7 questions with Craig Kerstiens: AI is like 100 farm hands you don't have to feed

00:00:00,600 --> 00:00:15,840 [Eric Dodds] [instrumental music] Welcome back to the Token Intelligence Show. John, number one, we're both wearing shorts. That's because it's August- 00:00:15,840 --> 00:00:16,170 [John Wessel] It's so hot 00:00:16,170 --> 00:00:19,180 [Eric Dodds] ... in South Carolina. [chuckles] And... 00:00:19,180 --> 00:00:21,270 [John Wessel] 90, 97, I don't know. 00:00:21,270 --> 00:00:21,280 [Eric Dodds] Yeah. 00:00:21,280 --> 00:00:22,180 [John Wessel] It feels like 97- 00:00:22,180 --> 00:00:22,320 [Eric Dodds] Yeah 00:00:22,320 --> 00:00:23,360 [John Wessel] ... even though it's only at 95, but... 00:00:23,360 --> 00:00:24,600 [Eric Dodds] Yeah, that's the humidity. 00:00:24,600 --> 00:00:25,480 [John Wessel] Yeah, right. 00:00:25,480 --> 00:00:31,870 [Eric Dodds] But we have an exciting announcement. We are actually going to do our 1st guest show, uh, 00:00:32,880 --> 00:00:35,050 [Eric Dodds] and we have a series of guest shows coming up, 00:00:36,160 --> 00:00:46,320 [Eric Dodds] and we're going to do something interesting with our guests. We're going to ask all of the guests that we have on Token Intelligence the same basic set of questions, 00:00:47,350 --> 00:00:57,940 [Eric Dodds] so that we can see how people use AI, how they think about AI, and how they think AI is going to change the way that the future looks, which will be super interesting. 00:00:57,940 --> 00:00:58,460 [John Wessel] Yep. 00:00:58,460 --> 00:01:13,000 [Eric Dodds] And so after a couple dozen of these, we can identify, you know, trends and opinions from people that we think are interesting and whose opinions we respect. So, we started at the Carolina Code Conference, uh, which happened last weekend. 00:01:13,000 --> 00:01:13,780 [John Wessel] Yep. 00:01:13,780 --> 00:01:39,720 [Eric Dodds] I wasn't able to attend, but you were there, and we talked with some really interesting people. The 1st guest that we had on the show at the conference was Craig Kerstiens. And what's really interesting to me about Craig, I mean, a number of things, but multi-time, uh, founder with exits at every major step around the same technology. So- 00:01:39,720 --> 00:01:40,600 [John Wessel] Rinse and repeat. 00:01:40,600 --> 00:01:46,190 [Eric Dodds] Rinse and repeat. Postgres, one of the most popular database projects in the entire world. Um, 00:01:47,320 --> 00:02:02,280 [Eric Dodds] so he was, uh, working on Postgres at Heroku, which is a hosting platform. They sold to Salesforce. Uh, his 2nd company, um, that he was part of sold to Microsoft, and then his latest company sold to Snowflake- 00:02:02,280 --> 00:02:03,100 [John Wessel] Yep 00:02:03,100 --> 00:02:05,030 [Eric Dodds] ... uh, which is a pretty incredible- 00:02:05,030 --> 00:02:05,460 [John Wessel] It's a pretty great run 00:02:05,460 --> 00:02:07,300 [Eric Dodds] ... [chuckling] pretty incredible run. 00:02:07,300 --> 00:02:15,080 [John Wessel] Yeah. Yeah, I had a blast talking to Craig. Um, one of the things I think you guys will find really interesting on the episode is his team 00:02:16,220 --> 00:02:22,900 [John Wessel] works with, um, their everyday workflow involves software that they've written as desktop apps for their Macs. 00:02:22,900 --> 00:02:25,570 [Eric Dodds] Really? They themselves? Not like they download from the app store? 00:02:25,570 --> 00:02:28,190 [John Wessel] Right. And this is a product team too. These aren't nec- these aren't necessarily- 00:02:28,190 --> 00:02:28,200 [Eric Dodds] Really? 00:02:28,200 --> 00:02:28,980 [John Wessel] ... engineers. Yeah. 00:02:28,980 --> 00:02:29,220 [Eric Dodds] Okay. 00:02:29,220 --> 00:02:32,140 [John Wessel] I mean, it's a, it's a mixed team, but like, they're more product. 00:02:32,140 --> 00:02:39,320 [Eric Dodds] And are their apps... are the apps per individual, or are they apps that are built that the team uses? 00:02:39,320 --> 00:02:40,920 [John Wessel] I think they're customized per individual. 00:02:40,920 --> 00:02:41,160 [Eric Dodds] Per individual. 00:02:41,160 --> 00:02:45,120 [John Wessel] We didn't dig super deep into it, but I'm excited for you guys to hear about that and a lot more. 00:02:45,120 --> 00:02:50,880 [Eric Dodds] All right, fascinating. Well, we're gonna jump into 6 AI questions with Craig Kerstiens. 00:02:50,880 --> 00:02:51,680 [John Wessel] Yep. Enjoy, everybody. 00:02:52,760 --> 00:02:58,760 [John Wessel] All right, we're here live with Craig Kerstiens at the Carolina Code Conference. Craig, welcome to the show. 00:02:58,760 --> 00:03:00,200 [Craig Kerstiens] Thanks. Excited to be here. 00:03:00,200 --> 00:03:11,060 [John Wessel] Awesome. Um, let's start out for, for the people that don't know you, um, just tell me a little bit about your background and, um, kind of how you, how you got into, to Postgres. 00:03:11,060 --> 00:03:32,300 [Craig Kerstiens] Yeah. I s- sort of stumbled into Postgres, actually. So I, uh, originally am from Alabama, came out to California about 20 years ago. Uh, was at a, a big consulting company, went to a small startup. Um, from there, ended up at Heroku. And at Heroku was one of the 1st product managers there. Uh, actually focused on, like, launching Python at 1st. 00:03:32,300 --> 00:03:32,980 [John Wessel] Oh, cool. 00:03:32,980 --> 00:03:47,010 [Craig Kerstiens] Um, but while there, like, we had launched a Postgres service, were running it, and I found our engineers were just, like, just using it like, um, I think DHH famously said, like, "The database is just like a dumb hash in the sky." Right? 00:03:47,010 --> 00:03:47,680 [John Wessel] Okay. Yeah. 00:03:47,680 --> 00:03:57,040 [Craig Kerstiens] Like, just put data in. Don't think about anything about it. And, uh, I accidentally found myself evangelizing, like, "Why aren't you taking advantage of everything that Postgres can do?" 00:03:57,040 --> 00:03:57,200 [John Wessel] Right. 00:03:57,200 --> 00:03:58,320 [Craig Kerstiens] To our internal engineers. 00:03:58,320 --> 00:03:59,140 [John Wessel] Right. 00:03:59,140 --> 00:04:04,400 [Craig Kerstiens] And then our, uh, the, the GM of that org was like, "Hey, you should come do marketing for us." Like- 00:04:04,400 --> 00:04:04,620 [John Wessel] Oh, wow 00:04:04,620 --> 00:04:05,980 [Craig Kerstiens] ... I, I don't do marketing. I'm like- 00:04:05,980 --> 00:04:06,010 [John Wessel] Yeah 00:04:06,010 --> 00:04:07,390 [Craig Kerstiens] ... that's, that's like ad- 00:04:07,390 --> 00:04:07,660 [John Wessel] Yeah 00:04:07,660 --> 00:04:09,160 [Craig Kerstiens] ... sleazy. Like, I'm a developer. 00:04:09,160 --> 00:04:09,420 [John Wessel] Yeah. 00:04:09,420 --> 00:04:09,680 [Craig Kerstiens] Like- 00:04:09,680 --> 00:04:10,240 [John Wessel] Right 00:04:10,240 --> 00:04:11,400 [Craig Kerstiens] ... I'm a purist. No. 00:04:11,400 --> 00:04:11,580 [John Wessel] Yeah. 00:04:11,580 --> 00:04:30,280 [Craig Kerstiens] And so, uh, there was an opening there to then come run product and do the marketing of that team. I'm like, "All right, I'll come run product. Maybe I'll do a little bit of this marketing stuff on the side." Um, fast-forward 5 years, like, helped build and grow that team. The tail end of my time at Heroku, we were managing just over 1.5 million databases- 00:04:30,320 --> 00:04:30,500 [John Wessel] Right 00:04:30,500 --> 00:04:31,480 [Craig Kerstiens] ... across all our customers. 00:04:31,480 --> 00:04:31,760 [John Wessel] Wow. Yeah. 00:04:31,760 --> 00:04:36,200 [Craig Kerstiens] Um, went from there to Citus Data, which was sharded Postgres. 00:04:36,200 --> 00:04:36,880 [John Wessel] Yep. 00:04:36,880 --> 00:04:37,460 [Craig Kerstiens] Um, so I kind of- 00:04:37,460 --> 00:04:39,040 [John Wessel] It's like Postgres at scale, right? 00:04:39,040 --> 00:04:44,880 [Craig Kerstiens] Yeah, exactly. Like, our average customer was 40 terabytes, the largest 960 terabytes. 00:04:44,880 --> 00:04:46,620 [John Wessel] Okay. Yeah, that's big. 00:04:46,620 --> 00:05:01,680 [Craig Kerstiens] So not quite petabyte, but I like to round up. Um, they were acquired by Microsoft. Found myself running all Azure Postgres, uh, and then was ready to kind of just get back to the roots of, like, running a really good Postgres service and joined Crunchy Data about 6 years ago. 00:05:01,680 --> 00:05:02,180 [John Wessel] Okay. 00:05:02,180 --> 00:05:05,900 [Craig Kerstiens] We were acquired about a year ago to build the foundation for Snowflake Postgres. So- 00:05:05,900 --> 00:05:06,340 [John Wessel] Yeah 00:05:06,340 --> 00:05:12,920 [Craig Kerstiens] ... it's funny, I, like, I've been in Postgres for about 15 years now deeply. I consider myself more of a DevTools guy. 00:05:12,920 --> 00:05:13,440 [John Wessel] Okay. 00:05:13,440 --> 00:05:15,660 [Craig Kerstiens] So, like, I can't do it again. Like, next time I'm- 00:05:15,660 --> 00:05:15,800 [John Wessel] Yeah 00:05:15,800 --> 00:05:16,360 [Craig Kerstiens] ... done with Postgres- 00:05:16,360 --> 00:05:16,600 [John Wessel] Okay 00:05:16,600 --> 00:05:17,360 [Craig Kerstiens] ... I'm done with Postgres. 00:05:17,360 --> 00:05:18,020 [John Wessel] Yeah. 00:05:18,020 --> 00:05:20,520 [Craig Kerstiens] Uh, but it was kind of by accident that I fell into it. 00:05:20,520 --> 00:05:29,940 [John Wessel] Yeah. Yeah, that's so interesting. All right, so what is... From your technical background, what's, what's an interest or hobby that you have that maybe would be surprising to people? 00:05:31,500 --> 00:05:33,660 [Craig Kerstiens] Sur- surprising to people. Um, 00:05:35,800 --> 00:05:40,720 [Craig Kerstiens] I don't... You know, uh, I've got young kids, so, like, hobbies mostly don't exist. 00:05:40,720 --> 00:05:41,520 [John Wessel] I agree with that. Yeah. 00:05:41,520 --> 00:05:46,400 [Craig Kerstiens] Um, like, my wine collection's pretty good. There's a, there's probably 5 or 600 bottles of wine. 00:05:46,400 --> 00:05:46,780 [John Wessel] Okay. 00:05:46,780 --> 00:05:54,580 [Craig Kerstiens] Um, so, uh, usually when people come over for dinner and bring a bottle of wine, they're like, "Oops, I shouldn't have done that." Like, just bring the flowers. That's totally fine. 00:05:54,580 --> 00:06:02,159 [John Wessel] That's super funny. Yeah. What, what's your region? What's your, like, go-to region? I mean, you're in California, so are you a California wine guy, or is there another region you like better? 00:06:02,160 --> 00:06:04,430 [Craig Kerstiens] I generally also love French wines as well. 00:06:04,430 --> 00:06:04,540 [John Wessel] Okay. 00:06:04,540 --> 00:06:06,220 [Craig Kerstiens] Um, Italy frustrates me- 00:06:06,220 --> 00:06:06,400 [John Wessel] Okay 00:06:06,400 --> 00:06:15,260 [Craig Kerstiens] ... because there's like 500 varietals, and they don't follow rules, and it's just like, like it's a lot of work. Uh, maybe when I retire, I'll get into Italians more just because- 00:06:15,260 --> 00:06:15,560 [John Wessel] Okay 00:06:15,560 --> 00:06:17,664 [Craig Kerstiens] ... after work. 00:06:17,664 --> 00:06:22,494 [Craig Kerstiens] Being in California, we've got really good wine there, good access to it. But 00:06:24,304 --> 00:06:33,203 [Craig Kerstiens] when I'm on the East Coast, there's better access to the French and European. So it kind of depends on where I'm at on the path. It's like French versus California, really. 00:06:33,204 --> 00:06:40,484 [John Wessel] So if I'm at a kind of a standard grocery store that has a decent wine selection, how do I pick something that's good? 00:06:41,644 --> 00:06:42,884 [John Wessel] Okay to you even, without- 00:06:42,884 --> 00:06:46,164 [Craig Kerstiens] I think it's know the basics of what you like, right? 00:06:46,164 --> 00:06:46,484 [John Wessel] Okay. 00:06:46,484 --> 00:06:51,184 [Craig Kerstiens] Like, do you like Merlot or Cab? Do you like white or Pinot? Like, if you know 00:06:52,344 --> 00:06:54,344 [Craig Kerstiens] generally what you like- 00:06:54,344 --> 00:06:54,844 [John Wessel] Right 00:06:54,844 --> 00:06:54,844 [Craig Kerstiens] ... 00:06:55,944 --> 00:07:01,344 [Craig Kerstiens] I think I maybe wrote a blog post, like, trying to demystify. My mom texts me. She's like, "I'm at a store." 00:07:01,344 --> 00:07:01,794 [John Wessel] There you go. 00:07:01,794 --> 00:07:04,144 [Craig Kerstiens] "There's like 50 bottles. Which one?" I'm like, "Okay." 00:07:04,144 --> 00:07:10,824 [John Wessel] Because I assume you can't just pick by price, and if you don't know a lot of the labels, like you're not really a connoisseur, like what? Yeah. 00:07:10,824 --> 00:07:12,744 [Craig Kerstiens] I mean, generally price is an okay- 00:07:12,744 --> 00:07:12,834 [John Wessel] It's 00:07:13,864 --> 00:07:14,134 [John Wessel] a decent- 00:07:14,134 --> 00:07:14,784 [Craig Kerstiens] It's a decent indicator. 00:07:15,804 --> 00:07:17,424 [Craig Kerstiens] There's certain wines like 00:07:18,884 --> 00:07:22,424 [Craig Kerstiens] a Viognier, like a great white, 00:07:23,604 --> 00:07:37,984 [Craig Kerstiens] not like between a Chardonnay and a Pinot gris, but like a nice balance. Like, there's not going to be 100 of them there. There's going to be like 5 Viogniers. There's certain, like, hidden regions, like France Gigondas. 00:07:37,984 --> 00:07:38,224 [John Wessel] Okay. 00:07:38,224 --> 00:07:40,324 [Craig Kerstiens] Like, whatever they have is going to be a great value for it. 00:07:40,324 --> 00:07:40,924 [John Wessel] Oh, cool. Okay. Yeah. 00:07:40,924 --> 00:07:46,524 [Craig Kerstiens] So, like, it's easy to find a couple of these, like, hiddens ones off the main path. 00:07:46,584 --> 00:07:48,144 [John Wessel] Okay, cool. Nice. 00:07:49,254 --> 00:08:02,424 [John Wessel] All right, so I got some AI questions for you. This is probably my favorite. So if you were to time travel to 1900, just to set the context, that's like Rockefeller, Carnegie Steel, 00:08:03,764 --> 00:08:08,884 [John Wessel] Gilded Age, I think is around 1900. Describe AI to someone that lived in that era. 00:08:10,464 --> 00:08:15,104 [Craig Kerstiens] Yeah. I think I'd start with something around just, like, the basics of automation, right? 00:08:15,104 --> 00:08:15,464 [John Wessel] Okay. Yeah. 00:08:15,464 --> 00:08:18,054 [Craig Kerstiens] Like, I think about what I have for today, and it just... 00:08:19,364 --> 00:08:32,284 [Craig Kerstiens] My senior engineers describe it as like, now I have a team of junior engineers. Like a team of 10 junior engineers for me. So, I mean, back then it's like, great, what if you could have 100 kids doing all the work on the farm for you, right? 00:08:32,284 --> 00:08:32,624 [John Wessel] Yeah. Sure. 00:08:33,684 --> 00:08:38,574 [Craig Kerstiens] Like, I have to go back and think a little bit of what kind of things they were automating at that point, right? 00:08:38,574 --> 00:08:38,584 [John Wessel] Right. Sure. 00:08:38,584 --> 00:08:40,944 [Craig Kerstiens] Because automation still existed in that age. 00:08:40,944 --> 00:08:41,304 [John Wessel] Right. 00:08:41,304 --> 00:08:44,824 [Craig Kerstiens] It's like, hey, what if you had unlimited automation? Then what else would you be doing? 00:08:44,824 --> 00:08:45,424 [John Wessel] Right. 00:08:45,424 --> 00:08:45,494 [Craig Kerstiens] Right? 00:08:45,494 --> 00:08:45,504 [John Wessel] Right. 00:08:47,324 --> 00:08:52,584 [Craig Kerstiens] And I think it's funny because now we're in this time of like, "Uh-oh, is AI going to take our jobs?" 00:08:52,584 --> 00:08:53,223 [John Wessel] Right. 00:08:54,224 --> 00:08:58,403 [Craig Kerstiens] But the industrial revolution came, and we got more productive, and we had more jobs. 00:08:58,404 --> 00:08:58,964 [John Wessel] Right. 00:08:58,964 --> 00:09:00,844 [Craig Kerstiens] It's like, no, we can just get more done now. 00:09:02,164 --> 00:09:08,504 [Craig Kerstiens] So really, like, yeah, instead of having 10 kids to run the farm, you had 100 kids, but you didn't have to feed them. 00:09:08,504 --> 00:09:08,644 [John Wessel] Right. 00:09:08,644 --> 00:09:09,244 [Craig Kerstiens] Raise them. 00:09:09,244 --> 00:09:11,524 [John Wessel] Right. [chuckling] Which is the hard part. 00:09:13,684 --> 00:09:14,384 [John Wessel] All right. So, 00:09:16,104 --> 00:09:26,184 [John Wessel] what is the most positive impact that you personally have seen from AI, and then maybe the most negative impact? Could be personally or professionally. 00:09:26,184 --> 00:09:35,884 [Craig Kerstiens] I think the most positive is, like, it feels like a lot... When I see engineers become AI-pilled, like, they're having fun again. 00:09:35,884 --> 00:09:36,744 [John Wessel] Yeah. 00:09:36,744 --> 00:09:37,254 [Craig Kerstiens] Like, 00:09:38,264 --> 00:09:47,904 [Craig Kerstiens] I got into this because it was just so fun to build, debug, create things, right? I don't know that the industry felt that way for the last 10 years. 00:09:49,084 --> 00:09:50,364 [Craig Kerstiens] Yeah, we were building software, 00:09:51,444 --> 00:09:57,364 [Craig Kerstiens] but just creating things just for the sake of it, for fun because we can- 00:09:57,364 --> 00:09:57,713 [John Wessel] Yeah 00:09:57,713 --> 00:09:59,504 [Craig Kerstiens] ... is kind of a beautiful thing. 00:09:59,504 --> 00:09:59,664 [John Wessel] Yeah. 00:10:00,724 --> 00:10:02,684 [Craig Kerstiens] I mean, I'm also loving how much, like... 00:10:03,784 --> 00:10:08,784 [Craig Kerstiens] Back at Heroku, we had this thought of, like, the idea of personalized software and everyone should be a developer. 00:10:10,204 --> 00:10:22,144 [Craig Kerstiens] For an interview question, we had everyone that interviewed at Heroku build an app and deploy to Heroku. When I say everyone, I mean, like, HR, like our receptionist, our marketing folks. And they're like, "No, I'm not a developer." 00:10:22,144 --> 00:10:23,004 [John Wessel] Yeah. 00:10:23,004 --> 00:10:25,344 [Craig Kerstiens] It's like, no, this is our... 00:10:27,104 --> 00:10:28,704 [Craig Kerstiens] I mean, it was almost cult-like, right? 00:10:28,704 --> 00:10:29,444 [John Wessel] Yeah. 00:10:29,444 --> 00:10:36,623 [Craig Kerstiens] And I think now we're at that stage where everyone can be a developer, and we now have personalized software. 00:10:36,623 --> 00:10:37,764 [John Wessel] Yeah. 00:10:37,764 --> 00:10:48,844 [Craig Kerstiens] Most of my team has, like, individual Mac apps that, like, review pull requests. Like, basically they manage their work in a personalized Mac app they built. 00:10:48,844 --> 00:10:49,073 [John Wessel] Yeah. 00:10:49,073 --> 00:10:53,874 [Craig Kerstiens] And instead of standardizing on that, I'm like, "No, build your own for your own workflow." For me, right, 00:10:54,924 --> 00:11:05,364 [Craig Kerstiens] I have a work in progress Mac app that analyzes all my Slack messages, all my emails, give me a summary of my day heading into my day, which things do I need to respond to, like, perfectly tailored to how I work. 00:11:05,364 --> 00:11:06,224 [John Wessel] Yeah. 00:11:06,224 --> 00:11:26,624 [Craig Kerstiens] So I think it's both productive and just fun that we're creating new stuff. I mean, my talk was largely written by... It was, like, pulled from my content, largely written by AI. And then I'm like, "By the way, I want to play, like, an ASCII version of Star Wars in PostgreSQL." 00:11:26,624 --> 00:11:27,964 [John Wessel] Yeah. 00:11:27,964 --> 00:11:29,884 [Craig Kerstiens] Do I need to do that? No, but it's fun. 00:11:29,884 --> 00:11:30,464 [John Wessel] Right. 00:11:30,464 --> 00:11:33,584 [Craig Kerstiens] If it took me 10 hours personally to do it, you know- 00:11:33,584 --> 00:11:34,424 [John Wessel] Yeah 00:11:34,424 --> 00:11:35,044 [Craig Kerstiens] ... it's not worth it. 00:11:35,044 --> 00:11:35,564 [John Wessel] Yeah. 00:11:38,984 --> 00:11:45,573 [Craig Kerstiens] The pain or the hard or the negative, I don't know that I see too many. I think it's like the pace of change, right? 00:11:45,573 --> 00:11:45,604 [John Wessel] Yeah. 00:11:45,604 --> 00:11:51,294 [Craig Kerstiens] Like, it's hard to adapt to it. We're changing, like, every few months now, and 00:11:52,404 --> 00:11:54,784 [Craig Kerstiens] you've got to get used to it. And change is hard for everyone. 00:11:54,784 --> 00:11:54,884 [John Wessel] Yeah. 00:11:54,884 --> 00:12:00,204 [Craig Kerstiens] I think that's the biggest part is, like, if we could go slower, it might be nice, but that's not the reality of the world we're in. 00:12:00,204 --> 00:12:10,984 [John Wessel] Right. Yeah. That has been one of my personal frustrations of doing all this optimization, because I love optimizing workflow, which has always been one of my passions. 00:12:12,264 --> 00:12:35,820 [John Wessel] So I'll get this workflow that I love and it's optimized, and then, like, there's a new change, like a new model comes out. Even if you pick an ecosystem, there's still constant change. Even if you pick, "I'm going to be in the Codex or Cloud or Pi or whatever you want ecosystem," there's still constant change. And I think I've experienced that, like, I just don't want to re-optimize this today, even though I love it, you know? 00:12:35,820 --> 00:12:36,540 [Craig Kerstiens] Yeah. 00:12:36,540 --> 00:12:51,060 [John Wessel] So which, again, and it's a push and pull because sometimes you can over-optimize and you can spend too much time because it's fun or because you're just able to now and you weren't able to. So yeah, I see that. 00:12:53,360 --> 00:12:54,360 [John Wessel] Okay. So, 00:12:55,620 --> 00:12:59,280 [John Wessel] kind of moving to the future, like things that you see as far as trends. 00:13:00,460 --> 00:13:05,160 [John Wessel] What's exciting to you if you were to kind of look out? 00:13:06,180 --> 00:13:06,320 [John Wessel] I mean, 00:13:07,480 --> 00:13:11,000 [John Wessel] 5 years even seems like forever at the current pace of change. Look out a few years. 00:13:12,340 --> 00:13:16,380 [John Wessel] What are some things that maybe are a little bit scary, and then what are some things that are exciting to you? 00:13:16,380 --> 00:13:28,160 [Craig Kerstiens] Yeah, I mean, I think the biggest meta is the age of actual usage of AI in production settings and enterprises is here now. It's not like, "Oh, it's coming and we're going to get there." 00:13:28,160 --> 00:13:28,170 [John Wessel] Right. 00:13:28,170 --> 00:13:30,020 [Craig Kerstiens] It's like literally here- 00:13:30,020 --> 00:13:30,180 [John Wessel] Yeah 00:13:30,180 --> 00:13:30,780 [Craig Kerstiens] ... right now. 00:13:30,780 --> 00:13:31,280 [John Wessel] Right. 00:13:31,280 --> 00:13:39,520 [Craig Kerstiens] We had an inflection point, I think, with a lot of the models back in October, November of last year, right? It's like these are good enough. The pace just keeps up. 00:13:39,520 --> 00:13:39,770 [John Wessel] Right. 00:13:39,770 --> 00:13:40,900 [Craig Kerstiens] The tooling around them. 00:13:42,700 --> 00:13:48,390 [Craig Kerstiens] I think, one is that we're here now. It's like not, "Oh, what's the future for AGI?" 00:13:48,390 --> 00:13:48,420 [John Wessel] Yeah. 00:13:48,420 --> 00:13:54,240 [Craig Kerstiens] I think the other piece is, again, I kind of hit on that, like this age of personalized software. 00:13:56,000 --> 00:14:00,410 [Craig Kerstiens] I don't have to share the same dashboard as you. I can easily build my own dashboard and achieve. 00:14:01,480 --> 00:14:04,820 [Craig Kerstiens] We can really curate software to our own liking and usage, 00:14:06,160 --> 00:14:13,860 [Craig Kerstiens] which is pretty fascinating. No one ever said before like, "Oh, code's the barrier," right? 00:14:13,860 --> 00:14:13,980 [John Wessel] Right. 00:14:13,980 --> 00:14:21,680 [Craig Kerstiens] So now it's been proven that it kind of is. Now it's like, no, the fastest we can create ideas and iterate and understand the business, 00:14:22,720 --> 00:14:29,700 [Craig Kerstiens] it's a pretty crazy industrial revolution-esque time right now, I think, in terms of software. 00:14:29,700 --> 00:14:31,019 [John Wessel] Yeah, for sure. 00:14:32,140 --> 00:14:33,580 [John Wessel] I think, 00:14:35,180 --> 00:14:37,820 [John Wessel] yeah. I mean, I think the change is so rapid right now. 00:14:39,780 --> 00:14:43,140 [John Wessel] So if we kind of take that same thing and apply it to 00:14:44,300 --> 00:14:44,940 [John Wessel] data, 00:14:46,580 --> 00:15:06,300 [John Wessel] because I was telling you before the show, I've spent most of my career in data, lots of kind of analytics, database-related things. And what do you think that looks like? How would you apply that personalization thing to data? Because we talked a little bit about workflow, but what is it? 00:15:06,300 --> 00:15:36,080 [Craig Kerstiens] Yeah. So I think there's 2 trends in the data space right now, right, that have changed, and it's not that they weren't present before, but the AI has made those needs way more acute, right? And so in the data space, it's an age-old problem. You've got your transactional system and you've got your analytics system. You've got your system of record, your system of engagement, right? The access patterns are completely different. The way they're designed is different, right? One is a row-based where you're inserting single records. The other is I'm scanning all of the orders yesterday and I want to know the report, right? 00:15:36,080 --> 00:15:36,700 [John Wessel] Right. 00:15:36,700 --> 00:15:46,240 [Craig Kerstiens] Row-based, column-based, like single point lookup, mass scans and reports, right? Like, this is 30-year-old database foundations that have not changed. 00:15:46,240 --> 00:15:47,260 [John Wessel] Right. 00:15:47,260 --> 00:15:55,920 [Craig Kerstiens] And yet the need is, like, AI doesn't want stale data. It wants those orders over in the analytical system immediately. 00:15:55,920 --> 00:15:56,220 [John Wessel] Right. 00:15:56,220 --> 00:16:03,340 [Craig Kerstiens] It wants to have that tight feedback loop of I make a decision on my analytic systems, it feeds back into my operational system. 00:16:05,840 --> 00:16:13,420 [Craig Kerstiens] That latency and delay and pain of like, I've got a colleague that jokes like the E in ETL stands for evil. 00:16:13,420 --> 00:16:14,889 [John Wessel] Yeah. [laughs] That's funny. 00:16:14,889 --> 00:16:22,660 [Craig Kerstiens] And I love it because it's also a necessary evil. Like, you can't be like, "No, I don't want to ETL the data between the systems because it's hard and I don't like it." 00:16:22,660 --> 00:16:23,020 [John Wessel] Right. 00:16:23,020 --> 00:16:23,829 [Craig Kerstiens] No, you have to do it. 00:16:23,829 --> 00:16:24,120 [John Wessel] Right. 00:16:24,120 --> 00:16:25,520 [Craig Kerstiens] You don't have a choice. It sucks. 00:16:25,520 --> 00:16:26,010 [John Wessel] Right. 00:16:26,010 --> 00:16:26,840 [Craig Kerstiens] It sucks for everyone. 00:16:28,060 --> 00:16:40,560 [John Wessel] Yeah, and it's because you've got 2 different optimizations, like one for transactional, like working system, and then one for analytics. And so far, we don't quite have that merge, which I think there's some people working on that. 00:16:40,560 --> 00:16:42,499 [Craig Kerstiens] Yeah. So I mean, [both laughing] 00:16:43,680 --> 00:16:50,560 [Craig Kerstiens] we're working on our own, like what we're calling mirroring, where you can single button mirror the tables you want in real time from Postgres into Snowflake. 00:16:50,560 --> 00:16:51,160 [John Wessel] Yep. 00:16:51,160 --> 00:16:51,570 [Craig Kerstiens] And 00:16:52,840 --> 00:17:03,410 [Craig Kerstiens] I was talking with someone the other day that's like, "I think I have my architectural foundation for years to come." And another customer that had spent 3 years working on their pipelines for this. 00:17:03,410 --> 00:17:04,200 [John Wessel] Yeah, sure. 00:17:05,320 --> 00:17:19,060 [Craig Kerstiens] Sorry, 3 months working on their pipelines. And they were like, "About to flip it to production." And I demoed this to them, and they're like, "I get to delete 3 months of code." I'm like, "I feel so bad you wasted 3 months." They weren't upset at all. They're like, "I didn't even know I was going to manage this." 00:17:19,060 --> 00:17:23,240 [John Wessel] Yeah. Which just a few years ago may have been 3 years of code, you know? 00:17:23,240 --> 00:17:24,640 [Craig Kerstiens] Yeah. And so I think, 00:17:25,700 --> 00:17:29,320 [Craig Kerstiens] so what's that mean? Like before, the business couldn't react fast enough. 00:17:29,320 --> 00:17:29,640 [John Wessel] Right. 00:17:29,640 --> 00:17:31,120 [Craig Kerstiens] We couldn't make decisions, right? 00:17:31,120 --> 00:17:31,280 [John Wessel] Right. 00:17:31,280 --> 00:17:32,439 [Craig Kerstiens] We couldn't change the business. 00:17:32,440 --> 00:17:32,860 [John Wessel] Right. 00:17:32,860 --> 00:17:38,379 [Craig Kerstiens] And with AI and how fast we're able to move now, that's why that matters, I think, more than ever. 00:17:38,380 --> 00:17:38,640 [John Wessel] Yeah. 00:17:38,640 --> 00:17:40,620 [Craig Kerstiens] And so it's a very foundational piece. 00:17:41,700 --> 00:17:46,890 [Craig Kerstiens] It's funny, I've run Postgres for folks for a very long time, and it's kind of a thankless job. 00:17:46,890 --> 00:17:46,900 [John Wessel] Oh, okay. 00:17:46,900 --> 00:17:48,620 [Craig Kerstiens] It can take a while to get used to it. 00:17:48,620 --> 00:17:48,780 [John Wessel] Yeah. 00:17:50,060 --> 00:17:54,980 [Craig Kerstiens] As a PM, you go and ask someone if they have feedback about your website. 00:17:54,980 --> 00:17:55,780 [John Wessel] Mm-hmm. 00:17:55,780 --> 00:17:58,260 [Craig Kerstiens] And you're like, "I don't like your font." 00:17:58,260 --> 00:17:58,500 [John Wessel] Yeah. 00:17:58,500 --> 00:17:59,720 [Craig Kerstiens] "I don't like the shape of the button." 00:17:59,720 --> 00:18:00,280 [John Wessel] Yeah. 00:18:00,280 --> 00:18:04,380 [Craig Kerstiens] You're like, "I just wanted to know if your package shipped okay was all I was asking," right? 00:18:04,380 --> 00:18:05,220 [John Wessel] Right. 00:18:05,220 --> 00:18:09,720 [Craig Kerstiens] And so for your transactional system, it's similar. 00:18:09,720 --> 00:18:10,920 [John Wessel] Yeah. 00:18:10,920 --> 00:18:16,180 [Craig Kerstiens] It's like, how do you feel about your bank? When's the last time you went into anything and you're like, "Thank you for keeping my money safe." 00:18:16,180 --> 00:18:18,140 [John Wessel] Yeah, exactly. [laughs] 00:18:18,140 --> 00:18:21,240 [Craig Kerstiens] Now they lose your Social Security number or, you know- 00:18:21,240 --> 00:18:21,439 [John Wessel] Right 00:18:21,924 --> 00:18:24,004 [Craig Kerstiens] ... miss a deposit, you're pretty unhappy- 00:18:24,004 --> 00:18:24,013 [John Wessel] Yeah, right 00:18:24,013 --> 00:18:25,004 [Craig Kerstiens] ... and you talk to them. 00:18:25,004 --> 00:18:25,184 [John Wessel] 100%. 00:18:25,184 --> 00:18:26,244 [Craig Kerstiens] It's the same thing with- 00:18:26,244 --> 00:18:26,504 [John Wessel] Yeah 00:18:26,504 --> 00:18:27,623 [Craig Kerstiens] ... your transactional side. 00:18:27,624 --> 00:18:27,944 [John Wessel] Mm-hmm. 00:18:27,944 --> 00:18:29,764 [Craig Kerstiens] Your analytical side's completely different. 00:18:29,764 --> 00:18:30,524 [John Wessel] Yeah. 00:18:30,524 --> 00:18:38,924 [Craig Kerstiens] But now with the need for these 2 to kind of unify and tie together, it's really like AI has just accelerated how fast we need to move and react- 00:18:38,924 --> 00:18:39,054 [John Wessel] Yeah 00:18:39,054 --> 00:18:44,283 [Craig Kerstiens] ... and lowered that tolerance for stale data. Like, AI doesn't want to wait a day- 00:18:44,284 --> 00:18:44,384 [John Wessel] Right 00:18:44,384 --> 00:18:45,804 [Craig Kerstiens] ... for those pipelines to run. 00:18:45,804 --> 00:18:46,744 [John Wessel] Right. 00:18:46,744 --> 00:18:53,304 [Craig Kerstiens] So, I think in the data space, it's foundational where AI has just surfaced this need and reduced, like, 00:18:54,364 --> 00:18:56,504 [Craig Kerstiens] the time that we can wait- 00:18:56,504 --> 00:18:56,564 [John Wessel] Yeah 00:18:56,564 --> 00:18:57,264 [Craig Kerstiens] ... for those kind of decisions. 00:18:57,264 --> 00:19:00,724 [John Wessel] Yeah, the friction. Yeah. All right, kind of last question here. 00:19:01,824 --> 00:19:02,244 [John Wessel] What are... 00:19:03,304 --> 00:19:15,964 [John Wessel] You can go personal or professional here. What are some tips or tricks that you've seen personally, professionally in working with AI on a daily basis of things, maybe things that are easy to miss or things that weren't intuitive to you at 1st? 00:19:18,524 --> 00:19:24,564 [Craig Kerstiens] I think the biggest thing is that mental mind shift. Like, I could do this before and keep doing it the same way- 00:19:24,564 --> 00:19:24,574 [John Wessel] Mm-hmm 00:19:24,574 --> 00:19:26,923 [Craig Kerstiens] ... or like, "No, I'm going to do it the AI way 1st." 00:19:26,923 --> 00:19:26,943 [John Wessel] Right. Yeah. 00:19:26,944 --> 00:19:32,154 [Craig Kerstiens] Like, can I do every bit of... Like, can I use AI for literally everything that I'm doing? 00:19:32,154 --> 00:19:32,644 [John Wessel] Mm-hmm. Right. 00:19:32,644 --> 00:19:35,004 [Craig Kerstiens] Like, if I'm doing something, can I do it in an AI way? 00:19:35,004 --> 00:19:36,444 [John Wessel] Like, yeah. 00:19:36,444 --> 00:19:39,944 [Craig Kerstiens] And that, like we talked about, that constant change, it's hard. 00:19:39,944 --> 00:19:40,024 [John Wessel] Right. 00:19:40,024 --> 00:19:43,504 [Craig Kerstiens] Like, I've got my workflow. I've used my workflow for 10 years. I don't want to change it. 00:19:43,504 --> 00:19:44,104 [John Wessel] Exactly. 00:19:45,344 --> 00:19:57,904 [John Wessel] That's why I got into databases, is like, I don't want to do front end because it changes too fast. That was one of my actual reasons. Because the SQL, like SQL, what you use to query a database with, like, the evolution of that's slow. It's like tweaks every year- 00:19:57,904 --> 00:19:58,024 [Craig Kerstiens] Yeah 00:19:58,024 --> 00:20:02,424 [John Wessel] ... very slow. I picked that on purpose. And now [chuckling] I feel like, what's up? 00:20:02,424 --> 00:20:03,724 [Craig Kerstiens] Now it's out the window. Yeah. 00:20:03,724 --> 00:20:03,884 [John Wessel] Yeah. 00:20:03,884 --> 00:20:10,124 [Craig Kerstiens] You have no hope now. Yeah, no, it's funny. I feel very similarly about, like, front-end development. I'm like, "I don't know what the latest framework is." 00:20:10,124 --> 00:20:10,784 [John Wessel] Right. 00:20:10,784 --> 00:20:12,244 [Craig Kerstiens] It changes every 6 weeks. 00:20:12,244 --> 00:20:12,583 [John Wessel] Right. 00:20:14,424 --> 00:20:15,504 [Craig Kerstiens] Postgres is Postgres. 00:20:15,504 --> 00:20:15,614 [John Wessel] Yeah. 00:20:15,614 --> 00:20:16,583 [Craig Kerstiens] It's the same thing for- 00:20:16,584 --> 00:20:17,024 [John Wessel] Right, right. 00:20:17,024 --> 00:20:24,014 [Craig Kerstiens] It's gotten better, it's gotten improvements, but not at the same scale. So, unfortunately, I think it is the reality of what we're in. 00:20:24,014 --> 00:20:24,044 [John Wessel] Right. 00:20:24,044 --> 00:20:34,464 [Craig Kerstiens] And the faster people adapt to that and say, "You know what? I get it. I'm embracing the change versus fearing it." Because pretending the Industrial Revolution wasn't going to happen, like- 00:20:34,464 --> 00:20:35,244 [John Wessel] Right 00:20:35,244 --> 00:20:36,444 [Craig Kerstiens] ... wasn't reality, right? 00:20:36,444 --> 00:20:36,543 [John Wessel] Right. 00:20:38,084 --> 00:20:47,064 [Craig Kerstiens] And so, it's a big shift, I think just personally for people. How much can you willingly embrace that change is the biggest thing. 00:20:47,064 --> 00:21:01,924 [John Wessel] Yeah. Yeah, and I agree with the mindset. For me, it's been, I need to try essentially everything with AI 1st and disprove that it doesn't work versus the opposite, right? And that's a big shift for a lot of people. 00:21:01,924 --> 00:21:03,984 [Craig Kerstiens] And I think it's retry it again 3 months later. 00:21:03,984 --> 00:21:38,924 [John Wessel] Yeah, right. That's the other thing. Yeah, that's a good point. You'll love this. So, in setting up for the podcast here, never done really a mobile setup. So we have a Mac Mini that's all set up and ready in our studio. So I had to set everything up on my laptop, which if you've ever messed with AV stuff, the software's awful. Like, it's super hard. It's a big pain. So, rather than just tinkering with it, I used AI on all the config and settings, and it, like, had to restart my computer a bunch of times, of course, and enable, like, legacy system this and that. But it worked. And I tried it before, and it didn't work 6 months ago. 00:21:38,924 --> 00:21:39,284 [Craig Kerstiens] Yeah. 00:21:39,284 --> 00:21:39,984 [John Wessel] So. 00:21:39,984 --> 00:21:44,764 [Craig Kerstiens] Yeah, I think it's like, don't make an assumption, "I tried this. It doesn't..." Like, the pace of change is- 00:21:44,764 --> 00:21:44,904 [John Wessel] Right 00:21:44,904 --> 00:21:47,224 [Craig Kerstiens] ... every few weeks, every few months. 00:21:47,224 --> 00:21:48,044 [John Wessel] Right. 00:21:48,104 --> 00:21:50,404 [Craig Kerstiens] And I think a lot of people that still, 00:21:51,704 --> 00:21:59,624 [Craig Kerstiens] especially in any technical realm, right, even whether it's on the business side or the marketing side, that had an opinion of it 12 months ago and wrote it off, like- 00:21:59,624 --> 00:22:00,064 [John Wessel] Right 00:22:00,064 --> 00:22:01,763 [Craig Kerstiens] ... it's a very, very stale opinion. 00:22:01,764 --> 00:22:01,953 [John Wessel] Yeah. 00:22:01,953 --> 00:22:04,324 [Craig Kerstiens] It feels like it's 10 years old versus a year old. 00:22:04,324 --> 00:22:08,684 [John Wessel] Yeah. Yeah, for sure. Awesome. Well, thanks for joining us. Thanks for being here, Greg. 00:22:08,684 --> 00:22:09,414 [Craig Kerstiens] Yeah, thanks so much for having me. 00:22:09,414 --> 00:22:18,074 [John Wessel] And we'll catch you on the next one. [upbeat music]