WEBVTT 00:00:00.160 --> 00:00:01.600 You look like you've been in the sun. 00:00:01.600 --> 00:00:02.080 >> I do. 00:00:02.080 --> 00:00:03.439 >> Or maybe it's just your camera. I don't 00:00:03.439 --> 00:00:04.000 know. 00:00:04.000 --> 00:00:06.000 >> I'm using my iPhone camera, so I 00:00:06.000 --> 00:00:07.759 shouldn't wobble when I move the move 00:00:07.759 --> 00:00:08.559 the desk. 00:00:08.559 --> 00:00:10.160 >> So, sorry to 00:00:10.160 --> 00:00:11.679 >> to be so late. I'm good. 00:00:11.679 --> 00:00:13.280 >> That's all right. I mean, to be honest, 00:00:13.280 --> 00:00:16.960 I was just reading about Opus 4.6 and 00:00:16.960 --> 00:00:20.640 Android agent CLI and the new codec Mac 00:00:20.640 --> 00:00:23.199 OS app. And there's there's always quite 00:00:23.199 --> 00:00:24.400 a lot of news in the morning, isn't 00:00:24.400 --> 00:00:24.720 there? 00:00:24.720 --> 00:00:26.080 >> I haven't really been reading anything 00:00:26.080 --> 00:00:30.160 for the last six days. I was so sick. I 00:00:30.160 --> 00:00:32.000 didn't feel like reading anything. 00:00:32.000 --> 00:00:33.920 >> Yeah, you're pretty quiet. You're better 00:00:33.920 --> 00:00:34.800 now or 00:00:34.800 --> 00:00:36.320 >> Yeah, hopefully. 00:00:36.320 --> 00:00:37.280 >> Surprisingly, 00:00:37.280 --> 00:00:40.239 >> hopefully you're not full of FOMO and 00:00:40.239 --> 00:00:41.760 you're feeling positive and things like 00:00:41.760 --> 00:00:42.000 that. 00:00:42.000 --> 00:00:44.320 >> I mean, I was just scrolling on Reddit 00:00:44.320 --> 00:00:45.600 and seeing how many people are 00:00:45.600 --> 00:00:49.120 complaining about the usage limits and 00:00:49.120 --> 00:00:51.039 model performance degradation and the 00:00:51.039 --> 00:00:54.000 director of AMD doing a full breakdown 00:00:54.000 --> 00:00:58.079 of the sessions of Opus 4.6 six being 00:00:58.079 --> 00:00:59.920 deprecated and then apparently the 00:00:59.920 --> 00:01:02.320 caching being reduced from 1 hour to 5 00:01:02.320 --> 00:01:04.080 minutes in an update in what was it 00:01:04.080 --> 00:01:06.159 Febru was it March 3rd or February 3rd I 00:01:06.159 --> 00:01:06.560 don't know 00:01:06.560 --> 00:01:09.360 >> yeah I didn't understand what that meant 00:01:09.360 --> 00:01:11.439 what does caching mean in the scheme of 00:01:11.439 --> 00:01:12.560 things 00:01:12.560 --> 00:01:15.119 >> if I understand correctly it means that 00:01:15.119 --> 00:01:17.680 entropic is keeping things like like 00:01:17.680 --> 00:01:19.840 answers or tokens that it has already 00:01:19.840 --> 00:01:22.560 processed in cash much longer like for 1 00:01:22.560 --> 00:01:23.920 hour that means they have to store a lot 00:01:23.920 --> 00:01:25.759 more on their side right so if they 00:01:25.759 --> 00:01:28.799 reduce it to 5 It means that things that 00:01:28.799 --> 00:01:31.040 like are no longer in the cache and you 00:01:31.040 --> 00:01:33.520 are being instead I think it costs less 00:01:33.520 --> 00:01:35.759 to get a cache response than for it to 00:01:35.759 --> 00:01:37.439 be reprocessed and regenerate the 00:01:37.439 --> 00:01:39.600 response and so that's where the usage 00:01:39.600 --> 00:01:41.759 was being affected or the performance 00:01:41.759 --> 00:01:44.159 was also worse. I mean that's how caches 00:01:44.159 --> 00:01:46.320 work, right? You either recalculate the 00:01:46.320 --> 00:01:48.880 result or you fetch it from cache and it 00:01:48.880 --> 00:01:50.320 doesn't cost any compute. 00:01:50.320 --> 00:01:53.119 >> I mean I I didn't notice anything except 00:01:53.119 --> 00:01:55.040 that people were moaning about it 00:01:55.040 --> 00:01:57.200 >> like proper upset. Yeah, morning is the 00:01:57.200 --> 00:02:00.399 word is the right word. Oh my goodness. 00:02:00.399 --> 00:02:02.479 >> There there surely were other things to 00:02:02.479 --> 00:02:05.920 talk about 4.7. What do you what have 00:02:05.920 --> 00:02:07.600 people been saying about 4.7? 00:02:07.600 --> 00:02:09.599 >> I don't even want to read it because 00:02:09.599 --> 00:02:11.840 people are so 00:02:11.840 --> 00:02:13.599 how you say they want to give their hot 00:02:13.599 --> 00:02:14.640 takes I suppose. 00:02:14.640 --> 00:02:17.840 >> Yeah. And people will just constantly 00:02:17.840 --> 00:02:21.120 say like well they degraded 4.6 so hard 00:02:21.120 --> 00:02:24.400 so that everyone will be happy with 4.7 00:02:24.400 --> 00:02:25.920 whereas the performance is just similar 00:02:25.920 --> 00:02:28.160 and blah blah blah like all this every 00:02:28.160 --> 00:02:30.480 single time people come up with the same 00:02:30.480 --> 00:02:33.680 theories and 00:02:33.680 --> 00:02:35.680 >> yeah there's a there's a lot of like 00:02:35.680 --> 00:02:37.840 >> I don't think it's completely untrue I 00:02:37.840 --> 00:02:40.160 think there is some I mean obviously 00:02:40.160 --> 00:02:42.000 there are things that weren't working 00:02:42.000 --> 00:02:44.400 well and people were complaining and 00:02:44.400 --> 00:02:45.920 then people are like oh it works fine 00:02:45.920 --> 00:02:47.360 for me and then the next day they're 00:02:47.360 --> 00:02:49.120 like oh my god I said it was fine for me 00:02:49.120 --> 00:02:51.760 and now I'm affected. And then 00:02:51.760 --> 00:02:53.599 >> and then other people coming into Reddit 00:02:53.599 --> 00:02:55.200 saying like, "Ah, I told you guys you 00:02:55.200 --> 00:02:56.319 didn't believe me, but now you're 00:02:56.319 --> 00:02:57.840 affected." And I'm just reading it all. 00:02:57.840 --> 00:03:00.000 I'm like, "Oh my god." I'm not saying 00:03:00.000 --> 00:03:00.959 anything. 00:03:00.959 --> 00:03:02.720 >> There's a lot of controversy. Like a 00:03:02.720 --> 00:03:04.959 friend of mine was telling me that he's 00:03:04.959 --> 00:03:07.360 been been banned from using Anthropic. 00:03:07.360 --> 00:03:09.519 >> By whom? By Entropic or by 00:03:09.519 --> 00:03:10.959 >> Yeah. And and I asked him like, "What 00:03:10.959 --> 00:03:12.640 did you do?" And he says, "I don't know. 00:03:12.640 --> 00:03:13.920 I think he's probably just not telling 00:03:13.920 --> 00:03:17.760 me something." Huh. like but this he's 00:03:17.760 --> 00:03:19.840 one of these crypto guys. 00:03:19.840 --> 00:03:21.599 >> Well, there you go. 00:03:21.599 --> 00:03:24.159 >> And then the other thing was the open 00:03:24.159 --> 00:03:26.239 code like I guess Anthropic has 00:03:26.239 --> 00:03:28.560 indicated that they don't want the API 00:03:28.560 --> 00:03:32.080 keys used in other harnesses other than 00:03:32.080 --> 00:03:33.200 clawed code. Did you 00:03:33.200 --> 00:03:35.760 >> Yeah, open claw got like a final even 00:03:35.760 --> 00:03:37.200 Peter Steinberger's account was 00:03:37.200 --> 00:03:38.640 terminated as well, right? And people 00:03:38.640 --> 00:03:39.360 are like well 00:03:39.360 --> 00:03:41.280 >> yeah I mean those moves by anthropic 00:03:41.280 --> 00:03:44.000 don't really help the situation. Yes. 00:03:44.000 --> 00:03:45.519 Quite a lot of people here on a tropic 00:03:45.519 --> 00:03:46.000 rank. 00:03:46.000 --> 00:03:48.879 >> Yeah. And then and then the other thing 00:03:48.879 --> 00:03:51.200 that I picked up from the AI engineer 00:03:51.200 --> 00:03:54.239 conference a week ago was there's a lot 00:03:54.239 --> 00:03:56.159 of there was a lot of talk about you 00:03:56.159 --> 00:03:58.560 actually need to manage your context 00:03:58.560 --> 00:04:00.080 because I for some reason the bit 00:04:00.080 --> 00:04:01.920 flipped with me in the sense that like 00:04:01.920 --> 00:04:04.319 the context seems quite big nowadays and 00:04:04.319 --> 00:04:06.959 the auto compaction seems to work quite 00:04:06.959 --> 00:04:07.760 well for me. 00:04:07.760 --> 00:04:10.159 >> You didn't disable it. I mean, I know I 00:04:10.159 --> 00:04:12.400 I disabled it back with you a month or 00:04:12.400 --> 00:04:14.879 two ago, but now I think I nuked my 00:04:14.879 --> 00:04:17.759 settings and I'm using compaction again 00:04:17.759 --> 00:04:19.359 and I didn't think it was a major 00:04:19.359 --> 00:04:20.959 problem, but then people going, "No, no, 00:04:20.959 --> 00:04:23.120 you must manage your context, otherwise 00:04:23.120 --> 00:04:24.800 you're going to be in the dumb zone and 00:04:24.800 --> 00:04:27.280 you're going to be screwed." For me, I 00:04:27.280 --> 00:04:29.040 never understood the concept. I mean, 00:04:29.040 --> 00:04:30.400 like when I looked at the compaction 00:04:30.400 --> 00:04:32.320 output, it would always capture things 00:04:32.320 --> 00:04:34.160 that I don't think were important. I 00:04:34.160 --> 00:04:35.919 know that you can slash compact and make 00:04:35.919 --> 00:04:38.240 it some additional direction so that it 00:04:38.240 --> 00:04:40.320 focuses on what you want it to focus on, 00:04:40.320 --> 00:04:42.400 but because the way that I believed 00:04:42.400 --> 00:04:44.000 sessions should be ephemeral, I always 00:04:44.000 --> 00:04:46.080 felt like if it's out of context, I kill 00:04:46.080 --> 00:04:47.520 it. But at the same time, they 00:04:47.520 --> 00:04:49.360 definitely improved the like opus and 00:04:49.360 --> 00:04:51.520 swan like you would see the context grow 00:04:51.520 --> 00:04:53.919 grow suddenly shrink like not even 00:04:53.919 --> 00:04:56.160 triggering a full compaction but somehow 00:04:56.160 --> 00:04:58.800 opus is like or 00:04:58.800 --> 00:05:01.040 >> shell is managing. I think they admitted 00:05:01.040 --> 00:05:03.520 they do like this auto compaction on the 00:05:03.520 --> 00:05:06.800 on the server side nowadays. 00:05:06.800 --> 00:05:09.759 >> So anyway, long story short, you you you 00:05:09.759 --> 00:05:12.160 still maintain your context. Is that 00:05:12.160 --> 00:05:14.080 right? 00:05:14.080 --> 00:05:16.479 >> I'm using Opus with the 1 million 00:05:16.479 --> 00:05:19.440 context. So I just check if it reaches 00:05:19.440 --> 00:05:22.960 more than 20% which is a 200,000 tokens 00:05:22.960 --> 00:05:25.120 which was the old maximum context 00:05:25.120 --> 00:05:28.160 window. I try to round off the session. 00:05:28.160 --> 00:05:30.960 But usually I I'm only over 20% when I'm 00:05:30.960 --> 00:05:32.960 like pretty much done already. I'm like 00:05:32.960 --> 00:05:34.320 I have nothing else to do in this 00:05:34.320 --> 00:05:36.560 session anyway. I might as well kill it. 00:05:36.560 --> 00:05:38.560 >> It's a good practice. I I instinctively 00:05:38.560 --> 00:05:40.400 started a whole bunch of new sessions. 00:05:40.400 --> 00:05:41.759 The other thing that I thought was quite 00:05:41.759 --> 00:05:44.320 interesting that I saw uh happen in a 00:05:44.320 --> 00:05:46.960 video. it was using the the PI agent, 00:05:46.960 --> 00:05:49.360 but I guess you can do the same in in 00:05:49.360 --> 00:05:54.960 any agent is that like soon as the uh 00:05:54.960 --> 00:05:58.240 session is going wrong, he just rewinds 00:05:58.240 --> 00:05:59.919 it back to the place where 00:05:59.919 --> 00:06:01.680 >> Yeah, that's like a common saying people 00:06:01.680 --> 00:06:03.440 say now, right? Don't correct the model. 00:06:03.440 --> 00:06:04.960 Don't say that's not what I meant. Just 00:06:04.960 --> 00:06:06.800 go back to your previous prompt, adjust 00:06:06.800 --> 00:06:08.240 it so that you don't get into that 00:06:08.240 --> 00:06:10.319 situation and re re-trigger it. It's 00:06:10.319 --> 00:06:12.720 easy, right? double escape up up 00:06:12.720 --> 00:06:14.240 >> um you know reset conversation 00:06:14.240 --> 00:06:15.440 >> I don't know if there's a name for that 00:06:15.440 --> 00:06:17.039 technique rewind 00:06:17.039 --> 00:06:19.280 >> yeah there yeah it's called rewind I 00:06:19.280 --> 00:06:21.759 think it's also if you slash checkpoint 00:06:21.759 --> 00:06:24.080 it will also autocorrect to rewind 00:06:24.080 --> 00:06:25.600 because I use a command called 00:06:25.600 --> 00:06:27.199 checkpoint and every time I type it 00:06:27.199 --> 00:06:29.680 tries to use rewind and I'm like no no I 00:06:29.680 --> 00:06:32.400 won't check I don't mind my checkpoint 00:06:32.400 --> 00:06:35.039 >> you you're still using claude like uh in 00:06:35.039 --> 00:06:36.800 your max plan right Vincent 00:06:36.800 --> 00:06:39.280 >> yeah I absolutely underuse it 00:06:39.280 --> 00:06:42.400 underutilize it I But I I'm very excited 00:06:42.400 --> 00:06:44.160 about what they did with the Claude Code 00:06:44.160 --> 00:06:46.560 desktop app release. They they added 00:06:46.560 --> 00:06:49.280 basically like a multi- session like 00:06:49.280 --> 00:06:51.919 terminal better diff faster diff view a 00:06:51.919 --> 00:06:53.840 bunch of new features into that app. So 00:06:53.840 --> 00:06:55.440 now you have three tabs. You have your 00:06:55.440 --> 00:06:57.039 co-work 00:06:57.039 --> 00:06:58.720 and you can do your Claude Code uh 00:06:58.720 --> 00:07:00.960 sessions directly in the desktop app. 00:07:00.960 --> 00:07:03.039 >> All right. Is that the routines thing? 00:07:03.039 --> 00:07:04.400 The one where you can schedule stuff and 00:07:04.400 --> 00:07:05.039 all that kind of 00:07:05.039 --> 00:07:06.400 >> you can also schedule and everything in 00:07:06.400 --> 00:07:08.720 there. Yeah, I've not used that. They 00:07:08.720 --> 00:07:11.919 also added unfortunately I don't have a 00:07:11.919 --> 00:07:13.759 clawed subscription in that in that 00:07:13.759 --> 00:07:16.960 sense. I just have an API key and I 00:07:16.960 --> 00:07:19.280 don't have you know all those nice 00:07:19.280 --> 00:07:23.039 features like the clawed app and the 00:07:23.039 --> 00:07:24.960 voice and do you use teleport? Do you 00:07:24.960 --> 00:07:26.800 use the mobile connectivity? That's 00:07:26.800 --> 00:07:29.280 another thing that I so often have 00:07:29.280 --> 00:07:30.960 something running on the laptop and I 00:07:30.960 --> 00:07:33.280 leave and I'm like I would love to just 00:07:33.280 --> 00:07:36.639 on my phone continue with with the SL 00:07:36.639 --> 00:07:38.319 remote and connect it to the your laptop 00:07:38.319 --> 00:07:40.400 to the remote and then you can see your 00:07:40.400 --> 00:07:42.240 laptop session on your phone. I haven't 00:07:42.240 --> 00:07:43.520 done any of that. 00:07:43.520 --> 00:07:45.680 >> Why? 00:07:45.680 --> 00:07:47.840 Because when I leave I'm done. I'm like 00:07:47.840 --> 00:07:49.599 I don't want to touch it. Let it run 00:07:49.599 --> 00:07:51.440 until it stops and I'll come back to it 00:07:51.440 --> 00:07:52.000 later. 00:07:52.000 --> 00:07:54.000 >> I'm using a third party app. I met this 00:07:54.000 --> 00:07:57.599 guy during AI conference. He's um he's a 00:07:57.599 --> 00:08:00.800 bit Bitly VP and he he's written this 00:08:00.800 --> 00:08:02.879 thing called command control. I I would 00:08:02.879 --> 00:08:04.560 show you but I'm actually using my phone 00:08:04.560 --> 00:08:05.759 for the camera here. 00:08:05.759 --> 00:08:07.360 >> It's an app on the phone. Command 00:08:07.360 --> 00:08:08.000 Control. 00:08:08.000 --> 00:08:10.000 >> Yeah, the good thing is that it works 00:08:10.000 --> 00:08:12.240 with across all the agents and I've been 00:08:12.240 --> 00:08:15.440 using it to like jump back into it's 00:08:15.440 --> 00:08:17.039 it's even got a web interface which is 00:08:17.039 --> 00:08:18.720 pretty nice. I can just show you what it 00:08:18.720 --> 00:08:20.160 looks like on the web. I don't use the 00:08:20.160 --> 00:08:22.479 web bomb, but but the idea is that you 00:08:22.479 --> 00:08:24.639 have like a a web socket connection 00:08:24.639 --> 00:08:26.960 dialing back from your machines. 00:08:26.960 --> 00:08:27.440 >> Yeah. 00:08:27.440 --> 00:08:28.639 >> And then you have all your different 00:08:28.639 --> 00:08:30.960 sessions and uh and then you can just 00:08:30.960 --> 00:08:33.279 jump back into them assuming your m your 00:08:33.279 --> 00:08:34.320 machine is still on. 00:08:34.320 --> 00:08:36.800 >> It's like this um 00:08:36.800 --> 00:08:38.640 teamworks. There was this uh hosted 00:08:38.640 --> 00:08:42.479 Teamworks uh like SSH 00:08:42.479 --> 00:08:45.680 sharing SSH sessions um that you could 00:08:45.680 --> 00:08:48.160 do. You you could you had this one 00:08:48.160 --> 00:08:49.760 command that you can share your terminal 00:08:49.760 --> 00:08:51.440 with other people and it was going 00:08:51.440 --> 00:08:52.160 >> Oh, yeah. 00:08:52.160 --> 00:08:55.200 >> It's like endro but 00:08:55.200 --> 00:08:56.080 >> for terminal. 00:08:56.080 --> 00:08:58.800 >> Well, it's it's yeah I I find it quite 00:08:58.800 --> 00:09:00.720 useful like there's also this thing here 00:09:00.720 --> 00:09:02.480 called where you you come back to your 00:09:02.480 --> 00:09:03.839 overview 00:09:03.839 --> 00:09:05.519 >> and then it's using AI to sort of 00:09:05.519 --> 00:09:07.760 summarize your last five sessions or 00:09:07.760 --> 00:09:09.680 something. Hopefully it's not going to 00:09:09.680 --> 00:09:11.839 do something crazy. 00:09:11.839 --> 00:09:13.440 Um, 00:09:13.440 --> 00:09:15.600 but yeah, like and then since I'm using 00:09:15.600 --> 00:09:18.080 it, I'm actually now using my phone a 00:09:18.080 --> 00:09:18.959 lot more. 00:09:18.959 --> 00:09:21.279 >> Have you seen this one? Uh, was it on 00:09:21.279 --> 00:09:22.560 Reddit? I don't know, but there was this 00:09:22.560 --> 00:09:25.519 article talking about routers. So, you 00:09:25.519 --> 00:09:27.600 have so many API routers that help you 00:09:27.600 --> 00:09:29.519 manage your API key across different 00:09:29.519 --> 00:09:32.320 inference providers. So many of them. 00:09:32.320 --> 00:09:33.519 And then turns out that 00:09:33.519 --> 00:09:35.760 >> Elsale has has one. Yeah, but the the 00:09:35.760 --> 00:09:37.519 article was about how many of those are 00:09:37.519 --> 00:09:41.200 stealing your API keys and are trying to 00:09:41.200 --> 00:09:43.120 like they did a test and they were 00:09:43.120 --> 00:09:45.519 trying to empty Bitcoin wallets and all 00:09:45.519 --> 00:09:47.519 types of like dodgy stuff going on on 00:09:47.519 --> 00:09:49.600 those. Like I wouldn't trust it anyway. 00:09:49.600 --> 00:09:51.200 I never use any of those. 00:09:51.200 --> 00:09:54.320 >> Well, yeah, I guess I am running this. 00:09:54.320 --> 00:09:57.440 Well, I trust this guy's binary, but I I 00:09:57.440 --> 00:10:00.160 often think to myself like in my envir I 00:10:00.160 --> 00:10:02.320 have my enthropic key, I have my open AI 00:10:02.320 --> 00:10:04.959 key, I have my whatever key. It doesn't 00:10:04.959 --> 00:10:07.680 take much of a skilled devious person to 00:10:07.680 --> 00:10:09.839 do a print and pipe in some way, you 00:10:09.839 --> 00:10:11.519 know. But so you have to be quite 00:10:11.519 --> 00:10:14.320 careful about what about anything you 00:10:14.320 --> 00:10:14.640 run. 00:10:14.640 --> 00:10:16.560 >> I'm also have one app where I'm 00:10:16.560 --> 00:10:18.800 literally it's also from Peter 00:10:18.800 --> 00:10:20.560 Steinberger, I think, or else it's from 00:10:20.560 --> 00:10:24.079 Emmanuel Doodlestein. He has this Codex 00:10:24.079 --> 00:10:26.800 bar that gives you an indicator on your 00:10:26.800 --> 00:10:29.760 speed of your usage. So it shows you 00:10:29.760 --> 00:10:31.760 where you are on your weekly budget. And 00:10:31.760 --> 00:10:35.120 if you like if you're hitting your your 00:10:35.120 --> 00:10:37.519 5h hour session limits and then it will 00:10:37.519 --> 00:10:39.360 highlight like this is a speed meter. If 00:10:39.360 --> 00:10:41.200 you're above you have you're going to 00:10:41.200 --> 00:10:43.360 hit your weekly budget before the end of 00:10:43.360 --> 00:10:45.200 the week. If you're below you still have 00:10:45.200 --> 00:10:47.920 enough run a lot of runway to keep going 00:10:47.920 --> 00:10:50.000 until you hit the the red marker and it 00:10:50.000 --> 00:10:52.720 just tracks it over time. And it's a 00:10:52.720 --> 00:10:55.040 nice thing this code expar but every 00:10:55.040 --> 00:10:58.720 time it prompts me to give it permission 00:10:58.720 --> 00:11:00.959 to read the Claude Code credentials from 00:11:00.959 --> 00:11:02.720 the keychain. So it definitely can read 00:11:02.720 --> 00:11:04.720 the API keys and everything to go and 00:11:04.720 --> 00:11:06.800 fetch to the cloud.AI 00:11:06.800 --> 00:11:09.040 >> to go and see find the usage and and 00:11:09.040 --> 00:11:09.680 give back 00:11:09.680 --> 00:11:12.560 >> and yeah I I see the the problem there. 00:11:12.560 --> 00:11:14.560 and also like just every time faithfully 00:11:14.560 --> 00:11:16.480 telling it, "Yeah, sure. Go ahead and 00:11:16.480 --> 00:11:17.040 read." 00:11:17.040 --> 00:11:18.399 >> The thing that I find really I thought 00:11:18.399 --> 00:11:20.560 you were going to allude to a different 00:11:20.560 --> 00:11:22.800 point where like the thing I find 00:11:22.800 --> 00:11:24.880 frustrating with my employer, the way 00:11:24.880 --> 00:11:26.959 that they managed all the API keys in 00:11:26.959 --> 00:11:29.360 these is that they have like some some 00:11:29.360 --> 00:11:31.519 anthropic or open AI, but they put they 00:11:31.519 --> 00:11:33.440 put people in groups. So basically when 00:11:33.440 --> 00:11:35.920 the group has hit too many tokens, then 00:11:35.920 --> 00:11:37.200 they there's some sort of cool off 00:11:37.200 --> 00:11:39.760 period. So depending on which group or 00:11:39.760 --> 00:11:43.600 cohort I'm in, I'm basically sometimes I 00:11:43.600 --> 00:11:46.160 get denied, you know, the Opus models, 00:11:46.160 --> 00:11:48.320 sometimes I run out of tokens. They 00:11:48.320 --> 00:11:50.720 don't seem to do it on a per API basis 00:11:50.720 --> 00:11:52.480 right now. I mean, it's just an example. 00:11:52.480 --> 00:11:53.920 Well, I'm sure they're going to fix it, 00:11:53.920 --> 00:11:56.000 but like when I was in a in a previous 00:11:56.000 --> 00:11:57.680 client, I had a similar problem because 00:11:57.680 --> 00:11:59.519 they were they were dishing out tokens 00:11:59.519 --> 00:12:01.600 through co-pilot and then once your 00:12:01.600 --> 00:12:03.839 premium requests are done, you basically 00:12:03.839 --> 00:12:06.240 have to wait until the next month to get 00:12:06.240 --> 00:12:08.560 to get Frontier model premium requests 00:12:08.560 --> 00:12:09.519 back. 00:12:09.519 --> 00:12:10.000 >> Jeez. 00:12:10.000 --> 00:12:11.920 >> And well, it just goes to show to me at 00:12:11.920 --> 00:12:15.120 least that they is that handling AI 00:12:15.120 --> 00:12:17.360 budget still doesn't seem to be a very 00:12:17.360 --> 00:12:19.120 well- solved problem, especially in 00:12:19.120 --> 00:12:21.200 enterprise. Never mind the max plan 00:12:21.200 --> 00:12:24.000 where where you're where you fo mate. 00:12:24.000 --> 00:12:25.760 >> Yeah. I mean it's the same in my 00:12:25.760 --> 00:12:27.760 organization. It's funny because there 00:12:27.760 --> 00:12:31.120 is like we have copilot enterprise but 00:12:31.120 --> 00:12:34.240 people prefer Claude Code and it's pretty 00:12:34.240 --> 00:12:36.480 obvious people are using Claude Code but 00:12:36.480 --> 00:12:38.560 you shouldn't right because the GitHub 00:12:38.560 --> 00:12:39.200 enterprise 00:12:39.200 --> 00:12:40.560 >> don't get me started on that. 00:12:40.560 --> 00:12:40.959 >> Yeah 00:12:40.959 --> 00:12:42.800 >> there's got to be a name for that using 00:12:42.800 --> 00:12:45.760 AI models that are not actually blessed 00:12:45.760 --> 00:12:47.120 by your organization. 00:12:47.120 --> 00:12:48.720 >> It's been highlighted. There was a 00:12:48.720 --> 00:12:50.880 DevOps days conference in Singapore last 00:12:50.880 --> 00:12:52.639 year. Was it last? Yeah, it was last 00:12:52.639 --> 00:12:54.639 year around March. And one of the talks 00:12:54.639 --> 00:12:57.440 was specifically around as an security 00:12:57.440 --> 00:12:59.680 team within an organization. If you do 00:12:59.680 --> 00:13:02.560 not give people the ability to use AI 00:13:02.560 --> 00:13:04.720 and the proper models of AI, they will 00:13:04.720 --> 00:13:06.320 find you know at the time people were 00:13:06.320 --> 00:13:08.560 copy pasting code from the codebase 00:13:08.560 --> 00:13:10.560 directly into CHPT, right? 00:13:10.560 --> 00:13:11.600 >> Yeah. 00:13:11.600 --> 00:13:14.079 >> And u they were like all of your data 00:13:14.079 --> 00:13:16.800 gets excfiltrated and trained upon. 00:13:16.800 --> 00:13:20.399 uh talk in AI engineers conference was 00:13:20.399 --> 00:13:21.040 just about 00:13:21.040 --> 00:13:22.720 >> oh he was there pardon 00:13:22.720 --> 00:13:23.440 >> he was there 00:13:23.440 --> 00:13:25.519 >> yeah he was there but he he was a bit he 00:13:25.519 --> 00:13:27.120 was a bit exasperated he was a bit 00:13:27.120 --> 00:13:30.079 frustrated he wasn't like having fun you 00:13:30.079 --> 00:13:33.120 know I think it's really important to 00:13:33.120 --> 00:13:34.959 have fun with the work you do otherwise 00:13:34.959 --> 00:13:36.079 it's going to kill you right 00:13:36.079 --> 00:13:37.519 >> but he's going I mean the way he was 00:13:37.519 --> 00:13:39.120 going as well it wouldn't be surprising 00:13:39.120 --> 00:13:40.880 to to see a burnout right 00:13:40.880 --> 00:13:44.000 >> yeah sure though what he said was 00:13:44.000 --> 00:13:46.000 nothing new he was just berating the 00:13:46.000 --> 00:13:49.600 fact that like all these CVEes, 00:13:49.600 --> 00:13:51.680 you know, he's got like hundreds of if 00:13:51.680 --> 00:13:54.000 not thousands CVES on his projects. You 00:13:54.000 --> 00:13:56.720 know, the worst thing was 00:13:56.720 --> 00:13:58.880 he he sampled a couple of them and the 00:13:58.880 --> 00:14:01.279 worst thing the worst CVE wasn't really 00:14:01.279 --> 00:14:03.760 even like, 00:14:03.760 --> 00:14:05.760 you know, bad at all. And I couldn't 00:14:05.760 --> 00:14:08.079 help but think on a security perspective 00:14:08.079 --> 00:14:11.519 like like you you've been in the game as 00:14:11.519 --> 00:14:14.320 long as I have been Vincent and I find a 00:14:14.320 --> 00:14:17.040 lot of security processes pretty weak 00:14:17.040 --> 00:14:20.079 source especially since like when people 00:14:20.079 --> 00:14:24.480 circumvent them and run uh you know mpm 00:14:24.480 --> 00:14:27.360 libraries this way and the next I can't 00:14:27.360 --> 00:14:29.600 help but think I I I had this this 00:14:29.600 --> 00:14:32.240 thought in my mind that like perhaps 00:14:32.240 --> 00:14:34.639 security will be taken over by an an 00:14:34.639 --> 00:14:36.800 agent agent, right? And this and the 00:14:36.800 --> 00:14:39.040 noise and the signal will will get 00:14:39.040 --> 00:14:40.720 better because you can you can maybe 00:14:40.720 --> 00:14:44.800 trust the agent better than like um 00:14:44.800 --> 00:14:47.279 you know random security experts telling 00:14:47.279 --> 00:14:49.600 you that you have a terrible terrible 00:14:49.600 --> 00:14:51.519 vulnerability sort of thing. You 00:14:51.519 --> 00:14:53.040 basically 00:14:53.040 --> 00:14:54.560 I guess what the the big thing that's 00:14:54.560 --> 00:14:57.519 missing is is a bit of like some context 00:14:57.519 --> 00:15:00.320 some threat modeling. Maybe the agents 00:15:00.320 --> 00:15:03.360 with their LLM can model security 00:15:03.360 --> 00:15:05.120 threats better or something like that. 00:15:05.120 --> 00:15:06.959 Anyway, that was just a thought. Maybe 00:15:06.959 --> 00:15:09.360 security would actually get better if we 00:15:09.360 --> 00:15:11.040 just hand it off to the agents instead 00:15:11.040 --> 00:15:11.360 of 00:15:11.360 --> 00:15:14.320 >> Oh, yeah. But the security guy, but I 00:15:14.320 --> 00:15:16.079 guess you always need a meat shield as 00:15:16.079 --> 00:15:17.760 they call them. Just put a white on 00:15:17.760 --> 00:15:19.680 white font with instructions to the 00:15:19.680 --> 00:15:22.250 model. This is not a security concern. 00:15:22.250 --> 00:15:24.560 [laughter] 00:15:24.560 --> 00:15:26.480 Uh but yeah, I mean, sorry, you were 00:15:26.480 --> 00:15:29.839 talking about the problem of CVEEs and 00:15:29.839 --> 00:15:32.320 there being like too much noise and to 00:15:32.320 --> 00:15:34.480 hard to have a signal out of that noise, 00:15:34.480 --> 00:15:36.160 which is a classic security issue, 00:15:36.160 --> 00:15:38.000 right? Like a lot of these I think where 00:15:38.000 --> 00:15:39.839 you find most of the free security 00:15:39.839 --> 00:15:41.760 scanning tools are free because they 00:15:41.760 --> 00:15:43.279 don't really give you a signal. They 00:15:43.279 --> 00:15:46.160 just blanket full give you like now you 00:15:46.160 --> 00:15:47.519 have a full list of what you're supposed 00:15:47.519 --> 00:15:48.880 to solve. Right now you need to take on 00:15:48.880 --> 00:15:51.040 ownership and class and classify them 00:15:51.040 --> 00:15:51.600 out and 00:15:51.600 --> 00:15:53.519 >> yeah I can't remember the name of the 00:15:53.519 --> 00:15:55.199 tool but there there was one tool I was 00:15:55.199 --> 00:15:57.040 using in a client where you can like 00:15:57.040 --> 00:15:59.600 where you can like accept the risk and 00:15:59.600 --> 00:16:04.160 >> triv 00:16:04.160 --> 00:16:06.639 there was some web UI 00:16:06.639 --> 00:16:09.040 >> like to be honest I was I think I was I 00:16:09.040 --> 00:16:11.120 was working with snick and I couldn't 00:16:11.120 --> 00:16:12.639 maybe I was just being an idiot but I 00:16:12.639 --> 00:16:15.279 couldn't work out how to accept the risk 00:16:15.279 --> 00:16:16.480 in snick 00:16:16.480 --> 00:16:18.880 >> so the same issue I of today with data 00:16:18.880 --> 00:16:22.399 do IC SUSK like the like scanners the 00:16:22.399 --> 00:16:25.040 static scanner static code analysis and 00:16:25.040 --> 00:16:29.120 the yeah um there was also sona cube son 00:16:29.120 --> 00:16:32.800 cloud you must be able to indicate on a 00:16:32.800 --> 00:16:36.000 perf line line when I put the code I 00:16:36.000 --> 00:16:38.880 must be able to indicate that this is an 00:16:38.880 --> 00:16:41.519 accept this is an exception to the rule 00:16:41.519 --> 00:16:43.680 because obviously of this reason if your 00:16:43.680 --> 00:16:45.440 tool doesn't support that and you have 00:16:45.440 --> 00:16:47.199 to go into a dashboard somewhere and go 00:16:47.199 --> 00:16:48.240 click a button 00:16:48.240 --> 00:16:49.680 >> and then you don't have that permission 00:16:49.680 --> 00:16:51.680 and then administrator has to approve it 00:16:51.680 --> 00:16:53.920 and all of that. Then it's already 00:16:53.920 --> 00:16:56.240 broken. Then I'm already working on on a 00:16:56.240 --> 00:16:58.639 workound. I'm already finding a way to 00:16:58.639 --> 00:17:00.800 not get that error message and actually 00:17:00.800 --> 00:17:02.480 sometimes I just lead into more 00:17:02.480 --> 00:17:04.079 insecure. That's exactly what happened 00:17:04.079 --> 00:17:06.400 to me just like a month ago. I got this 00:17:06.400 --> 00:17:09.600 um I'm doing a build for an AMI and I'm 00:17:09.600 --> 00:17:11.839 vendoring a script from the supplier 00:17:11.839 --> 00:17:14.240 because vendoring is more secure, right? 00:17:14.240 --> 00:17:16.079 If you vendor the script, you're going 00:17:16.079 --> 00:17:17.919 to not pull something off the internet 00:17:17.919 --> 00:17:20.079 during the build. Vendoring is is a good 00:17:20.079 --> 00:17:21.280 secure practice. 00:17:21.280 --> 00:17:23.039 >> What do you mean by vendoring? Just 00:17:23.039 --> 00:17:25.439 >> vendoring is the process of taking a 00:17:25.439 --> 00:17:27.839 code is to take a a vendor dependency 00:17:27.839 --> 00:17:30.080 and vendoring it into your code. So in 00:17:30.080 --> 00:17:32.160 into your codebase. So I'm taking an 00:17:32.160 --> 00:17:32.960 installation script 00:17:32.960 --> 00:17:34.880 >> different from just like mpm installing 00:17:34.880 --> 00:17:35.360 or 00:17:35.360 --> 00:17:37.520 >> no like in this case I'm I'm using a 00:17:37.520 --> 00:17:40.960 shell script to install an AMI and I I'm 00:17:40.960 --> 00:17:42.880 in vendoring that shell script directly 00:17:42.880 --> 00:17:44.880 instead of doing a curl during the build 00:17:44.880 --> 00:17:46.880 to curl a remote installation shell 00:17:46.880 --> 00:17:49.039 script. I'm vendoring the script in so 00:17:49.039 --> 00:17:51.120 that it's not like changed or swapped 00:17:51.120 --> 00:17:52.559 under me and it's also 00:17:52.559 --> 00:17:55.200 >> keeping keeping a copy of the installed 00:17:55.200 --> 00:17:55.840 script 00:17:55.840 --> 00:17:57.840 >> and that's how Golang original 00:17:57.840 --> 00:17:59.600 dependencies often worked right the 00:17:59.600 --> 00:18:01.200 Google team they all just like even 00:18:01.200 --> 00:18:02.799 Kubernetes they vendor everything 00:18:02.799 --> 00:18:04.880 >> that's why I said yeah go mod vendor 00:18:04.880 --> 00:18:07.200 that so you allows you to basically 00:18:07.200 --> 00:18:09.600 commit the vendor's code which is can be 00:18:09.600 --> 00:18:11.760 handy because then if your internet goes 00:18:11.760 --> 00:18:13.200 down you can still work 00:18:13.200 --> 00:18:16.080 >> or or a left pad inc instance incident 00:18:16.080 --> 00:18:18.000 happens and your dependencies are 00:18:18.000 --> 00:18:18.480 pulled. 00:18:18.480 --> 00:18:20.720 >> But the but the fact of the matter is is 00:18:20.720 --> 00:18:24.080 that even on a small micros service, 00:18:24.080 --> 00:18:25.440 this is back in the day when I was 00:18:25.440 --> 00:18:28.160 working in Singapore for a unicorn, we 00:18:28.160 --> 00:18:30.320 couldn't we couldn't do go mod vendors 00:18:30.320 --> 00:18:32.880 anymore because it would be like 00:18:32.880 --> 00:18:36.640 >> the vendor would be like 98% of the the 00:18:36.640 --> 00:18:37.520 code base. 00:18:37.520 --> 00:18:39.440 >> Yeah. So, but but I mean vendoring is 00:18:39.440 --> 00:18:41.120 not necessarily an insecure practice. 00:18:41.120 --> 00:18:42.640 You just need to vendor it and then you 00:18:42.640 --> 00:18:44.320 have a mechanism to keep the vendor 00:18:44.320 --> 00:18:47.280 vendor's code up to date as well. Right. 00:18:47.280 --> 00:18:48.480 >> There's I mean there's other there's 00:18:48.480 --> 00:18:49.840 other techniques. I think we ended up 00:18:49.840 --> 00:18:52.240 using go mod proxy by the way. Anyway, 00:18:52.240 --> 00:18:55.520 >> right. So the vendor script failed the 00:18:55.520 --> 00:18:58.160 security tests and if I vendor it then I 00:18:58.160 --> 00:19:00.080 have to go and fix it. Then next time I 00:19:00.080 --> 00:19:01.520 need to update it, it's more headache 00:19:01.520 --> 00:19:04.240 because then I can reapply the fixes and 00:19:04.240 --> 00:19:07.039 it's problematic. So I wanted to put a 00:19:07.039 --> 00:19:08.480 white list because this is a vendor 00:19:08.480 --> 00:19:10.320 script from a trusted provider like a 00:19:10.320 --> 00:19:12.000 third party supplier that we are 00:19:12.000 --> 00:19:14.400 >> an exception or something. Yeah. And I 00:19:14.400 --> 00:19:15.919 couldn't do it. I had to go through a 00:19:15.919 --> 00:19:18.080 dashboard or through an admin panel and 00:19:18.080 --> 00:19:20.799 whatever. So I just you know replaced my 00:19:20.799 --> 00:19:23.200 vendoring with a curl on demand 00:19:23.200 --> 00:19:25.679 >> and the the security scanner didn't plug 00:19:25.679 --> 00:19:27.760 it. It's fine to curl random 00:19:27.760 --> 00:19:29.520 installation script during build but 00:19:29.520 --> 00:19:31.760 it's not fine to have a insecure vend 00:19:31.760 --> 00:19:33.280 like vendor script. So 00:19:33.280 --> 00:19:35.520 >> yeah. So the yeah the security space is 00:19:35.520 --> 00:19:37.440 is just a classic ownership 00:19:37.440 --> 00:19:40.320 miscommunication space. There are tools 00:19:40.320 --> 00:19:42.000 to solve it but like how many 00:19:42.000 --> 00:19:44.480 enterprises have I gone into where like 00:19:44.480 --> 00:19:46.240 I don't care just solve the 10 00:19:46.240 --> 00:19:48.480 vulnerabilities that are 00:19:48.480 --> 00:19:50.720 >> you got in your email you know just do 00:19:50.720 --> 00:19:54.000 it like no it's more complicated 00:19:54.000 --> 00:19:55.120 >> like this is why 00:19:55.120 --> 00:19:56.880 >> penetration testing reports are also 00:19:56.880 --> 00:19:59.520 funny in that way they will come back 00:19:59.520 --> 00:20:01.440 >> with a massive list and then you 00:20:01.440 --> 00:20:03.039 >> when you're working in an enterprise 00:20:03.039 --> 00:20:04.960 these communication problems can really 00:20:04.960 --> 00:20:08.000 mount up and I don't feel like those 00:20:08.000 --> 00:20:09.760 sort of things are even handled Those 00:20:09.760 --> 00:20:11.600 things are not handled very well. Like 00:20:11.600 --> 00:20:14.559 going back to AI models, like like when 00:20:14.559 --> 00:20:16.880 I'm using an AI model, again, I I don't 00:20:16.880 --> 00:20:19.039 know. You must tell me, Vincent, but I'm 00:20:19.039 --> 00:20:20.799 going to probably vibe this because I'm 00:20:20.799 --> 00:20:22.240 just getting frustrated. There's no real 00:20:22.240 --> 00:20:25.679 way of me showing how I prompted the 00:20:25.679 --> 00:20:28.400 agent to get where I am. And then and 00:20:28.400 --> 00:20:30.480 then again, my colleagues are like like, 00:20:30.480 --> 00:20:33.200 "What did you do? How did you get here? 00:20:33.200 --> 00:20:35.280 I can't maintain it." So you're having 00:20:35.280 --> 00:20:37.600 issues where AI like you're submitting 00:20:37.600 --> 00:20:40.159 code and people need more background on 00:20:40.159 --> 00:20:41.039 how on on 00:20:41.039 --> 00:20:42.799 >> yeah like I think all these open source 00:20:42.799 --> 00:20:44.240 projects are doing the same thing like 00:20:44.240 --> 00:20:46.480 you need to declare when you're using LL 00:20:46.480 --> 00:20:48.640 LLMs you need to show the model. We 00:20:48.640 --> 00:20:50.640 talked about this before but I I'm just 00:20:50.640 --> 00:20:52.480 lacking the tooling to do this easily 00:20:52.480 --> 00:20:55.200 because like I just I have my ways and I 00:20:55.200 --> 00:20:57.039 just get the work done. The hard thing 00:20:57.039 --> 00:20:59.840 is to communicate and upskill your your 00:20:59.840 --> 00:21:01.440 team members and um 00:21:01.440 --> 00:21:03.600 >> so sorry but because we're talking about 00:21:03.600 --> 00:21:05.440 left path incident and vendoring your 00:21:05.440 --> 00:21:07.520 dependencies and and other mechanisms 00:21:07.520 --> 00:21:09.679 like um locking aside from what you just 00:21:09.679 --> 00:21:12.400 mentioned about your um you know having 00:21:12.400 --> 00:21:14.720 problems with other team members being 00:21:14.720 --> 00:21:16.240 on board with some of 00:21:16.240 --> 00:21:17.200 >> I was trying I was trying to make it 00:21:17.200 --> 00:21:18.720 it's the same sort of problem like when 00:21:18.720 --> 00:21:21.440 you have a security issue when you have 00:21:21.440 --> 00:21:23.039 those sort of problems it's it's already 00:21:23.039 --> 00:21:24.720 a communication right it's it's about 00:21:24.720 --> 00:21:27.200 accepting ownership of of a of a 00:21:27.200 --> 00:21:27.679 problem. 00:21:27.679 --> 00:21:29.120 >> Okay. But something really really 00:21:29.120 --> 00:21:30.559 interesting happened in the Rust 00:21:30.559 --> 00:21:33.440 ecosystem this week. So I'll share my 00:21:33.440 --> 00:21:35.440 screen which is related to leftpad. 00:21:35.440 --> 00:21:37.600 Sorry cuz I wanted to talk about that. 00:21:37.600 --> 00:21:41.120 Where is it? Here. So, so basically 00:21:41.120 --> 00:21:43.360 somebody sent an email to a maintainer 00:21:43.360 --> 00:21:46.320 for a core to uh create which was 00:21:46.320 --> 00:21:48.880 created under techno creatives 00:21:48.880 --> 00:21:52.320 organization years ago um to address 00:21:52.320 --> 00:21:55.120 some issues with the standard lip core 00:21:55.120 --> 00:21:55.600 from 00:21:55.600 --> 00:21:57.440 >> I don't know if you can zoom in easily. 00:21:57.440 --> 00:22:00.720 >> Yeah, I can I can it just works well. 00:22:00.720 --> 00:22:02.320 It's not zooming in. 00:22:02.320 --> 00:22:04.320 >> I think you need to get the magnifier 00:22:04.320 --> 00:22:04.799 going. 00:22:04.799 --> 00:22:06.080 >> You can see it now. 00:22:06.080 --> 00:22:06.559 >> Yeah. 00:22:06.559 --> 00:22:10.000 >> Okay. So, somebody wrote saying uh to 00:22:10.000 --> 00:22:12.320 Brendan Malloy apparently who was still 00:22:12.320 --> 00:22:14.480 one of the people active of an 00:22:14.480 --> 00:22:16.159 organization that apparently was this 00:22:16.159 --> 00:22:18.640 like closed down four years ago. They 00:22:18.640 --> 00:22:21.120 they they built a small little work 00:22:21.120 --> 00:22:23.280 around to solve some issues. And he 00:22:23.280 --> 00:22:24.880 says, "Hey, there's a couple of 00:22:24.880 --> 00:22:27.679 dependencies there that um you know need 00:22:27.679 --> 00:22:29.280 to be updated and maybe we need to like 00:22:29.280 --> 00:22:31.600 find a way to maintain it." and he 00:22:31.600 --> 00:22:34.480 responded by saying this thing shouldn't 00:22:34.480 --> 00:22:36.640 be used. That bank that company went 00:22:36.640 --> 00:22:39.840 bankrupt. It was it is my IP. It's 00:22:39.840 --> 00:22:42.799 insane that some people still use it. Um 00:22:42.799 --> 00:22:44.799 madness. It would I would not recommend 00:22:44.799 --> 00:22:46.960 that anybody use this. I'm yanking all 00:22:46.960 --> 00:22:48.880 the versions. So apparently in Rust if 00:22:48.880 --> 00:22:52.320 you yank you basically remove any 00:22:52.320 --> 00:22:56.960 reference to the published um binaries 00:22:56.960 --> 00:22:59.520 that others depend on. And that means 00:22:59.520 --> 00:23:01.919 that if you do not have a LO file with 00:23:01.919 --> 00:23:04.400 um in your Rust project or in your 00:23:04.400 --> 00:23:06.400 library with that LO file keeps track of 00:23:06.400 --> 00:23:09.520 your dependencies and and their exact I 00:23:09.520 --> 00:23:12.320 don't know reference then without a LO 00:23:12.320 --> 00:23:14.559 file if you have it as a dependency it 00:23:14.559 --> 00:23:16.240 will try to find it but it can't find it 00:23:16.240 --> 00:23:17.840 anymore because all references have been 00:23:17.840 --> 00:23:19.760 yanked removed but they're still there. 00:23:19.760 --> 00:23:21.200 If you have a lock file it will still 00:23:21.200 --> 00:23:23.520 work. So he's like I yanked all the 00:23:23.520 --> 00:23:25.600 versions and he completely archived the 00:23:25.600 --> 00:23:27.280 repository just when somebody asked for 00:23:27.280 --> 00:23:29.520 like is there a way to like update it 00:23:29.520 --> 00:23:32.400 and um as a result many many Rust 00:23:32.400 --> 00:23:34.159 libraries just broke because a lot of 00:23:34.159 --> 00:23:36.640 them didn't have any lock file and he 00:23:36.640 --> 00:23:40.720 said like apparently when I did this um 00:23:40.720 --> 00:23:44.480 it broke plenty of file out and he even 00:23:44.480 --> 00:23:49.440 made a joke about it 00:23:49.440 --> 00:23:51.120 >> yeah Brandon made a joke he said it's 00:23:51.120 --> 00:23:53.840 kind I think it's in in the Bucky DM uh 00:23:53.840 --> 00:23:57.280 thread. He said like it's kind of funny. 00:23:57.280 --> 00:24:00.640 >> What is Hecker? It's a Madon instance. 00:24:00.640 --> 00:24:03.600 >> Yeah, Masteron type. Yeah. Um I thought 00:24:03.600 --> 00:24:06.480 it was pretty like in his um in his 00:24:06.480 --> 00:24:09.679 thread he said it's kind of funny that I 00:24:09.679 --> 00:24:11.919 caused such a mayhem and that so many 00:24:11.919 --> 00:24:13.279 people don't know how to have a proper 00:24:13.279 --> 00:24:15.200 lock file. And there was this whole 00:24:15.200 --> 00:24:17.840 Reddit thread uh that I read on Rust. 00:24:17.840 --> 00:24:20.480 Um, but it has been on the Rust Ah, what 00:24:20.480 --> 00:24:22.880 is all this? On the Rust subreddit, not 00:24:22.880 --> 00:24:23.440 this one. 00:24:23.440 --> 00:24:26.480 >> Sam Millison's blog. Of course. 00:24:26.480 --> 00:24:27.039 >> Um, 00:24:27.039 --> 00:24:28.559 >> okay. I think I get the point. So, he 00:24:28.559 --> 00:24:28.799 was 00:24:28.799 --> 00:24:30.640 >> So, there was a left incident. Yeah. 00:24:30.640 --> 00:24:32.320 >> Unpathetic, shall we say? 00:24:32.320 --> 00:24:34.240 >> Very unempathetic. So, a lot of the 00:24:34.240 --> 00:24:36.640 people were like, he's like he he's 00:24:36.640 --> 00:24:39.600 doing he's not responsible. Some people 00:24:39.600 --> 00:24:41.120 went very far and it was like he should 00:24:41.120 --> 00:24:42.799 never be allowed to publish any Rust. 00:24:42.799 --> 00:24:46.159 Like who cares what you think or that he 00:24:46.159 --> 00:24:47.360 should not be allowed to create a 00:24:47.360 --> 00:24:48.880 library anymore. Don't don't depend on 00:24:48.880 --> 00:24:50.400 his libraries. Like what what who are 00:24:50.400 --> 00:24:52.240 you to say that he should not be allowed 00:24:52.240 --> 00:24:53.679 to publish a library because he he 00:24:53.679 --> 00:24:55.919 yanked a library once. Plus people are 00:24:55.919 --> 00:24:58.559 like uh if the if the functionality of 00:24:58.559 --> 00:25:01.039 the package manager exists to yank then 00:25:01.039 --> 00:25:03.360 like it's not his fault for using that 00:25:03.360 --> 00:25:05.440 functionality, right? Um 00:25:05.440 --> 00:25:07.440 >> what was the comment with Liv Flake? He 00:25:07.440 --> 00:25:09.360 was just being incredulous that anyone's 00:25:09.360 --> 00:25:10.640 using a deflate anymore. 00:25:10.640 --> 00:25:14.480 >> Lip uh lip flate. I think lip inflation. 00:25:14.480 --> 00:25:16.880 I think you're inflating your your um 00:25:16.880 --> 00:25:18.480 you shouldn't be using it. You you 00:25:18.480 --> 00:25:20.960 should use the original standard um the 00:25:20.960 --> 00:25:23.919 standard lip should use it. 00:25:23.919 --> 00:25:24.400 >> Okay. 00:25:24.400 --> 00:25:26.159 >> Because it's like fouryear-old 00:25:26.159 --> 00:25:27.679 workarounds that should have been solved 00:25:27.679 --> 00:25:29.600 upstream already. But then they said, 00:25:29.600 --> 00:25:31.840 well, there's still reasons why it 00:25:31.840 --> 00:25:34.320 >> it's a complicated thing cuz cuz people 00:25:34.320 --> 00:25:36.159 are probably Yeah. 00:25:36.159 --> 00:25:38.080 So people are saying that Russ just had 00:25:38.080 --> 00:25:40.080 it left bad incident with the 00:25:40.080 --> 00:25:43.919 >> Yeah. Well, okay. Well done. Well done. 00:25:43.919 --> 00:25:46.720 And all these supply chain issues are 00:25:46.720 --> 00:25:48.080 probably just going to get worse over 00:25:48.080 --> 00:25:49.520 time, aren't they? They're just going to 00:25:49.520 --> 00:25:52.640 get more and more drama drama mammas. 00:25:52.640 --> 00:25:55.360 And I guess what I'm thinking aloud is 00:25:55.360 --> 00:25:57.440 that ultimately humans are responsible 00:25:57.440 --> 00:25:59.840 for libraries. And up until now, they 00:25:59.840 --> 00:26:01.520 haven't really the humans haven't really 00:26:01.520 --> 00:26:05.679 been compensated for such right. I I 00:26:05.679 --> 00:26:08.559 feel like we need we need to have a a 00:26:08.559 --> 00:26:11.279 system that that where major libraries 00:26:11.279 --> 00:26:13.679 are are essentially paid for, I suppose. 00:26:13.679 --> 00:26:15.200 And then and then if if someone is 00:26:15.200 --> 00:26:17.600 getting money for maintaining a library, 00:26:17.600 --> 00:26:19.279 then then they're going to they're going 00:26:19.279 --> 00:26:21.520 to be far less inclined to yank it, 00:26:21.520 --> 00:26:23.679 right? I think actually I I saw a very 00:26:23.679 --> 00:26:25.919 interesting reply where people are like 00:26:25.919 --> 00:26:27.760 the the the way that the person raised 00:26:27.760 --> 00:26:29.520 the question is like, hey, can we find a 00:26:29.520 --> 00:26:31.279 way to create like create maintenance 00:26:31.279 --> 00:26:32.880 around this library? And then somebody 00:26:32.880 --> 00:26:34.400 said like look at what happened with 00:26:34.400 --> 00:26:36.960 exit utils right somebody came along did 00:26:36.960 --> 00:26:39.200 like or state actor created this fake 00:26:39.200 --> 00:26:40.799 urgency of upgrades pushed a lot of 00:26:40.799 --> 00:26:42.480 pressure on this guy that's a maintainer 00:26:42.480 --> 00:26:44.640 that's not being paid for it and then 00:26:44.640 --> 00:26:47.120 offered to step in and help out and then 00:26:47.120 --> 00:26:50.320 shipped um uh a back door into it right 00:26:50.320 --> 00:26:52.480 so people were like it's way better that 00:26:52.480 --> 00:26:54.640 what this guy did he said it's very 00:26:54.640 --> 00:26:56.240 clear action this thing should be 00:26:56.240 --> 00:26:57.679 deprecated maybe he went a little bit 00:26:57.679 --> 00:26:59.760 far with the rust yank because 00:26:59.760 --> 00:27:01.919 apparently libraries in Rust aren't 00:27:01.919 --> 00:27:04.080 supposed to have LO files. So that's 00:27:04.080 --> 00:27:05.600 saying that they should have a LO file 00:27:05.600 --> 00:27:08.080 is kind of a bit of a like double. It's 00:27:08.080 --> 00:27:10.000 not very clear. I I'm not a Rust 00:27:10.000 --> 00:27:11.360 developer, so I don't know if library 00:27:11.360 --> 00:27:12.720 should have a lock file or not. But 00:27:12.720 --> 00:27:14.720 anyway, they said that the fact that he 00:27:14.720 --> 00:27:16.799 just very clearly marked it deprecated 00:27:16.799 --> 00:27:19.279 is way better than just handing it off 00:27:19.279 --> 00:27:20.880 to somebody that comes along and then 00:27:20.880 --> 00:27:22.640 really creating a supply chain attack, 00:27:22.640 --> 00:27:24.559 right? if that person cannot be trusted 00:27:24.559 --> 00:27:26.080 and then goes and inject in injects a 00:27:26.080 --> 00:27:27.360 back door and then all of those things 00:27:27.360 --> 00:27:30.159 that depend on it get backed whereas it 00:27:30.159 --> 00:27:32.720 should just be deprecated. So I think in 00:27:32.720 --> 00:27:34.400 terms of what you said like yes there 00:27:34.400 --> 00:27:36.480 should be a way to make sure that people 00:27:36.480 --> 00:27:38.320 that are maintaining things get 00:27:38.320 --> 00:27:40.640 compensation for it that's a very 00:27:40.640 --> 00:27:43.520 complicated topic. Yeah, like during the 00:27:43.520 --> 00:27:45.200 conference last week, I don't I don't 00:27:45.200 --> 00:27:49.200 know who mentioned it, but uh there was 00:27:49.200 --> 00:27:52.320 a best practice suggestion where like if 00:27:52.320 --> 00:27:54.080 you're if you're if you're developing a 00:27:54.080 --> 00:27:56.080 piece of software, you you want to 00:27:56.080 --> 00:27:58.399 modulize it so that it's m it's sort of 00:27:58.399 --> 00:28:01.200 maintainable beyond uh you know a green 00:28:01.200 --> 00:28:03.679 field oneshotted application. uh and 00:28:03.679 --> 00:28:05.279 that person suggested that you should 00:28:05.279 --> 00:28:08.559 modulize it and think of it like if you 00:28:08.559 --> 00:28:11.919 were using a NodeJS app what what could 00:28:11.919 --> 00:28:14.320 the boundaries 00:28:14.320 --> 00:28:16.960 could be like what what would what would 00:28:16.960 --> 00:28:19.360 typically be packages there's lots of 00:28:19.360 --> 00:28:20.880 schools of thought about this I don't 00:28:20.880 --> 00:28:24.720 know if you have have dabbled in uh what 00:28:24.720 --> 00:28:26.799 do you call it domain driven development 00:28:26.799 --> 00:28:28.640 but like but like if you're if you're 00:28:28.640 --> 00:28:30.640 developing your own software uh Vincent 00:28:30.640 --> 00:28:32.640 like do you ever think to yourself well 00:28:32.640 --> 00:28:35.039 this should a library or this could be a 00:28:35.039 --> 00:28:37.520 package or you're like monor repo all 00:28:37.520 --> 00:28:41.600 the way um you know drop drop depend I 00:28:41.600 --> 00:28:43.200 don't care for dependencies or something 00:28:43.200 --> 00:28:45.279 >> it depends on what I'm trying to build 00:28:45.279 --> 00:28:48.799 and like how much time I want to put it 00:28:48.799 --> 00:28:51.039 around it for me right now whenever I 00:28:51.039 --> 00:28:53.279 work on a project with AI I just give it 00:28:53.279 --> 00:28:55.520 yagy kiss you ain't going to need it 00:28:55.520 --> 00:28:57.520 keep it simple stupid I actually used 00:28:57.520 --> 00:29:00.480 this in a in a conversation the other 00:29:00.480 --> 00:29:03.600 time with a group of people And I I 00:29:03.600 --> 00:29:05.919 wanted to avoid complexity where they 00:29:05.919 --> 00:29:07.520 were like we need to put the GitHub 00:29:07.520 --> 00:29:09.120 workflow and the CI/CD in a separate 00:29:09.120 --> 00:29:10.799 dedicated repo and and I'm like dude 00:29:10.799 --> 00:29:12.559 this is like one workflow for one 00:29:12.559 --> 00:29:14.960 project. Let's keep it together in one 00:29:14.960 --> 00:29:17.919 in the same project. And so I wrote down 00:29:17.919 --> 00:29:20.799 I propose we kis and then later when I 00:29:20.799 --> 00:29:22.720 was rereading conversation I was like 00:29:22.720 --> 00:29:24.640 that's a bit weird thing to say to a 00:29:24.640 --> 00:29:26.080 group of people that you work with like 00:29:26.080 --> 00:29:28.399 I propose we kissed. 00:29:28.399 --> 00:29:29.520 >> Yeah. I was like, 00:29:29.520 --> 00:29:32.799 >> well, like the problem I feel is that I 00:29:32.799 --> 00:29:35.840 I in in some clients like that and of 00:29:35.840 --> 00:29:37.279 course we're consultants, we're sort of 00:29:37.279 --> 00:29:38.960 under pressure to staff more people, but 00:29:38.960 --> 00:29:40.799 like sometimes the teams are quite big, 00:29:40.799 --> 00:29:42.559 you know, you have sometimes the teams 00:29:42.559 --> 00:29:45.279 have to be decided up front, right? And 00:29:45.279 --> 00:29:47.039 then and then and then you have this 00:29:47.039 --> 00:29:48.720 problem where like who does what, who 00:29:48.720 --> 00:29:51.279 does this, who's that. So it's like a a 00:29:51.279 --> 00:29:53.840 non evolutionary way of doing it. So 00:29:53.840 --> 00:29:57.600 this is why people domain model and it 00:29:57.600 --> 00:29:59.840 can often be a bit weird to say the 00:29:59.840 --> 00:30:01.360 least. of you had 00:30:01.360 --> 00:30:03.840 >> absolutely like the whole idea the whole 00:30:03.840 --> 00:30:06.399 rage or hype of microservices was 00:30:06.399 --> 00:30:08.880 because of teams being split up and 00:30:08.880 --> 00:30:10.799 having to find better ways of like 00:30:10.799 --> 00:30:13.440 working together and the whole Conway 00:30:13.440 --> 00:30:16.480 law of you organize your services or 00:30:16.480 --> 00:30:18.080 your product within your organization 00:30:18.080 --> 00:30:19.840 based on the teams that you have right 00:30:19.840 --> 00:30:22.799 like what's the what are the the domains 00:30:22.799 --> 00:30:26.000 that that each team is responsible for 00:30:26.000 --> 00:30:27.120 >> and yeah 00:30:27.120 --> 00:30:28.880 >> but like I'm actually a fan of 00:30:28.880 --> 00:30:30.799 microservices at all. I've mentioned 00:30:30.799 --> 00:30:32.559 this publicly before. 00:30:32.559 --> 00:30:34.240 >> It always goes in circles, right? I 00:30:34.240 --> 00:30:37.279 mean, everybody goes, it goes like RPC, 00:30:37.279 --> 00:30:39.279 what's it? We have like monolith, then 00:30:39.279 --> 00:30:41.279 we have like microservices. Let's do 00:30:41.279 --> 00:30:43.679 event orchestration. We create an event 00:30:43.679 --> 00:30:46.000 bus, enterprise event bus from back. 00:30:46.000 --> 00:30:46.720 >> Oh my god. 00:30:46.720 --> 00:30:49.840 >> Kfka, not Kafka. Long before that with 00:30:49.840 --> 00:30:52.640 Java ERB, isn't it? That was like the 00:30:52.640 --> 00:30:55.039 pattern. The enterprise software pattern 00:30:55.039 --> 00:30:56.720 was an enterprise 00:30:56.720 --> 00:30:58.880 >> the enterprise robust ESB. 00:30:58.880 --> 00:31:01.360 >> ESB. Yeah. Not BRB. 00:31:01.360 --> 00:31:03.520 >> Yeah. Yeah. I'm definitely 00:31:03.520 --> 00:31:05.440 >> And then we go back, right? Because And 00:31:05.440 --> 00:31:07.760 then we go microservices. Then we go 00:31:07.760 --> 00:31:09.600 back. No, it needs to be monolithic. 00:31:09.600 --> 00:31:11.760 Modularize monoliths. Now let's do it 00:31:11.760 --> 00:31:13.200 with Kafka. And now we do event 00:31:13.200 --> 00:31:16.320 sourcing. Now we do event orchestration. 00:31:16.320 --> 00:31:18.159 >> I'm definitely in team monolith, but 00:31:18.159 --> 00:31:20.799 like I want to express that I want I 00:31:20.799 --> 00:31:22.960 want the app to be a mon monolith, but 00:31:22.960 --> 00:31:25.919 the dependencies can be different teams. 00:31:25.919 --> 00:31:28.880 If your if your app is a booking a 00:31:28.880 --> 00:31:31.279 booking app then perhaps there could be 00:31:31.279 --> 00:31:33.840 a dependency around I don't know what 00:31:33.840 --> 00:31:35.600 storage or something like that that can 00:31:35.600 --> 00:31:37.279 be handled by another team. I don't know 00:31:37.279 --> 00:31:39.360 if that that makes any sense but you 00:31:39.360 --> 00:31:41.039 know what I mean like it's better than 00:31:41.039 --> 00:31:43.360 it's better to have a monolith and 00:31:43.360 --> 00:31:46.880 different modular dependencies than an 00:31:46.880 --> 00:31:48.960 almighty enterprise serial bus 00:31:48.960 --> 00:31:51.200 architecture. I feel I mean like it 00:31:51.200 --> 00:31:52.960 depends on because when you earlier you 00:31:52.960 --> 00:31:54.720 were saying like your teams they get 00:31:54.720 --> 00:31:56.480 pre-allocated and designed who's going 00:31:56.480 --> 00:31:58.799 to work on what and it depends on the 00:31:58.799 --> 00:32:01.600 size the project what is the the target 00:32:01.600 --> 00:32:03.519 and then it will you know the best 00:32:03.519 --> 00:32:05.440 architectural patterns need to fit I'm 00:32:05.440 --> 00:32:06.960 on a project right now that's been 00:32:06.960 --> 00:32:09.919 dragging for 2 years and they had this 00:32:09.919 --> 00:32:12.399 massive like almost waterfall like 00:32:12.399 --> 00:32:14.880 planning and design and architecture and 00:32:14.880 --> 00:32:17.919 then teams set up and the biggest 00:32:17.919 --> 00:32:19.760 problem that they had after 2 years was 00:32:19.760 --> 00:32:23.360 that people took 2 months to build a PC 00:32:23.360 --> 00:32:25.840 or to to prove that it would work and 00:32:25.840 --> 00:32:28.399 then they would get stuck and so on. So 00:32:28.399 --> 00:32:30.559 I'm lucky when I came into the project 00:32:30.559 --> 00:32:32.960 they literally on day one when they 00:32:32.960 --> 00:32:34.640 talked to me was like it takes us 2 00:32:34.640 --> 00:32:36.000 months to build something and then we 00:32:36.000 --> 00:32:38.159 don't even know if it works. So u we 00:32:38.159 --> 00:32:40.080 need people that can go run off and do a 00:32:40.080 --> 00:32:43.360 P within within a sprint or two sprints 00:32:43.360 --> 00:32:45.840 and verify that it works. So I landed in 00:32:45.840 --> 00:32:46.240 the team 00:32:46.240 --> 00:32:47.760 >> most of the way it should be done really 00:32:47.760 --> 00:32:51.120 but carry on. Yes. And I landed and then 00:32:51.120 --> 00:32:53.679 it they reorganized into smaller work 00:32:53.679 --> 00:32:56.720 streams. There was refocus on doing 00:32:56.720 --> 00:32:59.519 proper scrum methodologies and it's a 00:32:59.519 --> 00:33:02.159 very interesting project really in the 00:33:02.159 --> 00:33:04.240 end because I saw something that wasn't 00:33:04.240 --> 00:33:06.240 working when I joined. I wasn't part of 00:33:06.240 --> 00:33:08.559 I'm not a a scrum master. I saw scrum 00:33:08.559 --> 00:33:10.880 masters joining the same day I started 00:33:10.880 --> 00:33:13.519 and then suggesting changes in how we 00:33:13.519 --> 00:33:15.519 should work together creating a scrum of 00:33:15.519 --> 00:33:17.039 scrums and all that. I I don't really 00:33:17.039 --> 00:33:18.320 don't care too much. 00:33:18.320 --> 00:33:19.679 >> I don't even know what that means, but 00:33:19.679 --> 00:33:21.360 I'm I'm I'm guessing it it's just some 00:33:21.360 --> 00:33:24.159 weird way of basically doing agile. 00:33:24.159 --> 00:33:25.760 >> Yeah. Um but it's also 00:33:25.760 --> 00:33:28.480 >> iterations. It's also decoupling like 00:33:28.480 --> 00:33:30.399 the teams into smaller teams and 00:33:30.399 --> 00:33:32.240 reducing the dependencies between people 00:33:32.240 --> 00:33:34.480 by by making more individual work 00:33:34.480 --> 00:33:36.159 streams. And it gave me the opportunity 00:33:36.159 --> 00:33:38.720 to like for example when I I told you 00:33:38.720 --> 00:33:40.559 about let's keep it simple and let's do 00:33:40.559 --> 00:33:43.200 this. I've been able to just build stuff 00:33:43.200 --> 00:33:45.600 and within one sprint deliver and show 00:33:45.600 --> 00:33:48.080 that it works and they are much happier 00:33:48.080 --> 00:33:50.000 with this because they see progress. 00:33:50.000 --> 00:33:52.240 They don't wait one month for it to be 00:33:52.240 --> 00:33:54.399 you know ready. So um it's been an 00:33:54.399 --> 00:33:56.080 interesting experience but I can 00:33:56.080 --> 00:33:58.320 definitely see that like I'm usually I 00:33:58.320 --> 00:34:00.159 definitely was was the one that would be 00:34:00.159 --> 00:34:01.919 this is tech dab we cannot do this it 00:34:01.919 --> 00:34:03.679 had to be done proper at this point I'm 00:34:03.679 --> 00:34:06.399 way more um maybe also because of AI I'm 00:34:06.399 --> 00:34:08.879 like you know with AI help me identify 00:34:08.879 --> 00:34:10.639 the shortest way to get this done and 00:34:10.639 --> 00:34:12.079 then we just take a couple of shortcuts 00:34:12.079 --> 00:34:12.960 we write them down 00:34:12.960 --> 00:34:16.079 >> I think DH his the latest podcast with 00:34:16.079 --> 00:34:18.800 him and Greg Gori he was he was 00:34:18.800 --> 00:34:20.800 basically saying that like he thinks 00:34:20.800 --> 00:34:22.960 that it just needs one person to 00:34:22.960 --> 00:34:28.079 bootstrap a project with PC but but okay 00:34:28.079 --> 00:34:29.599 we talked about like green field 00:34:29.599 --> 00:34:32.399 projects here for brownfield projects 00:34:32.399 --> 00:34:34.720 like say if there's an existing 00:34:34.720 --> 00:34:37.280 enterprise you know a massive app that's 00:34:37.280 --> 00:34:40.480 already working I guess it's easier then 00:34:40.480 --> 00:34:43.280 to shape the teams I suppose but even 00:34:43.280 --> 00:34:45.359 then it can go wrong but I mean it's 00:34:45.359 --> 00:34:46.879 it's easier to shape the team because 00:34:46.879 --> 00:34:49.919 they they they organically evolved to 00:34:49.919 --> 00:34:52.560 have teams behind 00:34:52.560 --> 00:34:54.480 the enterprise app already. Yeah. I 00:34:54.480 --> 00:34:56.720 think what I'm trying to what what I'm a 00:34:56.720 --> 00:34:58.400 little bit just bewildered by is just 00:34:58.400 --> 00:35:00.480 it's just that like the the brownfield 00:35:00.480 --> 00:35:03.839 and AI agents still hasn't really 00:35:03.839 --> 00:35:06.400 um hit the ground running for me. 00:35:06.400 --> 00:35:08.720 >> For me it actually because again I come 00:35:08.720 --> 00:35:10.480 into this project a 2-year-old project 00:35:10.480 --> 00:35:12.640 that people's been working on and I have 00:35:12.640 --> 00:35:14.800 to land in two years of I don't know 00:35:14.800 --> 00:35:16.160 people come and go. It's like a 00:35:16.160 --> 00:35:19.520 revolving door. So things change and the 00:35:19.520 --> 00:35:21.119 whole setup this is definitely a 00:35:21.119 --> 00:35:23.200 brownfield setup right it's definitely 00:35:23.200 --> 00:35:24.720 brown it's not green field you say it's 00:35:24.720 --> 00:35:26.640 green field but it's brownfield because 00:35:26.640 --> 00:35:29.040 it's 2 years of people coming in and out 00:35:29.040 --> 00:35:31.200 and the whole thing being a massive mess 00:35:31.200 --> 00:35:31.920 honestly 00:35:31.920 --> 00:35:35.839 >> well let's talk I mean brown I mean it's 00:35:35.839 --> 00:35:38.320 a a brownfield software project is do 00:35:38.320 --> 00:35:40.400 you have existing users do you have 00:35:40.400 --> 00:35:42.079 transactions and 00:35:42.079 --> 00:35:43.920 >> in that in that way it's not because 00:35:43.920 --> 00:35:45.599 it's still under under 00:35:45.599 --> 00:35:47.599 >> development at the time I'm I'm I what I 00:35:47.599 --> 00:35:50.079 want to say is that AI has been a 00:35:50.079 --> 00:35:52.960 godsend to identify things because when 00:35:52.960 --> 00:35:55.040 I'm not sure where to land a change I 00:35:55.040 --> 00:35:56.880 just tell cloud go and launch and 00:35:56.880 --> 00:35:58.880 explore agents and go answer this qu to 00:35:58.880 --> 00:36:01.200 answer this particular question and I 00:36:01.200 --> 00:36:02.880 maybe know some history after working on 00:36:02.880 --> 00:36:05.119 the codebase for 2 weeks and I know that 00:36:05.119 --> 00:36:06.640 there's a migration between this and 00:36:06.640 --> 00:36:09.119 that and so cloud will just suggest do 00:36:09.119 --> 00:36:11.040 this over here and I'm like hold on 00:36:11.040 --> 00:36:13.440 because I think there's multiple paths 00:36:13.440 --> 00:36:15.040 identify the different ways that thing 00:36:15.040 --> 00:36:16.640 that things are being done right now in 00:36:16.640 --> 00:36:18.800 aspect and it just goes off several 00:36:18.800 --> 00:36:20.960 explore agents. It then collates the 00:36:20.960 --> 00:36:21.839 result. 00:36:21.839 --> 00:36:24.800 >> Your software is your the software that 00:36:24.800 --> 00:36:26.720 you're working with is bloody complex. 00:36:26.720 --> 00:36:30.480 >> It's it's it's it's a classic like um a 00:36:30.480 --> 00:36:32.800 project with too much money that gets 00:36:32.800 --> 00:36:34.880 too many vendors involved and that buys 00:36:34.880 --> 00:36:36.480 every solution under the sun. I think 00:36:36.480 --> 00:36:38.000 it's way over complicated for what it 00:36:38.000 --> 00:36:40.160 has to do. But then I don't know. I 00:36:40.160 --> 00:36:42.800 don't even want to know the full 00:36:42.800 --> 00:36:44.480 >> I just want to be focused on my thing 00:36:44.480 --> 00:36:46.320 and deliver what I need to deliver. 00:36:46.320 --> 00:36:48.720 >> Yeah. True. True. I guess if I'm going 00:36:48.720 --> 00:36:50.640 into a brownfield project, I guess in 00:36:50.640 --> 00:36:52.480 the short term I can just expect agents 00:36:52.480 --> 00:36:54.800 or AI just to help me easier make 00:36:54.800 --> 00:36:57.200 changes. Though 00:36:57.200 --> 00:36:59.280 I guess I'm always looking for the next 00:36:59.280 --> 00:37:01.520 big thing. 00:37:01.520 --> 00:37:02.240 Uh 00:37:02.240 --> 00:37:03.599 >> I think that's what they say about the 00:37:03.599 --> 00:37:05.839 adoption of AI. Like at first people use 00:37:05.839 --> 00:37:09.280 AI to ask um like as an autocomplete 00:37:09.280 --> 00:37:12.480 alternative. Then they use AI to just 00:37:12.480 --> 00:37:14.240 ask questions about the codebase to 00:37:14.240 --> 00:37:15.440 understand where things live. 00:37:15.440 --> 00:37:16.960 >> Great for learning like if I don't 00:37:16.960 --> 00:37:18.480 understand something I'm like what is 00:37:18.480 --> 00:37:20.240 this? And then it just tells me oh thank 00:37:20.240 --> 00:37:21.599 god 00:37:21.599 --> 00:37:24.079 >> you know no need to Google or 00:37:24.079 --> 00:37:26.160 >> I'm working on infrastructure. 00:37:26.160 --> 00:37:26.640 >> Pardon? 00:37:26.640 --> 00:37:28.720 >> I'm working on infrastructure. So I also 00:37:28.720 --> 00:37:31.680 give it like AWS CLI access and then it 00:37:31.680 --> 00:37:34.079 collates the infrastructure 00:37:34.079 --> 00:37:35.839 configuration against what's actually in 00:37:35.839 --> 00:37:38.000 in the cloud. And so sometimes it tells 00:37:38.000 --> 00:37:39.839 me something and I'm like can you go 00:37:39.839 --> 00:37:42.000 check is that really true? Is there no 00:37:42.000 --> 00:37:44.000 other instance of this or whatever? And 00:37:44.000 --> 00:37:45.599 it just goes and does a couple of AWS 00:37:45.599 --> 00:37:47.760 CLI queries and finds out what's already 00:37:47.760 --> 00:37:48.720 there. 00:37:48.720 --> 00:37:52.640 >> AWS CLI the AWS MCPS are so useful. I 00:37:52.640 --> 00:37:53.040 find 00:37:53.040 --> 00:37:55.599 >> I'm not using any MCP for some reason it 00:37:55.599 --> 00:37:57.920 works. Yeah, maybe I should. 00:37:57.920 --> 00:37:58.960 >> You should you should definitely have 00:37:58.960 --> 00:38:01.920 the AWS docs MC. I think it it needs to 00:38:01.920 --> 00:38:04.000 be a protocol because navigating AWS 00:38:04.000 --> 00:38:05.359 documentation is pretty hard. Oh, that 00:38:05.359 --> 00:38:06.880 reminds me. I need to do my AWS 00:38:06.880 --> 00:38:09.359 certifications. Oh my god, I'm not 00:38:09.359 --> 00:38:10.480 looking forward to doing that. 00:38:10.480 --> 00:38:12.160 >> You need to pay for your work. 00:38:12.160 --> 00:38:14.240 >> Yeah, cuz we're a partner and the more 00:38:14.240 --> 00:38:16.240 points the more SS I have, the more 00:38:16.240 --> 00:38:17.839 points we have, the more likely they're 00:38:17.839 --> 00:38:19.200 going to give us deals. There's actually 00:38:19.200 --> 00:38:21.520 an AWS summit next week which I'm 00:38:21.520 --> 00:38:22.800 probably going to go to 00:38:22.800 --> 00:38:24.560 >> in London. Yeah, 00:38:24.560 --> 00:38:27.119 >> there's one summit in Singapore soon as 00:38:27.119 --> 00:38:28.400 well, but 00:38:28.400 --> 00:38:28.880 >> yeah, 00:38:28.880 --> 00:38:30.560 >> I have a very fun story. I have to 00:38:30.560 --> 00:38:32.960 request a new Belgian passport and I 00:38:32.960 --> 00:38:34.320 live in Ho Chi Min City. 00:38:34.320 --> 00:38:35.760 >> Oh, no. This doesn't sound like a fun 00:38:35.760 --> 00:38:36.160 story. 00:38:36.160 --> 00:38:39.040 >> It's a super fun story because 00:38:39.040 --> 00:38:41.119 >> of course it's not fun to request a new 00:38:41.119 --> 00:38:43.599 passport. So, you find a way to send the 00:38:43.599 --> 00:38:46.640 birth certificate. No. So there's a 00:38:46.640 --> 00:38:50.079 embassy in Hanoi and I was going to go 00:38:50.079 --> 00:38:52.560 there for a month and stay in a hotel 00:38:52.560 --> 00:38:53.680 while I requested 00:38:53.680 --> 00:38:55.839 >> Hanoi is so different to Ho Chi Min. But 00:38:55.839 --> 00:38:57.359 anyway, you know this, I'm sure. 00:38:57.359 --> 00:38:59.920 >> So, but they're also doing missions down 00:38:59.920 --> 00:39:02.560 to the south with their equipment to do 00:39:02.560 --> 00:39:06.000 biometrical data um collection. So then 00:39:06.000 --> 00:39:07.520 you can actually 00:39:07.520 --> 00:39:08.160 >> thing, right? 00:39:08.160 --> 00:39:10.079 >> Yeah. But honestly, if you already have 00:39:10.079 --> 00:39:12.320 a biometrical passport, do you need to 00:39:12.320 --> 00:39:14.640 retake those biometrical data every time 00:39:14.640 --> 00:39:15.760 you request a passport? 00:39:15.760 --> 00:39:17.520 >> Well, I should hope so in a way. 00:39:17.520 --> 00:39:20.240 >> Why? My biometricals aren't changing. 00:39:20.240 --> 00:39:22.160 Still the same person unless I lost a 00:39:22.160 --> 00:39:23.359 finger. I need another one. 00:39:23.359 --> 00:39:26.000 >> I guess I'm always thinking of it like 00:39:26.000 --> 00:39:28.720 like my iPhone because with iPhone you 00:39:28.720 --> 00:39:30.400 have to reset it up, right? 00:39:30.400 --> 00:39:31.839 >> But I guess Huh. 00:39:31.839 --> 00:39:33.839 >> I don't know. I don't buy a new iPhone 00:39:33.839 --> 00:39:35.599 >> because because the idea with the secure 00:39:35.599 --> 00:39:37.599 enclave is that your biometrics are not 00:39:37.599 --> 00:39:38.320 leaked. 00:39:38.320 --> 00:39:41.280 >> Okay. So, so there there was an email in 00:39:41.280 --> 00:39:43.040 my Gmail that's registered with the 00:39:43.040 --> 00:39:45.599 embassy that there will be a mission and 00:39:45.599 --> 00:39:47.200 I have to make an appointment. Yeah. 00:39:47.200 --> 00:39:48.720 They call it the biometrical mission 00:39:48.720 --> 00:39:49.680 down. 00:39:49.680 --> 00:39:50.079 >> Yeah. 00:39:50.079 --> 00:39:51.520 >> So, you have to make an appointment. 00:39:51.520 --> 00:39:53.040 >> They call these things missions. 00:39:53.040 --> 00:39:55.200 >> The email is in Dutch and French, not in 00:39:55.200 --> 00:39:58.000 English. Okay. And I don't want Yeah. 00:39:58.000 --> 00:39:59.359 Yeah. And the email has a whole bunch of 00:39:59.359 --> 00:40:01.040 links and like way too much information 00:40:01.040 --> 00:40:03.200 for me to read what I have to do because 00:40:03.200 --> 00:40:04.160 I just want the stupid. 00:40:04.160 --> 00:40:06.000 >> This is where you pasted into the agent. 00:40:06.000 --> 00:40:08.079 >> No, this is where I wanted to test cloud 00:40:08.079 --> 00:40:13.200 code work co-work sorry. So I made sure 00:40:13.200 --> 00:40:15.440 it has access to my Gmail as everyone 00:40:15.440 --> 00:40:17.599 should give AI access to you their Gmail 00:40:17.599 --> 00:40:20.400 and I took a couple of pictures of my 00:40:20.400 --> 00:40:22.000 passport because I want of course 00:40:22.000 --> 00:40:23.680 Antropic to have my passport and 00:40:23.680 --> 00:40:26.160 everything about me and grace ideas, 00:40:26.160 --> 00:40:28.320 right? That's why it became fun. It's 00:40:28.320 --> 00:40:30.560 exciting. Waiting for the identity 00:40:30.560 --> 00:40:32.400 theft. Maybe I should ask 00:40:32.400 --> 00:40:34.640 >> then Claude became Vincent Dismmet and 00:40:34.640 --> 00:40:36.480 you're like, "No, I'm Vincent Dismmet." 00:40:36.480 --> 00:40:38.480 And then say, "No, I'm Vincent." 00:40:38.480 --> 00:40:40.240 >> Maybe I should ask Cloud to also set up 00:40:40.240 --> 00:40:42.560 an identity theft monitor at the same 00:40:42.560 --> 00:40:44.400 time as you give it all of that 00:40:44.400 --> 00:40:46.320 information. And I just told it, look, 00:40:46.320 --> 00:40:48.480 there is in my Gmail an email from the 00:40:48.480 --> 00:40:49.920 embassy that they're coming on a 00:40:49.920 --> 00:40:51.839 biometrical mission. But again, 00:40:51.839 --> 00:40:54.000 biometrical mission is the English term. 00:40:54.000 --> 00:40:57.440 the the Duchess bometrica missi so it's 00:40:57.440 --> 00:40:59.920 different words and I said look I need 00:40:59.920 --> 00:41:02.079 to request a new passport I put some 00:41:02.079 --> 00:41:04.240 pictures in this folder that I put into 00:41:04.240 --> 00:41:06.640 your workspace and help me do this so 00:41:06.640 --> 00:41:08.319 that I don't have to do the boring 00:41:08.319 --> 00:41:09.920 things as we all know working with 00:41:09.920 --> 00:41:13.200 embassies and so on is delightful so 00:41:13.200 --> 00:41:14.960 searched and initially didn't find but 00:41:14.960 --> 00:41:16.880 then did expand the search parameters 00:41:16.880 --> 00:41:18.720 and I don't know how but it found the 00:41:18.720 --> 00:41:21.119 email from the Belgian embassy maybe the 00:41:21.119 --> 00:41:22.560 email address Belgian embassy kind of 00:41:22.560 --> 00:41:24.960 gave it Okay. But it it it read the 00:41:24.960 --> 00:41:27.040 whole thing in in Dutch and French and 00:41:27.040 --> 00:41:29.520 it then gave me a little like qu like 00:41:29.520 --> 00:41:31.280 you know it gives these questions. It 00:41:31.280 --> 00:41:33.280 was like first of which um which 00:41:33.280 --> 00:41:34.880 appointment do you want because they 00:41:34.880 --> 00:41:36.400 only have two time slots. It's either 00:41:36.400 --> 00:41:39.119 Thursday afternoon or Friday morning. I 00:41:39.119 --> 00:41:40.480 hadn't even figured that out. I thought 00:41:40.480 --> 00:41:41.920 it was either the whole Thursday or 00:41:41.920 --> 00:41:43.839 Friday. And inside email it's like 00:41:43.839 --> 00:41:45.520 dunder 00:41:45.520 --> 00:41:48.240 which is Thursday afternoon in Dutch or 00:41:48.240 --> 00:41:50.720 Friday morning vak morgan. And so it had 00:41:50.720 --> 00:41:52.800 translated everything, put it nicely for 00:41:52.800 --> 00:41:55.200 me to choose which appointment I wanted 00:41:55.200 --> 00:41:57.760 and downloaded the forms through the PDF 00:41:57.760 --> 00:41:59.040 from the email. It was going to the 00:41:59.040 --> 00:42:01.359 Belgian embassy website. It opened a 00:42:01.359 --> 00:42:03.680 Chrome session. It tried to curl it, but 00:42:03.680 --> 00:42:05.599 the sandbox didn't give it access. So 00:42:05.599 --> 00:42:07.760 then it used computer use to launch a 00:42:07.760 --> 00:42:10.640 Chrome session and then navigate to the 00:42:10.640 --> 00:42:13.359 embassy website. Found the PDF forms and 00:42:13.359 --> 00:42:14.720 I thought I was going to help it, but I 00:42:14.720 --> 00:42:16.560 downloaded like all the forms and then I 00:42:16.560 --> 00:42:18.400 said I confused it. It was like why does 00:42:18.400 --> 00:42:19.839 it keep downloading the forms? I'm just 00:42:19.839 --> 00:42:21.599 trying to I was I was interfering with 00:42:21.599 --> 00:42:23.280 its computer use and then I said okay 00:42:23.280 --> 00:42:25.839 stop. I was trying to be helpful and I 00:42:25.839 --> 00:42:27.680 downloaded the forms here and then it 00:42:27.680 --> 00:42:29.440 told me you downloaded the wrong form. 00:42:29.440 --> 00:42:30.800 You should have downloaded only this 00:42:30.800 --> 00:42:32.240 form. You downloaded the other form 00:42:32.240 --> 00:42:34.640 which is another procedure that you 00:42:34.640 --> 00:42:36.480 don't need. I was like okay sorry it was 00:42:36.480 --> 00:42:38.480 all in Dutch by the way. And then it I 00:42:38.480 --> 00:42:40.160 had forwarded the Belgian embassy email 00:42:40.160 --> 00:42:43.040 to my wife's email and it said is your 00:42:43.040 --> 00:42:44.800 wife's name because on the PDF document 00:42:44.800 --> 00:42:46.560 you have to say marital statement and 00:42:46.560 --> 00:42:49.119 and so on. Is this your wife's name? And 00:42:49.119 --> 00:42:50.640 in between bracket it says I saw you 00:42:50.640 --> 00:42:52.640 forward the email to this person so I 00:42:52.640 --> 00:42:54.960 assume that's your wife. Yes. I was like 00:42:54.960 --> 00:42:56.720 what? And then it just filled out 00:42:56.720 --> 00:42:59.119 everything from my passport picture. It 00:42:59.119 --> 00:43:01.680 filled out my name, asked for my my uh 00:43:01.680 --> 00:43:04.000 home address. I provided it 00:43:04.000 --> 00:43:05.839 >> and it filled out all of the PDFs. 00:43:05.839 --> 00:43:08.000 Marked it that my marital statements. 00:43:08.000 --> 00:43:09.839 Marked everything on the PDF form with 00:43:09.839 --> 00:43:11.440 little crosses. Filled out the whole 00:43:11.440 --> 00:43:12.880 because you know how annoying it is to 00:43:12.880 --> 00:43:14.480 fill out a PDF, right? I I don't know 00:43:14.480 --> 00:43:15.599 what you have to do. 00:43:15.599 --> 00:43:16.240 >> Yeah. Yeah. 00:43:16.240 --> 00:43:18.560 >> Acrobat reader. A a lot of the time I 00:43:18.560 --> 00:43:21.520 use AI as like an OCR, right? I just 00:43:21.520 --> 00:43:24.079 copy and paste some screenshot like what 00:43:24.079 --> 00:43:26.640 >> the whole form is in in in Dutch and it 00:43:26.640 --> 00:43:28.160 it asks you how many pages do you want 00:43:28.160 --> 00:43:29.599 your passport? 30 32. 00:43:29.599 --> 00:43:31.680 >> Did it actually fill in a PDF for you? 00:43:31.680 --> 00:43:33.520 >> It filled out the PDF. It asked me how 00:43:33.520 --> 00:43:35.040 many pages do you want in your passport? 00:43:35.040 --> 00:43:35.920 >> Sign at the bottom. 00:43:35.920 --> 00:43:37.359 >> I didn't check if I had to sign 00:43:37.359 --> 00:43:39.040 anything. You're right. Actually, 00:43:39.040 --> 00:43:40.800 when I do a passport, you have to sign. 00:43:40.800 --> 00:43:41.920 But I think that's going to be on the 00:43:41.920 --> 00:43:43.040 appointment. I'm probably going to have 00:43:43.040 --> 00:43:44.720 to sign there because I do sign the 00:43:44.720 --> 00:43:47.040 passport. But anyway, it then drafted 00:43:47.040 --> 00:43:49.119 the whole email for me, wrote it all out 00:43:49.119 --> 00:43:51.040 in Dutch as a reply to the original 00:43:51.040 --> 00:43:53.200 Dutch email. So, it asked me like, 00:43:53.200 --> 00:43:54.800 "What's your preferred language, Dutch 00:43:54.800 --> 00:43:56.480 or French?" And I was like, "Yeah, let's 00:43:56.480 --> 00:43:58.400 do Dutch." And and it wrote out the 00:43:58.400 --> 00:44:00.400 email in Dutch better than I could at 00:44:00.400 --> 00:44:02.800 this moment. Fully written out. 00:44:02.800 --> 00:44:04.560 >> That was Claude Co-work, right? 00:44:04.560 --> 00:44:06.319 >> Co-work. Yeah. I just 00:44:06.319 --> 00:44:10.400 >> feature I don't have 00:44:10.400 --> 00:44:12.720 >> I think OpenAI released Codex today 00:44:12.720 --> 00:44:14.560 which does something similar. Yeah. I 00:44:14.560 --> 00:44:16.560 mean it's amazing. It's amazing. 00:44:16.560 --> 00:44:18.240 >> I mean the the most amazing thing was 00:44:18.240 --> 00:44:20.880 that it it filtered out the information 00:44:20.880 --> 00:44:23.200 needed. It highlighted and got the 00:44:23.200 --> 00:44:25.280 clarification for me that it needed. It 00:44:25.280 --> 00:44:27.119 asked me also, have you registered with 00:44:27.119 --> 00:44:28.960 the Belgian embassy in Vietnam? And I 00:44:28.960 --> 00:44:32.560 said like I did back in 2006. 00:44:32.560 --> 00:44:34.720 I even got married. So I definitely have 00:44:34.720 --> 00:44:36.880 like all of my legal stuff 00:44:36.880 --> 00:44:38.720 >> with the Belgian embassy in Vietnam. But 00:44:38.720 --> 00:44:40.800 I did live five years in Singapore. And 00:44:40.800 --> 00:44:42.640 I don't think I signed out. I think I 00:44:42.640 --> 00:44:45.359 kept my how you say residency with the 00:44:45.359 --> 00:44:45.920 Vietnamese. 00:44:45.920 --> 00:44:47.680 >> That's usually a huge mistake. 00:44:47.680 --> 00:44:49.520 >> Yeah, but I never signed out of Vietnam. 00:44:49.520 --> 00:44:52.000 I I lived in in Singapore, didn't sign 00:44:52.000 --> 00:44:53.839 out and then I moved back to Vietnam so 00:44:53.839 --> 00:44:55.440 I didn't have to go procedure again. 00:44:55.440 --> 00:44:56.560 >> Sorry, I thought you're talking about 00:44:56.560 --> 00:45:00.160 tax residency for Europe. I don't 00:45:00.160 --> 00:45:02.000 >> I was going to I just had the thought 00:45:02.000 --> 00:45:04.800 like if agents are so good at all these 00:45:04.800 --> 00:45:07.760 like really horrible tedious processes 00:45:07.760 --> 00:45:10.240 these horrible tedious processes are 00:45:10.240 --> 00:45:13.119 probably not incentivized to improve are 00:45:13.119 --> 00:45:13.440 they? 00:45:13.440 --> 00:45:15.520 >> So the embassy right that's also what I 00:45:15.520 --> 00:45:17.760 was thinking because I I looked at the 00:45:17.760 --> 00:45:19.440 form that AI had completed and I was 00:45:19.440 --> 00:45:22.160 like this is the same form from like 40 00:45:22.160 --> 00:45:24.319 years 30 years ago I'm not that old 20 00:45:24.319 --> 00:45:27.680 years ago. Okay. So yeah, 2006 it's the 00:45:27.680 --> 00:45:29.680 same bloody form. Yeah. So they are not 00:45:29.680 --> 00:45:32.319 changing anything in the embassy way of 00:45:32.319 --> 00:45:34.319 working collecting that. Although the 00:45:34.319 --> 00:45:36.960 Vietnam is doing a lot of electronic ID. 00:45:36.960 --> 00:45:38.400 Maybe in Belgium they're also doing it. 00:45:38.400 --> 00:45:39.920 Just just the Belgian embassy in Vietnam 00:45:39.920 --> 00:45:41.920 not doing it. Vietnam is doing a lot in 00:45:41.920 --> 00:45:43.760 terms of like even there's a hackathon 00:45:43.760 --> 00:45:45.680 ongoing right now which is in 00:45:45.680 --> 00:45:48.240 collaboration with some government and 00:45:48.240 --> 00:45:50.800 like in transportation education. 00:45:50.800 --> 00:45:52.400 >> They're doing big hackathons in Vietnam. 00:45:52.400 --> 00:45:54.960 But like but let's say let's go from 00:45:54.960 --> 00:45:57.440 maybe let's be hypothetically let's go 00:45:57.440 --> 00:45:58.880 from the other angle in the sense that 00:45:58.880 --> 00:46:01.119 like say everything's agent friendly 00:46:01.119 --> 00:46:02.880 super agent friendly surely it's going 00:46:02.880 --> 00:46:04.480 to get to the point where like agents 00:46:04.480 --> 00:46:06.960 are going to start creating passports 00:46:06.960 --> 00:46:08.720 for themselves or something like that 00:46:08.720 --> 00:46:10.000 that's what they already do right with 00:46:10.000 --> 00:46:12.079 agent mail and that's so silly I didn't 00:46:12.079 --> 00:46:14.560 realize but agent mail the proposition 00:46:14.560 --> 00:46:16.720 that they do and why they got like 00:46:16.720 --> 00:46:18.480 millions of dollars of investment and 00:46:18.480 --> 00:46:20.720 why combinator incubated is because 00:46:20.720 --> 00:46:22.560 email addresses are the way that you do 00:46:22.560 --> 00:46:23.839 everything and authenticate yourself. 00:46:23.839 --> 00:46:26.160 It's an identity. And so they provide 00:46:26.160 --> 00:46:28.640 email like a a mail system for agents to 00:46:28.640 --> 00:46:30.400 request their identity, an email 00:46:30.400 --> 00:46:32.560 address, and then interact with it, see 00:46:32.560 --> 00:46:35.200 the emails all via the API or via skills 00:46:35.200 --> 00:46:38.560 or via MCPS. So agent mail is now worth 00:46:38.560 --> 00:46:40.319 so much money because they provide 00:46:40.319 --> 00:46:41.359 identity to agents. 00:46:41.359 --> 00:46:44.319 >> The DHH uh podcast, he was saying that 00:46:44.319 --> 00:46:46.240 >> yeah, that hey works really well with 00:46:46.240 --> 00:46:47.760 agents and that he was surprised how 00:46:47.760 --> 00:46:49.599 open claw could create the account and 00:46:49.599 --> 00:46:51.839 that he created it and then he saw the 00:46:51.839 --> 00:46:53.599 Open claw do it all. Yeah, that was 00:46:53.599 --> 00:46:55.359 really it was very interesting because I 00:46:55.359 --> 00:46:56.640 thought it's funny because he's talking 00:46:56.640 --> 00:46:57.839 about 00:46:57.839 --> 00:46:58.960 >> Have you? 00:46:58.960 --> 00:47:00.720 >> No, I also haven't. But he the way he 00:47:00.720 --> 00:47:02.640 talked about it is like we saw email as 00:47:02.640 --> 00:47:04.480 a problem. We didn't think Gmail was the 00:47:04.480 --> 00:47:06.079 final solution. So we came up with a 00:47:06.079 --> 00:47:07.920 much better email system which is hey. 00:47:07.920 --> 00:47:09.200 And he says it works really well with 00:47:09.200 --> 00:47:09.839 agents too. 00:47:09.839 --> 00:47:11.839 >> Yeah. The way that he was trying to sell 00:47:11.839 --> 00:47:15.119 it like oh hey allows you to screen 00:47:15.119 --> 00:47:16.800 people who email you. I mean like 00:47:16.800 --> 00:47:18.079 >> and never have to 00:47:18.079 --> 00:47:20.240 >> don't people just hit exclamation mark 00:47:20.240 --> 00:47:22.800 if they come across some spam? I mean in 00:47:22.800 --> 00:47:25.599 Gmail I don't get any spam because I 00:47:25.599 --> 00:47:29.280 pretty much mark it as spam if I see 00:47:29.280 --> 00:47:31.280 >> he was saying that because some if 00:47:31.280 --> 00:47:33.839 somebody has your email they can contact 00:47:33.839 --> 00:47:36.240 you. you can decline to talk to them but 00:47:36.240 --> 00:47:38.319 they can still contact you and with hey 00:47:38.319 --> 00:47:40.960 you one time approve it and if you don't 00:47:40.960 --> 00:47:42.880 want to speak to them ever again 00:47:42.880 --> 00:47:43.440 >> I don't know if 00:47:43.440 --> 00:47:44.480 >> I don't want to speak to everyone again 00:47:44.480 --> 00:47:46.400 I just go you know shift exclamation 00:47:46.400 --> 00:47:47.839 mark and I'll never see that person 00:47:47.839 --> 00:47:48.319 again 00:47:48.319 --> 00:47:50.720 >> I haven't tried hey and I don't really 00:47:50.720 --> 00:47:53.440 if you check my inbox check my email so 00:47:53.440 --> 00:47:55.680 >> okay I I have two questions for you I've 00:47:55.680 --> 00:47:58.319 written it down and I' I god knows how 00:47:58.319 --> 00:47:59.680 long I've been speaking for and I still 00:47:59.680 --> 00:48:01.920 haven't asked you what since you're like 00:48:01.920 --> 00:48:03.760 a speck kick guy and you're maintaining 00:48:03.760 --> 00:48:05.520 spec tech ledger I believe I don't know 00:48:05.520 --> 00:48:08.000 if you still are or what the deal is but 00:48:08.000 --> 00:48:10.240 what do you think about code to spec 00:48:10.240 --> 00:48:11.839 other way around say you have a 00:48:11.839 --> 00:48:14.960 brownfield project like a booking site 00:48:14.960 --> 00:48:18.000 or what what is a very enterprisey thing 00:48:18.000 --> 00:48:20.400 I don't know a payment system I I don't 00:48:20.400 --> 00:48:22.800 know and you want to basically document 00:48:22.800 --> 00:48:26.240 that system in your spec ledger way can 00:48:26.240 --> 00:48:28.880 you go from from code to spec or you 00:48:28.880 --> 00:48:30.079 haven't really thought about it 00:48:30.079 --> 00:48:32.960 >> I don't really know I haven't really 00:48:32.960 --> 00:48:34.559 thought about it But also 00:48:34.559 --> 00:48:36.720 >> it's funny because my thesis as as 00:48:36.720 --> 00:48:39.680 university master's thesis was about 00:48:39.680 --> 00:48:42.559 visualizing legacy code systems about 00:48:42.559 --> 00:48:43.599 cobalt. So 00:48:43.599 --> 00:48:46.000 >> wow how old are you? 00:48:46.000 --> 00:48:48.559 >> Well I was it was old when I was at uni. 00:48:48.559 --> 00:48:50.160 Okay it's not that I was old but 00:48:50.160 --> 00:48:52.000 basically the I was lazy. I didn't come 00:48:52.000 --> 00:48:53.440 up with my own idea. So I had to pick 00:48:53.440 --> 00:48:56.559 whatever crumbs were left and I just 00:48:56.559 --> 00:48:57.760 picked a 00:48:57.760 --> 00:48:59.680 >> visualizing legacy system. Yeah sure. 00:48:59.680 --> 00:49:01.920 Why not? So you know these systems are 00:49:01.920 --> 00:49:04.640 worth a lot of money. And so my thesis 00:49:04.640 --> 00:49:07.359 promoter I don't know he actually wrote 00:49:07.359 --> 00:49:09.520 he was very much into talking and Lord 00:49:09.520 --> 00:49:11.839 of the Rings. So he built a tool called 00:49:11.839 --> 00:49:15.200 Yetaurunia which is looking at runes. 00:49:15.200 --> 00:49:17.920 Yeta is Elvish for looking apparently. 00:49:17.920 --> 00:49:20.400 And um runes is like footprints. So 00:49:20.400 --> 00:49:22.800 Yetunia was his project and it was 00:49:22.800 --> 00:49:26.880 basically um a parser for cobalt and 00:49:26.880 --> 00:49:29.440 then it he had already built some 00:49:29.440 --> 00:49:31.599 visualizations which is like you know in 00:49:31.599 --> 00:49:33.680 cobalt you can do gotos to jump to 00:49:33.680 --> 00:49:35.119 different sections 00:49:35.119 --> 00:49:38.559 >> so jumps and you can do like loops. So 00:49:38.559 --> 00:49:40.640 he would visualize those like he would 00:49:40.640 --> 00:49:43.520 parse the cobalt into an XML a and then 00:49:43.520 --> 00:49:45.520 the xml was converted into 00:49:45.520 --> 00:49:47.440 >> but then you get like a fancy graph. I 00:49:47.440 --> 00:49:49.680 often find code visualizations like the 00:49:49.680 --> 00:49:51.680 worst thing ever. But 00:49:51.680 --> 00:49:54.319 >> I I put I put on my thesis as like a 00:49:54.319 --> 00:49:56.319 picture is worth a thousand words as one 00:49:56.319 --> 00:49:56.880 of the reason 00:49:56.880 --> 00:49:59.760 >> like a like this like graph is meth. 00:49:59.760 --> 00:50:02.319 >> Yeah. But like again I didn't really did 00:50:02.319 --> 00:50:04.559 a really good job of my thesis. Honestly 00:50:04.559 --> 00:50:06.720 not very proud. So don't go look. So 00:50:06.720 --> 00:50:08.480 whatever I I delivered. 00:50:08.480 --> 00:50:11.839 >> My thesis is on um mobile web 00:50:11.839 --> 00:50:13.119 engineering and the 00:50:13.119 --> 00:50:15.200 >> I wish I I picked something dude. I wish 00:50:15.200 --> 00:50:16.480 I picked I would and you know to be 00:50:16.480 --> 00:50:18.079 honest and with how much I went into 00:50:18.079 --> 00:50:20.640 Kubernetes later when I and I then see 00:50:20.640 --> 00:50:22.880 people posting like I'm doing a thesis 00:50:22.880 --> 00:50:24.079 on Kubernetes and they're learning the 00:50:24.079 --> 00:50:25.920 step and I'm like I wish I had the 00:50:25.920 --> 00:50:28.240 opportunity to do at the time you know 00:50:28.240 --> 00:50:30.160 to to be because I would have been so 00:50:30.160 --> 00:50:32.319 deep into it. I'm I loved it you know 00:50:32.319 --> 00:50:33.599 exploring and breaking it down. 00:50:33.599 --> 00:50:35.839 >> Are you are you a Kubernetes fanboy? I I 00:50:35.839 --> 00:50:37.440 thought you just used it in anger. 00:50:37.440 --> 00:50:39.839 >> I organized the meetups right. I 00:50:39.839 --> 00:50:42.079 organized the Docker meetups in Okay. in 00:50:42.079 --> 00:50:43.280 Vietnam and then I organized the 00:50:43.280 --> 00:50:45.680 Kubernetes meetups. But but back to the 00:50:45.680 --> 00:50:47.839 point of of visualizing code and and 00:50:47.839 --> 00:50:51.760 going from code to spec um 00:50:51.760 --> 00:50:54.240 I okay so I think the problem with spec 00:50:54.240 --> 00:50:56.559 kit specifically and I haven't tried 00:50:56.559 --> 00:50:59.839 open spec but with specit creates your 00:50:59.839 --> 00:51:02.800 specs your user stories and then you 00:51:02.800 --> 00:51:05.040 deliver code. I actually read a very 00:51:05.040 --> 00:51:07.040 interesting post earlier today from 00:51:07.040 --> 00:51:08.319 someone. He says like the problem is 00:51:08.319 --> 00:51:10.400 those spec folders become a graveyard. 00:51:10.400 --> 00:51:11.760 And this is something that I also 00:51:11.760 --> 00:51:13.920 explore with my friends is 00:51:13.920 --> 00:51:15.040 >> become a graveyard. 00:51:15.040 --> 00:51:16.720 >> So what happens in specit is every time 00:51:16.720 --> 00:51:19.200 you want to add a new feature you create 00:51:19.200 --> 00:51:22.480 a branch for that feature and with that 00:51:22.480 --> 00:51:24.960 branch comes a little folder and inside 00:51:24.960 --> 00:51:26.880 that folder you generate your artifacts 00:51:26.880 --> 00:51:29.200 which are your your specification file 00:51:29.200 --> 00:51:31.040 which identifies your user stories that 00:51:31.040 --> 00:51:33.119 you're working on for that feature. 00:51:33.119 --> 00:51:35.119 >> Yeah. So once that feature is delivered 00:51:35.119 --> 00:51:37.200 in the codebase and it works or test 00:51:37.200 --> 00:51:41.040 pass you merge it right but then the 00:51:41.040 --> 00:51:43.440 specs the actual user stories for that 00:51:43.440 --> 00:51:45.520 feature live into that spec folder. 00:51:45.520 --> 00:51:48.079 There's no folder at the root that keeps 00:51:48.079 --> 00:51:50.640 track of what is the actual live spec 00:51:50.640 --> 00:51:52.880 across all of your application. Instead 00:51:52.880 --> 00:51:54.079 every single 00:51:54.079 --> 00:51:56.160 >> it's lacking a source of truth I suppose 00:51:56.160 --> 00:51:56.480 right? 00:51:56.480 --> 00:51:59.119 >> Yeah but I mean nothing stops you from 00:51:59.119 --> 00:52:01.680 as part of your constitution. I think we 00:52:01.680 --> 00:52:02.800 might have discussed this before because 00:52:02.800 --> 00:52:04.960 I remember saying this before. Nothing 00:52:04.960 --> 00:52:07.200 stops you from making sure in your 00:52:07.200 --> 00:52:09.119 constitution that not only do you 00:52:09.119 --> 00:52:11.440 create, you know, your spec inside this 00:52:11.440 --> 00:52:13.760 feature folder, but then when you want 00:52:13.760 --> 00:52:15.200 to close it off, you're landing the 00:52:15.200 --> 00:52:17.839 plane as as Tifit calls it when you want 00:52:17.839 --> 00:52:20.640 to land the feature, you at least as one 00:52:20.640 --> 00:52:22.480 of the steps because you generate a task 00:52:22.480 --> 00:52:24.000 list, right? And the task list is what 00:52:24.000 --> 00:52:25.920 gets checked off. So in one of those 00:52:25.920 --> 00:52:29.040 tasks, you should have an work item to 00:52:29.040 --> 00:52:31.440 go and update your central 00:52:31.440 --> 00:52:33.440 representation of your app and update 00:52:33.440 --> 00:52:35.200 all the user stories. Now, but what I 00:52:35.200 --> 00:52:37.280 read on the on on Reddit this morning 00:52:37.280 --> 00:52:40.720 was that and it's kind of like for me 00:52:40.720 --> 00:52:42.720 what really validates that the 00:52:42.720 --> 00:52:45.440 application works the way I expect it to 00:52:45.440 --> 00:52:47.839 work and there's no regression is tests, 00:52:47.839 --> 00:52:50.319 right? An end toend tests. So usually I 00:52:50.319 --> 00:52:53.680 have an E2E directory where I have my 00:52:53.680 --> 00:52:56.559 E2A flows and that's the actual stories 00:52:56.559 --> 00:52:58.559 that still work. I know because I run 00:52:58.559 --> 00:52:59.040 them 00:52:59.040 --> 00:53:03.440 >> actually that's test like instead of 00:53:03.440 --> 00:53:05.359 going code to spec you should go code to 00:53:05.359 --> 00:53:07.280 test. Yeah, you should you should have 00:53:07.280 --> 00:53:08.960 your regression test and I think in any 00:53:08.960 --> 00:53:10.800 system brown field system that wants to 00:53:10.800 --> 00:53:12.559 adopt AI, the first thing that you 00:53:12.559 --> 00:53:14.240 should adopt with AI is you write the 00:53:14.240 --> 00:53:16.160 tests because AI can write tests very 00:53:16.160 --> 00:53:17.920 quickly and as long as you can build a 00:53:17.920 --> 00:53:20.079 test harness because a lot of systems 00:53:20.079 --> 00:53:21.760 they don't have a proper test harness. 00:53:21.760 --> 00:53:23.359 You write all of the tests, make sure 00:53:23.359 --> 00:53:25.440 that they cover all of your code before 00:53:25.440 --> 00:53:27.680 you let AI work on the actual features 00:53:27.680 --> 00:53:28.800 or the code 00:53:28.800 --> 00:53:29.839 >> because then you know it's not going to 00:53:29.839 --> 00:53:31.200 break one of the previous tests. 00:53:31.200 --> 00:53:32.880 >> Yeah, that's that seems that seems like 00:53:32.880 --> 00:53:35.200 a no-brainer to me. I don't see people 00:53:35.200 --> 00:53:37.440 talking about that really. I think I 00:53:37.440 --> 00:53:39.520 need I need to I mean I I work with 00:53:39.520 --> 00:53:41.280 enterprise brownfield projects. I need 00:53:41.280 --> 00:53:43.839 to do it and report back. 00:53:43.839 --> 00:53:45.680 >> I think I I didn't come up with this. I 00:53:45.680 --> 00:53:47.599 think reading this as well like the 00:53:47.599 --> 00:53:50.240 first thing of AI adoption in any big 00:53:50.240 --> 00:53:52.079 enterprise project should be in the test 00:53:52.079 --> 00:53:54.240 environment because you takes so much 00:53:54.240 --> 00:53:54.720 off your head. 00:53:54.720 --> 00:53:56.559 >> I totally agree. Yeah. Yeah. I I need to 00:53:56.559 --> 00:53:58.160 explore this a bit more. Just changing 00:53:58.160 --> 00:54:00.480 the topic. Not changing the topic ever 00:54:00.480 --> 00:54:02.559 so slightly. So one one of the the 00:54:02.559 --> 00:54:05.839 Claude uh 00:54:05.839 --> 00:54:06.800 >> workshops 00:54:06.800 --> 00:54:11.440 >> quick starts quick shots quick starts is 00:54:11.440 --> 00:54:15.200 uh a project where they define 00:54:15.200 --> 00:54:17.599 the spec in this text file which is 00:54:17.599 --> 00:54:20.640 actually more like an XML file but I was 00:54:20.640 --> 00:54:22.880 just curious same with Symphony from 00:54:22.880 --> 00:54:23.920 open air 00:54:23.920 --> 00:54:25.280 >> I I don't know I don't know if you've 00:54:25.280 --> 00:54:27.119 seen this before but like I was just 00:54:27.119 --> 00:54:29.359 curious 00:54:29.359 --> 00:54:32.319 what you thought, well, maybe just maybe 00:54:32.319 --> 00:54:34.000 I just copy and paste this in the chat 00:54:34.000 --> 00:54:36.480 to you. I was curious if you thought the 00:54:36.480 --> 00:54:38.880 way that they define the project uh have 00:54:38.880 --> 00:54:41.680 you seen Symfony from Open EI in I don't 00:54:41.680 --> 00:54:45.119 know Haskell or or uh not Haskell uh 00:54:45.119 --> 00:54:48.240 Elixir, but they they have a similar 00:54:48.240 --> 00:54:51.040 like setup for for the for the Symfony 00:54:51.040 --> 00:54:53.839 project to to build. They said you want 00:54:53.839 --> 00:54:55.920 to build it, you just use the spec and 00:54:55.920 --> 00:54:59.040 then the spec is like an XML document. 00:54:59.040 --> 00:55:01.280 I don't know why they chose XML. I guess 00:55:01.280 --> 00:55:02.720 it doesn't really matter. 00:55:02.720 --> 00:55:05.760 >> I do whenever I want to give the AI an 00:55:05.760 --> 00:55:08.160 example, I use a little tag and say like 00:55:08.160 --> 00:55:11.119 example one and then slash example one. 00:55:11.119 --> 00:55:13.359 I do that. It's it's it's been 00:55:13.359 --> 00:55:16.160 documented that using XML inside your 00:55:16.160 --> 00:55:18.240 prompts is actually a good way to 00:55:18.240 --> 00:55:19.040 structure them. 00:55:19.040 --> 00:55:21.920 >> Okay. Whatever. So, here is the project 00:55:21.920 --> 00:55:23.760 specification. I was just curious if you 00:55:23.760 --> 00:55:25.520 think it it's a good one. I don't know 00:55:25.520 --> 00:55:26.960 if you've seen it before. like they have 00:55:26.960 --> 00:55:28.480 a project name, they have an overview, 00:55:28.480 --> 00:55:30.880 they have the technology stack. Oh, by 00:55:30.880 --> 00:55:32.640 the way, what does this thing do? It's 00:55:32.640 --> 00:55:35.280 basically a 00:55:35.280 --> 00:55:38.160 it's it's a clone of Claude AI. So, you 00:55:38.160 --> 00:55:41.280 can just chat to, you know, the model. 00:55:41.280 --> 00:55:43.119 Then in the technology stack, it sort of 00:55:43.119 --> 00:55:44.640 breaks down that there's going to be 00:55:44.640 --> 00:55:48.160 React and Node and endpoints 00:55:48.160 --> 00:55:51.040 and uh 00:55:51.040 --> 00:55:52.880 and it has a breakdown of the core 00:55:52.880 --> 00:55:54.799 features. I like Yeah, I'm just curious. 00:55:54.799 --> 00:55:57.520 Is this the way you would do it or do 00:55:57.520 --> 00:55:59.839 you think they're 00:55:59.839 --> 00:56:03.359 they're missing like for example I'm I'm 00:56:03.359 --> 00:56:05.599 guessing your does your spec kit go as 00:56:05.599 --> 00:56:07.040 far as defining the end point? 00:56:07.040 --> 00:56:09.359 >> Yes. Yeah. Yeah. So but the speckit does 00:56:09.359 --> 00:56:12.480 this across multiple artifacts and again 00:56:12.480 --> 00:56:15.839 speckit has these prompts um that the 00:56:15.839 --> 00:56:19.040 the I mean then the guy who built it um 00:56:19.040 --> 00:56:20.559 created together with the team at 00:56:20.559 --> 00:56:23.200 Microsoft originally almost a year ago. 00:56:23.200 --> 00:56:25.760 Then um it creates data model, it 00:56:25.760 --> 00:56:29.760 creates data contracts. Um so your AI I 00:56:29.760 --> 00:56:32.079 mean you I don't tell it what are the 00:56:32.079 --> 00:56:35.040 endpoints or the rest uh setup or the 00:56:35.040 --> 00:56:36.480 routes or anything like that like 00:56:36.480 --> 00:56:38.240 >> but you get but you get there eventually 00:56:38.240 --> 00:56:38.880 I'm sure. 00:56:38.880 --> 00:56:41.839 >> Yeah. Yeah. I like I I say with the user 00:56:41.839 --> 00:56:43.680 stories and then it goes and it says 00:56:43.680 --> 00:56:46.000 okay so this is a data contract like one 00:56:46.000 --> 00:56:49.599 time I used Protobuff. Um 00:56:49.599 --> 00:56:51.680 >> and then it it wrote down these are the 00:56:51.680 --> 00:56:55.520 services these are the RPC functions. 00:56:55.520 --> 00:56:57.839 >> What I'm trying to say is that when when 00:56:57.839 --> 00:57:00.640 you do spec it it asks you questions you 00:57:00.640 --> 00:57:03.599 go through a process but uh like what 00:57:03.599 --> 00:57:06.079 what is the final artifact in your in 00:57:06.079 --> 00:57:08.480 your spec ledger is is it like one file 00:57:08.480 --> 00:57:11.119 or is it like five markdown files or 00:57:11.119 --> 00:57:13.280 something? So it's a whole folder right 00:57:13.280 --> 00:57:16.640 it creates the spec.mmd it creates the 00:57:16.640 --> 00:57:18.640 um plan which is more like a research 00:57:18.640 --> 00:57:21.280 document about spec is just user stories 00:57:21.280 --> 00:57:23.280 no tech plan is where you put the 00:57:23.280 --> 00:57:25.760 technical stack definitions and data 00:57:25.760 --> 00:57:27.119 contracts and then 00:57:27.119 --> 00:57:28.640 >> it's probably just a matter of taste but 00:57:28.640 --> 00:57:30.960 I I do like the idea of just having like 00:57:30.960 --> 00:57:33.200 like one one text file 00:57:33.200 --> 00:57:35.920 >> and safe from openi does the same thing 00:57:35.920 --> 00:57:39.119 um so like if both entropic and open eye 00:57:39.119 --> 00:57:41.920 openai are giving us examples of how to 00:57:41.920 --> 00:57:44.000 define your products in a single file 00:57:44.000 --> 00:57:46.480 like this, right? Must be that's what 00:57:46.480 --> 00:57:47.680 their models are trained on. 00:57:47.680 --> 00:57:49.760 >> Yeah. So, yeah, I'm just thinking, 00:57:49.760 --> 00:57:51.520 wouldn't it be best if you if you sort 00:57:51.520 --> 00:57:53.760 of tweak your your spec ledger to to 00:57:53.760 --> 00:57:55.760 produce a single page artifact? 00:57:55.760 --> 00:57:56.880 >> No, I don't care. 00:57:56.880 --> 00:57:58.799 >> Is this is this talk about Symphony? 00:57:58.799 --> 00:58:00.640 Looks a bit strange. Looks like a 00:58:00.640 --> 00:58:02.720 >> So, the the critical thing that I found 00:58:02.720 --> 00:58:04.400 and I think is very interesting. I'm not 00:58:04.400 --> 00:58:05.920 sure if we did talk about this before, 00:58:05.920 --> 00:58:08.880 but and I think we did. 00:58:08.880 --> 00:58:10.559 they talk about the English lang 00:58:10.559 --> 00:58:12.319 language. There's a paper from I think 00:58:12.319 --> 00:58:15.119 it's Donald Kut who who wrote down there 00:58:15.119 --> 00:58:17.920 is a reason why um the English language 00:58:17.920 --> 00:58:19.520 is never a going to be a programming 00:58:19.520 --> 00:58:21.440 language because it does not formally 00:58:21.440 --> 00:58:24.319 define very precisely what is it that 00:58:24.319 --> 00:58:26.559 need to be done. So the whole idea of 00:58:26.559 --> 00:58:27.839 like we're never going to write a 00:58:27.839 --> 00:58:29.599 program in programming language anymore. 00:58:29.599 --> 00:58:30.720 We're actually just going to reuse 00:58:30.720 --> 00:58:32.960 English. It's not possible according to 00:58:32.960 --> 00:58:34.480 this 00:58:34.480 --> 00:58:36.559 thesis from Donald is it Donald? 00:58:36.559 --> 00:58:39.200 >> What about German? But like, yeah, fair 00:58:39.200 --> 00:58:40.480 enough. But that's the beauty of 00:58:40.480 --> 00:58:42.319 English. You can you have the freedom to 00:58:42.319 --> 00:58:43.040 express. 00:58:43.040 --> 00:58:44.799 >> Yeah, but it's not expressed. 00:58:44.799 --> 00:58:47.119 >> It's not a literal programming tool. 00:58:47.119 --> 00:58:50.400 It's not precise enough to to define a 00:58:50.400 --> 00:58:52.720 program. I I wrote this blog post 00:58:52.720 --> 00:58:54.799 looking at Symfony, which you just saw, 00:58:54.799 --> 00:58:56.160 but even the one that you just saw, the 00:58:56.160 --> 00:58:58.559 XML with every endpoint, and their their 00:58:58.559 --> 00:59:00.880 theirsis is basically we're saying that 00:59:00.880 --> 00:59:03.280 these specs are the English language 00:59:03.280 --> 00:59:05.839 representation of a program. But if you 00:59:05.839 --> 00:59:08.240 look at them, the actual symfony spec is 00:59:08.240 --> 00:59:09.920 almost as long as the elixir code 00:59:09.920 --> 00:59:12.319 implementing it. It's a is it because of 00:59:12.319 --> 00:59:14.240 like we're actually almost writing the 00:59:14.240 --> 00:59:16.640 program itself right there. Like 00:59:16.640 --> 00:59:18.240 ultimately it's going to boil down to 00:59:18.240 --> 00:59:20.160 the programming language itself is what 00:59:20.160 --> 00:59:21.599 they were saying. And I thought that's 00:59:21.599 --> 00:59:23.119 kind of an interesting way of looking at 00:59:23.119 --> 00:59:24.960 it. But that's why I like this user 00:59:24.960 --> 00:59:27.359 stories which are very clearly English 00:59:27.359 --> 00:59:29.839 like as a user I want this because this 00:59:29.839 --> 00:59:31.920 they're very decoupled from the actual 00:59:31.920 --> 00:59:33.839 programming code under the hood. 00:59:33.839 --> 00:59:35.839 Actually, this spec I got to look at the 00:59:35.839 --> 00:59:37.359 spec here. I mean, the spec is quite 00:59:37.359 --> 00:59:41.440 long. How many What does Symphfony do? 00:59:41.440 --> 00:59:43.599 Sorry. It's like 2,000 lines long. 00:59:43.599 --> 00:59:46.160 >> It's like um task tracker system if I 00:59:46.160 --> 00:59:46.799 don't remember. 00:59:46.799 --> 00:59:48.400 >> It's like the camboard. Oh, I see. I 00:59:48.400 --> 00:59:48.720 see. 00:59:48.720 --> 00:59:50.960 >> Issue execution. It's more like you you 00:59:50.960 --> 00:59:52.880 you create issues and you delegate to 00:59:52.880 --> 00:59:54.640 agents type of thing. 00:59:54.640 --> 00:59:57.119 >> The other thing that bothered me when we 00:59:57.119 --> 00:59:58.960 were in the workshop is like that 00:59:58.960 --> 01:00:00.240 anthropic was saying like, "Yeah, this 01:00:00.240 --> 01:00:03.599 is your app spec now. Now, you use uh 01:00:03.599 --> 01:00:05.440 your your builder agent and your 01:00:05.440 --> 01:00:07.599 evaluator agent to build this out. And 01:00:07.599 --> 01:00:09.760 then they said, "Well, this is but but 01:00:09.760 --> 01:00:11.920 beware. This is going to cost about $100 01:00:11.920 --> 01:00:13.680 worth of credits." So, I'm just thinking 01:00:13.680 --> 01:00:15.920 to myself, this this is probably not a 01:00:15.920 --> 01:00:17.440 very good way to develop software 01:00:17.440 --> 01:00:19.520 because if if if you're if you're if 01:00:19.520 --> 01:00:21.359 you're going to oneshot this thing from 01:00:21.359 --> 01:00:24.480 a spec and it's going to go wrong, 01:00:24.480 --> 01:00:26.559 then you've just wasted $100. You know 01:00:26.559 --> 01:00:27.920 what I mean, Vincent? 01:00:27.920 --> 01:00:30.319 >> Yeah. Well, it's it's the argument 01:00:30.319 --> 01:00:32.400 against people saying you can use it 01:00:32.400 --> 01:00:34.079 just to distribute the programming like 01:00:34.079 --> 01:00:37.200 a software because everyone that's what 01:00:37.200 --> 01:00:39.359 what like when Stripe talks about 01:00:39.359 --> 01:00:40.480 minions, they say we're not going to 01:00:40.480 --> 01:00:42.319 open open source minions and how it 01:00:42.319 --> 01:00:45.200 works. You just point cloud at this blog 01:00:45.200 --> 01:00:46.720 post and it will build your own version 01:00:46.720 --> 01:00:49.040 of it. Uh it's the same thing as what 01:00:49.040 --> 01:00:50.000 you just said, right? 01:00:50.000 --> 01:00:52.319 >> Yeah. 01:00:52.319 --> 01:00:54.079 I mean, I'm I'm I'm basically I'm 01:00:54.079 --> 01:00:56.720 basically just doubting specd driven 01:00:56.720 --> 01:00:58.480 development because like 01:00:58.480 --> 01:00:59.839 >> well, you're not alone. Like a lot of 01:00:59.839 --> 01:01:01.760 people there are just like um you need 01:01:01.760 --> 01:01:04.240 superpowers now. Don't you need to use 01:01:04.240 --> 01:01:06.400 GSD? Get done. If you're still 01:01:06.400 --> 01:01:09.359 using specit like 01:01:09.359 --> 01:01:11.760 what can you explain get done to 01:01:11.760 --> 01:01:12.400 me? I think 01:01:12.400 --> 01:01:15.359 >> GSD I have not used it but like I was um 01:01:15.359 --> 01:01:18.319 showing and explaining um I think we 01:01:18.319 --> 01:01:20.160 talked about this as well in the past. 01:01:20.160 --> 01:01:22.079 Yeah, I think you mentioned it before 01:01:22.079 --> 01:01:23.920 >> that I mentioned that I was showing 01:01:23.920 --> 01:01:26.640 specket to my friends and then they 01:01:26.640 --> 01:01:28.240 tried it and then they came back to me 01:01:28.240 --> 01:01:30.480 and said uh I stopped using specket. GSD 01:01:30.480 --> 01:01:32.160 is is is really good. 01:01:32.160 --> 01:01:34.240 >> It looks more like beads or something. 01:01:34.240 --> 01:01:36.559 >> No, it's more uh people say it's like it 01:01:36.559 --> 01:01:38.160 interviews you. It asks you the 01:01:38.160 --> 01:01:40.480 clarifying questions only and then it 01:01:40.480 --> 01:01:41.920 just goes off in a loop and then when 01:01:41.920 --> 01:01:43.440 you come back it gives you exactly what 01:01:43.440 --> 01:01:44.960 you want. 01:01:44.960 --> 01:01:46.480 >> So you build you build basically 01:01:46.480 --> 01:01:48.400 incrementally. I I guess what I'm 01:01:48.400 --> 01:01:51.839 incredulous about is that is okay, it's 01:01:51.839 --> 01:01:54.160 really good to have a spec. I mean, I'm 01:01:54.160 --> 01:01:56.640 a huge fan of the HML spec. It's great 01:01:56.640 --> 01:01:58.720 to have a spec, but like you're not 01:01:58.720 --> 01:02:00.400 going to bloody build the thing from the 01:02:00.400 --> 01:02:03.760 spec. That sounds that sounds crazy. I 01:02:03.760 --> 01:02:06.079 mean, am I am I losing it? Like, you 01:02:06.079 --> 01:02:08.400 can't build you can't expect to build 01:02:08.400 --> 01:02:12.480 workable software from Huh? It's Dextra. 01:02:12.480 --> 01:02:15.119 Kut actually said the opposite. Dextra 01:02:15.119 --> 01:02:18.799 said that the English is imprecise. Um 01:02:18.799 --> 01:02:20.880 he said that natural language is too 01:02:20.880 --> 01:02:22.720 ambiguous for reasoning about 01:02:22.720 --> 01:02:24.960 correctness. Programming requires formal 01:02:24.960 --> 01:02:26.960 systems to avoid misunderstanding. And 01:02:26.960 --> 01:02:29.920 then Kut actually created literal 01:02:29.920 --> 01:02:32.319 English as a programming language that 01:02:32.319 --> 01:02:34.240 is closer to English to 01:02:34.240 --> 01:02:34.559 >> okay 01:02:34.559 --> 01:02:36.960 >> a literal programming. Yeah. It's really 01:02:36.960 --> 01:02:39.119 >> like say so Vincent with your specs that 01:02:39.119 --> 01:02:41.680 you've derived through your your spec 01:02:41.680 --> 01:02:45.119 ledger, how many times have you sort of 01:02:45.119 --> 01:02:47.599 uh implemented it? Do you implement it 01:02:47.599 --> 01:02:49.920 multiple times? Do you It's like do you 01:02:49.920 --> 01:02:52.240 compile every version of your spec? 01:02:52.240 --> 01:02:56.799 >> No, like I said, um I use I I had a very 01:02:56.799 --> 01:02:58.799 good experience last week before I got 01:02:58.799 --> 01:03:01.599 sick. Basically, we have a lot of 01:03:01.599 --> 01:03:03.760 terraform states that have dependencies 01:03:03.760 --> 01:03:06.960 on each other and we were merging some 01:03:06.960 --> 01:03:08.720 states like refactoring, moving states 01:03:08.720 --> 01:03:11.280 to another place, moving resources. And 01:03:11.280 --> 01:03:14.000 as we move resources, it kind of 01:03:14.000 --> 01:03:16.240 exemplified that there was like a cycle 01:03:16.240 --> 01:03:18.079 hidden because like we have one state 01:03:18.079 --> 01:03:19.920 here, one state here, one state here. 01:03:19.920 --> 01:03:21.920 And when we move things from this state 01:03:21.920 --> 01:03:23.920 which is going to go away down to this 01:03:23.920 --> 01:03:25.920 state, this state depended on that and 01:03:25.920 --> 01:03:27.680 that state depended on this. And when we 01:03:27.680 --> 01:03:29.839 moved it, I mean the state depended on 01:03:29.839 --> 01:03:31.839 each other that we created a cycle by 01:03:31.839 --> 01:03:33.920 moving resources. So we had to like 01:03:33.920 --> 01:03:35.440 identify like where do we need to move 01:03:35.440 --> 01:03:37.680 them and so on. So I told him we 01:03:37.680 --> 01:03:40.079 actually have a whole on disk we have a 01:03:40.079 --> 01:03:41.920 whole map of what depends on what we 01:03:41.920 --> 01:03:43.760 just need to get like create a diagram. 01:03:43.760 --> 01:03:47.599 So he my my junior he went and vcoded 01:03:47.599 --> 01:03:51.280 with co-pilot and um it took him 2 hours 01:03:51.280 --> 01:03:53.520 and he says here's this a CLI here's a 01:03:53.520 --> 01:03:55.599 dot graph representation of all of our 01:03:55.599 --> 01:03:56.960 terapform code and dependencies between 01:03:56.960 --> 01:03:58.400 them like what the grid you know 01:03:58.400 --> 01:03:59.920 supposed to give me but we're not 01:03:59.920 --> 01:04:02.079 migrating to the grid so and then I said 01:04:02.079 --> 01:04:05.760 okay but it's dot format I cannot relay 01:04:05.760 --> 01:04:08.880 out the the graph it's very confusing it 01:04:08.880 --> 01:04:10.799 doesn't show exactly like it shows 01:04:10.799 --> 01:04:12.480 there's a dependency between them but it 01:04:12.480 --> 01:04:14.240 doesn't really show what are the outputs 01:04:14.240 --> 01:04:16.400 between the states that we depend on. So 01:04:16.400 --> 01:04:18.720 I I then introduced uh specdriven 01:04:18.720 --> 01:04:21.599 development to him and I say we take the 01:04:21.599 --> 01:04:24.000 idea of what we want with the ability to 01:04:24.000 --> 01:04:26.000 find a graph and so on. So that's a user 01:04:26.000 --> 01:04:28.559 story as a as a platform engineer I need 01:04:28.559 --> 01:04:31.839 to identify a graph and it in cycles and 01:04:31.839 --> 01:04:34.559 then we we write down and and basically 01:04:34.559 --> 01:04:36.559 I reimplemented it using specdriven 01:04:36.559 --> 01:04:38.799 development with him and he can 01:04:38.799 --> 01:04:41.039 understand hey yeah that's great now all 01:04:41.039 --> 01:04:43.599 of these things that were assumptions 01:04:43.599 --> 01:04:45.599 from the model and that we did not align 01:04:45.599 --> 01:04:47.359 on because he spent two hours to build 01:04:47.359 --> 01:04:49.039 it and I said look there's a problem 01:04:49.039 --> 01:04:50.799 because 01:04:50.799 --> 01:04:52.720 >> I had a whole bunch of comments and he 01:04:52.720 --> 01:04:54.319 we never discussed it because within one 01:04:54.319 --> 01:04:55.520 hour he had a solution. 01:04:55.520 --> 01:04:58.640 So specit allows me to build that first 01:04:58.640 --> 01:05:01.440 version of the CLI but the vcoded 01:05:01.440 --> 01:05:04.160 version that he provided also allowed us 01:05:04.160 --> 01:05:06.160 to discuss where the problems were with 01:05:06.160 --> 01:05:07.119 that vibe coded version. 01:05:07.119 --> 01:05:10.079 >> Okay so off to each other but 01:05:10.079 --> 01:05:12.319 >> and then you keep iterating because then 01:05:12.319 --> 01:05:14.000 I got the first version and then we get 01:05:14.000 --> 01:05:15.920 it in our hands and then we see oh 01:05:15.920 --> 01:05:18.640 actually I used Excaliraw instead of 01:05:18.640 --> 01:05:20.720 graph so I thought exceal is going to be 01:05:20.720 --> 01:05:22.480 much better because then I can just drag 01:05:22.480 --> 01:05:24.880 them and it stores the xy position. So 01:05:24.880 --> 01:05:26.960 when it regenerates it just overwrites 01:05:26.960 --> 01:05:28.880 the nodes, adds new nodes and doesn't 01:05:28.880 --> 01:05:31.599 rewrite the graph. But then it's like 01:05:31.599 --> 01:05:35.280 400 different arrows and 40 states and 01:05:35.280 --> 01:05:37.520 when I try to change it, it crashes the 01:05:37.520 --> 01:05:39.440 machine. So then excalate draw is too 01:05:39.440 --> 01:05:41.520 slow. So we're looking at a JavaScript 01:05:41.520 --> 01:05:43.280 library for visualizing and then 01:05:43.280 --> 01:05:45.359 modifying the graph. So when I drag it, 01:05:45.359 --> 01:05:47.520 it can drag all the arrows with it. So 01:05:47.520 --> 01:05:49.920 we iterate, right? That's a second user 01:05:49.920 --> 01:05:51.039 story. 01:05:51.039 --> 01:05:53.200 I'm just I'm just mapping this into my 01:05:53.200 --> 01:05:54.640 own experience because like in thought 01:05:54.640 --> 01:05:57.280 works when we when we uh first engage 01:05:57.280 --> 01:05:59.200 with the client we have this like 01:05:59.200 --> 01:06:01.599 discovery phase and what I'm 01:06:01.599 --> 01:06:03.760 interpreting is that you use tools like 01:06:03.760 --> 01:06:06.319 spec kits or spec leisure rather to 01:06:06.319 --> 01:06:09.440 basically capture 01:06:09.440 --> 01:06:11.599 or or better define what you're going to 01:06:11.599 --> 01:06:13.920 do in this discovery phase. But then 01:06:13.920 --> 01:06:15.599 after the discovery phase when you're 01:06:15.599 --> 01:06:17.760 actually going down to implement and and 01:06:17.760 --> 01:06:20.079 things like this, you your your your 01:06:20.079 --> 01:06:23.520 spec um well hopefully will but still be 01:06:23.520 --> 01:06:25.359 a good reference point but you're not 01:06:25.359 --> 01:06:27.520 necessarily developing from the spec. 01:06:27.520 --> 01:06:28.480 Did that make sense? 01:06:28.480 --> 01:06:29.920 >> Well, you said something about using 01:06:29.920 --> 01:06:32.319 spec to explore but you use vibe to 01:06:32.319 --> 01:06:35.440 explore, right? You you I understand you 01:06:35.440 --> 01:06:37.760 what you're saying vibe there's there's 01:06:37.760 --> 01:06:39.839 different f potentially different phases 01:06:39.839 --> 01:06:42.079 like you have a vibe phase and then you 01:06:42.079 --> 01:06:43.680 have maybe 01:06:43.680 --> 01:06:45.119 um a more uh 01:06:45.119 --> 01:06:47.200 >> take your learnings you you maybe 01:06:47.200 --> 01:06:49.920 highlight a bit more disciplined uh 01:06:49.920 --> 01:06:53.119 driven um development phase and then you 01:06:53.119 --> 01:06:55.520 have I don't know the the actual thing 01:06:55.520 --> 01:06:58.000 that's developed phase and then you you 01:06:58.000 --> 01:07:00.000 evolve that thing that's been developed 01:07:00.000 --> 01:07:02.000 from the spec or something. Yeah, 01:07:02.000 --> 01:07:04.720 there's I I think of it like that now. I 01:07:04.720 --> 01:07:05.920 mean that's the way I'm thinking of it 01:07:05.920 --> 01:07:07.760 right now. I I think people are getting 01:07:07.760 --> 01:07:10.720 hung up on methods. I mean whatever 01:07:10.720 --> 01:07:13.599 works. Specit worked really well for me 01:07:13.599 --> 01:07:16.400 when I was coming from chat GPT saying 01:07:16.400 --> 01:07:18.240 like I want to build this system. it has 01:07:18.240 --> 01:07:20.319 these features and then Chad GPT gives 01:07:20.319 --> 01:07:22.319 me like a PRD product requirement 01:07:22.319 --> 01:07:24.160 document and then I'm like I agree with 01:07:24.160 --> 01:07:26.079 most of it but this one part we need to 01:07:26.079 --> 01:07:28.160 clarify and then it gives me completely 01:07:28.160 --> 01:07:29.839 different PRD where the original stuff 01:07:29.839 --> 01:07:31.839 is all gone and then it focuses on one 01:07:31.839 --> 01:07:33.039 area and I'm like getting super 01:07:33.039 --> 01:07:35.760 frustrated and I'm like I need a better 01:07:35.760 --> 01:07:38.240 way to to evolve this and then I use 01:07:38.240 --> 01:07:40.480 spec it and it just did perfectly what I 01:07:40.480 --> 01:07:41.520 wanted. 01:07:41.520 --> 01:07:42.000 >> Yeah, 01:07:42.000 --> 01:07:43.599 >> I guess of course 01:07:43.599 --> 01:07:45.839 >> I fell victim to this. I guess what what 01:07:45.839 --> 01:07:46.880 I'm thinking is that you 01:07:46.880 --> 01:07:48.960 >> and that was a year ago. So it changes 01:07:48.960 --> 01:07:51.520 like now you use GSD now you use power 01:07:51.520 --> 01:07:53.599 skills and now you do an introduc to 01:07:53.599 --> 01:07:55.760 it's just a tool and the the idea of the 01:07:55.760 --> 01:07:57.839 tool is to help you engineer the 01:07:57.839 --> 01:07:58.799 software. 01:07:58.799 --> 01:08:00.319 >> Yeah. So a lot of people are like 01:08:00.319 --> 01:08:03.119 walking back from specit and I mean like 01:08:03.119 --> 01:08:05.280 obviously specit is just one like open 01:08:05.280 --> 01:08:07.039 spec has another way of managing the 01:08:07.039 --> 01:08:09.200 specs and they have underactive chains 01:08:09.200 --> 01:08:10.960 versus what has been archived. 01:08:10.960 --> 01:08:12.400 Apparently they have like a really good 01:08:12.400 --> 01:08:15.280 way of identifying what has already 01:08:15.280 --> 01:08:16.799 landed and I think that's what I'm 01:08:16.799 --> 01:08:18.319 missing in specket. So open spec 01:08:18.319 --> 01:08:20.799 apparently does that. So I I mean I had 01:08:20.799 --> 01:08:22.480 people reach out to me like they said 01:08:22.480 --> 01:08:24.480 they want to use speckit for their 01:08:24.480 --> 01:08:26.480 organization and I was like I don't 01:08:26.480 --> 01:08:28.560 think we have enough capacity right now 01:08:28.560 --> 01:08:31.679 me and my friends to to like support. So 01:08:31.679 --> 01:08:34.560 there is this open spec spaces thing. 01:08:34.560 --> 01:08:36.560 There's like a wait list. Go and try 01:08:36.560 --> 01:08:38.400 that. And then they go back to me and 01:08:38.400 --> 01:08:39.920 they say yeah that's great. It works. 01:08:39.920 --> 01:08:41.839 And I'm like, I wish I I could 01:08:41.839 --> 01:08:43.040 have involved. 01:08:43.040 --> 01:08:45.279 >> It would be good to run through 01:08:45.279 --> 01:08:47.040 something with you when you have the 01:08:47.040 --> 01:08:49.440 time, Vincent, just so that I get more 01:08:49.440 --> 01:08:52.080 of your uh experience because I I guess 01:08:52.080 --> 01:08:54.000 I'm I'm going to be starting on a new 01:08:54.000 --> 01:08:56.640 project. Uh I think using something like 01:08:56.640 --> 01:08:59.839 Spec Ledger would help Yeah. discover to 01:08:59.839 --> 01:09:02.400 capture the the client's requirements 01:09:02.400 --> 01:09:04.239 for me to deliver better work. 01:09:04.239 --> 01:09:06.239 >> Yeah. I mean like the most important 01:09:06.239 --> 01:09:08.080 thing about specledger is and when I 01:09:08.080 --> 01:09:10.159 talk to my friends they're usually 01:09:10.159 --> 01:09:11.839 working on their own and they're like 01:09:11.839 --> 01:09:13.440 what does the does spec ledger solve my 01:09:13.440 --> 01:09:15.199 problem in getting the model to keep 01:09:15.199 --> 01:09:17.199 working and it's it's done and I'm like 01:09:17.199 --> 01:09:20.000 no the whole point of specledger for me 01:09:20.000 --> 01:09:22.159 was so that I could discuss the user 01:09:22.159 --> 01:09:24.719 story with other people before we go and 01:09:24.719 --> 01:09:27.279 implement it. It's not an agent runner. 01:09:27.279 --> 01:09:30.000 It's not it's not like it's just a 01:09:30.000 --> 01:09:33.040 formal way. It does I addit stuff that I 01:09:33.040 --> 01:09:35.359 want that I use which is like after you 01:09:35.359 --> 01:09:38.480 implement I click SL checkpoint which 01:09:38.480 --> 01:09:40.880 will then go adversarial review. It's 01:09:40.880 --> 01:09:42.319 not a celebration of what we have 01:09:42.319 --> 01:09:44.640 accomplished. It's more of like have we 01:09:44.640 --> 01:09:45.920 met what we needed to do 01:09:45.920 --> 01:09:47.679 >> I guess. Yeah it's it's a bit of a 01:09:47.679 --> 01:09:49.279 misnomer like spectrum of development 01:09:49.279 --> 01:09:51.679 because I I guess I mean I assume that 01:09:51.679 --> 01:09:54.320 it's it's basically a way of driving the 01:09:54.320 --> 01:09:56.000 agents to implement it but it's not 01:09:56.000 --> 01:09:58.000 really. It's it's more of a of a 01:09:58.000 --> 01:09:59.760 structure. It's it's more of a planning 01:09:59.760 --> 01:10:00.480 tool, right? 01:10:00.480 --> 01:10:01.840 >> It's a planning tool. And I think the 01:10:01.840 --> 01:10:04.400 biggest critic that you can get against 01:10:04.400 --> 01:10:06.800 it is that it's too waterfall like if 01:10:06.800 --> 01:10:08.400 you don't use it the right. I mean, if 01:10:08.400 --> 01:10:10.388 you get started like the way I started, 01:10:10.388 --> 01:10:12.080 [snorts] it's very easy to create this 01:10:12.080 --> 01:10:14.239 beautiful big plan and then hit all 01:10:14.239 --> 01:10:16.960 kinds of unknowns along the way and then 01:10:16.960 --> 01:10:18.719 realize that you have to change things. 01:10:18.719 --> 01:10:21.679 Whereas people with models are way more 01:10:21.679 --> 01:10:23.440 like like you said now they are like 01:10:23.440 --> 01:10:25.280 models are very good at creating a plan 01:10:25.280 --> 01:10:27.120 and if you are not getting what you want 01:10:27.120 --> 01:10:28.960 rather than to tell it to change it it's 01:10:28.960 --> 01:10:32.159 hard you go back and you restart from a 01:10:32.159 --> 01:10:34.400 known good state and start again with 01:10:34.400 --> 01:10:35.920 the knowledge that you have acquired 01:10:35.920 --> 01:10:37.600 that that didn't work and we need to 01:10:37.600 --> 01:10:39.440 watch out for that. I can 01:10:39.440 --> 01:10:42.080 >> I mean I'm I'm a big agile proponent but 01:10:42.080 --> 01:10:43.920 like if you start a pro I mean 01:10:43.920 --> 01:10:45.679 especially a green field project if you 01:10:45.679 --> 01:10:48.880 waterfall the initial design uh and 01:10:48.880 --> 01:10:51.199 since it's so quick to do it really if 01:10:51.199 --> 01:10:52.400 the stakeholders all there 01:10:52.400 --> 01:10:54.480 >> openspec calls it mini waterfall and I 01:10:54.480 --> 01:10:56.320 think it's the right way to to to think 01:10:56.320 --> 01:10:56.960 about it 01:10:56.960 --> 01:10:58.880 >> it's it's probably a good idea to to 01:10:58.880 --> 01:11:01.440 start with a with a and and and yeah you 01:11:01.440 --> 01:11:03.360 could throw away a few of these mini 01:11:03.360 --> 01:11:05.520 waterfalls and then and then when when 01:11:05.520 --> 01:11:07.360 the tie hits the road and there's some 01:11:07.360 --> 01:11:09.840 sort of traction I think it won't be 01:11:09.840 --> 01:11:11.600 waterfall anymore. It would be agile 01:11:11.600 --> 01:11:12.159 surely 01:11:12.159 --> 01:11:14.800 >> because DHA's also highlights a lot of 01:11:14.800 --> 01:11:16.480 the lessons that we've learned is based 01:11:16.480 --> 01:11:18.239 on the way that humans work with code 01:11:18.239 --> 01:11:20.320 and collaborate on code and like the 01:11:20.320 --> 01:11:22.400 problem with waterfall was that it took 01:11:22.400 --> 01:11:24.000 two months to get to solution. And I 01:11:24.000 --> 01:11:25.840 don't remember but I remember I don't 01:11:25.840 --> 01:11:27.520 remember who but I remember seeing 01:11:27.520 --> 01:11:30.080 somebody talk about that but like what 01:11:30.080 --> 01:11:32.560 DHA is saying is like he's it's a single 01:11:32.560 --> 01:11:34.560 person exploring and coming up with the 01:11:34.560 --> 01:11:36.800 design before handing it off and then 01:11:36.800 --> 01:11:38.960 letting the team finune it right but 01:11:38.960 --> 01:11:41.760 like DHS example isn't really realistic 01:11:41.760 --> 01:11:43.199 because like if you if you go into like 01:11:43.199 --> 01:11:45.760 an enterprise I'm just thinking of a a 01:11:45.760 --> 01:11:48.640 big enterprise with I don't know um 01:11:48.640 --> 01:11:51.920 entrenched domains you need a lot of 01:11:51.920 --> 01:11:54.239 like uh uh what do you call when the 01:11:54.239 --> 01:11:55.600 information is sort of hidden, you know, 01:11:55.600 --> 01:11:57.360 like there's lots of little you need to 01:11:57.360 --> 01:11:59.120 talk with different product owners and 01:11:59.120 --> 01:12:01.360 different teams in order to build 01:12:01.360 --> 01:12:03.520 something across it. You you you just 01:12:03.520 --> 01:12:06.880 can't do it in that DHH fashion just 01:12:06.880 --> 01:12:08.719 because just because you just don't know 01:12:08.719 --> 01:12:10.880 the requirements and you you don't know 01:12:10.880 --> 01:12:12.480 the system's very complex or something 01:12:12.480 --> 01:12:13.120 like that. 01:12:13.120 --> 01:12:13.920 >> Yeah. 01:12:13.920 --> 01:12:14.400 >> Okay. 01:12:14.400 --> 01:12:16.080 >> I think I think we're at the end of our 01:12:16.080 --> 01:12:19.280 tether, aren't we? At least I am. 01:12:19.280 --> 01:12:21.040 >> Um yeah, I think that was thanks for the 01:12:21.040 --> 01:12:23.360 good conversation as usual, Vincent. And 01:12:23.360 --> 01:12:25.440 I hope anyone listening enjoyed that 01:12:25.440 --> 01:12:28.719 too. If not, please comment below. Or if 01:12:28.719 --> 01:12:30.880 you did enjoy it, comment below. 01:12:30.880 --> 01:12:32.560 Whatever. Just say hi. 01:12:32.560 --> 01:12:34.480 >> I mean, what I like about once it's 01:12:34.480 --> 01:12:35.920 posted on YouTube and you run the 01:12:35.920 --> 01:12:38.080 summarize on it, you you actually get 01:12:38.080 --> 01:12:40.000 like a full breakdown of 01:12:40.000 --> 01:12:41.360 >> Oh, yeah. It works quite well. 01:12:41.360 --> 01:12:43.040 >> Some of the although that sometimes when 01:12:43.040 --> 01:12:45.199 I read the summarize, I do see it like 01:12:45.199 --> 01:12:47.199 summarize in a certain way where like 01:12:47.199 --> 01:12:48.960 that's not really what we were saying. 01:12:48.960 --> 01:12:50.480 There was definitely there's some 01:12:50.480 --> 01:12:51.760 mistakes in there I saw. 01:12:51.760 --> 01:12:54.400 >> Yeah. I I I just like the fact that we 01:12:54.400 --> 01:12:56.719 have a record of what we talked about 01:12:56.719 --> 01:12:58.480 because 01:12:58.480 --> 01:13:01.440 because I don't know if it's my brain or 01:13:01.440 --> 01:13:03.199 whatever. It's like sometimes I have 01:13:03.199 --> 01:13:05.040 interesting conversations quite often 01:13:05.040 --> 01:13:07.760 with you and I and yeah like I forget 01:13:07.760 --> 01:13:10.080 what we talked about and and now we have 01:13:10.080 --> 01:13:12.719 a good I can refer to it like oh yeah we 01:13:12.719 --> 01:13:15.040 talked about this. What did we say? 01:13:15.040 --> 01:13:17.440 Yeah. Yeah. We said about this and uh we 01:13:17.440 --> 01:13:19.920 talked about that. Yeah. just just 01:13:19.920 --> 01:13:22.080 having a record I found quite useful. 01:13:22.080 --> 01:13:23.679 Yeah, these summarizes are pretty 01:13:23.679 --> 01:13:24.320 impressive 01:13:24.320 --> 01:13:25.840 >> and the fact that Google spits it right 01:13:25.840 --> 01:13:28.320 back at me when I search something. 01:13:28.320 --> 01:13:31.360 Sometimes it show it sources this 01:13:31.360 --> 01:13:34.239 >> um one time I was searching for 01:13:34.239 --> 01:13:37.199 >> something related to CDKTF or ter CDK 01:13:37.199 --> 01:13:37.760 terrain. 01:13:37.760 --> 01:13:39.520 >> It came it came back from my blog. 01:13:39.520 --> 01:13:42.159 >> Yeah. 01:13:42.159 --> 01:13:45.679 >> What should we call this one? Um 01:13:45.679 --> 01:13:48.800 I I felt I I think in hinds I I really 01:13:48.800 --> 01:13:50.480 thought what you what we just discussing 01:13:50.480 --> 01:13:52.960 about expect driven development was 01:13:52.960 --> 01:13:55.840 quite interesting but I would hate for 01:13:55.840 --> 01:13:58.000 anyone to read the title and then have 01:13:58.000 --> 01:14:00.719 to scan in have to listen to us for for 01:14:00.719 --> 01:14:02.400 an hour and 10 minutes until we start 01:14:02.400 --> 01:14:05.360 talking about it. I mean we talked about 01:14:05.360 --> 01:14:08.239 the security implications and you know 01:14:08.239 --> 01:14:11.040 noise getting signal from noise and AI 01:14:11.040 --> 01:14:14.000 agents hopefully helping with that. We 01:14:14.000 --> 01:14:16.000 also talked about a left bat incident in 01:14:16.000 --> 01:14:20.800 Rust. Um and you talked about some AI 01:14:20.800 --> 01:14:23.679 engineer and then I think the last bit 01:14:23.679 --> 01:14:26.560 was about how 01:14:26.560 --> 01:14:28.960 um these examples from entropic use XML 01:14:28.960 --> 01:14:30.719 and the way that they organize their 01:14:30.719 --> 01:14:32.800 specs are very different from the way 01:14:32.800 --> 01:14:36.239 specit does and um I think it's 01:14:36.239 --> 01:14:38.400 interesting. I definitely think I want 01:14:38.400 --> 01:14:43.600 to try open spec uh power skills and um 01:14:43.600 --> 01:14:45.679 or is it superpowers? I think it's super 01:14:45.679 --> 01:14:47.360 I don't know. It's like this cloud 01:14:47.360 --> 01:14:49.600 superpowers thing apparently has a bunch 01:14:49.600 --> 01:14:52.400 of skills that make it really good at um 01:14:52.400 --> 01:14:54.320 helping you organize and build. And then 01:14:54.320 --> 01:14:55.760 there's get done that everyone 01:14:55.760 --> 01:14:57.920 talks about. But it's so funny because 01:14:57.920 --> 01:14:59.440 my friend is like, "Oh, you know, 01:14:59.440 --> 01:15:03.040 Speckit is so um not good and I will use 01:15:03.040 --> 01:15:05.440 uh GSD." And then I go back to him a 01:15:05.440 --> 01:15:07.120 week later. I said, "Well, look at this 01:15:07.120 --> 01:15:09.360 Reddit comment. GSD is so last month. 01:15:09.360 --> 01:15:12.560 You should be using superpowers." And 01:15:12.560 --> 01:15:14.080 >> apparently 01:15:14.080 --> 01:15:17.040 >> I actually need to set myself a goal 01:15:17.040 --> 01:15:19.920 like I need to build my own little SAS 01:15:19.920 --> 01:15:22.880 application some some way to have a side 01:15:22.880 --> 01:15:25.840 income or something like this you know 01:15:25.840 --> 01:15:26.640 have a product 01:15:26.640 --> 01:15:28.640 >> a SAS for who? For yourself. 01:15:28.640 --> 01:15:31.040 >> Well I build one of my side projects 01:15:31.040 --> 01:15:33.440 which I never really launched was to 01:15:33.440 --> 01:15:35.760 have like a YouTube backup service so 01:15:35.760 --> 01:15:37.440 that you like you pay 10 bucks a month 01:15:37.440 --> 01:15:39.760 and I back your YouTube up. But there 01:15:39.760 --> 01:15:41.360 are like three or four other ones like 01:15:41.360 --> 01:15:43.600 that already. Mine would be better. Of 01:15:43.600 --> 01:15:44.560 course. 01:15:44.560 --> 01:15:46.960 >> Of course. You know what? I think we 01:15:46.960 --> 01:15:48.239 talked about this last time, which is 01:15:48.239 --> 01:15:50.640 how do you check if you hire someone if 01:15:50.640 --> 01:15:53.679 they are like good working with AI, 01:15:53.679 --> 01:15:55.360 right? We talked about this last time as 01:15:55.360 --> 01:15:55.840 well. 01:15:55.840 --> 01:15:56.800 >> Build a product. 01:15:56.800 --> 01:15:58.480 >> Yeah. But you know how flight 01:15:58.480 --> 01:16:00.960 controllers they they check if you if 01:16:00.960 --> 01:16:03.360 you can do multitasking like you are cuz 01:16:03.360 --> 01:16:05.360 when you're in a in an ATC tower at the 01:16:05.360 --> 01:16:07.199 airport and flights are coming in you 01:16:07.199 --> 01:16:08.560 have to like keep track of five 01:16:08.560 --> 01:16:09.760 different flights. You have those games 01:16:09.760 --> 01:16:11.679 on the iPhone as well where you need to 01:16:11.679 --> 01:16:13.360 land the planes. Uh it's probably not 01:16:13.360 --> 01:16:14.480 the same thing but 01:16:14.480 --> 01:16:16.960 >> Oh yeah. Oh yeah. I love that game. It 01:16:16.960 --> 01:16:20.400 was it was this guy from Mass Driver who 01:16:20.400 --> 01:16:22.480 posted on LinkedIn talking about how he 01:16:22.480 --> 01:16:24.960 has to work with cloud and go into 01:16:24.960 --> 01:16:26.560 different sessions and then go and pick 01:16:26.560 --> 01:16:28.480 up again the context and then figure out 01:16:28.480 --> 01:16:30.560 what he wanted to do next and then go 01:16:30.560 --> 01:16:32.480 switch on the next one. And then his 01:16:32.480 --> 01:16:34.000 father who used to be a air traffic 01:16:34.000 --> 01:16:36.239 controller told him like it's what I did 01:16:36.239 --> 01:16:38.000 every day you know go back to this 01:16:38.000 --> 01:16:39.600 flight figure out where they are if 01:16:39.600 --> 01:16:41.040 they're still on course for the landing 01:16:41.040 --> 01:16:43.360 on this trip and constantly switching 01:16:43.360 --> 01:16:45.840 between. So I was like what are the 01:16:45.840 --> 01:16:48.480 interview processes for ATC you know 01:16:48.480 --> 01:16:49.840 roles like jobs 01:16:49.840 --> 01:16:52.239 >> when you're focusing on on a product I 01:16:52.239 --> 01:16:53.440 don't think you need to switch between 01:16:53.440 --> 01:16:55.520 agents that much think I think when you 01:16:55.520 --> 01:16:58.000 do your day-to-day sort of job then 01:16:58.000 --> 01:16:59.440 you're switching between agents 01:16:59.440 --> 01:17:01.520 >> but I think you are either working on 01:17:01.520 --> 01:17:03.600 multiple features on the same product at 01:17:03.600 --> 01:17:05.280 the same time because I've definitely 01:17:05.280 --> 01:17:08.159 had like um with spec ledger where I was 01:17:08.159 --> 01:17:10.719 working on a feature for the spec ledger 01:17:10.719 --> 01:17:12.800 and then I when I was using spec ledger 01:17:12.800 --> 01:17:14.960 it errored or have a bug or I can see 01:17:14.960 --> 01:17:17.040 the model is not optim optimally using 01:17:17.040 --> 01:17:18.960 this and it got confused. So then I tell 01:17:18.960 --> 01:17:20.880 it what in that session I say create an 01:17:20.880 --> 01:17:23.040 issue about this this problem that we 01:17:23.040 --> 01:17:25.040 just faced and why did it happen and 01:17:25.040 --> 01:17:26.719 then it goes and creates a GitHub issue. 01:17:26.719 --> 01:17:28.640 Then I tell another Claude Code session 01:17:28.640 --> 01:17:30.080 create a work tree and work on that 01:17:30.080 --> 01:17:32.239 GitHub issue and then in the meantime I 01:17:32.239 --> 01:17:33.679 go back to my main session where I'm 01:17:33.679 --> 01:17:35.520 working on a new feature and then if I 01:17:35.520 --> 01:17:37.360 have another issue I create another one. 01:17:37.360 --> 01:17:39.280 So then I have like three work trees 01:17:39.280 --> 01:17:40.960 working on three different things. I 01:17:40.960 --> 01:17:41.920 think you're doing too much. But you 01:17:41.920 --> 01:17:42.080 can't 01:17:42.080 --> 01:17:43.600 >> No, that was like two two weeks ago 01:17:43.600 --> 01:17:45.760 before I got sick. 01:17:45.760 --> 01:17:47.199 >> Yeah, like you can't work on three 01:17:47.199 --> 01:17:48.640 features in parallel. I think that's 01:17:48.640 --> 01:17:49.760 kind of madness. 01:17:49.760 --> 01:17:52.400 >> I mean, if I if I'm doing my product, if 01:17:52.400 --> 01:17:54.159 I'm doing my YouTube backup thing, I'm 01:17:54.159 --> 01:17:55.440 not going to work on three features in 01:17:55.440 --> 01:17:57.280 parallel. That'd be crazy. And I think 01:17:57.280 --> 01:17:58.960 that's the problem between people that 01:17:58.960 --> 01:18:02.239 are AI active and people that are not. I 01:18:02.239 --> 01:18:04.000 think that's what we need to be able to 01:18:04.000 --> 01:18:06.080 do. So I think we need to be able to and 01:18:06.080 --> 01:18:07.280 that's what a lot of people say like 01:18:07.280 --> 01:18:09.120 Simon as well on this on this um 01:18:09.120 --> 01:18:11.600 interview he says like I'm working on so 01:18:11.600 --> 01:18:13.600 many things at the same time because I 01:18:13.600 --> 01:18:15.040 just come up with the idea give it to 01:18:15.040 --> 01:18:15.679 the agent 01:18:15.679 --> 01:18:18.320 >> so many packages he's got so many 01:18:18.320 --> 01:18:20.480 >> yeah but not just him people that are 01:18:20.480 --> 01:18:22.640 successful here they say that this is 01:18:22.640 --> 01:18:24.560 what we are required to do now we are 01:18:24.560 --> 01:18:27.120 able to switch context contently and 01:18:27.120 --> 01:18:29.280 then by lunchtime we are exhausted we 01:18:29.280 --> 01:18:29.920 are depleted 01:18:29.920 --> 01:18:33.120 >> yeah yeah true like I I honestly well I 01:18:33.120 --> 01:18:35.040 push back on that in sense of like if if 01:18:35.040 --> 01:18:37.440 I'm building my own product then I would 01:18:37.440 --> 01:18:40.800 be more dh I think said it well you know 01:18:40.800 --> 01:18:42.719 you're you're focused on that particular 01:18:42.719 --> 01:18:44.880 thing and using AI to polish the hell 01:18:44.880 --> 01:18:48.159 out of it you're not like this whole um 01:18:48.159 --> 01:18:50.000 but when you work on a project Kai there 01:18:50.000 --> 01:18:51.600 will always be things that you want to 01:18:51.600 --> 01:18:53.040 lend right now a feature that you're 01:18:53.040 --> 01:18:54.719 working on and things that you postpone 01:18:54.719 --> 01:18:57.040 and say later we will polish this and 01:18:57.040 --> 01:18:59.280 with AI you can say later I will polish 01:18:59.280 --> 01:19:00.880 this but at the same time go in a work 01:19:00.880 --> 01:19:01.920 tree and go and do it 01:19:01.920 --> 01:19:03.520 >> when I when I personally develop 01:19:03.520 --> 01:19:05.520 software. I I'm very much in the Unix 01:19:05.520 --> 01:19:07.440 philosophy like do one thing and do one 01:19:07.440 --> 01:19:09.920 thing well. I always want to step away 01:19:09.920 --> 01:19:11.440 from like I don't know a G. 01:19:11.440 --> 01:19:14.159 >> What I'm saying is that in the future 01:19:14.159 --> 01:19:16.719 when AI is doing the work, you're not 01:19:16.719 --> 01:19:18.560 going to sit there and twiddle your 01:19:18.560 --> 01:19:20.880 thumb well while it's doing the work and 01:19:20.880 --> 01:19:21.840 wait for it to come back. 01:19:21.840 --> 01:19:24.000 >> I think it's going to implement quickly. 01:19:24.000 --> 01:19:25.920 The important the important thing is to 01:19:25.920 --> 01:19:27.040 >> it takes time. 01:19:27.040 --> 01:19:28.400 >> It always takes time. 01:19:28.400 --> 01:19:29.840 >> Yeah. The important thing is to have 01:19:29.840 --> 01:19:32.159 that that that human taste, that human 01:19:32.159 --> 01:19:34.640 like how is this going to work? How is 01:19:34.640 --> 01:19:36.239 this going to feel? 01:19:36.239 --> 01:19:37.360 >> I think 01:19:37.360 --> 01:19:40.960 >> I'm at only 40% of my max usage mostly 01:19:40.960 --> 01:19:43.760 at every week. 40%. And I am running 01:19:43.760 --> 01:19:45.760 >> You got to stop You got to stop FOMOing 01:19:45.760 --> 01:19:46.400 about that. 01:19:46.400 --> 01:19:48.080 >> I'm not fomoing. I'm just saying that 01:19:48.080 --> 01:19:50.725 I'm I'm running, 01:19:50.725 --> 01:19:51.360 [snorts] 01:19:51.360 --> 01:19:53.440 you know, maybe four or five projects at 01:19:53.440 --> 01:19:55.360 the same time. So, I have like two 01:19:55.360 --> 01:19:57.520 laptops and I have different cloud s 01:19:57.520 --> 01:20:00.080 sessions running and then I just go and 01:20:00.080 --> 01:20:02.000 check what it's blocked on and unblock 01:20:02.000 --> 01:20:04.480 it and and I'm only at 40%. And there's 01:20:04.480 --> 01:20:06.320 people on Reddit complaining that 01:20:06.320 --> 01:20:09.040 they're hitting the budget the the 5 01:20:09.040 --> 01:20:10.960 hour session budget on a max plan or on 01:20:10.960 --> 01:20:12.000 a 01:20:12.000 --> 01:20:14.719 >> 5x like they have like 10x max plan 01:20:14.719 --> 01:20:16.159 which is double of what I have and 01:20:16.159 --> 01:20:17.520 they're running out in 5 hours. 01:20:17.520 --> 01:20:19.920 >> I I I admit I mean you saw my command 01:20:19.920 --> 01:20:21.360 control. I mean I think I had 20 01:20:21.360 --> 01:20:23.360 sessions yesterday. Not to say that I'm 01:20:23.360 --> 01:20:24.480 going back and forth all the time 01:20:24.480 --> 01:20:26.800 between 20 though at the moment I I need 01:20:26.800 --> 01:20:29.199 to basically build out a product. I 01:20:29.199 --> 01:20:30.640 think when you build a product and you 01:20:30.640 --> 01:20:33.120 and you do more focused work, you're 01:20:33.120 --> 01:20:35.040 going to be doing less of that is just 01:20:35.040 --> 01:20:35.679 my thought. 01:20:35.679 --> 01:20:37.440 >> Okay. But I think you will notice that 01:20:37.440 --> 01:20:39.199 you will be waiting a lot and decide 01:20:39.199 --> 01:20:41.040 that you can start a little session on 01:20:41.040 --> 01:20:42.719 the side and end up being with four 01:20:42.719 --> 01:20:45.199 sessions once you do it. I I would 01:20:45.199 --> 01:20:47.040 hopefully have the I hope to go for a 01:20:47.040 --> 01:20:48.719 cup of tea and a biscuit in that time 01:20:48.719 --> 01:20:52.880 >> like in the AI is a vampire. 01:20:52.880 --> 01:20:53.440 >> Yeah. 01:20:53.440 --> 01:20:54.880 >> Story. 01:20:54.880 --> 01:20:57.440 >> It is a trap to do too many things and I 01:20:57.440 --> 01:20:59.280 don't want to I mean I' I've been very 01:20:59.280 --> 01:21:01.280 close to burned out a few times. I mean 01:21:01.280 --> 01:21:03.280 I mean like God knows how many GitHub 01:21:03.280 --> 01:21:04.960 repos I maintain badly. 01:21:04.960 --> 01:21:06.640 >> But this is a problem that I also had 01:21:06.640 --> 01:21:08.800 after being sick for you know 5 days on 01:21:08.800 --> 01:21:11.600 flu where you have body aches all over. 01:21:11.600 --> 01:21:13.600 You're burning up with fever. You're 01:21:13.600 --> 01:21:15.920 tired, no energy. I was the whole time 01:21:15.920 --> 01:21:18.480 thinking I really don't care about those 01:21:18.480 --> 01:21:20.480 three projects that I was doing if they 01:21:20.480 --> 01:21:22.560 are now continuing or not. 01:21:22.560 --> 01:21:24.560 >> Yeah. 01:21:24.560 --> 01:21:26.480 >> Well, the way I'm thinking is that 01:21:26.480 --> 01:21:29.920 whatever I do um 01:21:29.920 --> 01:21:32.239 especially outside work, it needs to 01:21:32.239 --> 01:21:33.920 perhaps make me more financially 01:21:33.920 --> 01:21:35.360 independent. I think it's going to be 01:21:35.360 --> 01:21:37.120 very hard in the future with AI, you 01:21:37.120 --> 01:21:39.760 know, with like I was talking to an old 01:21:39.760 --> 01:21:42.080 friend who works at the startup that is 01:21:42.080 --> 01:21:45.199 quite a big like app on the app store 01:21:45.199 --> 01:21:48.239 and the founder was freaking out about 01:21:48.239 --> 01:21:50.880 you see like when Entropic announces 01:21:50.880 --> 01:21:54.639 with their new designer and then Figma 01:21:54.639 --> 01:21:57.120 stock drops I don't know how many% by 01:21:57.120 --> 01:21:58.960 the announcement alone of Entropic 01:21:58.960 --> 01:22:00.880 building a design environment. 01:22:00.880 --> 01:22:03.120 >> Yeah. Well, like DHA said it like the 01:22:03.120 --> 01:22:05.040 just putting a spin on on an existing 01:22:05.040 --> 01:22:07.199 idea should be enough to make have a 01:22:07.199 --> 01:22:08.080 business in a way. 01:22:08.080 --> 01:22:10.159 >> By the way, when you earlier said that 01:22:10.159 --> 01:22:12.560 as we make agents more efficient at 01:22:12.560 --> 01:22:14.080 dealing with the processes that are 01:22:14.080 --> 01:22:15.920 working there, we're not changing the 01:22:15.920 --> 01:22:17.679 processes and they're not improving. But 01:22:17.679 --> 01:22:19.199 I think that's one of the problems that 01:22:19.199 --> 01:22:21.679 I or one of the mistakes I made assuming 01:22:21.679 --> 01:22:23.520 that we're going to make processes more 01:22:23.520 --> 01:22:25.520 aligned for agents. I think what every 01:22:25.520 --> 01:22:28.000 AI organ uh company is doing is the 01:22:28.000 --> 01:22:30.560 exact opposite. making agents more fit 01:22:30.560 --> 01:22:32.960 in to every human process. Live in 01:22:32.960 --> 01:22:35.440 Slack, live in Confluence, be part of 01:22:35.440 --> 01:22:37.679 the procedure, not change the procedure 01:22:37.679 --> 01:22:40.239 for agents. This is a big play of OpenAI 01:22:40.239 --> 01:22:41.199 and everything. 01:22:41.199 --> 01:22:42.880 >> I mean, it's being it's being pushed 01:22:42.880 --> 01:22:44.480 from both sides, right? Like I was 01:22:44.480 --> 01:22:46.480 impressed by Google's announcements with 01:22:46.480 --> 01:22:49.280 all the Android CLI stuff cuz like I 01:22:49.280 --> 01:22:50.400 don't know if you've ever developed an 01:22:50.400 --> 01:22:52.159 Android app. It's it's a bit painful, 01:22:52.159 --> 01:22:54.080 but now you can you can you can develop 01:22:54.080 --> 01:22:56.400 an app through their CLI which is 01:22:56.400 --> 01:22:58.960 amazing, right? Yeah, for sure it's 01:22:58.960 --> 01:23:00.719 being pushed on both sides. But I think 01:23:00.719 --> 01:23:02.960 like for example with co-work when they 01:23:02.960 --> 01:23:05.360 are pushing this to to people that are 01:23:05.360 --> 01:23:07.600 not techsavvy, they are not living on 01:23:07.600 --> 01:23:09.520 the terminal, but they want people to 01:23:09.520 --> 01:23:11.920 use AI, they create co-work that works 01:23:11.920 --> 01:23:14.239 inside Excel, that does machine use, 01:23:14.239 --> 01:23:15.840 that just does all of the work like an 01:23:15.840 --> 01:23:18.080 assistant. Why is open cloud famous? 01:23:18.080 --> 01:23:19.679 It's not famous because it's such a 01:23:19.679 --> 01:23:21.520 great like technological thing. It's 01:23:21.520 --> 01:23:22.880 because you just tell it something and 01:23:22.880 --> 01:23:25.440 it goes does everything for you like a 01:23:25.440 --> 01:23:27.040 personal assistant. 01:23:27.040 --> 01:23:28.960 >> It becomes like a human that does the 01:23:28.960 --> 01:23:30.639 work for you. Like my experience when I 01:23:30.639 --> 01:23:32.880 have to go and apply for a passport, it 01:23:32.880 --> 01:23:35.760 wasn't me like figuring out how to give 01:23:35.760 --> 01:23:37.679 Claude Code access to the, you know, all 01:23:37.679 --> 01:23:39.920 of the necessary areas and make it fill 01:23:39.920 --> 01:23:41.920 out the PDF and all that. No, it was all 01:23:41.920 --> 01:23:44.000 in the Claude Code desktop app. It had 01:23:44.000 --> 01:23:45.520 the connector and the skills to manage 01:23:45.520 --> 01:23:47.920 PDFs. it understood how to go and use my 01:23:47.920 --> 01:23:50.000 Gmail and it had the ability to go to 01:23:50.000 --> 01:23:51.760 the Belgian government website and get 01:23:51.760 --> 01:23:53.840 everything and that's where it the power 01:23:53.840 --> 01:23:54.560 came from. 01:23:54.560 --> 01:23:58.080 >> Yeah. just just to uh talk about like my 01:23:58.080 --> 01:24:00.000 my silly like YouTube backup 01:24:00.000 --> 01:24:02.159 application. I guess the way the way to 01:24:02.159 --> 01:24:04.320 validate that idea or the way to affirm 01:24:04.320 --> 01:24:06.080 that idea in the market is that if you 01:24:06.080 --> 01:24:09.760 tell an agent to back up YouTube, will 01:24:09.760 --> 01:24:12.560 it do it manually with the AWS CLI or 01:24:12.560 --> 01:24:15.920 would it somehow find my app and use my 01:24:15.920 --> 01:24:19.040 CLI to back up YouTube? That's probably 01:24:19.040 --> 01:24:20.960 the only way that that software is going 01:24:20.960 --> 01:24:24.159 to like survive in the future. if is is 01:24:24.159 --> 01:24:26.400 is a are agents going to pick it? 01:24:26.400 --> 01:24:27.920 >> I think that's that's that's the biggest 01:24:27.920 --> 01:24:30.400 issue with the way that the LLM agents 01:24:30.400 --> 01:24:32.239 or the way that Google grounds its 01:24:32.239 --> 01:24:35.280 models, it relies so heavily on SEO. So 01:24:35.280 --> 01:24:37.600 if you like have one blog post that 01:24:37.600 --> 01:24:39.280 talks about this or that, 01:24:39.280 --> 01:24:39.760 >> right? 01:24:39.760 --> 01:24:42.000 >> And it just picks that up and now I ask 01:24:42.000 --> 01:24:44.159 a question, it just surfaces at the 01:24:44.159 --> 01:24:46.239 truth, which is just a random blog post. 01:24:46.239 --> 01:24:48.800 Sorry, no offense, but like I'm it's 01:24:48.800 --> 01:24:51.679 just very surprising that within one or 01:24:51.679 --> 01:24:54.159 two days of that being there now the the 01:24:54.159 --> 01:24:54.560 model 01:24:54.560 --> 01:24:56.480 >> That's because I have clout Vincent. 01:24:56.480 --> 01:24:58.719 That's because I have years of 01:24:58.719 --> 01:25:01.280 publishing on the internet. I have trust 01:25:01.280 --> 01:25:01.840 built up. 01:25:01.840 --> 01:25:04.239 >> I mean, it's part of your domain name 01:25:04.239 --> 01:25:07.360 and it's standing with SEO. So, you're 01:25:07.360 --> 01:25:09.280 you're I think this is one of the 01:25:09.280 --> 01:25:11.360 biggest 01:25:11.360 --> 01:25:13.120 >> Yeah. You know, sometimes people go on 01:25:13.120 --> 01:25:13.920 Reddit and say like 01:25:13.920 --> 01:25:15.760 >> SEO is probably going to become a a 01:25:15.760 --> 01:25:17.600 whole field unto itself. 01:25:17.600 --> 01:25:20.159 >> Yeah. So on Reddit people go like, "Oh, 01:25:20.159 --> 01:25:22.320 you know, I just this small little trick 01:25:22.320 --> 01:25:24.800 I wrote on my website exactly these 01:25:24.800 --> 01:25:28.159 words and now whenever somebody goes and 01:25:28.159 --> 01:25:30.960 um ask Chachi PT, it just recommends my 01:25:30.960 --> 01:25:33.440 tool straight out of the box and and 01:25:33.440 --> 01:25:35.679 people just go and and use it. 01:25:35.679 --> 01:25:38.480 >> I mean, I think it's pretty clear. I 01:25:38.480 --> 01:25:40.159 don't know for how long uh before they 01:25:40.159 --> 01:25:42.159 make it the money game where they go 01:25:42.159 --> 01:25:43.920 like well you want to be your tool on 01:25:43.920 --> 01:25:45.520 top then you're gonna have to pay 01:25:45.520 --> 01:25:48.080 >> like you know the you know Primagen and 01:25:48.080 --> 01:25:49.760 uh and friends they made that like 01:25:49.760 --> 01:25:52.719 coffee ordering business 01:25:52.719 --> 01:25:53.920 >> and I I can't help but think that's 01:25:53.920 --> 01:25:55.600 actually kind of brilliant because now 01:25:55.600 --> 01:25:57.920 because now if you want like in the 01:25:57.920 --> 01:26:00.000 future people are going to go like uh I 01:26:00.000 --> 01:26:02.320 I want some Ethiopian coffee and they 01:26:02.320 --> 01:26:04.239 smash enter and then the agent is going 01:26:04.239 --> 01:26:06.960 to find that coffee and and make the 01:26:06.960 --> 01:26:08.800 order and That's the way it's probably 01:26:08.800 --> 01:26:11.520 going to be in the future. 01:26:11.520 --> 01:26:11.920 >> Yeah, 01:26:11.920 --> 01:26:13.679 >> like shopping. Sh shopping is going to 01:26:13.679 --> 01:26:15.600 be all agent driven. Okay, this is a lot 01:26:15.600 --> 01:26:17.920 of food for thought here. Okay, I need a 01:26:17.920 --> 01:26:19.840 break. I need a break. Good chat, 01:26:19.840 --> 01:26:21.199 Vincent. Thanks very much. 01:26:21.199 --> 01:26:23.679 >> Yeah, good chat the video. 01:26:23.679 --> 01:26:24.320 >> Okay, 01:26:24.320 --> 01:26:27.040 >> boy. What?