WEBVTT 00:00:00.000 --> 00:00:01.680 What about yourself, Vincent? What's 00:00:01.680 --> 00:00:02.760 news with you? 00:00:02.760 --> 00:00:04.800 >> Well, Fable is back, so that was 00:00:04.800 --> 00:00:07.080 exciting. 00:00:07.080 --> 00:00:09.120 >> So, you're smashing it. 00:00:09.120 --> 00:00:12.160 >> Yeah. So, not like 00:00:12.160 --> 00:00:14.240 I just last time Fable was here, I asked 00:00:14.240 --> 00:00:15.760 him to create a proposal on how we 00:00:15.760 --> 00:00:18.680 should implement newer Terraform 00:00:18.680 --> 00:00:20.800 provider features in in a way that's 00:00:20.800 --> 00:00:23.160 compatible both with Terraform and Open 00:00:23.160 --> 00:00:25.560 Tofu, given they have diverged in the 00:00:25.560 --> 00:00:27.960 protocol they implement. And Fable 00:00:27.960 --> 00:00:30.280 created a beautiful proposal of how we 00:00:30.280 --> 00:00:31.800 should implement this. 00:00:31.800 --> 00:00:33.640 And I was like 00:00:33.640 --> 00:00:34.715 you know, at some point I'll implement 00:00:34.715 --> 00:00:35.960 [clears throat] it. 00:00:35.960 --> 00:00:37.640 Uh but now Fable came back. So, I was 00:00:37.640 --> 00:00:39.200 like, "Okay, Fable. 00:00:39.200 --> 00:00:41.480 Um I tried this theory. 00:00:41.480 --> 00:00:43.440 Um so, I I I thought that if you give 00:00:43.440 --> 00:00:45.520 Fable as the orchestrator the 00:00:45.520 --> 00:00:48.440 instruction to use Sonnet 5, which is 00:00:48.440 --> 00:00:50.440 almost Opus quality at a much higher 00:00:50.440 --> 00:00:53.520 speed um and just basically create 00:00:53.520 --> 00:00:55.840 dynamic workflows. So, I for the first 00:00:55.840 --> 00:00:57.760 time in my life used the /goal, 00:00:57.760 --> 00:00:59.680 which is a Claude Code feature to set a 00:00:59.680 --> 00:01:01.520 goal. And the model will keep going 00:01:01.520 --> 00:01:03.800 until it reaches that goal. So, I said 00:01:03.800 --> 00:01:05.720 So, I thought this was a perfect um 00:01:05.720 --> 00:01:07.760 perfect scenario, right? I I I used 00:01:07.760 --> 00:01:09.840 /goal. I pointed to its proposal 00:01:09.840 --> 00:01:12.520 it created a month ago. And I told it 00:01:12.520 --> 00:01:14.720 >> like a a page a one-pager, cuz I thought 00:01:14.720 --> 00:01:16.480 goal was supposed to be like one prompt 00:01:16.480 --> 00:01:17.240 or something. 00:01:17.240 --> 00:01:19.680 >> No, I said the goal is to land this 00:01:19.680 --> 00:01:21.920 proposal. So, it could explore the 00:01:21.920 --> 00:01:23.120 proposal. 00:01:23.120 --> 00:01:23.320 >> I see. 00:01:23.320 --> 00:01:25.320 >> And I made sure that it was on the 00:01:25.320 --> 00:01:27.320 latest head. So, the CDK terrain 00:01:27.320 --> 00:01:29.880 project, which is the, you know, 00:01:29.880 --> 00:01:31.800 community fork of CDK for Terraform 00:01:31.800 --> 00:01:34.400 after HashiCorp deprecated it, has moved 00:01:34.400 --> 00:01:36.720 on significantly. We have removed a 00:01:36.720 --> 00:01:38.240 whole bunch of like deprecated 00:01:38.240 --> 00:01:40.200 dependencies, reduced the binary size, 00:01:40.200 --> 00:01:43.120 addressed several CVEs, and landed like 00:01:43.120 --> 00:01:44.880 functional support like function 00:01:44.880 --> 00:01:47.800 support. So, newer Terraform core and 00:01:47.800 --> 00:01:49.200 OpenTofu, they have a whole bunch of 00:01:49.200 --> 00:01:51.240 like additional functions like cider, 00:01:51.240 --> 00:01:52.680 subnet calculations, and things like 00:01:52.680 --> 00:01:54.120 that. All of that has been now 00:01:54.120 --> 00:01:56.240 integrated within CDK terrain. 00:01:56.240 --> 00:01:57.120 >> of improvements. 00:01:57.120 --> 00:01:59.440 >> A lot. And another thing, somebody's 00:01:59.440 --> 00:02:02.360 >> rewriting the whole CLI to not longer 00:02:02.360 --> 00:02:06.000 use or depend on react for the UI. So 00:02:06.000 --> 00:02:08.360 initially the project creators they used 00:02:08.360 --> 00:02:10.479 react library so that you can render 00:02:10.479 --> 00:02:12.480 components react components in in the 00:02:12.480 --> 00:02:14.880 terminal and it's going it's being 00:02:14.880 --> 00:02:17.120 completely replaced. That one isn't I'm 00:02:17.120 --> 00:02:18.240 I'm not think I don't think we're going 00:02:18.240 --> 00:02:20.000 to cut that right now because we have a 00:02:20.000 --> 00:02:22.800 release PR ready ready to go but we keep 00:02:22.800 --> 00:02:24.840 landing more features and and replacing 00:02:24.840 --> 00:02:28.000 the the user interface like that we is 00:02:28.000 --> 00:02:29.880 quite complicated. So 00:02:29.880 --> 00:02:30.200 >> Wait. 00:02:30.200 --> 00:02:32.720 >> Yeah. So oh he has no access. Okay. 00:02:32.720 --> 00:02:34.800 Anyway, so yeah that that it's been it's 00:02:34.800 --> 00:02:37.480 very interesting. 00:02:37.480 --> 00:02:40.400 >> So you're living you're you're I'm 00:02:40.400 --> 00:02:42.080 assuming you you're doing all these 00:02:42.080 --> 00:02:45.959 development on using Fable now and 00:02:45.959 --> 00:02:47.840 >> So today Fable became available like 00:02:47.840 --> 00:02:49.680 it's supposed to be from July 1st but in 00:02:49.680 --> 00:02:51.760 Vietnam yesterday even at 10:00 p.m. 00:02:51.760 --> 00:02:53.680 before I sleep I was checking cuz I 00:02:53.680 --> 00:02:55.840 already had the direct goal in mind and 00:02:55.840 --> 00:02:57.520 you only get 7 days. 00:02:57.520 --> 00:02:57.840 >> Yeah. 00:02:57.840 --> 00:02:59.840 >> I know that my limit was resetting this 00:02:59.840 --> 00:03:02.200 morning. So I had like 12 hours for my 00:03:02.200 --> 00:03:03.800 weekly limit. I thought I could like 00:03:03.800 --> 00:03:05.480 maximize because then I know my weekly 00:03:05.480 --> 00:03:06.400 limit resets. 00:03:06.400 --> 00:03:07.080 >> Yeah. 00:03:07.080 --> 00:03:08.880 >> But they do track Fable separately so 00:03:08.880 --> 00:03:11.120 I'm not sure how they're tracking it. So 00:03:11.120 --> 00:03:13.200 I had a very clear like goal of what to 00:03:13.200 --> 00:03:14.920 build with Fable the moment it came 00:03:14.920 --> 00:03:15.200 available. 00:03:15.200 --> 00:03:16.320 >> Those features that you mentioned all 00:03:16.320 --> 00:03:17.720 those improvements you've done to city 00:03:17.720 --> 00:03:19.760 cater rain like 00:03:19.760 --> 00:03:22.160 this is only really possible with Claude 00:03:22.160 --> 00:03:23.880 right? This is what I was going to 00:03:23.880 --> 00:03:24.519 allude to. 00:03:24.519 --> 00:03:26.480 >> of it comes from community contributors. 00:03:26.480 --> 00:03:28.320 I know that some of them are using AI. I 00:03:28.320 --> 00:03:30.280 think the the the most difficult bit is 00:03:30.280 --> 00:03:32.400 like having a reliable reviewers. 00:03:32.400 --> 00:03:34.800 There's one person John Steinek who has 00:03:34.800 --> 00:03:36.280 worked on the project a long time ago. 00:03:36.280 --> 00:03:38.959 He's also a Java user so he has a lot of 00:03:38.959 --> 00:03:41.160 requirements related to how the Java 00:03:41.160 --> 00:03:43.200 part of it works. He's been reviewing 00:03:43.200 --> 00:03:45.239 and I really appreciate his reviews 00:03:45.239 --> 00:03:46.880 because they have deep technical 00:03:46.880 --> 00:03:48.920 references like we used to do it this 00:03:48.920 --> 00:03:50.720 way. We decided not to do this. This 00:03:50.720 --> 00:03:53.200 could be a problem. He he goes in a very 00:03:53.200 --> 00:03:55.360 good detail. And I feel a little bit 00:03:55.360 --> 00:03:59.560 guilty if I send a PR and it's all AI 00:03:59.560 --> 00:04:03.600 and honestly, I I just ask AI to 00:04:03.600 --> 00:04:05.520 validate it, which is bad, right? 00:04:05.520 --> 00:04:06.959 Because I come up with a whole bunch of 00:04:06.959 --> 00:04:09.600 scenarios and when it all passes, I do 00:04:09.600 --> 00:04:12.080 like cursory reviews. Um, I do look at 00:04:12.080 --> 00:04:14.480 the the code a little, but not not in 00:04:14.480 --> 00:04:16.440 great detail. So, I feel a bit guilty 00:04:16.440 --> 00:04:17.920 there because I don't want to be the guy 00:04:17.920 --> 00:04:21.280 keeps contributing AI slop. And and and 00:04:21.280 --> 00:04:23.160 and that's like the biggest problem. 00:04:23.160 --> 00:04:23.560 >> Yeah. 00:04:23.560 --> 00:04:25.360 >> One thing that's I feel one of my 00:04:25.360 --> 00:04:26.880 tenants that I've had 00:04:26.880 --> 00:04:28.080 at 00:04:28.080 --> 00:04:30.040 that I've had throughout my life is that 00:04:30.040 --> 00:04:32.040 if I don't know what something is, I do 00:04:32.040 --> 00:04:33.000 try to learn. 00:04:33.000 --> 00:04:33.520 >> Of course. 00:04:33.520 --> 00:04:35.320 >> And but of course with with AI it makes 00:04:35.320 --> 00:04:37.040 it easy just to ship. Like it's kind of 00:04:37.040 --> 00:04:39.360 working. Yeah, go go go. But like 00:04:39.360 --> 00:04:40.600 if there's something in the in the 00:04:40.600 --> 00:04:42.240 programming language like I don't really 00:04:42.240 --> 00:04:43.280 understand like what's the difference 00:04:43.280 --> 00:04:44.480 between 00:04:44.480 --> 00:04:45.919 I don't know this read byte thing and 00:04:45.919 --> 00:04:49.160 this raw thing. I actually usually just 00:04:49.160 --> 00:04:51.520 I try take the time to and and this is 00:04:51.520 --> 00:04:53.040 where again where AI is damn useful. 00:04:53.040 --> 00:04:54.360 Like, can you explain to me the 00:04:54.360 --> 00:04:56.360 difference between this, you know, byte 00:04:56.360 --> 00:04:59.120 array and this raw representation of a 00:04:59.120 --> 00:05:01.000 string? Uh 00:05:01.000 --> 00:05:02.160 And then it and then it tells me and 00:05:02.160 --> 00:05:04.360 then I know and then I can basically not 00:05:04.360 --> 00:05:06.200 understand 00:05:06.200 --> 00:05:08.160 or keep understanding what I'm doing. 00:05:08.160 --> 00:05:10.200 And I feel I mean there must be people 00:05:10.200 --> 00:05:12.200 who have just uh 00:05:12.200 --> 00:05:13.440 Oh yeah, there's been blogs written 00:05:13.440 --> 00:05:15.480 about this cognitive uh 00:05:15.480 --> 00:05:16.680 >> offload your brain 00:05:16.680 --> 00:05:17.840 >> term? 00:05:17.840 --> 00:05:19.400 At At the osmotic cognitive 00:05:19.400 --> 00:05:20.000 >> serenade 00:05:20.000 --> 00:05:21.440 >> serenade or something. When you just 00:05:21.440 --> 00:05:23.520 basically like, okay, now the AI knows 00:05:23.520 --> 00:05:25.120 how to do this. I don't need to know 00:05:25.120 --> 00:05:27.120 this. No, no, you know how to you need 00:05:27.120 --> 00:05:29.440 to know how it works. Oh, yeah. Oh, I I 00:05:29.440 --> 00:05:31.880 wanted to share that um I tried a new 00:05:31.880 --> 00:05:34.800 feature yesterday. Have you heard of 00:05:34.800 --> 00:05:36.800 Hold on. I have to find it. This might 00:05:36.800 --> 00:05:38.400 take me some time. Give me a second. 00:05:38.400 --> 00:05:39.600 >> Yeah, I need just needed to kick off 00:05:39.600 --> 00:05:41.840 SonarQube 5 to do some validation. So so 00:05:41.840 --> 00:05:43.880 one of the main things that I wanted to 00:05:43.880 --> 00:05:45.200 validate 00:05:45.200 --> 00:05:47.040 was explain able to orchestrate 00:05:47.040 --> 00:05:48.960 SonarQube five like I told you right. 00:05:48.960 --> 00:05:50.760 >> Let me share something with you. I think 00:05:50.760 --> 00:05:51.920 I'm I'm going to make a video about 00:05:51.920 --> 00:05:54.680 this. Okay, so you know I'm a number one 00:05:54.680 --> 00:05:57.360 AWS fan for many years and I've been 00:05:57.360 --> 00:05:58.880 doing serverless stuff for a while. But 00:05:58.880 --> 00:05:59.960 when you use 00:05:59.960 --> 00:06:02.040 >> Yeah, I've read this. 00:06:02.040 --> 00:06:04.200 This is nonsense. I mean, they're 00:06:04.200 --> 00:06:06.520 basically getting on par with Terraform. 00:06:06.520 --> 00:06:08.280 >> Well, yeah, yeah, there there are tricks 00:06:08.280 --> 00:06:10.240 to update the binary and this I guess 00:06:10.240 --> 00:06:11.520 this is what they're doing right? 00:06:11.520 --> 00:06:13.640 They're using some sort of some sort of 00:06:13.640 --> 00:06:15.160 I don't know exactly I don't I don't 00:06:15.160 --> 00:06:16.360 think they go into detail about how 00:06:16.360 --> 00:06:19.200 Express works. But I'm I'm assuming it 00:06:19.200 --> 00:06:20.919 does what Terraform 00:06:20.919 --> 00:06:22.120 could do. 00:06:22.120 --> 00:06:24.919 >> No, the the main thing that Express does 00:06:24.919 --> 00:06:28.880 is to avoid the validation phase, right? 00:06:28.880 --> 00:06:30.280 I I read one blog post about 00:06:30.280 --> 00:06:32.880 CloudFormation. They added the 00:06:32.880 --> 00:06:34.560 deployment validation. I thought this 00:06:34.560 --> 00:06:35.880 was Express. 00:06:35.880 --> 00:06:37.520 >> Is it that simple? They just removed the 00:06:37.520 --> 00:06:39.280 bloody validation step. They don't 00:06:39.280 --> 00:06:40.760 mention that in the blog. They don't 00:06:40.760 --> 00:06:42.680 mention that but like yeah, I was I'm 00:06:42.680 --> 00:06:44.320 not showing my whole screen. I tried it 00:06:44.320 --> 00:06:45.800 out and uh 00:06:45.800 --> 00:06:47.360 it is much faster to deploy. 00:06:47.360 --> 00:06:48.520 >> Yeah, it's about the resource 00:06:48.520 --> 00:06:50.120 stabilization. So basically 00:06:50.120 --> 00:06:52.680 CloudFormation is famous for like it's 00:06:52.680 --> 00:06:54.720 very famous it does something Terraform 00:06:54.720 --> 00:06:57.280 doesn't do which is it does 00:06:57.280 --> 00:06:58.840 validates your deployment and if it 00:06:58.840 --> 00:07:00.680 doesn't pass the validation it rolls 00:07:00.680 --> 00:07:03.800 everything back which is a massive 00:07:03.800 --> 00:07:06.160 feature of CloudFormation that Terraform 00:07:06.160 --> 00:07:08.360 misses and to be honest because I was 00:07:08.360 --> 00:07:10.640 building a startup with my friend to 00:07:10.640 --> 00:07:13.919 automate and he was using CloudFormation 00:07:13.919 --> 00:07:15.400 and then we were looking at like uh 00:07:15.400 --> 00:07:17.480 Azure and the biggest issue was how do 00:07:17.480 --> 00:07:19.640 we ensure that the Terraform deployment 00:07:19.640 --> 00:07:21.760 is correct? It's hard. CloudFormation 00:07:21.760 --> 00:07:23.640 does that out of the box but that's also 00:07:23.640 --> 00:07:25.720 the main gripe people have against 00:07:25.720 --> 00:07:28.160 CloudFormation because it goes into this 00:07:28.160 --> 00:07:30.880 crash loop back up not sorry that's 00:07:30.880 --> 00:07:32.960 it goes into these like rollback revert 00:07:32.960 --> 00:07:35.280 failures where you know it deployed like 00:07:35.280 --> 00:07:38.200 an S3 bucket or a DynamoDB table and 00:07:38.200 --> 00:07:39.720 that is flagged as like something 00:07:39.720 --> 00:07:41.680 stateful and then in the rollback it 00:07:41.680 --> 00:07:44.120 refuses to remove those. Those are all 00:07:44.120 --> 00:07:45.960 issues that were like very common in the 00:07:45.960 --> 00:07:47.640 past. I'm not sure if they introduced 00:07:47.640 --> 00:07:50.440 more like fixes for that. But the whole 00:07:50.440 --> 00:07:53.120 CloudFormation validation step was the 00:07:53.120 --> 00:07:54.160 main issue like when people say 00:07:54.160 --> 00:07:56.040 CloudFormation is slow, it's like 00:07:56.040 --> 00:07:57.840 Terraform is so much faster and easier, 00:07:57.840 --> 00:07:59.000 it's because Terraform doesn't do any 00:07:59.000 --> 00:08:00.680 validation. And Terraform also you have 00:08:00.680 --> 00:08:03.200 to build your own runtime around it. Or 00:08:03.200 --> 00:08:05.080 you have to adopt a cloud for it. Like 00:08:05.080 --> 00:08:06.960 CloudFormation is a runtime provided by 00:08:06.960 --> 00:08:09.040 AWS for you. Gives you a whole bunch of 00:08:09.040 --> 00:08:11.600 like um validation mechanisms. The 00:08:11.600 --> 00:08:13.040 runtime is cross-regional, so you don't 00:08:13.040 --> 00:08:14.720 have to go and boost everything. It's 00:08:14.720 --> 00:08:15.200 all there. 00:08:15.200 --> 00:08:17.200 >> For you, I like I'm just try Sorry, I'm 00:08:17.200 --> 00:08:18.440 trying to summarize and make sense of 00:08:18.440 --> 00:08:20.320 this. Do you find the validation useful? 00:08:20.320 --> 00:08:21.760 To be To be honest, I don't find it 00:08:21.760 --> 00:08:23.800 useful. But like am I an idiot? 00:08:23.800 --> 00:08:25.760 >> Well, if you are having to guarantee 00:08:25.760 --> 00:08:27.120 that something It's like a database 00:08:27.120 --> 00:08:28.760 transaction, right? Is a database 00:08:28.760 --> 00:08:31.600 transaction useful? You can you start 00:08:31.600 --> 00:08:33.120 transaction and then you commit 00:08:33.120 --> 00:08:35.560 transaction. It's all or nothing, right? 00:08:35.560 --> 00:08:37.400 That's very hard to guarantee in 00:08:37.400 --> 00:08:39.680 Terraform. Terraform can go go and go 00:08:39.680 --> 00:08:41.320 off and apply a bunch of things and 00:08:41.320 --> 00:08:43.120 leave it like all of these resources 00:08:43.120 --> 00:08:45.040 were updated, but then it failed because 00:08:45.040 --> 00:08:47.320 of the the API endpoint returned an 00:08:47.320 --> 00:08:48.360 error 00:08:48.360 --> 00:08:50.040 and it's blocked. You can't roll back. 00:08:50.040 --> 00:08:53.240 >> like many many times I've deployed 00:08:53.240 --> 00:08:55.560 something and and the rollback feature 00:08:55.560 --> 00:08:57.480 caused more problems than 00:08:57.480 --> 00:08:59.240 it it should have solved. As you were 00:08:59.240 --> 00:09:00.120 saying, right? 00:09:00.120 --> 00:09:02.080 >> I mean, it's it's really up to how you 00:09:02.080 --> 00:09:04.560 like your tolerance for and how you 00:09:04.560 --> 00:09:06.240 approach the problem, right? In some 00:09:06.240 --> 00:09:09.200 scenarios, a full validation is what you 00:09:09.200 --> 00:09:11.440 want. So it's all like it depends. And 00:09:11.440 --> 00:09:13.240 to be honest, if you try to build a full 00:09:13.240 --> 00:09:15.440 validation mechanism, the failure modes 00:09:15.440 --> 00:09:17.800 like you I think this is probably also 00:09:17.800 --> 00:09:19.680 why CloudFormation fails. The failure 00:09:19.680 --> 00:09:22.400 modes in Terraform are very hard to 00:09:22.400 --> 00:09:24.800 recover from sometimes because it goes 00:09:24.800 --> 00:09:25.839 and creates like half of the 00:09:25.839 --> 00:09:27.880 configuration as it should be. And then 00:09:27.880 --> 00:09:29.000 sometimes 00:09:29.000 --> 00:09:30.880 the other parts are not there. It's not 00:09:30.880 --> 00:09:32.760 simple like take back the old config and 00:09:32.760 --> 00:09:34.280 reapply it to go back to your working 00:09:34.280 --> 00:09:36.680 state. It could be completely broken and 00:09:36.680 --> 00:09:38.400 requires you to manually go and figure 00:09:38.400 --> 00:09:40.440 out what actually worked, how do I 00:09:40.440 --> 00:09:41.680 recover from this? Like, is it roll 00:09:41.680 --> 00:09:43.400 forward? Is there a way? Cuz cuz my 00:09:43.400 --> 00:09:45.040 friend had these ideas, like he was 00:09:45.040 --> 00:09:46.760 using full cloud formation, it was very 00:09:46.760 --> 00:09:48.320 reliable, and he was like, "How do I do 00:09:48.320 --> 00:09:49.600 this with Terraform?" He was trying to 00:09:49.600 --> 00:09:51.080 like, "I'm going to snapshot every 00:09:51.080 --> 00:09:52.760 Terraform configuration before we do an 00:09:52.760 --> 00:09:54.480 apply, and then if it fails, we do a 00:09:54.480 --> 00:09:55.880 rollback by just reapplying the old 00:09:55.880 --> 00:09:57.160 one." And I told him, "That's not going 00:09:57.160 --> 00:09:58.720 to work. Terraform doesn't work like 00:09:58.720 --> 00:10:00.280 that. Terraform providers don't work 00:10:00.280 --> 00:10:02.920 like that, and the the the APIs that 00:10:02.920 --> 00:10:04.240 they talk to don't work like that." 00:10:04.240 --> 00:10:04.760 >> Yeah. 00:10:04.760 --> 00:10:06.320 >> And I think that's the main thing that 00:10:06.320 --> 00:10:08.360 people say when they say uh you can't 00:10:08.360 --> 00:10:10.480 treat infrastructure as code, it's in 00:10:10.480 --> 00:10:12.760 configuration, it's not the same 00:10:12.760 --> 00:10:14.880 deployment pipeline as you do with like, 00:10:14.880 --> 00:10:17.000 you know, it's like database migrations, 00:10:17.000 --> 00:10:19.240 right? You can't treat it the same. 00:10:19.240 --> 00:10:20.920 Like, you can't auto approve. That's 00:10:20.920 --> 00:10:22.720 also why people say you can't have like 00:10:22.720 --> 00:10:24.760 a cross plane controller reconciling 00:10:24.760 --> 00:10:27.640 your infrastructure all the time because 00:10:27.640 --> 00:10:28.920 all of these scenarios where the cloud 00:10:28.920 --> 00:10:31.120 goes and upgrades a database engine uh 00:10:31.120 --> 00:10:32.640 behind the scene, and then your 00:10:32.640 --> 00:10:34.520 reconciler tries to downgrade, which is 00:10:34.520 --> 00:10:36.520 not allowed by the API. There's a whole 00:10:36.520 --> 00:10:39.120 bunch of like edge cases that that that 00:10:39.120 --> 00:10:41.280 these like that doesn't fit with ice 00:10:41.280 --> 00:10:42.800 like infrastructure. But 00:10:42.800 --> 00:10:45.240 >> Yeah. Well, I I mean, I know you said it 00:10:45.240 --> 00:10:47.360 depends, but like the majority of the 00:10:47.360 --> 00:10:50.280 time I feel like the Terraform flow and 00:10:50.280 --> 00:10:53.120 the fixed forward approach has been good 00:10:53.120 --> 00:10:55.000 to me. Like, I I don't think I 00:10:55.000 --> 00:10:56.240 >> I wouldn't I wouldn't make a blanket 00:10:56.240 --> 00:10:57.520 statement like you. I would I would 00:10:57.520 --> 00:10:59.200 never. Because 00:10:59.200 --> 00:11:00.320 >> It's all about I I think it's all about 00:11:00.320 --> 00:11:00.680 the 00:11:00.680 --> 00:11:02.960 >> database migrations are also not useful. 00:11:02.960 --> 00:11:04.760 I think you will find if you claim to be 00:11:04.760 --> 00:11:07.080 a developer, a lot of developers upset 00:11:07.080 --> 00:11:08.920 if you say all database migrations are 00:11:08.920 --> 00:11:11.120 not useful. I find that I should just 00:11:11.120 --> 00:11:12.800 always fail forward and never use a 00:11:12.800 --> 00:11:14.680 database migration. Like, no, it 00:11:14.680 --> 00:11:15.280 depends. 00:11:15.280 --> 00:11:17.240 >> True, true, true. I I guess I'm just 00:11:17.240 --> 00:11:19.360 trying to debate with you 00:11:19.360 --> 00:11:21.960 that uh at least the majority of the 00:11:21.960 --> 00:11:25.960 cases don't need that level of support, 00:11:25.960 --> 00:11:29.040 but yeah. I think your point is is 00:11:29.040 --> 00:11:30.320 better made than mine. 00:11:30.320 --> 00:11:33.200 >> Hello and welcome to the AI and podcast. 00:11:33.200 --> 00:11:35.280 In this podcast we like to discuss our 00:11:35.280 --> 00:11:38.200 experience using AI as infrastructure 00:11:38.200 --> 00:11:41.760 engineers and try to hit on some of the 00:11:41.760 --> 00:11:44.560 implications towards how it will impact 00:11:44.560 --> 00:11:46.480 software delivery life cycle and 00:11:46.480 --> 00:11:48.920 security at our job. I think the main 00:11:48.920 --> 00:11:50.280 topic or the 00:11:50.280 --> 00:11:52.080 one of the main highlights of today is 00:11:52.080 --> 00:11:54.280 that Fable is back as you can see on the 00:11:54.280 --> 00:11:56.520 shared screen with lots of tabs open 00:11:56.520 --> 00:11:57.840 which is another 00:11:57.840 --> 00:12:00.200 >> because it requires to turn off 00:12:00.200 --> 00:12:02.560 zero data retention which is not allowed 00:12:02.560 --> 00:12:04.720 in the enterprise environment for fear 00:12:04.720 --> 00:12:07.440 of our our amazing source code being 00:12:07.440 --> 00:12:08.720 leaked to 00:12:08.720 --> 00:12:10.560 >> Yeah, it makes total sense. These models 00:12:10.560 --> 00:12:12.520 are being trained on our usage, right? 00:12:12.520 --> 00:12:14.800 So I am using it for an open source 00:12:14.800 --> 00:12:16.560 project mainly. Some things that I 00:12:16.560 --> 00:12:18.680 really wanted to land into this project 00:12:18.680 --> 00:12:20.760 and I think regular listeners will be 00:12:20.760 --> 00:12:22.640 familiar with the fact with Terraform 00:12:22.640 --> 00:12:24.880 CDK and that I keep talking about how it 00:12:24.880 --> 00:12:26.480 was deprecated and how I forked it with 00:12:26.480 --> 00:12:28.160 a couple of people. And in this 00:12:28.160 --> 00:12:30.640 particular case Fable helped me plan out 00:12:30.640 --> 00:12:33.600 some features. And I was actually before 00:12:33.600 --> 00:12:36.040 we go into that rabbit hole asked a a 00:12:36.040 --> 00:12:39.000 question to Kai earlier which was is he 00:12:39.000 --> 00:12:41.680 actually still using cloud Claude Code 00:12:41.680 --> 00:12:43.360 plan mode which is like one of the 00:12:43.360 --> 00:12:45.320 famous things that drew people into 00:12:45.320 --> 00:12:48.640 cloud coming from chat GPT experiencing 00:12:48.640 --> 00:12:51.080 um you know, an unstructured chat mode 00:12:51.080 --> 00:12:53.360 where you're very hard to set targets 00:12:53.360 --> 00:12:56.040 and getting snippets and bits back or 00:12:56.040 --> 00:12:57.760 evolving into cursor where the agent 00:12:57.760 --> 00:12:59.640 goes and modifies your files and you 00:12:59.640 --> 00:13:02.160 look at typing and re-rewriting the 00:13:02.160 --> 00:13:04.480 whole file originally and then cloud 00:13:04.480 --> 00:13:06.760 code coming up with this plan mode 00:13:06.760 --> 00:13:08.480 laying out cleanly what it is going to 00:13:08.480 --> 00:13:11.240 do and and then people like TV Egge 00:13:11.240 --> 00:13:13.880 coming up with task trackers to do cross 00:13:13.880 --> 00:13:16.160 session content like re-reloading the 00:13:16.160 --> 00:13:16.839 session. 00:13:16.839 --> 00:13:18.480 >> I don't use plan mode 00:13:18.480 --> 00:13:21.440 uh as much anymore. I like I I I do find 00:13:21.440 --> 00:13:23.600 I like yesterday I put it into plan mode 00:13:23.600 --> 00:13:26.480 as a as a as a safety mechanism not for 00:13:26.480 --> 00:13:28.760 it to start editing my code. Like I just 00:13:28.760 --> 00:13:29.600 had 00:13:29.600 --> 00:13:31.440 That's what I use plan mode for. Like 00:13:31.440 --> 00:13:34.360 don't edit my code little idea here, but 00:13:34.360 --> 00:13:35.720 then whatever. 00:13:35.720 --> 00:13:36.440 >> That's right. 00:13:36.440 --> 00:13:38.280 >> I don't use it very much to 00:13:38.280 --> 00:13:40.680 >> I actually see also in open code, it's 00:13:40.680 --> 00:13:42.840 kind of like plan versus build. And I 00:13:42.840 --> 00:13:44.440 see when I work with other people, they 00:13:44.440 --> 00:13:47.000 very heavily use plan mode to put it in 00:13:47.000 --> 00:13:48.520 a read only mode. 00:13:48.520 --> 00:13:50.040 To me it was always like it's going to 00:13:50.040 --> 00:13:51.800 explore and find some things. I think it 00:13:51.800 --> 00:13:53.880 was like you say a guardrail. 00:13:53.880 --> 00:13:55.800 Even if the model decides to to write 00:13:55.800 --> 00:13:57.640 something down, plan mode often blocks 00:13:57.640 --> 00:13:59.160 it saying like you know, you can't. The 00:13:59.160 --> 00:14:01.680 only allowed path for you to write is in 00:14:01.680 --> 00:14:04.200 the plan uh file uh path and using the 00:14:04.200 --> 00:14:06.400 planning tool, I guess. But plan mode 00:14:06.400 --> 00:14:08.160 was great until you had to plan 00:14:08.160 --> 00:14:10.120 something complicated that involved 00:14:10.120 --> 00:14:12.120 doing research on certain libraries or 00:14:12.120 --> 00:14:13.800 even getting buying with other people. 00:14:13.800 --> 00:14:15.640 So then something that became quite 00:14:15.640 --> 00:14:17.560 famous is spectrum in development. And I 00:14:17.560 --> 00:14:19.560 would I adopted it back in September and 00:14:19.560 --> 00:14:21.920 I found it much more effective. I think 00:14:21.920 --> 00:14:23.760 one of the big differences between then 00:14:23.760 --> 00:14:25.240 and now is that we now have 1 million 00:14:25.240 --> 00:14:28.880 context uh tokens and also the agent 00:14:28.880 --> 00:14:31.120 like interactions are much better. The 00:14:31.120 --> 00:14:33.920 main models tend to create complicated 00:14:33.920 --> 00:14:36.160 workflows with sub-agents or even just 00:14:36.160 --> 00:14:38.200 branch out to agents to keep the context 00:14:38.200 --> 00:14:40.400 focused and allow you to go a lot 00:14:40.400 --> 00:14:42.120 further without having to resort to 00:14:42.120 --> 00:14:43.839 these elaborate plans. 00:14:43.839 --> 00:14:45.320 >> I always thought spectrum in development 00:14:45.320 --> 00:14:46.839 to me 00:14:46.839 --> 00:14:48.520 more to do with team alignment. Have you 00:14:48.520 --> 00:14:50.960 Have you used a slash team onboarding 00:14:50.960 --> 00:14:53.000 shortcut or whatever command? 00:14:53.000 --> 00:14:55.280 >> I tried it a long time ago and I didn't 00:14:55.280 --> 00:14:58.280 like it. I don't know. But But like I've 00:14:58.280 --> 00:15:00.320 also got some experience like Originally 00:15:00.320 --> 00:15:02.440 the way I saw uh spectrum in development 00:15:02.440 --> 00:15:05.880 was to create like a a breakdown of why 00:15:05.880 --> 00:15:07.920 I I built something. Sometimes first to 00:15:07.920 --> 00:15:10.120 get buying for other people to comment 00:15:10.120 --> 00:15:13.320 on. So there are very nice tools like 00:15:13.320 --> 00:15:16.040 crit uh and others that allow you to 00:15:16.040 --> 00:15:18.160 collect feedback from people and I think 00:15:18.160 --> 00:15:19.560 that's a lot of these tools are still 00:15:19.560 --> 00:15:22.080 >> You like crit? Okay. I thought you were 00:15:22.080 --> 00:15:22.240 >> I 00:15:22.240 --> 00:15:24.880 >> see very like nice tools like crit. Crit 00:15:24.880 --> 00:15:27.120 became absolute slop. Sorry. Like I 00:15:27.120 --> 00:15:29.160 looked at it and then in just in a span 00:15:29.160 --> 00:15:31.440 of two weeks, the whole root of the 00:15:31.440 --> 00:15:34.040 repository was filled with markdown and 00:15:34.040 --> 00:15:36.040 >> I think crit I mean Thomas is still 00:15:36.040 --> 00:15:38.560 working on it and he's still on X and 00:15:38.560 --> 00:15:40.320 he's very approachable. So I'm going to 00:15:40.320 --> 00:15:41.840 give let's give him the benefit of the 00:15:41.840 --> 00:15:44.080 doubt and say it there is tools like 00:15:44.080 --> 00:15:46.400 crit, but it might have been 00:15:46.400 --> 00:15:47.920 a bad experience. 00:15:47.920 --> 00:15:49.240 >> and I think some of them do it a lot 00:15:49.240 --> 00:15:51.080 better. There's Anyway, I don't want to 00:15:51.080 --> 00:15:52.800 go too much into that details because I 00:15:52.800 --> 00:15:54.480 think what we concluded after you 00:15:54.480 --> 00:15:56.240 answered the question and I answered the 00:15:56.240 --> 00:15:58.280 question was that with the state of the 00:15:58.280 --> 00:15:59.960 tools, I don't see the point of it that 00:15:59.960 --> 00:16:02.280 much anymore. Definitely not for like I 00:16:02.280 --> 00:16:04.880 think we've gone back to Yes, I do read 00:16:04.880 --> 00:16:06.680 the source code and we need to break 00:16:06.680 --> 00:16:08.440 things down in small PR's like what you 00:16:08.440 --> 00:16:09.920 were saying like you you spend a lot of 00:16:09.920 --> 00:16:11.840 time focused on how do I make sure that 00:16:11.840 --> 00:16:14.680 the PR is small enough and reviewable, 00:16:14.680 --> 00:16:16.480 right? 00:16:16.480 --> 00:16:19.680 >> My sense has come in like I I used it in 00:16:19.680 --> 00:16:21.920 my project that I mentioned yesterday to 00:16:21.920 --> 00:16:24.600 break up the commit. So now 00:16:24.600 --> 00:16:25.839 can I share the screen now? 00:16:25.839 --> 00:16:27.720 >> Just just no. Because breaking up 00:16:27.720 --> 00:16:29.839 commits, I I don't see the point and 00:16:29.839 --> 00:16:31.480 also actually Fable is doing this really 00:16:31.480 --> 00:16:33.320 >> makes it really good not not just for 00:16:33.320 --> 00:16:34.640 PR's, but for me. 00:16:34.640 --> 00:16:35.680 >> So 00:16:35.680 --> 00:16:37.200 if I walk you through this, I don't 00:16:37.200 --> 00:16:38.240 think 00:16:38.240 --> 00:16:40.640 Okay. So I didn't see the point of 00:16:40.640 --> 00:16:42.720 breaking down individual commits, but I 00:16:42.720 --> 00:16:45.480 I do see Fable doing this really well. 00:16:45.480 --> 00:16:47.760 Like when you do a plan like when in 00:16:47.760 --> 00:16:49.600 this case, okay, let let's just walk 00:16:49.600 --> 00:16:50.960 away from all of this history of 00:16:50.960 --> 00:16:52.760 spectrum development goal. It's not 00:16:52.760 --> 00:16:54.839 really goal. It's like building 00:16:54.839 --> 00:16:57.480 elaborate documents for planning like 00:16:57.480 --> 00:16:59.920 data models and so on. I think they have 00:16:59.920 --> 00:17:01.680 absolutely have a purpose in certain 00:17:01.680 --> 00:17:03.760 scenarios, but if you're doing open 00:17:03.760 --> 00:17:05.760 source work and you're depending on 00:17:05.760 --> 00:17:08.000 other people's time to do the review and 00:17:08.000 --> 00:17:10.800 you need to land small little PR's that 00:17:10.800 --> 00:17:12.839 work. But they don't care about going 00:17:12.839 --> 00:17:15.520 through massive amount of markdown. And 00:17:15.520 --> 00:17:18.000 and actually when the Vogel he posted 00:17:18.000 --> 00:17:20.600 this uh this famous AWS working back 00:17:20.600 --> 00:17:23.040 from from the customer and user 00:17:23.040 --> 00:17:23.800 experience. 00:17:23.800 --> 00:17:24.120 >> that. 00:17:24.120 --> 00:17:26.280 >> I love that. He He posted, "How does 00:17:26.280 --> 00:17:28.199 this evolve in the age of AI when you 00:17:28.199 --> 00:17:30.120 can build a prototype? Like, when it 00:17:30.120 --> 00:17:31.920 used to take months to build software, 00:17:31.920 --> 00:17:34.280 now you can build a prototype in days." 00:17:34.280 --> 00:17:36.280 And so, they don't just write the 00:17:36.280 --> 00:17:38.400 document any anymore from just a press 00:17:38.400 --> 00:17:40.520 release and FAQ, but they actually also 00:17:40.520 --> 00:17:43.440 build a POC, a prototype. And there were 00:17:43.440 --> 00:17:44.480 two articles, [clears throat] Werner 00:17:44.480 --> 00:17:46.800 Vogels shared one article, and then he 00:17:46.800 --> 00:17:49.120 he pointed to uh I think the team lead 00:17:49.120 --> 00:17:51.880 of Kiro who who who who did that with 00:17:51.880 --> 00:17:53.920 with with her team, how they developed 00:17:53.920 --> 00:17:55.600 uh Kiro. Those are very interesting to 00:17:55.600 --> 00:17:57.360 read. What does it like what what does 00:17:57.360 --> 00:17:59.080 this like working back from the customer 00:17:59.080 --> 00:18:00.960 requirements? But, what I really wanted 00:18:00.960 --> 00:18:02.680 to to dive into right now without going 00:18:02.680 --> 00:18:05.800 too much in any other details there is 00:18:05.800 --> 00:18:08.600 that I tried a Fable 5 with this new 00:18:08.600 --> 00:18:10.600 thing called goal. 00:18:10.600 --> 00:18:13.560 >> Stashka, could you just control plus 00:18:13.560 --> 00:18:14.600 whatever, command plus? 00:18:14.600 --> 00:18:16.880 >> Okay. So, so I opened Cloud, and I do 00:18:16.880 --> 00:18:20.240 have Fable active. And there's a lot of 00:18:20.240 --> 00:18:22.280 people complaining about the the usage 00:18:22.280 --> 00:18:23.960 that you get. You only get it for 7 00:18:23.960 --> 00:18:26.640 days, and it's very hard limited. But, 00:18:26.640 --> 00:18:28.920 the way I saw it is like Fable shouldn't 00:18:28.920 --> 00:18:30.600 be the one doing most of the work, 00:18:30.600 --> 00:18:32.800 right? It should just be the expert like 00:18:32.800 --> 00:18:35.160 overhead orchestrator. And then, I 00:18:35.160 --> 00:18:36.680 thought because Suno AI 5 came out as 00:18:36.680 --> 00:18:39.320 well, and it's like almost Opus level at 00:18:39.320 --> 00:18:41.480 a fraction of the cost and much faster 00:18:41.480 --> 00:18:44.000 than Opus 4.8, they say. Therefore, I 00:18:44.000 --> 00:18:47.240 thought, "Hey, I have an RFC here that 00:18:47.240 --> 00:18:50.480 was built by Fable a month ago when it 00:18:50.480 --> 00:18:53.200 when it was available to 00:18:53.200 --> 00:18:55.520 Yeah, for It was available I I used it 00:18:55.520 --> 00:18:57.800 for two two or three days, I think. And 00:18:57.800 --> 00:19:00.120 basically, I asked it to to do a full 00:19:00.120 --> 00:19:02.160 sweep of all of the Terraform provider 00:19:02.160 --> 00:19:04.800 capabilities across OpenTofu and 00:19:04.800 --> 00:19:08.040 Terraform because each of them land the 00:19:08.040 --> 00:19:09.760 support different at different versions 00:19:09.760 --> 00:19:11.440 at different releases. So, we want to 00:19:11.440 --> 00:19:13.760 make sure that if somebody uses our meta 00:19:13.760 --> 00:19:15.560 language, which is in TypeScript, to 00:19:15.560 --> 00:19:17.440 configure your Terraform, basically gets 00:19:17.440 --> 00:19:19.760 like a lot of niceties like if you use 00:19:19.760 --> 00:19:22.320 this feature, you got a little synth 00:19:22.320 --> 00:19:23.760 warning like the configuration you're 00:19:23.760 --> 00:19:25.520 generating right now, it will work with 00:19:25.520 --> 00:19:28.720 OpenTofu 1.10. You are currently on on 00:19:28.720 --> 00:19:30.800 1.5. It doesn't It won't work or 00:19:30.800 --> 00:19:32.560 something like that. They started at 1.6 00:19:32.560 --> 00:19:34.640 actually, I think. So, yeah, that that's 00:19:34.640 --> 00:19:35.960 kind of like the things that that Fable 00:19:35.960 --> 00:19:38.320 actually landed a month ago, the ability 00:19:38.320 --> 00:19:40.520 to and we went through some iterations. 00:19:40.520 --> 00:19:42.200 I I let it do completely organically. 00:19:42.200 --> 00:19:44.240 Fable, just tell me how you would do it. 00:19:44.240 --> 00:19:45.800 Initially, Fable quickly highlighted 00:19:45.800 --> 00:19:47.360 like, "Hey, we're doing a config 00:19:47.360 --> 00:19:49.360 generation and we're we're calling out 00:19:49.360 --> 00:19:51.160 to a binary, which is kind of bad 00:19:51.160 --> 00:19:53.280 because the binary callout could be on a 00:19:53.280 --> 00:19:54.920 synthesis machine that doesn't have 00:19:54.920 --> 00:19:56.480 access to the right binary, right? The 00:19:56.480 --> 00:19:57.920 binary only gets called when you do the 00:19:57.920 --> 00:20:00.920 apply." So, we we we flipped it around 00:20:00.920 --> 00:20:02.640 like Fable kind of highlighted it that 00:20:02.640 --> 00:20:04.560 it would make more sense to let the the 00:20:04.560 --> 00:20:06.480 people writing the configuration declare 00:20:06.480 --> 00:20:08.520 what they support and then when you call 00:20:08.520 --> 00:20:10.000 those configurations, you get this 00:20:10.000 --> 00:20:11.880 validation happening when the binary is 00:20:11.880 --> 00:20:13.880 actually being executed when you do the 00:20:13.880 --> 00:20:15.600 plan and the apply and then it will 00:20:15.600 --> 00:20:17.040 highlight, "Okay, you know, there's a 00:20:17.040 --> 00:20:19.320 couple of functions in here that that 00:20:19.320 --> 00:20:20.840 are not going to work uh because it's 00:20:20.840 --> 00:20:22.520 not supported in Terraform, it's only 00:20:22.520 --> 00:20:24.400 OpenTofu available, for example." So, 00:20:24.400 --> 00:20:26.440 anyway, yeah, so that that Fable was 00:20:26.440 --> 00:20:28.720 able to land before it was taken away 00:20:28.720 --> 00:20:30.600 and then it also did a full sweep of all 00:20:30.600 --> 00:20:33.040 of the provider schemas. So, looking at 00:20:33.040 --> 00:20:35.360 at the plugin protocol, so if you're 00:20:35.360 --> 00:20:37.120 familiar with Go lang and Terraform, Go 00:20:37.120 --> 00:20:38.800 lang doesn't really support like plugins 00:20:38.800 --> 00:20:41.200 well unless you do some type of GRPC or 00:20:41.200 --> 00:20:42.400 HTTP 00:20:42.400 --> 00:20:43.120 um 00:20:43.120 --> 00:20:45.400 in like interaction between the main Go 00:20:45.400 --> 00:20:47.120 lang process and then other processes, 00:20:47.120 --> 00:20:49.120 which are the plugins. So, this protocol 00:20:49.120 --> 00:20:50.760 is is is defined. 00:20:50.760 --> 00:20:52.960 >> Okay, like like binaries type stuff, 00:20:52.960 --> 00:20:53.560 right? 00:20:53.560 --> 00:20:54.520 >> Yes. 00:20:54.520 --> 00:20:55.360 >> Okay. 00:20:55.360 --> 00:20:57.120 >> Go lang is not good with like in-process 00:20:57.120 --> 00:20:59.000 plugins system. 00:20:59.000 --> 00:21:00.600 >> that's the benefit of Go is everything's 00:21:00.600 --> 00:21:02.160 statically built, but yeah, okay, I get 00:21:02.160 --> 00:21:03.560 where you're going at. Yeah. 00:21:03.560 --> 00:21:06.720 >> Yeah, so so so the Terraform um 00:21:06.720 --> 00:21:08.880 works through protocols like the plugin 00:21:08.880 --> 00:21:12.280 protocol and Fable did the full matrix 00:21:12.280 --> 00:21:14.160 built of what are the the different 00:21:14.160 --> 00:21:15.920 versions that the port which plugin 00:21:15.920 --> 00:21:17.160 protocols 00:21:17.160 --> 00:21:18.440 between 00:21:18.440 --> 00:21:20.720 both of them. And then I just kept it 00:21:20.720 --> 00:21:22.880 very simple. Like I didn't even pointed 00:21:22.880 --> 00:21:24.600 at the proposal. There's an RFS it's 00:21:24.600 --> 00:21:26.480 like an RFC. There's a proposal in there 00:21:26.480 --> 00:21:27.840 but there's like a lot of like a static 00:21:27.840 --> 00:21:30.000 schema everything. I said look, our goal 00:21:30.000 --> 00:21:31.920 is to land this proposal and we want to 00:21:31.920 --> 00:21:34.240 land it into this into this library. And 00:21:34.240 --> 00:21:36.400 I I did want it to like not do bulk of 00:21:36.400 --> 00:21:38.200 the work and use iterative dynamic 00:21:38.200 --> 00:21:40.400 workflows with Sonar 5 agents because I 00:21:40.400 --> 00:21:41.640 thought that's going to save me cost, 00:21:41.640 --> 00:21:43.560 right? Fable is going to do the like the 00:21:43.560 --> 00:21:46.280 overall project status planning and then 00:21:46.280 --> 00:21:47.720 Sonar is going to do bulk of the bulk of 00:21:47.720 --> 00:21:47.840 the 00:21:47.840 --> 00:21:49.720 >> I got deja vu. This was the case with 00:21:49.720 --> 00:21:52.160 Opus and Sonar months back, right? But 00:21:52.160 --> 00:21:52.680 okay. 00:21:52.680 --> 00:21:53.080 >> Right. 00:21:53.080 --> 00:21:53.760 >> Here we go. 00:21:53.760 --> 00:21:54.440 >> So 00:21:54.440 --> 00:21:56.360 So this and then actually funny because 00:21:56.360 --> 00:21:58.040 I had to leave the house. So I really 00:21:58.040 --> 00:21:59.640 wanted to kick this off this morning 00:21:59.640 --> 00:22:01.320 because Fable was available. You only 00:22:01.320 --> 00:22:03.720 have 7 days. I have to go out of the 00:22:03.720 --> 00:22:05.160 house for like 2 3 hours. 00:22:05.160 --> 00:22:07.080 >> hell for you, man. 00:22:07.080 --> 00:22:08.520 >> Yeah, so 00:22:08.520 --> 00:22:10.640 there's auto mode on and goal was 00:22:10.640 --> 00:22:12.280 acknowledged and it just went off, 00:22:12.280 --> 00:22:14.080 right? It started doing commits and this 00:22:14.080 --> 00:22:16.080 is like it makes very small commits, 00:22:16.080 --> 00:22:18.560 very detailed and it even when it 00:22:18.560 --> 00:22:20.360 finishes it gives you an overview and 00:22:20.360 --> 00:22:22.680 we'll see that. So it it it generates 00:22:22.680 --> 00:22:24.960 like the the mappers, blah blah blah, 00:22:24.960 --> 00:22:26.400 whole bunch of stuff. I didn't even read 00:22:26.400 --> 00:22:28.160 it. And I think this is another pattern 00:22:28.160 --> 00:22:30.720 that I started to follow is that I'm no 00:22:30.720 --> 00:22:32.440 longer following the stream of what the 00:22:32.440 --> 00:22:35.080 model is doing. Like I I usually am not 00:22:35.080 --> 00:22:36.400 at the at the at the chair and I'm 00:22:36.400 --> 00:22:37.760 scrolling back and trying to figure out 00:22:37.760 --> 00:22:40.320 what happened is kind of annoying. So 00:22:40.320 --> 00:22:42.200 what I I I tend to do now is if I have a 00:22:42.200 --> 00:22:45.080 question if my ADHD pulled my attention 00:22:45.080 --> 00:22:47.000 away before I was able to read the whole 00:22:47.000 --> 00:22:49.200 summary that that that Fable provided, I 00:22:49.200 --> 00:22:51.400 just go by the way, was there anything 00:22:51.400 --> 00:22:53.040 unusual or was there something wrong in 00:22:53.040 --> 00:22:55.000 the proposal that that was that you 00:22:55.000 --> 00:22:56.920 needed to work around, things like that 00:22:56.920 --> 00:22:58.040 that I wanted to highlight. 00:22:58.040 --> 00:22:58.560 >> Isn't [clears throat] there isn't there 00:22:58.560 --> 00:23:00.960 like a slash recap command, too? 00:23:00.960 --> 00:23:03.440 >> There is like it always it 00:23:03.440 --> 00:23:04.920 shows up, but it's like just one line. 00:23:04.920 --> 00:23:06.560 It doesn't answer my questions 00:23:06.560 --> 00:23:08.480 >> Yeah, true. I I 00:23:08.480 --> 00:23:10.760 I'm just like you, Vincent. I It does 00:23:10.760 --> 00:23:12.440 some stuff, and I'm like, I'm switch 00:23:12.440 --> 00:23:14.040 context, and then I'm back there like, 00:23:14.040 --> 00:23:16.000 "What did it do what I wanted it to do?" 00:23:16.000 --> 00:23:17.680 >> I think this is normal. I We're at the 00:23:17.680 --> 00:23:19.760 point where we're no longer driving the 00:23:19.760 --> 00:23:22.120 shell. We're not at the seat. We're We 00:23:22.120 --> 00:23:23.760 may have five or six running at the same 00:23:23.760 --> 00:23:25.720 time, and we're just switching between 00:23:25.720 --> 00:23:27.040 them, and we need to very quickly pick 00:23:27.040 --> 00:23:28.000 up what happened. 00:23:28.000 --> 00:23:30.080 >> Yeah. 00:23:30.080 --> 00:23:32.080 >> posted there is definitely a feature 00:23:32.080 --> 00:23:33.680 that's exactly aimed at that. Like, 00:23:33.680 --> 00:23:35.200 you're coming back into this What Where 00:23:35.200 --> 00:23:37.240 are we? What did we just do? What So, 00:23:37.240 --> 00:23:38.920 it's fully landed, it's told me, but 00:23:38.920 --> 00:23:40.480 that didn't give me much. So, then it 00:23:40.480 --> 00:23:41.880 also told me like all of the things it 00:23:41.880 --> 00:23:43.160 did. Um 00:23:43.160 --> 00:23:44.360 >> What landed? 00:23:44.360 --> 00:23:46.120 >> This was the This is review. Hold on. 00:23:46.120 --> 00:23:48.280 We're not yet at the review. Here, this 00:23:48.280 --> 00:23:50.440 is This is where it goes. So, done. It's 00:23:50.440 --> 00:23:52.960 landed, and there's seven commits. And 00:23:52.960 --> 00:23:54.360 here are each one of them, and what is 00:23:54.360 --> 00:23:56.120 your focus for the review? If you review 00:23:56.120 --> 00:23:57.240 these commits, this is what you need to 00:23:57.240 --> 00:23:58.000 focus on. 00:23:58.000 --> 00:24:00.440 >> Following conventional commits, is that 00:24:00.440 --> 00:24:02.280 Is that out of the box, or did you have 00:24:02.280 --> 00:24:03.680 that in your claw.md? 00:24:03.680 --> 00:24:05.680 >> I can't I don't know because does it 00:24:05.680 --> 00:24:07.600 does it look at the git log? I deployed 00:24:07.600 --> 00:24:09.160 release please from Google, which 00:24:09.160 --> 00:24:10.680 requires conventional commits. Maybe 00:24:10.680 --> 00:24:13.680 that drove it. So, I can't say if this 00:24:13.680 --> 00:24:14.240 is a 00:24:14.240 --> 00:24:16.280 >> Should Should ask it why How do you know 00:24:16.280 --> 00:24:16.880 you're 00:24:16.880 --> 00:24:18.440 >> I'm I'm 100% sure it's actually 00:24:18.440 --> 00:24:20.200 following the convention because there's 00:24:20.200 --> 00:24:22.640 a contributing.md from the original 00:24:22.640 --> 00:24:25.360 Terraform CDK, which which clearly says 00:24:25.360 --> 00:24:27.200 you need to put the the convention with 00:24:27.200 --> 00:24:29.480 the library or the area of effect. This 00:24:29.480 --> 00:24:31.200 is how contributing guideline tells it 00:24:31.200 --> 00:24:32.920 to do. So, it it followed that, yeah. 00:24:32.920 --> 00:24:35.200 So, it did four iterative workflows, and 00:24:35.200 --> 00:24:36.880 I sent you a screenshot of how much it 00:24:36.880 --> 00:24:39.040 used of my 5-hour session, right? It 00:24:39.040 --> 00:24:41.440 used about, I would say, 60% of my 00:24:41.440 --> 00:24:43.440 5-hour session in just 1 hour. Yeah, 00:24:43.440 --> 00:24:45.640 this is the cool thing here. Shown Shown 00:24:45.640 --> 00:24:47.720 for 1 hour and 14 minutes. Now, this 00:24:47.720 --> 00:24:50.400 isn't a very big I Actually, I think 00:24:50.400 --> 00:24:52.400 it's 4,000 line diff or something, so it 00:24:52.400 --> 00:24:53.960 is not small. Uh but it's not like 00:24:53.960 --> 00:24:55.800 people that used to show Fable building 00:24:55.800 --> 00:24:58.360 like whole 3D games in one shot, right? 00:24:58.360 --> 00:24:58.840 >> Yeah. 00:24:58.840 --> 00:24:59.680 >> So, I don't think it's 00:24:59.680 --> 00:25:01.080 >> I don't quite understand your point. I 00:25:01.080 --> 00:25:03.120 mean, it's still you 00:25:03.120 --> 00:25:04.720 >> said it kept running for like a day or 00:25:04.720 --> 00:25:07.160 so or like 12 hours. For me, I was I was 00:25:07.160 --> 00:25:08.840 happy to go away for 2 hours and come 00:25:08.840 --> 00:25:11.520 back to it having run 1 hour and 14 00:25:11.520 --> 00:25:11.800 minutes. 00:25:11.800 --> 00:25:14.120 >> get that point. That for it to to to 00:25:14.120 --> 00:25:16.280 work away for an hour and get some great 00:25:16.280 --> 00:25:18.360 results is fine by me, too. Yeah. 00:25:18.360 --> 00:25:21.080 >> Yeah. And then, because I I also want to 00:25:21.080 --> 00:25:22.880 use the models I efficiently and I had 00:25:22.880 --> 00:25:25.800 to also go eat and stuff. So, I I then 00:25:25.800 --> 00:25:28.440 asked my Hermes agent while I was on the 00:25:28.440 --> 00:25:31.520 go, uh which is using GPT-5.5, 00:25:31.520 --> 00:25:34.360 to basically build a demo repository. 00:25:34.360 --> 00:25:36.000 So, this was the PR it created. It's 00:25:36.000 --> 00:25:37.800 quite massive. So, then I was like, now 00:25:37.800 --> 00:25:39.200 I have to read all of this, you know? 00:25:39.200 --> 00:25:40.680 And I was like, I don't 00:25:40.680 --> 00:25:41.760 What do you mean? I don't really want to 00:25:41.760 --> 00:25:42.240 read it. 00:25:42.240 --> 00:25:44.280 >> did you get Codex to pick up the the 00:25:44.280 --> 00:25:46.240 code session? I don't quite follow. 00:25:46.240 --> 00:25:48.280 >> So, no, no, no. This this is a draft PR 00:25:48.280 --> 00:25:51.240 open on on there. And then I just asked, 00:25:51.240 --> 00:25:52.800 um where is it? 00:25:52.800 --> 00:25:54.840 >> Well, you use spaces like a fiend. I 00:25:54.840 --> 00:25:56.720 should I should maybe use spaces, 00:25:56.720 --> 00:25:58.280 actually. 00:25:58.280 --> 00:25:59.760 Do you do you use your trackpad or do 00:25:59.760 --> 00:26:01.120 you use the keyboard to switch? 00:26:01.120 --> 00:26:04.400 >> I I use my my uh MacBook uh laptop 00:26:04.400 --> 00:26:06.120 trackpad. Well, you know, with the three 00:26:06.120 --> 00:26:08.360 fingers, like you can do so much more. 00:26:08.360 --> 00:26:10.480 Um so, yeah, I 00:26:10.480 --> 00:26:13.720 I basically asked Hermes to 00:26:13.720 --> 00:26:14.920 when I was at the keyboard still, I 00:26:14.920 --> 00:26:16.440 should go to the top of the thread. 00:26:16.440 --> 00:26:18.600 Yeah, I say, "Hey, can you review this 00:26:18.600 --> 00:26:20.160 draft PR?" How do I make this? 00:26:20.160 --> 00:26:22.080 >> Sorry, and this was the PR created by 00:26:22.080 --> 00:26:22.640 Sona. 00:26:22.640 --> 00:26:24.640 >> Uh by yeah, Fable. So, this is 00:26:24.640 --> 00:26:25.320 >> Fable. 00:26:25.320 --> 00:26:27.080 >> This is the draft PR. 00:26:27.080 --> 00:26:27.560 >> Okay, so it's 00:26:27.560 --> 00:26:28.880 >> And I have some tooling called review 00:26:28.880 --> 00:26:31.280 polar. So, the review polar is a tool 00:26:31.280 --> 00:26:32.840 that runs on a cron and always prepares 00:26:32.840 --> 00:26:34.800 a work tree for certain repositories 00:26:34.800 --> 00:26:37.160 that I'm reviewing. I set this like it 00:26:37.160 --> 00:26:39.160 did a full review. Oh, it actually also 00:26:39.160 --> 00:26:41.520 read contributing, everything. So, these 00:26:41.520 --> 00:26:43.320 ones pop up, like you like you 00:26:43.320 --> 00:26:44.560 mentioned, this is annoying. I think 00:26:44.560 --> 00:26:46.160 this one I missed. So, this one didn't 00:26:46.160 --> 00:26:49.440 execute. So, Codex is like GPT is trying 00:26:49.440 --> 00:26:50.960 to do a review, but half of the things 00:26:50.960 --> 00:26:52.840 it's trying to see are are like timing 00:26:52.840 --> 00:26:54.360 out because I'm not paying attention to 00:26:54.360 --> 00:26:56.840 my phone and I didn't approve it. So, it 00:26:56.840 --> 00:26:59.240 actually found a few like concerns. I 00:26:59.240 --> 00:27:01.440 wasn't sure if they were very relevant. 00:27:01.440 --> 00:27:03.560 But, what I really wanted this reviewer 00:27:03.560 --> 00:27:05.280 to do is 00:27:05.280 --> 00:27:08.040 is to build a demo, you know? So, 00:27:08.040 --> 00:27:10.680 basically, where is my I wanted it to 00:27:10.680 --> 00:27:13.080 build a a demo of this new feature. So, 00:27:13.080 --> 00:27:15.000 this is not yet landed in into the 00:27:15.000 --> 00:27:17.840 binary and I wanted to build examples 00:27:17.840 --> 00:27:19.960 using this feature before it landed. Of 00:27:19.960 --> 00:27:21.560 course, I do think we have full 00:27:21.560 --> 00:27:24.320 integration tests inside this as well, 00:27:24.320 --> 00:27:26.240 but I wanted something very simple that 00:27:26.240 --> 00:27:28.800 I could easily review. So, this actually 00:27:28.800 --> 00:27:31.480 showed like it's importing the the 00:27:31.480 --> 00:27:33.680 Terraform provider AWS functions and 00:27:33.680 --> 00:27:35.680 it's using the ARN parse, it's using the 00:27:35.680 --> 00:27:38.480 ARN build, it's using trim IAM role 00:27:38.480 --> 00:27:40.120 path. So, it's using all of these 00:27:40.120 --> 00:27:41.600 functions that the Terraform provider 00:27:41.600 --> 00:27:43.800 AWS now ships with it so that you can 00:27:43.800 --> 00:27:45.640 use in your Terraform configuration. So, 00:27:45.640 --> 00:27:47.080 it's actually came up with a couple of 00:27:47.080 --> 00:27:49.840 like basic examples of how to use it. 00:27:49.840 --> 00:27:51.720 And then it did the whole Terraform plan 00:27:51.720 --> 00:27:53.440 against my AWS account. That's what the 00:27:53.440 --> 00:27:55.720 Hermes agent did. So, my Hermes agent 00:27:55.720 --> 00:27:57.560 had a Of course, Claude Code on my laptop 00:27:57.560 --> 00:27:59.560 has a lot of capabilities as well, but 00:27:59.560 --> 00:28:01.320 my Hermes agent was accessible while I 00:28:01.320 --> 00:28:05.040 was outside with Discord and able to 00:28:05.040 --> 00:28:06.960 >> Yeah, that seems pretty cool. Like 00:28:06.960 --> 00:28:09.160 getting your getting your Hermes agent 00:28:09.160 --> 00:28:13.440 to to review Claude output. So, long 00:28:13.440 --> 00:28:16.560 story short, was a bit But, I mean, what 00:28:16.560 --> 00:28:19.000 I often think is lacking here is is the 00:28:19.000 --> 00:28:21.160 back and forth. Like, is there some back 00:28:21.160 --> 00:28:23.040 and forth? Is there some iterations 00:28:23.040 --> 00:28:24.360 going on? 00:28:24.360 --> 00:28:26.960 >> So, after the Hermes agent did a bunch 00:28:26.960 --> 00:28:29.960 of tests and executed, one of the things 00:28:29.960 --> 00:28:31.760 I asked it like, "Hey, tell me how this 00:28:31.760 --> 00:28:34.440 feature felt from a UX." Cuz the LLM is 00:28:34.440 --> 00:28:36.440 making assumptions on how it works and 00:28:36.440 --> 00:28:38.160 it writes an example, then it runs and 00:28:38.160 --> 00:28:41.000 it fails. And so so so so Hermes said, 00:28:41.000 --> 00:28:43.080 "I don't know. I saw that it tried to do 00:28:43.080 --> 00:28:44.800 something and it failed." And I said, 00:28:44.800 --> 00:28:46.840 "Can And write down everything that you 00:28:46.840 --> 00:28:49.160 assumed was going to work but failed. 00:28:49.160 --> 00:28:50.960 Because that's important UX as well. 00:28:50.960 --> 00:28:53.120 >> All right. That's interesting. 00:28:53.120 --> 00:28:55.240 >> It did actually say here, UX integration 00:28:55.240 --> 00:28:56.560 notes from the demo. I don't know if 00:28:56.560 --> 00:28:58.280 they're that useful, but it did say 00:28:58.280 --> 00:29:00.280 like, "Hey, I tried to import from a 00:29:00.280 --> 00:29:04.040 barrel import and Node.js 00:29:04.040 --> 00:29:05.920 went ran ran out of memory. So, this 00:29:05.920 --> 00:29:09.080 this Hermes agent had 8 GB of memory and 00:29:09.080 --> 00:29:09.720 um 00:29:09.720 --> 00:29:11.680 and it doing a barrel import didn't 00:29:11.680 --> 00:29:13.480 work. So, I think that's an like very 00:29:13.480 --> 00:29:15.080 relevant UX 00:29:15.080 --> 00:29:16.760 feedback. And another one was like, I 00:29:16.760 --> 00:29:19.040 tried to do things while the rest is, 00:29:19.040 --> 00:29:20.640 you know, fallout. Another thing is 00:29:20.640 --> 00:29:23.760 like, ephemeral as null had to be used 00:29:23.760 --> 00:29:25.480 with sensitive true. So, this is like 00:29:25.480 --> 00:29:27.400 very Terraform specific. Didn't really 00:29:27.400 --> 00:29:29.360 care too much about that. But, one of 00:29:29.360 --> 00:29:30.760 the things that that that made me 00:29:30.760 --> 00:29:33.360 annoyed was like, I removed an example 00:29:33.360 --> 00:29:36.360 of user agent. I was like, why? Why do 00:29:36.360 --> 00:29:38.040 you need to remove an example? What what 00:29:38.040 --> 00:29:40.520 went wrong? Why why didn't it work? And 00:29:40.520 --> 00:29:42.200 so, I I spent like a lot of time trying 00:29:42.200 --> 00:29:44.080 to figure out what is this user agent 00:29:44.080 --> 00:29:47.040 function for? And even asked at ChatGPT, 00:29:47.040 --> 00:29:49.000 what are the common use cases and how do 00:29:49.000 --> 00:29:50.720 you use this? So, I went a bit down a 00:29:50.720 --> 00:29:53.160 rabbit hole of this one particular thing 00:29:53.160 --> 00:29:54.560 that the Terraform provider AWS 00:29:54.560 --> 00:29:55.960 provides. But, it was interesting 00:29:55.960 --> 00:29:58.240 because it showed, for example, that you 00:29:58.240 --> 00:30:00.080 can use this feature in your in your 00:30:00.080 --> 00:30:02.800 module so that CloudTrail shows which 00:30:02.800 --> 00:30:04.720 Terraform module is creating these 00:30:04.720 --> 00:30:07.320 resources. So, you can attribute things. 00:30:07.320 --> 00:30:10.040 So, where is the actual usage of it? 00:30:10.040 --> 00:30:15.000 Here. So, in in the database version.tf 00:30:15.000 --> 00:30:15.920 >> small screen here. 00:30:15.920 --> 00:30:17.640 >> Trying to zoom in. So, here is the 00:30:17.640 --> 00:30:22.240 Terraform provider AWS module and they 00:30:22.240 --> 00:30:25.400 insert metadata to highlight that that 00:30:25.400 --> 00:30:28.440 it's the AWS modules creating this. And 00:30:28.440 --> 00:30:31.080 and if you go back to the reason that 00:30:31.080 --> 00:30:34.040 feature is there is because CloudTrail 00:30:34.040 --> 00:30:36.840 shows you generic Terraform execution, 00:30:36.840 --> 00:30:40.880 but this augments your logs showing you 00:30:40.880 --> 00:30:44.200 which module is creating those things. 00:30:44.200 --> 00:30:44.720 >> That's cool. 00:30:44.720 --> 00:30:46.040 >> Yeah, it's it's really cool and it's 00:30:46.040 --> 00:30:47.840 very useful also if you build 00:30:47.840 --> 00:30:49.440 complex constructs, right? If you have 00:30:49.440 --> 00:30:50.440 like 00:30:50.440 --> 00:30:51.440 L2 constructs. 00:30:51.440 --> 00:30:52.960 >> I'm now I'm now curious what the log 00:30:52.960 --> 00:30:55.640 looks like. Does it also put a version 00:30:55.640 --> 00:30:57.000 number in, I wonder? 00:30:57.000 --> 00:30:59.400 >> Yeah, I need to find back to the 00:30:59.400 --> 00:31:01.000 >> Cuz version numbers are so key in my 00:31:01.000 --> 00:31:03.400 opinion. Like every software needs to 00:31:03.400 --> 00:31:05.560 show me the version number. And and 00:31:05.560 --> 00:31:07.320 >> By default, yeah. The annoying thing 00:31:07.320 --> 00:31:10.000 about on about Jet GPT is when I ask it 00:31:10.000 --> 00:31:11.920 to give examples, it goes from all of 00:31:11.920 --> 00:31:13.240 the memories it knows about me. Like 00:31:13.240 --> 00:31:15.680 this is a tool I bit built. And it goes 00:31:15.680 --> 00:31:17.560 like, "Oh, this is how you could use it. 00:31:17.560 --> 00:31:19.600 You could 00:31:19.600 --> 00:31:21.800 >> You could use your your binary show the 00:31:21.800 --> 00:31:23.240 the binary and the version. 00:31:23.240 --> 00:31:25.080 >> memory for these very reasons. Like it 00:31:25.080 --> 00:31:27.280 just picks up something random. Like, 00:31:27.280 --> 00:31:29.240 "Oh, I know you have two kids and this 00:31:29.240 --> 00:31:31.240 is why I've done it." Like, "No, I mean, 00:31:31.240 --> 00:31:32.000 you shouldn't 00:31:32.000 --> 00:31:34.360 >> This is the Yeah, this was the one why 00:31:34.360 --> 00:31:36.160 they added They added I think the main 00:31:36.160 --> 00:31:38.560 function is like provider meta support 00:31:38.560 --> 00:31:40.280 and and it shows here like can show 00:31:40.280 --> 00:31:43.360 example module version, example comment. 00:31:43.360 --> 00:31:45.360 >> Okay, so you went down some rabbit hole 00:31:45.360 --> 00:31:47.960 because of your adversarial review. So, 00:31:47.960 --> 00:31:49.720 okay, so you you went down that rabbit 00:31:49.720 --> 00:31:51.480 hole. That then what? I'm still curious 00:31:51.480 --> 00:31:53.960 if your I mean, your your PR review 00:31:53.960 --> 00:31:56.880 thing with with Codex or looks really 00:31:56.880 --> 00:31:59.240 cool, but is it going back and forth? 00:31:59.240 --> 00:32:01.560 This is the key that I feel is missing, 00:32:01.560 --> 00:32:03.640 but I guess being able to review on the 00:32:03.640 --> 00:32:07.000 go from Discord is is pretty cool. 00:32:07.000 --> 00:32:08.280 >> Yeah. 00:32:08.280 --> 00:32:10.560 Um so, it was able to like build all of 00:32:10.560 --> 00:32:12.680 this and then I told it, "But it doesn't 00:32:12.680 --> 00:32:14.760 comment until I tell it to." So, then I 00:32:14.760 --> 00:32:16.160 had I had I was on my phone, so I have 00:32:16.160 --> 00:32:18.080 to like copy these messages and you 00:32:18.080 --> 00:32:19.800 can't select things in Discord, which is 00:32:19.800 --> 00:32:21.320 very annoying. So, I have to like right 00:32:21.320 --> 00:32:23.120 click, copy text. 00:32:23.120 --> 00:32:23.960 >> Oh my god. 00:32:23.960 --> 00:32:26.040 That kills me. I hate Discord. I hate 00:32:26.040 --> 00:32:27.640 it. Okay, but 00:32:27.640 --> 00:32:28.720 >> Yeah, so anyway, 00:32:28.720 --> 00:32:29.560 >> I'm I'm going to 00:32:29.560 --> 00:32:32.080 >> back and forth that happened is that I 00:32:32.080 --> 00:32:33.520 asked Hermes to 00:32:33.520 --> 00:32:35.160 comment this and then I went back to 00:32:35.160 --> 00:32:37.800 Fable and I said, "Do you think some of 00:32:37.800 --> 00:32:39.880 this should be in this PR? Is it out of 00:32:39.880 --> 00:32:41.600 scope? Should we create follow-up 00:32:41.600 --> 00:32:42.280 tickets?" 00:32:42.280 --> 00:32:44.080 >> So, you just basically copied and pasted 00:32:44.080 --> 00:32:45.600 some feedback in there, right? 00:32:45.600 --> 00:32:48.440 >> No, I told Hermes that these particular 00:32:48.440 --> 00:32:50.240 things it found, I found that those were 00:32:50.240 --> 00:32:52.640 relevant and asked it to to make the 00:32:52.640 --> 00:32:54.200 comment. So, it actually didn't even 00:32:54.200 --> 00:32:55.680 make the comment because I think I I 00:32:55.680 --> 00:32:57.920 made it very like don't post anything 00:32:57.920 --> 00:32:59.840 because if it posts something stupid 00:32:59.840 --> 00:33:01.680 >> and then Fable is able to suck that 00:33:01.680 --> 00:33:02.200 down. 00:33:02.200 --> 00:33:04.080 >> So, the thing is with Hermes, I don't 00:33:04.080 --> 00:33:07.080 want it to post comments at willy-nilly. 00:33:07.080 --> 00:33:07.480 >> Yeah. 00:33:07.480 --> 00:33:09.000 >> Because sometimes I'm like conversing 00:33:09.000 --> 00:33:11.200 with it and it and it says I 00:33:11.200 --> 00:33:12.680 misunderstands me and it goes and posts 00:33:12.680 --> 00:33:14.760 a comment. As discussed, blah blah blah. 00:33:14.760 --> 00:33:16.400 And it makes no sense because nobody has 00:33:16.400 --> 00:33:19.000 that comment thread. So, I'm like very I 00:33:19.000 --> 00:33:20.520 told it several times don't comment 00:33:20.520 --> 00:33:22.680 until I tell you. And so, it actually 00:33:22.680 --> 00:33:24.640 even created like a a draft review 00:33:24.640 --> 00:33:27.360 comment and and and tells me exactly 00:33:27.360 --> 00:33:29.200 what it's going to post and then I said, 00:33:29.200 --> 00:33:31.480 "Okay, go ahead. Like looks good." 00:33:31.480 --> 00:33:34.760 >> That that that sounds like a good thing. 00:33:34.760 --> 00:33:36.480 >> Yeah, it's it's because you will lose 00:33:36.480 --> 00:33:39.200 trust. Like people filter out. 00:33:39.200 --> 00:33:39.480 >> Yeah. 00:33:39.480 --> 00:33:40.320 >> If it's 00:33:40.320 --> 00:33:41.000 >> I 00:33:41.000 --> 00:33:43.160 I already on my on my work thing there's 00:33:43.160 --> 00:33:44.720 there's some agent that's running 00:33:44.720 --> 00:33:46.680 reviewing things and it creates so much 00:33:46.680 --> 00:33:49.200 noise. I'm like and the trouble is is 00:33:49.200 --> 00:33:52.160 that like I reply to the bloody comment 00:33:52.160 --> 00:33:54.600 and the agent doesn't pick it up. So, 00:33:54.600 --> 00:33:55.120 anyway. 00:33:55.120 --> 00:33:55.720 >> Yeah. 00:33:55.720 --> 00:33:58.120 So, the back and forth is is really 00:33:58.120 --> 00:34:00.760 plain. You can ask either on Claude Code 00:34:00.760 --> 00:34:03.200 in the web or in here, you can ask it to 00:34:03.200 --> 00:34:05.640 monitor the PR and look for comments and 00:34:05.640 --> 00:34:06.400 wake up. 00:34:06.400 --> 00:34:06.920 >> Interesting. 00:34:06.920 --> 00:34:08.320 >> In this case, yeah, you can say 00:34:08.320 --> 00:34:10.240 >> How did you get it to propose it? 00:34:10.240 --> 00:34:11.679 >> It even proposes it. So, if you use 00:34:11.679 --> 00:34:13.360 Claude Code in the web, it will say, 00:34:13.360 --> 00:34:14.679 "I've created a PR. Do you want me to 00:34:14.679 --> 00:34:16.360 watch it?" And then you can say, "Yes, 00:34:16.360 --> 00:34:17.159 go watch it." 00:34:17.159 --> 00:34:19.040 >> Holy I mean, 00:34:19.040 --> 00:34:20.800 that's amazing. 00:34:20.800 --> 00:34:21.919 Though, 00:34:21.919 --> 00:34:24.639 can I debate with you here because you 00:34:24.639 --> 00:34:26.520 you have shown me a cool workflow for 00:34:26.520 --> 00:34:28.560 dealing with a 4,000 whatever line 00:34:28.560 --> 00:34:29.200 change. 00:34:29.200 --> 00:34:30.800 >> Well, I still need to read it. Like I 00:34:30.800 --> 00:34:33.399 can't dump this on other people. 00:34:33.399 --> 00:34:34.600 >> I still 00:34:34.600 --> 00:34:37.080 I would I wouldn't do that. I I I'm 00:34:37.080 --> 00:34:40.520 still in camp tiny change, tiny change. 00:34:40.520 --> 00:34:43.240 >> I I 100% expect somebody to come back 00:34:43.240 --> 00:34:44.960 and say, "Can you break this up?" And 00:34:44.960 --> 00:34:46.840 I'm this close of asking Fable to do the 00:34:46.840 --> 00:34:48.480 same, to say, "Hey, can we break this up 00:34:48.480 --> 00:34:49.879 into smaller PRs?" 00:34:49.879 --> 00:34:51.679 >> sensor because the brilliant thing is 00:34:51.679 --> 00:34:54.639 like I was vibing yesterday on that on 00:34:54.639 --> 00:34:57.320 that tennis membership thing. And it 00:34:57.320 --> 00:35:00.600 made 1,000 2,000 whatever li- lines of 00:35:00.600 --> 00:35:02.840 thing. And then right And then when I 00:35:02.840 --> 00:35:04.600 told it to commit, only then when it 00:35:04.600 --> 00:35:06.880 committed, then it was like working out 00:35:06.880 --> 00:35:09.720 how to sequence it. And then it built a 00:35:09.720 --> 00:35:12.120 really good good history. All because 00:35:12.120 --> 00:35:14.040 all because of my pre-commit hook 00:35:14.040 --> 00:35:16.040 limiting it to like 100 lines of change 00:35:16.040 --> 00:35:18.240 all the time. It was It was fantastic. 00:35:18.240 --> 00:35:19.920 >> Yeah, but the thing is you get a lot of 00:35:19.920 --> 00:35:22.080 overhead in that, right? I mean, it 00:35:22.080 --> 00:35:24.320 again, it's all depends if how many 00:35:24.320 --> 00:35:26.880 people you're working with. 00:35:26.880 --> 00:35:28.600 If I'm working alone on something and I 00:35:28.600 --> 00:35:30.640 don't have to get a PR review, I will 00:35:30.640 --> 00:35:33.040 100% do just get it through. Definitely 00:35:33.040 --> 00:35:35.200 if I'm vibe coding. But but like to be 00:35:35.200 --> 00:35:36.680 honest, in this project where there's a 00:35:36.680 --> 00:35:38.040 lot of other people and they have a lot 00:35:38.040 --> 00:35:40.320 of opinions of how it should be, that's 00:35:40.320 --> 00:35:41.840 where it that's where I'm forced to 00:35:41.840 --> 00:35:42.880 break it things down. 00:35:42.880 --> 00:35:45.520 >> I'm This is like the software 00:35:45.520 --> 00:35:48.600 development life cycle is about if I 00:35:48.600 --> 00:35:50.920 mean, this is the agile approach and PRs 00:35:50.920 --> 00:35:53.240 are always in the bloody way of this. 00:35:53.240 --> 00:35:55.360 Like, I hate Like, I'm I'm working with 00:35:55.360 --> 00:35:57.280 with with a guy on a project right now 00:35:57.280 --> 00:36:00.400 and it's just typical PR ping pong where 00:36:00.400 --> 00:36:02.720 like, "Oh, I made a change." And then 00:36:02.720 --> 00:36:05.360 And then like, you know, a day goes by, 00:36:05.360 --> 00:36:07.240 I review it and I say, "No, it's wrong." 00:36:07.240 --> 00:36:09.320 Then a day goes by and he fixes it. I 00:36:09.320 --> 00:36:12.680 mean, it it doesn't work. I feel it's 00:36:12.680 --> 00:36:14.480 really ineffective to do this in a in a 00:36:14.480 --> 00:36:16.520 in a business. I I know it makes sense 00:36:16.520 --> 00:36:18.520 in open source because it's a low trust 00:36:18.520 --> 00:36:21.080 environment, but but in business it's 00:36:21.080 --> 00:36:22.240 just it's pissing me off. 00:36:22.240 --> 00:36:23.400 >> a whole lot of discussion because 00:36:23.400 --> 00:36:24.640 there's quite a few discussions on 00:36:24.640 --> 00:36:27.440 LinkedIn people say like PRs are broken 00:36:27.440 --> 00:36:29.320 and it took agents for us to realize it, 00:36:29.320 --> 00:36:31.640 how broken they are. But, um but I just 00:36:31.640 --> 00:36:32.760 wanted to highlight here, and then I'll 00:36:32.760 --> 00:36:33.920 stop sharing my screen. 00:36:33.920 --> 00:36:36.160 >> how did agents help us realize that PRs 00:36:36.160 --> 00:36:38.120 are broken? Because you To me, you 00:36:38.120 --> 00:36:40.480 showed me agents loving the whole PR 00:36:40.480 --> 00:36:41.840 process because it gives you an 00:36:41.840 --> 00:36:45.640 opportunity to work your Hermes PR 00:36:45.640 --> 00:36:48.360 review agent, you know. 00:36:48.360 --> 00:36:50.520 It doesn't really show It doesn't show 00:36:50.520 --> 00:36:53.160 show off PRs being broken to me. 00:36:53.160 --> 00:36:54.800 >> I think I think the whole industry is 00:36:54.800 --> 00:36:56.080 questioning Git as a versioning 00:36:56.080 --> 00:36:57.840 mechanism. There's at least three, you 00:36:57.840 --> 00:36:59.520 know, well-funded startups that are 00:36:59.520 --> 00:37:01.600 reinventing Git or reinventing version 00:37:01.600 --> 00:37:02.600 control, 00:37:02.600 --> 00:37:04.040 um that are more that is more agent 00:37:04.040 --> 00:37:04.520 native. 00:37:04.520 --> 00:37:05.280 >> You mean pull requests? 00:37:05.280 --> 00:37:06.760 >> top of that, there was also LinkedIn 00:37:06.760 --> 00:37:09.520 post about trunk-based development and 00:37:09.520 --> 00:37:11.680 the way that PRs like work. I need to 00:37:11.680 --> 00:37:13.080 find it. Let me try and find it. That 00:37:13.080 --> 00:37:16.080 explains the problem. I think it was 00:37:16.080 --> 00:37:17.760 mainly on the review bottleneck that 00:37:17.760 --> 00:37:19.320 they were trying to highlight. And And 00:37:19.320 --> 00:37:20.680 it was also focused on trunk-based, 00:37:20.680 --> 00:37:22.240 yeah. Let me try and find that. 00:37:22.240 --> 00:37:23.840 >> Maybe let's take a little break cuz I 00:37:23.840 --> 00:37:24.560 need a pee. 00:37:24.560 --> 00:37:26.400 >> Actually, yeah. Maybe this is something 00:37:26.400 --> 00:37:28.920 I want to share is while I found the 00:37:28.920 --> 00:37:32.000 Hermes agent really good at finding 00:37:32.000 --> 00:37:33.840 things that I read on LinkedIn. For 00:37:33.840 --> 00:37:36.920 example, I asked ChatGPT about a post I 00:37:36.920 --> 00:37:39.120 read a day ago, and it couldn't find it. 00:37:39.120 --> 00:37:40.400 Like, it came up with generic Google 00:37:40.400 --> 00:37:41.560 results. 00:37:41.560 --> 00:37:43.440 And then I went back to 00:37:43.440 --> 00:37:47.440 Hermes on on Discord, and the agent has 00:37:47.440 --> 00:37:50.560 capabilities to scrape the web on demand 00:37:50.560 --> 00:37:52.840 more than than ChatGPT. So, I don't 00:37:52.840 --> 00:37:55.280 know. So, this was So, this was Yeah, so 00:37:55.280 --> 00:37:57.000 you're back. Okay, so I want What I was 00:37:57.000 --> 00:37:58.520 saying while you were gone is that to 00:37:58.520 --> 00:38:00.480 find back something I read on LinkedIn, 00:38:00.480 --> 00:38:03.160 I found ChatGPT very bad, very poor. I 00:38:03.160 --> 00:38:05.040 tried to find like, "Hey, I read this 00:38:05.040 --> 00:38:07.240 thing. It was about something." And 00:38:07.240 --> 00:38:09.160 ChatGPT can't find it, or it comes up 00:38:09.160 --> 00:38:10.600 with generic Google results or web 00:38:10.600 --> 00:38:13.080 search results. And this is an example 00:38:13.080 --> 00:38:15.360 where I just read the LinkedIn post of a 00:38:15.360 --> 00:38:17.920 end-to-hand hug end-to-hand hugging face 00:38:17.920 --> 00:38:20.360 model provisioning. So, it was a post 00:38:20.360 --> 00:38:22.440 that talked about you can pick any model 00:38:22.440 --> 00:38:24.760 card from hugging face and it will look 00:38:24.760 --> 00:38:26.600 at the GPU requirements. It will then go 00:38:26.600 --> 00:38:29.120 and find the properly sized GPU to rent 00:38:29.120 --> 00:38:30.600 it from like an 00:38:30.600 --> 00:38:32.560 like an a GPU 00:38:32.560 --> 00:38:34.480 inference provider and it would deploy 00:38:34.480 --> 00:38:36.560 the model on top of it and it would land 00:38:36.560 --> 00:38:39.040 you with Claude Code router configuration 00:38:39.040 --> 00:38:40.440 so that you could immediately launch 00:38:40.440 --> 00:38:42.000 your Claude Code using that model on 00:38:42.000 --> 00:38:44.680 rented GPU so that you can basically try 00:38:44.680 --> 00:38:46.920 an open weights model and before you buy 00:38:46.920 --> 00:38:49.280 all the GPU and hardware for it, you can 00:38:49.280 --> 00:38:51.720 experience what is it tokens per second 00:38:51.720 --> 00:38:53.040 is it is it like 00:38:53.040 --> 00:38:56.520 >> With Claude Code not using pie or open 00:38:56.520 --> 00:38:56.640 AI. 00:38:56.640 --> 00:38:58.800 >> Yeah, so so so the post was very cool, 00:38:58.800 --> 00:39:00.880 right? I just wanted to find it back and 00:39:00.880 --> 00:39:02.680 when I asked her GPT it did an absolute 00:39:02.680 --> 00:39:04.960 horrible job about how to take hugging 00:39:04.960 --> 00:39:07.640 face models run GPU and deploy them and 00:39:07.640 --> 00:39:10.080 here I ran out the gateway actually was 00:39:10.080 --> 00:39:11.920 was crashing and then Sony told me how 00:39:11.920 --> 00:39:13.800 to fix it. But then I came back and said 00:39:13.800 --> 00:39:14.200 hey 00:39:14.200 --> 00:39:15.880 >> You're using fire crawl, aren't you? 00:39:15.880 --> 00:39:16.920 >> Yeah, 00:39:16.920 --> 00:39:18.200 and and and the funny thing is it 00:39:18.200 --> 00:39:20.440 actually found him like it found GA 00:39:20.440 --> 00:39:23.480 Ahama who posted this this this thing 00:39:23.480 --> 00:39:25.640 and then when I asked it to use web 00:39:25.640 --> 00:39:27.520 search to to scrape and you whatever 00:39:27.520 --> 00:39:29.720 scraping skills you have to find it and 00:39:29.720 --> 00:39:31.480 then it very quickly found the exact 00:39:31.480 --> 00:39:34.040 post which was then posted a day ago and 00:39:34.040 --> 00:39:36.640 found the actual AI on demand GitHub 00:39:36.640 --> 00:39:38.960 repository and then we went ahead and we 00:39:38.960 --> 00:39:41.000 built we wrote a little blog post about 00:39:41.000 --> 00:39:41.320 it. 00:39:41.320 --> 00:39:43.600 >> fire crawl use your LinkedIn credentials 00:39:43.600 --> 00:39:44.560 to do the scraping? 00:39:44.560 --> 00:39:45.760 >> I think I 00:39:45.760 --> 00:39:48.000 not fire crawl but you can install 00:39:48.000 --> 00:39:50.600 custom tools. I I tried for a week to 00:39:50.600 --> 00:39:53.600 use XAI with an API key and I put $10 in 00:39:53.600 --> 00:39:56.280 it and it was it was it was empty within 00:39:56.280 --> 00:39:58.280 like two or four like within five days. 00:39:58.280 --> 00:40:00.720 So XAI is extremely expensive 00:40:00.720 --> 00:40:04.160 to use AI like API tools to to to scrape 00:40:04.160 --> 00:40:06.120 things. So don't use that. 00:40:06.120 --> 00:40:08.560 >> I bust through like a $10 budget within 00:40:08.560 --> 00:40:11.200 a a week and I was like I can't afford 00:40:11.200 --> 00:40:13.920 this even though XAI web search seems 00:40:13.920 --> 00:40:14.520 awesome. 00:40:14.520 --> 00:40:17.160 >> So so I wanted to I wanted to highlight 00:40:17.160 --> 00:40:18.960 actually if I want to find that LinkedIn 00:40:18.960 --> 00:40:20.760 post that we just discussed, it's better 00:40:20.760 --> 00:40:23.160 for me to ask the Hermes agent, sorry. I 00:40:23.160 --> 00:40:24.760 think Firecrawl with the web search is 00:40:24.760 --> 00:40:26.520 better than ChatGPT. 00:40:26.520 --> 00:40:28.040 >> It's like a a personal chat. 00:40:28.040 --> 00:40:30.160 >> Yeah, but but still I can't have a thing 00:40:30.160 --> 00:40:32.080 you need that most sites are getting 00:40:32.080 --> 00:40:34.680 good at at blocking it. Robot, so you 00:40:34.680 --> 00:40:36.800 have to use your logged-in creden- you 00:40:36.800 --> 00:40:38.680 have to use your your you have to like 00:40:38.680 --> 00:40:42.080 offer Chrome remote debugging port to 00:40:42.080 --> 00:40:44.320 your Firecrawl instance just to get 00:40:44.320 --> 00:40:45.040 anywhere. 00:40:45.040 --> 00:40:47.280 >> Yeah, maybe. I I think one thing I I I 00:40:47.280 --> 00:40:49.040 want to do is I know that somebody built 00:40:49.040 --> 00:40:51.800 a LinkedIn CLI tool for AI agents, so I 00:40:51.800 --> 00:40:53.800 could try to use that, I think. 00:40:53.800 --> 00:40:56.040 From Ion Murdock, he posted it. And 00:40:56.040 --> 00:40:57.320 there's also 00:40:57.320 --> 00:40:59.080 >> That sounds like familiar. It's a guy 00:40:59.080 --> 00:41:00.440 from Singapore, right? 00:41:00.440 --> 00:41:01.040 >> Yeah, yeah. 00:41:01.040 --> 00:41:03.240 >> That's our friend. He used to attend the 00:41:03.240 --> 00:41:04.120 meetings. 00:41:04.120 --> 00:41:06.040 I just had the same experience that you 00:41:06.040 --> 00:41:08.640 did about uh you know, Claude asking me 00:41:08.640 --> 00:41:10.640 a whole bunch of questions, and then it 00:41:10.640 --> 00:41:13.280 sort of no response. This is really 00:41:13.280 --> 00:41:14.480 interesting development. 00:41:14.480 --> 00:41:17.000 >> Yeah, but also I also had a few times, I 00:41:17.000 --> 00:41:19.040 think mostly in Claude Code web uh or 00:41:19.040 --> 00:41:21.480 Claude web, where I say use the ask user 00:41:21.480 --> 00:41:23.920 tool. No, even in in Claude Code on on 00:41:23.920 --> 00:41:25.880 the on the on the CLI, I had a quite a 00:41:25.880 --> 00:41:28.480 few times where it gives me a proposal, 00:41:28.480 --> 00:41:30.680 and I I I can add notes into it, and I 00:41:30.680 --> 00:41:32.680 add some notes, and then it come and I 00:41:32.680 --> 00:41:34.480 press enter and it goes like, "No 00:41:34.480 --> 00:41:36.880 response. Okay, I'll just assume." And 00:41:36.880 --> 00:41:40.080 I'm like, "I did respond." Like, "What" 00:41:40.080 --> 00:41:42.240 And then I I escape like, "I told you." 00:41:42.240 --> 00:41:43.520 Maybe I should double escape and go 00:41:43.520 --> 00:41:44.000 back. 00:41:44.000 --> 00:41:45.880 >> Do you have Claude on your on your 00:41:45.880 --> 00:41:47.560 mobile phone? Because I can't help but 00:41:47.560 --> 00:41:49.160 think it would be cool if it just 00:41:49.160 --> 00:41:50.880 notified you while you're having your 00:41:50.880 --> 00:41:53.040 lunch like, "Okay, these three things, 00:41:53.040 --> 00:41:54.120 yeah. 00:41:54.120 --> 00:41:55.640 Yeah, the second one." 00:41:55.640 --> 00:41:58.520 >> So, what what what I did see happen is 00:41:58.520 --> 00:42:00.240 if you have the Claude app on your on 00:42:00.240 --> 00:42:02.320 your laptop, and you're doing something 00:42:02.320 --> 00:42:04.000 in Claude Code on the web or even on my 00:42:04.000 --> 00:42:06.160 phone, and then that that Claude Code 00:42:06.160 --> 00:42:07.800 instance finishes, you get a 00:42:07.800 --> 00:42:09.600 notification on your laptop saying like 00:42:09.600 --> 00:42:11.480 that Claude Code instance finished, you 00:42:11.480 --> 00:42:13.400 can go it's like it's waiting for your 00:42:13.400 --> 00:42:16.000 response. So, it does from the app go 00:42:16.000 --> 00:42:18.160 back from your like from the web. I'm 00:42:18.160 --> 00:42:20.280 maybe on the phone as well, but maybe I 00:42:20.280 --> 00:42:21.800 don't have notifications. But, there 00:42:21.800 --> 00:42:23.760 were quite a few interesting posts about 00:42:23.760 --> 00:42:25.960 the Claude Code web app and also Cloud 00:42:25.960 --> 00:42:27.800 Code sorry 00:42:27.800 --> 00:42:31.720 about the Cloud 00:42:31.720 --> 00:42:33.280 >> I'm I'm starting to think I need my own 00:42:33.280 --> 00:42:35.720 Max plan. I got I got serious FOMO 00:42:35.720 --> 00:42:37.760 because I can't get Fable because of my 00:42:37.760 --> 00:42:39.440 enterprise stuff. 00:42:39.440 --> 00:42:41.640 >> But, then you have to like find a way to 00:42:41.640 --> 00:42:43.160 like toggle accounts. Like you said you 00:42:43.160 --> 00:42:44.760 were you were telling me that you were 00:42:44.760 --> 00:42:47.120 doing all of that with Docker sandboxes 00:42:47.120 --> 00:42:49.360 to set up like individual sessions with 00:42:49.360 --> 00:42:50.680 different authentications. 00:42:50.680 --> 00:42:53.080 >> Yep. I I I I think I'm I'll have to do 00:42:53.080 --> 00:42:55.040 it I'll have to do it I'll have to do 00:42:55.040 --> 00:42:57.760 it. I got my FOMO's too real. I need 00:42:57.760 --> 00:42:59.400 Fable. 00:42:59.400 --> 00:43:00.880 Okay, I think it's lunchtime here. I 00:43:00.880 --> 00:43:03.160 think we have enough content for for a 00:43:03.160 --> 00:43:05.560 for a podcast. It was a bit haphazard 00:43:05.560 --> 00:43:07.960 cuz I was getting interrupted, but if 00:43:07.960 --> 00:43:10.040 you Yeah, we should maybe plan out what 00:43:10.040 --> 00:43:11.320 we're going to talk about a bit better, 00:43:11.320 --> 00:43:13.280 okay? Cuz I did want to talk about the 00:43:13.280 --> 00:43:16.080 charity majors thing, too. The the 00:43:16.080 --> 00:43:17.680 >> What? The tennis? For the charity No, I 00:43:17.680 --> 00:43:19.000 definitely wanted to talk about that, 00:43:19.000 --> 00:43:20.760 but have you read it? Have you all read 00:43:20.760 --> 00:43:21.000 it? 00:43:21.000 --> 00:43:22.520 >> I I just read the headline. 00:43:22.520 --> 00:43:24.240 >> Okay, so I don't think we can talk about 00:43:24.240 --> 00:43:24.760 it much. 00:43:24.760 --> 00:43:25.080 >> Okay. 00:43:25.080 --> 00:43:26.880 >> I send you the link and I'm like I want 00:43:26.880 --> 00:43:28.080 to read it, but 00:43:28.080 --> 00:43:30.000 >> Okay, we should have we should we should 00:43:30.000 --> 00:43:32.160 have an agenda where we have have things 00:43:32.160 --> 00:43:33.960 and then have we read it and then only 00:43:33.960 --> 00:43:36.840 when when we both have click tick, we 00:43:36.840 --> 00:43:39.440 can get back to it. Okay, I'm going to 00:43:39.440 --> 00:43:41.240 sign off now. Hopefully, Riverside too 00:43:41.240 --> 00:43:43.840 can make sense of this recording and if 00:43:43.840 --> 00:43:46.360 you like the podcast, like it. In fact, 00:43:46.360 --> 00:43:48.800 if you didn't like it, like it. Comment 00:43:48.800 --> 00:43:50.200 why you didn't like it. Yeah, that's a 00:43:50.200 --> 00:43:51.400 good idea. Geez. 00:43:51.400 --> 00:43:53.400 >> You're a genius. Okay, see you. I'm 00:43:53.400 --> 00:43:54.720 going to take some lunch. Bye. 00:43:54.720 --> 00:43:57.200 >> Okay, bye.