WEBVTT 00:00:00.560 --> 00:00:02.560 How does it go? How did it go? Did you 00:00:02.560 --> 00:00:04.319 have a play? Did you do what you said 00:00:04.319 --> 00:00:04.880 you would? 00:00:04.880 --> 00:00:08.400 >> To be honest, I I wanted to I did, but I 00:00:08.400 --> 00:00:11.040 kept on running into issues because the 00:00:11.040 --> 00:00:14.080 project that I had the std spec driven 00:00:14.080 --> 00:00:16.000 development in mind was a mapping 00:00:16.000 --> 00:00:18.320 project which has a ton of dependencies 00:00:18.320 --> 00:00:20.480 in order for it to work with the open 00:00:20.480 --> 00:00:22.080 street map. And I kept on having 00:00:22.080 --> 00:00:23.840 problems with Docker and 00:00:23.840 --> 00:00:25.199 >> yeah because I needed Docker to build 00:00:25.199 --> 00:00:27.199 the sort of dev container with all the 00:00:27.199 --> 00:00:29.920 with all the open street map. 00:00:29.920 --> 00:00:32.800 C++ boost balls. But let's not let's not 00:00:32.800 --> 00:00:35.600 talk about that because 00:00:35.600 --> 00:00:37.120 >> but I think it highlights a very 00:00:37.120 --> 00:00:40.320 important thing, right? Um you you 00:00:40.320 --> 00:00:42.000 speced it out. You had the user stories 00:00:42.000 --> 00:00:43.680 but you got stuck on technical problems 00:00:43.680 --> 00:00:44.399 like the 00:00:44.399 --> 00:00:45.920 >> No, I didn't even I to be honest I 00:00:45.920 --> 00:00:47.520 didn't even get that far. 00:00:47.520 --> 00:00:50.160 >> Okay, I need to revisit that. 00:00:50.160 --> 00:00:54.719 >> But today I was going to conveniently go 00:00:54.719 --> 00:00:57.039 around that way. I hear this echo. 00:00:57.039 --> 00:00:58.640 >> Let's Hold on. Let me just see if I can 00:00:58.640 --> 00:01:00.559 >> probably my my head too loud or 00:01:00.559 --> 00:01:01.199 something. 00:01:01.199 --> 00:01:02.480 >> Oh, echo cancellation. 00:01:02.480 --> 00:01:04.320 >> You mean you hear yourself? Maybe from 00:01:04.320 --> 00:01:06.159 my can back into the mic. 00:01:06.159 --> 00:01:07.200 >> Maybe it's me. 00:01:07.200 --> 00:01:10.000 >> Yeah, the I mean changing the topic cuz 00:01:10.000 --> 00:01:13.680 the spectrum I mean need to I I just 00:01:13.680 --> 00:01:16.640 I've had a few just abortive attempts to 00:01:16.640 --> 00:01:18.080 spectrum development. I just need to 00:01:18.080 --> 00:01:20.000 give it another try. So have you been 00:01:20.000 --> 00:01:22.320 doing it? What what is your status with 00:01:22.320 --> 00:01:23.920 spectriven development before we move 00:01:23.920 --> 00:01:26.320 on? So I've been working on a rewrite an 00:01:26.320 --> 00:01:28.720 open source version of the back end of 00:01:28.720 --> 00:01:31.040 Spec Ledger. So currently Spec Ledger 00:01:31.040 --> 00:01:34.000 public repo is a Golang CLI and the 00:01:34.000 --> 00:01:37.360 actual web app is a Versel hosted Nex.js 00:01:37.360 --> 00:01:40.960 app with serverside rendering and I am 00:01:40.960 --> 00:01:43.759 not a fan of that. Um I also prefer 00:01:43.759 --> 00:01:46.240 monor repos and I don't mind. I think it 00:01:46.240 --> 00:01:47.600 will be more interesting if the back end 00:01:47.600 --> 00:01:50.399 is open source directly within the CLI 00:01:50.399 --> 00:01:53.759 repository. So, I've been specking out a 00:01:53.759 --> 00:01:56.880 new version of of the spec ledger with 00:01:56.880 --> 00:01:59.280 organization support like multi-tenant 00:01:59.280 --> 00:02:01.840 from the get- go. So, you can organize 00:02:01.840 --> 00:02:03.759 projects per organization and then you 00:02:03.759 --> 00:02:06.000 have your private workflows of if you 00:02:06.000 --> 00:02:07.680 want to do openspec or if you want to do 00:02:07.680 --> 00:02:09.280 specit or if you want to do Google 00:02:09.280 --> 00:02:11.039 conductor 00:02:11.039 --> 00:02:12.800 >> then you would should be able to choose 00:02:12.800 --> 00:02:14.800 internally with like or you can build 00:02:14.800 --> 00:02:16.400 your own version of your of your 00:02:16.400 --> 00:02:18.239 templates. Basically what I want is I 00:02:18.239 --> 00:02:21.360 don't want the prompts the um the 00:02:21.360 --> 00:02:23.120 workflow to be embedded within the 00:02:23.120 --> 00:02:26.319 Golang CLI. I want the Golang CLI to 00:02:26.319 --> 00:02:28.640 like you know you authenticate and you 00:02:28.640 --> 00:02:31.440 pull it from your organization or maybe 00:02:31.440 --> 00:02:32.959 maybe there is the first version is is 00:02:32.959 --> 00:02:35.200 is is available so when you 00:02:35.200 --> 00:02:37.200 >> organization organization why do you say 00:02:37.200 --> 00:02:39.920 organization is that just to set set the 00:02:39.920 --> 00:02:41.599 context or something? Yeah, because I 00:02:41.599 --> 00:02:43.760 believe the main purpose of or the main 00:02:43.760 --> 00:02:45.200 problems that I'm trying to solve is 00:02:45.200 --> 00:02:49.200 stakeholder buyin and sharing access to 00:02:49.200 --> 00:02:51.360 the artifacts that need to be reviewed 00:02:51.360 --> 00:02:54.160 and a review environment. And then the 00:02:54.160 --> 00:02:57.440 second thing I want to solve is keeping 00:02:57.440 --> 00:02:59.440 your specdriven development projects 00:02:59.440 --> 00:03:02.080 aligned in terms of workflows so that 00:03:02.080 --> 00:03:04.480 you can version and pull the workflows. 00:03:04.480 --> 00:03:06.319 And then because specdriven development 00:03:06.319 --> 00:03:08.000 workflows are heavy-handed, you should 00:03:08.000 --> 00:03:10.239 be able to customize your workflows to 00:03:10.239 --> 00:03:12.879 have maybe less steps for your for maybe 00:03:12.879 --> 00:03:15.440 the data team doesn't want to do certain 00:03:15.440 --> 00:03:17.360 steps of it. So you could customize 00:03:17.360 --> 00:03:19.040 basically right now on Reddit I see 00:03:19.040 --> 00:03:21.360 quite a lot of posts about we had the 00:03:21.360 --> 00:03:22.959 problem of distributing skills within 00:03:22.959 --> 00:03:25.040 our organization so we build an internal 00:03:25.040 --> 00:03:27.599 skills registry which I think is already 00:03:27.599 --> 00:03:28.159 hold on 00:03:28.159 --> 00:03:32.080 >> does mpx skills.sh SH support private 00:03:32.080 --> 00:03:35.040 ones I wonder cuz that's the way I I use 00:03:35.040 --> 00:03:36.879 to distribute my own skills. 00:03:36.879 --> 00:03:39.760 >> Yeah, so Versel is very simple the NPX 00:03:39.760 --> 00:03:42.159 skills from Versel. I actually ported 00:03:42.159 --> 00:03:44.080 that TypeScript source code because it's 00:03:44.080 --> 00:03:47.440 open source into Golang specure CLI. So 00:03:47.440 --> 00:03:49.280 um so basically because I I don't want 00:03:49.280 --> 00:03:51.280 you to have to run multiple right beats 00:03:51.280 --> 00:03:54.000 is inside the Golang CLI. the skills is 00:03:54.000 --> 00:03:56.560 inside the Golang CLIs manage. So you 00:03:56.560 --> 00:03:57.360 just download 00:03:57.360 --> 00:04:00.239 >> building your own bundle package of AI 00:04:00.239 --> 00:04:01.200 tooling or something. 00:04:01.200 --> 00:04:02.560 >> Yeah, I mean you could argue that you 00:04:02.560 --> 00:04:04.319 take the best tool for the job. But I I 00:04:04.319 --> 00:04:06.000 think all of this is related like by 00:04:06.000 --> 00:04:08.879 adding the skill search and skill 00:04:08.879 --> 00:04:10.480 install capabilities directly to the 00:04:10.480 --> 00:04:13.439 Golang CLI of the spec ledger. You can 00:04:13.439 --> 00:04:15.680 also trigger it by the agent. So when 00:04:15.680 --> 00:04:17.680 the agent goes down to the plan phase 00:04:17.680 --> 00:04:19.759 and identifies a certain text tech like 00:04:19.759 --> 00:04:21.440 hey you're building a react it then 00:04:21.440 --> 00:04:23.600 actually goes and searches using the SL 00:04:23.600 --> 00:04:26.160 skills any react skills that we can use 00:04:26.160 --> 00:04:28.000 and then it proposes like I found these 00:04:28.000 --> 00:04:30.639 skills and the versel uh registry has 00:04:30.639 --> 00:04:33.440 metadata from snik about vulnerability. 00:04:33.440 --> 00:04:35.759 So the agent does that 00:04:35.759 --> 00:04:37.360 >> so it sounds like every organ 00:04:37.360 --> 00:04:40.880 organization needs to heavily customize 00:04:40.880 --> 00:04:43.360 this spec ledger workflow or how do you 00:04:43.360 --> 00:04:44.160 describe it? Yeah. 00:04:44.160 --> 00:04:46.400 >> Yeah. Because I think skills are like I 00:04:46.400 --> 00:04:48.000 don't know my experience with skills is 00:04:48.000 --> 00:04:49.280 that they have some skills I cannot 00:04:49.280 --> 00:04:50.800 share because they have some hard like 00:04:50.800 --> 00:04:53.280 some information that is private. So yes 00:04:53.280 --> 00:04:56.160 you can use the versel skills dot you 00:04:56.160 --> 00:04:57.840 know skills registry which is actually 00:04:57.840 --> 00:04:59.680 it's interesting because after porting 00:04:59.680 --> 00:05:01.600 it from Typescript over in Golang cloud 00:05:01.600 --> 00:05:03.040 told me a couple of interesting things. 00:05:03.040 --> 00:05:05.040 Apparently the way that skills are 00:05:05.040 --> 00:05:07.360 indexed by that tool is if you use it to 00:05:07.360 --> 00:05:08.479 install a skill. 00:05:08.479 --> 00:05:09.680 >> Yeah. Then it comes up on the 00:05:09.680 --> 00:05:10.240 leaderboard. 00:05:10.240 --> 00:05:11.840 >> And then it comes up on the leaderboard. 00:05:11.840 --> 00:05:14.639 It's through metadata to usage metadata. 00:05:14.639 --> 00:05:17.199 So, so my Golang CLI is also sending 00:05:17.199 --> 00:05:18.960 usage like every time you run the specs 00:05:18.960 --> 00:05:21.360 ledger CLI to install a skill, it's 00:05:21.360 --> 00:05:24.000 sending the same metadata like this 00:05:24.000 --> 00:05:26.320 skill was installed. Yeah. And then the 00:05:26.320 --> 00:05:28.000 second part is the skills live inside 00:05:28.000 --> 00:05:29.199 GitHub, right? There nothing else. 00:05:29.199 --> 00:05:31.199 There's no there's no no other registry. 00:05:31.199 --> 00:05:32.880 It's just simple. A skill is referenced 00:05:32.880 --> 00:05:35.039 by uh a git repository. There's a 00:05:35.039 --> 00:05:37.520 shortcut for the Versel skills which are 00:05:37.520 --> 00:05:38.800 served directly from the registry 00:05:38.800 --> 00:05:41.120 endpoint. So for every or pretty much 00:05:41.120 --> 00:05:44.720 every other skill, the npx skills CLI 00:05:44.720 --> 00:05:46.880 actually does a get clone of the repo 00:05:46.880 --> 00:05:49.360 and and then my Golang CLI is like 00:05:49.360 --> 00:05:51.600 trying to use GitHub API, but then there 00:05:51.600 --> 00:05:53.520 was a bug because it wasn't like getting 00:05:53.520 --> 00:05:55.440 all of the references like we get get 00:05:55.440 --> 00:05:56.240 the main 00:05:56.240 --> 00:05:58.639 >> API is a nice way to get some private 00:05:58.639 --> 00:05:59.600 stuff. 00:05:59.600 --> 00:06:01.360 >> Yeah. And it would be that's my goal 00:06:01.360 --> 00:06:04.080 ultimately. My goal is SL CLI, you know, 00:06:04.080 --> 00:06:06.560 and as part of an organization, you have 00:06:06.560 --> 00:06:08.240 your own skills that you can't make 00:06:08.240 --> 00:06:09.840 public and that you don't want the users 00:06:09.840 --> 00:06:12.560 data to be registered. So, so then you 00:06:12.560 --> 00:06:14.240 you should be able within teams to run 00:06:14.240 --> 00:06:16.880 the SL binary and then uh hide or or 00:06:16.880 --> 00:06:18.639 just have private skills as well on top 00:06:18.639 --> 00:06:20.479 of the Versel registry hosted skills. 00:06:20.479 --> 00:06:21.280 >> Let me just 00:06:21.280 --> 00:06:22.400 >> But the same with agents. 00:06:22.400 --> 00:06:25.039 >> I'm trying to find um how to share 00:06:25.039 --> 00:06:26.639 something. Where's the chat? There's 00:06:26.639 --> 00:06:28.080 chat here somewhere. Oh, yeah, it's on 00:06:28.080 --> 00:06:30.319 the right on the side. So this this this 00:06:30.319 --> 00:06:32.479 rewrite of spec ledger is really a big 00:06:32.479 --> 00:06:34.080 like a thing that I wanted because the 00:06:34.080 --> 00:06:36.000 way I envisioned it, it always had like 00:06:36.000 --> 00:06:37.520 the ability to create your own workflow 00:06:37.520 --> 00:06:38.880 within your organization and share those 00:06:38.880 --> 00:06:40.639 across repositories and that was never 00:06:40.639 --> 00:06:41.840 possible because of the way it was 00:06:41.840 --> 00:06:43.600 implemented. So I'm really excited about 00:06:43.600 --> 00:06:47.039 this rewrite but um I haven't been able 00:06:47.039 --> 00:06:50.240 to actually you know complete it. 00:06:50.240 --> 00:06:52.319 >> You can see what I'm writing here. 00:06:52.319 --> 00:06:53.840 >> Yeah, it's a big box with spec ledger. 00:06:53.840 --> 00:06:55.280 Maybe you want to zoom in because when I 00:06:55.280 --> 00:06:57.759 watch these um recordings on YouTube the 00:06:57.759 --> 00:06:59.520 it's always super small. I usually watch 00:06:59.520 --> 00:07:00.160 from mobile. 00:07:00.160 --> 00:07:03.759 >> Okay. So, um I I love how like in the 00:07:03.759 --> 00:07:06.080 last one I'm we we called this this 00:07:06.080 --> 00:07:07.840 whole thing 00:07:07.840 --> 00:07:11.360 um you know this whole STD thing a 00:07:11.360 --> 00:07:13.759 planning tool uh you know driving 00:07:13.759 --> 00:07:16.960 alignment and now this thing like you're 00:07:16.960 --> 00:07:21.120 doing a package a bundle of of necessary 00:07:21.120 --> 00:07:24.479 AI tooling or something is the way 00:07:24.479 --> 00:07:26.639 thinking I'm processing what you just 00:07:26.639 --> 00:07:29.919 said to me. Uh yeah it's more like um 00:07:29.919 --> 00:07:32.080 boilerplate management. So if you want 00:07:32.080 --> 00:07:34.800 to bootstrap a repository for a certain 00:07:34.800 --> 00:07:37.440 type of development it will you know 00:07:37.440 --> 00:07:40.479 with SL init you will get the full spec 00:07:40.479 --> 00:07:42.240 like spec kit workflow which is the same 00:07:42.240 --> 00:07:43.599 spec ledger workflow 00:07:43.599 --> 00:07:45.440 >> and [snorts] then every prompt is 00:07:45.440 --> 00:07:47.599 telling the agent to hey if you run the 00:07:47.599 --> 00:07:51.120 SL binary you can find open comments 00:07:51.120 --> 00:07:52.880 that people has left on on these 00:07:52.880 --> 00:07:56.240 documents. you can find um issues that 00:07:56.240 --> 00:07:57.840 have been created and and task 00:07:57.840 --> 00:07:59.680 dependency graph and what is open what 00:07:59.680 --> 00:08:01.360 is available what's the definition of 00:08:01.360 --> 00:08:03.039 done has this been completed 00:08:03.039 --> 00:08:05.520 >> oh so do what are you using for like a 00:08:05.520 --> 00:08:06.560 beads visualizer 00:08:06.560 --> 00:08:09.280 >> no visual I yes sorry so when you push 00:08:09.280 --> 00:08:11.280 git there's a web hook that parses the 00:08:11.280 --> 00:08:12.800 JSNL 00:08:12.800 --> 00:08:15.120 >> uh and then visualizes into a conbon 00:08:15.120 --> 00:08:18.080 view and um a tree view in in the web 00:08:18.080 --> 00:08:19.039 app so 00:08:19.039 --> 00:08:21.120 >> Oh so so that's part of your your web 00:08:21.120 --> 00:08:21.440 app 00:08:21.440 --> 00:08:23.440 >> yeah that's the back end so the back can 00:08:23.440 --> 00:08:26.319 is the web app is is really so I I did 00:08:26.319 --> 00:08:29.039 um a little like brain 00:08:29.039 --> 00:08:31.840 storm with um I think it was with CHPT 00:08:31.840 --> 00:08:35.599 to identify if um Spec Ledger should 00:08:35.599 --> 00:08:40.080 really be a plug-in for backstage or a 00:08:40.080 --> 00:08:41.760 collection of plugins for backstage 00:08:41.760 --> 00:08:44.399 because I felt like backstage has is an 00:08:44.399 --> 00:08:46.000 internal developer platform that gives 00:08:46.000 --> 00:08:48.240 you a single pane to discover you know. 00:08:48.240 --> 00:08:51.279 >> Yeah, I'm not I'm not fan. Yes, I know a 00:08:51.279 --> 00:08:52.720 lot of people hate backstage because of 00:08:52.720 --> 00:08:54.640 past experience because it was hard to 00:08:54.640 --> 00:08:56.160 customize because it required you to 00:08:56.160 --> 00:08:58.000 write TypeScript plugins. I think a lot 00:08:58.000 --> 00:08:59.760 of those arguments and I'm very excited 00:08:59.760 --> 00:09:02.720 to challenge that opinion of me are gone 00:09:02.720 --> 00:09:05.760 with AI because now asking and also 00:09:05.760 --> 00:09:07.600 backstage has repositioned itself more 00:09:07.600 --> 00:09:10.560 as like a context delivery or like AI 00:09:10.560 --> 00:09:13.200 native tool and like remote MCP. So a 00:09:13.200 --> 00:09:14.560 friend told me like I think we discussed 00:09:14.560 --> 00:09:16.720 this last time. He said that having a 00:09:16.720 --> 00:09:18.640 remote MCP server unlike having all 00:09:18.640 --> 00:09:20.399 these skills inside repositories and 00:09:20.399 --> 00:09:22.080 distributed across machines is much 00:09:22.080 --> 00:09:24.320 harder to keep aligned than having a 00:09:24.320 --> 00:09:26.959 remote MCP. And a remote MCP is always 00:09:26.959 --> 00:09:28.240 up to date to the version that's 00:09:28.240 --> 00:09:30.880 deployed for everyone. So that made me 00:09:30.880 --> 00:09:33.920 think like actually um having this like 00:09:33.920 --> 00:09:36.399 this IDP as a context provider. 00:09:36.399 --> 00:09:38.000 Therefore specledger should really just 00:09:38.000 --> 00:09:41.200 be like a small AI workflow plugin to 00:09:41.200 --> 00:09:42.959 it. But at the same time, I don't want 00:09:42.959 --> 00:09:46.320 Spec Ledger to pre-require um uh 00:09:46.320 --> 00:09:47.760 backstage obviously because a lot of 00:09:47.760 --> 00:09:49.120 people don't want to run it. So that's 00:09:49.120 --> 00:09:50.800 why I was like, is there a way for me to 00:09:50.800 --> 00:09:53.120 build specger as a bunch of plugins and 00:09:53.120 --> 00:09:57.600 then mock the whole of backstage as just 00:09:57.600 --> 00:09:59.040 um you know, if you don't want 00:09:59.040 --> 00:10:01.040 backstage, you can just have like um a 00:10:01.040 --> 00:10:03.200 virtual one that just runs spec ledger 00:10:03.200 --> 00:10:06.000 only. Um but it looked like it's really 00:10:06.000 --> 00:10:07.760 not a good idea to do that. But it also 00:10:07.760 --> 00:10:09.519 surfaced very interesting stuff. So, 00:10:09.519 --> 00:10:12.240 backstage does have some document 00:10:12.240 --> 00:10:14.880 plug-in that basically uses MK docs to 00:10:14.880 --> 00:10:18.160 visualize the um the data and and I I 00:10:18.160 --> 00:10:20.079 was basically telling TGPD I feel like 00:10:20.079 --> 00:10:23.760 specledger is just um a like markdown 00:10:23.760 --> 00:10:25.760 render like a web app that renders 00:10:25.760 --> 00:10:28.000 markdown out of a git repository, right? 00:10:28.000 --> 00:10:30.160 But it's not really um because Specled 00:10:30.160 --> 00:10:31.760 Ledger gives you metadata about the 00:10:31.760 --> 00:10:34.079 ability to to leave comments. It also 00:10:34.079 --> 00:10:36.000 visualizes specifically the workflow 00:10:36.000 --> 00:10:37.760 because that's a common problem with 00:10:37.760 --> 00:10:39.360 specd driven development is like I've 00:10:39.360 --> 00:10:40.640 done this what's the next step I 00:10:40.640 --> 00:10:42.399 remember our early discussions every 00:10:42.399 --> 00:10:44.240 single time was like what's the next 00:10:44.240 --> 00:10:46.399 step so so spec ledger visualizes that 00:10:46.399 --> 00:10:49.920 process so it's it's a very customized 00:10:49.920 --> 00:10:52.320 um for the spec spec development 00:10:52.320 --> 00:10:53.839 >> and then how does GitHub fit in with 00:10:53.839 --> 00:10:54.880 this 00:10:54.880 --> 00:10:56.000 >> GitHub right 00:10:56.000 --> 00:10:59.360 >> is git is the source of truth right but 00:10:59.360 --> 00:11:01.440 you're assuming GitHub is there and 00:11:01.440 --> 00:11:04.320 going to be running like your CI/CD and 00:11:04.320 --> 00:11:06.079 the rest of it, right? I mean, just to 00:11:06.079 --> 00:11:07.920 build a complete picture here, 00:11:07.920 --> 00:11:10.320 >> um, I mean, Git is a source of truth. 00:11:10.320 --> 00:11:12.000 Specled Ledger just generates documents 00:11:12.000 --> 00:11:14.800 that are version controlled in git and 00:11:14.800 --> 00:11:17.680 the spec spec web app is currently only 00:11:17.680 --> 00:11:20.320 built for GitHub web hooks, but it could 00:11:20.320 --> 00:11:22.720 support any git version control system. 00:11:22.720 --> 00:11:24.959 Uh, and just visualizes the contents of 00:11:24.959 --> 00:11:26.959 your git repository also across 00:11:26.959 --> 00:11:29.360 repositories, right? um as an 00:11:29.360 --> 00:11:31.279 organization then you should be able to 00:11:31.279 --> 00:11:32.800 like one one other feature that specled 00:11:32.800 --> 00:11:35.680 ledger I feel must have is the ability 00:11:35.680 --> 00:11:39.040 to search like is there any like have we 00:11:39.040 --> 00:11:40.959 if this uses story somewhere is there 00:11:40.959 --> 00:11:42.959 like like you know to index and find and 00:11:42.959 --> 00:11:44.640 source like what what you get cuz 00:11:44.640 --> 00:11:46.000 basically it's kind of like a knowledge 00:11:46.000 --> 00:11:48.079 base like what do you need if you if 00:11:48.079 --> 00:11:50.800 you've replaced confluence with kit you 00:11:50.800 --> 00:11:52.720 need still the ability to see the 00:11:52.720 --> 00:11:54.720 documents in a nice uh printable view 00:11:54.720 --> 00:11:57.519 mobile so it's possible to you know have 00:11:57.519 --> 00:12:00.160 a mobile version to to easily review and 00:12:00.160 --> 00:12:02.160 look at documents. Um that's really bad 00:12:02.160 --> 00:12:03.600 on the GitHub interface because that's 00:12:03.600 --> 00:12:05.839 not what it it's written for, right? And 00:12:05.839 --> 00:12:07.519 uh the search functionality, right, with 00:12:07.519 --> 00:12:09.120 Confluence, you can very easily go and 00:12:09.120 --> 00:12:11.600 find related documents and things like 00:12:11.600 --> 00:12:14.399 that as a KB. I mean, in the end, it's a 00:12:14.399 --> 00:12:17.120 bit like Obsidian then, right? 00:12:17.120 --> 00:12:18.639 Creating my gosh, I'm really messing 00:12:18.639 --> 00:12:21.680 this up. Okay, so I think I I I 00:12:21.680 --> 00:12:23.440 understand where you're going here. I I 00:12:23.440 --> 00:12:26.880 think all this tooling 00:12:26.880 --> 00:12:31.920 um has its place and I think it's kind 00:12:31.920 --> 00:12:34.000 of interesting. But I just I wanted to 00:12:34.000 --> 00:12:37.279 just contrast this with with my reality. 00:12:37.279 --> 00:12:40.880 And my reality is like 00:12:40.880 --> 00:12:42.399 um 00:12:42.399 --> 00:12:45.519 I'm usually like in a in a in a team. Oh 00:12:45.519 --> 00:12:47.279 my god, what happened to my you know I'm 00:12:47.279 --> 00:12:49.839 usually in a team. I mean this is great 00:12:49.839 --> 00:12:52.560 for for alignment, right? But I'm I'm in 00:12:52.560 --> 00:12:54.320 a team with like a couple of other 00:12:54.320 --> 00:12:58.480 people. We're all generally using um but 00:12:58.480 --> 00:13:00.160 that's you know something like claw code 00:13:00.160 --> 00:13:01.279 or Codex or whatever 00:13:01.279 --> 00:13:02.079 >> or open code 00:13:02.079 --> 00:13:04.320 >> or even open code. Yeah. Well, we all 00:13:04.320 --> 00:13:06.160 Okay, let's let's see if I can draw a 00:13:06.160 --> 00:13:07.279 picture that's a bit more 00:13:07.279 --> 00:13:08.000 representative. 00:13:08.000 --> 00:13:09.839 >> There's a massive CDK terrain release 00:13:09.839 --> 00:13:11.519 coming by the way. Lots of fixes. 00:13:11.519 --> 00:13:13.600 >> Nice. like we we all have like a our 00:13:13.600 --> 00:13:18.480 individual uh uh 00:13:18.480 --> 00:13:21.600 sessions whatever and um and we all I 00:13:21.600 --> 00:13:23.200 don't know working on different sort of 00:13:23.200 --> 00:13:26.160 things 00:13:26.160 --> 00:13:29.360 I I guess I guess in reality this in a 00:13:29.360 --> 00:13:31.519 proper software project this turns out 00:13:31.519 --> 00:13:33.600 to be uh PRs I suppose 00:13:33.600 --> 00:13:35.920 >> yeah which is also I mean I mean as 00:13:35.920 --> 00:13:37.680 you're drawing arrows which is also a 00:13:37.680 --> 00:13:40.079 feature that specledger built which is a 00:13:40.079 --> 00:13:43.760 a session like um indexing engine, 00:13:43.760 --> 00:13:45.920 right? Every every time you do a commit 00:13:45.920 --> 00:13:49.040 spec ledger CLI is um triggered as a 00:13:49.040 --> 00:13:50.480 hook. And to be honest, right now that's 00:13:50.480 --> 00:13:51.839 not very configurable. It's a bit like 00:13:51.839 --> 00:13:53.600 beats. It's it's like a virus and it 00:13:53.600 --> 00:13:55.440 stalls themselves. But basically, when 00:13:55.440 --> 00:13:57.760 it detects a get commit, it will capture 00:13:57.760 --> 00:14:01.600 the human um conversation of what led to 00:14:01.600 --> 00:14:04.079 that commit inside. Yeah, it does. 00:14:04.079 --> 00:14:06.399 >> Okay, this is what I was looking for. A 00:14:06.399 --> 00:14:07.839 a get commit. 00:14:07.839 --> 00:14:09.600 >> That's how it has been implemented. I 00:14:09.600 --> 00:14:11.199 don't know like I let somebody else 00:14:11.199 --> 00:14:12.800 decide what's the best integration 00:14:12.800 --> 00:14:13.519 pattern. 00:14:13.519 --> 00:14:16.079 >> So it grabs the the transcript or 00:14:16.079 --> 00:14:18.560 >> it grabs the delta. So every like every 00:14:18.560 --> 00:14:22.240 time um it does it uses like a 00:14:22.240 --> 00:14:25.199 sophisticated regular expression 00:14:25.199 --> 00:14:29.519 uh and it it identifies um which context 00:14:29.519 --> 00:14:31.360 that you're on like for example are we 00:14:31.360 --> 00:14:33.360 working on this spec or that spec which 00:14:33.360 --> 00:14:36.240 branch is he on um and so on. And then 00:14:36.240 --> 00:14:38.880 it will capture all of the conversation 00:14:38.880 --> 00:14:41.760 since the last capture point uh and then 00:14:41.760 --> 00:14:43.519 put it inside superbase. And then the 00:14:43.519 --> 00:14:45.600 idea is then to surface it, right? To be 00:14:45.600 --> 00:14:47.519 able to see this comet was really good 00:14:47.519 --> 00:14:49.680 or this like this whole feature went 00:14:49.680 --> 00:14:51.680 wrong. So every single like ultimately 00:14:51.680 --> 00:14:54.320 all of the messages that led for one 00:14:54.320 --> 00:14:56.480 feature spec are all tagged by the 00:14:56.480 --> 00:14:58.000 feature spec and we can break it down by 00:14:58.000 --> 00:15:00.000 commit to see how did the user come up 00:15:00.000 --> 00:15:03.199 to that commit. Um but that part of 00:15:03.199 --> 00:15:05.360 actually mining the data displaying the 00:15:05.360 --> 00:15:07.839 data is not implemented. 00:15:07.839 --> 00:15:10.160 >> So we do have like data being in in 00:15:10.160 --> 00:15:11.839 superbase. I have never looked at the 00:15:11.839 --> 00:15:13.839 table. Somebody else implemented it. I'm 00:15:13.839 --> 00:15:15.199 not sure if this is the best way like 00:15:15.199 --> 00:15:16.720 one of the approaches we thought was 00:15:16.720 --> 00:15:19.600 like why don't we just you know just 00:15:19.600 --> 00:15:22.480 drop the whole session in inside S3 or 00:15:22.480 --> 00:15:23.920 something like that like obviously with 00:15:23.920 --> 00:15:27.120 the consent. Um yeah, but um yeah, 00:15:27.120 --> 00:15:28.399 that's not really there, but that's 00:15:28.399 --> 00:15:29.440 definitely one of the things that we 00:15:29.440 --> 00:15:30.160 want 00:15:30.160 --> 00:15:31.279 >> that we want to expand on. 00:15:31.279 --> 00:15:32.959 >> All right, cool. Cool. Cuz like I 00:15:32.959 --> 00:15:34.800 there's a few things that that that come 00:15:34.800 --> 00:15:35.120 to mind. 00:15:35.120 --> 00:15:36.560 >> I keep saying the same thing because I 00:15:36.560 --> 00:15:37.680 mean I also keep saying the same thing, 00:15:37.680 --> 00:15:39.519 but you shared Simon Willis like 00:15:39.519 --> 00:15:41.680 visualization of how you capture your 00:15:41.680 --> 00:15:43.600 sessions and put them into an HTML 00:15:43.600 --> 00:15:44.240 format. 00:15:44.240 --> 00:15:46.160 >> Yeah. I'm just revisiting it. I'm just 00:15:46.160 --> 00:15:48.480 revisiting it because there's still some 00:15:48.480 --> 00:15:52.959 questions in my mind like um 00:15:52.959 --> 00:15:57.279 like the PRs like like like guard rails 00:15:57.279 --> 00:16:00.000 here like like like things I think are 00:16:00.000 --> 00:16:02.399 quite sensible to do like like how how 00:16:02.399 --> 00:16:05.199 would you like limit the size of the PR 00:16:05.199 --> 00:16:06.880 because I mean I I mean I'm guilty of 00:16:06.880 --> 00:16:09.279 doing this myself when I'm I'm and I'm 00:16:09.279 --> 00:16:11.040 in a bit of a session with Claude or 00:16:11.040 --> 00:16:14.399 Codex and my PR that I just created is 00:16:14.399 --> 00:16:15.920 huge like wouldn't it make sense? 00:16:15.920 --> 00:16:18.480 >> I mean this is a common problem with AI 00:16:18.480 --> 00:16:20.399 development, right? That that it's very 00:16:20.399 --> 00:16:23.920 hard for humans to review. I I honestly 00:16:23.920 --> 00:16:25.839 think with 00:16:25.839 --> 00:16:28.079 that we have to trust our CI/CD and our 00:16:28.079 --> 00:16:30.000 tests more [laughter] that we cannot 00:16:30.000 --> 00:16:33.440 review the code. Um but I also I have 00:16:33.440 --> 00:16:35.360 two ways of working with with uh with 00:16:35.360 --> 00:16:37.360 the code, right? 00:16:37.360 --> 00:16:39.360 Um 00:16:39.360 --> 00:16:42.880 I have smaller PRs when I tend to use um 00:16:42.880 --> 00:16:45.440 cloud plan mode and I do that when I and 00:16:45.440 --> 00:16:47.360 so no spec driven development when I'm 00:16:47.360 --> 00:16:49.199 working on like an issue that was 00:16:49.199 --> 00:16:52.160 created. So I have two ways two work um 00:16:52.160 --> 00:16:54.000 streams. One is specd driven development 00:16:54.000 --> 00:16:56.480 creates the issue list next issue ready 00:16:56.480 --> 00:16:58.079 you know based on the spec and the user 00:16:58.079 --> 00:16:59.680 stories and everything that we agree the 00:16:59.680 --> 00:17:01.279 humans involvement is the review of the 00:17:01.279 --> 00:17:03.920 user stories and the validation that the 00:17:03.920 --> 00:17:05.679 end to end thing work and the user story 00:17:05.679 --> 00:17:08.160 is is implemented as is exe executable 00:17:08.160 --> 00:17:10.000 specs should be the ideal proof that it 00:17:10.000 --> 00:17:11.520 works we don't care about the code 00:17:11.520 --> 00:17:13.199 >> okay I don't think I've this before but 00:17:13.199 --> 00:17:15.520 that makes a lot of sense the human okay 00:17:15.520 --> 00:17:18.240 let me just let me just pen that before 00:17:18.240 --> 00:17:21.039 I lose it 00:17:21.039 --> 00:17:24.160 the human does the um I'm I send you the 00:17:24.160 --> 00:17:27.600 link in the chat by the way. Um a story 00:17:27.600 --> 00:17:30.160 review. So that comes I I'm not too sure 00:17:30.160 --> 00:17:32.480 what the the whole cycle looks like. And 00:17:32.480 --> 00:17:36.480 then I'm looking for a I don't know an 00:17:36.480 --> 00:17:40.720 AI. The AI does does the um or the agent 00:17:40.720 --> 00:17:43.840 does the the the code review is what 00:17:43.840 --> 00:17:44.960 you're saying, right? 00:17:44.960 --> 00:17:47.120 >> Uh 00:17:47.120 --> 00:17:50.799 code review. I mean awesome. I think 00:17:50.799 --> 00:17:52.240 right now people are going back and 00:17:52.240 --> 00:17:54.240 forth this a lot between because some 00:17:54.240 --> 00:17:55.919 people like I mean back in January a lot 00:17:55.919 --> 00:17:57.360 of people step up and says I'm an 00:17:57.360 --> 00:17:59.440 engineer that have like senior engineer 00:17:59.440 --> 00:18:01.280 lots of experience I don't look at my 00:18:01.280 --> 00:18:03.039 code anymore right lots of people said 00:18:03.039 --> 00:18:05.200 that back in January I think people now 00:18:05.200 --> 00:18:06.880 in May are coming back from that a 00:18:06.880 --> 00:18:08.640 little bit and saying yeah you know we 00:18:08.640 --> 00:18:10.480 did that we we didn't look at our code 00:18:10.480 --> 00:18:12.480 it wasn't really good we found like so 00:18:12.480 --> 00:18:15.760 much duplicated functions so much uh 00:18:15.760 --> 00:18:18.320 >> I don't know like running situations 00:18:18.320 --> 00:18:20.640 >> so you were just telling me I mean sorry 00:18:20.640 --> 00:18:22.720 I was maybe half listening because I was 00:18:22.720 --> 00:18:24.000 trying to 00:18:24.000 --> 00:18:25.679 >> cuz I was telling you that the human's 00:18:25.679 --> 00:18:27.760 involvement is purely and I don't user 00:18:27.760 --> 00:18:29.760 stories yeah on the user stories and 00:18:29.760 --> 00:18:32.400 then validating that like I mean you 00:18:32.400 --> 00:18:33.520 have you should have automated 00:18:33.520 --> 00:18:36.400 validation like we talked and endlessly 00:18:36.400 --> 00:18:39.120 about that um how do you really have 00:18:39.120 --> 00:18:41.200 fully automated tests and the ability to 00:18:41.200 --> 00:18:42.960 bootstrap a full ephemeral environment 00:18:42.960 --> 00:18:44.960 and validate everything works exactly 00:18:44.960 --> 00:18:46.559 the way you expect or how you have 00:18:46.559 --> 00:18:49.200 executable specs uh and then on the 00:18:49.200 --> 00:18:53.280 other ite you have the the code. Um, 00:18:53.280 --> 00:18:55.200 right now people are going crazy, right? 00:18:55.200 --> 00:18:56.960 You have agents reviewing agents. My 00:18:56.960 --> 00:18:58.559 friend was like that's that will never 00:18:58.559 --> 00:19:00.000 work. Agents will always say that they 00:19:00.000 --> 00:19:01.440 are doing the right job. But to be 00:19:01.440 --> 00:19:03.520 honest, you can do adversarial reviews 00:19:03.520 --> 00:19:05.360 and agents can be very critical. You can 00:19:05.360 --> 00:19:08.400 create dedicated skills that focus on 00:19:08.400 --> 00:19:10.720 identifying security issues and so on. 00:19:10.720 --> 00:19:12.480 There's a lot of startups right now with 00:19:12.480 --> 00:19:16.799 huge amount of investment um for 00:19:16.799 --> 00:19:20.240 for um you know for AI review systems, 00:19:20.240 --> 00:19:20.640 right? 00:19:20.640 --> 00:19:23.919 >> Yeah. Yep. So I mean I mean that's the 00:19:23.919 --> 00:19:26.240 next thing I wanted to get on to with 00:19:26.240 --> 00:19:28.240 Okay. So so I'm just I'm still trying to 00:19:28.240 --> 00:19:30.640 imagine this PR. So So you still have a 00:19:30.640 --> 00:19:33.520 huge PR and you just basically rely on 00:19:33.520 --> 00:19:37.520 your your CICD and your your a your 00:19:37.520 --> 00:19:39.840 adversarial agents to review it. But 00:19:39.840 --> 00:19:42.240 like and and then as you said like for 00:19:42.240 --> 00:19:44.480 security you also have a particular 00:19:44.480 --> 00:19:45.919 agent to to 00:19:45.919 --> 00:19:47.679 >> it doesn't all have to be agents right I 00:19:47.679 --> 00:19:50.400 mean right now you have very decent like 00:19:50.400 --> 00:19:52.720 Golang evaluators that say complexity of 00:19:52.720 --> 00:19:54.400 function is too is is too 00:19:54.400 --> 00:19:57.440 >> yeah like go sec and go 00:19:57.440 --> 00:19:59.840 um go lint something whatever 00:19:59.840 --> 00:20:01.760 >> yeah it will say you have three layer 00:20:01.760 --> 00:20:05.360 like you you this is there's a nested if 00:20:05.360 --> 00:20:09.360 branches um this is too complex uh fall 00:20:09.360 --> 00:20:10.640 is not clear. 00:20:10.640 --> 00:20:13.120 >> So there's some deterministic uh things 00:20:13.120 --> 00:20:14.160 you can run over it. 00:20:14.160 --> 00:20:16.880 >> Also test coverage even if you want to 00:20:16.880 --> 00:20:19.600 consider it as a valid metric. 00:20:19.600 --> 00:20:22.960 Um that will say like I don't know I 00:20:22.960 --> 00:20:24.880 really don't like test coverage after 00:20:24.880 --> 00:20:26.880 you know the problems I had with with AI 00:20:26.880 --> 00:20:28.960 because of the amount of refactoring you 00:20:28.960 --> 00:20:30.720 have to do and how I feel tests fail 00:20:30.720 --> 00:20:32.720 when you have to do massive refactoring. 00:20:32.720 --> 00:20:34.880 But maybe it's a it's a skill issue of 00:20:34.880 --> 00:20:38.080 initial planning. I I I I I'm I'm I'm 00:20:38.080 --> 00:20:39.760 not a fan of test coverage. I feel like 00:20:39.760 --> 00:20:41.520 almost ripping that out just because I 00:20:41.520 --> 00:20:42.960 don't want to encourage it because like 00:20:42.960 --> 00:20:46.320 sometimes I think the the Go core team 00:20:46.320 --> 00:20:48.320 they basically 00:20:48.320 --> 00:20:51.360 uh they have very light uh testing in a 00:20:51.360 --> 00:20:53.280 way. Their coverage is not for the full 00:20:53.280 --> 00:20:56.159 Golang source code. They they they test 00:20:56.159 --> 00:20:57.679 where they think it's meaningful to 00:20:57.679 --> 00:20:59.120 test. And I noticed that in my 00:20:59.120 --> 00:21:00.880 personally in my own projects when I get 00:21:00.880 --> 00:21:04.640 AI to do um you know try to get 100% 00:21:04.640 --> 00:21:06.240 coverage like it like completely 00:21:06.240 --> 00:21:08.880 re-engineers my my source code and 00:21:08.880 --> 00:21:11.200 completely uh you know creates these 00:21:11.200 --> 00:21:12.960 interfaces just so that they can 00:21:12.960 --> 00:21:15.039 properly be tested and it m it just 00:21:15.039 --> 00:21:17.440 balloons the source code in my it makes 00:21:17.440 --> 00:21:19.760 the source even simple projects kind of 00:21:19.760 --> 00:21:21.760 difficult in my opinion but I know it's 00:21:21.760 --> 00:21:23.600 probably better engineered but I just 00:21:23.600 --> 00:21:25.360 for a hobby project I just hate it. I 00:21:25.360 --> 00:21:26.960 had an interesting experience very early 00:21:26.960 --> 00:21:29.840 on when I started learning Golang. Um I 00:21:29.840 --> 00:21:31.520 was writing small little projects back 00:21:31.520 --> 00:21:34.880 like 10 years ago and I was trying to 00:21:34.880 --> 00:21:36.799 write go tests and it wasn't going very 00:21:36.799 --> 00:21:38.960 well and I ended up rewriting a lot of 00:21:38.960 --> 00:21:40.720 my like I implemented all the stuff and 00:21:40.720 --> 00:21:42.320 then I was like okay now let me write 00:21:42.320 --> 00:21:43.600 some tests like I haven't done that 00:21:43.600 --> 00:21:46.320 before. Uh I start writing tests and I 00:21:46.320 --> 00:21:47.919 can't test this thing. I can't test that 00:21:47.919 --> 00:21:49.760 thing because you you can't like execute 00:21:49.760 --> 00:21:51.440 it individually. you can't you you need 00:21:51.440 --> 00:21:54.000 to like the function needs to uh you 00:21:54.000 --> 00:21:56.159 know maybe the interface must be what 00:21:56.159 --> 00:21:57.679 you call it inversion of control must be 00:21:57.679 --> 00:22:00.080 injectable must be this type of patterns 00:22:00.080 --> 00:22:02.080 and I I was like wow after I finished 00:22:02.080 --> 00:22:03.919 writing the test I feel my code is so 00:22:03.919 --> 00:22:06.320 much nicer I've like and I go to my 00:22:06.320 --> 00:22:08.000 colleague and I say like well I just 00:22:08.000 --> 00:22:09.520 learned how to write go test it was 00:22:09.520 --> 00:22:11.120 pretty cool I rewrote all of my original 00:22:11.120 --> 00:22:12.799 code and he was like if you have to 00:22:12.799 --> 00:22:15.039 rewrite your original code to write go 00:22:15.039 --> 00:22:16.320 test that's probably not a good thing 00:22:16.320 --> 00:22:19.360 and I was like to be honest I believe 00:22:19.360 --> 00:22:21.039 the original code was really bad and I 00:22:21.039 --> 00:22:22.640 believe rewriting your code to implement 00:22:22.640 --> 00:22:24.159 the tests made it a lot better like 00:22:24.159 --> 00:22:26.320 you're just saying oh it creates I I I 00:22:26.320 --> 00:22:27.760 think the fact that you had trouble 00:22:27.760 --> 00:22:30.320 adding the tests or the AI ask you to 00:22:30.320 --> 00:22:32.640 you know to re interfaces because it was 00:22:32.640 --> 00:22:34.559 all tightly coupled and and not possible 00:22:34.559 --> 00:22:36.799 to test again where do you want to do 00:22:36.799 --> 00:22:38.400 the tests do you want to do unit test 00:22:38.400 --> 00:22:40.480 >> I think on on a on a client project I 00:22:40.480 --> 00:22:41.840 would have done the proper thing it's 00:22:41.840 --> 00:22:46.240 just it's just I find I find it's uh I 00:22:46.240 --> 00:22:48.320 was just trying to make a facicious I 00:22:48.320 --> 00:22:49.840 was trying to make an argument that when 00:22:49.840 --> 00:22:52.960 you go to 100% uh when when you want to 00:22:52.960 --> 00:22:54.880 achieve 100% coverage, weird things 00:22:54.880 --> 00:22:56.240 start to happen in my opinion. 00:22:56.240 --> 00:22:58.320 >> Yeah. No, like every metric once it 00:22:58.320 --> 00:22:59.760 starts to get gamed, you're not getting 00:22:59.760 --> 00:23:01.360 what you want. And I also had a 00:23:01.360 --> 00:23:03.679 colleague like who said my my coverage 00:23:03.679 --> 00:23:05.919 is super high, but that's usually that's 00:23:05.919 --> 00:23:07.600 like you have some organizations where 00:23:07.600 --> 00:23:09.600 you can't have more than X percentage of 00:23:09.600 --> 00:23:12.400 coverage. Can't be more than 70%. If if 00:23:12.400 --> 00:23:14.080 it's more than 70%, you're you're you're 00:23:14.080 --> 00:23:16.400 you're writing your test to technical 00:23:16.400 --> 00:23:18.000 implementation details. So I think 00:23:18.000 --> 00:23:19.600 there's there's some assumptions that I 00:23:19.600 --> 00:23:21.919 think that we need to capture here 00:23:21.919 --> 00:23:24.159 because I I heard a story where there's 00:23:24.159 --> 00:23:25.919 an organization that's using Claude 00:23:25.919 --> 00:23:28.320 everywhere and and in some of some of 00:23:28.320 --> 00:23:30.880 the you know the different teams the 00:23:30.880 --> 00:23:33.600 different skill levels and some people 00:23:33.600 --> 00:23:36.880 are basically producing like 00:23:36.880 --> 00:23:39.440 uh artifacts using claude or codecs 00:23:39.440 --> 00:23:42.480 which maybe don't go through a CI/CD you 00:23:42.480 --> 00:23:44.559 know like like like how do you prevent 00:23:44.559 --> 00:23:47.200 claude from like publishing to a public 00:23:47.200 --> 00:23:48.960 site or something like this these sort 00:23:48.960 --> 00:23:53.120 of things. Um like I feel I feel like we 00:23:53.120 --> 00:23:55.120 this is quite a rosy situation where 00:23:55.120 --> 00:23:56.559 everyone's like 00:23:56.559 --> 00:23:58.320 >> I disagree. I mean if you're telling me 00:23:58.320 --> 00:24:00.000 that the organization doesn't have 00:24:00.000 --> 00:24:03.919 proper uh you know development practices 00:24:03.919 --> 00:24:07.360 and people are like I think the biggest 00:24:07.360 --> 00:24:09.840 issue that people have with AI messing 00:24:09.840 --> 00:24:11.600 up things is that they are not engineers 00:24:11.600 --> 00:24:13.360 and they have never worked in in an 00:24:13.360 --> 00:24:14.960 organization that actually follows 00:24:14.960 --> 00:24:17.360 proper sane engineering practices. 00:24:17.360 --> 00:24:18.320 >> But this this is the 00:24:18.320 --> 00:24:20.240 >> having separate environments not getting 00:24:20.240 --> 00:24:22.400 giving access to production going and 00:24:22.400 --> 00:24:24.400 promoting artifacts and as you increase 00:24:24.400 --> 00:24:26.400 the trust. I mean a lot of these stories 00:24:26.400 --> 00:24:28.880 that I hear about AI is because you're 00:24:28.880 --> 00:24:31.520 not following the most basic DevOps or 00:24:31.520 --> 00:24:33.440 security practices that even the 00:24:33.440 --> 00:24:34.640 smallest startup has. 00:24:34.640 --> 00:24:35.840 >> Well, [clears throat] I mean I could 00:24:35.840 --> 00:24:37.279 come up I can come up with some 00:24:37.279 --> 00:24:40.720 scenarios like for example like for like 00:24:40.720 --> 00:24:43.440 you you in a in a even in a a small or 00:24:43.440 --> 00:24:45.120 larger organization you might have 00:24:45.120 --> 00:24:48.559 people who are like um you know working 00:24:48.559 --> 00:24:50.400 in finance and things like this and then 00:24:50.400 --> 00:24:52.400 they want to come up with some sales 00:24:52.400 --> 00:24:55.039 report. I mean to them this this whole 00:24:55.039 --> 00:24:57.360 engineering stuff is completely foreign 00:24:57.360 --> 00:24:59.039 to them and then they end up 00:24:59.039 --> 00:25:01.120 >> but but it is a problem where where 00:25:01.120 --> 00:25:04.000 you're not like an AI engineer or an AI 00:25:04.000 --> 00:25:06.559 um what what would you say an AI agent 00:25:06.559 --> 00:25:08.559 is basically need to be on boarded onto 00:25:08.559 --> 00:25:11.200 the systems right and the boarding 00:25:11.200 --> 00:25:14.400 practice must um must be solid if you do 00:25:14.400 --> 00:25:16.400 the same with a human you could argue 00:25:16.400 --> 00:25:18.320 that like you tell a human it's going to 00:25:18.320 --> 00:25:19.760 do it's going to make a mistake but you 00:25:19.760 --> 00:25:21.520 tell it not to do it again it's going to 00:25:21.520 --> 00:25:23.200 do it again and AI agent doesn't really 00:25:23.200 --> 00:25:24.640 learn unless you have like this like 00:25:24.640 --> 00:25:26.799 memory system or or so on, but like it's 00:25:26.799 --> 00:25:28.080 still going to keep making the same 00:25:28.080 --> 00:25:29.760 mistakes. I I would accept that as an 00:25:29.760 --> 00:25:32.240 argument, but still it's a problem with 00:25:32.240 --> 00:25:33.840 access. If you're saying like, oh, 00:25:33.840 --> 00:25:36.080 there's a finance guy now gets wants to 00:25:36.080 --> 00:25:38.240 get cloud access because um it's all the 00:25:38.240 --> 00:25:40.720 rage and he believes himself um an 00:25:40.720 --> 00:25:43.120 excellent engineer because AI tells you 00:25:43.120 --> 00:25:44.880 this is so easy and let's do it and then 00:25:44.880 --> 00:25:46.880 it just goes and publishes um a 00:25:46.880 --> 00:25:49.279 financial report which would be private. 00:25:49.279 --> 00:25:51.679 Uh I can agree with you that that that 00:25:51.679 --> 00:25:54.400 happens. Um and and we actually have a 00:25:54.400 --> 00:25:56.480 DevOps day talk about that last year in 00:25:56.480 --> 00:25:59.120 Singapore where if your organization 00:25:59.120 --> 00:26:00.880 doesn't have a proper policy on AI 00:26:00.880 --> 00:26:02.720 adoption already a year ago they were 00:26:02.720 --> 00:26:04.320 saying like well how are you going to 00:26:04.320 --> 00:26:07.360 govern um AI like if if you don't have a 00:26:07.360 --> 00:26:09.200 policy people are going to use their own 00:26:09.200 --> 00:26:10.400 subscription and their own tool. 00:26:10.400 --> 00:26:11.840 >> So for example the finance or the 00:26:11.840 --> 00:26:14.960 nontechnical people in an organization 00:26:14.960 --> 00:26:16.480 let me get this straight. So you're 00:26:16.480 --> 00:26:17.600 saying that you wouldn't give them 00:26:17.600 --> 00:26:20.080 access to to claude on AI tools? 00:26:20.080 --> 00:26:23.120 >> No, no. Like the talk that I refer of uh 00:26:23.120 --> 00:26:25.039 DevOps day Singapore last year 00:26:25.039 --> 00:26:27.039 explicitly said that as an organization 00:26:27.039 --> 00:26:29.200 you have to have an AI policy. You have 00:26:29.200 --> 00:26:32.000 to put in place um the contracts and 00:26:32.000 --> 00:26:33.520 also you have to give them access you 00:26:33.520 --> 00:26:35.360 have to give them a proper way to access 00:26:35.360 --> 00:26:37.520 it. That includes through a company 00:26:37.520 --> 00:26:40.320 subscription um with limitations that 00:26:40.320 --> 00:26:42.480 like will 00:26:42.480 --> 00:26:45.520 um you know first of like guarantee that 00:26:45.520 --> 00:26:46.720 there will be no training on this 00:26:46.720 --> 00:26:48.559 internal data. No internal data will be 00:26:48.559 --> 00:26:50.400 used by the inference provider. So 00:26:50.400 --> 00:26:52.320 that's a proper contract between the 00:26:52.320 --> 00:26:55.360 company and the um and the inference 00:26:55.360 --> 00:26:57.760 provider. and and then other suggestions 00:26:57.760 --> 00:27:00.320 that was in that talk which I'm off the 00:27:00.320 --> 00:27:02.159 top of my head was like to actually have 00:27:02.159 --> 00:27:06.000 a router that looks at the AI usage. Um 00:27:06.000 --> 00:27:08.159 basically everyone is using AI through 00:27:08.159 --> 00:27:11.120 this router and then that gives you a a 00:27:11.120 --> 00:27:13.440 layer of accountability and also ability 00:27:13.440 --> 00:27:14.320 to block things. 00:27:14.320 --> 00:27:16.960 >> So on on that topic how do you how do 00:27:16.960 --> 00:27:19.039 you route all the the AI network 00:27:19.039 --> 00:27:20.880 requests through a router? This is 00:27:20.880 --> 00:27:24.240 something I I I I was meaning to look at 00:27:24.240 --> 00:27:26.000 today. I I 00:27:26.000 --> 00:27:28.080 >> so I can I I could find you that talk 00:27:28.080 --> 00:27:29.679 because he he gives an example. It's a 00:27:29.679 --> 00:27:30.880 year old so obviously a lot have 00:27:30.880 --> 00:27:33.440 changed. Um and also funny I think we 00:27:33.440 --> 00:27:34.960 mentioned this in the previous call as 00:27:34.960 --> 00:27:36.799 well that a lot of these routers are 00:27:36.799 --> 00:27:39.039 actually like stealing um your your 00:27:39.039 --> 00:27:41.360 credentials and like trying like crypto 00:27:41.360 --> 00:27:43.360 wallet stealers and so on. So you have 00:27:43.360 --> 00:27:47.200 to be careful which one you use. But but 00:27:47.200 --> 00:27:49.200 I think another important part of that 00:27:49.200 --> 00:27:52.159 aspect is or another important aspect of 00:27:52.159 --> 00:27:53.679 that is something that's really 00:27:53.679 --> 00:27:55.200 happening a lot right now with agents 00:27:55.200 --> 00:27:57.279 messing up is like agent like you just 00:27:57.279 --> 00:27:59.679 mentioned guard rails right I talk 00:27:59.679 --> 00:28:01.919 frequently with George who works at 00:28:01.919 --> 00:28:03.679 stackpack and they kind of pivoted 00:28:03.679 --> 00:28:05.919 towards more of aentic uh because they 00:28:05.919 --> 00:28:07.919 are focused on devops tasks so they're 00:28:07.919 --> 00:28:10.080 focused specifically on how do we put 00:28:10.080 --> 00:28:13.679 like a harness that catches 00:28:13.679 --> 00:28:16.480 damaging actions and then uh stops them, 00:28:16.480 --> 00:28:18.159 right? I think there was recently an 00:28:18.159 --> 00:28:20.640 announcement from AWS or uh they said 00:28:20.640 --> 00:28:22.640 like hey here we give you a sandbox was 00:28:22.640 --> 00:28:24.559 it I think Verscell announced agent 00:28:24.559 --> 00:28:26.320 sandbox where they say now you give we 00:28:26.320 --> 00:28:28.399 give you a full sandbox and um we 00:28:28.399 --> 00:28:29.840 guarantee that the agent cannot do 00:28:29.840 --> 00:28:31.840 anything anything silly. I'm not sure if 00:28:31.840 --> 00:28:34.240 it was Versel or AWS. There's been a 00:28:34.240 --> 00:28:36.320 whole there has been a few announcements 00:28:36.320 --> 00:28:38.640 like I I thought I was of the opinion 00:28:38.640 --> 00:28:42.080 that AWS Bedrock had had a lot of stuff 00:28:42.080 --> 00:28:44.480 in this regard, but I'm I'm just talking 00:28:44.480 --> 00:28:45.279 up my ass really. 00:28:45.279 --> 00:28:46.399 >> I remember seeing an announcement 00:28:46.399 --> 00:28:49.760 somewhere on on on on um LinkedIn being 00:28:49.760 --> 00:28:52.799 like we we guarantee you that the agent 00:28:52.799 --> 00:28:55.760 cannot do something damaging. 00:28:55.760 --> 00:28:57.600 Like in my previous client, I was I 00:28:57.600 --> 00:28:59.279 think I was whining and bitching about 00:28:59.279 --> 00:29:01.840 it with you, but like it was heavily uh 00:29:01.840 --> 00:29:03.600 limited in the sense that we could only 00:29:03.600 --> 00:29:06.320 use it in co-pilot 00:29:06.320 --> 00:29:08.480 and I guess it was a bit for for 00:29:08.480 --> 00:29:11.840 fortuitous, but we were using uh dev 00:29:11.840 --> 00:29:13.840 containers. So when we were using dev 00:29:13.840 --> 00:29:15.760 containers, we were using co-pilot and 00:29:15.760 --> 00:29:17.679 we're using AI and dev containers and 00:29:17.679 --> 00:29:19.919 everything was very was sandboxed by 00:29:19.919 --> 00:29:23.120 default like without us even thinking 00:29:23.120 --> 00:29:26.880 about it. Um so 00:29:26.880 --> 00:29:28.960 of course 00:29:28.960 --> 00:29:31.360 of course most people who who using AI 00:29:31.360 --> 00:29:33.520 are not really using it in a sandbox 00:29:33.520 --> 00:29:35.840 right at this point in time. I I know 00:29:35.840 --> 00:29:38.080 maybe clawed code work tried to get 00:29:38.080 --> 00:29:39.760 there 00:29:39.760 --> 00:29:42.399 uh but like most people who who who just 00:29:42.399 --> 00:29:44.720 run clawed code are don't have a proper 00:29:44.720 --> 00:29:46.559 sandbox really especially when they when 00:29:46.559 --> 00:29:48.159 they go you know dangerously skip 00:29:48.159 --> 00:29:49.279 permissions or something. 00:29:49.279 --> 00:29:51.279 >> Yeah I think 00:29:51.279 --> 00:29:53.039 but so Claude Code added a new 00:29:53.039 --> 00:29:55.200 interesting thing which is auto mode. 00:29:55.200 --> 00:29:57.039 Um, so originally when you did a plan, 00:29:57.039 --> 00:29:58.720 you had the option to go auto accept 00:29:58.720 --> 00:30:00.720 edits and then it would still constantly 00:30:00.720 --> 00:30:02.480 ex ask you for permission because I I 00:30:02.480 --> 00:30:04.399 don't use I don't run dangerally skip 00:30:04.399 --> 00:30:06.320 permissions. I am the bottleneck. When 00:30:06.320 --> 00:30:07.520 it asks me a comment, I haven't 00:30:07.520 --> 00:30:09.200 whitelisted. I have to approve it or 00:30:09.200 --> 00:30:11.520 allow listed. Then now it has this allow 00:30:11.520 --> 00:30:13.520 mode which I've been trying to use a 00:30:13.520 --> 00:30:15.440 couple of times and successfully 00:30:15.440 --> 00:30:17.600 sometimes but lately I actually kind of 00:30:17.600 --> 00:30:19.760 avoid it. So auto mode is a separate 00:30:19.760 --> 00:30:22.080 eval agent that like a small little 00:30:22.080 --> 00:30:24.000 model that does a quick eval like is 00:30:24.000 --> 00:30:26.799 this a dangerous activity or not and it 00:30:26.799 --> 00:30:28.480 rejects things because it says things 00:30:28.480 --> 00:30:30.240 yeah but that's kind of like affecting a 00:30:30.240 --> 00:30:32.480 third party system and we should not be 00:30:32.480 --> 00:30:34.080 affecting third party system like it's 00:30:34.080 --> 00:30:37.360 very uh generic type of evaluator. It 00:30:37.360 --> 00:30:38.880 doesn't take into account the context of 00:30:38.880 --> 00:30:40.240 what we're trying to do. It just looks 00:30:40.240 --> 00:30:41.840 at the command and says could this 00:30:41.840 --> 00:30:43.679 potentially have a negative impact? If 00:30:43.679 --> 00:30:45.440 it does it just rejects it interesting 00:30:45.440 --> 00:30:46.880 and then it gives the reason. I thought 00:30:46.880 --> 00:30:48.320 we're going to maybe talk about hooks 00:30:48.320 --> 00:30:50.080 and things like that. Have you have you 00:30:50.080 --> 00:30:51.279 played with clawed hooks? 00:30:51.279 --> 00:30:54.159 >> Yeah. So the the spec ledger CLI is a 00:30:54.159 --> 00:30:57.840 hook that um that that acts whenever a 00:30:57.840 --> 00:30:59.120 command get executed. 00:30:59.120 --> 00:31:00.880 >> Yeah. But like are you talking about 00:31:00.880 --> 00:31:03.200 like a cloud hook or github? 00:31:03.200 --> 00:31:03.840 >> Yeah. 00:31:03.840 --> 00:31:06.080 >> Cloud. So it happens whenever the agent 00:31:06.080 --> 00:31:08.080 triggers a git commit because a lot of 00:31:08.080 --> 00:31:10.960 the prompts and a lot of the work I do 00:31:10.960 --> 00:31:13.200 um now is I I have no problem with the 00:31:13.200 --> 00:31:15.360 agent. Actually, before I never allow 00:31:15.360 --> 00:31:17.279 the agent to commit, but now I always 00:31:17.279 --> 00:31:18.640 ask the agent to commit because it 00:31:18.640 --> 00:31:20.320 writes a really good it stages the file 00:31:20.320 --> 00:31:21.760 correctly. It writes a really good 00:31:21.760 --> 00:31:23.919 commit message that I like to see. So, 00:31:23.919 --> 00:31:25.919 um, so that's why the agent it requires 00:31:25.919 --> 00:31:27.760 you to trust the agent to do the commit. 00:31:27.760 --> 00:31:30.240 So, when that happens, um, it's a bash 00:31:30.240 --> 00:31:31.919 call that that doesn't 00:31:31.919 --> 00:31:35.279 >> your claw hook uh sort of make sure that 00:31:35.279 --> 00:31:37.840 there's a good uh get. 00:31:37.840 --> 00:31:39.200 >> No, no, no. the only thing it does 00:31:39.200 --> 00:31:41.200 because it has to do something real fast 00:31:41.200 --> 00:31:42.960 uh and get out of the way because if the 00:31:42.960 --> 00:31:44.720 hook fails it it really breaks your 00:31:44.720 --> 00:31:46.320 agent's shell. We have experience where 00:31:46.320 --> 00:31:49.279 if CLI breaks for whatever reason and 00:31:49.279 --> 00:31:51.200 then everything the agent tries to do 00:31:51.200 --> 00:31:54.159 because it was a a bash tool hook. So 00:31:54.159 --> 00:31:56.159 every time the bash tool was called to 00:31:56.159 --> 00:31:58.320 do a git command or whatever to identify 00:31:58.320 --> 00:32:00.159 if it was a big git command. It was a 00:32:00.159 --> 00:32:01.919 nightmare. No. suddenly like I would 00:32:01.919 --> 00:32:04.720 download the new spec ledger CLI uh 00:32:04.720 --> 00:32:06.080 release and then I would try to use 00:32:06.080 --> 00:32:07.679 cloud and it would just break like the 00:32:07.679 --> 00:32:09.120 model would just constantly every 00:32:09.120 --> 00:32:10.799 command and it wouldn't tell you why it 00:32:10.799 --> 00:32:12.960 would just be failed failed failed and 00:32:12.960 --> 00:32:14.880 after investigation we figure I figured 00:32:14.880 --> 00:32:16.880 out like oh it's because of the SL 00:32:16.880 --> 00:32:19.919 binary hook is is exiting exit one and 00:32:19.919 --> 00:32:21.279 the hooks are failing and there was no 00:32:21.279 --> 00:32:23.279 information why 00:32:23.279 --> 00:32:24.559 >> that's the one thing I always get 00:32:24.559 --> 00:32:26.559 frustrated with uh all the tooling 00:32:26.559 --> 00:32:29.279 really is that like sometimes something 00:32:29.279 --> 00:32:32.240 hit something went into my global thing 00:32:32.240 --> 00:32:33.919 and then and then and then there's 00:32:33.919 --> 00:32:35.600 something in my project thing and I 00:32:35.600 --> 00:32:37.840 don't know what the precedence is like 00:32:37.840 --> 00:32:40.880 or what what's being used 00:32:40.880 --> 00:32:42.480 >> but that what I really like about that 00:32:42.480 --> 00:32:44.159 is that whenever something like that is 00:32:44.159 --> 00:32:46.320 failing I always ask cloud okay go and 00:32:46.320 --> 00:32:48.159 look what is going on and it goes 00:32:48.159 --> 00:32:49.840 through its local settings it goes 00:32:49.840 --> 00:32:51.440 through the global settings it goes here 00:32:51.440 --> 00:32:52.559 and there and says like these are all 00:32:52.559 --> 00:32:54.159 the things that are currently connected 00:32:54.159 --> 00:32:56.480 that could be impacting this and and 00:32:56.480 --> 00:32:57.919 here it is this is the fix 00:32:57.919 --> 00:32:59.519 >> so the next thing I want to talk about 00:32:59.519 --> 00:33:02.960 is things.json like 00:33:02.960 --> 00:33:05.360 um do you ever touch this file like I I 00:33:05.360 --> 00:33:06.880 know it's probably more of a Claude 00:33:06.880 --> 00:33:08.880 thing but like do you ever because in 00:33:08.880 --> 00:33:13.120 the settings.json you can like uh I 00:33:13.120 --> 00:33:15.360 think Claude also does it in skills in 00:33:15.360 --> 00:33:17.279 other places. Hold on I'll get to skills 00:33:17.279 --> 00:33:18.720 in a bit but like in in the 00:33:18.720 --> 00:33:20.240 settings.json you have your like 00:33:20.240 --> 00:33:22.000 permission sets you can say that like 00:33:22.000 --> 00:33:23.679 >> and also arguments for the model and 00:33:23.679 --> 00:33:25.519 like API endpoints for the model. Yeah, 00:33:25.519 --> 00:33:26.720 there's a bunch of stuff that goes in 00:33:26.720 --> 00:33:28.320 there. Like everything goes in there 00:33:28.320 --> 00:33:33.200 basically. Like would you say it's um it 00:33:33.200 --> 00:33:36.080 makes sense to ship a settings.json in a 00:33:36.080 --> 00:33:39.760 project as to you know maybe disallow 00:33:39.760 --> 00:33:42.399 users from I don't know RM or something 00:33:42.399 --> 00:33:44.240 like that 00:33:44.240 --> 00:33:46.080 RM minus RF. You know what I mean? So, 00:33:46.080 --> 00:33:48.480 so what what specledger uh the Golang 00:33:48.480 --> 00:33:50.960 CLI does is it does have the concept of 00:33:50.960 --> 00:33:53.919 agent config because one of the people 00:33:53.919 --> 00:33:55.519 that was working on the project was 00:33:55.519 --> 00:33:57.279 either using open code or Claude Code 00:33:57.279 --> 00:34:00.080 with quen local or a separate uh 00:34:00.080 --> 00:34:02.480 endpoint. So when he I built a feature 00:34:02.480 --> 00:34:04.559 that launch launches Claude Code and with 00:34:04.559 --> 00:34:06.399 a prepared prompt and then goes straight 00:34:06.399 --> 00:34:08.639 into some activity. So you would run SL 00:34:08.639 --> 00:34:10.480 and then it would bootstrap the repo and 00:34:10.480 --> 00:34:12.159 then would immediately launch Claude Code 00:34:12.159 --> 00:34:14.240 and says um we're bootstrapping the repo 00:34:14.240 --> 00:34:15.679 and it has the prompts and the skills 00:34:15.679 --> 00:34:17.200 and it would just go off and he'd be 00:34:17.200 --> 00:34:18.879 like that really breaks for me because I 00:34:18.879 --> 00:34:21.119 have to set a custom endpoint. I have to 00:34:21.119 --> 00:34:22.399 do all of these additional settings 00:34:22.399 --> 00:34:24.480 configuration when I launch Claude Code 00:34:24.480 --> 00:34:26.159 there's all this like arguments that I 00:34:26.159 --> 00:34:28.240 pass in and so he built a feature on the 00:34:28.240 --> 00:34:30.159 on the CLI that that gives you the 00:34:30.159 --> 00:34:32.480 ability to configure different agents uh 00:34:32.480 --> 00:34:34.800 profiles um like and and so your 00:34:34.800 --> 00:34:36.879 settings.json JSON but but with two 00:34:36.879 --> 00:34:39.440 sections like secrets versus non-secs 00:34:39.440 --> 00:34:41.919 overridable what is collected what you 00:34:41.919 --> 00:34:44.079 would commit into the repo because maybe 00:34:44.079 --> 00:34:46.000 you want some of those settings be get 00:34:46.000 --> 00:34:47.679 included and some of this like local 00:34:47.679 --> 00:34:50.800 that you don't want to include so so 00:34:50.800 --> 00:34:53.440 there is quite a extensive feature spec 00:34:53.440 --> 00:34:55.520 on on how to set up the Claude Code 00:34:55.520 --> 00:34:57.440 settings and so on but to be honest I 00:34:57.440 --> 00:34:59.520 don't use it because I I'm using cloud 00:34:59.520 --> 00:35:02.320 code as is I don't have any like custom 00:35:02.320 --> 00:35:03.920 flags or something so there was only one 00:35:03.920 --> 00:35:05.760 other person using completeness. Yeah, I 00:35:05.760 --> 00:35:09.440 wanted to mention like get um what is it 00:35:09.440 --> 00:35:10.480 called? Git hooks 00:35:10.480 --> 00:35:12.320 >> attribut hooks. 00:35:12.320 --> 00:35:14.160 >> What is have you heard of this new 00:35:14.160 --> 00:35:16.560 project called PR? Hold on. What is pre? 00:35:16.560 --> 00:35:17.760 Oh, pre-commit. Sorry. 00:35:17.760 --> 00:35:19.760 >> Yeah, of course. What do you mean new? 00:35:19.760 --> 00:35:21.280 >> Okay, whatever. 00:35:21.280 --> 00:35:21.760 >> Um 00:35:21.760 --> 00:35:24.880 >> I I've been using Precomit since 2019, I 00:35:24.880 --> 00:35:26.720 think seven years for all of my 00:35:26.720 --> 00:35:28.160 Terraform. 00:35:28.160 --> 00:35:29.280 >> Autofixers are like 00:35:29.280 --> 00:35:31.599 >> Yeah. Yeah. Yeah. But PRE is the new 00:35:31.599 --> 00:35:34.160 Rust one. Whatever. uh 00:35:34.160 --> 00:35:35.359 >> new wish. Okay, 00:35:35.359 --> 00:35:38.320 >> you just given me an idea uh Vincent 00:35:38.320 --> 00:35:40.880 like how about the vibe coding a tool 00:35:40.880 --> 00:35:43.280 like this a tool where you can go like 00:35:43.280 --> 00:35:46.560 um you know 00:35:46.560 --> 00:35:49.200 uh AI show guard rails or something like 00:35:49.200 --> 00:35:51.280 this and it would basically look through 00:35:51.280 --> 00:35:54.240 your your uh your settings your 00:35:54.240 --> 00:35:56.960 pre-commits 00:35:56.960 --> 00:35:59.839 um your your your I don't know your 00:35:59.839 --> 00:36:03.040 switches your flags and basically 00:36:03.040 --> 00:36:05.119 basically you know you know, give a 00:36:05.119 --> 00:36:09.440 score about about your your your 00:36:09.440 --> 00:36:11.280 session, how safe it is, you know what I 00:36:11.280 --> 00:36:13.280 mean? Cuz cuz a lot of people, including 00:36:13.280 --> 00:36:15.040 myself, to be honest, in a personal 00:36:15.040 --> 00:36:18.320 context, I go yolo in my little sandbox. 00:36:18.320 --> 00:36:21.280 But like from an org point of view, you 00:36:21.280 --> 00:36:24.640 you want people um well to to set up to 00:36:24.640 --> 00:36:27.200 to use AI safely. But in the in the case 00:36:27.200 --> 00:36:29.200 that they haven't for one reason or 00:36:29.200 --> 00:36:31.200 another, maybe it would be good to sort 00:36:31.200 --> 00:36:33.680 of gify it to say like, hey, you okay, 00:36:33.680 --> 00:36:36.079 you had a great session with AI, but you 00:36:36.079 --> 00:36:38.880 know what? um you know this could have 00:36:38.880 --> 00:36:40.560 gone wrong that could have gone wrong 00:36:40.560 --> 00:36:44.320 and you know your score is 58% 00:36:44.320 --> 00:36:47.920 you know next time um AI could have done 00:36:47.920 --> 00:36:51.200 this you could have improved and become 00:36:51.200 --> 00:36:54.000 you know 88% 00:36:54.000 --> 00:36:56.960 uh if you um had this guard rail in 00:36:56.960 --> 00:36:58.640 place or something like that you know 00:36:58.640 --> 00:36:59.119 what I mean 00:36:59.119 --> 00:37:01.839 >> yeah I would definitely ask clout or 00:37:01.839 --> 00:37:04.400 chat to do a explore if anything like 00:37:04.400 --> 00:37:07.839 that exists um it Sounds interesting, 00:37:07.839 --> 00:37:09.680 but it also sounds it also sounds like 00:37:09.680 --> 00:37:11.280 something that I would go like that's 00:37:11.280 --> 00:37:13.575 wrong. You don't trust me. I don't care 00:37:13.575 --> 00:37:14.480 [laughter] 00:37:14.480 --> 00:37:16.880 cuz yes. I mean I mean these tools are 00:37:16.880 --> 00:37:18.880 popular and people use them and I think 00:37:18.880 --> 00:37:22.320 they're good for um non-technical user 00:37:22.320 --> 00:37:24.240 but like for me I find them offensive. 00:37:24.240 --> 00:37:25.760 >> I guess for us we would find it 00:37:25.760 --> 00:37:27.440 frustrating and dumb or something like 00:37:27.440 --> 00:37:28.640 that. But I do 00:37:28.640 --> 00:37:30.079 >> Yeah. It would it would say like you 00:37:30.079 --> 00:37:32.079 only have a 50% secure setup and I would 00:37:32.079 --> 00:37:35.839 be like f you. I know what I'm doing. Uh 00:37:35.839 --> 00:37:38.800 but but but um while we were on the 00:37:38.800 --> 00:37:40.160 topic actually earlier about 00:37:40.160 --> 00:37:42.480 settings.json and when I was talking 00:37:42.480 --> 00:37:44.640 about doing I mean I think the AI show 00:37:44.640 --> 00:37:46.720 guard wheels is cool but what I wanted 00:37:46.720 --> 00:37:49.200 to to highlight is something that I I 00:37:49.200 --> 00:37:51.599 learned and I saw a funny meme. So I 00:37:51.599 --> 00:37:52.960 think it was on Reddit and people and 00:37:52.960 --> 00:37:54.800 people are like when you work with AI 00:37:54.800 --> 00:37:58.480 and um you say like commit this then 00:37:58.480 --> 00:38:00.240 cloud goes like here's the commit and 00:38:00.240 --> 00:38:01.839 then it like writes the whole commit 00:38:01.839 --> 00:38:03.680 message and then at the bottom it says 00:38:03.680 --> 00:38:07.520 co-authored by um cloud opus 1M or 00:38:07.520 --> 00:38:09.280 whatever like the model and the context 00:38:09.280 --> 00:38:12.480 and so on. a lot of times because maybe 00:38:12.480 --> 00:38:14.160 it's something that you don't want 00:38:14.160 --> 00:38:16.160 people to see that it was generated with 00:38:16.160 --> 00:38:18.800 AI. you may you know either I I I 00:38:18.800 --> 00:38:20.400 usually I don't give it a permission to 00:38:20.400 --> 00:38:22.160 do the git commit sometimes and I just 00:38:22.160 --> 00:38:23.680 in those projects I don't give it the 00:38:23.680 --> 00:38:26.000 permission and I would go like no I 00:38:26.000 --> 00:38:27.280 would copy the command remove the 00:38:27.280 --> 00:38:28.720 attribution and then run the command 00:38:28.720 --> 00:38:30.079 myself and I would be like I already 00:38:30.079 --> 00:38:31.599 committed it uh I had to fix the 00:38:31.599 --> 00:38:34.720 pre-commit um go continue [laughter] 00:38:34.720 --> 00:38:36.240 I would I would usually because actually 00:38:36.240 --> 00:38:38.079 it's in projects usually 00:38:38.079 --> 00:38:40.720 >> get commit gas amend right 00:38:40.720 --> 00:38:42.960 >> but the very funny thing is and it was a 00:38:42.960 --> 00:38:45.440 funny meme of a picture of of a of a guy 00:38:45.440 --> 00:38:47.119 uh I I forgot the picture. Okay, you got 00:38:47.119 --> 00:38:48.480 to find my picture for for 00:38:48.480 --> 00:38:50.000 >> I got to find it. I might I'm not sure 00:38:50.000 --> 00:38:52.400 if I saved the comment because there was 00:38:52.400 --> 00:38:54.000 a very interesting comment. Apparently, 00:38:54.000 --> 00:38:56.480 you can in the settings of cloud disable 00:38:56.480 --> 00:38:58.640 attribution. So, by default, yeah, by 00:38:58.640 --> 00:39:00.480 default, cloud will add its like 00:39:00.480 --> 00:39:02.079 co-authored by cloud in the commit 00:39:02.079 --> 00:39:03.920 message and also if you create a pull 00:39:03.920 --> 00:39:06.000 request, it will also say generated with 00:39:06.000 --> 00:39:08.480 cloud in the in the pull request. Uh you 00:39:08.480 --> 00:39:10.960 can disable both or one or the other. 00:39:10.960 --> 00:39:12.800 So, so you have all of that uh 00:39:12.800 --> 00:39:14.400 capability. So you can actually and then 00:39:14.400 --> 00:39:16.079 right after I read read that comment on 00:39:16.079 --> 00:39:18.000 on Reddit I went to cloud and I said can 00:39:18.000 --> 00:39:20.079 you disable attribution and it was like 00:39:20.079 --> 00:39:22.160 both PR or commit and I was like both 00:39:22.160 --> 00:39:23.599 and it was like okay done. I was like, 00:39:23.599 --> 00:39:25.359 "Holy shit." You know, so many times 00:39:25.359 --> 00:39:27.359 I've just like told it not to do that 00:39:27.359 --> 00:39:29.520 and then went in and and and and did it 00:39:29.520 --> 00:39:31.040 myself. And the funny thing is when I 00:39:31.040 --> 00:39:32.400 did that and it then shows the whole 00:39:32.400 --> 00:39:34.560 pre-commit like pass pass and then or 00:39:34.560 --> 00:39:36.160 maybe with an autofixer and then I go on 00:39:36.160 --> 00:39:37.520 stage and then do the commit again and 00:39:37.520 --> 00:39:38.800 then I have a reason, right? And I say, 00:39:38.800 --> 00:39:40.480 "Hey, I had to do it myself." And then 00:39:40.480 --> 00:39:42.240 cloud next time it's like here's the git 00:39:42.240 --> 00:39:43.520 command for you to run and I was like, 00:39:43.520 --> 00:39:45.200 "Okay, thank you." So it kind of 00:39:45.200 --> 00:39:46.800 remembers like this guy doesn't want me 00:39:46.800 --> 00:39:48.240 to commit. 00:39:48.240 --> 00:39:50.480 >> But but that's that's kind of a bad 00:39:50.480 --> 00:39:52.480 practice though. I mean we you want to 00:39:52.480 --> 00:39:54.720 show that you used AI, right? 00:39:54.720 --> 00:39:56.800 >> We depend on the project. 00:39:56.800 --> 00:39:59.520 >> For this for this AI show guardrails 00:39:59.520 --> 00:40:00.000 that 00:40:00.000 --> 00:40:01.599 >> I'm going to give you an example that 00:40:01.599 --> 00:40:04.240 somebody told me. Okay. Not me. 00:40:04.240 --> 00:40:06.000 >> It's a friend of a friend. Absolutely 00:40:06.000 --> 00:40:07.040 not me. 00:40:07.040 --> 00:40:08.000 >> Totally not you. 00:40:08.000 --> 00:40:09.359 >> Remember the whole thing I said about 00:40:09.359 --> 00:40:12.560 like AI routers, API uh AI router 00:40:12.560 --> 00:40:14.720 governance contracts in place. You 00:40:14.720 --> 00:40:16.240 cannot use like you even said it. You 00:40:16.240 --> 00:40:18.560 cannot use certain models. If I do a 00:40:18.560 --> 00:40:21.599 commit with Opus 4.71M, but it's not 00:40:21.599 --> 00:40:24.240 available on whatever the the plan that 00:40:24.240 --> 00:40:26.320 that that I'm supposed to use, that's a 00:40:26.320 --> 00:40:28.400 problem, right? I cannot have a commit 00:40:28.400 --> 00:40:31.680 that says Opus 4.71 did this commit. 00:40:31.680 --> 00:40:32.320 >> Yeah. 00:40:32.320 --> 00:40:34.079 >> I mean, not me, right? Somebody told me 00:40:34.079 --> 00:40:34.800 that. 00:40:34.800 --> 00:40:35.920 >> Yeah. 00:40:35.920 --> 00:40:38.240 >> Yeah. I mean, this I'm actually 00:40:38.240 --> 00:40:39.200 >> using the right. 00:40:39.200 --> 00:40:40.800 >> I want I want to write this program 00:40:40.800 --> 00:40:42.640 because I 00:40:42.640 --> 00:40:44.160 >> I know it's going to catch people like 00:40:44.160 --> 00:40:45.280 me. 00:40:45.280 --> 00:40:47.359 >> Catch people like me. catch people like 00:40:47.359 --> 00:40:49.920 you Vincent for example but this this is 00:40:49.920 --> 00:40:51.680 the problem I'm seeing all every day at 00:40:51.680 --> 00:40:54.320 work people are basically 00:40:54.320 --> 00:40:57.520 um maybe maybe token usage there's a lot 00:40:57.520 --> 00:40:58.960 of things that could go in here do you 00:40:58.960 --> 00:41:01.359 think this okay now do you think this 00:41:01.359 --> 00:41:04.400 makes sense to be a spec driven develop 00:41:04.400 --> 00:41:06.800 spec driven develop project yeah I mean 00:41:06.800 --> 00:41:08.480 I think this type of project is 00:41:08.480 --> 00:41:10.800 something I mean you could probably okay 00:41:10.800 --> 00:41:12.560 you could probably vibe code this with 00:41:12.560 --> 00:41:14.400 the going into plan mode and it will 00:41:14.400 --> 00:41:16.000 make a lot of assumptions 00:41:16.000 --> 00:41:17.839 Um, but if you want to have the real 00:41:17.839 --> 00:41:19.440 control of exactly what are the metrics 00:41:19.440 --> 00:41:21.280 that you want to capture, then doing it 00:41:21.280 --> 00:41:22.960 with spec driven development, it might 00:41:22.960 --> 00:41:25.520 actually help you more like spike on the 00:41:25.520 --> 00:41:26.960 different type of guard rails that you 00:41:26.960 --> 00:41:27.440 want to build. 00:41:27.440 --> 00:41:29.680 >> Okay, I'm I I just realized the time I 00:41:29.680 --> 00:41:31.760 need to take my kids to school in 10 00:41:31.760 --> 00:41:34.800 minutes, but okay, I I'm I'm going to 00:41:34.800 --> 00:41:38.480 record myself. I'm I'm gonna mark this 00:41:38.480 --> 00:41:40.880 the the places where I might be a bit 00:41:40.880 --> 00:41:42.960 confused. I'm going to record myself 00:41:42.960 --> 00:41:46.720 developing this in spec with spec uh the 00:41:46.720 --> 00:41:48.640 spectrum development on the specket and 00:41:48.640 --> 00:41:51.040 then and then maybe we can use that as a 00:41:51.040 --> 00:41:53.040 topic for the next pod, right? 00:41:53.040 --> 00:41:55.040 >> Yeah. Um it might be like if you're used 00:41:55.040 --> 00:41:57.040 to plan mode, it will be frustrating but 00:41:57.040 --> 00:41:58.720 it might also be rewarding because you 00:41:58.720 --> 00:42:00.480 might see all the assumptions that the 00:42:00.480 --> 00:42:03.599 AI gives you uh and then decide that yes 00:42:03.599 --> 00:42:05.440 this is a great assumption and actually 00:42:05.440 --> 00:42:07.599 no I want to go more in this direction. 00:42:07.599 --> 00:42:09.760 Um but you will it will definitely be 00:42:09.760 --> 00:42:12.480 frustrating in the token usage. Um you 00:42:12.480 --> 00:42:15.760 will definitely be like I have used 80 00:42:15.760 --> 00:42:18.560 like 30% of my session budget and I have 00:42:18.560 --> 00:42:20.160 not even written a line of code yet and 00:42:20.160 --> 00:42:21.839 I've only got these stupid spec 00:42:21.839 --> 00:42:23.520 documents. That will definitely be the 00:42:23.520 --> 00:42:24.720 first impression. 00:42:24.720 --> 00:42:26.960 >> Yeah, I I I like this. I mean this this 00:42:26.960 --> 00:42:29.200 seems like a 00:42:29.200 --> 00:42:31.280 I can't remember the name. Can can you 00:42:31.280 --> 00:42:33.920 can you dig up that talk from from 00:42:33.920 --> 00:42:35.280 Singapore Dev Days or whatever? 00:42:35.280 --> 00:42:37.680 >> Yeah, DevOps Day Singapore. um which I 00:42:37.680 --> 00:42:38.800 think this year is going to be by the 00:42:38.800 --> 00:42:40.960 end of the year. Um last year it was 00:42:40.960 --> 00:42:42.960 around May. So pretty much 00:42:42.960 --> 00:42:47.119 >> yeah the whole networking uh log VPC oh 00:42:47.119 --> 00:42:50.319 whatever capturing the the the the 00:42:50.319 --> 00:42:52.560 networking of um cuz what the lethal 00:42:52.560 --> 00:42:55.040 trifecta for for AI agents is private 00:42:55.040 --> 00:42:56.880 data access to the internet and 00:42:56.880 --> 00:42:59.119 untrusted content 00:42:59.119 --> 00:43:00.800 uh untrusted content if I remember 00:43:00.800 --> 00:43:02.720 correctly which is and untrusted content 00:43:02.720 --> 00:43:04.640 and internet access is almost the same 00:43:04.640 --> 00:43:07.119 thing. So I'm I'm thinking the network 00:43:07.119 --> 00:43:08.880 angle is something I want to investigate 00:43:08.880 --> 00:43:10.400 further. But in the meantime, I will 00:43:10.400 --> 00:43:12.480 specdriven development some sort of AI 00:43:12.480 --> 00:43:14.640 show guard rails. This is going to be my 00:43:14.640 --> 00:43:17.359 my my next vibe project. How do I I 00:43:17.359 --> 00:43:20.160 don't know how to use a cruel code 00:43:20.160 --> 00:43:21.440 anymore. What are you what are you 00:43:21.440 --> 00:43:22.960 doing? Thanks. Was there some other 00:43:22.960 --> 00:43:25.920 basic DevOps practices to add here that 00:43:25.920 --> 00:43:28.240 come to mind? I'm I'm drawing blanks 00:43:28.240 --> 00:43:30.000 here. 00:43:30.000 --> 00:43:34.960 or have having a a pipeline test gates 00:43:34.960 --> 00:43:37.280 can't hear you. You you muted yourself, 00:43:37.280 --> 00:43:38.000 Vincent. 00:43:38.000 --> 00:43:40.319 >> Ah, so earlier with my amazing jokes, 00:43:40.319 --> 00:43:41.440 you didn't hear any of them. 00:43:41.440 --> 00:43:44.000 >> Yeah, they were that good. Okay, thanks. 00:43:44.000 --> 00:43:45.200 I think that was a I think that was a 00:43:45.200 --> 00:43:47.839 productive session. I'm I'm I'm going to 00:43:47.839 --> 00:43:51.200 do some Spectre and development on a on 00:43:51.200 --> 00:43:53.680 a guardrails 00:43:53.680 --> 00:43:55.440 uh linting tool. Okay, 00:43:55.440 --> 00:43:57.119 >> I want to ramble off one more fun fact. 00:43:57.119 --> 00:43:57.440 Okay, 00:43:57.440 --> 00:44:00.560 >> there's a branch on the bun repository 00:44:00.560 --> 00:44:03.119 of a rewrite into Rust. Reddit and had 00:44:03.119 --> 00:44:04.880 like five or six threats going crazy 00:44:04.880 --> 00:44:06.880 about, oh my god, Antropic, they have so 00:44:06.880 --> 00:44:09.839 much money to have AI rewrite bun from 00:44:09.839 --> 00:44:11.599 Zigg into Rust. Haven't they ever heard 00:44:11.599 --> 00:44:14.640 of Dino? What's wrong with them? Um, and 00:44:14.640 --> 00:44:17.040 X I guess the same type of backlash. And 00:44:17.040 --> 00:44:19.520 the author of the branch came uh in one 00:44:19.520 --> 00:44:21.280 of those threads and said, "You guys are 00:44:21.280 --> 00:44:22.880 blowing this way out of proportion. I 00:44:22.880 --> 00:44:25.920 just want to know what a rust version of 00:44:25.920 --> 00:44:27.520 bun looks like. What's the memory 00:44:27.520 --> 00:44:29.839 profile? How does it act and so on. Uh 00:44:29.839 --> 00:44:31.440 but it's kind of interesting to that we 00:44:31.440 --> 00:44:32.800 are at a day and age that you can just 00:44:32.800 --> 00:44:34.720 go and let me rewrite this massive 00:44:34.720 --> 00:44:37.520 framework from zig into rust just to see 00:44:37.520 --> 00:44:38.800 what's the you know. 00:44:38.800 --> 00:44:39.119 >> Yeah. 00:44:39.119 --> 00:44:40.960 >> Yeah. But that those sort of experiments 00:44:40.960 --> 00:44:42.800 should be private. Why the hell is he 00:44:42.800 --> 00:44:43.760 publishing it for Christ 00:44:43.760 --> 00:44:46.319 >> sake? because um bun is an open source 00:44:46.319 --> 00:44:47.599 project and it's a branch on the 00:44:47.599 --> 00:44:50.640 opensource uh repository rewriting and 00:44:50.640 --> 00:44:52.319 he's doing and and I don't know I think 00:44:52.319 --> 00:44:54.240 I shared with you because he has this 00:44:54.240 --> 00:44:56.720 context document in the branch that says 00:44:56.720 --> 00:44:59.359 like this is how you rewrite um from 00:44:59.359 --> 00:45:01.680 zing 00:45:01.680 --> 00:45:04.400 >> super interesting because um that's one 00:45:04.400 --> 00:45:06.880 of my original AI that was my DevOps day 00:45:06.880 --> 00:45:09.520 talk last year by the way uh using AI to 00:45:09.520 --> 00:45:12.079 rewrite uh something from AWS CDK cloud 00:45:12.079 --> 00:45:15.760 form into CDK for Terraform. So to see 00:45:15.760 --> 00:45:17.599 that how the rules that he he defines 00:45:17.599 --> 00:45:20.720 are very very very cool. Um 00:45:20.720 --> 00:45:24.319 >> Right. Right. Right. Oh on the topic to 00:45:24.319 --> 00:45:26.240 have you just to rattle off a couple of 00:45:26.240 --> 00:45:28.160 things the 00:45:28.160 --> 00:45:30.960 for setting up um 00:45:30.960 --> 00:45:34.160 like a like a fresh AWS 00:45:34.160 --> 00:45:36.000 um 00:45:36.000 --> 00:45:38.640 landing zone like a whole organization, 00:45:38.640 --> 00:45:41.280 right? Would you say the Terraform 00:45:41.280 --> 00:45:44.400 tooling or the CDK uh tooling is about 00:45:44.400 --> 00:45:46.240 the same? It makes no difference like 00:45:46.240 --> 00:45:49.119 using AWS control tower, setting up your 00:45:49.119 --> 00:45:52.319 logging account and your cloud trail 00:45:52.319 --> 00:45:55.200 blah blah blah. Does like which IA 00:45:55.200 --> 00:45:57.599 solution would you go for? I think e 00:45:57.599 --> 00:45:59.680 either or would you say Terraform's 00:45:59.680 --> 00:46:01.359 better? Like especially if you had to 00:46:01.359 --> 00:46:03.920 like you know do do general things like 00:46:03.920 --> 00:46:06.880 uh you know link up with uh with octa 00:46:06.880 --> 00:46:08.319 and things like that. Does it make 00:46:08.319 --> 00:46:10.560 >> I think my my experience is dated when I 00:46:10.560 --> 00:46:13.839 use AWS control tower um I think four 00:46:13.839 --> 00:46:15.599 years ago it missed a lot of the 00:46:15.599 --> 00:46:18.640 features that I wanted. I have a mixed 00:46:18.640 --> 00:46:21.599 where I do like AWS cloud AWS 00:46:21.599 --> 00:46:23.520 organization and cloud formation stack 00:46:23.520 --> 00:46:26.640 set which allow you to target a child or 00:46:26.640 --> 00:46:29.599 member account based on the uh unit like 00:46:29.599 --> 00:46:32.160 organization unit ou. So you can apply a 00:46:32.160 --> 00:46:34.000 stack set basically you can define the 00:46:34.000 --> 00:46:35.599 cloud for stack at the organization 00:46:35.599 --> 00:46:38.000 level and then on these organization 00:46:38.000 --> 00:46:40.000 units automatically deploy them. 00:46:40.000 --> 00:46:41.839 >> Sure you're using cloud for I guess 00:46:41.839 --> 00:46:43.680 >> no bootstrapping. So I said I have a mix 00:46:43.680 --> 00:46:46.640 setup like I use AWS or I don't use 00:46:46.640 --> 00:46:49.359 control tower um because of the landing 00:46:49.359 --> 00:46:51.440 zone uh product was pretty bad 00:46:51.440 --> 00:46:53.680 >> as another project I can't remember. 00:46:53.680 --> 00:46:54.079 Sorry. 00:46:54.079 --> 00:46:56.160 >> What is AWS or? 00:46:56.160 --> 00:46:58.640 >> So AWS or is where you are able to 00:46:58.640 --> 00:47:00.240 manage multiple AWS account 00:47:00.240 --> 00:47:02.400 >> under Yeah. Yeah. Yeah. Sorry. I thought 00:47:02.400 --> 00:47:03.839 you were referring to some sort of 00:47:03.839 --> 00:47:05.040 framework or something. Sorry. 00:47:05.040 --> 00:47:07.280 >> No, no, no. But but that all of those 00:47:07.280 --> 00:47:10.000 services have evolved so much um that I 00:47:10.000 --> 00:47:10.720 think 00:47:10.720 --> 00:47:12.240 >> it's the kind of thing you don't do very 00:47:12.240 --> 00:47:14.880 often. um usually and also that's 00:47:14.880 --> 00:47:17.119 another very like an argument that that 00:47:17.119 --> 00:47:19.119 is very hard for me to argue against 00:47:19.119 --> 00:47:20.800 which is like yeah but we only have five 00:47:20.800 --> 00:47:22.240 accounts and we only on boarding them 00:47:22.240 --> 00:47:25.520 one one every year or one every quarter 00:47:25.520 --> 00:47:27.359 so why do we need to have this all 00:47:27.359 --> 00:47:29.839 automated so my my personal preference 00:47:29.839 --> 00:47:32.240 right now that I have that I use is I 00:47:32.240 --> 00:47:35.119 use stack set to bootstrap the terraform 00:47:35.119 --> 00:47:38.640 because I I don't use tacos uh hosted 00:47:38.640 --> 00:47:41.040 service I use Atlantis and Terraform 00:47:41.040 --> 00:47:43.280 with with my own setup you're using when 00:47:43.280 --> 00:47:44.880 you say stack set you're using like the 00:47:44.880 --> 00:47:47.119 AWS provided stuff I guess right 00:47:47.119 --> 00:47:50.240 >> so stack sets are an AWS orc cloud for 00:47:50.240 --> 00:47:52.560 feature you have to give the I think you 00:47:52.560 --> 00:47:54.480 have to give a service linked account uh 00:47:54.480 --> 00:47:56.800 linked role permission uh you have to 00:47:56.800 --> 00:47:59.359 enable cloud for service to run AC 00:47:59.359 --> 00:48:01.599 across your AWS org so that it can go in 00:48:01.599 --> 00:48:03.599 into different accounts and do things so 00:48:03.599 --> 00:48:05.839 I use that to bootstrap my accounts 00:48:05.839 --> 00:48:08.319 automatically they get bootstrapped with 00:48:08.319 --> 00:48:10.319 uh an IM role that has a trust 00:48:10.319 --> 00:48:12.800 relationship with the AWS account where 00:48:12.800 --> 00:48:15.280 my my Terraform automation lives. So I 00:48:15.280 --> 00:48:16.800 have one which usually is called like 00:48:16.800 --> 00:48:19.280 managed services or shared services 00:48:19.280 --> 00:48:21.760 account or they call it in in AWS CDK 00:48:21.760 --> 00:48:24.079 you have this as well. AWS CDK pipelines 00:48:24.079 --> 00:48:26.720 uses the DevOps account as as a as an 00:48:26.720 --> 00:48:28.079 idea. That's where your pipeline is 00:48:28.079 --> 00:48:29.839 defined and it runs across all of your 00:48:29.839 --> 00:48:31.280 other accounts. So I have the same 00:48:31.280 --> 00:48:33.359 setup. I have one account where where 00:48:33.359 --> 00:48:37.119 the um the pipeline lives and all of the 00:48:37.119 --> 00:48:38.640 other accounts when they're on boarded 00:48:38.640 --> 00:48:40.240 they trust they create a trust 00:48:40.240 --> 00:48:41.839 relationship with that account. 00:48:41.839 --> 00:48:44.480 Have you had any things that like like 00:48:44.480 --> 00:48:46.559 easily bootstrap Atlantis and things 00:48:46.559 --> 00:48:48.000 like that or is that just 00:48:48.000 --> 00:48:50.400 >> Yeah, I do. And I I so part of my terra 00:48:50.400 --> 00:48:52.880 constructs uh project was to basically 00:48:52.880 --> 00:48:55.200 come to the point that I publish a a 00:48:55.200 --> 00:48:58.800 terraform mod terf or a CDK terrain 00:48:58.800 --> 00:49:00.960 construct that just you just run it and 00:49:00.960 --> 00:49:03.839 it just bootstraps your whole um account 00:49:03.839 --> 00:49:06.000 and terraform automation. 00:49:06.000 --> 00:49:07.520 >> That would be cool. I mean like people 00:49:07.520 --> 00:49:09.680 ask me about this and I'm like yeah use 00:49:09.680 --> 00:49:11.839 control tower. How do you do Atlantis? 00:49:11.839 --> 00:49:13.920 Uh do it manually? 00:49:13.920 --> 00:49:16.240 >> I feel Atlantis is lagging behind in a 00:49:16.240 --> 00:49:18.319 lot of features and it's designed in a 00:49:18.319 --> 00:49:20.400 very rigid way. Like it's very hard to 00:49:20.400 --> 00:49:22.720 to make changes like even just making a 00:49:22.720 --> 00:49:24.720 drift detection with the API endpoint is 00:49:24.720 --> 00:49:26.640 kind of really is broken. So I use 00:49:26.640 --> 00:49:28.480 Atlantis but I don't think it's a very 00:49:28.480 --> 00:49:29.839 good solution. I think there might be 00:49:29.839 --> 00:49:31.440 other like I don't know maybe state 00:49:31.440 --> 00:49:33.680 graph if there but there should be more 00:49:33.680 --> 00:49:35.760 like a better automation solution for 00:49:35.760 --> 00:49:36.319 that. But 00:49:36.319 --> 00:49:38.400 >> yeah, agreed. But as you said, like you 00:49:38.400 --> 00:49:40.319 you generally only do this like once. 00:49:40.319 --> 00:49:41.920 How often do you need to do it? Okay, 00:49:41.920 --> 00:49:43.520 cool. Um, I thought that was a 00:49:43.520 --> 00:49:45.920 productive session. I will I will pledge 00:49:45.920 --> 00:49:49.839 to do some some sedd for you to review 00:49:49.839 --> 00:49:52.000 in a bit. Have a great day. See you. 00:49:52.000 --> 00:49:53.520 >> You would have a lot better experience 00:49:53.520 --> 00:49:55.359 if you use spec ledger, but maybe you 00:49:55.359 --> 00:49:57.119 should use specket first to know. 00:49:57.119 --> 00:49:58.960 >> Yeah, I'll use packet first and then 00:49:58.960 --> 00:50:03.839 then we can exchange notes. Okay. See?