1 00:00:00,040 --> 00:00:03,120 Coding Elixir with AI or building AI into your Elixir 2 00:00:03,120 --> 00:00:05,840 apps? The Elixir AI Collective Discord 3 00:00:05,840 --> 00:00:08,720 is your community for both. Link in the description below. 4 00:00:09,920 --> 00:00:12,520 Secure your Elixir apps with paraxial IO. 5 00:00:12,680 --> 00:00:15,720 From code analysis and bot defense to dependency checks, 6 00:00:15,720 --> 00:00:19,000 they've got you covered. Elixir first design with GitHub 7 00:00:19,000 --> 00:00:21,400 integration. Elevate your security at 8 00:00:21,400 --> 00:00:31,280 paraxial dot IO testing. What's up? 9 00:00:31,320 --> 00:00:34,120 I'm Jacob Litzow, your Elixir mentor and welcome to another 10 00:00:34,120 --> 00:00:36,520 exciting episode of the Elixir Mentor podcast. 11 00:00:36,520 --> 00:00:38,640 This is where we discuss everything related to Elixir, 12 00:00:38,640 --> 00:00:40,800 from interviews with enthusiasts, pioneers in the 13 00:00:40,800 --> 00:00:44,680 community to innovative projects and libraries shaping Elixir's 14 00:00:44,680 --> 00:00:46,920 future. Today I'm joined with Leandro, 15 00:00:46,920 --> 00:00:51,880 creator of MDX and Loomis, and he's also a returning guest on 16 00:00:51,880 --> 00:00:53,920 the podcast. So it's exciting to have you 17 00:00:53,920 --> 00:00:56,600 back. Thanks for joining me today. 18 00:00:56,600 --> 00:01:00,560 And yeah, I always have more fun when I have returning guests. 19 00:01:00,560 --> 00:01:03,560 I don't know why I feel like it's we already like broke the 20 00:01:03,560 --> 00:01:07,280 ice. Thank you for having me back on 21 00:01:07,440 --> 00:01:10,480 and excited to be talking about MBX and Lumis this time. 22 00:01:11,160 --> 00:01:14,200 Yeah. And I feel like last time what 23 00:01:14,200 --> 00:01:18,520 we talked about Beacon and I feel like what were you were 24 00:01:18,520 --> 00:01:25,960 working on, like the UI, was it not? 25 00:01:25,960 --> 00:01:28,160 No, what were you working on? I'm blanking now. 26 00:01:29,000 --> 00:01:33,280 It was about a year ago, so probably I was working on the 27 00:01:33,560 --> 00:01:36,800 admin interface for for Beacon. That's right. 28 00:01:36,920 --> 00:01:40,760 OK. Yeah, So we were working on the 29 00:01:41,200 --> 00:01:44,160 like the live editor of templates. 30 00:01:44,920 --> 00:01:47,960 That's right. And it and it became a Elixir 31 00:01:47,960 --> 00:01:55,360 conf talk in 2000, 24 as well. So yeah, for a little like quick 32 00:01:55,360 --> 00:01:58,560 refresh for those watching that might not know you, do you want 33 00:01:58,560 --> 00:02:00,200 to give like a just a brief background? 34 00:02:01,240 --> 00:02:03,560 Oh, yeah, sure. Yeah, yeah, yeah, sure. 35 00:02:03,600 --> 00:02:08,600 So, yeah, I've been working with Elixir for 10 years now. 36 00:02:09,600 --> 00:02:13,280 You know, I used to work with Rios, you know, back in the day. 37 00:02:13,640 --> 00:02:16,560 And I'm from Brazil, same as Joseph Vallin. 38 00:02:16,560 --> 00:02:22,120 So it was kind of natural for me to get into Elixir, you know, 39 00:02:22,680 --> 00:02:26,520 those 10 years trying to make my my contributions to the 40 00:02:26,520 --> 00:02:30,400 community. And some of you might know me 41 00:02:30,400 --> 00:02:33,920 from Beacon when I was working at Dockyard. 42 00:02:34,760 --> 00:02:38,800 Now I'm working at Nutella Spark, it's an ad tech solution 43 00:02:38,800 --> 00:02:47,280 for US based schools and also working on the MDX and Lumis 44 00:02:47,280 --> 00:02:49,880 projects, Markdown and Syntex highlighter. 45 00:02:50,400 --> 00:02:56,520 It's a you know it's it's a mix of Elixir, Rust and someone now 46 00:02:56,520 --> 00:03:01,880 is working the looms project to Java and I'm also working to 47 00:03:01,880 --> 00:03:05,120 port the looms project to Jean Crete as well. 48 00:03:05,120 --> 00:03:08,520 So there is a lot we can we can discuss. 49 00:03:09,040 --> 00:03:12,520 Yeah, that's really cool. I'm curious what what kind of 50 00:03:12,520 --> 00:03:17,440 what sparked like MDX? Like what made you want to make 51 00:03:17,440 --> 00:03:19,760 a Markdown library? And I'm guessing it's written in 52 00:03:19,760 --> 00:03:24,360 Elixir, right? Yeah, that was like, you know, 53 00:03:24,800 --> 00:03:30,080 it start about two and I don't maybe three years ago ish. 54 00:03:30,680 --> 00:03:33,520 So I was working on Beacons Ms. right? 55 00:03:33,640 --> 00:03:39,720 And we had to migrate all the blog posts from the old Docker 56 00:03:39,720 --> 00:03:45,600 website to Beacon. And that means migrate about 800 57 00:03:46,000 --> 00:03:49,040 blog posts just, you know, markdown. 58 00:03:49,320 --> 00:03:54,000 And as you can imagine, like 10 years of history, you know, 59 00:03:54,000 --> 00:03:57,800 different people writing blog posts, markdown files. 60 00:03:57,800 --> 00:04:02,400 So we had a bunch of different markdowns to migrate from the 61 00:04:02,400 --> 00:04:08,280 old one to Beacon. And that's when we realized that 62 00:04:08,920 --> 00:04:14,720 Earmark at that time was not generating the output we were 63 00:04:14,720 --> 00:04:18,920 expecting. And so just just to be clear, 64 00:04:18,959 --> 00:04:21,040 Earmark is a really amazing library. 65 00:04:21,399 --> 00:04:26,040 You know it, it has been serving the community for so many years 66 00:04:26,040 --> 00:04:27,640 now. So it's a really, really hard 67 00:04:27,640 --> 00:04:30,480 work to do. So if Earmarks works for you, 68 00:04:30,960 --> 00:04:32,720 you know, keep you using, it's amazing. 69 00:04:33,320 --> 00:04:39,080 But for us back then, we had an issue with some blog posts 70 00:04:39,600 --> 00:04:43,800 because the the problem we've marked down is marked down is a 71 00:04:43,800 --> 00:04:48,800 permissive format. Like you can write whatever you 72 00:04:48,800 --> 00:04:52,960 want, but there's no guarantee the the output is what you 73 00:04:52,960 --> 00:04:56,640 expect, right? So let's say for example, you 74 00:04:56,640 --> 00:05:00,800 can write like a table or a lease, but the generate HML 75 00:05:00,800 --> 00:05:05,600 might not be what you expect. And that's the problem I found 76 00:05:05,880 --> 00:05:09,120 with some specific details with earmark. 77 00:05:09,560 --> 00:05:13,920 And at that time I opened an issue for the project. 78 00:05:13,960 --> 00:05:21,880 I was willing to contribute back upstream fix earmark, but the so 79 00:05:22,200 --> 00:05:29,840 the maintainer at that time he, you know, he just, he just 80 00:05:29,840 --> 00:05:33,400 realized the the earmark he could not change the earmark 81 00:05:33,400 --> 00:05:35,600 design to make that implementation. 82 00:05:35,600 --> 00:05:37,720 So I was stuck with this problem. 83 00:05:38,440 --> 00:05:44,920 I need earmark solution markdown solution to migrate all the blog 84 00:05:44,920 --> 00:05:50,400 posts the dockyard, but I was not finding a mixer solution 85 00:05:50,880 --> 00:05:55,240 right. So coming back from Elixir Conf 86 00:05:55,840 --> 00:06:02,000 in Florida, I was, I was stuck at the airport because of a 87 00:06:02,120 --> 00:06:06,680 thunderstorm. So there was no, no timeline for 88 00:06:06,680 --> 00:06:09,760 my, my flight. And you know, I had nothing to 89 00:06:09,760 --> 00:06:14,160 do essentially. So that's when I, you know, I 90 00:06:14,160 --> 00:06:18,720 opened the, the rust book. I started reading about Rust and 91 00:06:18,720 --> 00:06:22,040 decide to create a new proof of proof of concept. 92 00:06:23,440 --> 00:06:27,520 You know just a wrapper of the Conrac library. 93 00:06:27,720 --> 00:06:30,880 Conrac is Rust is the Rust parser that does all the 94 00:06:30,880 --> 00:06:33,800 markdown parsing and rendering for MDX. 95 00:06:34,680 --> 00:06:38,480 So that's when I start MDX. It was pretty much a proof of 96 00:06:38,480 --> 00:06:42,880 concept just to see like if you can use Conrac, a well 97 00:06:42,880 --> 00:06:47,960 established markdown parser for Rust, if you can use that in 98 00:06:47,960 --> 00:06:50,800 Elixir. So I was there at the airport 99 00:06:50,800 --> 00:06:53,680 just reading the books and, you know, writing some code. 100 00:06:53,680 --> 00:06:56,040 And that's how MDX started essentially. 101 00:06:56,840 --> 00:07:01,640 And yeah, so I use, I use most of my free time at that time to 102 00:07:01,640 --> 00:07:06,640 work on a proof of concept. And then eventually when I, you 103 00:07:06,640 --> 00:07:11,400 know, proved that was working, I started using MDX on Beacon. 104 00:07:11,720 --> 00:07:15,360 And that's how Beacon came to be, came to life. 105 00:07:15,960 --> 00:07:20,040 And I'm also grateful for Dockyard because I was also 106 00:07:20,040 --> 00:07:23,800 working on MDX on my work time sometimes. 107 00:07:23,800 --> 00:07:30,320 So that's that was let's say a sponsorship of Dockyard on the 108 00:07:30,320 --> 00:07:33,560 project as well. Yeah, that's really awesome. 109 00:07:34,000 --> 00:07:37,720 So I'm curious, are you just, are you basically just using a 110 00:07:37,720 --> 00:07:42,240 NIF then with with Comrack or how's that working? 111 00:07:43,040 --> 00:07:48,720 Yeah, it starts just like a niche, just a rocker essentially 112 00:07:49,160 --> 00:07:54,440 just passing the options from Elixir to Korak and, you know, 113 00:07:54,720 --> 00:07:59,000 return the HML. So it starts like this, very, 114 00:07:59,000 --> 00:08:03,560 very simple at that time. Now the project has grown to 115 00:08:04,200 --> 00:08:09,120 have way more features like plug insurance and streaming and heat 116 00:08:09,120 --> 00:08:13,440 support and more, more features that Correct does not have. 117 00:08:14,280 --> 00:08:17,720 It's just a Linuxer MBX solution. 118 00:08:19,000 --> 00:08:21,000 OK, cool. But but it's still using. 119 00:08:21,000 --> 00:08:24,640 Correct. OK, What's like, what's the most 120 00:08:24,640 --> 00:08:29,440 difficult thing you've ran into right now building MBX? 121 00:08:31,280 --> 00:08:35,960 Probably in terms of implementation, I'd say 122 00:08:35,960 --> 00:08:41,480 streaming, but that's that's the that's not the only difficult 123 00:08:41,480 --> 00:08:45,480 when you're working on an open source project right there is 124 00:08:46,480 --> 00:08:51,520 like maybe people don't realize how hard it is to work on open 125 00:08:51,520 --> 00:08:53,760 source project. And even though index is not 126 00:08:53,760 --> 00:08:58,720 super huge library, but you know, maintaining a project is 127 00:08:58,720 --> 00:09:02,400 really hard. But in terms of implementation 128 00:09:02,400 --> 00:09:06,960 now you'd say streaming. Streaming is streaming is the 129 00:09:06,960 --> 00:09:12,960 feature that allows you to heal or complete the fragments of 130 00:09:12,960 --> 00:09:16,520 markdown. And that's very useful for AI 131 00:09:16,960 --> 00:09:19,960 like applications where you have the tokens. 132 00:09:20,800 --> 00:09:23,880 So the the AI is feeding some tokens to you. 133 00:09:24,440 --> 00:09:28,400 Those tokens are incomplete fragments of markdown and then 134 00:09:28,400 --> 00:09:32,280 you have to render that, you know, those fragments. 135 00:09:33,040 --> 00:09:37,040 So that's really, really hard to do because you have many edge 136 00:09:37,040 --> 00:09:41,560 cases that you have to solve. So for example, let's say you 137 00:09:41,560 --> 00:09:46,840 have like a cold block or a table or you know, blocks of 138 00:09:46,840 --> 00:09:50,640 markdown. You have to complete those 139 00:09:50,640 --> 00:09:55,480 blocks and then you've parse and you have to parse and render 140 00:09:55,480 --> 00:09:58,800 those fragments. And there are many ways to do 141 00:09:58,800 --> 00:10:02,200 this. Probably the most popular 142 00:10:02,520 --> 00:10:05,920 popular library now is streamed down by Versailles. 143 00:10:06,680 --> 00:10:12,080 They they have like this kind of library library for Markdown as 144 00:10:12,080 --> 00:10:15,080 well. And there are some others 145 00:10:15,280 --> 00:10:19,440 implementing a different way creating new parsers for this 146 00:10:19,440 --> 00:10:24,000 specific use case. But for MDX it does not make 147 00:10:24,000 --> 00:10:27,560 sense to create a new parser because we are using call Rack 148 00:10:27,760 --> 00:10:32,840 and it's it's not an easy task to create a new markdown parser. 149 00:10:33,800 --> 00:10:40,000 So the way it works essentially is as we get new fragments, we 150 00:10:40,000 --> 00:10:46,680 keep placing those fragments into the stack and using new 151 00:10:46,680 --> 00:10:51,640 ball parsec to parse those fragments and then try to figure 152 00:10:51,640 --> 00:10:55,800 out which element we are working with. 153 00:10:55,880 --> 00:11:00,240 Like a table table or a paragraph or a lease. 154 00:11:00,360 --> 00:11:03,560 You know, what kind of element we have that fragment. 155 00:11:04,000 --> 00:11:09,320 And then we have to complete that fragment like a link or a 156 00:11:09,360 --> 00:11:12,280 table. For example, we've let's say we 157 00:11:12,280 --> 00:11:16,600 have just the heater of the table and just the first column 158 00:11:16,920 --> 00:11:19,480 of the table. Let's say we have those tokens 159 00:11:19,520 --> 00:11:23,080 in the stack. Then we have to complete the 160 00:11:23,080 --> 00:11:27,280 rest of the table in order to render that table. 161 00:11:28,000 --> 00:11:32,720 So so yeah, streaming is. I'm curious with, with 162 00:11:32,720 --> 00:11:36,080 streaming, you might not always have like you might not know 163 00:11:36,080 --> 00:11:37,640 that the table's finished yet, right? 164 00:11:37,640 --> 00:11:39,520 Because you can't control what's coming through. 165 00:11:39,840 --> 00:11:44,600 Or is that interesting and so you just kind of assume it's 166 00:11:44,600 --> 00:11:51,120 finished and you have to end the table and then as more data 167 00:11:51,120 --> 00:11:54,240 streams in you, you build on it? Or how does that work? 168 00:11:54,840 --> 00:11:58,600 Yeah, that that's pretty much how you, you know how that's 169 00:11:58,600 --> 00:12:01,920 that's the problem we have because most of the parsers, the 170 00:12:01,920 --> 00:12:05,480 markdown parsers, they need the full document, the full 171 00:12:05,480 --> 00:12:08,760 markdown. So they only they only work with 172 00:12:08,760 --> 00:12:12,720 the full document, not with fragments of or pieces of 173 00:12:12,720 --> 00:12:16,840 markdown. So as we receive fragments we 174 00:12:16,840 --> 00:12:21,320 need we need to complete them. So for example string DAO is 175 00:12:21,320 --> 00:12:28,000 using regex, so they just try to find the element by regex. 176 00:12:28,360 --> 00:12:30,480 So let's say there is a asterisk. 177 00:12:31,080 --> 00:12:36,280 So you you know, that's, you know, you know, if that's like a 178 00:12:36,280 --> 00:12:39,680 bold statement or something like that. 179 00:12:39,680 --> 00:12:45,080 And then you have to place the the paint off the fragment so 180 00:12:45,080 --> 00:12:49,520 you can render that. So let's say you receive like a 181 00:12:49,520 --> 00:12:54,520 dash for a lease, right? And then you have to render that 182 00:12:54,600 --> 00:12:56,920 as a lease. So you have to complete that 183 00:12:56,920 --> 00:13:01,800 element as a lease. And as you receive more of that, 184 00:13:02,360 --> 00:13:05,760 more of those elements, you keep, you know, you, you keep 185 00:13:05,760 --> 00:13:10,760 just feeding the, the streaming parser and, and just like like a 186 00:13:10,760 --> 00:13:13,040 buffer. And then eventually you just 187 00:13:13,040 --> 00:13:18,640 flush and render. OK, Yeah, how do you, how do you 188 00:13:18,640 --> 00:13:23,600 keep track of syntax? Like you're obviously what's the 189 00:13:23,600 --> 00:13:28,600 word I'm looking for. You're you're basically like 190 00:13:28,920 --> 00:13:31,440 optimistically completing things, right? 191 00:13:31,720 --> 00:13:34,760 So how do you know when it's like you can't just like insert 192 00:13:34,760 --> 00:13:36,320 that optimistic completion, right? 193 00:13:36,320 --> 00:13:39,800 Because you have to know it it was you that added it as more 194 00:13:39,800 --> 00:13:43,520 data streams in Or am I over complicating things? 195 00:13:44,200 --> 00:13:49,040 No, that that's, you know, some, some, some implementations they 196 00:13:49,720 --> 00:13:54,880 that's that's how index works. So let's think about an example 197 00:13:54,880 --> 00:13:56,440 here. Let's say you have a code. 198 00:13:56,720 --> 00:14:01,400 Code block is probably the most, you know the most the better 199 00:14:01,400 --> 00:14:06,240 example here, let's say you you start code block, you know, you 200 00:14:06,240 --> 00:14:11,080 have the back dates and you have like the Elixir info streaming. 201 00:14:11,080 --> 00:14:15,240 And then you start streaming some code like the first line, 202 00:14:15,240 --> 00:14:22,440 the second line and so on. So when when MDMDX receives a 203 00:14:22,440 --> 00:14:28,520 new fragment, it just plays the the block and the end of the 204 00:14:28,520 --> 00:14:35,320 block to close the decode block. And so as it it's get as it gets 205 00:14:35,320 --> 00:14:39,680 more and more code lines, it just completes the end of the 206 00:14:39,680 --> 00:14:44,120 block and render that piece. So I think eventually it's going 207 00:14:44,120 --> 00:14:46,040 to receive the the end of the block. 208 00:14:46,400 --> 00:14:51,440 So it just discard that's temporary block and then 209 00:14:51,440 --> 00:14:55,800 continue. But the only thing that MDX does 210 00:14:55,800 --> 00:15:00,920 not do now is reverting a decision. 211 00:15:01,200 --> 00:15:04,880 So let's say you receive a piece of markdown and for some reason 212 00:15:04,880 --> 00:15:08,600 they streaming parts of things that leak that's a lease or a 213 00:15:08,600 --> 00:15:11,840 code block or whatever. And then eventually it realized 214 00:15:11,840 --> 00:15:16,080 that was not. OK, so it can't go back and 215 00:15:16,080 --> 00:15:20,120 correct its mistake. No, not right now, because 216 00:15:20,120 --> 00:15:23,760 that's really, really complex to do, because you have to keep not 217 00:15:23,760 --> 00:15:27,920 only the future like fragments, but have to keep the stack of 218 00:15:27,920 --> 00:15:31,240 the whole document. And it can go back to the 219 00:15:31,240 --> 00:15:33,280 beginning of the document, right? 220 00:15:33,280 --> 00:15:39,120 Because Markdown you have blocks and inline elements and then you 221 00:15:39,120 --> 00:15:43,120 have nested elements. So you have to go back to the 222 00:15:43,120 --> 00:15:48,120 root of the decision. So that's really hard to do and 223 00:15:48,120 --> 00:15:51,560 that's something that is not worth doing right now because 224 00:15:51,560 --> 00:15:54,760 I'm curious can do that. Yeah, I'm curious, do you have 225 00:15:54,760 --> 00:15:58,400 like a solution mapped out in your head or not? 226 00:15:58,400 --> 00:16:01,800 Is it just too complicated or you haven't spent much time on 227 00:16:01,800 --> 00:16:05,760 it? No I I had some ideas but it 228 00:16:05,760 --> 00:16:09,600 gets so complex it's not worth it. 229 00:16:09,600 --> 00:16:12,640 Well, and like you, performance becomes a problem, right? 230 00:16:12,640 --> 00:16:16,080 Like everyone expects like a markdown to be rendered 231 00:16:16,400 --> 00:16:19,320 relatively fast, right? Especially when you're streaming 232 00:16:19,320 --> 00:16:22,160 the data and that should be almost instant. 233 00:16:23,480 --> 00:16:27,360 Yeah, it also depends on the room time where you are calling 234 00:16:27,360 --> 00:16:30,360 MDX. If you have a live view where 235 00:16:30,360 --> 00:16:35,160 you can def the, you know, you can define the content, then 236 00:16:35,160 --> 00:16:38,920 that helps a lot, right? Because as you get more 237 00:16:39,400 --> 00:16:44,840 fragments and then you can just def the the reindeer to a live 238 00:16:44,840 --> 00:16:49,800 view page, you can just leverage the define algorithm of live 239 00:16:49,800 --> 00:16:52,680 view. So it it it really depends where 240 00:16:52,720 --> 00:16:55,280 we you are streaming that markdown. 241 00:16:56,320 --> 00:17:01,040 I'm curious like once you have like all the data like you know 242 00:17:01,040 --> 00:17:04,839 well at least like a portion and you realize like you know some 243 00:17:04,839 --> 00:17:08,960 syntax was assumed incorrectly, couldn't you just re render 244 00:17:08,960 --> 00:17:12,599 everything you have in memory at that point or it does it not 245 00:17:12,599 --> 00:17:17,319 really work that? You can, you can, but there's 246 00:17:17,319 --> 00:17:20,040 going to be like a flag on the page, right? 247 00:17:20,040 --> 00:17:24,880 Let's say you had like a list and then you just re render the 248 00:17:24,880 --> 00:17:28,079 whole document. If it changed the elements, 249 00:17:28,079 --> 00:17:31,000 you're going to see some changes on the page, like you're going 250 00:17:31,000 --> 00:17:36,200 to see like the page changing. It's not a big deal, but it's 251 00:17:36,200 --> 00:17:40,440 possible to just, you know, the documents finish, try to render 252 00:17:40,440 --> 00:17:44,640 the whole document you know and make and fix any mistakes. 253 00:17:45,240 --> 00:17:51,440 But the idea is to reduce the mistakes as much as possible 254 00:17:51,440 --> 00:17:56,880 that you don't need to do that. There was there was a 255 00:17:56,880 --> 00:18:02,880 contribution recently, someone sent a fix for the streaming 256 00:18:02,880 --> 00:18:09,080 parser adding more patterns because we are using Invoke 257 00:18:09,080 --> 00:18:11,440 Parsec. So that's how we identify the 258 00:18:11,440 --> 00:18:15,840 patterns and someone sent a fix to improve it. 259 00:18:16,600 --> 00:18:19,080 So I think it's pretty stable right now. 260 00:18:20,240 --> 00:18:27,720 You know, some tools are using, we have Ashy Jido and others 261 00:18:27,960 --> 00:18:32,880 libraries using MDXI. Don't know who is using the 262 00:18:32,880 --> 00:18:38,480 streaming feature to be honest, but I know some, some, some 263 00:18:38,480 --> 00:18:40,880 companies, some tools, some people are using this streaming 264 00:18:40,880 --> 00:18:43,760 feature now. So it it should be stable 265 00:18:43,760 --> 00:18:45,360 enough. That's awesome. 266 00:18:46,000 --> 00:18:48,680 I won't be probably using the streaming feature, but I am 267 00:18:48,680 --> 00:18:52,720 going to check it out and implement MDX into kill switch. 268 00:18:52,720 --> 00:18:57,200 I think it'll be cool. Yeah, just like if you do just, 269 00:18:57,480 --> 00:19:02,200 you know, just pay attention to the to the render HML, something 270 00:19:02,200 --> 00:19:06,560 that is does not look right, just open an issue and then we 271 00:19:06,560 --> 00:19:09,960 can like we can work on that because there are many, many, 272 00:19:09,960 --> 00:19:15,080 many edge cases with streaming because you you don't control 273 00:19:15,080 --> 00:19:19,760 the the fragments, right. And you know, it can be like a 274 00:19:19,880 --> 00:19:22,080 whole lease or just a piece of that lease. 275 00:19:22,080 --> 00:19:26,400 And that lease can be like a task or no task. 276 00:19:26,480 --> 00:19:31,280 And then you don't know if you have to render a text or a task 277 00:19:31,680 --> 00:19:34,520 like like those tasks you have in GitHub. 278 00:19:35,400 --> 00:19:40,120 So it's really hard to, you know, to get all the elements 279 00:19:40,120 --> 00:19:43,760 correct. But you know, it's, it's kind 280 00:19:43,760 --> 00:19:46,080 of, it should be stable now, yeah. 281 00:19:46,880 --> 00:19:48,400 I like that. How long have you been working 282 00:19:48,400 --> 00:19:54,600 on MDX? On MDX let me check the first 283 00:19:55,320 --> 00:19:59,880 version was September 2023. OK. 284 00:20:01,760 --> 00:20:06,000 It was right after Elixir conf. It was just the first version 285 00:20:06,000 --> 00:20:09,440 like the the proof of concept was working and then I released 286 00:20:09,440 --> 00:20:12,400 the first the first. It all stemmed from you getting 287 00:20:12,400 --> 00:20:16,600 bored at an airport. Yes, it's like that, Yeah. 288 00:20:17,360 --> 00:20:22,480 I had no idea how wrestler worked. 289 00:20:22,480 --> 00:20:26,440 Like I knew about the wrestler but never had used it in a 290 00:20:26,440 --> 00:20:29,480 project before, so I started just learning there and, you 291 00:20:29,480 --> 00:20:34,240 know, and that's how it started. Yeah I have never used rustler 292 00:20:34,240 --> 00:20:38,840 or nifs and like in my own project so I'm aware of them. 293 00:20:38,960 --> 00:20:41,240 I just have I have 0 experience with them. 294 00:20:41,600 --> 00:20:44,240 What what was the learning curve like do you think? 295 00:20:44,240 --> 00:20:47,800 Is rustler pretty easy or was it complicated? 296 00:20:48,960 --> 00:20:53,280 I to be honest, I think Rustler deserves more attention from the 297 00:20:53,280 --> 00:20:56,320 community because a really, really good piece of software, 298 00:20:56,960 --> 00:21:00,240 it solves a lot of problems because sometimes you don't have 299 00:21:00,240 --> 00:21:03,480 a library in the Elixir community, so you can just 300 00:21:03,480 --> 00:21:06,760 leverage all their ecosystems, right? 301 00:21:06,800 --> 00:21:11,960 And Rus has Rus has a lot of libraries for a lot. 302 00:21:12,360 --> 00:21:13,360 Of different. Things. 303 00:21:13,720 --> 00:21:17,360 If you're doing heavy computing or anything, you're better off 304 00:21:17,360 --> 00:21:20,800 using Rust instead of Elixir too, right? 305 00:21:20,800 --> 00:21:24,600 Just for speed of processing. It depends. 306 00:21:25,200 --> 00:21:29,200 It depends what you if you need to exchange data, there's going 307 00:21:29,200 --> 00:21:34,240 to be overhead between Elixir and Rust. 308 00:21:34,880 --> 00:21:38,480 If you can, if you can share the same space of memory like some 309 00:21:38,480 --> 00:21:47,120 projects do, then it's it's OK. But if you need to exchange data 310 00:21:47,120 --> 00:21:51,680 and see realize the data, then it's going to have some 311 00:21:51,720 --> 00:21:53,520 overhead. But it really depends on the 312 00:21:53,520 --> 00:21:56,920 project. But The thing is, you can share 313 00:21:56,920 --> 00:21:59,640 the same space of memory and that's a really, really cool 314 00:21:59,640 --> 00:22:03,240 feature. Then it makes it way more. 315 00:22:03,240 --> 00:22:06,200 Performing. Yeah, I can see that. 316 00:22:06,240 --> 00:22:09,680 That's cool. Maybe someday I'll like come up 317 00:22:09,680 --> 00:22:13,280 with an excuse to try Rustler and put in a project. 318 00:22:14,440 --> 00:22:17,040 Yeah, it's really nice, really nice project. 319 00:22:19,120 --> 00:22:26,240 So now like I feel like MDX and then it's Loomis, right? 320 00:22:26,240 --> 00:22:30,800 I'm saying that correctly. Yes, I I I think so. 321 00:22:30,880 --> 00:22:33,400 I to be honest, I renamed the project. 322 00:22:33,400 --> 00:22:37,280 It used to be called another name which I don't know how to 323 00:22:37,280 --> 00:22:40,840 pronounce. So that's the reason I I renamed 324 00:22:40,840 --> 00:22:44,480 to Loomis and I just say Loomis, but you know. 325 00:22:45,400 --> 00:22:47,480 Nice. Now I feel like I don't know. 326 00:22:48,520 --> 00:22:53,040 I would always kind of assume the like Loomis would just be 327 00:22:53,040 --> 00:22:57,520 integrated into a markdown, but there's probably a lot more 328 00:22:57,920 --> 00:23:02,680 complicated aspects to syntax highlighting for code blocks. 329 00:23:02,680 --> 00:23:08,560 Yeah. Yeah, Loomis also start like to 330 00:23:08,560 --> 00:23:13,560 solve a problem on the Dockyard blog post because we had to, you 331 00:23:13,560 --> 00:23:16,640 know, code. We have to have synthetic 332 00:23:16,640 --> 00:23:21,200 highlighting for different types of languages and code because 333 00:23:21,200 --> 00:23:25,840 the Dockyard blog, the blog has, you know, Ruby, JavaScript, 334 00:23:26,040 --> 00:23:31,880 Swift and a lot of other stuff. And in the mixer we have the 335 00:23:31,880 --> 00:23:37,760 makeup library, you know, create by just evaline and, and then it 336 00:23:37,920 --> 00:23:42,360 didn't have all the parsers that I needed for this, this problem. 337 00:23:43,080 --> 00:23:50,120 So then I, I started researching the options I had and I had two 338 00:23:50,120 --> 00:23:55,720 options essentially, either I, I had to create new parsers for 339 00:23:55,720 --> 00:23:58,680 makeup and I had to maintain those parsers. 340 00:23:58,680 --> 00:24:01,640 And it's not easy to create those parsers either. 341 00:24:02,600 --> 00:24:07,680 Or I could just leverage 3 seater. 3 seater is being used 342 00:24:07,680 --> 00:24:12,080 by a lot of editors now. It's just a parser for. 343 00:24:12,640 --> 00:24:15,360 It's essentially A parser for languages. 344 00:24:15,360 --> 00:24:17,920 It's defines the grammar for languages. 345 00:24:18,920 --> 00:24:22,480 And then one of the modules is the syntax highlight because it 346 00:24:22,480 --> 00:24:25,600 identifies the tokens on the search code and then you can 347 00:24:25,600 --> 00:24:29,720 apply like colors on those tokens. 348 00:24:30,760 --> 00:24:33,880 So that those are the two options I had. 349 00:24:34,360 --> 00:24:40,600 And since I was already working with Rust in the MDX, I decided 350 00:24:40,600 --> 00:24:45,840 to just give a try for, you know, for these, for three 351 00:24:45,840 --> 00:24:49,920 seater, I was using a library called Inkyjet. 352 00:24:50,320 --> 00:24:54,520 Back back then he was also Rustler, pretty much the same 353 00:24:54,520 --> 00:24:57,280 idea. And inkjet. 354 00:24:57,280 --> 00:25:02,160 The Internet used to work with the Helix editor, right? 355 00:25:02,160 --> 00:25:05,800 So we had the three pseudo grammars on one side and the 356 00:25:05,800 --> 00:25:09,520 theme, the color schemes from the Helix editor. 357 00:25:10,480 --> 00:25:15,760 The problem though, is, was that the colors were not matching 358 00:25:16,480 --> 00:25:21,880 because many themes from the Helix editor was, they were like 359 00:25:22,240 --> 00:25:27,440 outdated or incomplete. So it was not working very well 360 00:25:27,440 --> 00:25:32,200 for my specific use case. And I was contributing to that 361 00:25:32,200 --> 00:25:37,120 project, the inkjet project. But it came to a point where I 362 00:25:37,120 --> 00:25:41,680 decided to create my own project so I could have the three serial 363 00:25:41,680 --> 00:25:47,800 grammars and the Neofine themes. Because in Neofine you have like 364 00:25:48,120 --> 00:25:52,200 all the themes, all the color schemes are very well maintained 365 00:25:52,200 --> 00:25:57,480 and we have a lot of options. So that's when the synthetic 366 00:25:57,480 --> 00:26:02,360 highlighter is started. But it start as like internal 367 00:26:02,360 --> 00:26:07,880 code inside MDX and then it start growing to the point where 368 00:26:07,880 --> 00:26:13,560 I create the, the Alton. It used to be called Alton in 369 00:26:13,560 --> 00:26:16,600 Elixir. It was another, I separate the 370 00:26:16,640 --> 00:26:20,320 two projects, so MDX and Alton back in the day. 371 00:26:21,480 --> 00:26:25,480 And then eventually I realized the Alton project was more worse 372 00:26:25,480 --> 00:26:30,040 than Elixer because it was because there is, there is no 373 00:26:30,240 --> 00:26:34,080 bindings for three seater for three Elixer, no official 374 00:26:34,080 --> 00:26:35,680 bindings. It's just Rust. 375 00:26:37,440 --> 00:26:41,720 I mean it's Rust in some other like languages, but no, no 376 00:26:41,720 --> 00:26:44,400 Elixer. So the project was pretty much 377 00:26:44,400 --> 00:26:49,120 just Rust, right? And then I realized that I had 378 00:26:49,120 --> 00:26:52,440 to it was better to just release a Russ library. 379 00:26:53,560 --> 00:26:58,360 The the problem was the problem though, is like there wasn't the 380 00:26:58,560 --> 00:27:03,920 autumn name was not available to publish as a crate, as a Russ 381 00:27:03,920 --> 00:27:07,160 library. And the best name I could I 382 00:27:07,160 --> 00:27:11,760 could, I could find was Autonomous, which is Latin, 383 00:27:11,760 --> 00:27:16,320 Latin for autumn. It's not working very well 384 00:27:16,320 --> 00:27:18,160 because you know, it's hard to pronounce. 385 00:27:18,520 --> 00:27:20,760 People don't understand it, including me. 386 00:27:22,400 --> 00:27:24,800 So eventually I just renamed Loomis. 387 00:27:24,920 --> 00:27:29,200 And now we have Loomis for lexer, Loomis for Rust. 388 00:27:29,200 --> 00:27:33,360 We have Acli, the the maintainer of. 389 00:27:34,280 --> 00:27:37,280 I don't know how to pronounce this correctly, but Shikari or 390 00:27:37,280 --> 00:27:41,080 something like that is the the was me runtime for Java. 391 00:27:41,680 --> 00:27:46,800 He also ported movies for for Java, so we also have a Java 392 00:27:46,800 --> 00:27:48,480 library now. I was. 393 00:27:48,600 --> 00:27:52,480 Going to So are you supporting all these multiple run times or 394 00:27:53,280 --> 00:27:56,280 do you just let other people support? 395 00:27:56,320 --> 00:28:01,560 Well, I'm working on the JS runtime or JS version, so yes. 396 00:28:02,720 --> 00:28:06,320 OK, so yeah. So the core engine is just Rust 397 00:28:06,600 --> 00:28:11,720 and CLI is also Rust. Elixir is a Rustler rustler 398 00:28:11,920 --> 00:28:15,120 wrapper, so it's not that much code right now. 399 00:28:17,120 --> 00:28:23,360 JS in TypeScript on the other hand is the most complex one I'd 400 00:28:23,360 --> 00:28:29,720 say because the biggest problem is with three seater is all the 401 00:28:29,720 --> 00:28:34,520 languages that you have. They are defined the gram, the 402 00:28:34,520 --> 00:28:39,880 grammar is defined like a like a Jason JavaScript file and then 403 00:28:39,880 --> 00:28:44,880 it compiles to AC and transpires to C and then you compile that. 404 00:28:46,200 --> 00:28:51,320 The problem is a parser for a Rus or Elixir. 405 00:28:51,320 --> 00:28:55,120 They have like 1 megabyte or two maybe. 406 00:28:55,640 --> 00:29:00,400 So if you include all the languages like let's say 100 407 00:29:00,400 --> 00:29:05,920 different languages, each one consuming like 1 megabyte or 408 00:29:05,920 --> 00:29:10,440 two, it does not work. Like you're not going to 409 00:29:10,440 --> 00:29:15,920 download 200 megabytes for just a Syntex highlight library. 410 00:29:16,920 --> 00:29:19,920 What you don't want just bloat in your project. 411 00:29:21,160 --> 00:29:23,920 Yeah, so that's the biggest problem right now in. 412 00:29:24,360 --> 00:29:26,840 Keeping projects small? Yeah, that makes sense. 413 00:29:28,400 --> 00:29:32,240 So the way we I'm trying to solve these now is using OASM. 414 00:29:33,080 --> 00:29:39,640 SO3 seater has a target. You can compile the parsers into 415 00:29:39,640 --> 00:29:46,640 OASM modules and then the JavaScript libraries can just 416 00:29:46,640 --> 00:29:50,520 load those modules, those languages dynamically at 417 00:29:50,520 --> 00:29:54,960 runtime. So the core is to TypeScript, 418 00:29:55,360 --> 00:29:58,920 but the languages, the parsers are loaded dynamically. 419 00:29:59,840 --> 00:30:01,800 So that's a way to solve this problem. 420 00:30:01,840 --> 00:30:05,840 Although it's not perfect because there is an overhead to 421 00:30:06,080 --> 00:30:11,040 load those languages, but that's the best we can do right now, at 422 00:30:11,040 --> 00:30:15,000 least until you we have the the water components. 423 00:30:15,800 --> 00:30:19,400 Water components might solve this problem, but it's not 424 00:30:19,400 --> 00:30:21,320 something we can use right now. I. 425 00:30:23,880 --> 00:30:29,960 See I I don't know if I fully understand how wasm works. 426 00:30:33,520 --> 00:30:34,840 Can you explain that a little bit? 427 00:30:35,640 --> 00:30:38,640 Yeah, yeah, sure. So essentially you can think 428 00:30:38,640 --> 00:30:44,360 about WASM like having like for example, we have the Bing VM, 429 00:30:44,720 --> 00:30:46,920 right? And then you run some code into 430 00:30:47,240 --> 00:30:51,400 into the VM. WASM is a WASM requires a 431 00:30:51,400 --> 00:30:57,000 runtime as well. So, so in Ross you have wasn't 432 00:30:57,000 --> 00:30:59,880 time. In Elixir we have some as well. 433 00:31:00,280 --> 00:31:04,160 For Java there is this chikori chikori something like that. 434 00:31:04,400 --> 00:31:10,440 It's a runtime as well. For JS we also have some 435 00:31:10,480 --> 00:31:14,240 runtimes. So you you need to load at 436 00:31:14,240 --> 00:31:19,400 runtime and then you can load the wasn't modules into that 437 00:31:19,400 --> 00:31:23,640 runtime. And it works for the user and 438 00:31:23,800 --> 00:31:27,520 and also for the developers. It works like transparent, like 439 00:31:27,560 --> 00:31:30,800 it's transparent. You can just call the functions 440 00:31:30,800 --> 00:31:34,880 from the from the new OSM module. 441 00:31:35,720 --> 00:31:39,240 So if you create like a hello word function and then compile 442 00:31:39,240 --> 00:31:44,440 into a OSME module, you can call that very same function in any 443 00:31:44,440 --> 00:31:49,000 runtime in real S or JS or That's pretty cool. 444 00:31:49,000 --> 00:31:53,800 So it, yeah, yeah, Wasmi is really, really nice. 445 00:31:53,920 --> 00:31:58,680 I had never used it before for anything, like for real. 446 00:31:59,120 --> 00:32:02,320 It's the first time I'm using Rawsmi for something that, you 447 00:32:02,320 --> 00:32:05,520 know, I have to make it work correctly. 448 00:32:06,760 --> 00:32:10,640 It seems to be really a really good solution. 449 00:32:11,040 --> 00:32:13,280 I know. And some projects are using it 450 00:32:13,280 --> 00:32:18,600 now, For example, the popcorn. I, I know it's on the news now. 451 00:32:19,080 --> 00:32:25,680 Popcorn is just using OASM as well load Elixir on the browser. 452 00:32:26,800 --> 00:32:31,280 So yeah, it's, it's a really nice solution, but there are 453 00:32:31,280 --> 00:32:34,280 some limitations. It's not a, not a perfect 454 00:32:34,280 --> 00:32:41,240 solution, but my, my idea with this is having the same API, the 455 00:32:41,240 --> 00:32:45,240 same syntax highlight API for different platforms. 456 00:32:45,240 --> 00:32:50,720 So we have the same API for JS, same API for reals and elixirs. 457 00:32:50,720 --> 00:32:54,800 So you can have the same syndex highlighter, the same colors, 458 00:32:54,800 --> 00:33:01,240 the same languages parsers, all the ecosystem in different 459 00:33:01,240 --> 00:33:05,960 languages and including including the CLI. 460 00:33:06,440 --> 00:33:08,960 Yeah, that's really cool. And then it kind of work. 461 00:33:09,080 --> 00:33:12,880 It'll work the same way as you kind of change technologies and 462 00:33:12,880 --> 00:33:15,320 tech stacks too, right? You'd be pretty familiar with 463 00:33:15,320 --> 00:33:19,400 the library. Yeah, because right now you have 464 00:33:19,480 --> 00:33:23,000 really good libraries, but if you go to JavaScript you have to 465 00:33:23,000 --> 00:33:27,000 choose one, if you go to also have to choose another one and 466 00:33:27,000 --> 00:33:29,440 so on. So sometimes you have a back 467 00:33:29,440 --> 00:33:33,560 end, a front end and you have to choose different Syndex 468 00:33:33,560 --> 00:33:36,240 highlighter libraries, including the CLI. 469 00:33:36,760 --> 00:33:41,320 You know all those coding agents now have a Syntex highlighter 470 00:33:42,160 --> 00:33:46,480 and and actually tree seater works really well for it. 471 00:33:46,960 --> 00:33:52,080 You know AI streaming like feature shutdown interfaces 472 00:33:52,080 --> 00:33:56,560 because Tracer. It works really well with 473 00:33:56,800 --> 00:33:59,440 incomplete code. Yeah, that's cool. 474 00:34:00,240 --> 00:34:02,320 The same problem we had with Mark now. 475 00:34:03,160 --> 00:34:04,160 Interesting. All right. 476 00:34:04,240 --> 00:34:07,680 I'm going to put this on hold for a second and we'll be right 477 00:34:07,680 --> 00:34:11,080 back in like, I don't know. 30 seconds. 478 00:34:11,080 --> 00:34:14,560 Hey, I'm Kimberly with Elixir mentor Jacob recently added 479 00:34:14,560 --> 00:34:17,880 recruitment services to help companies hire pre vetted Elixir 480 00:34:17,880 --> 00:34:20,880 devs. I help with company outreach, so 481 00:34:20,880 --> 00:34:24,159 if you're hiring, I may be reaching out, or you can connect 482 00:34:24,159 --> 00:34:26,360 with us anytime at elixirmentor.com. 483 00:34:28,560 --> 00:34:33,480 And we're back, all right. So I'm curious between like 484 00:34:33,480 --> 00:34:38,880 having two, like you're kind of the solo person in charge of MDX 485 00:34:38,880 --> 00:34:42,639 and Loomis, right? How, how do you balance the 486 00:34:42,639 --> 00:34:45,000 tube? Just having two open source 487 00:34:45,000 --> 00:34:46,600 projects? I feel like that's a big 488 00:34:46,600 --> 00:34:54,960 undertaking. I mean, you have to choose the 489 00:34:54,960 --> 00:35:00,720 most important features. And one thing I learned I was 490 00:35:00,720 --> 00:35:07,720 talking to you Zach from Ash a while ago and one thing he said 491 00:35:07,720 --> 00:35:11,480 to me that it's really important for open source project is to be 492 00:35:11,840 --> 00:35:18,880 like should be as responsive as as possible people open issue. 493 00:35:18,960 --> 00:35:21,760 Zach is the most responsive person I've ever met. 494 00:35:21,760 --> 00:35:23,400 That's. Impressive. 495 00:35:24,640 --> 00:35:27,520 He and Valene, they are like AI don't know. 496 00:35:27,760 --> 00:35:30,240 I think they are some kind of botch because. 497 00:35:30,240 --> 00:35:31,400 I know. I don't think they're real 498 00:35:31,400 --> 00:35:34,880 people. Yeah, they can just spawn up new 499 00:35:34,880 --> 00:35:37,400 versions of themselves and they're just everywhere. 500 00:35:38,120 --> 00:35:43,240 Yeah, it's, you know, it's unbelievable, but the idea is to 501 00:35:43,240 --> 00:35:46,840 be as responsible as possible. If you have a someone open an 502 00:35:46,840 --> 00:35:49,200 issue. If you don't have a solution, at 503 00:35:49,200 --> 00:35:52,560 least just reply and say, hey, I'm going to work on this, it's 504 00:35:52,560 --> 00:35:54,480 on the backlog. I have some ideas. 505 00:35:55,400 --> 00:35:58,920 And then you just kind of, you know, gives. 506 00:35:59,640 --> 00:36:03,800 So give something to the person. Sometimes that person has some 507 00:36:03,800 --> 00:36:06,880 idea, he's just not sure what to do. 508 00:36:06,960 --> 00:36:10,840 And that's how you also get contributions back, right? 509 00:36:11,640 --> 00:36:14,320 So the idea should be as responsible as possible. 510 00:36:14,320 --> 00:36:20,280 That's the first rule. And you know, using AI nowadays 511 00:36:20,280 --> 00:36:27,560 is, is a way to help with the backlog, even though like vibe 512 00:36:27,760 --> 00:36:31,640 coding is not a solution. You know, you have to be very 513 00:36:32,080 --> 00:36:34,240 careful. Yeah. 514 00:36:34,240 --> 00:36:37,880 Have you seen any like nasty messy contribution attempts? 515 00:36:37,880 --> 00:36:43,280 Like does it increase your PRS from contributors or not really? 516 00:36:44,040 --> 00:36:47,240 No, not really. MDX and looms are not products 517 00:36:47,240 --> 00:36:51,360 that are receiving a lot of PRS, I would say. 518 00:36:53,080 --> 00:36:57,400 But it's interesting because we are getting, we are getting more 519 00:36:57,400 --> 00:37:00,560 contributions. The last version, the last MDX 520 00:37:00,560 --> 00:37:05,920 version was was pretty much just external contributors. 521 00:37:07,320 --> 00:37:12,160 So that's really cool when you have people contributing, right. 522 00:37:12,360 --> 00:37:17,000 People are not only using the project, but also helping to 523 00:37:17,000 --> 00:37:21,640 improve it. And I, I saw yesterday on the 524 00:37:21,640 --> 00:37:26,800 leak surf forum the hope to be pronounced his name correctly. 525 00:37:26,840 --> 00:37:30,720 Austin Ziegler. He's building some plug 526 00:37:30,720 --> 00:37:37,400 insurance for for MDX. So it's it's yeah, it's really 527 00:37:37,400 --> 00:37:40,400 nice because, you know, the plug in system. 528 00:37:40,920 --> 00:37:48,120 I was talking to you why tech Mac like 2 years, two years ago, 529 00:37:48,120 --> 00:37:52,800 because the plug in system is based on the rack system, plug 530 00:37:52,800 --> 00:37:55,000 in system. It's pretty much the same idea. 531 00:37:55,640 --> 00:37:59,520 And he was really, you know, cool to, you know, just, you 532 00:37:59,520 --> 00:38:01,920 know, use it. And he gave some ideas, some 533 00:38:01,920 --> 00:38:08,840 feedback as well for MDX. And now you can build plug 534 00:38:08,840 --> 00:38:13,600 insurance. I built, I built the JFM, the 535 00:38:13,600 --> 00:38:22,000 good Hub, McDonald, I built the Mermaid as well and a few 536 00:38:22,000 --> 00:38:25,200 others. And now people are also building 537 00:38:25,440 --> 00:38:29,240 plug insurance, right? So it's not just the MDX like we 538 00:38:29,240 --> 00:38:36,240 have MDX, the plug insurance Loomis and also Conrac, Conrac 539 00:38:36,240 --> 00:38:39,960 which you know the upstream markdown parser. 540 00:38:40,880 --> 00:38:45,160 I am also contributor to Conrac because I need to make changes 541 00:38:45,320 --> 00:38:52,000 on CONRAC first to support, you know features downstream in MDX. 542 00:38:52,880 --> 00:38:58,360 The the most impactful contribution on CONRAC from my 543 00:38:58,360 --> 00:39:04,480 end was the Higgs parser. So the markdown can work with 544 00:39:04,560 --> 00:39:07,720 hex components and Elixir expressions. 545 00:39:09,120 --> 00:39:13,920 So, but yeah, back to your to your question. 546 00:39:13,920 --> 00:39:19,800 Essentially you have to be very pragmatic like you have an 547 00:39:19,800 --> 00:39:21,960 issue, you have APR, you have to take a decision. 548 00:39:21,960 --> 00:39:26,800 Don't, don't, don't let it file. Otherwise it'd be, you know, 549 00:39:26,840 --> 00:39:30,000 there's going to be a lot of PRS and issues. 550 00:39:30,000 --> 00:39:35,920 You have to be responsive. It's hard, but at least for now, 551 00:39:35,960 --> 00:39:41,760 MDX and gums are not like huge projects like you know, like Ash 552 00:39:41,760 --> 00:39:45,840 or Phoenix or whatever. So it's fine for now. 553 00:39:45,840 --> 00:39:51,760 But getting, if it does get more contributions and issues, I 554 00:39:51,760 --> 00:39:54,720 don't know. I just, you know, you have to 555 00:39:54,720 --> 00:39:57,080 find some time. Yeah, for sure. 556 00:39:59,360 --> 00:40:05,400 Yeah, recently I opened the spot the sponsors sponsorship on you 557 00:40:05,440 --> 00:40:10,960 hub and yeah, I got some some sponsors like it's not something 558 00:40:10,960 --> 00:40:14,960 I'm looking for like my you know, it's not not my main 559 00:40:14,960 --> 00:40:16,440 source of commerce. Hey. 560 00:40:16,440 --> 00:40:18,680 Guys, head over there and sponsor this project. 561 00:40:20,000 --> 00:40:22,440 Yeah, like. If you're using MDX, you know, 562 00:40:22,520 --> 00:40:27,720 feel free and you know, it's a, it's a really nice way to, to 563 00:40:27,720 --> 00:40:32,400 help as well. And I'm grateful to have some, 564 00:40:32,400 --> 00:40:37,880 like I had one, I had some one time contributors. 565 00:40:38,280 --> 00:40:43,240 Mike from Dido is now contributing monthly as well. 566 00:40:43,680 --> 00:40:45,760 So that's really, really nice to have. 567 00:40:45,760 --> 00:40:49,400 Like, I mean, open source is about community. 568 00:40:49,560 --> 00:40:52,240 That's really what open source is. 569 00:40:53,080 --> 00:40:56,720 Jeto is something like they're making new projects every day. 570 00:40:56,720 --> 00:40:59,600 I feel like they they're, I can't keep up. 571 00:40:59,600 --> 00:41:01,560 I'm in their Discord and I can't keep up. 572 00:41:02,160 --> 00:41:05,640 Yeah, you open Twitter X or whatever and there's a new 573 00:41:05,640 --> 00:41:08,960 project. Like OK, I didn't have time to 574 00:41:08,960 --> 00:41:11,160 see the other one, now there's a new one. 575 00:41:11,760 --> 00:41:14,040 I'm making, I'm going to make Mike come on and do like a 576 00:41:14,040 --> 00:41:16,240 coding live stream with me because I was like, dude, I 577 00:41:16,240 --> 00:41:17,720 can't keep up. So you have to teach me 578 00:41:17,720 --> 00:41:19,440 everything that you're doing in Jeto. 579 00:41:21,440 --> 00:41:25,440 Yeah, it's really it. I mean, now AI, it's really hard 580 00:41:25,440 --> 00:41:28,880 to keep up because you know. You feel like everyone's pedal 581 00:41:28,880 --> 00:41:32,800 to the metal now and it's like new stuff multiple times a day. 582 00:41:32,800 --> 00:41:37,800 Like we thought it was hard to keep up in in tech before AI and 583 00:41:37,800 --> 00:41:42,000 now now look at it. Yeah, we should. 584 00:41:43,040 --> 00:41:45,600 I guess we still need to learn how to navigate this. 585 00:41:45,600 --> 00:41:49,640 This time is because, you know, new projects, you know, 586 00:41:49,640 --> 00:41:52,480 contributions and you know, a lot of stuff happening at the 587 00:41:52,480 --> 00:41:57,440 same time. In Markdown, for example, we 588 00:41:57,440 --> 00:42:01,760 have so many new projects because, you know, Markdown has 589 00:42:01,760 --> 00:42:05,320 a specification. So I'd say using AI to create a 590 00:42:05,320 --> 00:42:10,120 decent parser from just, you know, from scratch, just using 591 00:42:10,120 --> 00:42:17,320 AI is I'd say it's possible. So I had, you know, I'm trying 592 00:42:17,320 --> 00:42:21,240 to follow some projects, but every day there's something new, 593 00:42:21,640 --> 00:42:24,440 even for just for markdown or ceilidh highlights. 594 00:42:26,280 --> 00:42:29,600 Yeah, that's crazy. One thing I'm noticing is there 595 00:42:29,600 --> 00:42:33,680 seems to be a growing gap of like especially so like Ioffer 596 00:42:33,680 --> 00:42:35,560 like pre vetted recruitment stuff. 597 00:42:36,200 --> 00:42:42,080 And so I'm reviewing a lot of code and I think we have the 598 00:42:42,080 --> 00:42:46,040 issue that I'm starting to see and it's kind of annoying is I 599 00:42:46,040 --> 00:42:49,440 think you have a lot of people that are like very like junior, 600 00:42:49,440 --> 00:42:54,120 mid level developers trying to claim senior level now because 601 00:42:54,120 --> 00:42:57,400 they think they can code so much better and faster with LLMS. 602 00:42:58,120 --> 00:43:02,960 But it you're just like wading through so much LLM slop, it's 603 00:43:02,960 --> 00:43:07,400 like they don't know how to actually create, create well 604 00:43:07,400 --> 00:43:10,440 structured code with LLMS. And there's like a there's a big 605 00:43:10,440 --> 00:43:13,760 gap because like I see, you know, like senior level guys 606 00:43:13,760 --> 00:43:16,680 that I work with on the daily on a daily basis. 607 00:43:16,680 --> 00:43:19,640 I know they're using LLMS to write their code, but it still 608 00:43:19,640 --> 00:43:22,640 looks like good code. And you know, you're not over 609 00:43:22,640 --> 00:43:25,840 engineering solutions, but then you have like you're less 610 00:43:25,840 --> 00:43:29,520 talented developers that are just like spitting out these 611 00:43:29,520 --> 00:43:33,680 crazy over engineered solutions. And it's like almost too much. 612 00:43:34,080 --> 00:43:37,040 And you're like, come on, guys. You got like, there's got to be, 613 00:43:37,480 --> 00:43:40,960 I don't know, we need like a, a better push to teach people how 614 00:43:40,960 --> 00:43:45,000 to use LLMS in their workflow. Yeah, yeah. 615 00:43:45,000 --> 00:43:47,680 And that reminds me of another Jido project. 616 00:43:47,680 --> 00:43:51,640 The there's like a Jido spec or something like the spec driven 617 00:43:51,640 --> 00:43:54,880 development. That's a new, I mean it's not a 618 00:43:54,880 --> 00:44:01,520 new thing, but seems like to be born again because of AI. 619 00:44:01,640 --> 00:44:06,680 So you because there's a big difference between the AI 620 00:44:06,680 --> 00:44:11,440 driving the code driving you and you driving the AI code. 621 00:44:11,440 --> 00:44:13,600 Yeah, it's 100%. Right. 622 00:44:13,840 --> 00:44:18,480 And seems like spec driven development is kind of, you 623 00:44:18,480 --> 00:44:24,040 know, the, it's not, I'm not going to see the solution 624 00:44:24,040 --> 00:44:28,160 because we are still learning, you know, but the spectrum 625 00:44:28,160 --> 00:44:32,480 driven development might be a way that people can, you know, 626 00:44:33,000 --> 00:44:36,120 just stop for a second, think about the problem 1st. 627 00:44:36,360 --> 00:44:42,360 And to realize a bad, a good design, you have to be a senior 628 00:44:42,360 --> 00:44:46,040 developer, you have to have experience, right? 629 00:44:46,200 --> 00:44:51,360 And so AI alone is not enough. You need to understand what's 630 00:44:51,360 --> 00:44:55,920 happening. So I was able to. 631 00:44:56,240 --> 00:44:58,960 Catch architectural mistakes too, right? 632 00:44:58,960 --> 00:45:05,000 If you just are just generically asking an LLM to create a 633 00:45:05,000 --> 00:45:08,840 feature and there's no guiding or planning to that feature, 634 00:45:09,080 --> 00:45:12,240 you're going to get over engineered slop and like unused 635 00:45:12,520 --> 00:45:14,800 do. The amount of unused code in 636 00:45:14,800 --> 00:45:17,960 coding projects is baffling to me too. 637 00:45:18,720 --> 00:45:25,920 Have have you ever tried to ask Claude to review the codecs or 638 00:45:25,920 --> 00:45:30,600 GPT code or vice versa? They, you know, put the two 639 00:45:30,640 --> 00:45:35,320 together 1 reveal the code of the other and they always find 640 00:45:35,320 --> 00:45:40,360 bugs and like a lot of stuff. I'm like, I'll be, it'll be like 641 00:45:40,360 --> 00:45:42,440 hit or miss. I feel like sometimes it'll just 642 00:45:42,440 --> 00:45:45,240 be a total garbage review and then sometimes it'll actually 643 00:45:45,240 --> 00:45:48,920 find stuff that's like super helpful, but it's not consistent 644 00:45:48,920 --> 00:45:53,480 enough that I trust it. Like using like Copilot to do 645 00:45:53,480 --> 00:45:55,760 code reviews or clawed to do code reviews. 646 00:45:56,280 --> 00:46:00,280 Like I feel like it hallucinates more than it helps but I still 647 00:46:00,280 --> 00:46:04,040 use them. Yeah, I I'm finding that having 648 00:46:04,040 --> 00:46:07,200 skills, really well written skills. 649 00:46:07,200 --> 00:46:11,320 Yeah, skills are important. Yeah, it does help a lot. 650 00:46:11,800 --> 00:46:14,920 It reduce hallucinations. It, you know, it's a way to like 651 00:46:15,800 --> 00:46:22,040 a constraint for the, the AI for the Asians, but it's, I've been 652 00:46:22,360 --> 00:46:26,560 doing this experiment like running cloud and codecs at the 653 00:46:26,560 --> 00:46:30,920 same time and asking 1 to reveal the code of the other. 654 00:46:31,680 --> 00:46:36,200 And it's funny because when I need to go back to the, to cloud 655 00:46:36,200 --> 00:46:38,840 and say, Hey, that's what GPT found about your code. 656 00:46:38,840 --> 00:46:43,520 And then it's like, Oh yeah, that's read a book like, so why 657 00:46:43,520 --> 00:46:49,240 didn't you get the book before? So you you cannot trust for 658 00:46:49,240 --> 00:46:50,840 sure. Like you have to review the 659 00:46:50,840 --> 00:46:52,960 code, you have to know what's happening. 660 00:46:53,680 --> 00:46:58,440 So it does require some, you know, experience to use AII. 661 00:46:59,040 --> 00:47:04,760 Think the the big mistake a lot of like mid level and under 662 00:47:04,760 --> 00:47:07,440 developers are making too is they're trying to one shot these 663 00:47:07,440 --> 00:47:10,840 things instead of like stepping through a problem. 664 00:47:11,480 --> 00:47:14,400 Because I like the smaller you break things down, the higher 665 00:47:14,400 --> 00:47:17,240 quality of the code can because you're actually paying attention 666 00:47:17,240 --> 00:47:20,800 to it more carefully too. So if it spits out something 667 00:47:20,800 --> 00:47:23,680 that's not quite right, you're going to correct it and then 668 00:47:23,680 --> 00:47:27,400 move on, right. So I feel like the one shot in 669 00:47:27,400 --> 00:47:31,720 giant features is not the smartest move without plan. 670 00:47:31,760 --> 00:47:37,560 So like I have like 1 shotted features before but like I have 671 00:47:37,560 --> 00:47:42,560 a whole planning phase and I create like an epic and issues 672 00:47:42,560 --> 00:47:44,440 and like a checklist essentially right? 673 00:47:45,000 --> 00:47:48,680 So but that works. Because you have experience in 674 00:47:48,680 --> 00:47:50,320 that, right? Yeah, that works because you 675 00:47:50,320 --> 00:47:52,360 know what's happening. Yeah, yeah. 676 00:47:52,720 --> 00:47:57,000 So by the coding's a mistake, it does not work. 677 00:47:57,600 --> 00:48:01,080 Yeah, I think of it more as like a junior's writing the stuff you 678 00:48:01,080 --> 00:48:03,640 don't want to write, but you still are reviewing it and 679 00:48:03,640 --> 00:48:07,200 putting your name on it when you hit that merge button, you know? 680 00:48:08,640 --> 00:48:11,880 Yeah, I mean, it's really cool to have contributions on on 681 00:48:11,880 --> 00:48:17,800 projects, but you have to be careful, you know, especially on 682 00:48:17,800 --> 00:48:22,600 bigger projects and even like bigger projects now they just 683 00:48:22,600 --> 00:48:27,320 reject AI contributions. So they just. 684 00:48:27,320 --> 00:48:28,680 Sometimes it's hard to tell though. 685 00:48:29,480 --> 00:48:33,080 It's getting harder and harder. Yes, that that's true. 686 00:48:33,200 --> 00:48:34,120 That's true. But. 687 00:48:34,120 --> 00:48:38,600 But I think like the the giveaways are, I mean, if you're 688 00:48:38,600 --> 00:48:41,880 not careful, you get random really weird comments in your 689 00:48:41,880 --> 00:48:44,120 code. And I don't think any well 690 00:48:44,120 --> 00:48:46,960 written code needs comments. So you shouldn't have commented 691 00:48:46,960 --> 00:48:50,440 code And then I would say over engineering problems. 692 00:48:50,680 --> 00:48:56,440 That's like a big giveaway. Yeah, you, you, you, you can, 693 00:48:56,440 --> 00:48:58,960 you can tell. Sometimes it's hard, but you can 694 00:48:58,960 --> 00:49:04,760 tell when it's AI generic code like comments and it's not it 695 00:49:04,760 --> 00:49:09,000 does not go straight to the problem or it does not it's not 696 00:49:09,000 --> 00:49:13,600 using the right boundaries on on your code that that a lot of you 697 00:49:13,600 --> 00:49:18,120 know, or for example, adding new libraries that are outdated and 698 00:49:18,680 --> 00:49:19,480 stuff like that. So. 699 00:49:19,560 --> 00:49:24,200 It loves using old like old libraries like on everything. 700 00:49:25,440 --> 00:49:28,720 Which actually an open source project makes sense because it 701 00:49:28,720 --> 00:49:35,320 cannot depends on the ecosystem. But usually you have to declare 702 00:49:35,440 --> 00:49:39,280 older versions, so otherwise people cannot use your. 703 00:49:39,280 --> 00:49:40,880 Project. Yeah, that makes sense. 704 00:49:41,480 --> 00:49:45,760 But you know, regular project, usually you want to, you know, 705 00:49:45,880 --> 00:49:50,600 you want to have the latest right and AI they fail, usually 706 00:49:50,600 --> 00:49:53,000 they fail to do that. There are always a couple of 707 00:49:53,000 --> 00:49:56,320 versions off, you know. Yeah, yes. 708 00:49:56,320 --> 00:50:01,200 So there are, there are ways to you know, that you can tell if 709 00:50:01,200 --> 00:50:05,000 it's AAI generic code or not. But I guess sometimes it's 710 00:50:05,000 --> 00:50:07,880 really hard and just, you know, just goes. 711 00:50:07,880 --> 00:50:11,360 But but for open source maintainers, I can see the the 712 00:50:11,400 --> 00:50:15,480 problem. Although although I'm pro pro 713 00:50:15,480 --> 00:50:18,920 AII like to use AIII think it helps. 714 00:50:18,920 --> 00:50:21,240 But you need to understand what's happening right? 715 00:50:21,360 --> 00:50:23,560 Because the the code is is still yours. 716 00:50:23,560 --> 00:50:28,560 It's your PR, it's your brain, it's your code even though it 717 00:50:28,560 --> 00:50:31,920 was generated by AI. Yeah, I agree. 718 00:50:32,080 --> 00:50:34,080 I'm I'm a big, you know, I'm a fan. 719 00:50:34,200 --> 00:50:38,680 It took me, I was like a late, late to the partier but late to 720 00:50:38,680 --> 00:50:41,320 the party. But now I like it because it, it 721 00:50:41,320 --> 00:50:48,960 just improves my efficiency. And once you kind of know how to 722 00:50:48,960 --> 00:50:52,040 work it into your own like workflow and planning, I think 723 00:50:52,040 --> 00:50:55,360 it's pretty powerful tool. Like I'm not a good front end 724 00:50:55,360 --> 00:50:59,120 developer and it helps me be a lot better front end developer. 725 00:50:59,960 --> 00:51:03,560 So it's beneficial for me in those ways. 726 00:51:04,320 --> 00:51:07,160 Yeah, same here. If you go to the Lumis website, 727 00:51:07,160 --> 00:51:10,920 that was the itinerary like the front end is. 728 00:51:11,520 --> 00:51:16,000 I have no idea to how to write that code because I'm not a 729 00:51:16,000 --> 00:51:20,560 front end developer either. So for that stuff it's really 730 00:51:20,560 --> 00:51:24,000 nice. And also, for example, the Lumis 731 00:51:24,000 --> 00:51:29,880 requires a lot of, a lot of like bureaucracy code, like defining 732 00:51:29,880 --> 00:51:33,280 the languages, repos and all that stuff. 733 00:51:34,360 --> 00:51:39,760 That's the kind of, that's the kind of code AI just shines, you 734 00:51:39,760 --> 00:51:42,640 know, just repeated code, you know, just get this block and 735 00:51:42,640 --> 00:51:46,200 repeat and change just just the name or just the URL or 736 00:51:46,200 --> 00:51:49,960 something. You know, it takes forever to 737 00:51:50,360 --> 00:51:54,400 maintain this kind of code. But AI is just 5 minutes and 738 00:51:54,400 --> 00:51:58,840 then you reveal and it's done. So there are many, many users 739 00:51:58,840 --> 00:52:02,360 for, for AI for sure. You just have to be careful. 740 00:52:03,120 --> 00:52:08,000 You definitely do. And yeah, yeah. 741 00:52:08,640 --> 00:52:12,880 And guys, if you're not truly like a mid level or senior, 742 00:52:13,200 --> 00:52:16,000 don't pretend you are because you think you are with AI just, 743 00:52:16,720 --> 00:52:19,800 you know, take take. You got to go through the the 744 00:52:19,800 --> 00:52:23,120 same ladders and steps everyone else has and you'll get there 745 00:52:23,120 --> 00:52:26,120 eventually. Yeah, but you can use AI for 746 00:52:26,120 --> 00:52:28,400 learning as well. Yes, absolutely. 747 00:52:28,680 --> 00:52:30,680 I use it all the time for learning honestly. 748 00:52:30,680 --> 00:52:33,440 Like if I ever see like a block of code or like some kind of 749 00:52:33,440 --> 00:52:39,040 like flow in the code base that I'm not quite sure how it works. 750 00:52:39,040 --> 00:52:41,520 Dude I I have AI break it down for me. 751 00:52:42,600 --> 00:52:46,800 Yeah, I I do the same like for research and you know, you can 752 00:52:46,800 --> 00:52:50,000 just use it. I just like, hey, give a talk 753 00:52:50,000 --> 00:52:55,440 and how to brainstorm something or research it's or even 754 00:52:55,440 --> 00:52:58,000 learning about a new product, about a new code base. 755 00:52:58,000 --> 00:53:00,840 It's really helpful. Yeah, for sure. 756 00:53:01,080 --> 00:53:05,640 I'm curious, with like MDX and Loomis now, do you have any 757 00:53:05,640 --> 00:53:08,200 projects that you're building where you're using them 758 00:53:08,440 --> 00:53:12,040 consistently or not? Not really anymore. 759 00:53:12,880 --> 00:53:19,320 I was using for the Wazmi research because I had never 760 00:53:19,320 --> 00:53:23,440 used Wazmi before even though I, you know, I knew about about it 761 00:53:23,440 --> 00:53:29,120 but never used. So Wazmi research and these new 762 00:53:29,320 --> 00:53:33,600 JavaScript library requires a lot of, you know, copy and paste 763 00:53:33,600 --> 00:53:37,600 code essentially. So I'm using for that as well 764 00:53:38,280 --> 00:53:40,360 and using for code review as well. 765 00:53:40,360 --> 00:53:44,920 Even though it's not perfect, but sometimes it catch some bugs 766 00:53:44,920 --> 00:53:49,280 or something that you, you like, you really, sometimes you miss 767 00:53:49,280 --> 00:53:52,200 something, right? And then you can use for code 768 00:53:52,200 --> 00:53:56,920 review and give some hints about something you for you're 769 00:53:56,920 --> 00:53:59,240 forgetting or you have to take another look. 770 00:54:00,120 --> 00:54:04,600 So for code review as well and for for Lumis, for Lumis and 771 00:54:04,600 --> 00:54:08,560 MEDX, that's the, you know, that's where I'm using AI now 772 00:54:08,920 --> 00:54:11,200 essentially. OK, nice. 773 00:54:11,520 --> 00:54:16,720 Wait, I, I feel like it's, there's some sort of benefit 774 00:54:16,840 --> 00:54:20,480 like if you don't to have another engineer look at code 775 00:54:21,120 --> 00:54:24,600 that like not everyone can pick up everything, like you'll miss 776 00:54:24,600 --> 00:54:28,120 things from time to time. And I feel like on solo 777 00:54:28,120 --> 00:54:31,960 projects, AI has a huge benefit there because it's kind of like 778 00:54:32,640 --> 00:54:35,400 an extra pair of eyes. At least it makes you look 779 00:54:35,400 --> 00:54:38,320 closer at parts of code, whether you agree with what it finds or 780 00:54:38,320 --> 00:54:42,280 not. Yeah, that's that's true because 781 00:54:42,640 --> 00:54:46,520 most open source projects are just just one person like 782 00:54:47,720 --> 00:54:52,160 working on a lot of stuff at the same at same time, you know, on 783 00:54:52,400 --> 00:54:57,400 on his free time or or you know, short limited time. 784 00:54:58,440 --> 00:55:04,440 So it's, it's a way to unblock new features to help with code 785 00:55:04,440 --> 00:55:09,120 review. Yeah, it's for, I mean, there 786 00:55:09,120 --> 00:55:15,120 are downsides of, you know, using or abusing AI in the open 787 00:55:15,120 --> 00:55:17,920 source community, but it can also help. 788 00:55:19,040 --> 00:55:24,560 Yeah, 100% that like now that you're not, you're not like 789 00:55:24,880 --> 00:55:28,320 working on the Beacon project and that's where you're kind of 790 00:55:28,320 --> 00:55:31,600 consuming these libraries. Are you consuming your libraries 791 00:55:31,600 --> 00:55:37,760 anywhere else currently or no? No, not really. 792 00:55:39,160 --> 00:55:43,040 Although I really enjoy working on MDX and Loomis, I'm not. 793 00:55:43,400 --> 00:55:49,280 I'm not going to stop, but I had some ideas to use MDX or Loomis. 794 00:55:49,280 --> 00:55:52,800 But you know, I was trying to find some ideas where I could 795 00:55:52,800 --> 00:55:58,800 monetize the projects, right? Like a private project, like a 796 00:55:58,880 --> 00:56:02,960 private closed source project, something I could sell. 797 00:56:04,480 --> 00:56:08,680 But for now, I'm just trying to get the, you know, the bigger 798 00:56:08,680 --> 00:56:11,880 features that I want to have on those libraries. 799 00:56:12,800 --> 00:56:16,680 And then eventually maybe you know, create something like a 800 00:56:17,120 --> 00:56:20,480 like a side project, not something big or whatever, 801 00:56:21,040 --> 00:56:24,920 something small that I could try to monetize and help, you know, 802 00:56:25,000 --> 00:56:27,640 to help fund the project. Yeah, Yeah. 803 00:56:27,640 --> 00:56:32,720 I'm a big believer in doing like like micro products and just 804 00:56:32,720 --> 00:56:36,920 have, you know, a couple micro products and you know, try to 805 00:56:36,920 --> 00:56:41,400 grow them to a couple 1000 users and then that's good, that's 806 00:56:41,400 --> 00:56:44,520 good like side income and eventually like it could 807 00:56:44,520 --> 00:56:47,160 replace, you know, your day job, whatever. 808 00:56:47,160 --> 00:56:50,720 And then you just have all these cool side projects that actually 809 00:56:50,720 --> 00:56:55,400 monetize. Yeah, I mean, I'm not looking to 810 00:56:55,400 --> 00:57:01,080 keeping keeping my my day, day job, but it'll be nice to have 811 00:57:01,080 --> 00:57:06,800 like a side project that I could main mainly helped fund the 812 00:57:06,800 --> 00:57:09,600 project. And yeah, for sure, that would 813 00:57:09,600 --> 00:57:11,080 be nice. What? 814 00:57:11,240 --> 00:57:13,360 What do you have any ideas that you can share? 815 00:57:14,840 --> 00:57:17,800 No, not really. I was, you know that that's the 816 00:57:17,800 --> 00:57:20,320 problem. Every day you open Twitter and 817 00:57:20,320 --> 00:57:23,800 there are like a bunch of new projects, a bunch of new ideas 818 00:57:23,800 --> 00:57:26,400 and stuff in. There's no original idea at this 819 00:57:26,400 --> 00:57:29,240 point anymore, is there? Yeah, yeah. 820 00:57:29,240 --> 00:57:34,840 But you can always innovate even in well established, you know, 821 00:57:34,840 --> 00:57:37,480 areas. But the problem is there's so 822 00:57:37,480 --> 00:57:42,160 many, so much happening at the same time, like and you know, 823 00:57:42,160 --> 00:57:49,120 markdown has been used in projects for like MCP projects 824 00:57:49,120 --> 00:57:52,760 or that kind of stuff. But it's moving so fast that, 825 00:57:52,800 --> 00:57:58,160 you know, your idea became like, your idea becomes stale like in 826 00:57:58,160 --> 00:58:02,480 a week. So it's like it's, it's hard to 827 00:58:02,480 --> 00:58:04,080 decide. It's hard to find something that 828 00:58:04,080 --> 00:58:09,280 you can like commit to because if you start some side project, 829 00:58:09,280 --> 00:58:11,960 you have to be certain about that project, right? 830 00:58:12,520 --> 00:58:14,080 Because you're going to spend some time. 831 00:58:14,960 --> 00:58:19,480 So it's yeah for now just working on the working search 832 00:58:19,480 --> 00:58:22,240 projects. Fair enough. 833 00:58:22,920 --> 00:58:24,840 What? What kind of things do you work 834 00:58:24,840 --> 00:58:27,800 on with your day job? Yeah. 835 00:58:27,800 --> 00:58:30,200 So right now I'm working at Tallis Park. 836 00:58:30,200 --> 00:58:37,560 They are a tech solution for, you know, for US based schools. 837 00:58:38,600 --> 00:58:46,560 And so we, I can say for sure, but we were one of the first to 838 00:58:46,560 --> 00:58:49,240 use live view, you know. In production. 839 00:58:49,240 --> 00:58:51,840 I was just going to ask, are you writing Elixir every day? 840 00:58:51,920 --> 00:58:54,440 Yeah, it's Elixir essentially. Yeah. 841 00:58:55,600 --> 00:59:00,920 So we have been using Phoenix live view for like for years now 842 00:59:01,640 --> 00:59:05,920 and seems like the first published, like one of the first 843 00:59:06,200 --> 00:59:10,480 published versions. And yeah, so it's pretty much 844 00:59:10,720 --> 00:59:16,600 Phoenix and Live view. And we do have some markdown in 845 00:59:16,600 --> 00:59:22,400 the product, but not, not so much we, we, we use MDX here, 846 00:59:22,400 --> 00:59:24,800 but it's not the core of the product. 847 00:59:26,280 --> 00:59:31,920 But most of the projects just Phoenix and live view and that 848 00:59:32,080 --> 00:59:35,800 yeah, that's how I got into live view. 849 00:59:35,800 --> 00:59:40,640 And that's, yeah. So that's the, the, the project 850 00:59:40,640 --> 00:59:43,560 we are. It's pretty much just live view, 851 00:59:43,560 --> 00:59:45,000 yeah. OK, cool. 852 00:59:46,160 --> 00:59:50,280 Any like interesting fun project or problems that you've recently 853 00:59:50,440 --> 00:59:58,320 ran into that you had to solve? Like here the OR like any of 854 00:59:59,840 --> 01:00:02,160 like you mean here in the like in the? 855 01:00:02,160 --> 01:00:05,800 Company or yeah, in the company just on the platform, like do 856 01:00:05,800 --> 01:00:09,000 you have like a pretty large user base? 857 01:00:09,000 --> 01:00:12,040 Do you have performance issues you're dealing with? 858 01:00:13,320 --> 01:00:19,360 Yeah, I guess, I guess now it's more about scaling the company. 859 01:00:19,440 --> 01:00:22,760 You know, when you have like like you have a product and you 860 01:00:22,760 --> 01:00:27,440 have a bunch of clients and you have to scale like the you know, 861 01:00:27,600 --> 01:00:30,480 the infrastructure that you have, you need to scale and then 862 01:00:30,480 --> 01:00:34,040 the the cold base. So you are that at this point 863 01:00:34,040 --> 01:00:38,480 now. But over the years, that's how I 864 01:00:38,480 --> 01:00:42,240 got into Phoenix Live your contributions, because we 865 01:00:42,240 --> 01:00:48,200 started using live view on this product and also Ecto and all 866 01:00:48,200 --> 01:00:55,240 the other libraries and we had to like, we needed some, you 867 01:00:55,920 --> 01:00:59,600 know, some features that, you know, we didn't have back in the 868 01:00:59,600 --> 01:01:02,320 day. So for example, when we start 869 01:01:02,320 --> 01:01:07,800 scaling and having more customers, we had to analyze the 870 01:01:07,800 --> 01:01:11,800 performance of the database and the queries on the database. 871 01:01:12,480 --> 01:01:16,160 So that's, for example, when I contribute the EXPLAIN function 872 01:01:16,160 --> 01:01:20,880 in Ecto, so we can run the EXPLAIN query on Ecto. 873 01:01:22,000 --> 01:01:25,760 And so, yeah, I was able to do some, you know, contribution 874 01:01:25,760 --> 01:01:31,160 here and there, the ecosystem because of, you know, the 875 01:01:31,160 --> 01:01:33,080 company's growth. So that's another way that you 876 01:01:33,080 --> 01:01:37,080 can also, in my opinion, you can contribute to open service, 877 01:01:37,120 --> 01:01:38,480 right? You have a problem in your 878 01:01:38,480 --> 01:01:41,760 project, like don't, don't wait for a solution. 879 01:01:41,760 --> 01:01:44,120 Just, you know, you can solve it. 880 01:01:44,840 --> 01:01:49,240 Yeah, I know it's hard, but it's another way that you can, you 881 01:01:49,240 --> 01:01:52,880 know, contribute to open source. Well I always find it funny when 882 01:01:52,880 --> 01:01:55,200 you see like open source projects and someone's 883 01:01:55,200 --> 01:01:59,760 complaining about a feature not working or wanting a new 884 01:01:59,760 --> 01:02:03,520 feature. Like guys you could just ask if 885 01:02:03,520 --> 01:02:06,120 it's on the road map and then submit the PR. 886 01:02:06,120 --> 01:02:07,960 Do it. Like if you need it, you should 887 01:02:07,960 --> 01:02:11,720 submit the PR. Just just open Cloud Code and. 888 01:02:13,120 --> 01:02:16,040 Vibe it out, put on your sunglasses and go. 889 01:02:18,080 --> 01:02:22,200 And just submit the PR, Yeah, but you can, you can do like you 890 01:02:22,200 --> 01:02:25,040 can, you can for sure contribute. 891 01:02:26,480 --> 01:02:33,120 It's not super hard I'd say. Just make sure that it will be 892 01:02:33,120 --> 01:02:35,200 accepted before you put it in the work though. 893 01:02:36,360 --> 01:02:42,200 Yeah, it it depends when it's a feature request, depending on 894 01:02:42,200 --> 01:02:45,960 the project, you have to use the issue tracker or there's a core 895 01:02:45,960 --> 01:02:49,160 or you can chat about our discord server. 896 01:02:50,440 --> 01:02:54,480 When it's a bug fix, I'd say don't wait, just send the patch, 897 01:02:54,880 --> 01:03:00,800 just send the PR. And yeah, that's, you know, the 898 01:03:00,800 --> 01:03:04,920 good thing about Elixir community is that we are pretty 899 01:03:04,920 --> 01:03:09,800 much pretty much everyone is very friend friendly and open. 900 01:03:10,080 --> 01:03:17,600 So it's AI would say Elixir is a we we could say it's a small 901 01:03:17,600 --> 01:03:23,480 community compared to others. So we some we kind of know each 902 01:03:23,560 --> 01:03:27,680 other, if you know what I mean. And then it's it's easier to get 903 01:03:27,680 --> 01:03:30,440 contributions in this environment. 904 01:03:30,640 --> 01:03:35,760 It's more friendly than others. Yeah, I'd say for the most part, 905 01:03:35,760 --> 01:03:37,720 everyone's really nice in the electric community. 906 01:03:39,200 --> 01:03:40,920 I think that's one of the bonuses, right? 907 01:03:42,440 --> 01:03:49,360 Yes, yeah, yeah, and yeah. For so for index and, and and 908 01:03:49,360 --> 01:03:56,080 Loomis, we got some contributions recently and it's, 909 01:03:57,040 --> 01:04:00,760 you know, it's nice to see the the library growing and 910 01:04:00,760 --> 01:04:04,480 contribution. So what I sometimes people ask 911 01:04:04,480 --> 01:04:07,440 me like, hey, can I contribute or can do something like just 912 01:04:07,720 --> 01:04:11,200 just open the issue, like open an issue we can discuss. 913 01:04:11,200 --> 01:04:14,120 We can, you know, sometimes it does not work out, but 914 01:04:15,160 --> 01:04:18,200 eventually most of them does work. 915 01:04:18,800 --> 01:04:22,080 And that's how you can get, you know, contributions done. 916 01:04:23,960 --> 01:04:27,680 Yeah. Any, any lessons that you can 917 01:04:27,680 --> 01:04:34,160 share having open source projects because like did you, 918 01:04:34,360 --> 01:04:37,000 did you do anything before Beacon or was Beacon kind of 919 01:04:37,000 --> 01:04:43,640 your first open source project? I used to do some contributions 920 01:04:43,880 --> 01:04:47,680 back in the day for angry JS and radios. 921 01:04:49,760 --> 01:04:56,560 That's when I started and then, you know, eventually after 922 01:04:56,560 --> 01:05:01,160 migrating to Elixir, getting some contributions to Elixir 923 01:05:01,160 --> 01:05:08,280 projects. But I don't know, maybe maybe 924 01:05:08,400 --> 01:05:13,640 Beacon was, you know, the first big one, I'd say, if I can say. 925 01:05:13,720 --> 01:05:15,480 That you were like a main contributor on. 926 01:05:16,160 --> 01:05:19,920 Yeah, yeah. But I'd say if you're starting a 927 01:05:19,920 --> 01:05:26,240 new project, if you want to succeed, I'd say marketing is 928 01:05:26,240 --> 01:05:30,400 very important, even though no one likes to do that. 929 01:05:30,720 --> 01:05:34,960 No, it's the hardest thing ever. I always tell people the easiest 930 01:05:34,960 --> 01:05:40,560 part of, like, building like a SAS product or like a library is 931 01:05:40,800 --> 01:05:43,280 coding. And that's just because my 932 01:05:43,280 --> 01:05:46,520 brain's not, Yeah, my brain's not wired for the marketing and 933 01:05:46,520 --> 01:05:50,920 the sales. It's like you wish the like the 934 01:05:50,920 --> 01:05:52,560 statement, build it and they will come. 935 01:05:52,560 --> 01:05:55,440 You wish that was true, but it's the farthest thing from the 936 01:05:55,440 --> 01:05:58,480 truth. And, and I'm guilt of that. 937 01:05:58,480 --> 01:06:02,640 I'm guilt of that as well because that's, you know, MDX, 938 01:06:02,640 --> 01:06:05,760 for example, I never talk about MDX in public. 939 01:06:07,240 --> 01:06:08,880 I. Have Here we are, though you're 940 01:06:08,880 --> 01:06:11,800 in public. Yeah, that's the first time I'm 941 01:06:11,800 --> 01:06:16,520 talking about MDX in English. Like you know before that I had 942 01:06:16,520 --> 01:06:22,400 a meet up, remote meet up in Brazil, but it was Portuguese. 943 01:06:22,480 --> 01:06:27,040 Right now is the first time I'm talking about MDX in English. 944 01:06:27,040 --> 01:06:31,920 Like the library, he's like 2 years old more and that that's 945 01:06:31,920 --> 01:06:35,360 the first time I'm talking about MDX in English. 946 01:06:35,360 --> 01:06:42,600 So marketing is important. So right now we have 500,000 947 01:06:42,600 --> 01:06:47,080 downloads and I had. I have never talked about index 948 01:06:47,240 --> 01:06:50,920 in public, only like Twitter. Sometimes I just post some 949 01:06:50,920 --> 01:06:52,840 releases. I was going to ask, how do you 950 01:06:52,840 --> 01:06:56,880 do you think people are finding it through the hex library or 951 01:06:56,880 --> 01:07:00,800 through the hex website? Probably following you on on 952 01:07:00,800 --> 01:07:04,400 Twitter. I'm I'm not sure to be honest. 953 01:07:05,160 --> 01:07:08,760 I know some people react on the Elixir forum. 954 01:07:08,760 --> 01:07:13,840 People use the Elixir forum as a sirs new libraries and news on 955 01:07:13,840 --> 01:07:17,440 the community, the thinking Lexer podcast. 956 01:07:17,440 --> 01:07:22,280 They have been covering the MDX releases for some time now. 957 01:07:22,280 --> 01:07:25,160 I'm very grateful because that's a really good source. 958 01:07:25,600 --> 01:07:32,720 Yeah, I know the I know Mark is using MDX, so he's always, you 959 01:07:32,720 --> 01:07:35,840 know, following the development of MDX. 960 01:07:36,120 --> 01:07:40,920 So that's probably a good source, but I'm, I'm not sure, 961 01:07:42,160 --> 01:07:44,120 you know, people just starting to use it. 962 01:07:44,280 --> 01:07:48,640 And sometimes people just say, Hey, I'm using index on this and 963 01:07:48,640 --> 01:07:53,480 I, I had no idea. I I don't know, to be honest, 964 01:07:53,720 --> 01:07:58,440 where people are finding index. But but again, marketing is 965 01:07:58,440 --> 01:08:01,160 important. So, you know, I'm trying to be 966 01:08:01,160 --> 01:08:05,160 more open about the development. There are a lot of challenges of 967 01:08:05,240 --> 01:08:09,240 the, you know, development of people's people sometimes don't 968 01:08:09,240 --> 01:08:13,240 realize how hard it is to, you know, mark down and Syntex 969 01:08:13,240 --> 01:08:17,120 highlight are they look like a simple problem to solve, but 970 01:08:17,640 --> 01:08:19,680 it's not. I mean, just talking to you 971 01:08:19,680 --> 01:08:23,200 about like the streaming aspects and like things that I never 972 01:08:23,200 --> 01:08:27,120 even thought about. There's a lot of complex 973 01:08:27,120 --> 01:08:30,000 problems under the hood for something that you just want to 974 01:08:30,000 --> 01:08:32,560 work perfectly right as a end user. 975 01:08:32,560 --> 01:08:34,479 You just think, oh, it's simple. It's just. 976 01:08:35,960 --> 01:08:37,520 Just. Markdown, dude, it's just 977 01:08:37,520 --> 01:08:40,560 markdown. Yeah, it's it's not like 978 01:08:41,560 --> 01:08:44,399 streaming is just one of the features. 979 01:08:44,399 --> 01:08:48,800 We have a bunch of more stuff. Is there any benefit do you 980 01:08:48,800 --> 01:08:53,479 think, like building a markdown library in Elixir versus other 981 01:08:55,000 --> 01:09:01,720 technologies or languages? There's another Elixir library, 982 01:09:02,240 --> 01:09:04,840 I think it's called Med if not mistaken. 983 01:09:05,600 --> 01:09:09,840 And it's a really nice idea because it's pretty much even 984 01:09:09,880 --> 01:09:15,960 based, not even like process, but you know, more like a 985 01:09:15,960 --> 01:09:20,960 functional design. So you can leverage this kind of 986 01:09:21,960 --> 01:09:27,880 functional design in mixer. I, I wish I I wish I could read. 987 01:09:28,640 --> 01:09:36,359 I wish I could write a parser, a pure Elixer parser for for MDXI. 988 01:09:36,359 --> 01:09:40,120 Don't know, maybe someday, but what is the? 989 01:09:41,040 --> 01:09:43,319 What's the most difficult part there? 990 01:09:43,319 --> 01:09:48,080 Just finding the time to do it or are there limitations that 991 01:09:48,080 --> 01:09:51,439 you know of? Markdown has many many edge 992 01:09:51,439 --> 01:09:56,280 cases you have to like. The parsing is not easy. 993 01:09:56,280 --> 01:09:59,080 You have to deal with a lot of complication. 994 01:10:00,560 --> 01:10:03,760 Even though you have the specification like you have the 995 01:10:03,760 --> 01:10:11,040 common mark specification, which is like 10-11 years old now 996 01:10:11,120 --> 01:10:15,760 this, this specification, so it is, it is well established, but 997 01:10:15,760 --> 01:10:19,400 there are many sometimes some, some parts of this 998 01:10:19,400 --> 01:10:22,000 specification. You, you have to understand what 999 01:10:22,000 --> 01:10:24,480 they mean. Like it's not strict 1000 01:10:25,600 --> 01:10:29,040 specification sometimes. And there are many edge cases 1001 01:10:29,040 --> 01:10:33,520 that you have to deal with. So if if you look at like if you 1002 01:10:33,520 --> 01:10:37,560 look at the correct source code, it's like really complex source 1003 01:10:37,560 --> 01:10:42,320 code. And the the Conrad author, I 1004 01:10:42,320 --> 01:10:49,600 think they, they used to work on the GitHub specification at 1005 01:10:49,600 --> 01:10:55,440 GitHub and now working at GitLab and GitLab is kind of sponsoring 1006 01:10:55,440 --> 01:11:00,720 the Conrad development. And Conrad is used to generate 1007 01:11:00,720 --> 01:11:06,800 all the Rus docs like like we have the hack stocks where else 1008 01:11:06,800 --> 01:11:11,400 has documentations and they use they use conrac so it's a well 1009 01:11:11,480 --> 01:11:17,520 well established parser. So there are some reasons to 1010 01:11:17,520 --> 01:11:23,800 create a pure Elixir parser, but it's not an easy task. 1011 01:11:24,120 --> 01:11:27,360 Yeah, I'm sure. Sounds like a lot of work. 1012 01:11:29,080 --> 01:11:33,280 Yeah, it it like Conrac. Conrac has a lot of contributors 1013 01:11:33,280 --> 01:11:40,320 now, so it's the only downside of using correct? 1014 01:11:40,600 --> 01:11:43,600 My opinion is really having like. 1015 01:11:44,840 --> 01:11:50,320 It's not a blocker, but it's a REST library so if you want to 1016 01:11:50,320 --> 01:11:54,240 use it you have to either download the pre compiled binary 1017 01:11:54,800 --> 01:11:57,880 or compile compile the REST code. 1018 01:11:58,800 --> 01:12:01,760 Right. Right, which is not a big deal, 1019 01:12:01,760 --> 01:12:05,520 but there might be some scenarios or environments where 1020 01:12:05,720 --> 01:12:10,760 that's not allowed. But otherwise it's fine. 1021 01:12:11,920 --> 01:12:16,400 The the roster pre compiled project is like amazing. 1022 01:12:16,400 --> 01:12:20,640 It's so a lot of you solve this problem of confining the roast 1023 01:12:20,640 --> 01:12:23,120 code. Man you can just have the pre 1024 01:12:23,120 --> 01:12:25,400 compiled binary. You also have to kind of like 1025 01:12:25,400 --> 01:12:27,600 pick and choose how you want to spend your time, right? 1026 01:12:27,600 --> 01:12:30,360 For an open source project that's not really making you any 1027 01:12:30,360 --> 01:12:34,200 money except for, you know, a few sponsorships, Do you really 1028 01:12:34,200 --> 01:12:38,200 want to build a whole parser from scratch or just leverage a 1029 01:12:38,200 --> 01:12:40,320 tool that already works really well, you know? 1030 01:12:41,760 --> 01:12:48,280 Yeah, like I have so many, I have so many ideas to do with 1031 01:12:48,760 --> 01:12:53,200 MDX and Lumis. The next one I'm going to work 1032 01:12:53,200 --> 01:12:58,920 is the Higgs parser, right? Right now we do have Higgs 1033 01:12:58,920 --> 01:13:01,680 support. You can write Higgs components 1034 01:13:01,680 --> 01:13:05,400 and mixer expressions inside the markdown, and it does generate 1035 01:13:05,560 --> 01:13:07,120 a. It compiles. 1036 01:13:09,760 --> 01:13:11,520 I forgot the name of that struct. 1037 01:13:11,520 --> 01:13:16,280 Nowadays struct that is used by live view, so you can just bring 1038 01:13:16,280 --> 01:13:21,560 their mark down side live view page like like if you had a hex 1039 01:13:21,560 --> 01:13:24,960 template. So you can have components and 1040 01:13:24,960 --> 01:13:31,680 expressions inside markdown. But there's a problem the when 1041 01:13:31,680 --> 01:13:36,480 you the way it works now, it just identifies the boundaries 1042 01:13:36,640 --> 01:13:40,440 like what is markdown, what is hex inside the markdown 1043 01:13:40,440 --> 01:13:47,440 document, and it keeps the Higgs part out of the markdown parser. 1044 01:13:48,240 --> 01:13:52,000 And then we do a second pass compilation of using the Higgs 1045 01:13:52,000 --> 01:13:55,200 compiler. So first it parse the markdown, 1046 01:13:55,560 --> 01:13:58,800 keep the Higgs like, does that change that part? 1047 01:13:59,680 --> 01:14:02,040 And then the second pass with the Higgs compiler. 1048 01:14:02,520 --> 01:14:08,120 The problem though is the inner content of those Higgs parts, 1049 01:14:08,120 --> 01:14:11,640 Higgs blocks. We need to parse that as 1050 01:14:11,640 --> 01:14:14,440 markdown as well. Because if you have markdown 1051 01:14:15,040 --> 01:14:17,960 inside the component, which is inside the markdown inside 1052 01:14:17,960 --> 01:14:20,280 another component, you know you can be nested. 1053 01:14:20,960 --> 01:14:25,600 We need to go down all the children and parse those small 1054 01:14:25,600 --> 01:14:32,000 pieces of markdown as well. So that's how other libraries 1055 01:14:32,000 --> 01:14:37,920 work like MDX, the JavaScript library and that that was the 1056 01:14:37,920 --> 01:14:42,200 main inspiration for MDEX. There's MDXJS. 1057 01:14:43,000 --> 01:14:48,000 They work this way if you have JavaScript inside components and 1058 01:14:48,000 --> 01:14:51,960 markdown as well. It can it, it does render the 1059 01:14:51,960 --> 01:14:56,400 markdown, it does parse the markdown inside components and 1060 01:14:56,400 --> 01:15:00,440 that's really hard to do. I'm working on the Rose code. 1061 01:15:00,440 --> 01:15:03,640 Make that happen on the current parser. 1062 01:15:04,440 --> 01:15:09,400 As I get that done, I can just use it on on the MDX. 1063 01:15:10,320 --> 01:15:16,760 And then finally I can say the Heeks support is like 100% done 1064 01:15:16,800 --> 01:15:20,400 and working. So we can have Mcdowie Felixer 1065 01:15:21,600 --> 01:15:23,920 for, you know, for Elixer projects. 1066 01:15:24,360 --> 01:15:32,480 Yeah, that's awesome and right. So right now Heeks isn't fully 1067 01:15:32,480 --> 01:15:36,440 supported or it is. It works, but if you have 1068 01:15:36,440 --> 01:15:41,760 markdown side the components it's going to be like just raw 1069 01:15:41,760 --> 01:15:44,280 text. So OK, I see. 1070 01:15:45,400 --> 01:15:49,000 Because what it does is all the text, you know all the 1071 01:15:49,000 --> 01:15:54,760 information inside the, you know the block is just hex. 1072 01:15:55,160 --> 01:16:00,080 And then when you send that to the hex compiler, if you have 1073 01:16:00,080 --> 01:16:05,240 like a title or a bold text or something like a markdown text, 1074 01:16:06,280 --> 01:16:10,640 the H compiler has no idea that's a markdown, right? 1075 01:16:10,840 --> 01:16:14,720 Then it's just reindeers as a a raw text. 1076 01:16:15,880 --> 01:16:17,680 I see. So that's that's the problem we 1077 01:16:17,680 --> 01:16:20,280 have right now. I'm trying to solve on the 1078 01:16:20,320 --> 01:16:25,680 correct side first the the worst Spicer and then I can use them 1079 01:16:25,680 --> 01:16:28,000 in the X. OK. 1080 01:16:28,360 --> 01:16:31,640 That's cool, yeah. So yeah, you never would think 1081 01:16:31,760 --> 01:16:35,520 the complicated under the hood problems you have. 1082 01:16:37,720 --> 01:16:39,440 And. Now everyone knows that listens 1083 01:16:39,440 --> 01:16:42,000 to this podcast. They know how hard creating a 1084 01:16:42,000 --> 01:16:46,440 markdown library is. Yeah, it's no easy task like and 1085 01:16:46,440 --> 01:16:50,640 that's why maybe I, I mean, again, I should be talking more 1086 01:16:50,640 --> 01:16:55,840 about those those problems because you know, it's really 1087 01:16:55,840 --> 01:17:01,440 fun to solve those problems, but it does take time and you know, 1088 01:17:01,720 --> 01:17:04,520 I could be sharing more about this those problems. 1089 01:17:05,040 --> 01:17:08,960 Yeah, yeah, I know it's good and I'm glad that you're on on the 1090 01:17:08,960 --> 01:17:12,760 podcast talking about it. And I just have to have you back 1091 01:17:12,760 --> 01:17:16,520 on more often to keep keep us updated on all your projects. 1092 01:17:17,240 --> 01:17:22,000 Yeah, that would be nice because like if I get the three seater, 1093 01:17:22,240 --> 01:17:26,680 you know loom is alone has a bunch of complex like 3 seater 1094 01:17:26,680 --> 01:17:30,400 Novi cleaning that we need to extract and now wasn't 1095 01:17:30,400 --> 01:17:35,160 compilation, you know bunch of bunch of stuff happening there. 1096 01:17:36,040 --> 01:17:41,000 So you can so you can just to have like nice colors. 1097 01:17:41,160 --> 01:17:46,280 The source code you know the the the output seems simple but the 1098 01:17:46,440 --> 01:17:49,720 out engineering for is the hard part. 1099 01:17:49,920 --> 01:18:00,520 Yeah, for sure. And for for MDX, you know, MDX 1100 01:18:00,520 --> 01:18:07,080 has other features as well, but the main, one of the main 1101 01:18:07,080 --> 01:18:10,600 concerns is also the performance. 1102 01:18:11,560 --> 01:18:16,800 We, we had this problem back in the day with Beacon because if 1103 01:18:17,280 --> 01:18:21,960 because we had to compile a bunch of pages at the same time, 1104 01:18:22,920 --> 01:18:24,800 right? So if you take like a second, 1105 01:18:25,880 --> 01:18:30,600 like a second to compile a page, compile, I mean from markdown to 1106 01:18:30,600 --> 01:18:35,560 HTML to live view. If you take a second or two or 1107 01:18:35,560 --> 01:18:39,840 whatever and you have like 100 or 1000 pages, performance 1108 01:18:39,840 --> 01:18:43,040 matters. And that was one of the reasons 1109 01:18:43,200 --> 01:18:49,160 I chose Korac as well because you know, on a benchmark it's 1110 01:18:49,200 --> 01:18:54,560 way, way faster than any other Elixir library. 1111 01:18:55,240 --> 01:18:59,280 And and Korac is not even the fastest implementation. 1112 01:18:59,520 --> 01:19:05,280 There are others way even more faster than than Korac, but 1113 01:19:05,280 --> 01:19:08,720 conrac is is fast is fastest, you know. 1114 01:19:09,640 --> 01:19:12,400 So what's the fastest? Like breaking down to like AC 1115 01:19:12,400 --> 01:19:14,480 language or something like that or getting. 1116 01:19:15,560 --> 01:19:19,520 Yeah, there is. I mean again there's there's the 1117 01:19:19,640 --> 01:19:24,920 I mean there is a pull down C mark, which I believe is faster 1118 01:19:24,920 --> 01:19:31,120 than Conrac, OK. There is also Med for C which is 1119 01:19:31,120 --> 01:19:35,800 marked down for C It's also really really fast. 1120 01:19:36,680 --> 01:19:40,040 But the problem is correct. It was the only one that had 1121 01:19:40,040 --> 01:19:46,360 extensions that I could support GitHub and like discord, flavor 1122 01:19:46,360 --> 01:19:51,240 markdown, GitLab as well. And the IT had extensions so I 1123 01:19:51,240 --> 01:19:55,080 could plug in a syntex highlighter and a bunch of other 1124 01:19:55,080 --> 01:19:59,240 stuff. So and since the outer used to 1125 01:19:59,240 --> 01:20:02,480 work. On the specification, the GitHub 1126 01:20:02,480 --> 01:20:06,440 specification and now working on the GitLab specification. 1127 01:20:08,160 --> 01:20:10,720 I mean, they, they, they really know how it works. 1128 01:20:10,840 --> 01:20:13,120 So they have hands on experience. 1129 01:20:13,640 --> 01:20:18,840 So it was the safest choice. And I was glad to yeah. 1130 01:20:19,560 --> 01:20:24,920 And I was glad to know they, the rest team chose correct to 1131 01:20:24,920 --> 01:20:27,440 generate the the documentation as well. 1132 01:20:27,440 --> 01:20:32,440 So there was like a, you know, prove the the chart that you 1133 01:20:32,440 --> 01:20:35,800 know, that was a good choice. Yeah, cool. 1134 01:20:38,160 --> 01:20:41,280 Well, but I, I really appreciate you coming on and talking about 1135 01:20:41,680 --> 01:20:45,240 MDXI think we can like slowly start to wrap this up, but we 1136 01:20:45,240 --> 01:20:49,280 have a pretty good question and I think both of us can answer. 1137 01:20:49,280 --> 01:20:51,880 It has nothing to do with what we were talking about, but I 1138 01:20:51,880 --> 01:20:56,120 feel like we've both been on this developer journey. 1139 01:20:56,480 --> 01:21:00,960 And so Desmond asks, how has the journey been transitioning from 1140 01:21:00,960 --> 01:21:03,320 a junior to mid, then senior dev? 1141 01:21:03,720 --> 01:21:06,200 Did you have mentors along the way? 1142 01:21:06,600 --> 01:21:09,520 And then how about like coding in multiple languages? 1143 01:21:12,360 --> 01:21:16,040 Oh yeah, that's a really. I mean, we could record a 1144 01:21:16,040 --> 01:21:18,040 podcast just. I know we really could. 1145 01:21:19,080 --> 01:21:21,880 We could I'll, I'll give it a stab. 1146 01:21:22,000 --> 01:21:25,160 I, I think one, it's really important to have mentors, 1147 01:21:25,160 --> 01:21:26,600 right? You need to work with people 1148 01:21:26,600 --> 01:21:31,320 that are smarter than you and, but also support you like 1149 01:21:31,400 --> 01:21:34,640 learning and kind of like trust in your ability to learn and 1150 01:21:34,640 --> 01:21:38,760 produce, right. And I would say if I were to 1151 01:21:38,760 --> 01:21:42,440 think back when I first started, I was a very junior developer. 1152 01:21:42,440 --> 01:21:46,080 I was mostly like an iOS developer dipping my toes into 1153 01:21:46,200 --> 01:21:48,800 like node JS and some back end API stuff. 1154 01:21:49,680 --> 01:21:53,880 And the second I kind of felt like I could build a full app, 1155 01:21:53,880 --> 01:21:56,240 right? I can do the front end app work 1156 01:21:56,480 --> 01:22:01,360 and the back end rest APII was really cocky and thought like I 1157 01:22:01,360 --> 01:22:04,840 knew everything. And I think like I talk about it 1158 01:22:04,840 --> 01:22:07,320 a lot on this podcast, but the Dunning Kruger curve. 1159 01:22:08,760 --> 01:22:11,840 I think juniors have a really good habit of once they feel 1160 01:22:11,840 --> 01:22:14,160 like they can build a whole app, you're at the top of mount 1161 01:22:14,160 --> 01:22:17,280 stupid, but you like you feel like you could like conquer the 1162 01:22:17,280 --> 01:22:19,720 world, right? And then you Start learning 1163 01:22:19,720 --> 01:22:21,680 about like security vulnerabilities. 1164 01:22:21,680 --> 01:22:25,000 Your app might have gotten hacked and yet you fall down 1165 01:22:25,000 --> 01:22:26,840 real quick and you're like humbled. 1166 01:22:27,560 --> 01:22:31,640 And then from that point on, I would say you start becoming mid 1167 01:22:31,640 --> 01:22:33,960 and you slowly are building your way up to a senior. 1168 01:22:34,320 --> 01:22:40,840 But you always know how much more you don't know the more you 1169 01:22:40,840 --> 01:22:42,960 learn, right? So I feel like you're always on 1170 01:22:42,960 --> 01:22:45,720 this chase to know it all and it's impossible. 1171 01:22:45,720 --> 01:22:48,680 You never will. Yeah, it's it's interesting 1172 01:22:48,680 --> 01:22:52,080 because when you are junior, you don't know much you so you can 1173 01:22:52,280 --> 01:22:56,360 you can not do much yet. And then you Start learning and 1174 01:22:56,360 --> 01:23:00,040 then you start producing a lot of code, a lot of stuff. 1175 01:23:00,560 --> 01:23:04,000 And then when you are the middle is like you are on the top of 1176 01:23:04,000 --> 01:23:09,080 the curve like over engineering everything, like writing a lot 1177 01:23:09,080 --> 01:23:12,720 of stuff. And then senior for me is when 1178 01:23:12,720 --> 01:23:17,840 you realize that you don't like cold is not the end goal like 1179 01:23:18,240 --> 01:23:21,240 coding is just, you know, part of the process. 1180 01:23:21,680 --> 01:23:24,960 And then removing code is actually better than adding 1181 01:23:24,960 --> 01:23:29,440 code, which is kind of the opposite when you're learning. 1182 01:23:29,880 --> 01:23:33,200 But there is this difference when you get into like more 1183 01:23:33,200 --> 01:23:39,240 senior position and it. Becomes the simplest solution as 1184 01:23:39,240 --> 01:23:42,080 the best solution, not the over engineered like. 1185 01:23:42,080 --> 01:23:46,360 I feel like there's a weird thing and among engineers, not 1186 01:23:46,360 --> 01:23:50,960 just software engineers, but you, you want to like flex, how 1187 01:23:50,960 --> 01:23:53,720 smart you are with a real complex solution. 1188 01:23:53,880 --> 01:23:56,240 And that's the last thing you want in your code base is a 1189 01:23:56,240 --> 01:23:59,200 complex solution. You want the simplest version of 1190 01:23:59,200 --> 01:24:01,080 that solution. Yeah. 1191 01:24:01,360 --> 01:24:04,680 And two more things. One about mentorship is it's 1192 01:24:04,680 --> 01:24:07,480 really, really good to have mentorship, but don't wait for 1193 01:24:07,640 --> 01:24:11,000 mentorship. You have to build something like 1194 01:24:11,120 --> 01:24:13,600 yeah, you don't need, you don't need permission to build 1195 01:24:13,600 --> 01:24:15,280 something. You can just build something, 1196 01:24:15,680 --> 01:24:19,920 publish like just you know, that's true for our concerns, 1197 01:24:19,920 --> 01:24:21,760 for mentorship. Don't wait. 1198 01:24:22,200 --> 01:24:25,440 Just go and build something. Yeah, like I was really 1199 01:24:25,440 --> 01:24:27,800 fortunate. Like I, I worked like, you know, 1200 01:24:27,800 --> 01:24:32,760 I like my first developer job. I just was like, you know, the 1201 01:24:32,800 --> 01:24:35,760 CEO liked me, he liked my drive and motivation. 1202 01:24:35,760 --> 01:24:39,120 So I kind of like luckily fell into that and had like a really 1203 01:24:39,120 --> 01:24:42,760 good mentor there. But at the same time, like if 1204 01:24:42,760 --> 01:24:45,960 you don't have that, say you don't have a dev job, but you 1205 01:24:45,960 --> 01:24:50,600 have a, an idea, build that. And then, you know, watch 1206 01:24:50,800 --> 01:24:54,720 podcasts like this, watch my tutorials, follow other people 1207 01:24:54,720 --> 01:24:56,880 building stuff in the Elixir community. 1208 01:24:57,080 --> 01:25:00,160 And those people become your mentors and you can learn from 1209 01:25:00,160 --> 01:25:02,640 them 'cause they, they'll post about either what they're 1210 01:25:02,640 --> 01:25:05,200 building, maybe a mistake that they made. 1211 01:25:05,240 --> 01:25:08,920 And like so you have it, you're just acquiring it a little 1212 01:25:08,920 --> 01:25:12,600 differently. Yeah, and don't be afraid to do 1213 01:25:12,600 --> 01:25:17,880 this stuff like 2019. I had a talk at Elixir Conf and 1214 01:25:17,880 --> 01:25:22,600 the talk start with a mistake I made like really, really hooky 1215 01:25:22,600 --> 01:25:27,560 mistake. I was trying to contribute just 1216 01:25:27,560 --> 01:25:32,280 a fix on the documentation on on their Airlink and I completely 1217 01:25:32,280 --> 01:25:36,560 messed up the geek, you know, the geek branches or whatever. 1218 01:25:37,200 --> 01:25:42,040 It was like a really rookie mistake and like it happens. 1219 01:25:42,240 --> 01:25:45,400 And that was part of my talk because I was trying to say, 1220 01:25:45,400 --> 01:25:49,800 hey, just go and do stuff and. You know, just. 1221 01:25:50,080 --> 01:25:53,080 Don't don't need to wait. And you know, some, sometimes 1222 01:25:53,080 --> 01:25:56,520 something bad happens, but it's, it's fine. 1223 01:25:57,480 --> 01:26:00,800 Yeah, I mean, it doesn't matter where you are on your career. 1224 01:26:00,800 --> 01:26:04,960 Like you can be a staff level engineer and you're still going 1225 01:26:04,960 --> 01:26:06,760 to run into things you've never seen before. 1226 01:26:08,080 --> 01:26:10,960 Yeah. I feel like what makes what what 1227 01:26:10,960 --> 01:26:14,280 levels up an engineer is the ability to solve a problem that 1228 01:26:14,280 --> 01:26:16,440 they've never seen before, you know? 1229 01:26:16,840 --> 01:26:21,680 Yeah, yeah, it's I feel like that now working with me was me 1230 01:26:22,240 --> 01:26:27,680 stuff like I I have been reading about was me for a long time, 1231 01:26:27,680 --> 01:26:30,360 but never use it. So now I I feel like a junior 1232 01:26:30,440 --> 01:26:34,120 working was me because that's new to me, right. 1233 01:26:34,200 --> 01:26:39,920 And that's, that's, that's fine. Yeah, I mean, I'm so like to go 1234 01:26:39,920 --> 01:26:42,640 on his like, what about coding in multiple languages? 1235 01:26:43,080 --> 01:26:46,560 I feel like the the basic principles of programming, 1236 01:26:46,560 --> 01:26:49,640 they're all the same. You're just changing syntax 1237 01:26:49,640 --> 01:26:52,480 essentially, right? I mean object oriented 1238 01:26:52,480 --> 01:26:55,800 programming versus functional and immutable obviously have 1239 01:26:55,800 --> 01:26:58,840 their differences but for the most part like writing a 1240 01:26:58,840 --> 01:27:05,000 programming language comes down to syntax and I context switch a 1241 01:27:05,000 --> 01:27:08,320 lot less. Now I write a lot of Elixir and 1242 01:27:08,520 --> 01:27:11,160 a little bit of JavaScript. I don't do a lot of other things 1243 01:27:11,160 --> 01:27:15,720 anymore, but when I was jumping between like Swift, JavaScript, 1244 01:27:16,200 --> 01:27:20,280 C#, an that was the at the beginning and like, I learned a 1245 01:27:20,280 --> 01:27:21,560 lot. I'm real thankful. 1246 01:27:21,560 --> 01:27:25,640 But it, it, it wears on you, but at the same time you just end up 1247 01:27:25,640 --> 01:27:29,720 like I like, I would like forget how to do like for loops in 1248 01:27:29,920 --> 01:27:31,880 languages, right? Because I was so I had like 1249 01:27:31,880 --> 01:27:35,720 Google syntax all the time. But like for the most part, like 1250 01:27:35,720 --> 01:27:39,360 the concept of programming is relatively the same between all 1251 01:27:39,360 --> 01:27:40,800 languages. You just will. 1252 01:27:41,520 --> 01:27:44,840 You just have to remember the syntax or like the language you 1253 01:27:44,840 --> 01:27:48,280 know. Understand the fundamentals and 1254 01:27:48,280 --> 01:27:50,640 you'll be fine. You can jump into another 1255 01:27:50,640 --> 01:27:55,320 language and you'll be fine. I guess for me the the hardest 1256 01:27:55,320 --> 01:27:59,160 like move was from Ruby and Rails to Elixir. 1257 01:27:59,640 --> 01:28:04,160 My first year or so working for Elixir I was still trying to 1258 01:28:04,160 --> 01:28:08,360 write like objects, like an instance and state and using 1259 01:28:08,360 --> 01:28:12,560 gene servers as a class. You know, that kind of, it was 1260 01:28:12,600 --> 01:28:15,400 a, it was a mess. Coming from an object oriented 1261 01:28:15,400 --> 01:28:18,400 programming language, like that's all I knew. 1262 01:28:18,640 --> 01:28:20,680 And then I came to Elixir. It was hard. 1263 01:28:20,800 --> 01:28:22,760 It's hard to change your mindset. 1264 01:28:23,520 --> 01:28:26,280 Yeah. But also it once it clicks, it 1265 01:28:26,360 --> 01:28:30,440 makes it's it's way easier I think. 1266 01:28:30,720 --> 01:28:33,760 I believe like I have a feeling that Elixir is way easier than 1267 01:28:33,760 --> 01:28:38,240 other languages, but you have first the functional program has 1268 01:28:38,360 --> 01:28:41,320 clicked on your mind first and then I. 1269 01:28:41,320 --> 01:28:45,520 Just if you started with it, that's how your brain works 1270 01:28:45,520 --> 01:28:47,160 solving problems naturally, right? 1271 01:28:47,160 --> 01:28:50,200 Like you kind of transform a thing and pass it along. 1272 01:28:50,800 --> 01:28:54,160 But once you have objects oriented cloud in your brain, 1273 01:28:54,400 --> 01:28:57,520 it's hard to go back to that and relearn that that way of. 1274 01:28:57,520 --> 01:29:02,320 Thinking yeah, but if you start with functional programming you 1275 01:29:02,320 --> 01:29:05,280 have to learn Ruby shack oriented program as well because 1276 01:29:05,560 --> 01:29:08,680 you become a better better engineer when you understand 1277 01:29:08,680 --> 01:29:10,640 both they are. Completely. 1278 01:29:10,920 --> 01:29:13,960 They're not. It's not an alternative to 1 to 1279 01:29:13,960 --> 01:29:15,200 the. No, you can. 1280 01:29:15,320 --> 01:29:17,480 You can learn from both and use both. 1281 01:29:18,160 --> 01:29:20,840 I agree depend on the language but the concept. 1282 01:29:20,840 --> 01:29:23,560 The idea is you have you need to move. 1283 01:29:24,560 --> 01:29:28,200 Yeah, well, like just the idea. Well, when you're in the object 1284 01:29:28,200 --> 01:29:31,920 oriented world, you have a lot more architectural patterns, 1285 01:29:31,920 --> 01:29:34,760 right? And you know, you learn about 1286 01:29:34,760 --> 01:29:39,720 like singletons and you can pass state around and objects and you 1287 01:29:39,720 --> 01:29:44,520 know it has its superpowers, but it also leads to a lot of really 1288 01:29:44,520 --> 01:29:49,840 hard traceable bugs. Yeah, that that's one thing with 1289 01:29:49,840 --> 01:29:53,520 functional programming, at least for me, it leads to a simple 1290 01:29:53,520 --> 01:29:57,160 code way more than object program, Yeah. 1291 01:29:57,320 --> 01:29:59,360 It sets you up to write clean code. 1292 01:30:00,920 --> 01:30:02,760 Yeah. And think of not in the sense of 1293 01:30:02,760 --> 01:30:08,480 like the clean organization style, like I mean clean coding, 1294 01:30:08,480 --> 01:30:11,440 like writing less. Code like the simplest. 1295 01:30:11,520 --> 01:30:13,280 Yeah, the less code. Simplest. 1296 01:30:13,320 --> 01:30:16,040 Yeah. Simplest approach to a solution, 1297 01:30:16,120 --> 01:30:18,920 right? Yeah, and I'm not. 1298 01:30:19,000 --> 01:30:23,600 Yeah, that reminds me of another stuff like when I was every time 1299 01:30:23,600 --> 01:30:26,280 I tried to contribute something, I tried to learn from the 1300 01:30:26,280 --> 01:30:29,520 maintainer. And you can learn from people 1301 01:30:29,520 --> 01:30:32,400 like Joseph Alim, Charisma Card and others. 1302 01:30:33,240 --> 01:30:37,120 They have like a really good approach about, you know, 1303 01:30:37,680 --> 01:30:41,480 thinking design and, you know, realizing what is needed and 1304 01:30:41,480 --> 01:30:44,960 what is not needed, which is probably more important. 1305 01:30:45,640 --> 01:30:48,800 And then you can learn. And I see that, you know those 1306 01:30:48,840 --> 01:30:52,680 guys, they try to make the most simple solution and the most 1307 01:30:52,680 --> 01:30:56,800 effective one, which is really, really hard, but you can learn. 1308 01:30:57,680 --> 01:31:01,560 Yeah, I think it's harder to know what solutions shouldn't 1309 01:31:01,560 --> 01:31:05,720 even be brought into a conversation to like, you know, 1310 01:31:05,720 --> 01:31:09,640 how many features get thrown at any kind of like open source 1311 01:31:09,640 --> 01:31:13,600 project or language or anything, honestly, even like even if 1312 01:31:13,600 --> 01:31:16,640 you're building a SAS product, right, you know how many feature 1313 01:31:16,640 --> 01:31:20,440 requests you're going to get and the ability to know what 1314 01:31:20,520 --> 01:31:24,000 strengthens your product versus what, doesn't it? 1315 01:31:24,320 --> 01:31:26,200 I think that's the probably the hardest piece. 1316 01:31:26,920 --> 01:31:31,960 Yeah, the the balance of, you know, the amount of code you 1317 01:31:31,960 --> 01:31:37,280 need to write to support new features because the code has to 1318 01:31:37,280 --> 01:31:42,280 be flexible enough to grow and support new features, but not 1319 01:31:42,280 --> 01:31:45,800 over complex. So this balance is really, 1320 01:31:45,800 --> 01:31:48,760 really hard to get. You know, you have to write just 1321 01:31:48,760 --> 01:31:54,520 the minimum amount of code that is flexible, scarce enough to 1322 01:31:54,520 --> 01:31:57,840 continue. But that's really hard to do 1323 01:31:57,880 --> 01:32:01,120 that. I don't know if there's formal 1324 01:32:01,120 --> 01:32:05,640 education on that, but at least for me that was like like years 1325 01:32:05,640 --> 01:32:10,200 of experience, making mistakes, really time. 1326 01:32:10,200 --> 01:32:13,240 In the saddle, making mistakes and learning from them. 1327 01:32:13,240 --> 01:32:20,640 Like I I think my most like leveling up events in my life 1328 01:32:20,640 --> 01:32:26,600 were really bad mistakes. Yeah, you, you can learn from 1329 01:32:26,600 --> 01:32:29,960 manufacturing. Or hearing about other people's 1330 01:32:29,960 --> 01:32:35,520 mistakes, like I just had Thomas from level all on and you know, 1331 01:32:35,560 --> 01:32:41,120 staff level engineer, like super smart guy and they just had like 1332 01:32:41,120 --> 01:32:46,560 a rate limiting issue with Oauth and like so all their users 1333 01:32:46,560 --> 01:32:49,920 couldn't use their platform all of a sudden. 1334 01:32:50,160 --> 01:32:52,320 And it's like one of those things where you don't think 1335 01:32:52,320 --> 01:32:56,880 about and like they, they have like a school system platform, 1336 01:32:56,880 --> 01:32:58,160 right? A learning platform. 1337 01:32:58,560 --> 01:33:01,960 And all the students from a school district were on the same 1338 01:33:01,960 --> 01:33:05,480 IP address. And so they, they hit a problem 1339 01:33:05,480 --> 01:33:08,240 that they never knew they had because they were hitting a rate 1340 01:33:08,240 --> 01:33:12,000 limit on authentication. And that's just like things like 1341 01:33:12,000 --> 01:33:14,320 that. Yeah, that's another way to 1342 01:33:14,320 --> 01:33:21,000 learn is exposure to real world situations like, and you know, 1343 01:33:21,160 --> 01:33:24,160 if you're a junior, you're not going to have so much exposure, 1344 01:33:24,160 --> 01:33:28,440 but you can like other people do and you can learn from them. 1345 01:33:28,440 --> 01:33:30,600 And then you grow, you keep going. 1346 01:33:30,960 --> 01:33:35,120 And as you get more exposure and that's where you you learn. 1347 01:33:35,640 --> 01:33:40,080 So more conversations and like meetups are really important. 1348 01:33:40,080 --> 01:33:43,120 Go meet people, talk about your projects, talk about your 1349 01:33:43,120 --> 01:33:45,760 problems. And you know 9 times out of 10 1350 01:33:45,760 --> 01:33:49,840 someone else in that room that's listening already have the same 1351 01:33:49,840 --> 01:33:52,560 problem you're experiencing and they can tell you how they 1352 01:33:52,560 --> 01:33:55,760 solved it. Yeah, we have a really good 1353 01:33:56,720 --> 01:34:02,680 elixir Metap nearby in Toronto and they, you know, we have 1354 01:34:02,680 --> 01:34:05,680 metaps, I don't know, twice a year, maybe not. 1355 01:34:05,800 --> 01:34:09,080 It's not so frequent, but every time it's, you know, packing 1356 01:34:09,080 --> 01:34:12,240 with a lot of people and really good talks. 1357 01:34:12,240 --> 01:34:16,640 It's another way that you can, you know, you can, you can also 1358 01:34:16,800 --> 01:34:20,720 share like you can give a talk in the meet up if you, you know, 1359 01:34:20,840 --> 01:34:24,840 if you feel like it's another way to, I feel it's like another 1360 01:34:24,840 --> 01:34:29,560 way to learn as well is sharing about what you know, sharing 1361 01:34:29,560 --> 01:34:33,680 about your experience. So yeah, there are many ways to 1362 01:34:33,680 --> 01:34:38,480 grow in the career, but yeah, it's a very long. 1363 01:34:38,920 --> 01:34:40,880 Talk. Yeah, you just got to keep doing 1364 01:34:40,880 --> 01:34:42,960 it, keep showing up every day, right? 1365 01:34:42,960 --> 01:34:45,640 And then eventually, before you know it, you're like, wow, I've 1366 01:34:45,640 --> 01:34:48,720 learned a lot and I still don't know a lot. 1367 01:34:48,920 --> 01:34:51,160 And you just kind of have to accept that. 1368 01:34:51,160 --> 01:34:53,840 And then you just. That's it. 1369 01:34:53,840 --> 01:34:56,120 Yeah, it yeah. It's impossible to know 1370 01:34:56,360 --> 01:34:59,160 everything. If someone says he knows 1371 01:34:59,160 --> 01:35:01,920 everything, he's lying because it's impossible. 1372 01:35:02,040 --> 01:35:03,840 I'm going to tell you where he is right now. 1373 01:35:03,840 --> 01:35:08,360 He's on the top of the Dunning Crew Kruger curve if someone 1374 01:35:08,360 --> 01:35:12,880 claims to know it all. Yeah, yeah. 1375 01:35:12,960 --> 01:35:15,800 We are always learning. For sure. 1376 01:35:15,880 --> 01:35:20,360 Absolutely. All right, Leandro, thank you so 1377 01:35:20,360 --> 01:35:24,720 much for joining me. This was a fun conversation as 1378 01:35:24,720 --> 01:35:26,920 always. And yeah, anytime you want to 1379 01:35:26,920 --> 01:35:28,480 come back on, just please reach out. 1380 01:35:29,520 --> 01:35:34,360 Yeah, thank you for having me. Thanks to everyone for tuning in 1381 01:35:34,360 --> 01:35:37,120 and listening, I appreciate it and hope you guys have a good 1382 01:35:37,120 --> 01:35:42,840 weekend testing.