a signal to say 561 00:33:03,160 --> 00:33:06,080 we are the bottleneck here. Like we're outgunned. 562 00:33:06,320 --> 00:33:10,680 Like for sure we've got intelligence and you know, it's 563 00:33:10,680 --> 00:33:13,560 non deterministic and so it's not always right. 564 00:33:14,000 --> 00:33:16,560 But have you met people? Have you met the? 565 00:33:17,160 --> 00:33:23,440 Internet nobody's right. And I think, you know, what 566 00:33:23,440 --> 00:33:26,880 prepared me most for a lot of things in this life was I had a 567 00:33:26,880 --> 00:33:31,440 brief period of playing poker online and back in like probably 568 00:33:31,880 --> 00:33:35,160 what year was it, 2003 when poker got really popular. 569 00:33:35,760 --> 00:33:39,120 And I learnt, I took, I was in between sort of jobs and I was 570 00:33:39,120 --> 00:33:41,280 like kind of disillusioned with IT. 571 00:33:41,280 --> 00:33:45,160 And I was at the end of my 20s going, is this really what I 572 00:33:45,160 --> 00:33:48,200 signed up for building software. And I, I didn't get it. 573 00:33:48,320 --> 00:33:51,720 Like I didn't get why we're building this, this garbage 574 00:33:51,720 --> 00:33:53,480 stuff. And it was overly complicated. 575 00:33:54,440 --> 00:33:57,880 And I thought it was me. Then I realised, oh, actually 576 00:33:57,880 --> 00:34:00,520 it's everyone else. Like why are you making this 577 00:34:00,520 --> 00:34:02,000 more complicated than it needs to be? 578 00:34:02,440 --> 00:34:05,720 And simplicity is hard. Like you know, it sounds easy, 579 00:34:05,720 --> 00:34:07,760 but it actually I always more more effort. 580 00:34:08,120 --> 00:34:11,960 I always found like when I write my own code, it's easier to 581 00:34:11,960 --> 00:34:14,960 follow right? And I also feel like my 582 00:34:14,960 --> 00:34:18,560 solutions are always more simple than other people's, but I don't 583 00:34:18,560 --> 00:34:19,440 know if that's 'cause I've. Really. 584 00:34:19,440 --> 00:34:21,480 So says you. Right, right. 585 00:34:21,800 --> 00:34:25,000 Like I always I have to talk to the guy that is working on my 586 00:34:25,000 --> 00:34:27,960 code years later when I'm no longer part of that company, you 587 00:34:27,960 --> 00:34:32,000 know, and see how they really. Well, exactly. 588 00:34:32,000 --> 00:34:35,600 And I think, you know, coding is a communication exercise and 589 00:34:35,600 --> 00:34:38,920 it's a thinking exercise. And so really a lot of the 590 00:34:38,920 --> 00:34:41,920 things that we're doing is communicating and thinking. 591 00:34:41,920 --> 00:34:45,760 And if you think it's about typing, you probably need to 592 00:34:46,040 --> 00:34:47,199 check. For now, it's. 593 00:34:47,280 --> 00:34:49,760 Really what programming and software development is all 594 00:34:49,760 --> 00:34:51,360 about. You're solving problems. 595 00:34:51,920 --> 00:34:54,719 You're doing like hard thinking about how to solve those 596 00:34:54,719 --> 00:34:57,920 problems and communicating the solutions. 597 00:34:58,960 --> 00:35:03,360 Well, even before we, even even before we had LLMS writing a lot 598 00:35:03,360 --> 00:35:08,120 of our code for us, you still spent 90% of your time thinking 599 00:35:08,120 --> 00:35:11,400 through solutions and they're like and then you write the 600 00:35:11,400 --> 00:35:15,240 code. But it's 10% is probably on the 601 00:35:15,240 --> 00:35:18,200 high side of where you spend your time writing code. 602 00:35:19,200 --> 00:35:23,280 Right. And look, I think just the the 603 00:35:23,320 --> 00:35:26,280 poker story was all about just playing probabilistic games. 604 00:35:27,040 --> 00:35:29,280 I've always loved backgammon and I like poker. 605 00:35:29,480 --> 00:35:31,280 I've never been that great at chess. 606 00:35:31,720 --> 00:35:35,200 I mean, chess has some probabilistic stuff to it, but 607 00:35:35,200 --> 00:35:38,640 if you know your opponent and you know what level they are and 608 00:35:38,640 --> 00:35:42,000 you're reasonably good, you know, chess is pretty 609 00:35:42,000 --> 00:35:44,040 deterministic. There's only a certain set of 610 00:35:44,040 --> 00:35:45,600 moves that your opponents like, right? 611 00:35:45,600 --> 00:35:47,600 Well, and they might come up with something good that you 612 00:35:47,600 --> 00:35:50,280 didn't think about, but you largely know within sort of. 613 00:35:50,560 --> 00:35:52,840 I feel like. The best what they're going to 614 00:35:52,840 --> 00:35:54,560 do. The best chess players are just 615 00:35:54,560 --> 00:35:57,560 really good at memorizing tactics, right? 616 00:35:57,560 --> 00:35:59,600 Like, there's only so many movements you can make on a 617 00:35:59,600 --> 00:36:02,000 board and when. Yeah. 618 00:36:02,320 --> 00:36:04,360 And so it's like memorizing what's been done. 619 00:36:05,320 --> 00:36:08,520 And that's why, like the speed chess is interesting because you 620 00:36:08,520 --> 00:36:11,160 can't deeply think about it. You're really just falling back 621 00:36:11,160 --> 00:36:13,160 to just pure speed pattern matching. 622 00:36:13,600 --> 00:36:16,640 But also, have you seen chess? I think it's called 960. 623 00:36:17,400 --> 00:36:22,880 And what they do is you pick a random distribution of how the 624 00:36:22,880 --> 00:36:27,920 pieces on the back are arranged. And I'm not a chess expert, but 625 00:36:27,920 --> 00:36:31,120 basically randomizes, so you can't memorize all the moves. 626 00:36:31,120 --> 00:36:34,280 So you're actually going back to sort of those principles and it 627 00:36:34,280 --> 00:36:37,480 sort of levels the playing field away from just pure memorization 628 00:36:37,680 --> 00:36:39,360 and memory. That's kind of cool. 629 00:36:39,360 --> 00:36:42,280 So that's a a more pure sort of thinking game. 630 00:36:42,680 --> 00:36:44,720 But yeah, I always like probabilistic games. 631 00:36:44,720 --> 00:36:49,560 And so thinking about how to constrain the probability fields 632 00:36:49,560 --> 00:36:52,520 or the range of outputs of an LM certainly suits me. 633 00:36:52,960 --> 00:36:55,040 And Poker's an interesting game because you don't even know what 634 00:36:55,040 --> 00:36:56,640 game you're playing and you're playing the maths. 635 00:36:57,120 --> 00:37:00,200 Are you playing like the the game theory optimal, like no one 636 00:37:00,200 --> 00:37:01,920 can exploit me? Or are you playing an 637 00:37:01,920 --> 00:37:04,600 exploitative game or exploitative defence? 638 00:37:05,120 --> 00:37:09,000 And every hand is is different and every session is completely 639 00:37:09,000 --> 00:37:11,800 different and you're changing up the levels of where you're 640 00:37:11,800 --> 00:37:13,960 actually playing. And, and it's a beautiful, 641 00:37:14,000 --> 00:37:17,000 beautiful game for that reason. It doesn't look beautiful on the 642 00:37:17,000 --> 00:37:19,040 surface, but it's like, are you playing the long game? 643 00:37:19,040 --> 00:37:22,800 Are you playing the short game? Are you trying to five in a 644 00:37:22,800 --> 00:37:26,280 tournament? Like it's a crazy, crazy game 645 00:37:26,280 --> 00:37:28,680 with lots of variants and you have to manage the probability 646 00:37:28,680 --> 00:37:31,200 and you know, you can have the situation where you can make the 647 00:37:31,200 --> 00:37:35,200 right decision and be and lose a lot of money, possibly all of 648 00:37:35,200 --> 00:37:37,520 it. And you know, you still would 649 00:37:37,520 --> 00:37:41,720 make have to make the same decision every time and learning 650 00:37:41,720 --> 00:37:45,720 from things where you're right but but you lost, and vice versa 651 00:37:45,760 --> 00:37:48,680 like you were wrong but you won. And you have to know that you 652 00:37:48,680 --> 00:37:51,400 were right and still lost, so you don't like change that 653 00:37:51,400 --> 00:37:54,760 behavior. Right, exactly. 654 00:37:55,600 --> 00:37:58,160 So that's a crazy game. And that look, that applies 655 00:37:58,160 --> 00:38:01,360 perfectly to business, and it also applies perfectly to, you 656 00:38:01,360 --> 00:38:06,760 know, managing the probabilistic output range of an LLM as well. 657 00:38:07,360 --> 00:38:11,440 So really our jobs as engineers working with LLMS is trying to 658 00:38:11,880 --> 00:38:15,720 narrow the range of outputs that it can come up with. 659 00:38:16,400 --> 00:38:19,720 And then we assemble, you know, committees and hierarchies of 660 00:38:19,720 --> 00:38:26,000 LLMS to make sure that, you know, humans less involved 661 00:38:26,000 --> 00:38:28,760 because we can't do human review on everything. 662 00:38:29,360 --> 00:38:34,600 And really what's maybe missing on most engineering teams is 663 00:38:34,600 --> 00:38:38,640 like the idea that humans have to review everything. 664 00:38:38,640 --> 00:38:42,840 That's like, well, do they, you know, because think of the, you 665 00:38:42,840 --> 00:38:44,880 know, the stakes. When the stakes are low, you 666 00:38:44,880 --> 00:38:47,440 should let AI do it. When the stakes are high, you 667 00:38:47,440 --> 00:38:50,560 should absolutely have humans spend the most time there and 668 00:38:50,680 --> 00:38:53,440 not all all things to review are equal. 669 00:38:53,720 --> 00:38:56,560 Yeah, that definitely terrifies me because like I, I'm 670 00:38:56,560 --> 00:39:00,640 definitely still stuck in the mindset that I don't want to 671 00:39:00,640 --> 00:39:04,280 allow the possibility of bad code going to my code base. 672 00:39:04,280 --> 00:39:07,840 So I'm the bottleneck reviewing everything, right? 673 00:39:08,040 --> 00:39:10,160 And I don't know how to get past that either. 674 00:39:10,160 --> 00:39:12,880 Like how did what? Like, what safeguards do you put 675 00:39:12,880 --> 00:39:17,000 in place to make sure the right code is being generated, the 676 00:39:17,000 --> 00:39:20,600 right kind of tests are being written without verifying it, 677 00:39:20,600 --> 00:39:24,280 you know? Well, and you know, I don't know 678 00:39:24,280 --> 00:39:26,240 which project you're talking about, but you're talking about, 679 00:39:26,360 --> 00:39:28,280 you know, your you've got a few side projects. 680 00:39:28,280 --> 00:39:31,440 Any project really. But yeah, like for instance, 681 00:39:31,440 --> 00:39:34,040 like Killswitch is a good example because, you know, I'm 682 00:39:34,040 --> 00:39:36,680 using encryption. People have to trust me with 683 00:39:37,160 --> 00:39:41,440 properly encrypting files and text and all of that good stuff 684 00:39:41,800 --> 00:39:44,880 and yeah, I can't not look at that. 685 00:39:45,680 --> 00:39:48,120 What is What is Killswitch for those who don't know? 686 00:39:49,360 --> 00:39:51,880 I think everyone that watches this podcast probably knows, but 687 00:39:51,880 --> 00:39:56,280 it it's, it's essentially like an encrypted document storage 688 00:39:56,520 --> 00:40:00,400 with deadman switches. So you can add beneficiaries to 689 00:40:00,560 --> 00:40:03,880 any of your files, notes, whatever it may be. 690 00:40:03,920 --> 00:40:07,840 And then if you stop checking in and you can set that window 691 00:40:07,840 --> 00:40:11,560 whatever you want it to be. Then it will send out to your 692 00:40:11,560 --> 00:40:15,560 beneficiaries and don't. Go offline. 693 00:40:16,080 --> 00:40:22,720 Don't go offline for too long, but it's been a lot of fun. 694 00:40:23,000 --> 00:40:27,720 So it's sort of like, you know, so yeah, well, like when the 695 00:40:27,720 --> 00:40:30,240 time when the the switch triggers, 'cause you haven't 696 00:40:30,240 --> 00:40:34,800 responded, like, you know, your special set of secret stuff is 697 00:40:34,800 --> 00:40:36,440 released. Yeah, that's. 698 00:40:36,480 --> 00:40:38,240 Be careful. And so forth. 699 00:40:38,480 --> 00:40:43,920 Well, so one thing that's cool, I basically I create like 700 00:40:44,080 --> 00:40:49,800 signing tokens for each share link that's created. 701 00:40:49,800 --> 00:40:53,400 So like every person that receives access to your 702 00:40:53,400 --> 00:40:55,960 document, you can revoke at your level. 703 00:40:55,960 --> 00:40:59,440 So like if things accidentally go out, you can revoke 704 00:40:59,440 --> 00:41:02,280 everything and then they lose access as long as they haven't 705 00:41:02,280 --> 00:41:03,520 already downloaded it. But. 706 00:41:04,440 --> 00:41:06,160 Right. Yeah, I don't think we've 707 00:41:06,400 --> 00:41:13,040 managed to create the, what's the word, the the encryption for 708 00:41:13,040 --> 00:41:15,200 like once his hit human eyeballs. 709 00:41:15,200 --> 00:41:16,720 Yeah, yeah, with. Basically. 710 00:41:16,720 --> 00:41:23,080 Unencrypted black flash just. Are you flashy theme me? 711 00:41:23,440 --> 00:41:27,480 Yeah. So all right, I'm going to just 712 00:41:27,480 --> 00:41:30,840 put this on hold for one second and play a quick add and then 713 00:41:31,000 --> 00:41:34,080 we'll be right back. Hey, I'm Kimberly with Elixir 714 00:41:34,080 --> 00:41:37,560 mentor Jacob recently added recruitment services to help 715 00:41:37,560 --> 00:41:39,760 companies hire pre vetted Elixir devs. 716 00:41:40,280 --> 00:41:43,520 I help with company outreach, so if you're hiring, I may be 717 00:41:43,520 --> 00:41:46,520 reaching out, or you can connect with us anytime at 718 00:41:46,520 --> 00:41:51,000 elixirmentor.com. Coding Elixir with AI or 719 00:41:51,000 --> 00:41:53,000 building AI into your Elixir apps? 720 00:41:53,240 --> 00:41:56,720 The Elixir AI Collective Discord is your community for both. 721 00:41:56,840 --> 00:42:02,320 Link in the description below and we're back now. 722 00:42:02,320 --> 00:42:03,560 I forgot what we were talking about. 723 00:42:03,600 --> 00:42:05,360 I always. Well, we were talking about kill 724 00:42:05,360 --> 00:42:09,720 switch and agentic engineering. Yeah, so like. 725 00:42:10,160 --> 00:42:12,600 Probabilistic stuff. Well, I have a question for you 726 00:42:12,600 --> 00:42:17,800 first, you're using, you're using Ash and Elixer and Live 727 00:42:17,800 --> 00:42:19,000 View I'm. Assuming Yep. 728 00:42:19,240 --> 00:42:21,840 Live view Elixer Ash. Yep. 729 00:42:24,080 --> 00:42:30,000 And how are you finding that? Oh, I mean I love it now I would 730 00:42:30,000 --> 00:42:34,760 say so Killswitch, I feel like I went into kind of already like 731 00:42:35,040 --> 00:42:39,400 get I I got it like I know how to use ash now, but like with 732 00:42:39,400 --> 00:42:42,520 projects prior to it, I was like, I know how to do this the 733 00:42:42,520 --> 00:42:45,520 normal way. How do I do this the ash way and 734 00:42:45,520 --> 00:42:48,720 that was like the hardest thing right 'cause then you're like 735 00:42:48,720 --> 00:42:50,920 OK, I can do this multi tenant app. 736 00:42:50,920 --> 00:42:55,880 I know how to do this with just migrations and my database, but 737 00:42:55,880 --> 00:42:59,040 how the F do I do this in Ash? And like so I would have moments 738 00:42:59,040 --> 00:43:00,880 like that where I would I just beat my head on the table and 739 00:43:00,880 --> 00:43:05,160 then it's as simple as adding a freaking one liner to to your 740 00:43:05,200 --> 00:43:09,000 resource, you know, and you have a multi tenant application 741 00:43:09,000 --> 00:43:12,480 essentially. And but so like once you know 742 00:43:12,480 --> 00:43:14,560 those tricks, Ash is amazing, right? 743 00:43:14,560 --> 00:43:19,400 Because now I can make things multi tenant or, you know, yeah, 744 00:43:19,520 --> 00:43:23,360 multi tenancy ability with one liners and config files 745 00:43:23,360 --> 00:43:27,120 essentially. Yeah. 746 00:43:27,240 --> 00:43:29,960 And the idea is that, you know, like in 30 years, I've built a 747 00:43:29,960 --> 00:43:33,400 lot of apps, I've seen a lot of apps being built and you see the 748 00:43:33,400 --> 00:43:34,960 same patterns over and over again. 749 00:43:35,640 --> 00:43:38,680 And so it was always funny that, you know, it was always the the 750 00:43:38,680 --> 00:43:43,840 Super, you know, battle scarred super senior engineers that got 751 00:43:43,840 --> 00:43:46,320 ash straight away because it was like, oh God, I don't have to do 752 00:43:46,320 --> 00:43:50,880 this thing again because it's like someone solved this 753 00:43:50,880 --> 00:43:53,240 problem, like for real this time. 754 00:43:54,600 --> 00:44:01,000 And, you know, it's funny that you know, as we, you know, tried 755 00:44:01,000 --> 00:44:05,920 to explain Ash to people, Zach and I sort of went to Elixir 756 00:44:05,920 --> 00:44:10,440 conf in the US and I think it was 22 and we were trying to 757 00:44:10,440 --> 00:44:13,520 explain it to people and it just didn't land at all. 758 00:44:13,920 --> 00:44:16,400 And it's, it's like the people that really got it were like, 759 00:44:16,480 --> 00:44:19,160 oh, you solve these problems so I don't have to do it again. 760 00:44:20,000 --> 00:44:24,200 And then the, you know, the people that are less experienced 761 00:44:24,200 --> 00:44:27,440 actually have an easier time going, Oh, like, I don't have to 762 00:44:27,920 --> 00:44:29,840 know how to do that. Like, it just does it for me. 763 00:44:29,960 --> 00:44:32,360 Oh, awesome. And so it ended up being like 764 00:44:32,360 --> 00:44:35,840 almost like a Midwick meme. So like the the people in the 765 00:44:35,840 --> 00:44:39,720 middle, like who are still cutting their teeth learning how 766 00:44:39,720 --> 00:44:42,960 to code and haven't made all the mistakes yet, actually want to 767 00:44:42,960 --> 00:44:46,960 write code and want to learn by making the mistakes themselves 768 00:44:47,240 --> 00:44:49,960 without having it spoon fed to them. 769 00:44:49,960 --> 00:44:55,200 And so actually in the middle of the sort of range of experience 770 00:44:55,200 --> 00:44:59,160 were the most resistant to ash ideas because it's interesting. 771 00:44:59,160 --> 00:45:02,480 Well I I just got my coding badge and I need to learn this 772 00:45:02,480 --> 00:45:05,520 stuff myself so. You feel like like the like mid 773 00:45:05,520 --> 00:45:08,480 level developers are the least likely to like Ash? 774 00:45:09,400 --> 00:45:12,520 And just hate it in our experience. 775 00:45:12,520 --> 00:45:15,240 Well, like, obviously that's a generalization, but we've talked 776 00:45:15,240 --> 00:45:18,360 to a lot of people and the most resistance comes from, you know, 777 00:45:18,360 --> 00:45:21,240 the middle. Well, I also feel like like 778 00:45:21,240 --> 00:45:26,040 Elixir is stuffed full of a lot of senior level people as well. 779 00:45:26,120 --> 00:45:30,200 Like, yeah, we have our beginners, but a lot of us come 780 00:45:30,200 --> 00:45:34,960 from the battles and headaches of other tech stacks. 781 00:45:34,960 --> 00:45:37,160 And then when we find Elixir, we're like, we're never leaving. 782 00:45:38,440 --> 00:45:40,040 So I think there's that too, right? 783 00:45:40,040 --> 00:45:42,560 I don't have any plans to leave. I don't want to. 784 00:45:43,720 --> 00:45:47,600 I made-up my other job, I made-up my own jobs just so I 785 00:45:47,600 --> 00:45:50,920 could write Elixir before I, you know, got my official first 786 00:45:50,920 --> 00:45:54,560 Elixir paying job, you know? Hey I started a company so I 787 00:45:54,560 --> 00:45:57,040 could do. This Yeah, I did the same thing, 788 00:45:57,040 --> 00:46:00,000 but I struggled. I tried to start a consultancy 789 00:46:00,000 --> 00:46:03,760 and it was hard. I just like, I suck at getting 790 00:46:03,760 --> 00:46:05,320 clients and it was a lot of hard work. 791 00:46:05,320 --> 00:46:08,280 And then I ended up just like getting a lot of website clients 792 00:46:08,280 --> 00:46:09,520 and I was like, well, this sucks. 793 00:46:09,520 --> 00:46:11,320 I don't want to just make websites all day. 794 00:46:12,480 --> 00:46:15,400 Yeah, I mean, look, running your own business is hard. 795 00:46:16,200 --> 00:46:19,520 You know the founder out there. You know found a life. 796 00:46:19,520 --> 00:46:22,320 If you know, you know. I'm stuck in it for some reason 797 00:46:22,320 --> 00:46:27,160 I can't stop torturing myself. Well, once it, it's a double 798 00:46:27,160 --> 00:46:30,880 edged sword, right? It's sort of like you get more 799 00:46:30,880 --> 00:46:37,880 freedom from the traditional employment relationship, but you 800 00:46:37,880 --> 00:46:43,640 know, at a cost of there's a lot of you actually get more 801 00:46:43,640 --> 00:46:48,040 freedom, but less because, you know, your boss just shifts to, 802 00:46:48,360 --> 00:46:51,200 you know, something else. I don't know for my own, for all 803 00:46:51,200 --> 00:46:53,920 my projects, my boss is a real asshole. 804 00:46:55,000 --> 00:46:57,360 Right. And sometimes you are your own 805 00:46:57,360 --> 00:46:58,720 worst boss. Exactly. 806 00:46:58,960 --> 00:47:01,880 Like I'm probably harder on myself than oh dude. 807 00:47:02,320 --> 00:47:05,640 I don't give myself any breaks. It's just like, OK, day job is 808 00:47:05,640 --> 00:47:08,160 over, now you're grinding until you can't stay awake. 809 00:47:09,600 --> 00:47:12,800 Right. So yes, the, the grind is real. 810 00:47:12,800 --> 00:47:16,520 And look, nine years of, of running a consultancy, yeah, 811 00:47:16,520 --> 00:47:19,960 it's been a challenge. And look, it's been, I mean, I, 812 00:47:19,960 --> 00:47:24,840 I know a ton of the Elixir agencies and certainly it's been 813 00:47:24,840 --> 00:47:27,080 a while since I caught up with like a lot of the other 814 00:47:27,120 --> 00:47:30,640 consultants, more consultant consultancy and sort of studios 815 00:47:30,640 --> 00:47:32,440 and agencies around Sydney and Australia. 816 00:47:32,920 --> 00:47:35,440 But I know a bunch of them and not everyone made it through the 817 00:47:35,440 --> 00:47:38,840 last sort of 3-4 years. I feel like the last four years 818 00:47:38,840 --> 00:47:41,720 have been rough. Brutal. 819 00:47:41,720 --> 00:47:45,520 I think brutal. So, you know, like even just 820 00:47:45,520 --> 00:47:48,920 surviving through that is definitely success. 821 00:47:49,560 --> 00:47:54,640 And, you know, because not everyone made it and certainly, 822 00:47:55,240 --> 00:48:00,200 you know, wallets and purses snapshot about yeah, after post 823 00:48:00,200 --> 00:48:03,760 COVID and like, you know, there was free money for a bit and 824 00:48:03,760 --> 00:48:07,600 then suddenly there wasn't. So the 0 the zerp or the 0 825 00:48:07,600 --> 00:48:10,360 interest rate meant that basically borrowing money was 826 00:48:10,480 --> 00:48:13,560 effectively free. And so there was a little party 827 00:48:13,560 --> 00:48:16,360 for a while and there was lots of money and lots of, you know, 828 00:48:16,400 --> 00:48:18,840 craziness going on. And then reality bites. 829 00:48:18,840 --> 00:48:20,680 And you know versus a snapshot the. 830 00:48:20,680 --> 00:48:23,680 Fake growth bubble or whatever you want to call it. 831 00:48:25,000 --> 00:48:29,480 Because I feel like even like like SAS companies or tech 832 00:48:29,480 --> 00:48:32,680 companies in general, not just consultancies, they all just 833 00:48:32,680 --> 00:48:36,680 bloated their payroll and then they mass laid off and we're 834 00:48:36,680 --> 00:48:41,440 still seeing that now. Yeah, I mean, the the news 835 00:48:41,440 --> 00:48:43,760 recently was blocked doing exactly that. 836 00:48:44,080 --> 00:48:47,480 Yeah. And you know, funnily enough, I 837 00:48:47,480 --> 00:48:53,960 do know the head of engineering of Block and he came out, he 838 00:48:53,960 --> 00:48:58,320 lives in overseas and came out for Christmas one time. 839 00:48:58,320 --> 00:49:00,440 And he's talking about just how many engineers they had. 840 00:49:00,440 --> 00:49:03,920 And I'm like, I can't even reason about how many engineers 841 00:49:03,920 --> 00:49:05,120 you have, buddy. Like, what do you do with them 842 00:49:05,120 --> 00:49:10,720 all, you know? And so I guess I got my answer a 843 00:49:10,720 --> 00:49:14,000 week ago. It's like actually maybe you had 844 00:49:14,000 --> 00:49:18,040 a few too many. And so that wasn't, I think, 845 00:49:18,040 --> 00:49:22,720 just an AI story, although that's a nice take for, you 846 00:49:22,720 --> 00:49:25,520 know, the investor crowd. And it's a it's an easy story to 847 00:49:25,520 --> 00:49:27,640 swallow, but I don't think that was the whole story. 848 00:49:28,080 --> 00:49:32,680 Yeah, No, I agree. I think well, I think so like 849 00:49:32,680 --> 00:49:36,800 Collects is my day job really awesome company, right. 850 00:49:36,800 --> 00:49:42,280 And they're very bold like our CEO and CTO are very bold on 851 00:49:42,280 --> 00:49:46,400 like AI moving forward. But like no one at our company 852 00:49:46,400 --> 00:49:51,440 talks as AI is like a replacement, more just like like 853 00:49:51,440 --> 00:49:54,960 an accelerator for all of us. Like it adds super powers to all 854 00:49:54,960 --> 00:49:58,240 of us. And so we always talk about how 855 00:49:58,240 --> 00:50:01,040 we use it and what our workflows look like. 856 00:50:01,040 --> 00:50:04,480 And we have like weekly calls about it, just so like if 857 00:50:04,480 --> 00:50:06,920 someone starts doing something differently, but it works really 858 00:50:06,920 --> 00:50:12,520 well, we all talk about it. Yeah, yeah, we're doing the 859 00:50:12,520 --> 00:50:14,520 same. And you know, look, it's hard to 860 00:50:14,520 --> 00:50:17,280 keep up. There's 4 Claude releases of 861 00:50:17,280 --> 00:50:20,480 Claude code per day. It's like it's just keeping up 862 00:50:20,480 --> 00:50:25,320 with that is difficult. You know, I was describing it to 863 00:50:25,800 --> 00:50:32,640 my business partner and wife, you know, because she isn't like 864 00:50:32,640 --> 00:50:34,840 she's like, oh, why aren't we just sharing all this stuff? 865 00:50:34,880 --> 00:50:38,440 And I'm like, it's like having Christmas every hour and you're 866 00:50:38,560 --> 00:50:40,760 opening all your presents and you don't have time to get to 867 00:50:40,760 --> 00:50:42,120 them all. And certainly the ones you've 868 00:50:42,120 --> 00:50:44,040 opened, you don't even have time to play with. 869 00:50:44,480 --> 00:50:47,680 And so there's this giant pile of toys behind you and you're 870 00:50:47,680 --> 00:50:50,920 like, this Christmas sucks. Like, I don't even play with any 871 00:50:50,920 --> 00:50:53,480 of my toys. It's just all the dopamine of 872 00:50:53,480 --> 00:50:55,760 opening the presents and going, oh, maybe this is good. 873 00:50:55,880 --> 00:50:59,040 I don't know, I can't even keep up on like just reading the 874 00:50:59,040 --> 00:51:02,120 change logs and I'm trying to be a better person about that. 875 00:51:03,720 --> 00:51:06,600 Right. Oh, but you know, like, I think 876 00:51:06,800 --> 00:51:10,120 you can probably agree that we're able to do much more now. 877 00:51:11,000 --> 00:51:13,920 And we don't know where the limits are either. 878 00:51:14,200 --> 00:51:17,680 It's like, what are the limits? We don't know. 879 00:51:18,160 --> 00:51:20,160 And people are busy finding it out. 880 00:51:20,400 --> 00:51:27,080 And I think, you know, I think who said it, I think it was Karl 881 00:51:27,080 --> 00:51:32,520 Pathey may have said, like I've never felt so behind because you 882 00:51:32,520 --> 00:51:34,960 can sort of see where the possibilities are and you're not 883 00:51:34,960 --> 00:51:36,720 there. And it's almost like a a 884 00:51:36,720 --> 00:51:39,280 restatement of Dunning Kruger where the people that know the 885 00:51:39,280 --> 00:51:42,720 most actually think they're the dumbest, know the least because 886 00:51:42,720 --> 00:51:45,640 of they they're aware of how much there is to know and how 887 00:51:45,640 --> 00:51:49,920 little they actually know in relative amounts to to the 888 00:51:50,200 --> 00:51:52,240 total. The Dunning Kruger curve is 889 00:51:52,440 --> 00:51:52,800 real. It's. 890 00:51:54,080 --> 00:51:56,120 Totally real. And so the people that are most 891 00:51:56,800 --> 00:51:59,960 knowledgeable actually think they are the least knowledgeable 892 00:51:59,960 --> 00:52:02,080 because they know what's possible. 893 00:52:02,120 --> 00:52:05,200 And except for Mike Hostetler, I can't keep up with that guy. 894 00:52:05,320 --> 00:52:07,760 Can you keep up with him? No. 895 00:52:08,360 --> 00:52:11,640 I think he's running agents in his own brain and he's just 896 00:52:11,800 --> 00:52:13,280 always working. I don't even know. 897 00:52:14,000 --> 00:52:16,800 I I keep chatting to Mike and I'm like, Oh my God, dude. 898 00:52:16,840 --> 00:52:18,600 Like I don't know how you do it all. 899 00:52:19,160 --> 00:52:21,400 I don't either. Trying to figure it out myself. 900 00:52:23,520 --> 00:52:25,280 Some level of LLMS is the answer. 901 00:52:25,280 --> 00:52:29,920 But yeah, it's like, you know, I always thought Jose Velim was 902 00:52:29,920 --> 00:52:32,960 like 10 developers in a in a long trench coat. 903 00:52:33,640 --> 00:52:36,680 Like how does he keep doing what he's doing for as long as he's 904 00:52:36,680 --> 00:52:40,120 done it, dude. For him, Zach Daniel is another 905 00:52:40,120 --> 00:52:43,480 one, like I don't know how he does everything and like how 906 00:52:43,480 --> 00:52:46,520 responsive he is too with like Ash questions. 907 00:52:46,520 --> 00:52:50,080 Like dude, I don't even want to respond to people in my elixir 908 00:52:50,080 --> 00:52:51,960 discard. I mean, I love that everyone's 909 00:52:51,960 --> 00:52:54,800 in there and I like chatting, but it's like when there's like 910 00:52:54,800 --> 00:52:57,480 a question, it's like, oh, that's going to take me like 20 911 00:52:57,480 --> 00:53:00,440 minutes to actually thoughtfully answer and. 912 00:53:00,440 --> 00:53:05,320 I was like no like an SLA of milliseconds. 913 00:53:05,800 --> 00:53:08,280 Yeah. And actually, you know, Zach's 914 00:53:08,720 --> 00:53:13,000 now head of engineering, I think his his title and he's sort of 915 00:53:13,120 --> 00:53:15,640 running a pretty massive team at Remedy Meds. 916 00:53:16,320 --> 00:53:20,040 And yeah, like I think that has tied him up. 917 00:53:20,760 --> 00:53:22,680 And so he's. Slowing down a little, right, 918 00:53:22,920 --> 00:53:25,640 You know. Yeah, we just had to put some 919 00:53:25,760 --> 00:53:28,960 some anchors on Zach so that he could create some space so more 920 00:53:28,960 --> 00:53:31,480 contributors could come in because no one could beat Zach 921 00:53:31,480 --> 00:53:34,920 to answering questions or. Or even writing features. 922 00:53:37,120 --> 00:53:40,000 And so actually slowing Zach down was the best thing to 923 00:53:40,000 --> 00:53:42,600 happen to the Ash ecosystem, ironically. 924 00:53:42,600 --> 00:53:44,840 Like, you wouldn't think that would be true, but it's like, 925 00:53:45,560 --> 00:53:49,760 you know, I I definitely, you know, begged Zach to like, just 926 00:53:49,760 --> 00:53:52,600 create some room for other people to step in and he 927 00:53:52,600 --> 00:53:54,680 couldn't. And so actually like. 928 00:53:54,680 --> 00:53:58,080 I get it, when something is your baby, it's hard to it's your. 929 00:53:58,200 --> 00:54:00,160 Baby right? And you know, yeah. 930 00:54:00,760 --> 00:54:04,840 And he was like his I guess iterative OODA loop. 931 00:54:06,000 --> 00:54:11,400 If anyone knows that, it's the observe, Orient, decide and act 932 00:54:11,720 --> 00:54:14,240 his OODA loops, which is so fast no one could compete. 933 00:54:14,920 --> 00:54:19,080 And basically, yeah, there was an interesting and recent chat 934 00:54:19,080 --> 00:54:25,400 where he described the the OODA loop and it's kind of a 935 00:54:25,400 --> 00:54:29,720 generally applicable thing and it's worth calling out the 936 00:54:29,720 --> 00:54:32,280 story. But it was created or sort of 937 00:54:32,400 --> 00:54:36,640 coined by a fighter pilot. And basically this fighter pilot 938 00:54:36,640 --> 00:54:41,880 was just so fast at orienting, observing, orienting, deciding 939 00:54:41,880 --> 00:54:45,480 and acting that he won every dogfight. 940 00:54:45,840 --> 00:54:49,920 And so it was about having a tighter OODA loop than your 941 00:54:49,920 --> 00:54:55,080 competitors or adversaries in a, in a sort of aerial dogfight. 942 00:54:55,960 --> 00:55:01,120 Because if you can do the Orient, observe, Orient, decide 943 00:55:01,120 --> 00:55:04,320 and act faster, you've changed the state of the world. 944 00:55:04,680 --> 00:55:08,520 And if your opponent's OODA loop is slower, they have to start 945 00:55:08,520 --> 00:55:11,280 again. And so if they've only got to 946 00:55:11,280 --> 00:55:14,760 deciding and they haven't acted yet, if you've acted and they 947 00:55:14,760 --> 00:55:17,400 see that, then they have to start and observe and, or, and 948 00:55:17,800 --> 00:55:22,120 you actually create, if you're tight enough of a loop, you 949 00:55:22,120 --> 00:55:24,440 create a psychological break because you can never do 950 00:55:24,440 --> 00:55:27,360 anything because you have to keep starting again. 951 00:55:29,840 --> 00:55:33,160 And, and that was like a completely wild idea. 952 00:55:33,520 --> 00:55:37,920 And I, I think, you know, definitely applicable to a lot 953 00:55:37,920 --> 00:55:43,080 of scenarios, but I mean, the OODA loop is also the scientific 954 00:55:43,080 --> 00:55:46,200 method, right? It's sort of keep trying some 955 00:55:46,200 --> 00:55:50,240 things and seeing what happens and depending on the response 956 00:55:50,240 --> 00:55:53,520 from the world, learn some things and do it again. 957 00:55:54,240 --> 00:55:59,000 And you know, what is FAFO? Can we swear on this show? 958 00:55:59,520 --> 00:56:01,320 Yeah, I mean, it's YouTube I think. 959 00:56:01,520 --> 00:56:04,240 You know, it's fuck around and find out is the scientific 960 00:56:04,240 --> 00:56:06,800 method. It's like do some things, learn, 961 00:56:08,240 --> 00:56:09,880 rinse and repeat. That's all it is. 962 00:56:10,160 --> 00:56:13,600 And you know, it's interesting that we're seeing that, you 963 00:56:13,600 --> 00:56:20,160 know, you program LMS to do that and you get sort of evolutionary 964 00:56:20,160 --> 00:56:24,840 intelligence. And, you know, I was kind of 965 00:56:24,840 --> 00:56:29,080 joking about AGI getting closer because Jeff Huntley is 966 00:56:29,120 --> 00:56:32,120 rediscovering ash and elixir like he it's. 967 00:56:32,160 --> 00:56:37,280 Cool to see people talking about it at like, you know, the access 968 00:56:37,920 --> 00:56:41,560 of you know, the popularity of these people bringing elixir 969 00:56:41,560 --> 00:56:43,920 back into the game. I also. 970 00:56:43,920 --> 00:56:46,200 Well, I think. Someone like this isn't going to 971 00:56:46,200 --> 00:56:49,120 surprise you, but the Rusk community wants to rebuild 972 00:56:49,120 --> 00:56:52,640 Erlang and Rust. Like, well, I mean, I think 973 00:56:52,640 --> 00:56:55,720 we've tried that a few times ourselves, the Elixir community 974 00:56:55,720 --> 00:57:00,440 and look, I'm here for it. I mean, it's a difficult, it's a 975 00:57:00,440 --> 00:57:04,240 difficult thing to do because there's more to beam than meets 976 00:57:04,240 --> 00:57:06,040 the eye. And I think there's a pretty 977 00:57:06,040 --> 00:57:09,440 interesting Twitter, I'm not going to call it the new thing 978 00:57:09,880 --> 00:57:12,760 Twitter conversation recently about exactly that. 979 00:57:12,960 --> 00:57:15,040 And I think, you know, it's sort of like, oh, you don't 980 00:57:15,040 --> 00:57:18,200 understand the beam. You know, language wars. 981 00:57:18,200 --> 00:57:21,080 We don't have them very often, but you know, it helps to be 982 00:57:22,240 --> 00:57:25,920 unfailing and welcoming. I. 983 00:57:25,920 --> 00:57:28,400 Sound like a broken record on my podcast at least. 984 00:57:28,400 --> 00:57:32,480 Because what's wild to me is like Beam and Erlang solved all 985 00:57:32,480 --> 00:57:36,240 of our problems today back in the mid 80s. 986 00:57:37,360 --> 00:57:39,760 I'm. Having the same problems, it's 987 00:57:39,760 --> 00:57:43,560 just a different different mask on them or whatever you want to 988 00:57:43,560 --> 00:57:46,680 call it. Well, what I've been seeing the 989 00:57:46,680 --> 00:57:51,800 last sort of month or so is people basically putting an LLM 990 00:57:51,800 --> 00:57:56,080 inside their Kubernetes cluster and then going, oh, we can spot 991 00:57:56,080 --> 00:58:00,280 the problems and like fix them and like restart the thing. 992 00:58:00,280 --> 00:58:03,320 And I'm like, oh, like self healing systems. 993 00:58:04,120 --> 00:58:07,560 You know, Erlang's had that since 1986, man, 40 years of 994 00:58:07,560 --> 00:58:10,240 self healing systems. I want to get AT shirt actually 995 00:58:10,320 --> 00:58:15,400 self healing since 1986. And so what's interesting about 996 00:58:15,400 --> 00:58:17,520 that is part of that is self healing. 997 00:58:18,560 --> 00:58:21,360 And you know, obviously with supervision trees and limiting 998 00:58:22,080 --> 00:58:26,640 the blast radius, you know, you can fix the problem, limit its 999 00:58:27,080 --> 00:58:31,040 blast radius and not affect other things and restart it in a 1000 00:58:31,040 --> 00:58:35,840 known good state. So you know that is a a brutally 1001 00:58:35,840 --> 00:58:39,640 effective and simple strategy. It seems complicated, but it's 1002 00:58:39,640 --> 00:58:44,320 like no, just start, have a tree of things and then let something 1003 00:58:44,640 --> 00:58:47,240 fails. Crash and start it back up. 1004 00:58:47,960 --> 00:58:49,680 Start it back up in a known good state. 1005 00:58:49,760 --> 00:58:53,200 I mean, how do you fix most technical problems that it's, 1006 00:58:53,360 --> 00:58:55,440 yeah, you know, have? You tried unplugging it and 1007 00:58:55,440 --> 00:58:56,680 plugging it back in? Have you tried? 1008 00:58:57,080 --> 00:58:58,400 Turning it on and turning it off again. 1009 00:58:58,400 --> 00:59:03,160 Turning it off and turning it on again, that's basically the beam 1010 00:59:03,160 --> 00:59:06,160 in a nutshell. In terms of how you need to 1011 00:59:06,160 --> 00:59:09,320 understand supervision trees. It's like, you know, and the 1012 00:59:09,320 --> 00:59:13,720 bigger the tree and the smaller the pieces, the less, you know, 1013 00:59:13,840 --> 00:59:16,040 impact. Any failure at any part of the 1014 00:59:16,040 --> 00:59:18,000 system can affect any other part of the system. 1015 00:59:19,160 --> 00:59:24,080 And that's such a brutally simple idea, and I don't think 1016 00:59:24,080 --> 00:59:27,160 people appreciate the elegance and simplicity of that as much 1017 00:59:27,160 --> 00:59:29,320 as they should. And people think it's the actor 1018 00:59:29,320 --> 00:59:31,400 model, which, you know, like, is it the actor model? 1019 00:59:32,360 --> 00:59:37,360 Certainly BEAM was invented and Erlang sort of VM was invented 1020 00:59:38,440 --> 00:59:40,880 before the actor model existed. And they were like, Oh yeah, it 1021 00:59:40,880 --> 00:59:42,200 just happens to be the same thing. 1022 00:59:43,040 --> 00:59:45,880 But you know, what are Gen. servers other than OO? 1023 00:59:46,360 --> 00:59:49,080 You've got this container of state where you encapsulate 1024 00:59:49,160 --> 00:59:52,120 state and you talk to it through a message passing, which is what 1025 00:59:52,120 --> 00:59:55,440 Alan Kay actually meant by OO. He was like, Oh yeah that's what 1026 00:59:55,440 --> 00:59:57,520 I meant. Like not objects and an 1027 00:59:57,520 --> 00:59:59,760 inheritance. Dude, I wish. 1028 01:00:00,360 --> 01:00:02,680 Honestly I wish it would have saved me a lot of time. 1029 01:00:02,680 --> 01:00:05,680 My learning curve when I learned how to program with object 1030 01:00:05,680 --> 01:00:08,760 oriented programming was so difficult I could not wrap my 1031 01:00:08,760 --> 01:00:11,280 head around. It wasn't just you mate. 1032 01:00:11,680 --> 01:00:15,560 Yeah, I didn't understand. Like inheritance sort of made 1033 01:00:15,560 --> 01:00:18,760 this logical sense, but it didn't make any sense at all. 1034 01:00:19,360 --> 01:00:23,600 Like I'm like this seems simple but it's why is it so wrong? 1035 01:00:24,080 --> 01:00:26,040 And then like. Probably. 1036 01:00:26,400 --> 01:00:29,080 It probably took me over a decade to figure that out. 1037 01:00:29,640 --> 01:00:34,000 Well, and it, it, it has so many flaws, which is why we have so 1038 01:00:34,000 --> 01:00:38,880 many patterns to follow, right? Because if you don't follow 1039 01:00:38,880 --> 01:00:42,400 specific architectural patterns, you're going to have a hell of a 1040 01:00:42,400 --> 01:00:45,120 time debugging or having a stable project. 1041 01:00:45,800 --> 01:00:49,480 And then like all of these like patterns in Elixir and Erlang, 1042 01:00:50,280 --> 01:00:53,240 you just have to use the language and there's just a way 1043 01:00:53,240 --> 01:00:55,720 to use it. And you don't really follow a 1044 01:00:55,720 --> 01:00:58,200 specific patterns, you're just using patterns because that's 1045 01:00:58,200 --> 01:01:02,880 just how things work. And here's The thing is that you 1046 01:01:02,880 --> 01:01:07,080 know, you and I probably took a decade long plus detour into OO 1047 01:01:07,080 --> 01:01:11,600 and never felt like we got it and I probably beat myself up 1048 01:01:11,600 --> 01:01:13,640 going. You must be really dumb if you 1049 01:01:13,640 --> 01:01:17,200 can't understand dude. You know what I the most out of 1050 01:01:18,120 --> 01:01:20,280 OO world is? Singletons. 1051 01:01:21,560 --> 01:01:25,560 And just like shared state and the fact that you can fuck 1052 01:01:25,560 --> 01:01:29,680 around with an object in multiple places and have race 1053 01:01:29,680 --> 01:01:33,200 conditions that are almost impossible to track down and 1054 01:01:33,200 --> 01:01:36,440 just alter data without knowing where it's being altered from, 1055 01:01:36,440 --> 01:01:39,440 that drives me insane. It's funny that you mentioned 1056 01:01:39,440 --> 01:01:45,520 singletons, but you know, yes, scars were had because I did 1057 01:01:45,520 --> 01:01:50,320 like Java and Ejbs and like, you know, Spring came along as 1058 01:01:50,320 --> 01:01:53,400 dependency injection, which wasn't the best dependency 1059 01:01:53,400 --> 01:01:56,720 injection solution by any stretch because it was all XML 1060 01:01:56,720 --> 01:02:00,040 and like. And I'm like, what I realized 1061 01:02:00,120 --> 01:02:05,520 actually a long time later is that, yeah, when I joined the 1062 01:02:05,520 --> 01:02:09,440 mob in, in London hacking away, it was like the early Footworks 1063 01:02:09,440 --> 01:02:12,560 office and they were building their own dependency injection 1064 01:02:12,560 --> 01:02:15,600 framework. And what I realised a lot later 1065 01:02:15,600 --> 01:02:19,200 is that I'd actually been doing exactly the same thing in my 1066 01:02:19,200 --> 01:02:24,440 Java files because I'm like, oh, well, I've got 2 implementations 1067 01:02:24,440 --> 01:02:26,640 of this interface. And I'm like, you know, they're 1068 01:02:26,640 --> 01:02:30,200 sort of structural strategies. I'm like, I want to boot up my 1069 01:02:30,200 --> 01:02:35,920 app with one or the other, like depending on what I need when I 1070 01:02:35,920 --> 01:02:38,040 run it. And I'm like, well, how can you 1071 01:02:38,040 --> 01:02:42,440 do that? And what I realized was, oh, 1072 01:02:42,440 --> 01:02:45,800 well, you just need to be able to dynamically set from outside 1073 01:02:45,800 --> 01:02:48,880 what the implementation is. And you could pass that as a 1074 01:02:48,880 --> 01:02:52,920 parameter to your program. But what I decided to use was if 1075 01:02:52,920 --> 01:02:56,240 you remember Java properties files that basically were just 1076 01:02:56,560 --> 01:02:58,520 kind of like tunnel. Sadly, I do. 1077 01:02:58,520 --> 01:03:01,640 Remember it was just a key equals of that value. 1078 01:03:02,920 --> 01:03:04,720 Kind of like an environment file or whatever. 1079 01:03:04,720 --> 01:03:08,480 It was just key equals value. Dude, I remember. 1080 01:03:08,480 --> 01:03:12,880 And So what I did was I had like the interface full qualified 1081 01:03:12,880 --> 01:03:15,480 package name equals the implementation that I wanted. 1082 01:03:15,920 --> 01:03:19,880 And So what I didn't realise until way, way later is like 1083 01:03:19,920 --> 01:03:23,120 independently had invented dependency injection, but you 1084 01:03:23,120 --> 01:03:26,160 can do impendency dependency injection and everyone goes, oh, 1085 01:03:26,200 --> 01:03:28,160 what is that? It took me years to figure out 1086 01:03:28,160 --> 01:03:32,760 it was just pass some arguments to your functions and then this 1087 01:03:32,760 --> 01:03:35,800 is how I teach functional programming now is like, well 1088 01:03:36,320 --> 01:03:40,000 any high school kid, middle schools I think learn F of X. 1089 01:03:40,000 --> 01:03:42,800 When do you learn F of X in America you know? 1090 01:03:43,400 --> 01:03:46,280 That's like algebra. So right. 1091 01:03:46,800 --> 01:03:49,440 I mean, I taught my daughter that like F of. 1092 01:03:49,440 --> 01:03:52,320 XI want to say like freshman, sophomore, year of high school 1093 01:03:53,320 --> 01:03:54,520 probably. How old are you at that point? 1094 01:03:55,760 --> 01:04:00,520 You would be like 14. Right. 1095 01:04:00,720 --> 01:04:03,960 So you knew the secret of functional programming at age 1096 01:04:04,080 --> 01:04:07,800 14, and you were certainly capable of understanding it way 1097 01:04:07,800 --> 01:04:12,040 earlier than that. And F of X is it like you know, 1098 01:04:12,080 --> 01:04:15,720 it's a pure function, it takes arguments you don't suddenly 1099 01:04:15,720 --> 01:04:19,720 grab like other state or arguments from anywhere. 1100 01:04:19,720 --> 01:04:24,760 But X can change over here too. Right. 1101 01:04:24,760 --> 01:04:28,480 I mean I mean imagine maths how hard it would be if if suddenly 1102 01:04:28,480 --> 01:04:30,880 like stuff could change your F of X. 1103 01:04:31,200 --> 01:04:33,480 Yeah. Like or it started sending 1104 01:04:33,480 --> 01:04:38,680 emails or printing stuff out. Getting confusing, but yeah, 1105 01:04:38,680 --> 01:04:41,800 I've always said that. I've always said if I started 1106 01:04:41,800 --> 01:04:43,840 with a functional, like immutable language where you're 1107 01:04:43,840 --> 01:04:46,840 just transforming data and you know exactly where it's being 1108 01:04:46,840 --> 01:04:49,680 changed, it would have been life changing for me to learn. 1109 01:04:49,720 --> 01:04:51,160 I would have learned so much quicker. 1110 01:04:52,440 --> 01:04:56,400 And I think, you know, as functional programmers, you 1111 01:04:56,400 --> 01:05:00,240 know, it's probably, you know, on us to teach people just how 1112 01:05:00,240 --> 01:05:02,640 simple it is. And, you know, I've done this 1113 01:05:02,880 --> 01:05:05,880 because people have forgotten they got taught or like 1114 01:05:06,480 --> 01:05:08,920 overcomplicated stuff. And if you strip all the 1115 01:05:08,920 --> 01:05:13,360 nonsense away, it's just F of X. And that is the most brutally 1116 01:05:13,360 --> 01:05:16,600 simple and effective way to teach functional programming to 1117 01:05:16,720 --> 01:05:20,920 anyone, even children. And so I think actually, you 1118 01:05:20,920 --> 01:05:26,680 know, in this age, perhaps the people that have yet to learn 1119 01:05:26,680 --> 01:05:31,480 where the limits are and be sort of taught wrong things probably 1120 01:05:31,480 --> 01:05:35,200 have more scope to go further because they can skip to the end 1121 01:05:35,480 --> 01:05:37,520 without this detour that we've had. 1122 01:05:38,040 --> 01:05:41,680 As you know, we did like a decade wasted a decade in the OO 1123 01:05:41,680 --> 01:05:44,040 mines I. Mean yeah, I. 1124 01:05:44,400 --> 01:05:47,960 Mean wasted a decade not thinking F of X was the answer 1125 01:05:48,000 --> 01:05:48,800 just. But I also. 1126 01:05:48,920 --> 01:05:51,920 Think it functions. The problem with developers and 1127 01:05:51,920 --> 01:05:54,960 software engineers is we're like very hard headed and you never 1128 01:05:54,960 --> 01:05:59,440 want to be wrong. So shifting your mind like like 1129 01:05:59,680 --> 01:06:01,640 I ignore Elixir for like 2 years. 1130 01:06:01,640 --> 01:06:04,080 I had a buddy telling me I like, you got to check this out, 1131 01:06:04,080 --> 01:06:05,880 you're going to love it. It's like, dude, I don't have 1132 01:06:05,880 --> 01:06:09,120 time for that. I'm busy in node JS bro. 1133 01:06:10,960 --> 01:06:13,080 When your friends say that, you should probably listen. 1134 01:06:13,200 --> 01:06:15,680 They care. About them, well, I, I know, I 1135 01:06:15,680 --> 01:06:19,080 know and I say it all the time because then like I started my 1136 01:06:19,080 --> 01:06:21,520 own one of my like a bootstrapped business. 1137 01:06:21,520 --> 01:06:24,320 It was a marketplace and I was like, all right, let me look 1138 01:06:24,320 --> 01:06:28,320 into this elixir thing and like on paper I was like, OK, this is 1139 01:06:28,920 --> 01:06:31,080 this is what I need. This is perfect. 1140 01:06:31,440 --> 01:06:34,680 And like on paper, I understood the shiny things and then I 1141 01:06:34,680 --> 01:06:39,120 started learning it and I definitely didn't understand 1142 01:06:39,120 --> 01:06:44,000 supervision trees and Genservers and like OTP and all this stuff. 1143 01:06:44,000 --> 01:06:46,440 When I first started, I just like started kind of writing a 1144 01:06:46,440 --> 01:06:52,240 REST API like you would in JavaScript and I like relearned 1145 01:06:52,240 --> 01:06:56,000 and rewrote things over and over again and just learned the hard 1146 01:06:56,000 --> 01:06:58,520 way. And but I did it right away on a 1147 01:06:58,520 --> 01:07:01,000 project and I'm like really happy I did that. 1148 01:07:01,560 --> 01:07:03,880 But then I was like, dude, that was hard. 1149 01:07:04,440 --> 01:07:06,720 And that's actually what inspired me starting my YouTube 1150 01:07:06,720 --> 01:07:08,360 channel. I was like, I'm going to teach 1151 01:07:08,360 --> 01:07:12,240 people how to get like off the ground and running quicker than 1152 01:07:12,240 --> 01:07:15,720 I did because I just want people to build their projects. 1153 01:07:16,760 --> 01:07:23,080 You know, like as an older person and with more experience 1154 01:07:23,080 --> 01:07:25,840 and some learnings under the belt, you can definitely pave 1155 01:07:25,840 --> 01:07:28,920 the path behind you. So it's easier for sure. 1156 01:07:29,320 --> 01:07:33,080 With saying that, I need to get back in motion with my ash 1157 01:07:33,080 --> 01:07:35,320 course because I've just kind of ignored that. 1158 01:07:35,320 --> 01:07:36,560 For Oh yeah, what happened with that? 1159 01:07:36,880 --> 01:07:39,160 I just like, I saw the first lot and it was amazing. 1160 01:07:39,160 --> 01:07:41,080 I was like, this is very ambitious, Jacob. 1161 01:07:41,440 --> 01:07:43,040 I know. Well, it's really good. 1162 01:07:43,040 --> 01:07:48,160 It's a fun project. I just like a lot in life caught 1163 01:07:48,160 --> 01:07:51,560 up to me and I kind of just like burned out on making tutorials 1164 01:07:51,560 --> 01:07:54,320 and now I feel like things are slowing back down. 1165 01:07:54,320 --> 01:07:58,040 So I just have to like, get back in the motions of like recording 1166 01:07:58,040 --> 01:08:02,120 a video, like maybe just once a week and just get going. 1167 01:08:02,920 --> 01:08:05,280 Yeah, good, good habits. You know little. 1168 01:08:05,320 --> 01:08:08,120 Steps, it's easy to lose them. That's the thing. 1169 01:08:08,120 --> 01:08:12,440 And then I feel guilty, like it's an extra guilt because all 1170 01:08:12,440 --> 01:08:14,680 these people are like, when are the next videos coming out? 1171 01:08:14,680 --> 01:08:18,040 I'm so excited. And then like you feel obligated 1172 01:08:18,040 --> 01:08:20,840 to these people for your free content. 1173 01:08:20,840 --> 01:08:23,720 But like, I love helping people so I want to do it. 1174 01:08:25,600 --> 01:08:28,479 Right. Look, and it's the stuff that, 1175 01:08:28,680 --> 01:08:32,399 you know, you do for the love of it and because you want to and 1176 01:08:32,439 --> 01:08:35,640 because you care. And then, you know, life has a 1177 01:08:35,800 --> 01:08:38,359 funny way of paying that back sometimes. 1178 01:08:39,160 --> 01:08:42,359 And really it's about, you know, movement and emotion, just doing 1179 01:08:42,359 --> 01:08:43,880 stuff. I mean, it's I. 1180 01:08:43,880 --> 01:08:46,960 Wouldn't have. I wouldn't have any of my jobs 1181 01:08:47,000 --> 01:08:50,560 right now without starting my YouTube channel, you know, like, 1182 01:08:50,720 --> 01:08:54,479 yeah, that's kind of cool. And like this podcast, I get to 1183 01:08:54,479 --> 01:08:57,680 meet all the coolest people in the Elixir community because 1184 01:08:57,680 --> 01:08:59,319 they come on my podcast and talk to me. 1185 01:08:59,680 --> 01:09:03,479 Right. I mean, yeah, I think you've 1186 01:09:03,479 --> 01:09:07,479 collected a very interesting and amazing bunch of people. 1187 01:09:07,560 --> 01:09:10,319 I've been kind of jealous of your your guest list. 1188 01:09:10,319 --> 01:09:13,160 I'm delighted to be included on it. 1189 01:09:13,880 --> 01:09:16,760 I mean, you're one of them. You're one of the awesome people 1190 01:09:16,800 --> 01:09:20,520 on my guest list, so it's cool. I'm just surprised all these 1191 01:09:20,520 --> 01:09:24,760 people want to talk to me. You. 1192 01:09:24,800 --> 01:09:27,080 Know well, I think, you know, you're the lightning rod mate. 1193 01:09:27,319 --> 01:09:31,680 It's like, you know, someone has to do the organizing and, you 1194 01:09:31,760 --> 01:09:34,520 know, bring people together and that's a really valuable skill 1195 01:09:34,520 --> 01:09:38,080 because I think, you know, a language community is more than 1196 01:09:38,080 --> 01:09:41,840 the technical stuff. You know, it's about how do you 1197 01:09:42,240 --> 01:09:47,720 sustain and grow that community and, and keep it as great as it 1198 01:09:47,720 --> 01:09:49,040 is? Because I think, you know, like 1199 01:09:49,040 --> 01:09:51,640 the licks are maybe an amazing language and the beam may have 1200 01:09:51,640 --> 01:09:57,080 been self healing since 1986. But, and, and have some insanely 1201 01:09:57,240 --> 01:10:00,760 great ideas that, you know, maybe the words are complicated 1202 01:10:00,760 --> 01:10:04,000 and a bit scary, but the ideas are like elegantly simple. 1203 01:10:05,640 --> 01:10:09,680 And so, you know, like don't be put off by the, the crazy words. 1204 01:10:10,120 --> 01:10:13,840 But you know, like I've organized community stuff for 1205 01:10:14,280 --> 01:10:17,320 probably nearly a decade. I've sort of lapsed because I 1206 01:10:17,320 --> 01:10:24,080 did a lot of travel last year on organizing Elixir Sydney and 1207 01:10:24,560 --> 01:10:28,280 yeah, recently shut off mymeet-up.com subscription, 1208 01:10:28,280 --> 01:10:29,440 which was costing a lot of money. 1209 01:10:29,440 --> 01:10:31,280 You know, I still had meet ups that we weren't it. 1210 01:10:31,280 --> 01:10:34,840 Kind of sucks that it's so expensive to use meet up now. 1211 01:10:35,440 --> 01:10:36,640 I remember when. It was so. 1212 01:10:36,800 --> 01:10:39,960 Expensive. And it's so expensive. 1213 01:10:40,000 --> 01:10:42,520 And then they do that, like, blackmail thing. 1214 01:10:42,520 --> 01:10:45,760 So if anyone's listening on the Elixir City, it's like, still a 1215 01:10:45,760 --> 01:10:48,000 thing. We'll do some more things, I 1216 01:10:48,000 --> 01:10:52,640 promise. And yeah, like, they're going to 1217 01:10:52,640 --> 01:10:57,000 do the blackmail. Hey, your, your organizers step 1218 01:10:57,000 --> 01:10:58,320 down. Who wants to step up? 1219 01:10:58,320 --> 01:11:02,520 And it's like auctioning off the Elixir City thing. 1220 01:11:02,640 --> 01:11:05,520 Yeah, it's, you know, it would be a shame if something would 1221 01:11:05,520 --> 01:11:08,440 happen to your meet up and then it's sort of like extortion. 1222 01:11:09,280 --> 01:11:12,800 And so I'm like, no, I'm done. And look, I think, you know, we 1223 01:11:12,800 --> 01:11:16,600 should all, I haven't managed to achieve this yet, but if you're 1224 01:11:16,600 --> 01:11:19,240 listening and you run a meet up, you should probably move off 1225 01:11:19,240 --> 01:11:24,400 meetup.com and go to Luma because that is way, way better. 1226 01:11:24,840 --> 01:11:27,920 It's free unless you charge money, which typically meetups 1227 01:11:27,920 --> 01:11:31,200 do not charge money. And so it's free and the 1228 01:11:31,200 --> 01:11:34,800 experience is really good and you can have, it's a slightly 1229 01:11:34,800 --> 01:11:37,080 different thing. So it's not really designed, 1230 01:11:37,080 --> 01:11:40,560 it's designed for events rather than meetups, but perfectly 1231 01:11:40,560 --> 01:11:43,520 suits running meetups. But you can have a calendars 1232 01:11:43,520 --> 01:11:46,160 that contain other calendars. So you could have an Elixir 1233 01:11:46,560 --> 01:11:49,600 global calendar, you could have an Elixir Australia calendar or 1234 01:11:49,720 --> 01:11:52,640 sort of different hierarchies. And there's no sort of limits 1235 01:11:52,640 --> 01:11:56,680 about how you can constrain the containment of of different 1236 01:11:57,240 --> 01:12:00,760 meetups, for example. But the language does confuse 1237 01:12:00,760 --> 01:12:02,640 when you first see it. But yeah, we're going to be 1238 01:12:02,640 --> 01:12:05,640 using Luma from here on out, and I recommend that other people 1239 01:12:05,640 --> 01:12:07,920 do. Do you have meetups where you 1240 01:12:07,920 --> 01:12:14,440 are we? Do I'm horrible at get into 1241 01:12:14,440 --> 01:12:16,360 them. There's an elixir meet up that 1242 01:12:16,360 --> 01:12:19,680 tries to meet like once a month, like just South of Salt Lake 1243 01:12:19,680 --> 01:12:26,360 City, and I kind of slack it going, but when I do make it, 1244 01:12:26,360 --> 01:12:28,600 it's fun. Like it, you know, you see all 1245 01:12:28,600 --> 01:12:31,880 these guys that are all like working elixir developers for 1246 01:12:31,880 --> 01:12:35,800 the most part. And you know, I recognize I'm 1247 01:12:35,800 --> 01:12:38,920 horrible with names and faces. Oh, I'll remember faces horrible 1248 01:12:38,920 --> 01:12:41,160 with names, but it's. Cool. 1249 01:12:42,480 --> 01:12:45,400 It's cool when you like start seeing the same people at 1250 01:12:45,400 --> 01:12:49,200 everything. And I also like kind of enjoy 1251 01:12:49,920 --> 01:12:53,920 because like I do this, this is my spotlight and but when I'm in 1252 01:12:53,920 --> 01:12:56,800 person, I can like kind of take a back seat and just enjoy 1253 01:12:57,200 --> 01:12:59,440 speakers and things and it's kind of nice. 1254 01:13:01,320 --> 01:13:04,680 Yeah, for sure. And it like it's its own reward. 1255 01:13:04,840 --> 01:13:08,160 I mean, I used to sort of treat it as a monthly excuse to get 1256 01:13:08,160 --> 01:13:11,800 out of the office and just socialize with a a regular crew 1257 01:13:11,960 --> 01:13:16,360 and talk talk tech. I like when I'm like in it, but 1258 01:13:16,800 --> 01:13:19,600 getting there is the hard part. Like I'm like, do I really want 1259 01:13:19,600 --> 01:13:23,840 to drive and go socialize? Well, exactly. 1260 01:13:23,840 --> 01:13:27,840 And I think the the challenge is, you know, post COVID, 1261 01:13:27,840 --> 01:13:29,640 meetups were never quite the same. 1262 01:13:29,920 --> 01:13:34,960 And, you know, we kept running online stuff and then, you know, 1263 01:13:34,960 --> 01:13:37,360 geography doesn't really matter at that point. 1264 01:13:37,360 --> 01:13:43,200 It's just time zone. And so, you know, I think, yeah. 1265 01:13:43,200 --> 01:13:44,920 And meetups were never quite the same. 1266 01:13:44,920 --> 01:13:47,560 And certainly when they came back after COVID that like we 1267 01:13:47,560 --> 01:13:50,600 had some really big meetups in Sydney. 1268 01:13:51,560 --> 01:13:55,880 And then I think just everyone's not working in the city anymore. 1269 01:13:56,080 --> 01:13:59,240 Like, yeah, it used to be huge because everyone would be in the 1270 01:13:59,240 --> 01:14:00,880 city. It'd be like, oh, like after 1271 01:14:00,880 --> 01:14:03,560 work I'll go have a have a drink and. 1272 01:14:03,680 --> 01:14:05,600 I feel like that's what it was, right? 1273 01:14:05,600 --> 01:14:09,080 Like it was your commute home, so you didn't drive very far. 1274 01:14:09,800 --> 01:14:12,520 It like, you know, if you're driving 20 minutes to work but 1275 01:14:12,520 --> 01:14:16,080 your meet up's 5 minutes from the office, it's a lot easier 1276 01:14:16,080 --> 01:14:19,080 to, you know, make that excuse to swing by on your way home. 1277 01:14:19,800 --> 01:14:23,600 Yeah, exactly. And so, you know, like in 1278 01:14:23,600 --> 01:14:25,880 Australia, if you work in the city, you're probably not 1279 01:14:25,880 --> 01:14:28,760 driving your car and so you're getting public transport. 1280 01:14:28,760 --> 01:14:32,800 And so it was, you know, even easier if you can go from your 1281 01:14:32,800 --> 01:14:35,200 office, walk to the central meet up. 1282 01:14:35,200 --> 01:14:36,920 And, you know, we had a lot more people there. 1283 01:14:36,920 --> 01:14:39,280 And I think most people are sitting at home watching 1284 01:14:39,280 --> 01:14:42,200 YouTube, maybe this. And you know, that's why when 1285 01:14:42,600 --> 01:14:44,400 meet ups aren't quite the same, people can. 1286 01:14:44,400 --> 01:14:47,400 Chat with us live like you know, we got. 1287 01:14:47,520 --> 01:14:50,360 Let's see. Who's saying what? 1288 01:14:50,360 --> 01:14:52,400 I can't see any of the chat unfortunately. 1289 01:14:52,440 --> 01:14:55,040 If you went if you go over to the YouTube live stream, you can 1290 01:14:55,040 --> 01:14:58,520 see it, but Steve said he just picked up the ash book a lot of 1291 01:14:58,520 --> 01:15:02,040 fun so far and I was I think it's a great it's a great book 1292 01:15:02,600 --> 01:15:06,600 and Jura. Steve could probably find my my 1293 01:15:06,600 --> 01:15:09,880 quote about the the half. Oh yeah, there you go, the ash. 1294 01:15:09,960 --> 01:15:12,200 Thing it's in the Ford somewhere. 1295 01:15:12,720 --> 01:15:13,720 And then? What I call? 1296 01:15:14,680 --> 01:15:17,240 It Jura, sorry if I'm butchering your names guys. 1297 01:15:17,240 --> 01:15:21,400 He just said he loves Ash and then Victoria goes, dude I love 1298 01:15:21,400 --> 01:15:23,480 your channel. I'm studying to make my career 1299 01:15:23,480 --> 01:15:27,800 change from JS full stack to Elixir and he's from Peru. 1300 01:15:28,040 --> 01:15:30,360 It's a it's a good move dude, you won't regret it. 1301 01:15:30,600 --> 01:15:31,800 That was pretty much. You really. 1302 01:15:31,840 --> 01:15:34,920 You really weren't. I was JavaScript back end. 1303 01:15:34,920 --> 01:15:39,840 I always did back end stuff and I did Swift and a lot of iOS 1304 01:15:39,840 --> 01:15:42,720 development and I don't want to touch any of that stuff anymore. 1305 01:15:44,240 --> 01:15:45,920 Yeah, yeah. I mean. 1306 01:15:46,520 --> 01:15:49,680 Yeah. So you mentioned Micah Stetler, 1307 01:15:49,760 --> 01:15:54,160 who's a legend and he has just released Judo 2, which is super 1308 01:15:54,160 --> 01:16:00,040 exciting, but we we catch up fairly regularly and he, he told 1309 01:16:00,040 --> 01:16:03,760 me that he was actually part of the jQuery core team way back 1310 01:16:03,760 --> 01:16:05,360 when. Oh yeah, he was. 1311 01:16:05,480 --> 01:16:06,960 Isn't that crazy? Yeah. 1312 01:16:07,040 --> 01:16:09,920 I'm like, oh, what I love jQuery. 1313 01:16:09,960 --> 01:16:13,160 I like J when I discovered that dollar thing, I'm like, Oh my 1314 01:16:13,160 --> 01:16:15,400 God. And so for anyone who's sort of 1315 01:16:15,400 --> 01:16:19,520 new to the web, like jQuery was the giant Band-Aid that you put 1316 01:16:19,520 --> 01:16:24,440 across browsers so that any of your stuff would actually work. 1317 01:16:24,840 --> 01:16:28,680 And so if, if you're now used to document, you know, selectors 1318 01:16:28,680 --> 01:16:32,240 where you, everything's on document and you can find things 1319 01:16:32,240 --> 01:16:37,560 in a Dom by CSS selectors or, you know, do stuff that used to 1320 01:16:37,560 --> 01:16:40,600 not be true. And jQuery was the the solution. 1321 01:16:40,960 --> 01:16:44,560 And it's still around like jquery's used massively on, you 1322 01:16:44,560 --> 01:16:47,840 know, if you want to do vanilla JS with no framework, jQuery is 1323 01:16:47,840 --> 01:16:51,720 the thing that, yeah, wild. I, I cut my teeth on and you 1324 01:16:51,720 --> 01:16:54,400 know, it would get you very far and you'd be like reacting to 1325 01:16:54,400 --> 01:16:57,720 individual events. And then you'd end up going, oh, 1326 01:16:57,720 --> 01:17:00,200 I've just created the world's greatest spaghetti. 1327 01:17:00,960 --> 01:17:03,840 And then you really needed it because you'd be like, oh, how 1328 01:17:03,840 --> 01:17:07,760 do these things interact? And you couldn't reason about it 1329 01:17:07,800 --> 01:17:10,200 at a certain scale. It was great until it wasn't. 1330 01:17:10,600 --> 01:17:13,440 Yeah, yeah, most things are great. 1331 01:17:13,440 --> 01:17:17,320 Until anything that's good like gets hoovered into like. 1332 01:17:17,360 --> 01:17:20,680 A lot of what made jQuery great got hoovered into like actually 1333 01:17:20,680 --> 01:17:24,040 the the web APIs which. Is. 1334 01:17:24,040 --> 01:17:26,720 Yeah, I forget. Sometimes you serve as the the 1335 01:17:26,760 --> 01:17:28,560 Band-Aid until the thing gets fixed. 1336 01:17:29,560 --> 01:17:33,680 Yeah, Mike became my friend when I first started doing like Vibe 1337 01:17:33,680 --> 01:17:38,160 Coding live streams on Saturday, like when when LLMS just started 1338 01:17:38,160 --> 01:17:42,120 getting good enough to like kind of build small projects, I had 1339 01:17:42,120 --> 01:17:45,760 like a couple Vibe Coding live streams with guests on where I 1340 01:17:45,800 --> 01:17:47,760 made them teach me the tooling. It was great. 1341 01:17:48,160 --> 01:17:52,440 And then Mike reached out and I had them on and yeah, that was 1342 01:17:52,440 --> 01:17:54,600 the beginning of our friendship and it's been fun. 1343 01:17:55,240 --> 01:17:57,040 And now I can't. I can't keep up with the man. 1344 01:17:57,800 --> 01:18:00,160 Yeah, Mike, Mike's a legend. I haven't chatted to him in a 1345 01:18:00,160 --> 01:18:03,480 few weeks and. I want to try to get to Elixir 1346 01:18:03,480 --> 01:18:07,000 Conf in Chicago because he's in the Chicago area and it'd be a 1347 01:18:07,000 --> 01:18:10,800 lot of fun, I think. Yeah, like I'm looking forward 1348 01:18:10,800 --> 01:18:14,280 to that. I'm going to be going to Malaga 1349 01:18:14,360 --> 01:18:20,640 for Elixir Conf EU and I'll be swinging past San Fran and 1350 01:18:22,120 --> 01:18:26,480 Chicago along the way just because, yeah, I'm on this 1351 01:18:26,480 --> 01:18:28,280 constantly rotating round the world ticket. 1352 01:18:28,400 --> 01:18:29,960 Yeah, I guess so. I feel like you're always 1353 01:18:29,960 --> 01:18:32,880 travelling. Well, it's the cheapest way to 1354 01:18:32,880 --> 01:18:36,120 do it. It's by a long way. 1355 01:18:36,480 --> 01:18:38,240 Once you're doing, I just keep going. 1356 01:18:39,720 --> 01:18:41,680 Yeah, I'm in the middle of my trip right now. 1357 01:18:42,000 --> 01:18:45,240 I started in Oslo in September after Goatmeyer. 1358 01:18:45,880 --> 01:18:48,480 Nice. Made it home and I'm on a stop. 1359 01:18:48,520 --> 01:18:54,320 A very long stopover, but yeah, I caught up with Mike last time 1360 01:18:54,560 --> 01:18:58,040 as well and he took me out to his favorite steak restaurant 1361 01:18:58,040 --> 01:19:00,800 and that was nice. He is a scholar and a gentleman. 1362 01:19:00,800 --> 01:19:02,880 And yeah, you should definitely check out Jido. 1363 01:19:02,880 --> 01:19:07,680 It is, there's a lot there, but the way that it's been put 1364 01:19:07,680 --> 01:19:10,200 together is pretty amazing. So you've got first class 1365 01:19:10,320 --> 01:19:14,840 agentic programming, you've you've got just an insane amount 1366 01:19:14,840 --> 01:19:18,160 of tools in the tool belt. Possibly too many. 1367 01:19:18,320 --> 01:19:23,280 So many is overwhelming and certainly I I've only dabbled 1368 01:19:23,840 --> 01:19:26,600 but some of the team have been diving into it. 1369 01:19:26,600 --> 01:19:29,480 And I got to get Mike back on. He's supposed to be making 1370 01:19:29,480 --> 01:19:33,320 content for my channel for Jeto and he's been slacking. 1371 01:19:33,320 --> 01:19:35,040 Yeah, Mike, but. I can't throw him any stones. 1372 01:19:35,040 --> 01:19:37,880 Because I'm slacking. Well, I don't know whether 1373 01:19:37,880 --> 01:19:39,960 Mike's been slacking. Maybe on this particular talk. 1374 01:19:40,200 --> 01:19:43,480 He's not slacking on development, he's just slacking 1375 01:19:43,480 --> 01:19:46,240 on creating content for me. There you go. 1376 01:19:47,760 --> 01:19:49,200 Well, there you go. You heard it first, Mike. 1377 01:19:49,360 --> 01:19:52,320 It's it's time to come back. Yeah, Mike, stop slacking. 1378 01:19:53,240 --> 01:19:57,080 But yeah, join us Discord too, if you're interested in any like 1379 01:19:58,000 --> 01:20:01,120 agenic programming like the Geeto Discord's really awesome. 1380 01:20:01,680 --> 01:20:03,680 A lot of smart guys in there. I can't keep up. 1381 01:20:03,720 --> 01:20:07,040 I am so bad at keeping up in Discord communities but. 1382 01:20:07,440 --> 01:20:10,880 I can't keep up anymore, so I've reached my limit. 1383 01:20:11,040 --> 01:20:13,640 I need a box. Ash Ash is a crazy discord 1384 01:20:13,640 --> 01:20:15,840 community too. I can't keep up with that one 1385 01:20:15,880 --> 01:20:19,040 either. Yeah, there is a lot happening 1386 01:20:19,960 --> 01:20:24,400 and you know like there's a Geeto, I think it's Ash Geeto. 1387 01:20:24,480 --> 01:20:27,160 It's like first class connections to creating a 1388 01:20:27,200 --> 01:20:28,520 genetic programming on top of that. 1389 01:20:28,520 --> 01:20:30,160 I did see that. That's really cool. 1390 01:20:31,480 --> 01:20:33,920 So if you've got Ash, like there's a connector to Geeto 1391 01:20:33,920 --> 01:20:38,520 world for for that and. That's I need to bring Mike back 1392 01:20:38,520 --> 01:20:40,560 on and make him just teach me stuff again. 1393 01:20:40,600 --> 01:20:43,560 That's the only way I have time. I just make a live stream and I 1394 01:20:43,560 --> 01:20:47,160 box it out. Right This is why I did meet ups 1395 01:20:47,160 --> 01:20:50,160 dude it's like oh I want to learn stuff on a regular basis 1396 01:20:50,200 --> 01:20:53,560 I'll just set up the meet up so and if I couldn't find speakers 1397 01:20:53,600 --> 01:20:55,480 I would have to learn something and tell everyone. 1398 01:20:55,520 --> 01:20:56,480 There you go. Well. 1399 01:20:56,520 --> 01:20:57,880 Like it's really. Funny. 1400 01:20:58,520 --> 01:21:02,000 People always public speaking. People always be like, how do 1401 01:21:02,000 --> 01:21:05,200 you have time for your podcast? It's like, it's great 'cause I 1402 01:21:05,200 --> 01:21:07,840 just bring smart people on that are doing cool things and they 1403 01:21:07,840 --> 01:21:11,400 tell me about it. It's like my free education pass 1404 01:21:11,400 --> 01:21:16,320 and I have to pay attention 'cause I'm talking to them so. 1405 01:21:17,560 --> 01:21:19,600 The secret's out. Yeah, that's why I did meetups. 1406 01:21:19,680 --> 01:21:23,120 It's like I wanted to learn stuff and I wanted to catch up 1407 01:21:23,120 --> 01:21:25,640 with cool people every month, and so I just did that. 1408 01:21:26,040 --> 01:21:28,920 Yeah, my podcast is purely selfish, dude. 1409 01:21:28,920 --> 01:21:34,400 Like it's for me but. I mean, I I did meetups for 1410 01:21:34,400 --> 01:21:35,960 everyone, but it's mainly for me, yeah. 1411 01:21:37,120 --> 01:21:41,920 Yeah, I got it. You know, I I've got to find the 1412 01:21:41,920 --> 01:21:45,120 time to bring that back, but. Yeah, yeah. 1413 01:21:45,280 --> 01:21:46,920 I don't know. I don't know if the the age of 1414 01:21:46,920 --> 01:21:49,280 meet ups is really over it and I think maybe you think we'll 1415 01:21:49,480 --> 01:21:52,200 spend way too much time in front of our computers. 1416 01:21:52,200 --> 01:21:54,240 It's like actually healthy to go and meet in person. 1417 01:21:54,680 --> 01:21:56,520 I think so. I think it's good to meet in 1418 01:21:56,520 --> 01:22:00,080 person and just like who was I just talking with? 1419 01:22:00,880 --> 01:22:05,480 It's like just the like I was just down in Austin for the X 1420 01:22:05,480 --> 01:22:08,320 Max conference, the small. Oh yeah, how was that? 1421 01:22:08,640 --> 01:22:11,000 He was awesome. I got mad at Nathan for for 1422 01:22:11,000 --> 01:22:13,520 putting it at the same time. As yeah. 1423 01:22:13,560 --> 01:22:14,840 I couldn't. I couldn't go. 1424 01:22:15,040 --> 01:22:17,720 I was like same time as beam. Beam. 1425 01:22:17,720 --> 01:22:19,440 Beams or whatever code beam, yeah. 1426 01:22:22,040 --> 01:22:25,480 And like, I know this is just for me, but could you just shift 1427 01:22:25,480 --> 01:22:27,600 it by week? Yeah, he obviously told me to 1428 01:22:28,080 --> 01:22:28,680 get lost. I'm. 1429 01:22:28,960 --> 01:22:30,920 Like, but he did such a good job. 1430 01:22:30,920 --> 01:22:32,360 If you don't ask, you don't get right. 1431 01:22:32,640 --> 01:22:33,840 Right. Well, it was worth it. 1432 01:22:33,840 --> 01:22:35,640 It was worth asking. It was worth a try. 1433 01:22:35,640 --> 01:22:37,400 Was it great like? It was great. 1434 01:22:37,920 --> 01:22:41,560 Like honestly, like the talks were all great, right? 1435 01:22:41,560 --> 01:22:44,400 But the best part was like, it was such a small conference. 1436 01:22:44,400 --> 01:22:48,320 You had time to talk to everyone in between talks and at lunch. 1437 01:22:48,680 --> 01:22:52,800 And so you meet a lot of really cool people and that like, I 1438 01:22:52,800 --> 01:22:56,200 can't remember who I was just talking to, but it's like that, 1439 01:22:56,200 --> 01:22:59,840 like the hallway talk is how you solve problems. 1440 01:22:59,840 --> 01:23:03,480 Like you could just be like, hey, I'm like, I don't know, my 1441 01:23:03,520 --> 01:23:08,600 absinthe graph queries are like joining my tables in like a 1442 01:23:10,000 --> 01:23:12,040 horrible way. And I don't want it to do that. 1443 01:23:12,040 --> 01:23:14,520 Like you can talk about all these problems that you might 1444 01:23:14,520 --> 01:23:16,000 have that you don't know how to solve. 1445 01:23:16,040 --> 01:23:20,800 And nine times out of 10, someone you're talking to or 1446 01:23:20,800 --> 01:23:25,320 overhearing your conversation has been there and. 1447 01:23:25,520 --> 01:23:27,040 Well, exactly. And I, you know, like the 1448 01:23:27,040 --> 01:23:30,280 hallway track is if you haven't been to conferences or it's been 1449 01:23:30,280 --> 01:23:32,880 a while, like the hallway track is really where it's at. 1450 01:23:33,240 --> 01:23:36,120 I mean, talks are great, but you know, and people put a lot of 1451 01:23:36,120 --> 01:23:41,080 effort into talks, but you know, you can watch the video later, 1452 01:23:41,240 --> 01:23:45,320 but it is important to see the talks because, you know, you can 1453 01:23:45,320 --> 01:23:48,920 then, you know, chat about that stuff with folks in the hallway. 1454 01:23:49,440 --> 01:23:51,800 And so you're never going to, you know, it always feels 1455 01:23:51,800 --> 01:23:54,920 terrible when you, you've got 3 tracks and I'm like, if I see 1456 01:23:54,920 --> 01:23:57,280 one, I'm missing 2. And then it's like, I always 1457 01:23:57,280 --> 01:24:00,400 find that super overwhelming. Yeah, the single track's the way 1458 01:24:00,400 --> 01:24:03,160 to go. I really like the the smallness 1459 01:24:03,200 --> 01:24:05,280 of the single track. You don't have decisions to 1460 01:24:05,280 --> 01:24:08,960 make. And well, the decision is like, 1461 01:24:08,960 --> 01:24:12,160 am I in the thing or am I having chats outside? 1462 01:24:12,160 --> 01:24:16,640 Exactly. And so yeah, Goat Maya was the 1463 01:24:16,640 --> 01:24:19,680 pinnacle of small regional I. Heard that it was a lot of fun. 1464 01:24:19,680 --> 01:24:20,480 It was really well. Done. 1465 01:24:20,480 --> 01:24:22,560 It was so good. Yeah, that's awesome. 1466 01:24:22,680 --> 01:24:25,800 And Lars is cooking up another one. 1467 01:24:26,360 --> 01:24:29,520 Nice. Awesome already, and so you 1468 01:24:29,520 --> 01:24:33,800 should probably look at Goat Maya too, in the little town of 1469 01:24:33,800 --> 01:24:35,880 Varberg. And Western, that would be fun. 1470 01:24:36,000 --> 01:24:38,440 That would be fun. I want to, I got to start doing 1471 01:24:38,440 --> 01:24:41,000 making appearances because I think it'd also be fun to like 1472 01:24:41,000 --> 01:24:44,800 kind of change it up and do some like live podcast episodes with 1473 01:24:44,800 --> 01:24:47,360 people. Just snatch them out of the 1474 01:24:47,360 --> 01:24:50,760 hallway, throw a lapel mic on them and go I think. 1475 01:24:50,760 --> 01:24:53,560 It would be, yeah, yeah. I mean, look, capturing the 1476 01:24:53,560 --> 01:24:56,240 conversations that happen at those kind of conferences is 1477 01:24:56,320 --> 01:24:59,720 certainly that would be a lot of content gold. 1478 01:25:00,520 --> 01:25:00,840 Yeah. Yeah. 1479 01:25:00,840 --> 01:25:04,320 And look, you know, I, I was introduced to the small regional 1480 01:25:04,840 --> 01:25:10,840 conference certainly in Elixir, big Elixir, which was in New 1481 01:25:10,840 --> 01:25:14,640 Orleans in I don't know what year that was, 2022 I think it 1482 01:25:14,640 --> 01:25:17,720 was. And that was amazing. 1483 01:25:17,720 --> 01:25:23,920 It was, it was sort of in an old New Orleans theatre, the Petite 1484 01:25:23,920 --> 01:25:26,640 Theatre. You know, it wasn't really 1485 01:25:26,640 --> 01:25:28,160 catered. Like you went out for lunch and 1486 01:25:28,160 --> 01:25:30,640 you'd grab some folks and found some po boys or, you know, 1487 01:25:30,640 --> 01:25:32,600 whatever you find on the streets of New Orleans. 1488 01:25:32,600 --> 01:25:36,640 It was so fun like and it's a shame that hasn't happened 1489 01:25:36,800 --> 01:25:39,040 since. But yeah, that's how Max was. 1490 01:25:39,040 --> 01:25:41,200 It was nice. We would just grab a group of 1491 01:25:41,200 --> 01:25:44,720 guys and just go to a local restaurant at lunch and that was 1492 01:25:44,720 --> 01:25:46,120 it. And then we'd grab some drinks 1493 01:25:46,240 --> 01:25:51,040 afterwards at the end of the day and then, you know, rinse it. 1494 01:25:51,040 --> 01:25:53,200 I. Was sad that I couldn't make 1495 01:25:53,200 --> 01:25:55,960 that one. Austin is a fun town, but yeah, 1496 01:25:55,960 --> 01:25:57,680 New Orleans. Is a very fun town. 1497 01:25:57,760 --> 01:26:00,480 I've never been in New Orleans and I'm sure that's a fun town 1498 01:26:00,560 --> 01:26:02,800 it. Seems like it would be, you 1499 01:26:02,800 --> 01:26:06,080 know, yeah, it's definitely, definitely cool. 1500 01:26:06,560 --> 01:26:10,680 It's like, you know, yeah, I'm surprised that you haven't been 1501 01:26:10,680 --> 01:26:12,680 there. But it's like I had this sort of 1502 01:26:13,640 --> 01:26:17,240 very interesting idea of what in my mind of what New Orleans 1503 01:26:17,240 --> 01:26:21,280 would be like. And it it was quite different 1504 01:26:22,520 --> 01:26:25,400 current to what I imagined, but it was also similar in some. 1505 01:26:25,440 --> 01:26:27,160 I just imagine it's smelling like. 1506 01:26:27,160 --> 01:26:30,800 It's like super European. I don't know. 1507 01:26:30,800 --> 01:26:33,360 No, like I think, you know, Bourbon Street's Bourbon Street. 1508 01:26:33,360 --> 01:26:36,040 Yeah, there's a crazy on all the time. 1509 01:26:36,160 --> 01:26:40,400 And, you know, certainly they they like a a very liberal free 1510 01:26:40,400 --> 01:26:42,840 pour of odd spirits. Yeah. 1511 01:26:43,120 --> 01:26:45,760 Well, I'm from Wisconsin. We're the same way, you know. 1512 01:26:45,760 --> 01:26:48,880 The secrets in the name of Bourbon Street and then, you 1513 01:26:49,040 --> 01:26:53,040 know, Mardi Gras just sounded kind of crazy, but but you find 1514 01:26:53,040 --> 01:26:58,320 your own kind of fun and certainly like it's a super 1515 01:26:58,320 --> 01:27:03,280 European feeling town and it just feels old and different to 1516 01:27:03,320 --> 01:27:06,600 anywhere else in America. Yeah, it looks like it'd be fun. 1517 01:27:06,720 --> 01:27:08,040 One of these days, I'll get there. 1518 01:27:08,600 --> 01:27:11,080 I have not been to Nashville either and I need to. 1519 01:27:11,080 --> 01:27:12,280 Go. I haven't been to Nashville. 1520 01:27:14,440 --> 01:27:16,680 Closest I've been is Chattanooga I think. 1521 01:27:17,240 --> 01:27:21,320 OK, nice. That's Bruce. 1522 01:27:21,360 --> 01:27:23,960 Bruce and Maggie's conference, right? 1523 01:27:25,520 --> 01:27:28,000 Yeah, I heard rumors. I don't know if they're 1524 01:27:28,000 --> 01:27:31,480 substantiated that they're not doing Geek City this year, which 1525 01:27:31,520 --> 01:27:34,320 would be a shame, but that is another another amazing. 1526 01:27:35,080 --> 01:27:36,720 I've heard good things I wanted. To. 1527 01:27:36,720 --> 01:27:39,040 Like make it out to that one as well. 1528 01:27:40,520 --> 01:27:43,080 Yeah, I've, I've been to the last couple of amazing and 1529 01:27:43,320 --> 01:27:45,880 Maggie and Bruce are the consummate hosts. 1530 01:27:46,280 --> 01:27:50,600 And I've heard, you know, yeah, Chattanooga's a fun Old Town. 1531 01:27:51,360 --> 01:27:54,080 And yeah, well, if they're not doing it this year, then 1532 01:27:54,280 --> 01:27:58,440 certainly give them your support and make sure they come back 1533 01:27:58,440 --> 01:28:00,240 bigger and better next year. Yeah, for. 1534 01:28:00,320 --> 01:28:01,520 Sure. Because yeah, like, it's tough 1535 01:28:01,520 --> 01:28:05,320 running conferences, you know, there's not a lot of money in it 1536 01:28:05,320 --> 01:28:10,920 and it's hard work and, you know, you know, it's a roller 1537 01:28:10,920 --> 01:28:13,960 coaster. So really all I can do is say, 1538 01:28:14,000 --> 01:28:16,840 you know, throw your small regional conference, organize 1539 01:28:16,840 --> 01:28:19,840 your support because those are the things that make community, 1540 01:28:19,840 --> 01:28:22,880 I think. I agree, I've always. 1541 01:28:22,880 --> 01:28:26,320 Had people together trying. To do a Salt Lake City one but I 1542 01:28:26,400 --> 01:28:28,720 don't know it seems. Like, oh, you heard it here 1543 01:28:28,720 --> 01:28:32,560 first, folks. Salt Lake City 2027. 1544 01:28:32,840 --> 01:28:36,320 I I've been probably doing the same about Sydney 1. 1545 01:28:36,520 --> 01:28:39,920 We'll go on a camping trip in the mountains and then talk 1546 01:28:39,920 --> 01:28:41,120 Elixir. How about? 1547 01:28:41,120 --> 01:28:45,640 That hey, I have this. So we used to do elixir camps. 1548 01:28:45,720 --> 01:28:47,960 We've never done an Elixir conference in Australia, but we 1549 01:28:47,960 --> 01:28:52,680 used to do Elixir camps and that has been kind of between a meet 1550 01:28:52,680 --> 01:28:56,040 up and a conference. It's sort of like a weekend trip 1551 01:28:56,040 --> 01:28:59,360 away and we it was based off rails camps which we did from 1552 01:28:59,360 --> 01:29:03,000 like 2000 and five, 2006. Did you? 1553 01:29:03,400 --> 01:29:06,560 I feel like you shared this with me a few months. 1554 01:29:06,560 --> 01:29:07,720 I probably did. Yeah. 1555 01:29:07,800 --> 01:29:08,240 OK. Cool. 1556 01:29:08,240 --> 01:29:11,320 I may have showed you the the sort of thing like sort of. 1557 01:29:11,320 --> 01:29:14,640 Blueprint that was behind it. Well, OK. 1558 01:29:14,640 --> 01:29:16,320 And I may as well talk about it here. 1559 01:29:16,320 --> 01:29:17,920 It's like, yeah, of course. Elixir. 1560 01:29:18,240 --> 01:29:23,240 Elixir dot camp, the domain and that was where we ran five 1561 01:29:23,400 --> 01:29:26,360 really amazing elixir camps pre COVID. 1562 01:29:26,640 --> 01:29:30,880 So it sort of all fell apart in 2020. 1563 01:29:31,400 --> 01:29:34,400 And so we haven't done one since, but we would get like 50 1564 01:29:34,400 --> 01:29:39,560 plus people out to, you know, somewhere in the Bush camp if 1565 01:29:39,560 --> 01:29:44,280 you want, or sort of, you know, scout camp or sport and rec 1566 01:29:44,280 --> 01:29:49,160 camp, summer camp. I think the Americans, that kind 1567 01:29:49,160 --> 01:29:52,640 of vibe and then you had sort of bunk beds and dorm beds and like 1568 01:29:52,640 --> 01:29:55,080 just, you know, it wasn't enough people to fill it up. 1569 01:29:55,080 --> 01:29:58,480 So you'd sort of pick where you want to sleep or bring a tent if 1570 01:29:58,480 --> 01:30:00,960 you wanted to just be. Yeah, that would be fun minus 1571 01:30:00,960 --> 01:30:03,960 like, I don't want to bunk with anyone but camping and just 1572 01:30:03,960 --> 01:30:06,000 like, right. Having like so. 1573 01:30:06,200 --> 01:30:10,040 Mountain bike activities or rock climbing, and then some elixir 1574 01:30:10,040 --> 01:30:13,040 too. So you should totally organize 1575 01:30:13,040 --> 01:30:16,280 one of those because it's certainly easier than organizing 1576 01:30:16,320 --> 01:30:19,720 a whole conference. And you have a much smaller sort 1577 01:30:19,720 --> 01:30:26,360 of crowd of sort of 30 to 50 folks, you know, making friends 1578 01:30:26,360 --> 01:30:30,360 and talking about tech and just whatever and, you know, mixing 1579 01:30:30,360 --> 01:30:32,880 some hacking and working on some stuff. 1580 01:30:33,200 --> 01:30:38,120 I always thought it would get to do the angle of like, I know 1581 01:30:38,120 --> 01:30:40,360 there's like a lot of masterminds and groups out there 1582 01:30:40,360 --> 01:30:43,840 to try to help, but like being the technical founder and like 1583 01:30:43,840 --> 01:30:48,640 running a successful like SAS company, Like I struggle with 1584 01:30:48,640 --> 01:30:51,200 the actual sales of my SAS company, right? 1585 01:30:51,200 --> 01:30:54,680 I can build the product and then it's, it's hard to bring on the 1586 01:30:54,680 --> 01:30:57,280 users. And I think it'd be awesome to 1587 01:30:57,280 --> 01:31:01,120 have a group of the same kind of people, some more successful, 1588 01:31:01,120 --> 01:31:04,360 some last, but like have conversations around marketing 1589 01:31:04,360 --> 01:31:09,560 and user acquisition and like just like learn from a campfire 1590 01:31:09,560 --> 01:31:12,720 conversation, you know? I think that stuff exists, but 1591 01:31:12,720 --> 01:31:15,720 the format is, you know, very doable. 1592 01:31:15,880 --> 01:31:18,120 You know, there's a bit of organizing to do, but it's not a 1593 01:31:18,120 --> 01:31:21,080 whole conference, right. And it's, it's sort of like an 1594 01:31:22,280 --> 01:31:25,320 unconference, if you will. You can have, you know, people 1595 01:31:25,360 --> 01:31:29,760 sort of sign up for talks, but it can be very informal and ad 1596 01:31:29,760 --> 01:31:31,840 hoc and you don't have to be in the talks or whatever. 1597 01:31:31,880 --> 01:31:34,880 Yeah. And so look, yeah, I wanted to, 1598 01:31:34,880 --> 01:31:38,600 I've sort of published A proto blueprint for that, but I I will 1599 01:31:38,840 --> 01:31:41,080 publish that. But do you think there is like 1600 01:31:41,080 --> 01:31:43,960 micro SAS conferences? And there are. 1601 01:31:44,360 --> 01:31:47,720 Maybe hacking events and stuff? And like Rob Walling has a lot 1602 01:31:47,720 --> 01:31:50,000 of stuff out there, a lot of content. 1603 01:31:50,000 --> 01:31:52,920 He wrote the SAS playbook. I actually had him on as a guest 1604 01:31:52,920 --> 01:31:57,040 a couple weeks ago. I think now I it's all a blur. 1605 01:31:57,040 --> 01:32:00,080 But he's awesome and he's doing really cool stuff. 1606 01:32:00,080 --> 01:32:03,520 But I think it'd be like even cooler to like, you know, niche 1607 01:32:03,520 --> 01:32:07,800 down even more and like build that same kind of community for 1608 01:32:07,800 --> 01:32:11,920 Elixir developers that are running their own SAS and then. 1609 01:32:11,920 --> 01:32:12,600 Oh, right. Yeah. 1610 01:32:12,680 --> 01:32:17,200 Look, and I think you know in in any kind of distribution game 1611 01:32:18,280 --> 01:32:22,800 sales and marketing is definitely one of those and. 1612 01:32:22,800 --> 01:32:25,000 And I'd be curious to see, you know, like. 1613 01:32:25,000 --> 01:32:28,800 Developers are not great at. Well, we talk about how great, 1614 01:32:28,840 --> 01:32:32,920 you know, Clod Code is with helping us write code and, you 1615 01:32:32,920 --> 01:32:36,320 know, all of that. I'd love to see how like, 1616 01:32:36,560 --> 01:32:40,240 salespeople and marketers are using AI to make their lives 1617 01:32:40,240 --> 01:32:41,600 easier. You know, learn new. 1618 01:32:41,600 --> 01:32:45,440 Well, I, I, when you can sort of wave a magic wand and you've got 1619 01:32:45,720 --> 01:32:50,640 anything you can think of built for you to some degree, you 1620 01:32:50,640 --> 01:32:52,760 know, maybe it's not ready to put in front of a load of 1621 01:32:52,760 --> 01:32:55,000 people, but certainly, you know, you can. 1622 01:32:55,960 --> 01:32:59,440 Anything you can think of, you can basically build now. 1623 01:32:59,560 --> 01:33:01,400 And the limit is your imagination. 1624 01:33:01,840 --> 01:33:05,360 And so often when you can build something that quickly, the 1625 01:33:05,360 --> 01:33:08,680 bottleneck becomes, well, how do I get it out to people and 1626 01:33:08,680 --> 01:33:12,040 distribute it? And you know, that is largely 1627 01:33:12,040 --> 01:33:17,080 where a lot of the problems are going to be is in, well, if you 1628 01:33:17,080 --> 01:33:19,040 can build anything, how do you sell it to people? 1629 01:33:19,240 --> 01:33:21,600 And that that is what software developers are traditionally 1630 01:33:22,360 --> 01:33:25,960 extremely bad at. And, you know, I don't think 1631 01:33:25,960 --> 01:33:29,080 we're going to really want to know how to stand out in an 1632 01:33:29,080 --> 01:33:33,320 extremely crowded market because communicating that stuff is a 1633 01:33:33,320 --> 01:33:36,760 real skill, for sure. So yeah, connecting disparate 1634 01:33:36,760 --> 01:33:38,720 communities for sure would be very cool. 1635 01:33:39,120 --> 01:33:40,440 Yeah. But that would be fun. 1636 01:33:41,400 --> 01:33:44,600 I'm going to hold myself to putting like a blueprint out for 1637 01:33:44,600 --> 01:33:47,400 running Elixir camps because we learned a ton doing that and the 1638 01:33:47,400 --> 01:33:49,280 Rails camp. That would be awesome. 1639 01:33:49,640 --> 01:33:51,960 There was one in the US and like, people could just run 1640 01:33:51,960 --> 01:33:55,080 their own. And I think it's the perfect 1641 01:33:55,080 --> 01:33:59,160 sort of small regional conference thing. 1642 01:33:59,160 --> 01:34:00,480 And, you know, you can make some. 1643 01:34:00,600 --> 01:34:02,880 We'll have to. Make it like we'll have to time 1644 01:34:02,880 --> 01:34:06,800 it in with your with your pit stop before Chicago, right right 1645 01:34:06,800 --> 01:34:08,200 here in Salt Lake. So. 1646 01:34:08,520 --> 01:34:10,520 Oh yeah. We should, we should get cooking 1647 01:34:10,520 --> 01:34:13,200 on that. I'm about I'm about to have to 1648 01:34:13,200 --> 01:34:15,640 book my next round the world ticket because that's how I get 1649 01:34:15,640 --> 01:34:18,720 home from Malaga. That's because I've got to book 1650 01:34:18,720 --> 01:34:20,240 a whole nother round the world ticket. 1651 01:34:20,920 --> 01:34:23,320 So I have. You pick the longest flights to 1652 01:34:23,360 --> 01:34:29,720 go and you just go. Yeah, hey, how are we doing on 1653 01:34:29,720 --> 01:34:32,480 on time, by the way? I think we're good, like we can 1654 01:34:32,640 --> 01:34:34,800 wrap things up. This has been a lot of fun 1655 01:34:34,800 --> 01:34:37,600 though. Yeah, man, I kind of figured, 1656 01:34:37,600 --> 01:34:38,680 oh, what are we going to talk about? 1657 01:34:38,720 --> 01:34:40,400 And then? I was like, dude, you did. 1658 01:34:40,440 --> 01:34:42,160 You don't have to worry. It's going to go. 1659 01:34:43,120 --> 01:34:46,240 Just just, you know, pull the string and see what happens. 1660 01:34:46,320 --> 01:34:48,800 Exactly. That's the way to go and we'll 1661 01:34:48,800 --> 01:34:50,560 have to bring you back. On it feels like, you know, 1662 01:34:50,560 --> 01:34:52,160 we've got like another few hours to. 1663 01:34:52,280 --> 01:34:55,680 I know we'll have to bring you back on and dive into clarity 1664 01:34:55,680 --> 01:34:59,080 and maybe it'll be in a better, better spot in a. 1665 01:34:59,120 --> 01:35:00,560 Few months. Well, what I realized about 1666 01:35:00,560 --> 01:35:06,640 clarity, like so sort of my idea about bringing the sort of 1667 01:35:07,040 --> 01:35:11,240 introspection visualization of ash, so you can introspect 1668 01:35:11,240 --> 01:35:14,960 anything in ash. And we already had some mixed 1669 01:35:14,960 --> 01:35:18,320 tasks baked in very early on that you could create mermaid 1670 01:35:18,320 --> 01:35:24,360 diagrams for your, you know, the relationship diagrams or class 1671 01:35:24,360 --> 01:35:27,600 diagrams or you know, there's some C4 stuff there. 1672 01:35:27,600 --> 01:35:32,840 Now, it's not great because Mermaid C4I chatted to the guy 1673 01:35:32,840 --> 01:35:36,920 that made C4 at the Yale conference last year and he was 1674 01:35:36,920 --> 01:35:40,160 like, Nah, they just stole plant your melon and made it garbage. 1675 01:35:40,560 --> 01:35:44,760 But yeah, like there's this whole structurizer tool that 1676 01:35:44,760 --> 01:35:48,000 exists to do C4 properly, which is sort of zooming in. 1677 01:35:49,200 --> 01:35:51,600 But yeah, like there's introspection and you can create 1678 01:35:51,600 --> 01:35:54,360 diagrams from this stuff, but you can also just create 1679 01:35:54,400 --> 01:35:58,600 documentation like describing your, yeah. 1680 01:35:59,560 --> 01:36:01,840 I thought it was very cool. You kind of gave me a 1681 01:36:01,840 --> 01:36:04,120 walkthrough on Killswitch, actually. 1682 01:36:04,160 --> 01:36:05,320 I think that's what we ended. Up. 1683 01:36:05,320 --> 01:36:08,160 Yeah, that's right. That's yeah, that's a while ago 1684 01:36:08,160 --> 01:36:09,960 now. It feels like a while ago it 1685 01:36:09,960 --> 01:36:11,480 was. I feel like it was at least a 1686 01:36:11,480 --> 01:36:13,520 couple months ago. I don't know, maybe. 1687 01:36:13,560 --> 01:36:18,160 And so, yeah, I, I sort of put Jonathan onto like, hey, can you 1688 01:36:18,160 --> 01:36:19,720 build this out? And he did. 1689 01:36:19,720 --> 01:36:24,360 And we've sort of released it very, very early, rough version. 1690 01:36:24,360 --> 01:36:27,080 And I've had this branch going for like, yeah, at least a month 1691 01:36:27,080 --> 01:36:32,480 or two where I fixed a lot of the glitches and, you know, 1692 01:36:32,560 --> 01:36:36,400 added a few more features. But I haven't sort of twisted 1693 01:36:36,400 --> 01:36:38,360 the lid off the jar because I've been distracted by it. 1694 01:36:38,720 --> 01:36:40,320 Just a few too many other things. 1695 01:36:40,320 --> 01:36:43,920 But what I've realized this week, sort of getting back into 1696 01:36:43,920 --> 01:36:47,080 it, is that what Jonathan has actually done? 1697 01:36:47,760 --> 01:36:50,280 And it seemed overkill at the time, but what he's done is 1698 01:36:50,840 --> 01:36:53,680 hoove it in like everything in your Elixir app, like every 1699 01:36:53,680 --> 01:37:00,240 Elixir module and created an in memory graph data store. 1700 01:37:00,760 --> 01:37:04,560 And so it's all done in digraph, which is the the Erlang sort of 1701 01:37:04,560 --> 01:37:07,400 graph library which stores things away and ants and cases 1702 01:37:07,400 --> 01:37:09,720 it. But what it is is sort of like 1703 01:37:09,720 --> 01:37:13,840 an agentic knowledge graph of your entire Elixir application. 1704 01:37:14,160 --> 01:37:17,000 And you know when you have sort of graph relationships, you can 1705 01:37:17,000 --> 01:37:23,480 attach knowledge to the nodes and you can relate everything to 1706 01:37:23,480 --> 01:37:26,960 each other like not only within your ash domain but within your 1707 01:37:26,960 --> 01:37:30,200 entire application. So you could do like dependency 1708 01:37:30,200 --> 01:37:33,120 analysis because you've hoovered in all your dependencies. 1709 01:37:33,120 --> 01:37:38,680 You can do like xref graph things well and you can sort of 1710 01:37:39,120 --> 01:37:41,840 ask any question about. Your and then you got the 1711 01:37:41,840 --> 01:37:45,320 context of everything in your project which is. 1712 01:37:45,400 --> 01:37:49,960 And so I, Jonathan sort of buried that a little bit and and 1713 01:37:49,960 --> 01:37:51,760 maybe I did, I just didn't click. 1714 01:37:51,760 --> 01:37:55,240 The big idea was actually, oh, because we're looking at multi 1715 01:37:55,240 --> 01:38:00,320 model data stores where where they graph their vector time 1716 01:38:00,320 --> 01:38:03,640 series and we're looking at arcade DB, which is an 1717 01:38:03,640 --> 01:38:07,800 interesting open source Java thing that mixes all of these 1718 01:38:07,800 --> 01:38:11,360 models together because I think 1 model of data store isn't 1719 01:38:11,480 --> 01:38:14,720 going to cut it. And so you sort of need all of 1720 01:38:14,720 --> 01:38:18,720 those ways of navigating knowledge and context and you 1721 01:38:18,720 --> 01:38:22,600 want to be super queriable and fast. 1722 01:38:23,080 --> 01:38:26,200 And yeah, when I realized that Jonathan's just basically built 1723 01:38:26,880 --> 01:38:29,680 like an in memory knowledge store, I'm like, oh, what can we 1724 01:38:29,680 --> 01:38:32,360 do with this? It takes a while to ingest it 1725 01:38:32,360 --> 01:38:34,720 all, but not that long. It was very slow and then you 1726 01:38:34,720 --> 01:38:36,560 made it a lot faster. But that's awesome. 1727 01:38:36,560 --> 01:38:40,000 Very cool. You know what I showed you was 1728 01:38:40,000 --> 01:38:43,360 kind of like the very minimum viable demoable. 1729 01:38:43,680 --> 01:38:47,000 I saw the big picture though, like I saw where it could go and 1730 01:38:47,000 --> 01:38:50,920 it was cool, you know? And so you, you've got a sense 1731 01:38:50,920 --> 01:38:53,760 of like, oh, what does my domain actually look like and how does 1732 01:38:53,760 --> 01:38:56,840 it hang together? And if you put in all of the 1733 01:38:57,160 --> 01:39:01,160 annotations around documentation in your ash thing, not only can 1734 01:39:01,160 --> 01:39:04,080 they go out to your APIs. And in fact, I think I even 1735 01:39:04,080 --> 01:39:07,600 found some like messy resources I wanted to clean up after we 1736 01:39:07,600 --> 01:39:10,360 did that, remember? So it was good. 1737 01:39:10,360 --> 01:39:12,080 I was like why do I have this here? 1738 01:39:13,000 --> 01:39:17,080 And so really, it's hard to like, you know, like Ash was 1739 01:39:17,080 --> 01:39:19,360 hard to explain. Clarity's a little bit 1740 01:39:19,360 --> 01:39:24,360 interesting because it's sort of well, depending on the audience, 1741 01:39:24,360 --> 01:39:27,320 we had this lens idea of like, well, if you're thinking in 1742 01:39:27,320 --> 01:39:31,080 architecture mode, you want to see how my Ash resources and 1743 01:39:31,080 --> 01:39:33,080 domains sort of relate to each other. 1744 01:39:33,080 --> 01:39:35,200 What does my business logic look like? 1745 01:39:35,240 --> 01:39:38,320 Because that's really what you're bringing and what what 1746 01:39:38,320 --> 01:39:41,640 you're spending the effort to build is the the core platform. 1747 01:39:41,840 --> 01:39:46,040 And you know that is going to look really quite insightful. 1748 01:39:46,120 --> 01:39:49,760 If you saw my keynote in Berlin, I talk about why, like in Ash, 1749 01:39:49,760 --> 01:39:53,560 you're basically modelling your core system and then you derive 1750 01:39:53,560 --> 01:39:57,320 things like the database structure and the migrations to 1751 01:39:57,320 --> 01:40:00,000 get there because it doesn't really matter where you store 1752 01:40:00,000 --> 01:40:03,120 your data. Actually, you know, there's lots 1753 01:40:03,120 --> 01:40:06,880 of possible options. And I think we in 2026, we're 1754 01:40:06,880 --> 01:40:10,400 starting to see the real need and necessity to have different 1755 01:40:10,400 --> 01:40:14,360 ways of storing data because relational just doesn't cut it 1756 01:40:14,360 --> 01:40:16,720 for everything. And you know, I think we've been 1757 01:40:16,720 --> 01:40:20,000 seeing that for quite a while, but you know, like the 1758 01:40:20,000 --> 01:40:24,480 relational DB is a good run and Postgres is amazing. 1759 01:40:24,480 --> 01:40:25,920 I'm not saying don't use Postgres. 1760 01:40:25,920 --> 01:40:27,600 We. Just need a little more pain to 1761 01:40:27,600 --> 01:40:30,160 make it apparent, right? And then you're forced to make 1762 01:40:30,160 --> 01:40:33,280 the changes and we're. Seeing look and and ash deals 1763 01:40:33,280 --> 01:40:35,200 with a lot of that stuff for you, but there are different 1764 01:40:35,200 --> 01:40:37,320 ways to store data for different reasons. 1765 01:40:37,320 --> 01:40:41,120 And I think multi model in terms of storing data is is gonna be 1766 01:40:41,120 --> 01:40:44,560 certainly the way forward. But also, you know, I talked 1767 01:40:44,560 --> 01:40:49,200 about in my Berlin keynote code beam about like the 1768 01:40:49,200 --> 01:40:53,800 disappearance of the UI. Like I watch Lars do some stuff. 1769 01:40:54,080 --> 01:40:57,720 He he yeah, ping me on Discord going, Hey, this Ash AI stuff is 1770 01:40:57,720 --> 01:41:00,040 amazing. And he was basically doing 1771 01:41:00,040 --> 01:41:04,960 Goatmeyer organization through Claude by talking to an MCP that 1772 01:41:04,960 --> 01:41:07,720 was exposed to his goat Maya sort of platform. 1773 01:41:07,720 --> 01:41:10,680 And he didn't have AUI, it was just Ward and he was just 1774 01:41:10,680 --> 01:41:14,320 interacting. And so I'm doing the same kinds 1775 01:41:14,320 --> 01:41:17,160 of things. I've been seeing the big push of 1776 01:41:17,400 --> 01:41:23,200 worry more about creating good APIs now versus a full app like 1777 01:41:23,200 --> 01:41:25,800 give people access. And what is an API? 1778 01:41:25,800 --> 01:41:29,720 Like we had REST and we had Graph QL, but you know, MCP was 1779 01:41:29,720 --> 01:41:32,560 the stuff I've been talking about for quite a while now 1780 01:41:32,560 --> 01:41:34,760 since it came out like a year and a half or something. 1781 01:41:34,880 --> 01:41:39,000 It's November 2024, and I saw the promise of that because it's 1782 01:41:39,000 --> 01:41:42,480 like, oh, API is native to how LLMS work. 1783 01:41:42,760 --> 01:41:47,160 Right now we're seeing the MCPS don't work quite as well because 1784 01:41:47,160 --> 01:41:50,920 they have a lot of metadata that takes up context and figuring 1785 01:41:50,920 --> 01:41:55,160 out tool selection and making the LLM reliably pick a tool or 1786 01:41:55,160 --> 01:41:57,480 a skill or whatever. It's actually a really hard 1787 01:41:57,480 --> 01:41:59,760 problem, especially when you've got a lot of choice. 1788 01:42:00,120 --> 01:42:03,760 It doesn't know how to make the decisions to pick those. 1789 01:42:04,400 --> 01:42:06,560 And so I think that's that's still a problem. 1790 01:42:06,560 --> 01:42:12,240 And so CLI is, is actually a better deal because you get 1791 01:42:12,240 --> 01:42:15,680 progressive disclosure and, and Claude and and Co are really 1792 01:42:16,000 --> 01:42:17,920 trained on running bash commands. 1793 01:42:19,080 --> 01:42:23,680 And so you can do like man the CLI and get the man page and get 1794 01:42:23,680 --> 01:42:26,800 the high level idea. And so they know how to do this 1795 01:42:26,800 --> 01:42:28,960 stuff. And so CLI is in the last few 1796 01:42:28,960 --> 01:42:34,720 weeks has been like absolutely better than MCPS because Zolis 1797 01:42:34,720 --> 01:42:36,600 are just a standardized Unix format. 1798 01:42:36,600 --> 01:42:40,240 And the way that you get help and you understand, Oh yeah, 1799 01:42:40,240 --> 01:42:43,160 this is my top level menu with the one line description of each 1800 01:42:43,160 --> 01:42:45,240 command. And then I can go, you know, 1801 01:42:45,920 --> 01:42:50,560 command help, sub command and get more detailed information 1802 01:42:50,560 --> 01:42:53,680 about how that works. And so progressive disclosure is 1803 01:42:53,760 --> 01:42:57,160 absolutely where it's at for LLMS and humans, right? 1804 01:42:57,160 --> 01:42:59,080 That's how humans learn. Yeah. 1805 01:43:00,320 --> 01:43:02,680 And everyone's realizing that shoving everything into your 1806 01:43:02,680 --> 01:43:08,120 cloud MD is just a way to send that yeah to the to the bot on 1807 01:43:08,120 --> 01:43:09,760 the other. Side well, we also see the 1808 01:43:09,760 --> 01:43:13,520 limitation to that too, Like you can only throw so much in a 1809 01:43:13,520 --> 01:43:18,480 markdown file, you know? Right and well skills were and 1810 01:43:18,880 --> 01:43:20,480 usage rules if you haven't used it. 1811 01:43:20,720 --> 01:43:23,560 Usage rules is amazing. So it's probably got an upgrade. 1812 01:43:24,280 --> 01:43:27,560 Like it's crazy, like the simplicity behind something like 1813 01:43:27,560 --> 01:43:31,760 that, but like the amount of power you get from it. 1814 01:43:32,800 --> 01:43:34,400 Yeah. So if you haven't checked out 1815 01:43:34,400 --> 01:43:38,400 usage rules recently, when Claude announced skills, I was 1816 01:43:38,400 --> 01:43:41,240 like we should be able to take usage rules and turn it into 1817 01:43:41,240 --> 01:43:46,440 skills and did a very draft PR. Where I got stuck was we can't 1818 01:43:46,440 --> 01:43:49,800 auto generate the front matter description, which is how the L 1819 01:43:49,800 --> 01:43:52,760 lens see, because all it loads is the front matter for the 1820 01:43:52,760 --> 01:43:55,240 skill and then it figures out do I need that skill? 1821 01:43:55,560 --> 01:44:00,880 Then it loads the whole file. And so it's pretty critical that 1822 01:44:00,880 --> 01:44:05,120 your description for the tool or the yeah, the skill is is 1823 01:44:05,120 --> 01:44:07,680 perfect because otherwise the LM can't choose to use it. 1824 01:44:07,680 --> 01:44:11,200 It's like use this when and so that wasn't baked in. 1825 01:44:11,200 --> 01:44:14,720 So I think it was Zach that that rejected. 1826 01:44:16,240 --> 01:44:17,440 I'll have to check it out I feel. 1827 01:44:17,600 --> 01:44:20,880 Like you can now create skills and set your own descriptions 1828 01:44:20,880 --> 01:44:21,360 for. Nice. 1829 01:44:22,400 --> 01:44:26,240 Groups of skills like Ash skills or live view or Phoenix or what 1830 01:44:26,240 --> 01:44:29,680 have you and it's it's way, way better and you can sort of turn 1831 01:44:29,680 --> 01:44:33,320 it from shoving everything into clawed into more like 1832 01:44:33,320 --> 01:44:35,640 progressive disclosure through through skills. 1833 01:44:35,800 --> 01:44:39,120 I have been using skills heavily now in my daily workflow and 1834 01:44:39,120 --> 01:44:40,880 it's helpful. It's nice. 1835 01:44:42,200 --> 01:44:46,560 Yeah, I, I've, I've been using cmux for like, I love Ghosty, 1836 01:44:46,960 --> 01:44:51,320 but I've been using cmux and that's been really good around 1837 01:44:51,600 --> 01:44:55,400 just giving you like a Ghosty terminal with vertical tabs, 1838 01:44:55,400 --> 01:44:57,240 which so you can read what the tabs are. 1839 01:44:57,680 --> 01:45:00,800 So you have a lot of tabs. I've got a full side drawer of 1840 01:45:00,800 --> 01:45:02,520 those. And then every time Claude 1841 01:45:02,520 --> 01:45:06,480 stops, it puts a little, does a little Ding and puts a blue ring 1842 01:45:06,480 --> 01:45:10,840 around and yeah, just alerts you to what is needing because I'm 1843 01:45:10,840 --> 01:45:15,240 running like 20 plus sessions constantly and. 1844 01:45:15,320 --> 01:45:17,240 I feel like I had to step up my game. 1845 01:45:17,240 --> 01:45:19,880 I'm like a a single. So try Cmux. 1846 01:45:19,880 --> 01:45:23,920 I really like it. It's based on Ghosty and so it 1847 01:45:23,920 --> 01:45:25,640 just feels very familiar and it's got brown. 1848 01:45:27,080 --> 01:45:32,880 Oh yeah, it's cool looking. It's it's super cool someone Rob 1849 01:45:33,320 --> 01:45:36,360 at olembic also showed us super set SH. 1850 01:45:36,360 --> 01:45:39,960 I think it was was another one. I can't vouch for that one, but 1851 01:45:39,960 --> 01:45:42,720 Rob said it was pretty good. All right, cool. 1852 01:45:42,720 --> 01:45:46,720 And so similar ideas is like, you know, sort of agentic 1853 01:45:47,120 --> 01:45:51,200 workspace terminals and I was, you know, abusing VS Code as 1854 01:45:51,200 --> 01:45:52,960 basically like a team UX because I never got. 1855 01:45:52,960 --> 01:45:55,400 It yeah, I was going to say like I know a lot of people using 1856 01:45:55,400 --> 01:45:58,280 like team UX sessions and things and it seems to work pretty 1857 01:45:58,280 --> 01:46:00,760 well, but. Well, I just use VS Code with 1858 01:46:00,880 --> 01:46:03,400 the way that you can get terminal windows, and I had my 1859 01:46:03,400 --> 01:46:06,760 code and my terminals and my claws and my codex is all in one 1860 01:46:06,760 --> 01:46:07,280 space. Yeah. 1861 01:46:07,400 --> 01:46:11,000 And now I use CMUX for the stuff that it doesn't necessarily have 1862 01:46:11,000 --> 01:46:12,760 code attached attached to it. OK, cool. 1863 01:46:12,760 --> 01:46:14,480 But I use a combination of those I'm. 1864 01:46:15,720 --> 01:46:17,880 Gonna have to do this. I feel like I need a little 1865 01:46:17,880 --> 01:46:21,240 upgrade to my AI workflow so it'll be good to well. 1866 01:46:22,400 --> 01:46:24,240 There you go, CMOS is pretty good. 1867 01:46:26,040 --> 01:46:29,760 You know it's still new and so I found an issue the other day. 1868 01:46:29,760 --> 01:46:34,160 If I dragged a session with a clawed running inside of it over 1869 01:46:34,160 --> 01:46:39,160 to another workspace, it started like the the IT wraps clawed and 1870 01:46:39,160 --> 01:46:43,640 it wraps open something else so that it can inject the hooks 1871 01:46:43,640 --> 01:46:46,320 that it needs. And it's just a shell script 1872 01:46:46,320 --> 01:46:48,160 that you can look at if you're worried about the security. 1873 01:46:49,920 --> 01:46:54,560 And so yeah, it's pretty neat. And it works on Ghosty, which if 1874 01:46:54,560 --> 01:46:57,360 you haven't used Ghosty as a terminal Item 2, was my favorite 1875 01:46:57,360 --> 01:46:59,240 for a long time. And then I discovered Ghosty and 1876 01:46:59,480 --> 01:47:01,800 Mitchell Hashimoto. I think that's his name. 1877 01:47:03,240 --> 01:47:05,920 Right, I have not used it, but I have seen it. 1878 01:47:06,000 --> 01:47:07,480 Oh. Yeah, Ghosty's amazing. 1879 01:47:07,480 --> 01:47:12,080 And I think Zach was talking about, I think it was yesterday 1880 01:47:12,440 --> 01:47:15,680 that they got in in the latest release Applescript scripting 1881 01:47:15,680 --> 01:47:18,760 for Ghost. I did see his post on that. 1882 01:47:18,760 --> 01:47:19,920 Yeah, that's. Cool. 1883 01:47:19,920 --> 01:47:23,720 That's really cool, Apple. Script you know, you may not 1884 01:47:23,720 --> 01:47:27,440 have ever written it, but. I have actually I I have a 1885 01:47:27,480 --> 01:47:29,040 stupid. Script in your math. 1886 01:47:30,040 --> 01:47:34,360 I have a stupid script to mute my microphone at the hardware 1887 01:47:34,360 --> 01:47:38,000 level and I I did that with an Applescript. 1888 01:47:39,000 --> 01:47:41,080 There you go. So Applescript is probably the 1889 01:47:41,080 --> 01:47:44,080 the power tool that you've probably been sleeping on on 1890 01:47:44,080 --> 01:47:46,880 your Mac for. Oh, I I have absolutely. 1891 01:47:47,160 --> 01:47:50,600 I have like a wish list of, I have a wish list of things. 1892 01:47:50,600 --> 01:47:54,040 I want to just take the time and like write out and automate and 1893 01:47:54,040 --> 01:47:55,120 I. Have Well now you don't have to 1894 01:47:55,120 --> 01:47:56,840 take the time. Just ask Ford to do it. 1895 01:47:57,040 --> 01:47:59,040 Yeah, they're. Oh yeah, that's smart. 1896 01:48:00,040 --> 01:48:03,040 Right, Yeah, I see. Like I use cloud. 1897 01:48:03,040 --> 01:48:04,680 Daily. But then I don't think about 1898 01:48:05,200 --> 01:48:08,400 using it to simplify my like working tasks, you know? 1899 01:48:09,520 --> 01:48:12,240 So like I said before, and I think it's probably a good note 1900 01:48:12,240 --> 01:48:16,680 to to leave it on is that you know, you know, you asked like 1901 01:48:16,680 --> 01:48:18,120 where does software engineering go? 1902 01:48:18,160 --> 01:48:20,480 Well, we're not going to be doing any of the grind work 1903 01:48:20,720 --> 01:48:24,280 anymore, right? But we get lifted up to be 1904 01:48:24,360 --> 01:48:26,640 having more time to think about, well, what can we do? 1905 01:48:26,680 --> 01:48:28,080 Where are the limits here? We don't know. 1906 01:48:28,520 --> 01:48:29,920 And the only limits are your taste. 1907 01:48:29,920 --> 01:48:33,360 Now you know. The only limits are your taste 1908 01:48:33,360 --> 01:48:37,920 and judgement, you know, and your imagination. 1909 01:48:38,280 --> 01:48:40,840 Yeah. And time, right. 1910 01:48:41,120 --> 01:48:44,320 And like, you know, how many tokens can you get down from? 1911 01:48:44,360 --> 01:48:47,080 Yeah. How many tokens utilize your 1912 01:48:47,080 --> 01:48:51,880 monthly spend? I did accidentally buy personal 1913 01:48:51,880 --> 01:48:56,480 Max on top of the Olympic team plan premium seat, which is kind 1914 01:48:56,480 --> 01:48:59,080 of like the team version of Max plan. 1915 01:48:59,840 --> 01:49:05,040 And yeah, the whole Olympic team is on that now as of 2026. 1916 01:49:05,080 --> 01:49:07,920 Everyone got on there and earned their stripes. 1917 01:49:07,920 --> 01:49:10,960 They had to push through the API key to sort of figure out, oh, 1918 01:49:11,080 --> 01:49:14,600 I'm using more tokens than it would cost to do the, you know, 1919 01:49:14,640 --> 01:49:17,760 the team plan premium seat. And so, you know, you have to 1920 01:49:17,760 --> 01:49:24,800 earn new stripes and yeah, like, I need to figure out how to 1921 01:49:25,080 --> 01:49:28,160 utilize all the tokens I have at my disposal before the price 1922 01:49:28,160 --> 01:49:29,840 goes up because will the price go up? 1923 01:49:30,040 --> 01:49:32,680 I am, you know, yeah, I feel. Like where they get you on the 1924 01:49:32,680 --> 01:49:35,280 free credits and the cheap tokens and then we're going to 1925 01:49:35,280 --> 01:49:37,080 get. Trapped and all of a sudden 1926 01:49:37,080 --> 01:49:39,960 it'll be like, well I guess it's worth 2 grand a month or you 1927 01:49:39,960 --> 01:49:43,480 know what I mean, per seat. Right. 1928 01:49:44,080 --> 01:49:47,040 And then? If you're using it wisely, it 1929 01:49:47,040 --> 01:49:52,040 probably is. I I know folks using like cursor 1930 01:49:52,040 --> 01:49:55,600 through the like the betting. Plan where it's like his paper's 1931 01:49:55,600 --> 01:50:01,440 in What's Josh's handle on X? Oh, it's just my name. 1932 01:50:01,560 --> 01:50:03,160 Josh Price. Josh Price, right? 1933 01:50:03,200 --> 01:50:04,520 Yeah, there you. Go. 1934 01:50:05,440 --> 01:50:06,400 I'm pretty. Sure, yeah. 1935 01:50:07,080 --> 01:50:09,680 I I tagged him in some of my posts so you should be able to 1936 01:50:09,680 --> 01:50:14,240 find him. Yeah, yeah, I I stopped using X, 1937 01:50:14,880 --> 01:50:19,200 No Twitter. I had a hard time, I couldn't 1938 01:50:19,200 --> 01:50:21,240 accept it. Like I just, it doesn't flow 1939 01:50:21,240 --> 01:50:22,960 right. But like all of a sudden I feel 1940 01:50:22,960 --> 01:50:27,760 like I guess it's been X long enough but I I don't. 1941 01:50:27,880 --> 01:50:32,360 Know, I guess I I don't know, I like when Twitter came out and I 1942 01:50:32,360 --> 01:50:35,360 like Twitter doesn't. Sounds nicer tweets. 1943 01:50:35,360 --> 01:50:37,880 It just makes like. I don't like calling it posts. 1944 01:50:38,080 --> 01:50:40,960 I'd rather call them tweets, but. 1945 01:50:41,000 --> 01:50:43,120 Yeah, I, I'm still old school about that. 1946 01:50:43,560 --> 01:50:46,640 Yeah, I get it. 2008 when it came out, it was just absolutely 1947 01:50:46,640 --> 01:50:48,120 amazing. And it was like having 1948 01:50:48,120 --> 01:50:50,360 conversations with everyone around the world globally, like 1949 01:50:50,360 --> 01:50:52,280 they were they were next to you was amazing. 1950 01:50:53,360 --> 01:50:54,600 And I think it still is like that. 1951 01:50:54,600 --> 01:50:57,320 I enjoy it. There was a, there was a, there 1952 01:50:57,320 --> 01:51:00,040 was a dip for a bit and actually there's. 1953 01:51:00,360 --> 01:51:04,960 I feel like they cleaned up the the for me like. 1954 01:51:05,160 --> 01:51:07,840 I'm getting some really good stuff on there I've been doing. 1955 01:51:08,000 --> 01:51:11,720 Mine has been all elixir based, less politics and. 1956 01:51:12,520 --> 01:51:15,440 Yeah, yeah, no, I I've been getting into like 3 JS and what 1957 01:51:15,440 --> 01:51:16,720 you can do with that. Oh, nice. 1958 01:51:16,720 --> 01:51:22,440 And combining that with live view because we're building AI 1959 01:51:22,440 --> 01:51:26,600 built sort of vibe up a satellite tracker and it is 1960 01:51:26,600 --> 01:51:30,600 pretty amazing. I may launch that soon, but 1961 01:51:30,600 --> 01:51:34,920 we're, you know, about to start doing a interesting space 1962 01:51:34,920 --> 01:51:38,040 project. That's been something I've been 1963 01:51:38,040 --> 01:51:41,080 waiting for since my my teenage years. 1964 01:51:41,160 --> 01:51:45,600 Well, that's very into like 3D graphics and I've always wanted 1965 01:51:45,600 --> 01:51:48,840 for 30 years, my whole career. Like how can we bring game and 1966 01:51:48,840 --> 01:51:51,720 sci-fi interfaces into business applications? 1967 01:51:51,920 --> 01:51:55,920 And I think with LLMS now, the calculus is such that we can and 1968 01:51:55,920 --> 01:52:00,840 three JS is amazing and it works really really well with life. 1969 01:52:00,840 --> 01:52:02,840 I've seen demos, I've seen demos with it. 1970 01:52:02,840 --> 01:52:04,400 I've never like played around with it though. 1971 01:52:07,480 --> 01:52:09,960 Well, I mean, that's why I asked whether we could do some screen 1972 01:52:09,960 --> 01:52:12,840 sharing on this but. You know what we should? 1973 01:52:13,200 --> 01:52:15,720 Well, I said I'd prefer not to because then you have to like, 1974 01:52:15,720 --> 01:52:17,320 describe everything on the screen. 1975 01:52:17,680 --> 01:52:19,320 But. What we Well, this is a podcast 1976 01:52:19,360 --> 01:52:21,160 after all, so you're correct to say no. 1977 01:52:21,160 --> 01:52:24,320 Yeah. So, but I think what we should 1978 01:52:24,320 --> 01:52:29,160 do is maybe schedule like a coding session and then like I 1979 01:52:29,160 --> 01:52:31,800 don't put those on podcast platforms and we can screenshot. 1980 01:52:31,800 --> 01:52:33,280 Yeah, sure. Yeah. 1981 01:52:33,840 --> 01:52:36,360 So I'm definitely down for that. That would be a lot of fun. 1982 01:52:36,920 --> 01:52:40,320 Because I've been cooking some stuff that is pretty cool. 1983 01:52:40,840 --> 01:52:42,400 We have. And I feel like I'm just getting 1984 01:52:42,400 --> 01:52:44,960 started. Joshua Payne, he said the new 1985 01:52:44,960 --> 01:52:47,520 relational database structure really got me excited. 1986 01:52:47,520 --> 01:52:50,920 And now three JS combined with live view in space. 1987 01:52:51,200 --> 01:52:53,600 Let's go. Right. 1988 01:52:53,960 --> 01:52:57,760 I, I vibed up this satellite tracker in a couple of hours on 1989 01:52:57,760 --> 01:53:00,640 a weekend and I literally had a grin from ear to ear. 1990 01:53:00,640 --> 01:53:04,760 And I said I had so much fun because, yeah, in my uni days, 1991 01:53:04,760 --> 01:53:09,240 in my 3D graphics course, I built a ray tracer in C And 1992 01:53:09,240 --> 01:53:12,440 because I left it to the last minute, I was like, I was on my 1993 01:53:12,440 --> 01:53:16,040 modem telnetted to the university server. 1994 01:53:16,680 --> 01:53:22,600 And then I realized, oh, it's a, it's 10:00 at night and I can't 1995 01:53:22,600 --> 01:53:26,000 render the ray tracer output because I need a graphical 1996 01:53:26,000 --> 01:53:27,560 terminal. And I'm like, what do I do? 1997 01:53:28,120 --> 01:53:32,040 And I had this very slow modem. And so I had to downscale this 1998 01:53:32,880 --> 01:53:37,080 ray tracing output to like, you know, 10 by 10 or 20 by 20, so 1999 01:53:37,080 --> 01:53:39,160 it would fit on a screen and render it in ASCII. 2000 01:53:39,680 --> 01:53:42,200 So I wrote a ASCII renderer so I could keep trucking. 2001 01:53:42,200 --> 01:53:45,160 After hours, I couldn't get the graphics terminals. 2002 01:53:45,600 --> 01:53:48,080 That's awesome. And so yeah, like I've been 2003 01:53:48,080 --> 01:53:49,760 waiting for this moment to come back to the. 2004 01:53:49,880 --> 01:53:51,320 Yeah, for real. That's awesome. 2005 01:53:52,720 --> 01:53:55,600 But yeah, all right, Sport is quite good at 3GS and Live View 2006 01:53:55,600 --> 01:53:59,120 works with it a treat. Yeah, I think let's do a vibe 2007 01:53:59,120 --> 01:54:02,000 coding or just a coding session schedule, then we can play 2008 01:54:02,000 --> 01:54:03,440 around. I think that'd be fun. 2009 01:54:03,960 --> 01:54:05,480 Let's do it. All right, cool. 2010 01:54:05,480 --> 01:54:07,480 Well, thanks so much for joining me, Josh. 2011 01:54:07,560 --> 01:54:09,640 This has been a lot of fun. Thanks for having me and it's 2012 01:54:09,920 --> 01:54:12,480 always great to chat and I feel like we can go for hours. 2013 01:54:12,600 --> 01:54:16,200 I know we can just keep going and we'll just have to schedule 2014 01:54:16,200 --> 01:54:18,240 another one and just continue on. 2015 01:54:18,240 --> 01:54:20,160 Let's do it. Looking forward to the next one. 2016 01:54:20,800 --> 01:54:24,120 Alright guys, thanks for tuning in everyone and I'll, I'll see 2017 01:54:24,120 --> 01:54:30,280 you guys on Friday with another episode testing.