WEBVTT 00:00:02.399 --> 00:00:06.400 Wow, I can see everything clearly now. 00:00:06.400 --> 00:00:06.960 Are 00:00:06.960 --> 00:00:09.120 >> are you do you have uh what do you call 00:00:09.120 --> 00:00:11.200 it? By focal 00:00:11.200 --> 00:00:12.160 now because you're 00:00:12.160 --> 00:00:15.679 >> I have no idea. I have I I don't I need 00:00:15.679 --> 00:00:19.840 to wear these for um for at the desk. 00:00:19.840 --> 00:00:22.960 It's different from far sight. Like I I 00:00:22.960 --> 00:00:25.039 don't need them outside. I can perfectly 00:00:25.039 --> 00:00:27.039 see clear in the distance, but not not 00:00:27.039 --> 00:00:29.920 like 70 cm. 00:00:29.920 --> 00:00:30.960 Um, 00:00:30.960 --> 00:00:32.880 >> yeah, I actually need glasses for a 00:00:32.880 --> 00:00:35.520 while. But now, you know, when I put 00:00:35.520 --> 00:00:37.360 them on and I see the screen, then I 00:00:37.360 --> 00:00:40.879 realize how bad it was, like how blurry 00:00:40.879 --> 00:00:43.600 actually my I was looking at. 00:00:43.600 --> 00:00:45.360 >> There's a a way around that, like just 00:00:45.360 --> 00:00:47.200 having a bigger screen and bigger fonts 00:00:47.200 --> 00:00:47.840 and 00:00:47.840 --> 00:00:50.399 >> Yeah. Yeah. Yeah. I definitely had like 00:00:50.399 --> 00:00:52.239 situations where I woke up and I just 00:00:52.239 --> 00:00:53.920 was like, I can't be bothered with this 00:00:53.920 --> 00:00:56.399 right now. Just up the up the font size. 00:00:56.399 --> 00:00:58.079 And then later on during the day didn't 00:00:58.079 --> 00:00:59.760 even realize that I've been, you know, 00:00:59.760 --> 00:01:03.039 working with the font size like this 00:01:03.039 --> 00:01:06.080 big. 00:01:06.080 --> 00:01:09.360 >> Cool, man. So, uh, yeah. How so now 00:01:09.360 --> 00:01:11.360 you're you can see better, you can feel 00:01:11.360 --> 00:01:14.560 better, you can work better. 00:01:14.560 --> 00:01:16.240 >> It's like I'm wearing binoculars, dude. 00:01:16.240 --> 00:01:18.320 It's like like like the screen is like 00:01:18.320 --> 00:01:21.360 suddenly right in front of my face. 00:01:21.360 --> 00:01:23.200 >> What? I think I have the opposite 00:01:23.200 --> 00:01:26.400 problem. I I I see better near without 00:01:26.400 --> 00:01:29.759 my glasses. 00:01:29.759 --> 00:01:34.159 >> Yeah. H So I was just trying to add to 00:01:34.159 --> 00:01:37.600 the to the AI infrastructure podcast 00:01:37.600 --> 00:01:40.400 notes about what to talk about. I wish I 00:01:40.400 --> 00:01:42.400 did this a bit earlier. 00:01:42.400 --> 00:01:43.680 >> I haven't really been adding anything 00:01:43.680 --> 00:01:45.200 there for a while. 00:01:45.200 --> 00:01:45.920 >> There's been a lot 00:01:45.920 --> 00:01:47.439 >> quite a lot of announcements, right? 00:01:47.439 --> 00:01:48.640 Yeah, 00:01:48.640 --> 00:01:52.399 >> there's been a lot. Um, I guess you're 00:01:52.399 --> 00:01:54.640 >> I don't have a lot of time, though. 00:01:54.640 --> 00:01:56.720 >> Oh, you're you're all in on Claude, so I 00:01:56.720 --> 00:01:58.399 guess you haven't even bothered with uh 00:01:58.399 --> 00:02:00.320 Open AI stuff. 00:02:00.320 --> 00:02:02.960 >> They gave me cuz I was one of those 00:02:02.960 --> 00:02:05.759 people that unsubscribed after the whole 00:02:05.759 --> 00:02:09.759 um Department of War situation. 00:02:09.759 --> 00:02:10.800 >> Oh, 00:02:10.800 --> 00:02:13.680 >> I I went to Chat GPT and cancelled, but 00:02:13.680 --> 00:02:14.959 then they said, "We'll give you a month 00:02:14.959 --> 00:02:16.560 for free, so I still have access for a 00:02:16.560 --> 00:02:18.879 month." And now people are all raving 00:02:18.879 --> 00:02:22.319 about uh for 5.4 being um you know 00:02:22.319 --> 00:02:24.720 fine-tuned for coding and being so much 00:02:24.720 --> 00:02:25.440 better. 00:02:25.440 --> 00:02:27.920 >> Well well I don't think they've released 00:02:27.920 --> 00:02:31.840 GBT 5.4 for Codex. Codex is the one 00:02:31.840 --> 00:02:32.640 that's optimized, 00:02:32.640 --> 00:02:34.720 >> but they believe they believe it won't 00:02:34.720 --> 00:02:36.480 because they basically integrated the 00:02:36.480 --> 00:02:39.440 fine tuning from Codex into it and kind 00:02:39.440 --> 00:02:42.560 of reduce reduce the confusion 00:02:42.560 --> 00:02:45.200 >> of of the all of the Yeah, that's at 00:02:45.200 --> 00:02:48.080 least that's Tio's theory, right? I did 00:02:48.080 --> 00:02:50.160 watch The first 10 minutes. 00:02:50.160 --> 00:02:55.200 >> This is where you get the news. 00:02:55.200 --> 00:02:56.160 >> Yeah. Well, that's 00:02:56.160 --> 00:02:58.000 >> it. To me it makes sense 00:02:58.000 --> 00:03:00.080 >> because like all of this additional 00:03:00.080 --> 00:03:01.680 terminology people were getting super 00:03:01.680 --> 00:03:03.840 confused if you were open opening your 00:03:03.840 --> 00:03:07.440 your LLM model selection it was like 20 00:03:07.440 --> 00:03:09.519 GPT options 00:03:09.519 --> 00:03:14.239 >> GPT 5.2 two 5.3 5.3 thinking 5.3 codex 00:03:14.239 --> 00:03:17.120 5.3 codex ultra think 00:03:17.120 --> 00:03:19.519 >> it's like what are you what are you 00:03:19.519 --> 00:03:20.879 doing 00:03:20.879 --> 00:03:25.440 >> so on different models so so I wanted to 00:03:25.440 --> 00:03:28.560 ask you about do you use one model 00:03:28.560 --> 00:03:31.680 against another in your workflow I mean 00:03:31.680 --> 00:03:33.519 you showed me like months ago that you 00:03:33.519 --> 00:03:35.360 were doing something like that like 00:03:35.360 --> 00:03:38.720 >> yeah I was because I was getting 00:03:38.720 --> 00:03:39.920 >> uh sorry 00:03:39.920 --> 00:03:41.760 >> I was Because the problem was I was I 00:03:41.760 --> 00:03:43.280 was never I was always running out of 00:03:43.280 --> 00:03:44.959 credits on cloud, right? 00:03:44.959 --> 00:03:46.959 >> Oh, yeah. That's true. You're doing it 00:03:46.959 --> 00:03:48.959 as a means of necessity. 00:03:48.959 --> 00:03:52.239 >> Yeah. But now now I I I haven't even 00:03:52.239 --> 00:03:55.280 been able to, you know, hit my budget 00:03:55.280 --> 00:03:56.640 limits on cloud. 00:03:56.640 --> 00:03:59.200 >> Yeah. So that's interesting. So you were 00:03:59.200 --> 00:04:01.280 doing it you were using multi- aents 00:04:01.280 --> 00:04:03.840 because you're running out of credits 00:04:03.840 --> 00:04:05.599 though. So, I'm I'm thinking using 00:04:05.599 --> 00:04:07.120 multiple 00:04:07.120 --> 00:04:09.920 multiple different models might be 00:04:09.920 --> 00:04:11.760 >> constraint. 00:04:11.760 --> 00:04:13.599 >> Yeah, it might be a quality game because 00:04:13.599 --> 00:04:15.280 I'm I was actually a little bit shocked. 00:04:15.280 --> 00:04:19.440 I had this experience last week where I 00:04:19.440 --> 00:04:24.080 I uh I created a a PR. Of course, other 00:04:24.080 --> 00:04:27.120 people approved it without asking too 00:04:27.120 --> 00:04:29.120 many questions or I mean, it's human 00:04:29.120 --> 00:04:31.840 approval. What do you expect? 00:04:31.840 --> 00:04:33.360 your track record 00:04:33.360 --> 00:04:36.639 >> and and then I noticed that um like if I 00:04:36.639 --> 00:04:38.960 was if I was actually less trustful of 00:04:38.960 --> 00:04:41.759 AI and I honestly looked at this patch 00:04:41.759 --> 00:04:43.440 that I was contributing I would have 00:04:43.440 --> 00:04:45.440 realized it was a terrible mistake it 00:04:45.440 --> 00:04:47.360 was such an obvious mistake really like 00:04:47.360 --> 00:04:50.000 I I used env instead of environment for 00:04:50.000 --> 00:04:52.720 GitHub workflow and of course I didn't 00:04:52.720 --> 00:04:55.360 really test it because it was a kind of 00:04:55.360 --> 00:04:58.000 a trivial change but anyway long story 00:04:58.000 --> 00:05:02.080 short this this PR that I pushed that 00:05:02.080 --> 00:05:06.240 was AI generated was wrong and now 00:05:06.240 --> 00:05:08.639 that's just making me very it's just 00:05:08.639 --> 00:05:10.639 sort of like wrapped ripped my 00:05:10.639 --> 00:05:12.960 confidence apart 00:05:12.960 --> 00:05:14.960 and uh I'm just thinking like what can I 00:05:14.960 --> 00:05:16.960 what like what practical steps can I do 00:05:16.960 --> 00:05:19.600 to make sure I don't 00:05:19.600 --> 00:05:22.720 commit junk 00:05:22.720 --> 00:05:25.600 this has always been the job of uh CI/CD 00:05:25.600 --> 00:05:27.520 DevOps people like my job pretty much 00:05:27.520 --> 00:05:29.199 all the time is to set up llinters 00:05:29.199 --> 00:05:31.520 pre-commit and all this type of stuff. 00:05:31.520 --> 00:05:33.039 Whatever I can deterministically 00:05:33.039 --> 00:05:33.840 validate. 00:05:33.840 --> 00:05:36.639 >> Yeah, that that's that's the trick. 00:05:36.639 --> 00:05:39.919 Making AI deterministic or making what 00:05:39.919 --> 00:05:42.320 AI does deterministic though. The 00:05:42.320 --> 00:05:45.680 trouble is is that maybe I'm just 00:05:45.680 --> 00:05:48.639 casting shade on GitHub, but like GitHub 00:05:48.639 --> 00:05:52.320 actions is such a pain. 00:05:52.320 --> 00:05:55.280 The difficulty now there's action lint, 00:05:55.280 --> 00:05:57.199 but it doesn't capture like half the 00:05:57.199 --> 00:06:00.639 problems. it doesn't support like 00:06:00.639 --> 00:06:02.800 actions and 00:06:02.800 --> 00:06:06.000 >> I had the similar situation on Friday 00:06:06.000 --> 00:06:12.720 where I was like I was working on 00:06:12.720 --> 00:06:17.280 something related to a docker image and 00:06:17.280 --> 00:06:19.520 yes I remember uh we need to do 00:06:19.520 --> 00:06:24.319 immutable digest pins on the um docker 00:06:24.319 --> 00:06:26.720 base images so when we build using a 00:06:26.720 --> 00:06:29.919 >> file you're in a environment are what? 00:06:29.919 --> 00:06:33.120 >> Yeah. So, so we we we have like the .NET 00:06:33.120 --> 00:06:36.880 4 8.0 tag which is immutable because 00:06:36.880 --> 00:06:38.560 security patches and so on, right? When 00:06:38.560 --> 00:06:40.080 Microsoft builds a new version, they 00:06:40.080 --> 00:06:42.720 just push it under that 8.0 tag even 00:06:42.720 --> 00:06:43.600 though 00:06:43.600 --> 00:06:45.840 >> it's a different image. So, they don't 00:06:45.840 --> 00:06:47.680 accept that. So, they were like, you 00:06:47.680 --> 00:06:50.240 need to use an immutable digest. 00:06:50.240 --> 00:06:51.759 >> And then I was like, yeah, that's a pain 00:06:51.759 --> 00:06:53.360 because now we have to do a schedule to 00:06:53.360 --> 00:06:55.120 always check if there's a new digest and 00:06:55.120 --> 00:06:57.440 bump our base images, right? It's kind 00:06:57.440 --> 00:06:59.599 of like security like check checklist 00:06:59.599 --> 00:07:01.840 security like do we're consuming from 00:07:01.840 --> 00:07:04.800 Microsoft. Uh 00:07:04.800 --> 00:07:06.560 >> yeah sure there's supply chain attacks 00:07:06.560 --> 00:07:08.400 that can you know 00:07:08.400 --> 00:07:09.680 >> I'm not a I'm not a fan of these 00:07:09.680 --> 00:07:11.280 immutusable 00:07:11.280 --> 00:07:14.880 hashes on on docker things 00:07:14.880 --> 00:07:16.880 >> but like if Microsoft's been compromised 00:07:16.880 --> 00:07:19.520 and their and their 8.0.NET net image is 00:07:19.520 --> 00:07:21.919 is insecure and therefore you're saying 00:07:21.919 --> 00:07:23.759 we cannot have I think this is the type 00:07:23.759 --> 00:07:26.400 of nonsense um you know 00:07:26.400 --> 00:07:28.080 >> type of security because you're actually 00:07:28.080 --> 00:07:30.000 aggravating the issue because now you 00:07:30.000 --> 00:07:33.759 expect the client to adequately keep up 00:07:33.759 --> 00:07:34.400 up to 00:07:34.400 --> 00:07:38.000 >> yeah anyway the point was that um with 00:07:38.000 --> 00:07:40.240 Sun we had built a beautiful GitHub 00:07:40.240 --> 00:07:43.520 workflow that was tested against uh brew 00:07:43.520 --> 00:07:47.440 installed uh docker buildex image tools 00:07:47.440 --> 00:07:49.599 to inspect the digest for the remote 00:07:49.599 --> 00:07:51.440 image to figure out if we were using the 00:07:51.440 --> 00:07:54.400 if there was a new digest. And before I 00:07:54.400 --> 00:07:57.680 actually, you know, remove moved the PR 00:07:57.680 --> 00:08:00.639 from draft to review ready, I asked um 00:08:00.639 --> 00:08:03.520 Sun, hey, you know, we can test this. We 00:08:03.520 --> 00:08:06.639 can run the Docker Ubuntu image here and 00:08:06.639 --> 00:08:08.479 we can run the the commands that you 00:08:08.479 --> 00:08:10.319 have in the in the in the GitHub actions 00:08:10.319 --> 00:08:11.919 workflow. You can we can run it inside 00:08:11.919 --> 00:08:12.960 the local Docker image. 00:08:12.960 --> 00:08:13.280 >> Yeah. 00:08:13.280 --> 00:08:15.120 >> Ubuntu. and it did that and it 00:08:15.120 --> 00:08:17.120 immediately broke and there were all 00:08:17.120 --> 00:08:19.680 kinds of problems with it because um it 00:08:19.680 --> 00:08:23.199 was a multiarchc um digest and it wasn't 00:08:23.199 --> 00:08:24.879 detecting the digest properly. it was 00:08:24.879 --> 00:08:27.680 using uh a template string that didn't 00:08:27.680 --> 00:08:29.120 work or something or 00:08:29.120 --> 00:08:30.800 >> Yeah. And then also it was using the 00:08:30.800 --> 00:08:33.599 docker build 0.19 whereas in in the 00:08:33.599 --> 00:08:36.000 docker um that was on my local then it 00:08:36.000 --> 00:08:38.240 it says oh it fails I I should do an a 00:08:38.240 --> 00:08:39.760 instead of trying to use a temp because 00:08:39.760 --> 00:08:41.519 you know when you do golang like docker 00:08:41.519 --> 00:08:43.360 you can do these template strings to 00:08:43.360 --> 00:08:46.000 extract certain fields. So you can you 00:08:46.000 --> 00:08:48.000 can do docker inspect and then you can 00:08:48.000 --> 00:08:49.519 pass in a template string and it will 00:08:49.519 --> 00:08:52.240 expect it. Uh but the thing was it 00:08:52.240 --> 00:08:53.519 didn't work and then I assumed it was 00:08:53.519 --> 00:08:54.640 because it was using the wrong version 00:08:54.640 --> 00:08:56.640 and I said you're using 0.19 the latest 00:08:56.640 --> 00:08:58.720 is 0.26 and the GitHub action is using 00:08:58.720 --> 00:09:00.720 026 and they tried that it didn't work. 00:09:00.720 --> 00:09:02.720 Anyway the point is yeah I got so much 00:09:02.720 --> 00:09:05.279 things wrong and even even today when I 00:09:05.279 --> 00:09:07.600 was like um let's you use a make target. 00:09:07.600 --> 00:09:09.360 So this is another very stupid security 00:09:09.360 --> 00:09:11.920 problem where we have sonar cube that 00:09:11.920 --> 00:09:16.720 lints bashcript and I'm vendoring a v a 00:09:16.720 --> 00:09:18.399 third party supplier 00:09:18.399 --> 00:09:20.240 >> vendoring okay vendor 00:09:20.240 --> 00:09:22.000 >> vendoring meaning I'm inlining it into 00:09:22.000 --> 00:09:23.839 my git because I don't want to do a curl 00:09:23.839 --> 00:09:26.880 during my build I want to use an install 00:09:26.880 --> 00:09:29.279 script and when in my pull request it 00:09:29.279 --> 00:09:32.800 was flagged with 32 um sonar cube issues 00:09:32.800 --> 00:09:33.680 and 00:09:33.680 --> 00:09:35.760 >> if I fix them then the next time I pull 00:09:35.760 --> 00:09:37.920 the vendor next time I update from the 00:09:37.920 --> 00:09:39.040 vendor script, I'm going to have the 00:09:39.040 --> 00:09:41.040 same problem. Right? So, I want to like 00:09:41.040 --> 00:09:42.880 allow list it. I can't say whitelist, 00:09:42.880 --> 00:09:45.920 right? I have to say allow listed. 00:09:45.920 --> 00:09:46.880 >> Thank you. 00:09:46.880 --> 00:09:47.600 >> I'm very 00:09:47.600 --> 00:09:49.600 >> The audience appreciates this. 00:09:49.600 --> 00:09:52.240 >> Yeah. Um, so so I have to I have I I 00:09:52.240 --> 00:09:53.760 have to the I mean to me the right way 00:09:53.760 --> 00:09:56.560 is to vendor it and get track it allow 00:09:56.560 --> 00:09:59.279 list it. Um, but that's going to take 00:09:59.279 --> 00:10:01.279 time because I'm in this organization 00:10:01.279 --> 00:10:03.839 not an admin. So I I have to wait for 00:10:03.839 --> 00:10:05.440 someone else to whitelist it. So anyway, 00:10:05.440 --> 00:10:09.200 I says let's let's use a make target and 00:10:09.200 --> 00:10:11.279 um you know let's make sure that we 00:10:11.279 --> 00:10:12.880 actually do the insecure thing which is 00:10:12.880 --> 00:10:16.480 we we um we curl it and with make 00:10:16.480 --> 00:10:18.640 targets you can specify a file on disk 00:10:18.640 --> 00:10:20.320 and if it doesn't exist 00:10:20.320 --> 00:10:22.399 >> it will basically run the commands right 00:10:22.399 --> 00:10:24.480 I mean they it will keep it even like 00:10:24.480 --> 00:10:26.399 >> I'm the number one make file fan here 00:10:26.399 --> 00:10:28.240 >> yeah so but the thing is when I asked 00:10:28.240 --> 00:10:31.040 son or opus I think it was opus 4.6 six 00:10:31.040 --> 00:10:32.560 and I said, "Let's use a make target." 00:10:32.560 --> 00:10:35.360 It came up with all these phony make 00:10:35.360 --> 00:10:37.760 targets to do a unsure local or 00:10:37.760 --> 00:10:40.240 whatever. And I was like, you don't need 00:10:40.240 --> 00:10:41.680 to do that. That's not how make file 00:10:41.680 --> 00:10:44.000 works. You can point it to a file 00:10:44.000 --> 00:10:44.800 target. 00:10:44.800 --> 00:10:45.920 >> And then it was like, oh, you're 00:10:45.920 --> 00:10:47.680 absolutely right. But you see, like 00:10:47.680 --> 00:10:49.680 because I know make file, I know what to 00:10:49.680 --> 00:10:51.440 tell it. Like I know this this makes no 00:10:51.440 --> 00:10:52.160 sense. 00:10:52.160 --> 00:10:54.000 >> It's funny how make files trip up a lot 00:10:54.000 --> 00:10:57.120 of people and and AI in some ways. But 00:10:57.120 --> 00:10:59.120 >> so I was I was I was all mighty. I was 00:10:59.120 --> 00:11:01.040 like, "Haha, look at AI. You can't be 00:11:01.040 --> 00:11:02.640 just a stupid wipe coder. You have to 00:11:02.640 --> 00:11:04.320 know how make files work." At the same 00:11:04.320 --> 00:11:06.160 time, I was wondering the issue really 00:11:06.160 --> 00:11:08.240 is the make file, though. Like, if make 00:11:08.240 --> 00:11:10.320 file wasn't this so cryptic. 00:11:10.320 --> 00:11:12.240 >> Oh, have you heard of redo? By the way, 00:11:12.240 --> 00:11:15.279 I sometimes use redo in my projects. 00:11:15.279 --> 00:11:17.120 >> I you were just talking about you you 00:11:17.120 --> 00:11:19.040 did your all your skills with just or 00:11:19.040 --> 00:11:20.320 something like that, right? 00:11:20.320 --> 00:11:23.920 >> No, not just. I used redo. It's a DGP. 00:11:23.920 --> 00:11:28.240 Anyway, so going back to the the quality 00:11:28.240 --> 00:11:30.560 thing. Okay, I I'm with you a thousand% 00:11:30.560 --> 00:11:32.880 that we should we should do less things 00:11:32.880 --> 00:11:35.519 in GitHub actions, do more make files. 00:11:35.519 --> 00:11:37.120 That's what I'm that's my mantra. 00:11:37.120 --> 00:11:38.959 >> That's not what I said, though. I just 00:11:38.959 --> 00:11:41.519 said that that we can validate some of 00:11:41.519 --> 00:11:45.600 it. Um, for example, 00:11:45.600 --> 00:11:47.680 >> we can validate, we can we we can ask 00:11:47.680 --> 00:11:50.240 the AI to do a little extra and say 00:11:50.240 --> 00:11:52.320 like, you know, it all looks good and we 00:11:52.320 --> 00:11:54.480 can push it and see it pass. But even 00:11:54.480 --> 00:11:56.880 then, we're not 100% sure. We can still 00:11:56.880 --> 00:11:59.760 use local Docker VMs with with Linux and 00:11:59.760 --> 00:12:02.399 and and try to use that to reproduce in 00:12:02.399 --> 00:12:04.480 Ubuntu. Oh, by the way, I wanted to say 00:12:04.480 --> 00:12:06.480 on the subject of pinning, the one good 00:12:06.480 --> 00:12:08.480 thing about pinning in a Docker image is 00:12:08.480 --> 00:12:10.639 that is that you're likely to speed up 00:12:10.639 --> 00:12:12.720 your Docker builds because 00:12:12.720 --> 00:12:14.320 >> uh using caches and and things. 00:12:14.320 --> 00:12:18.240 >> Yeah, you're more that's one sort of 00:12:18.240 --> 00:12:20.880 >> Yeah. of the churn. 00:12:20.880 --> 00:12:22.800 >> But you were basically you started 00:12:22.800 --> 00:12:25.519 saying that um it's hard to trust the AI 00:12:25.519 --> 00:12:27.920 outputs because it can we can still miss 00:12:27.920 --> 00:12:29.920 things and how do we validate it? Um 00:12:29.920 --> 00:12:31.120 >> yeah. So, 00:12:31.120 --> 00:12:33.680 >> but you have you seen Adam Jacob's 00:12:33.680 --> 00:12:39.040 latest LinkedIn posts about that? 00:12:39.040 --> 00:12:40.399 I think he's been saying the same thing 00:12:40.399 --> 00:12:42.639 like quite a few times which is like 00:12:42.639 --> 00:12:45.600 it's it's difficult to be in this like 00:12:45.600 --> 00:12:50.800 what's it called the the twilight zone 00:12:50.800 --> 00:12:53.360 the where you are still validating like 00:12:53.360 --> 00:12:55.120 you have a lot of manual code and you 00:12:55.120 --> 00:12:57.360 still want to control what the AI output 00:12:57.360 --> 00:13:00.000 is um and you're actually validating the 00:13:00.000 --> 00:13:01.920 code versus you're not validating the 00:13:01.920 --> 00:13:04.320 output you're not validating the result 00:13:04.320 --> 00:13:06.480 and and he had quite a few posts where 00:13:06.480 --> 00:13:08.639 he was going on about how people have a 00:13:08.639 --> 00:13:11.360 difficulty letting go of the code and I 00:13:11.360 --> 00:13:12.959 mean something we've been discussing a 00:13:12.959 --> 00:13:14.639 lot as well and I kept telling you like 00:13:14.639 --> 00:13:16.480 you know as long as we put the 00:13:16.480 --> 00:13:18.240 deterministic checks around it then we 00:13:18.240 --> 00:13:20.000 can uh do a lot more 00:13:20.000 --> 00:13:22.720 >> I don't understand the oh I think this 00:13:22.720 --> 00:13:24.560 is how you do it 00:13:24.560 --> 00:13:27.680 >> I don't understand LinkedIn 00:13:27.680 --> 00:13:28.800 >> there was a 00:13:28.800 --> 00:13:30.720 >> yeah I mean there's definitely like 00:13:30.720 --> 00:13:34.160 people who are into AI and cranking it 00:13:34.160 --> 00:13:34.880 out 00:13:34.880 --> 00:13:36.959 >> this one about Nick's like I've been 00:13:36.959 --> 00:13:38.399 programming for literally my entire 00:13:38.399 --> 00:13:41.440 life. There's um that one I think he's 00:13:41.440 --> 00:13:44.079 talking about the fact that it's really 00:13:44.079 --> 00:13:47.760 a big mind shift. Uh it wasn't that Adam 00:13:47.760 --> 00:13:49.680 actually a couple of reflections that 00:13:49.680 --> 00:13:52.720 one couple of reflections the third one 00:13:52.720 --> 00:13:55.600 below from Adam that one. Yeah. Yeah. 00:13:55.600 --> 00:13:57.519 that that's where he goes on and says 00:13:57.519 --> 00:13:59.920 like it's hard to if you have an 00:13:59.920 --> 00:14:02.720 existing large code base um it's hard to 00:14:02.720 --> 00:14:05.839 let AI go wild on that code base 00:14:05.839 --> 00:14:07.680 >> and it's more easy to just start from 00:14:07.680 --> 00:14:09.839 scratch and let AI build everything and 00:14:09.839 --> 00:14:11.440 just build all the loops around it and 00:14:11.440 --> 00:14:12.880 that make me think I think the first 00:14:12.880 --> 00:14:14.480 thing we need to do when we are adopting 00:14:14.480 --> 00:14:17.600 AI into a codebase is to ask AI to 00:14:17.600 --> 00:14:19.519 validate all of the user stories and 00:14:19.519 --> 00:14:21.760 build a very strong build hardness don't 00:14:21.760 --> 00:14:24.880 let AI build like modify the code let AI 00:14:24.880 --> 00:14:27.199 build all of the validation and make 00:14:27.199 --> 00:14:28.959 sure that the validation is correct 00:14:28.959 --> 00:14:31.279 before you let AI loose on the code. 00:14:31.279 --> 00:14:32.959 >> I think that a way to 00:14:32.959 --> 00:14:36.320 >> my my colleague um was using AI with 00:14:36.320 --> 00:14:39.519 some database flows and what I and he's 00:14:39.519 --> 00:14:42.880 a very seasoned colleague um he's been 00:14:42.880 --> 00:14:44.480 around the block. He's been he's been 00:14:44.480 --> 00:14:48.000 hacking since the 80s sort of thing. And 00:14:48.000 --> 00:14:49.519 I don't know, maybe he's listening to 00:14:49.519 --> 00:14:51.199 different podcasts and getting his 00:14:51.199 --> 00:14:53.440 information from different sources, but 00:14:53.440 --> 00:14:56.000 I was really impressed that he like what 00:14:56.000 --> 00:14:59.199 what do you what he did was that he he 00:14:59.199 --> 00:15:03.120 set up like a flow where he was he was 00:15:03.120 --> 00:15:05.440 using different personas like he was 00:15:05.440 --> 00:15:07.760 saying like you know I am a business 00:15:07.760 --> 00:15:10.639 analyst and I want this thing and then I 00:15:10.639 --> 00:15:13.680 am um a database designer and I want 00:15:13.680 --> 00:15:16.320 this thing. I am a QA engineer and I 00:15:16.320 --> 00:15:18.240 want this thing. And then he basically 00:15:18.240 --> 00:15:21.519 ran the problem through these three 00:15:21.519 --> 00:15:24.959 agents profiles to do his work. And I 00:15:24.959 --> 00:15:26.399 thought it was very impressive the way 00:15:26.399 --> 00:15:28.880 he he did that. I mean, have you have 00:15:28.880 --> 00:15:31.760 you done have you thought about personas 00:15:31.760 --> 00:15:34.079 and in your spec ledger and things like 00:15:34.079 --> 00:15:34.959 that? Do you 00:15:34.959 --> 00:15:36.880 >> No, but there is quite a few. 00:15:36.880 --> 00:15:38.720 >> I am a quality assurance engineer. 00:15:38.720 --> 00:15:40.800 >> Yeah. I mean, a lot of people don't like 00:15:40.800 --> 00:15:42.399 it. They say it's like it's like 00:15:42.399 --> 00:15:44.160 theatrics. It's like you're going to a 00:15:44.160 --> 00:15:45.920 play and it's like, you know, I'm 00:15:45.920 --> 00:15:47.680 playing house and balls with my like 00:15:47.680 --> 00:15:48.320 daughter. 00:15:48.320 --> 00:15:49.839 >> Yeah. Yeah. 00:15:49.839 --> 00:15:50.240 >> But 00:15:50.240 --> 00:15:53.440 >> though I do I I think the the probably 00:15:53.440 --> 00:15:55.519 the benefit really is that you have like 00:15:55.519 --> 00:15:57.279 a fresh 00:15:57.279 --> 00:16:00.399 context in two or three different agents 00:16:00.399 --> 00:16:01.920 just to cover. 00:16:01.920 --> 00:16:04.320 >> Oh yeah, for sure. I think I think it I 00:16:04.320 --> 00:16:06.079 think people underestimate how effective 00:16:06.079 --> 00:16:08.160 it is. I think because of these things 00:16:08.160 --> 00:16:12.160 are language models the context and you 00:16:12.160 --> 00:16:14.079 know that you create around them and 00:16:14.079 --> 00:16:17.199 clearing the context and having it with 00:16:17.199 --> 00:16:19.440 different perspective like you know 00:16:19.440 --> 00:16:22.000 creating these personas I think it makes 00:16:22.000 --> 00:16:23.759 sense I don't think we understand 00:16:23.759 --> 00:16:25.759 >> you've not you've not uh implemented 00:16:25.759 --> 00:16:28.160 >> I don't do that no I' also many people 00:16:28.160 --> 00:16:29.680 are yeah like I just said people don't 00:16:29.680 --> 00:16:32.160 really like that but I think it makes 00:16:32.160 --> 00:16:33.680 sense that I'm not going to do that at 00:16:33.680 --> 00:16:36.240 the moment for me right now the because 00:16:36.240 --> 00:16:37.680 There's quite a few announcements lately 00:16:37.680 --> 00:16:40.000 that really affect uh Spec Ledger's 00:16:40.000 --> 00:16:42.959 position a lot. One is uh Antropic 00:16:42.959 --> 00:16:45.759 announcing announcing like a um config 00:16:45.759 --> 00:16:47.839 management as a service so that you can 00:16:47.839 --> 00:16:50.959 share configurations uh for Claude Code. 00:16:50.959 --> 00:16:52.639 That's something that they launched. 00:16:52.639 --> 00:16:55.600 They added the scheduling. So people are 00:16:55.600 --> 00:16:58.240 saying ohropic just launched openclaw 00:16:58.240 --> 00:17:00.560 clone by making Claude Code run 00:17:00.560 --> 00:17:02.800 independently and can run schedule task 00:17:02.800 --> 00:17:04.640 which is a bit like one of the aspects 00:17:04.640 --> 00:17:06.400 of open claw. 00:17:06.400 --> 00:17:08.160 I'm actually surprised how long that 00:17:08.160 --> 00:17:10.720 it's it's taken like I still think 00:17:10.720 --> 00:17:12.720 >> Claude and Codex doesn't make it super 00:17:12.720 --> 00:17:15.760 easy to share things in an organization. 00:17:15.760 --> 00:17:18.160 >> Yeah. So so that's what Speckled Ledger 00:17:18.160 --> 00:17:20.079 does, right? I mean we have a superbase 00:17:20.079 --> 00:17:23.039 metadata where we are um allowing 00:17:23.039 --> 00:17:25.360 organizations to 00:17:25.360 --> 00:17:27.439 create skill bundles. I think Tessle 00:17:27.439 --> 00:17:28.799 apparently does that as well. I didn't 00:17:28.799 --> 00:17:30.320 know but like tessel they are not like 00:17:30.320 --> 00:17:32.799 we are the scale bundle 00:17:32.799 --> 00:17:35.120 >> uh solution to belittle anything but I 00:17:35.120 --> 00:17:36.480 don't think it's like a huge problem to 00:17:36.480 --> 00:17:38.960 share the prompts but like it it 00:17:38.960 --> 00:17:40.640 absolutely it's become an industry in 00:17:40.640 --> 00:17:40.880 itself 00:17:40.880 --> 00:17:42.640 >> it's just it's just one of those little 00:17:42.640 --> 00:17:44.400 things you know it's just one of those 00:17:44.400 --> 00:17:46.240 things you easily tack tack on if you're 00:17:46.240 --> 00:17:48.640 if you're going to provide a spec ledger 00:17:48.640 --> 00:17:50.559 flow and you're going to create a UI 00:17:50.559 --> 00:17:53.440 that that allows you to to visualize the 00:17:53.440 --> 00:17:55.919 process might as well includes you know 00:17:55.919 --> 00:17:58.240 skill bundles might as well include you 00:17:58.240 --> 00:17:59.679 part of the bootstrap of the of the 00:17:59.679 --> 00:18:01.840 repo. That's part of like 00:18:01.840 --> 00:18:03.840 >> some organizations they call it a AI 00:18:03.840 --> 00:18:06.240 readiness of a repository. Um across 00:18:06.240 --> 00:18:08.799 teams you need to set set up your repos 00:18:08.799 --> 00:18:10.559 so that they are ready for AI 00:18:10.559 --> 00:18:12.320 contribution so that anyone who starts 00:18:12.320 --> 00:18:14.480 working into it gets the best practices 00:18:14.480 --> 00:18:17.360 gets the the skills to to set up the you 00:18:17.360 --> 00:18:19.760 know everything. about talking about 00:18:19.760 --> 00:18:21.679 that and best practices and 00:18:21.679 --> 00:18:24.480 documentation. The one good thing at 00:18:24.480 --> 00:18:26.480 work is that it just took some 00:18:26.480 --> 00:18:31.200 convincing I feel took some took some um 00:18:31.200 --> 00:18:33.840 posturing but I I think I have the 00:18:33.840 --> 00:18:35.440 mandate to move everything from 00:18:35.440 --> 00:18:40.080 confluence or p confluence to markdown. 00:18:40.080 --> 00:18:41.440 I don't know if I've shown you this 00:18:41.440 --> 00:18:44.400 before but like yeah basically I feel 00:18:44.400 --> 00:18:45.760 this is going to make my life a lot 00:18:45.760 --> 00:18:47.039 easier. 00:18:47.039 --> 00:18:49.760 I hope you agree 00:18:49.760 --> 00:18:52.080 >> moving away from Atlassian is uh the 00:18:52.080 --> 00:18:55.200 sweetest uh AI move you can make really. 00:18:55.200 --> 00:18:56.880 >> I say that but then I spend the whole 00:18:56.880 --> 00:18:59.919 morning um create like playing with the 00:18:59.919 --> 00:19:03.200 J CLI Golang and writing a skill for but 00:19:03.200 --> 00:19:04.720 I'm right I wrote a skill very 00:19:04.720 --> 00:19:06.799 explicitly customized for all the custom 00:19:06.799 --> 00:19:08.400 fields and all the custom workflows for 00:19:08.400 --> 00:19:10.720 one board and one team that I'm in. 00:19:10.720 --> 00:19:14.640 >> You use the Atlassian um MCP. I know you 00:19:14.640 --> 00:19:17.120 don't like MCP. I don't like MCPS. 00:19:17.120 --> 00:19:18.240 >> It does work really well. This 00:19:18.240 --> 00:19:21.360 particular one, this super set. 00:19:21.360 --> 00:19:22.320 >> Okay. 00:19:22.320 --> 00:19:24.320 >> Though it's got some blind size like it 00:19:24.320 --> 00:19:26.480 doesn't work with 00:19:26.480 --> 00:19:29.120 uh what do you call these these these 00:19:29.120 --> 00:19:31.919 boards? White spaces. Not too sure what 00:19:31.919 --> 00:19:32.240 they call. 00:19:32.240 --> 00:19:33.520 >> Yeah. So, I guess it wouldn't work for 00:19:33.520 --> 00:19:34.960 me because that's exactly what I just 00:19:34.960 --> 00:19:37.200 did. I just took the Golang CLI that's 00:19:37.200 --> 00:19:40.080 like the most popular one and apparently 00:19:40.080 --> 00:19:42.640 hasn't received PR merge. So, I I forked 00:19:42.640 --> 00:19:44.240 it. I added the PR. 00:19:44.240 --> 00:19:46.400 Is it able to understand um a 00:19:46.400 --> 00:19:47.600 blackboard? I don't know what it's 00:19:47.600 --> 00:19:49.520 called in Atlas. Yeah. 00:19:49.520 --> 00:19:50.160 >> A white board. 00:19:50.160 --> 00:19:51.360 >> Maybe we're talking about a different 00:19:51.360 --> 00:19:52.960 thing like a conbon board is not the 00:19:52.960 --> 00:19:53.919 same as a whiteboard. 00:19:53.919 --> 00:19:55.120 >> Oh, no, no, no, no, no, 00:19:55.120 --> 00:19:56.080 >> no. 00:19:56.080 --> 00:19:57.440 >> This supports everything except a 00:19:57.440 --> 00:19:59.840 whiteboard and it it's really good and 00:19:59.840 --> 00:20:01.679 but I do notice that sometimes it 00:20:01.679 --> 00:20:04.080 struggles to do the markdown inside Jira 00:20:04.080 --> 00:20:06.799 because Jirro markdown has always been 00:20:06.799 --> 00:20:09.120 >> insane. I'm I I was very surprised 00:20:09.120 --> 00:20:10.559 because I always struggle with Jira 00:20:10.559 --> 00:20:12.640 markdown and then I see Cloud Sonnet or 00:20:12.640 --> 00:20:14.960 Opus create Jira tickets and with all 00:20:14.960 --> 00:20:16.960 theseh4 and all M's like what the hell 00:20:16.960 --> 00:20:18.880 is that and and then it just creates the 00:20:18.880 --> 00:20:22.080 tickets and I go oh that looks good. 00:20:22.080 --> 00:20:23.520 >> Okay. 00:20:23.520 --> 00:20:25.200 >> Okay. You need to share this skill with 00:20:25.200 --> 00:20:27.600 me but I'm pretty happy. I 00:20:27.600 --> 00:20:30.000 >> I asked a I asked Opus to highlight any 00:20:30.000 --> 00:20:32.320 sensitive info and it was like um 00:20:32.320 --> 00:20:34.320 enumeration of projects, enumeration of 00:20:34.320 --> 00:20:36.240 of statuses, enumeration like don't 00:20:36.240 --> 00:20:37.679 share this. I was like, "Okay, I won't 00:20:37.679 --> 00:20:39.760 share it." 00:20:39.760 --> 00:20:41.440 I was like, "Ai, can I open source 00:20:41.440 --> 00:20:43.360 this?" And and and it was like, "No." I 00:20:43.360 --> 00:20:45.120 was like, "Okay. 00:20:45.120 --> 00:20:47.280 >> I don't even ask myself anymore." 00:20:47.280 --> 00:20:49.919 >> Well, I I'm not too sure I've shared 00:20:49.919 --> 00:20:51.280 this with you before, but I'm looking 00:20:51.280 --> 00:20:53.360 forward to rewriting a lot of well, not 00:20:53.360 --> 00:20:56.799 rewriting, just dumping Confluence to to 00:20:56.799 --> 00:20:57.520 markdown. 00:20:57.520 --> 00:21:01.919 >> Nice. And then I suppose just to make it 00:21:01.919 --> 00:21:04.159 agent friendly 00:21:04.159 --> 00:21:07.280 uh to to make it like goal orientated so 00:21:07.280 --> 00:21:09.440 that uh 00:21:09.440 --> 00:21:12.240 >> yeah so that the agents will know what 00:21:12.240 --> 00:21:15.760 to do considering this this uh platform 00:21:15.760 --> 00:21:17.520 that I'm working on is compated. 00:21:17.520 --> 00:21:18.799 >> That's a big improvement. 00:21:18.799 --> 00:21:20.880 >> It's going to be a massive improvement I 00:21:20.880 --> 00:21:23.520 feel. um trying not to big it up too 00:21:23.520 --> 00:21:26.240 much, but it's just it's just so nice to 00:21:26.240 --> 00:21:28.400 work in a in a sensible way like with 00:21:28.400 --> 00:21:31.520 markdown and uh being able to easily 00:21:31.520 --> 00:21:33.200 edit things because just just the act of 00:21:33.200 --> 00:21:34.720 edit updating documentation and 00:21:34.720 --> 00:21:37.039 confluence is such a nightmare. I'm just 00:21:37.039 --> 00:21:41.919 sick of it for me. Um right now my focus 00:21:41.919 --> 00:21:44.960 is to maximize my output but also stay 00:21:44.960 --> 00:21:47.360 reasonable healthy. So, I'm more of like 00:21:47.360 --> 00:21:50.000 in the AI of Empire chart. I'm close to 00:21:50.000 --> 00:21:52.080 the first group of I'm not I'm not 00:21:52.080 --> 00:21:54.400 trying to burn out, but I'm trying to 00:21:54.400 --> 00:21:56.880 achieve slightly more than what I would 00:21:56.880 --> 00:21:59.840 be able to do if I was working full-time 00:21:59.840 --> 00:22:02.000 uh myself. So, I'm trying to use AI to 00:22:02.000 --> 00:22:04.880 to just hit more than what I my my 00:22:04.880 --> 00:22:06.720 regular output, but not insane. Not like 00:22:06.720 --> 00:22:09.840 10x um for for different 00:22:09.840 --> 00:22:12.159 >> Yeah, I guess. And then I try to I try 00:22:12.159 --> 00:22:14.559 to do this for the next few months 00:22:14.559 --> 00:22:17.039 before I don't know organizations 00:22:17.039 --> 00:22:18.799 >> every I'm feeling pretty ragged really 00:22:18.799 --> 00:22:21.039 in uh 00:22:21.039 --> 00:22:23.360 >> um though 00:22:23.360 --> 00:22:25.840 I I'm I am sort of trying to take a step 00:22:25.840 --> 00:22:27.760 back. I'm trying to be that uh force 00:22:27.760 --> 00:22:30.400 multiplier where I I on board my 00:22:30.400 --> 00:22:33.440 colleagues and and and with my learnings 00:22:33.440 --> 00:22:36.080 of AI, you know, just simple stuff like 00:22:36.080 --> 00:22:38.320 if we just put our documentation AI and 00:22:38.320 --> 00:22:40.799 then reference that a that documentation 00:22:40.799 --> 00:22:44.000 when we want something to be done, then 00:22:44.000 --> 00:22:45.760 we'll get there in a more reliable 00:22:45.760 --> 00:22:48.960 fashion. As for like building things, 00:22:48.960 --> 00:22:51.520 I'm definitely not on your uh wavelength 00:22:51.520 --> 00:22:54.080 at the moment just because 00:22:54.080 --> 00:22:54.400 well, 00:22:54.400 --> 00:22:56.159 >> I've taken a step back, I think, for the 00:22:56.159 --> 00:22:56.960 last uh 00:22:56.960 --> 00:22:58.400 >> Oh, that reminds me. I need to do that 00:22:58.400 --> 00:23:00.960 blog. 00:23:00.960 --> 00:23:02.640 >> I'll I'll do that. I'll do that blog 00:23:02.640 --> 00:23:03.280 this morning. 00:23:03.280 --> 00:23:06.159 >> SEO uh strategy. 00:23:06.159 --> 00:23:06.799 >> So funny. 00:23:06.799 --> 00:23:08.960 >> You should focus on CDKTF because I 00:23:08.960 --> 00:23:11.280 think it's there is a need for it. 00:23:11.280 --> 00:23:13.360 Actually, there is a need. Um 00:23:13.360 --> 00:23:17.360 >> yeah, I I just spent a weekend Oh yeah, 00:23:17.360 --> 00:23:17.840 doing well 00:23:17.840 --> 00:23:22.240 >> brainstorming on how to incorporate 00:23:22.240 --> 00:23:24.880 the L2 library. So I have Terra 00:23:24.880 --> 00:23:27.039 constructs right but I feel the name is 00:23:27.039 --> 00:23:29.600 too close to Terraform and and I feel 00:23:29.600 --> 00:23:33.120 the CDK terrain should have its own L2 00:23:33.120 --> 00:23:35.440 library. I I intend so terra constructs 00:23:35.440 --> 00:23:38.320 is more like a it's more like a proof of 00:23:38.320 --> 00:23:40.159 concept. It it shows that if you build 00:23:40.159 --> 00:23:41.919 those L2 constructs and you can port 00:23:41.919 --> 00:23:44.400 them from AW CDK it works really well. 00:23:44.400 --> 00:23:46.559 And now I want to port that effort and 00:23:46.559 --> 00:23:48.480 maybe start again but under a new name 00:23:48.480 --> 00:23:50.640 which is aligned with with with CDK 00:23:50.640 --> 00:23:52.559 terrain. So I was like brainstorming 00:23:52.559 --> 00:23:55.039 names like what is the L2 concept? Is it 00:23:55.039 --> 00:23:55.840 stratum? 00:23:55.840 --> 00:23:57.600 >> You must have had some level twos in 00:23:57.600 --> 00:23:59.200 there already somewhere. I just haven't 00:23:59.200 --> 00:23:59.760 looked 00:23:59.760 --> 00:24:01.280 >> in terrain. 00:24:01.280 --> 00:24:02.720 >> Yeah. 00:24:02.720 --> 00:24:05.440 >> Syndicate terrain 00:24:05.440 --> 00:24:07.360 doesn't really do level twos, right? It 00:24:07.360 --> 00:24:09.520 just does the init sorry it just does 00:24:09.520 --> 00:24:12.320 the initial L1 bindings and then Terra 00:24:12.320 --> 00:24:14.159 Construct is the full L2 library. And I 00:24:14.159 --> 00:24:15.919 always wanted to keep them separate 00:24:15.919 --> 00:24:17.840 >> cuz everyone has built their own. 00:24:17.840 --> 00:24:19.360 >> Yeah, when I think about it, you're 00:24:19.360 --> 00:24:20.400 right. 00:24:20.400 --> 00:24:22.080 >> There's a few like there's there's the 00:24:22.080 --> 00:24:24.799 concept of an archive. Um Oh, but you 00:24:24.799 --> 00:24:26.480 asked me a question. You asked me about 00:24:26.480 --> 00:24:27.919 like what is the difference between 00:24:27.919 --> 00:24:30.960 CDKTF CLI and why would you run the CLI 00:24:30.960 --> 00:24:33.679 versus running the same directly? 00:24:33.679 --> 00:24:35.360 >> And I was going on about it's the asset 00:24:35.360 --> 00:24:37.440 pipeline. It's the asset pipeline. 00:24:37.440 --> 00:24:39.200 >> Yeah. And you were going to you're going 00:24:39.200 --> 00:24:44.400 to show me how dumb I am about this. 00:24:44.400 --> 00:24:47.039 You caught me in a bad time. 00:24:47.039 --> 00:24:51.600 >> Well, okay. C can I just explain my my 00:24:51.600 --> 00:24:54.240 problem just so that you just let me 00:24:54.240 --> 00:24:56.400 just explain what I'm trying to do. We 00:24:56.400 --> 00:24:59.120 have a whole bunch of CDK and I want to 00:24:59.120 --> 00:25:01.279 refactor it so that 00:25:01.279 --> 00:25:03.360 >> is it CDK AWS CDK. 00:25:03.360 --> 00:25:06.320 >> Yeah, it's a CDK but I feel this is 00:25:06.320 --> 00:25:09.200 >> a CDK a problem that h would probably 00:25:09.200 --> 00:25:12.720 happen in CDKF whatever. So we we have 00:25:12.720 --> 00:25:14.799 we have many business divisions. They 00:25:14.799 --> 00:25:16.960 say there's 10 like it could be I don't 00:25:16.960 --> 00:25:18.880 know HR, 00:25:18.880 --> 00:25:23.120 commercial, whatever. Um, and each of 00:25:23.120 --> 00:25:26.799 them has their own dev staging prod 00:25:26.799 --> 00:25:28.320 and and with all these different 00:25:28.320 --> 00:25:31.120 domains, there's lots of shared uh CDK 00:25:31.120 --> 00:25:34.320 stacks and and and while we've been 00:25:34.320 --> 00:25:35.679 bootstrapping them, we've been just 00:25:35.679 --> 00:25:38.880 copying files across because SIM links 00:25:38.880 --> 00:25:42.640 are unfortunately a dark art. Um, and 00:25:42.640 --> 00:25:44.640 people don't want to use SIM links. So 00:25:44.640 --> 00:25:46.320 we've been copying files and then over 00:25:46.320 --> 00:25:48.799 time each file has been slightly changed 00:25:48.799 --> 00:25:50.799 or got some wide space or what have you. 00:25:50.799 --> 00:25:54.640 So we have like 10 sort of duplicated 00:25:54.640 --> 00:25:56.320 CDK 00:25:56.320 --> 00:25:59.520 uh apps I think is the correct term and 00:25:59.520 --> 00:26:03.600 we want to dduplicate it. So but already 00:26:03.600 --> 00:26:05.679 each each domain is quite complicated. I 00:26:05.679 --> 00:26:07.840 wanted to create like a like a shared 00:26:07.840 --> 00:26:10.400 library and then when I refactor some 00:26:10.400 --> 00:26:12.640 stuff into a shared library that when I 00:26:12.640 --> 00:26:16.640 do a CDK synth that I know it's the same 00:26:16.640 --> 00:26:18.799 in a sense because one thing that bit me 00:26:18.799 --> 00:26:21.440 with CDK refactoring is that I didn't 00:26:21.440 --> 00:26:23.760 realize that at least in Python that if 00:26:23.760 --> 00:26:26.320 you like put it in a different class 00:26:26.320 --> 00:26:28.480 then the name changes because 00:26:28.480 --> 00:26:28.880 >> Yeah. 00:26:28.880 --> 00:26:31.279 >> I mean that caught me out. So basically 00:26:31.279 --> 00:26:34.000 I'm having to rejig it again. 00:26:34.000 --> 00:26:36.000 >> Yeah. But what I I I don't want I mean I 00:26:36.000 --> 00:26:36.240 think 00:26:36.240 --> 00:26:37.679 >> I think that will be 00:26:37.679 --> 00:26:40.960 >> very difficult with with AWS CDK. 00:26:40.960 --> 00:26:43.440 >> What I'm trying to do is not have all 00:26:43.440 --> 00:26:45.760 these updates and deletes because I I 00:26:45.760 --> 00:26:48.080 just hate hate doing that. So 00:26:48.080 --> 00:26:49.679 >> So 00:26:49.679 --> 00:26:52.080 basically I'm doing a CDK sent and I'm 00:26:52.080 --> 00:26:53.679 like comparing it before and after to 00:26:53.679 --> 00:26:56.400 make sure that my refactor is is is a 00:26:56.400 --> 00:26:56.799 good idea. 00:26:56.799 --> 00:26:58.720 >> Exactly. So yeah I see refactoring which 00:26:58.720 --> 00:27:00.799 in terraform is a non-issue, right? 00:27:00.799 --> 00:27:02.320 because you change it and then you just 00:27:02.320 --> 00:27:04.400 put in a refactoring configuration and 00:27:04.400 --> 00:27:06.960 it takes care of the logical ID changes. 00:27:06.960 --> 00:27:08.559 What you just described is your logical 00:27:08.559 --> 00:27:10.320 ID of the resources change but 00:27:10.320 --> 00:27:12.000 physically the resources shouldn't 00:27:12.000 --> 00:27:13.840 change. They shouldn't be deleted and 00:27:13.840 --> 00:27:15.120 red recreated. Right? 00:27:15.120 --> 00:27:18.400 >> Yeah. Yeah. What in terform I've seen 00:27:18.400 --> 00:27:18.880 some people 00:27:18.880 --> 00:27:21.039 >> because Terraform state is accessible 00:27:21.039 --> 00:27:23.440 you can modify it. Cloud form is not. 00:27:23.440 --> 00:27:24.400 They added a lot of 00:27:24.400 --> 00:27:25.520 >> What do you mean like modify with an 00:27:25.520 --> 00:27:26.720 import or something like that? What do 00:27:26.720 --> 00:27:27.039 you 00:27:27.039 --> 00:27:28.720 >> Yeah, you can like basically for me, 00:27:28.720 --> 00:27:31.440 right, when I change my CDKTF 00:27:31.440 --> 00:27:33.360 drastically and I get the problem that 00:27:33.360 --> 00:27:35.279 you have, which is it's going to like 00:27:35.279 --> 00:27:37.360 change all the logical identities. Yeah. 00:27:37.360 --> 00:27:40.400 >> Um I I can I actually have a script that 00:27:40.400 --> 00:27:43.440 reads um I have two. 00:27:43.440 --> 00:27:45.200 >> I have one that reads the terapform plan 00:27:45.200 --> 00:27:46.799 diff and sees what it's going to create 00:27:46.799 --> 00:27:48.480 and what it's going to delete. It parses 00:27:48.480 --> 00:27:50.799 out the resource type and the logic ID 00:27:50.799 --> 00:27:52.480 and it tries to map them. If you have 00:27:52.480 --> 00:27:54.159 one resource type, exactly one create, 00:27:54.159 --> 00:27:56.080 one delete, it will map it one to one, 00:27:56.080 --> 00:27:57.600 right? So say that's one that's that's 00:27:57.600 --> 00:27:58.880 exactly the same one that you deleted 00:27:58.880 --> 00:28:00.640 and created and it's going to generate a 00:28:00.640 --> 00:28:02.640 move block for you. It's going to go if 00:28:02.640 --> 00:28:04.320 you have three deletes and three 00:28:04.320 --> 00:28:05.520 creates, it's going to sort them 00:28:05.520 --> 00:28:07.200 alphabetically and try to map them and 00:28:07.200 --> 00:28:08.960 say like those three they match. Most 00:28:08.960 --> 00:28:10.159 likely these are the way that they 00:28:10.159 --> 00:28:12.080 match. But maybe that's wrong. It's put 00:28:12.080 --> 00:28:14.000 a big fat comment in the top says review 00:28:14.000 --> 00:28:16.399 this because I could get this wrong. And 00:28:16.399 --> 00:28:18.320 then the the one if it's asynchronous 00:28:18.320 --> 00:28:19.919 like there's two deletes and there's uh 00:28:19.919 --> 00:28:21.760 one create then one of them is deleted 00:28:21.760 --> 00:28:23.279 one is kept. So which one do you want to 00:28:23.279 --> 00:28:25.120 create? You have to valid validate that. 00:28:25.120 --> 00:28:26.880 So that's what my Python script does 00:28:26.880 --> 00:28:28.880 that I created with Opus like three 00:28:28.880 --> 00:28:30.399 years ago. And you can do that very 00:28:30.399 --> 00:28:32.399 easily with Terraform. Um you can parse 00:28:32.399 --> 00:28:34.159 the plans you can generate those those 00:28:34.159 --> 00:28:36.399 refactor blocks. If you do that with AWS 00:28:36.399 --> 00:28:39.360 CDK it was three years ago very very 00:28:39.360 --> 00:28:41.279 hard. I literally people usually just 00:28:41.279 --> 00:28:43.200 delete the stacks and recreate them 00:28:43.200 --> 00:28:46.080 because modifying a ter cloud for stack 00:28:46.080 --> 00:28:50.399 is extremely hard but they did create a 00:28:50.399 --> 00:28:52.399 whole bunch of refactoring tools around 00:28:52.399 --> 00:28:55.200 cloud for that I never used in the last 00:28:55.200 --> 00:28:56.080 oh this is interesting 00:28:56.080 --> 00:28:57.919 >> I have no idea but yes there there's a 00:28:57.919 --> 00:29:00.159 bunch of announcements around that um so 00:29:00.159 --> 00:29:02.000 so this is where I would be looking if I 00:29:02.000 --> 00:29:04.159 was you um if you need to do these 00:29:04.159 --> 00:29:06.080 massive code refactoring and you can 00:29:06.080 --> 00:29:08.559 also Google for people facing issues 00:29:08.559 --> 00:29:10.960 with AWS cloud automation around uh and 00:29:10.960 --> 00:29:14.080 CDK specifically about logical ids 00:29:14.080 --> 00:29:16.159 because the moment you change things the 00:29:16.159 --> 00:29:18.480 logic ID changes and then cloudformation 00:29:18.480 --> 00:29:20.159 loses track of it and it says that 00:29:20.159 --> 00:29:21.840 doesn't exist anymore. 00:29:21.840 --> 00:29:23.679 >> I feel like an idiot. I I mean I've had 00:29:23.679 --> 00:29:24.960 this problem for a while and I didn't 00:29:24.960 --> 00:29:29.440 actually care to to do a basic AWS CDK 00:29:29.440 --> 00:29:30.080 >> refactor 00:29:30.080 --> 00:29:31.360 >> refactor search. 00:29:31.360 --> 00:29:32.720 >> I said three years but it's from 00:29:32.720 --> 00:29:34.720 September so it's not it's not just 10 00:29:34.720 --> 00:29:36.799 months. 00:29:36.799 --> 00:29:38.720 So this is the thing about cloud for 00:29:38.720 --> 00:29:41.200 >> three years and AI. 00:29:41.200 --> 00:29:43.440 >> Yeah. This is the problem with with with 00:29:43.440 --> 00:29:44.799 cloud form playing catch up with 00:29:44.799 --> 00:29:46.480 Terraform, right? I mean that's why I 00:29:46.480 --> 00:29:48.960 convince people that I think CDKTF is 00:29:48.960 --> 00:29:51.200 better than AWS CDK because these things 00:29:51.200 --> 00:29:54.080 only come into AWS like in cloud form 00:29:54.080 --> 00:29:55.919 now where Terraform started out with 00:29:55.919 --> 00:29:57.760 these things. Well, not really. The 00:29:57.760 --> 00:29:59.679 refactoring has only been there maybe 00:29:59.679 --> 00:30:00.399 for four years. 00:30:00.399 --> 00:30:02.720 >> Holy moly. Why didn't Why didn't I 00:30:02.720 --> 00:30:06.399 bloody Google? Yeah, but but come back 00:30:06.399 --> 00:30:08.240 to the reason why I was being so 00:30:08.240 --> 00:30:11.760 annoying to you about the pipeline. 00:30:11.760 --> 00:30:13.200 Um I can show you. 00:30:13.200 --> 00:30:15.440 >> I don't get it. Just don't get it. 00:30:15.440 --> 00:30:17.440 >> Yeah, I I'll show you what I think. 00:30:17.440 --> 00:30:18.799 Maybe we'll explain it, but then I have 00:30:18.799 --> 00:30:20.799 to go because I have another call. Um so 00:30:20.799 --> 00:30:23.840 I'll just share this real quick. So um 00:30:23.840 --> 00:30:25.279 can you see this like 00:30:25.279 --> 00:30:27.840 >> Oh yeah. So what I'm doing here is a 00:30:27.840 --> 00:30:30.799 comparison between CDKTF and AWS CDK 00:30:30.799 --> 00:30:31.840 because a lot of people don't really 00:30:31.840 --> 00:30:33.120 understand what's the difference between 00:30:33.120 --> 00:30:36.720 them. Right? So AWS CDK and CDKTF they 00:30:36.720 --> 00:30:39.039 both are a comment line interface and 00:30:39.039 --> 00:30:42.000 one orchestrates terraform execution a 00:30:42.000 --> 00:30:43.279 little bit and the other one 00:30:43.279 --> 00:30:45.840 orchestrates cloud form right you will 00:30:45.840 --> 00:30:50.000 agree to that right 00:30:50.000 --> 00:30:52.320 >> both of them have L1 constructs so when 00:30:52.320 --> 00:30:54.799 you use AWS CDK inside there's a whole 00:30:54.799 --> 00:30:56.720 library of the L1 resources that are 00:30:56.720 --> 00:30:59.200 directly the cloud for resources for 00:30:59.200 --> 00:31:01.520 CDKTF you have the same thing you can 00:31:01.520 --> 00:31:04.399 build the um L1 construct which are 00:31:04.399 --> 00:31:06.640 basically one to one. So if you have 00:31:06.640 --> 00:31:08.480 Terraform provider AWS, it has a 00:31:08.480 --> 00:31:10.640 resource called AWS instance. There will 00:31:10.640 --> 00:31:13.039 be an L1 construct called AWS inst or 00:31:13.039 --> 00:31:15.200 just instance and you can use that to 00:31:15.200 --> 00:31:18.240 create an AWS instance um configuration 00:31:18.240 --> 00:31:20.399 block in terapform right the same thing 00:31:20.399 --> 00:31:23.360 CFN resources they also have an L1. So 00:31:23.360 --> 00:31:25.760 that's actually nothing special. Now the 00:31:25.760 --> 00:31:28.559 big difference is that AWS CDK has an 00:31:28.559 --> 00:31:30.799 advanced asset pipeline which includes 00:31:30.799 --> 00:31:34.240 bundling. what the asset pipeline which 00:31:34.240 --> 00:31:36.399 which sometimes I feel shouldn't be 00:31:36.399 --> 00:31:38.720 there because it creates really annoying 00:31:38.720 --> 00:31:40.320 divs but carry on 00:31:40.320 --> 00:31:42.399 >> the asset pipeline does things like if 00:31:42.399 --> 00:31:45.200 you have a lambda function and it needs 00:31:45.200 --> 00:31:46.000 u python 00:31:46.000 --> 00:31:47.360 >> it builds a layer and everything yeah 00:31:47.360 --> 00:31:47.919 it's nice 00:31:47.919 --> 00:31:50.240 >> it can do the layering it can package up 00:31:50.240 --> 00:31:52.960 like if you have JavaScript or NodeJS 00:31:52.960 --> 00:31:55.360 package JSON with dependencies it can 00:31:55.360 --> 00:31:56.960 identify automatically that you're 00:31:56.960 --> 00:31:59.440 deploying into a lambda layer or into a 00:31:59.440 --> 00:32:01.760 lambda function runtime that supports 00:32:01.760 --> 00:32:05.039 AWS SDK version 3 and it will make those 00:32:05.039 --> 00:32:06.799 external. So when you bundle things up, 00:32:06.799 --> 00:32:08.640 it's going to automatically make sure 00:32:08.640 --> 00:32:10.960 that it doesn't bundle in the AWS SDK 00:32:10.960 --> 00:32:12.480 for you. It does a lot of amazing 00:32:12.480 --> 00:32:14.080 things. You get none of that in 00:32:14.080 --> 00:32:15.679 Terraform world, right? Unless you build 00:32:15.679 --> 00:32:16.159 it. 00:32:16.159 --> 00:32:18.960 >> It could be more amazing. I feel like we 00:32:18.960 --> 00:32:21.840 at work it we have a few stacks that 00:32:21.840 --> 00:32:23.840 build a lot of lambda layers and and I 00:32:23.840 --> 00:32:25.600 like I I don't maybe there's a trick, 00:32:25.600 --> 00:32:27.279 but I would like to say like these 00:32:27.279 --> 00:32:29.120 lambda layers are all effectively the 00:32:29.120 --> 00:32:30.720 same. Can you just make them share 00:32:30.720 --> 00:32:31.200 ahead? 00:32:31.200 --> 00:32:33.039 >> Yeah. And if you look at to rebuild 00:32:33.039 --> 00:32:34.320 everyone. 00:32:34.320 --> 00:32:36.399 >> So, so there's an amazing open source 00:32:36.399 --> 00:32:38.640 project from the AWS CDK team, uh, an 00:32:38.640 --> 00:32:41.120 internal team at AWS that built the 00:32:41.120 --> 00:32:42.799 construct hub. So, if you go to 00:32:42.799 --> 00:32:45.279 constructs.dev or construct.dev, you can 00:32:45.279 --> 00:32:47.440 see every single package on npmgs that 00:32:47.440 --> 00:32:48.880 have constructs. If they are tacked and 00:32:48.880 --> 00:32:50.320 licensed properly, they get 00:32:50.320 --> 00:32:51.840 automatically indexed and they have a 00:32:51.840 --> 00:32:53.679 nice documentation page made available. 00:32:53.679 --> 00:32:56.000 So, this this hub thing, right, it runs 00:32:56.000 --> 00:32:58.559 an a data ETL pipeline. And I think I 00:32:58.559 --> 00:33:00.799 really have to go uh but that thing runs 00:33:00.799 --> 00:33:03.600 a pipeline and it runs off bunch of step 00:33:03.600 --> 00:33:05.760 functions and lambdas. And inside that 00:33:05.760 --> 00:33:08.320 repo the if you go to the source code 00:33:08.320 --> 00:33:09.600 you can actually see that they use 00:33:09.600 --> 00:33:11.840 progen to generate a special lambda 00:33:11.840 --> 00:33:14.000 function generator that then bundles the 00:33:14.000 --> 00:33:16.240 code appropriately. So exactly what you 00:33:16.240 --> 00:33:18.799 just said they build it in there. 00:33:18.799 --> 00:33:21.120 >> Yeah. Progen is is a tool that that that 00:33:21.120 --> 00:33:23.519 manages a it's like CDK for file 00:33:23.519 --> 00:33:25.840 systems. So you can define a file 00:33:25.840 --> 00:33:27.840 construct and you can define um 00:33:27.840 --> 00:33:29.039 >> okay I'll have to have a look at well 00:33:29.039 --> 00:33:31.039 thank well okay you have to go you have 00:33:31.039 --> 00:33:33.440 to go so 00:33:33.440 --> 00:33:35.679 >> so that that's the main difference AWS 00:33:35.679 --> 00:33:39.519 CDK does CDK does so much more and CDKTF 00:33:39.519 --> 00:33:41.840 does only the very basic it the asset 00:33:41.840 --> 00:33:43.760 pipeline is actually it has the concept 00:33:43.760 --> 00:33:46.399 of an archive but that's not bound to 00:33:46.399 --> 00:33:48.559 like any provider because Terapform is 00:33:48.559 --> 00:33:50.799 decoupled from AWS right so if you do 00:33:50.799 --> 00:33:53.279 AWS CDK if you say an asset it will use 00:33:53.279 --> 00:33:55.600 ECR S3 for you behind the behind the 00:33:55.600 --> 00:33:55.840 scenes. 00:33:55.840 --> 00:33:57.120 >> Yeah, I mean that could be that could be 00:33:57.120 --> 00:33:59.200 a benefit because I do I do prefer basic 00:33:59.200 --> 00:34:01.600 in a way. I do prefer zip files than 00:34:01.600 --> 00:34:02.640 docker images. 00:34:02.640 --> 00:34:04.880 >> You do until you don't you do until it's 00:34:04.880 --> 00:34:06.240 done for you and then you say, "Oh my 00:34:06.240 --> 00:34:07.519 god, I wish I had this sooner." 00:34:07.519 --> 00:34:08.079 >> Yeah. 00:34:08.079 --> 00:34:09.679 >> The trouble with zip files is that is 00:34:09.679 --> 00:34:11.679 that Snick doesn't seem to like them or 00:34:11.679 --> 00:34:13.839 whatever security bloody thing people 00:34:13.839 --> 00:34:14.159 use. 00:34:14.159 --> 00:34:16.560 >> Oh, then you just base 64, save them, 00:34:16.560 --> 00:34:19.200 zip them again, 64 them, zip them again. 00:34:19.200 --> 00:34:21.919 Then no security scanner will detect it. 00:34:21.919 --> 00:34:24.480 Okay. All right. I I have to drop. 00:34:24.480 --> 00:34:26.800 >> Okay. Thanks, Vincent. See you, man. 00:34:26.800 --> 00:34:29.520 >> See you. Bye. 00:34:29.520 --> 00:34:34.639 >> Uh well, it's just me and 00:34:34.639 --> 00:34:37.040 I will publish as as a podcast. And if 00:34:37.040 --> 00:34:38.480 you're listening to this podcast, thank 00:34:38.480 --> 00:34:44.159 you and comment and like. 00:34:44.159 --> 00:34:46.720 I wanted to remind you that if you got 00:34:46.720 --> 00:34:48.960 this far, well, well done. though we 00:34:48.960 --> 00:34:50.960 also do have summaries on debates.com 00:34:50.960 --> 00:34:53.040 podcast 00:34:53.040 --> 00:34:55.599 and uh yeah please comment please get in 00:34:55.599 --> 00:34:57.520 touch if you like the podcast so we I 00:34:57.520 --> 00:35:00.079 can keep on uh being motivated to get up 00:35:00.079 --> 00:35:04.320 really early and and do it though at the 00:35:04.320 --> 00:35:07.520 same time it's just great to chat with 00:35:07.520 --> 00:35:10.240 Vincent and have an excuse to do that 00:35:10.240 --> 00:35:12.000 because I'm always learning. I hope you 00:35:12.000 --> 00:35:13.680 agree 00:35:13.680 --> 00:35:15.599 um well that I'm learning and I hope 00:35:15.599 --> 00:35:18.640 that you picked up something yourself. 00:35:18.640 --> 00:35:21.040 Cool beans, have a great day. Have a 00:35:21.040 --> 00:35:23.280 great week. 00:35:23.280 --> 00:35:26.560 Get in touch.