From Coldplay to Code: AI Wins, Data Lakes, and the Age-Old Centralized vs. Distributed Debate

0:00 Scott Porad: I want our engineers to learn this for themselves and for their careers.

0:04 Intro/Outro: Theme Song

0:21 Scott Porad: Welcome everybody. To the latest and greatest episode of how Many CTOs Does it Take? There is no guest today. It's just me, Scott Porad and my trustee, co-host, Brad Hefta-Gaub, and we are here for the summer catch-up edition. Trying to find out what on earth is going on out there. It's just hot.

0:41 Brad Hefta-Gaub: What is going on out there?

0:42 Scott Porad: People are going crazy. I'll tell you what's going on out there. At the company I work at, we have some data pipelines, and we are in the progress of migrating those data pipelines from, I think it's something called DBT Cloud, to another service called astronomer. Have you heard of astronomer lately?

1:10 Brad Hefta-Gaub: I have not.

1:12 Scott Porad: You have not heard about the Coldplay CEO. I heard about that.

1:18 Brad Hefta-Gaub: I heard about that. Who was the

1:19 Scott Porad: CEO of Astronomer

1:21 Brad Hefta-Gaub: that, oh, oh, that astronomer.

1:23 Scott Porad: I assume it's the same company. I did not go investigate directly, Brad. You know that in my career, I have a stint of five year stint as a professional in internet memes.

1:38 Brad Hefta-Gaub: Yeah, of course.

1:40 Scott Porad: I, I can't, has, can't, has internet mem burger, I can't have cheeseburger, fail blog, all those things. We did that for five years. Uh, it grew like crazy. We grew 22 x in three years, the first three years, and then our growth started to plateau. But like the first three years were madness. And, um, I've never seen, I don't think I've ever seen.

2:08 Brad Hefta-Gaub: blow up like that.

2:09 Scott Porad: It's crazy.

2:11 Brad Hefta-Gaub: I, um, uh, there's, there's a couple things that come to mind. The first one, I think I, when I saw this, I was just like, look, like there's a few things, like my personal opinion is like, Hey, consenting adults, whatever, you know? Mm-hmm. Like, if that's, you know, if everybody, if everybody involved, you know, it's fine.

2:45 Scott Porad: right.

2:46 Brad Hefta-Gaub: So whatever

2:47 Scott Porad: the question related to our world has really nothing to do with Coldplay.

2:58 Brad Hefta-Gaub: Coldplay. Yeah. Where do

2:59 Scott Porad: you stand on Coldplay? I like 'em. Yeah, I do too.

3:03 Brad Hefta-Gaub: I like them. Yeah. I mean, nobody's, they're good musicians.

3:05 Scott Porad: Nobody sounds like Coldplay. They got a unique sound.

3:10 Brad Hefta-Gaub: but

3:10 Scott Porad: who sounds like Coldplay.

3:13 Brad Hefta-Gaub: I mean, there's plenty of other bands that try to sound like Coldplay. They try to, yeah.

3:19 Scott Porad: Okay. Well, here's the CTO question. A debate we're having at our office today. I had this debate today actually. It's been an ongoing dialogue. Distributed versus centralized data teams. So we've got a machine learning team, we've got a data science team, we've got a data engineering team, we've got analytics engineering, we've got business intelligence people, you know, all sprinkled all over the place.

4:13 Brad Hefta-Gaub: Man, this is, that's, it's so interesting you asking this question. Um, oh man. I think, I mean. Where's Kevin when we need him?

4:31 Scott Porad: to the Kevin Smith, Kevin Goldsmith episode for those of you exactly. Who aren't paying attention. Yeah.

4:36 Brad Hefta-Gaub: Kevin Goldsmith. It depends. Um, I think that like here, here's where, here's where I'm currently at on this I'm currently at isn't the whole point of.

6:32 Scott Porad: I'll be, I'll be truthful. This is not my area of expertise. Yeah. I. I have some experience doing the analytics work, 'cause I worked for a while as a product manager. Mm-hmm. Um, but building data pipelines, data lakes, all these different layers, et cetera. I've actually never done it before, so I don't really know.

7:40 Brad Hefta-Gaub: Yeah.

7:41 Scott Porad: Well, I don't know what I can trust in this data lake anywhere. Who knows what's right in here. Yeah. And then, then everyone wants to start over from scratch. Right. Yeah.

7:47 Brad Hefta-Gaub: But that's a, but. Okay. Couple quick reactions to this. First one is a hundred percent agree with you that like the definition of what the, the thing is like that's. Like super important. You gotta get that. You gotta nail that down somehow early. Like, I remember this, you know, from our time at 98.6, I remember there were these reports and I was looking at some report and it was like, wow, that looks problematic.

9:13 Scott Porad: and then it,

9:14 Brad Hefta-Gaub: it

9:14 Scott Porad: gets a step, you take a step further.

9:57 Brad Hefta-Gaub: might even have like some chargebacks. Yeah. Yeah. And like, there's all, all, I mean, all of our processes, like if you think about it, like even at the. You know, like we're running a chicken shack, you know? Mm-hmm. And it's like, you don't, it doesn't count the number of chicken patties that got used, right?

10:44 Scott Porad: So here's where I'm presently landing on it. I think this is a rule that could be applied broadly if kinda your scaffolding, so your data definitions, your analytics layers, if you have them, your infrastructure and pipelines and all those things are pretty dialed.

11:51 Brad Hefta-Gaub: My instinct is to say no to that.

11:53 Scott Porad: Okay?

11:54 Brad Hefta-Gaub: My instinct is to, my instinct is to argue with you on this one.

11:57 Scott Porad: Okay? Is that because you're naturally an argue viewer or you believe in the openness of the internet and you just believe? You just believe in open systems?

12:06 Brad Hefta-Gaub: Well. No, actually, my reason to question your, your, your kind of statement there is I think that one of, one of the things I'm seeing nowadays, I, I often see, and I feel like, I feel like, I feel like this is a pretty common story, is people don't know what they want Most of the time they don't know.

12:49 Scott Porad: Mm-hmm.

12:50 Brad Hefta-Gaub: But if you're building any kind of a new product offering, or you're even trying to like, uh, uh, you know, kind of.

13:01 Scott Porad: Mm-hmm.

13:02 Brad Hefta-Gaub: Or you're adding new products. I feel like when you first layer that new product on or that new approach on, you don't know what the metrics that are gonna matter are. Yeah.

13:16 Scott Porad: Hot take I agree with you and,

13:18 Brad Hefta-Gaub: okay.

13:18 Scott Porad: Yeah. I think it's, as you're, as you mature more and more metrics solidify, and then you kind of gotta do that.

13:29 Brad Hefta-Gaub: So I, maybe I spend too much time in the molten lava stage, right? No, like that. I'm dealing with these entrepreneurs that are building like new, fresh systems and they're constantly asking me questions like, wait a second.

14:03 Scott Porad: Yeah. Well I think when as, as you, as a company matures, then you need both. 'cause you'll, in theory, you're always evolving and learning if you know developing. Your business and we'll need to be learning molt and lava things, but as, as parts mature, then you need those to settle into rock, I guess, or something like that.

14:25 Brad Hefta-Gaub: It does turn into rock.

14:27 Scott Porad: Okay. So that's, that's okay. So, uh, data is one thing that's happening with us. We're trying to get our data house better. It's not bad. Um, we had a little bit of turnover in that. It sounds like

14:37 Brad Hefta-Gaub: you're doing a pretty major.

14:44 Scott Porad: Oh, that piece. That's just for, uh, I think that's just, you know, I, I actually don't know. I'm not that involved in it. Um, I actually think that's just sort of a small piece, which is whatever tool we're using for this portion right now just doesn't kind of have the monitoring and alerting sort of the, the management layer on top of.

16:04 Brad Hefta-Gaub: now. Yeah. One of the things that, uh, so here's my, here's my quick hot take on that question of like, together distributed, um. I, I feel like, um, you know, data engineering is a very real discipline.

17:32 Scott Porad: Mm-hmm. Mm-hmm.

17:33 Brad Hefta-Gaub: And. So just like when we talked about this before, like if you break up your team and you distribute them, that means each of those teams need to have one of those people that are really good mm-hmm.

18:54 Scott Porad: Yeah, yeah. Well that's what we got going on. What do you got going on over there? It's still stirred up.

19:02 Brad Hefta-Gaub: Oh man, we got lots of stuff going on. I've got, um, you know, I guess I, uh, wait, what did you say? How long is it gonna take before we talk about ai?

19:11 Scott Porad: Oh yeah. We're 20 minutes, took us 20 minutes to talk about ai.

19:26 Brad Hefta-Gaub: Yeah. I'm not wearing a hat. So maybe it's, maybe it's, maybe it's a, maybe it's a fake image of me.

21:25 Scott Porad: Hmm.

21:26 Brad Hefta-Gaub: And, you know, giving it problems to solve, telling it to come back to me with a plan before it does anything.

22:01 Scott Porad: So I have a few questions then for you. The first is the quickie. What is the, what is your tool set or stack? Are you using cloud code? Are you using GitHub or our, uh, open AI codex? Are you using, uh, cursor? Are you using augment code Windsor? What's your, what's your. As they say, rate. Rate my rig.

22:20 Brad Hefta-Gaub: Yeah. Rate my rig.

23:53 Scott Porad: Mm-hmm.

23:54 Brad Hefta-Gaub: The idea with another senior developer, right? Like, okay, this is the thing I'm trying to solve.

24:31 Scott Porad: Mm-hmm. Brilliant.

24:32 Brad Hefta-Gaub: And then I take that prompt and I go into cursor and I give it that prompt and then I again tell it don't code anything until I agree to your plan.

24:47 Scott Porad: Okay. I have a few questions. The first is in the phase where you were interacting with chat GPT and having the conversation about the solution, are you doing that by typing or actually doing it in voice?

24:59 Brad Hefta-Gaub: Oh, I type it.

25:01 Scott Porad: it. Um, rats. I just forgot my second question. A lot of

25:05 Brad Hefta-Gaub: time, by the way, the reason I type it is 'cause. Uh, a lot of times I've gotta like, give it code or give it, you know, I'm gonna, I'm gonna copy and paste stuff into it and say, this is what I'm doing. Here's an example of code. You know, here's the stack trace of a, of a stack overflow crash that hap, you know what I mean?

25:25 Scott Porad: So, um, just as a side note, it'd be interesting to get these ais to be more real. If you and I were in the, in. In the, if Youer and I were in an office Right. Or even online, it doesn't matter. Mm-hmm. And we had this situation, I would call you on a Zoom call or meet Google Meet or whatever, where I would, we'd get together in the same office, I would show you the error, you might look at it and then we would talk to each other.

26:45 Brad Hefta-Gaub: Well, and also it's fair for you to cr it's fair for you to say it's not rubber ducking because isn't the whole point of rubber ducking. It's more like I am saying it out loud to the rubber duck. And it's me hearing my own voice. Say it out loud. Yeah. Where in fact, I'm actually very much like going back and forth with this thing that's spitting back answers to me.

27:05 Scott Porad: Yeah. Would you want it to be talking to you like it? Would you want to hear a voice on the other end?

27:10 Brad Hefta-Gaub: I, for me personally, I don't think so. It's, that doesn't really, I mean that that feature exists, right? Like I could actually use that with chat GPT and just. Talk into it and then because it's got a little mic button and it'll read it back to me.

27:54 Scott Porad: Mm, yeah. You're the second person today to tell me that.

27:57 Brad Hefta-Gaub: Yeah, so for example, I just started working with a, a new, um, client that's like a couple hours a week of like just helping them get their, get their plan in place. And they already have this existing code base that was built by a external agency team.

29:58 Scott Porad: That tracks was something we talked about, I don't know, months ago, the study that I, we dug up from open AI where they had took these jobs off of.

30:15 Brad Hefta-Gaub: And then you can take it and say, okay, great. Now write a, now write a feature brief for this feature and here's like, here's the five sentences I know about it. Yeah. And write a tech plan based on the code you just looked at.

31:32 Scott Porad: you said. It's been working a lot better. Broadly speaking, it's been working a lot better for you. And I'm curious to know, proportionally, is it, the models are getting better, the tools are getting better, or just your understanding of how to use them is getting better?

31:58 Brad Hefta-Gaub: Well, I, I agree with you that all three of those things are part of the, the mix. I think, I think that, um, I'm getting better at understanding how to, what it's good at how to, how to, how to, so the prompt engineering part of things.

33:57 Scott Porad: That's definitely, I mean, I, I read about that or watched a video. I can't remember. You know, they do this, I don't remember the words that they use to describe it, but sort of, there's like the deep learning where they train the thing on the whole internet and then like.

34:29 Brad Hefta-Gaub: Oh, interesting. Yeah. Um, I, yeah, you are in a AB split 'cause I haven't seen that yet.

34:34 Scott Porad: Yeah. If you look, if you go to the homepage of Google, you might find a button or something to take you there. But it just called AI mode and it introduced itself like this is our new, our new search, like AI mode. And it was basically drop drop dropping you.

35:03 Brad Hefta-Gaub: Yeah, I, I think that, um, there's an element of this that's like, um. When, I mean, I'm, I'm assuming this will resonate for you, Scott, but it's like when you're solving problems quickly mm-hmm.

35:23 Scott Porad: Yeah. Feels good.

35:24 Brad Hefta-Gaub: That feels good. That feels good. Nothing feels better than

35:27 Scott Porad: crossing stuff off the list.

35:29 Brad Hefta-Gaub: It's a little bit like, I, I guess there's a part of me that's like, um, you know, I, uh. I, you know, I got the, I got a brand new Sawzall, and man, it is cutting through the drywall like crazy, you know?

35:48 Scott Porad: makes you wanna cut more things,

35:50 Brad Hefta-Gaub: makes you wanna cut more things, you know? Yeah. So that's what I think that's, that's, that's kind of my take on it. Um, and you know, I guess I, I kind of would, curiously, I'd pivot this back to. I'm obviously down in these small teams, these really small teams, you know, so that's why I have the, the luxury mm-hmm.

36:40 Scott Porad: We have a Slack channel. Um. It's not called AI wins, but it's basically called AI wins. I mean, it's got some funny, weird name 'cause an engineer named it, but, uh, it's like temp, POC, ai, whatever, whatever.

38:16 Brad Hefta-Gaub: yep. Anyway,

38:17 Scott Porad: we settled on this tool.

41:07 Brad Hefta-Gaub: You're like, you like felt like, was it like the episode of Star Trek where like, uh, in the movie where like Scotty says, wait a second. You're not actually supposed to tell the captain how fast it can go.

41:20 Scott Porad: So he didn't totally complete the job in that time, but he is like, yeah, he didn't get it all done, but like, yeah, this like sped this thing way up. And there's lots of examples of that kind of stuff Yeah. That we're seeing. And so, um, you know, in our organization there are early adopters, there are late adopters.

44:55 Brad Hefta-Gaub: I think that that is a, um, inspiring and, um. Compassionate and heartwarming perspective for you to have as a leader. And I think it's great. I'm sure that your team, uh, appreciates your perspective on that. And, um, you know, and I'll just, I'll just say that, you know, one of the things that I've seen recently on my relatively small team is some of the more experienced curmudgeonly mm-hmm.

45:46 Scott Porad: I think, I think that's awesome.

46:16 Brad Hefta-Gaub: I feel like I did write, I wrote some code in Notepad back in the day when, but there were these other

46:22 Scott Porad: guys that had Visual Studio. Right. And I really, I personally loved working with Visual Studio, um, because just, it made me powerful.

46:42 Brad Hefta-Gaub: I think that's what it was. That sounds right. Um.

46:46 Scott Porad: We were having this thing, I was just like, we were having this debate about it. I was like, why don't we just all agree to just hit that, that we don't even have to just let Visual Studio decide what the formatting's gonna be.

46:58 Brad Hefta-Gaub: right. Where the

46:59 Scott Porad: brackets, so I, I really liked Visual Studio. Yeah, exactly. I really liked Visual Studio. 'cause it may be powerful and eventually now I think everybody uses an IDE, you know? Um, and the, the earlier editors, like Vim and emax have really become tremendously more rich in terms of what they're basically IDs now and, um mm-hmm.

47:29 Brad Hefta-Gaub: Yeah. Yeah. I hope, I hope that's how it's going for you. You know, there's a part of me that does worry about, um, and does think, like, again, I said I, I, I, I feel like you are, uh, a very compassionate leader in your perspective that you shared, I think is really, you know, wonderful.

49:24 Scott Porad: Uh, yeah, I think we'll keep on celebrating it. Um, you know, I think we have in our career and leveling guide, you know, there is a flank in there about. Exploring, adopting new technologies, those kinds of things. That's, that's part of a job of an engineer.

51:16 Brad Hefta-Gaub: No, it does. Totally. A hundred percent. Yeah. And, and, uh, in the, in the spirit of like old guys with gray, gray hair and gray beards, I will say back when Visual Basic was first announced, everyone was convinced that that was gonna be the end of software engineering jobs because people were gonna just be able to use Visual Basic to, to build their business applications.

51:37 Scott Porad: nickel for every time. Exactly. Perhaps if the underlying world of technology wasn't changing so rapidly. That would be true, right? Like let's just fast forward some amount of years when you know AI has reached an equilibrium. Everyone's using it, we're getting all the value from it.

52:06 Brad Hefta-Gaub: Yeah. Alright. Yeah, that. Was a place to end. We, this was a wild ride. We started with Data Lakes and we ended up in back on AI again.

52:17 Scott Porad: Oh, actually there is some, can I give a little bonus a PS to this episode?

53:13 Brad Hefta-Gaub: I love it. It's a, it's a brave new world. I can't, it's a lot of cool stuff happening.

53:17 Scott Porad: Yeah.

53:18 Brad Hefta-Gaub: Alright, well, Brad, it's

53:19 Scott Porad: been fun

53:19 Brad Hefta-Gaub: talking this week. It's fun talking. Yeah. We'll, uh, talk to you next week. Don't forget to do what?

53:28 Scott Porad: review. It's true. If you are listening all the way into like the 50th minute of this podcast, you must like it. So you must like it. Go do, do us a favor. Go into Apple Podcasts, or Google Play or Spotify or wherever you go and to give it a, give it a five star rating.

53:50 Brad Hefta-Gaub: listening to this one podcast that had five stars and one of them was halfway filled in, so

53:58 Scott Porad: Oh, wow. So somebody has four and a half star.

54:01 Brad Hefta-Gaub: No. Oh, that was a bad joke. Bad joke.

54:07 Scott Porad: Alright. All right, thanks. Talk to you next week,

54:10 Brad Hefta-Gaub: This episode of How Many CTOs does It Take? Podcast was edited by Saffron Tigo. Our theme song was AI generated by Brad. Please follow us on all the social media platforms at at how many CTOs Pod you can find us on YouTube, LinkedIn, Instagram, Facebook Threads X, TikTok, and Patreon, or at our website at how many CTOs pod.com.