Welcome to Energy Bytes with John Kalfayan and Bobby Neelon. Your essential guide to the intersection of data and energy. This podcast dives deep into the world of energy, shedding light on how data, AI, and technology are revolutionizing this sector. Each episode equips listeners with insights into the most efficient tools and resources, paving the way for a data-driven future in energy. From technical nuances to broader industry trends, Energy Bytes offers an unparalleled perspective on the evolution of the energy industry. Join us as we decode the algorithms of energy, one byte at a time.
Welcome back everybody. Uh got another another exciting episode of energy bites. I've been looking at our transcripts and we say exhilarating almost every time. Switch it up with exciting. [laughter] Another hyperbole to uh the stories. This is not brought to you by chat GBT even though it may sound that like it is. Uh here we are guys. appreciate uh our guests today, John and Josh from uh Well Database coming on and and chatting with us. How you guys doing? Good. How you doing? Doing good, man. Thanks for having us on, man. Yeah, excited to have you here. We are. We uh been trying to get this for a while. Yeah. Yeah, mainly my fault, but finally got schedules all aligned, but you know, John's off living in the mountains and Josh is still almost an hour away, even though he's in the same town as us [laughter] in the woodlands. But that's what I was going to say. It's no different. It takes him as long to get there takes me. Yeah. Yeah. You fly into he drives around. Yeah. It would probably be faster to come here from from Bush than it would be from the woodlands, [laughter] I guess. [snorts] Oh man. Oh yeah. Anyway, how you doing, Bob? You good? Yeah, I'm great. Life's good. Yeah, Bobb's uh Bobby's swatting away all the the Grace and Mel rumors that just recently dropped. So, we're uh we're going to avoid that because of that. But, are you guys you guys are the founders of Well Database? Tell tell everybody kind of just what database is real quick if they're not familiar. I think most people heart because you guys have have built such a great product and yeah great following. One thing we've been asking people too when they come you know to our business lately um when they're presenting is like what was the problem statement too you know kind what were we trying to solve? Oh my god. You see, and and I've been thinking about this a little bit too because I people start companies for lots of different reasons. And I feel like people skim over that. Like the reason you start a company really kind of sheds light on your goals and where you're trying to go and you know where the customer lies within those goals and I think that says a lot about it. Um [snorts] but you know Josh and I worked together for since we were kids. I don't know how that's happened but like we met when we were really young. Now you're uh I am old. Yeah. Josh made the 40 under 40. I'm I aged out. So, [laughter] um but no, we we worked we worked at a few different places. We were together at Slumber. We worked at a consulting group. Um and uh you know, we were working at New Tech Energy at the time and we were, you know, using one of our competitors today um to load data into an ERP system. We had a staff of people loading and cleaning data. We had uh you know, an expensive contract and then they bumped us 50% on the contract, which was already high. Um, and I think, you know, Josh and I had been tossing around what we want to be doing because, you know, we knew we were game for starting a software company. Um, and so we just got sick of it and we were like, we don't have API connections. We are trying to jump through hoops. We're paying exorbitant fees for mediocre data and we're like, you screw this. These states are online. We're going to do this ourselves. We're going to provide data with better terms, better prices. We're going to automate this whole thing. And then yeah, uh those kind of fateful last words because it would be years later. [laughter] Oh god. Yeah. Years later, years later, we were finally done. We got the the automation system in place which runs today which is I mean people don't realize it and we go eager with it and on the tech side because we we run, you know, millions of jobs every single day and people like I mean what changes for millions? I mean there's six million wells in our database like how many how can you run so many jobs but we go through and just really eager for looking for updates changes anything and it's all automated so it doesn't cost us anything to do so we just we get really kind of overzealous with spinning off jobs and acquiring that data so we go out to all the sources though all the state sources regulatory bodies and you know our model that we built um is a little bit bigger and so we can go out to you know university systems you know the frack focus is an easy one USGS um university lands, you know, certain aspects of data out of the EIA and we pull it all together. We aggregate it all together and give it in one one kind of simple interface so that people can get stuff uh get the data on the public side that they need and then you know since it's all automated, we don't have the headcount the other guys have. We we don't have the delays the other guys have. You mean you don't have pass on manually doing data entry? Yeah. Yeah. We had we had somebody tell us we had somebody tell us once they came into our office and they said they had been to one of our competitor's office and they walked through uh I guess they were giving him a tour or something and they said that this dude had a stack of CDs on his desk and he asked what is that and he's like well this is data is how we load it. We pulled the CDs in and then we copied the data out and he was blown away by that and I was too. I don't I don't know what they were talking about. It doesn't make any sense. You ask him if they CDRs or RWS? [laughter] Yeah. Well, yeah. That brings up so many like nostalgic memories of me pirating music as a child and just burning DVDs and Yep. CDs. I you couldn't No. Yeah. We were having this conversation about this the other day with my son about the the PlayStation um and the Nintendo 64 and it was all around Final Fantasy moving to the PlayStation because if you guys played Final Fantasy 7, it's four discs and the cartridges on the N64 were like 400 meg. No, no, like 40 meg. Yeah, 40 meg. It was going to take 40 of those or something like that. Anyway, so yeah, the CDs, we we give talk trash about them now, but man, that was a big jump. Oh, I mean, I I will never forget wanting a zip drive so bad before CDs really kind of took off, right? Everybody did, right? Yeah. No, sloppies and then C those were like the next thing that But they were so much bigger and then Yeah. Almost. And then you had the jazz that was even bigger, right? Yeah. And then the CDs. Jazz Drive. Didn't the jazz drive, dude? [laughter] Yeah. I don't remember what that was. Was that a gigabyte of storage? I couldn't remember storage. I think how expensive was that? Oh, they were like it's just crazy to look at that. Like you're in Walgreens now and you can buy a 256 gig flash drive that fits on your keychain for 20 bucks. Well, even like I mean I mean I've talked to people who were in like you know hardware sales back in even like early 2000s and it was like right now we talk about you know you mentioned Snowflake. I'm sure we'll get a little bit into that later, but like it's 23 bucks per terabyte compressed and like they used to sell terabytes for like exorbitant amounts of money. Like it's just insane. Yeah. No, we uh Yeah. In the late 90s, I always remember this that it was a dollar per meg for memory. And uh that was, you know, you wanted a gig of memory. It was $1,000. Now, uh I just bought uh 64 gigs for a machine we built here. I think that did cost me $100 for that. So [laughter] insane. Oh man, we're dating ourselves. This is fun. Um, no, it's good. You got to know where you came from. These kids these days, they don't I'm just No, I don't feel that old. [laughter] No, I don't. I'm I'm just like I genuinely feel like Transistor Radio John, [laughter] I genuinely feel like though my eight track and my my CD. Anyway, our generation though time like I feel like got the opportunity to come through at a special time because I feel like the generation but like our parents they you know they don't they're not very good with technology at all. It's all very new to them but they never got the like opportunity to really try and learn it. And then the next generation like everything has gotten a lot better from a graphical interfa like all of the stuff on the technology side has gotten better and easier to use. And so like you grow like I I can't tell you how many computers I broke just messing around with stuff because that's all there was to do on the computer. Like you know you had your games on your floppy discs or whatever and that was really it. And then file sharing and all of that stuff started and it's like man I really learned a shitload about computers because of that like hacking kind of open source mentality from the beginning. And I wonder like if that ends up going away because of you know things get so much better or if it just shifts to you know a different area. But I feel like everything also got like we got to grow up with it getting progressively more difficult if that makes sense. Like everything was super simple in the early 90s both hardware and software you know and you got to toy around and then as you got older software got more and more complex and all these things and now you got these huge areas of software that are totally you know branches that didn't exist back then. So we kind of come along there and now if you jump in from like our current our kids age, you know, everything's already way up here. They don't hit that. Well, that's it. Yeah, that's what I think. They're just starting at a better place. And so like their hacking is a lot more interesting than our hacking was. Our hacking was, you know, can we make this DOSs prompt say something back to us? Like who [laughter] cares? Yeah. Reading computer manuals. Yeah. And our kids are now trying to like make videos of themselves and Yeah. My kids don't understand the concept of a mouse. like they don't understand why you would use a mouse when they can just touch everything. Well, my dad just brought this old camcorder he gave me as a gift when I started college and he brought it over and he's like, "I thought you might want this." Like, "I don't I will never use that." And my kids are like, "What's that?" Like, I mean, like it just thinks huge like and not nearly as big as the thing that they used to video my second birthday where my dad like rented it and was like on his shoulder. [laughter] Um, all those family moments on the No, I have a mini DV cam. Oh, yeah. I got a mini DV cam sitting somewhere in a box somewhere. I don't know why I still have it. It's complete garbage. But anyway, no, I I was going to say though, I remember my dad bringing home his first computer. It was in the 80s, late '8s, and he was showing me how you can save a Word document on a disc. They didn't have enough memory really to store or storage to store anything locally. And I was like, "This is the biggest waste of time." I think I said that to him. I was like, "Dad, this is dumb. This is a waste [laughter] of time. What is Who cares about computers? We'll never use this, Dad." Yeah, it's like like what was it today's show at Katie Kirk and then we're laughing at email like who's ever going to use this? You know, they thought it was so fun digital messaging. Yeah. Um well, cool. Like so let's jump into well databases. Um I guess I mean again maybe the tech story because I mean I'm really interested because I know I've talked to you over the last few years and I feel like even some things have changed maybe philosophically potentially or maybe they didn't. I just made some assumptions. But um let's talk about like how you know maybe where the tech stack started where you're at now and like you know why there were certain evolutions you know whether it was hardware and or software decisions and yeah even like a step before that too which is you know what were the like what were the the real hard pain points you were trying to to solve with the product because you know in my opinion there's you guys you know you've got combo curve you've got a bunch of these companies that have spun up the the root of what they're trying to do isn't different than what exists on the market, but how they're doing it is basically if you blew up the existing company and started from scratch at that point in time, right? And so there's a lot different ways and techniques that you guys use versus your competitors or even just regardless legacy software products that um I'm very curious to kind of hear y'all's how y'all strategized that, architected out, all that stuff. Yeah. Yeah. the very beginning number one everything had to be automated like we're a small team you know so that was like a foundational thing that has run all the way till today you know now we're starting to you know lean on even AI and all that to automate document processing and that kind of stuff but even back then everything always had to be automated and that drove a lot of our choices and then cost you know what drove more of our choices yeah being bootstrap we we didn't have any say besides that yeah And the the fact is that the the numbers most of the companies you're talking about there um and most companies today they they just they don't have any money. You hop on the cloud and you start some incremental expenses on you know because if you just are doing a web app on the cloud with some minimal user storage or whatever it may be it's it's cost is fine. It's not a big deal. You can squeak by for a couple hundred bucks a month and and be just fine. our product day one had to have the data and this data is not we joke about this it's not big data but it's not small data it's big-ish data um and so like we couldn't the the sheer and okay you also rewind this back to when we started back in you know very first in 2008 but really more in earnest in 10 or so um the storage it was going to cost us for um just the storage and the database access that we needed to function was going to be in the range of $10,000 per month. And so that automatically made that just a no-go for us when I mean we were bootstrapped and we don't Josh and I aren't wealthy people. We didn't come from money or anything like that. So this is a real bootstrap like work on it at nights and until it can make some money and then keep working on it at nights. But anyway, so you know, our goal was to peiemeal what we could get together, what we could pay for out of pocket and just kind of, you know, beg, steal, and borrow what we can to get off the ground. And so, [snorts] um, we've always hosted our own infrastructure because of that. I was going to say even when we started there, 2010, you didn't really have Azure cloud you were spinning up, right? You know, it was starting to become a thing, but when we started it wasn't. I mean, when we started Yeah. And so, and so cost was extremely high. And then uh so when we started you know we knew we we were Microsoft shop where we worked and so we kind of knew that and at the time Microsoft launched their bispark program which let you get in some licensing and all that so that helped us offset or get our database stuff and you know kind of drove choices along the way as well. Yeah there's a couple things there. Uh, one I've I've had this I had this conversation earlier this week with somebody as far as like well we need this function to happen and we need you know should we write a script or should we find you know a different way to do it. I was like, "Well, let's use and I it's funny because it's like a few years ago. You can ask Bobby. I would have never said this out loud, but I was like, well, let's just use like a Zapier or make or something like that to get to start it and get it in place because we need the speed of this feature and then we will turn it to the developers and get it on their backlog and then they will get to it when we have the time in our our backlog to get to it." But uh you know there's that like I I think bootstrap companies are so interesting because they the mentality is just so much different right it's like survive is literally the mentality and so like you learn these things of like oh well I can use a low code tool in the short term to get this moving and keep things rolling and put this in the backlog in the long term so that I can you know get to it when I need to get to it not have to drop everything today and or hire another developer like there's So, especially now, like I know back then it probably wasn't nearly as as uh fruitful as as it is now with all the different integration tools and and RPA and no code stuff, but I think that's a big thing that a lot of people just don't they don't think about it, right? Or like in principle, they're against it. Like I hated Zapier for the longest time because it's like I'm just paying you for these stupid API calls that I know I could write, but I don't have the time to write and maintain a hundred of them, right? And so it's like, okay, now I understand the value of what you do, right? Yeah. Yeah. No, we've definitely seen some of that. It's hard. Josh and I are both technical obviously. We u we're both programmers and so we definitely run more into, you know, you run into a problem, let's just write the code to solve it and do it once and be do it right. And which sounds nice except for those the tasks add up much faster and you can complete them. And u I mean it's a trade-off. I mean, one's a technical debt that you're taking on. Like, if you do that script or no code or whatever it is, it's it's a piece of debt and you know, one two as things go on these technical debt things start to stack up and you know, you have to manage that debt and your time and you know, find walk that fine balance. Yeah. No, that's it's actually a habit I've got into as piece of advice I'll throw out there. that if you do use these no code tools or whatever with your applications or software or whatever they may be when you set them up go and throw that function into your backlog as tech debt. I have a tech debt tag that I put in there now because well it you know like okay putting it in the backlog and not categorizing it in that way just means it's another backlog item and there's no like priority put on it or anything like that. But putting it as tech debt like just that word debt feels bad to humans, right? Like we don't like that. And so it's like, hey, these are things that are just flush it out one day, right? And so like at some point we're gonna need to fix this [ __ ] And so it's not going to just go get lost in the backlog somewhere. But [snorts] um Right. No, [clears throat] I think another thing you mentioned was like the uh the startup initiatives and programs that a lot of these guys have. Like I knew that they were there. I didn't realize how easy they were to like get and apply for. And you know, a lot of them out of the box are like5 $10,000 worth of credits most of the time now. Up to you know. Oh yeah. Yeah. In the credit side and this Bispark one, you know, again, being a Microsoft, Josh and I being Microsoft kind of in in our past corporate lives. Um, you know, Microsoft at the time and I don't know, it gets bad wraps here and there, but honestly, you know, it's most of the industry uses. [laughter] Yeah, exactly. I mean, this was the start of new Microsoft for what it's worth. So, you know, it was Yeah. So, I mean, I'm assuming and so, you know, being Go ahead. Sorry. No, no, go ahead. you I inter I interrupted you. Well, no, no, it's just the, you know, it's funny uh because even today you you see the breakdown of developers on Stack Overflow surveys and stuff like that. You know, these Python developers are, you know, ruling the world and then, you know, you got all the other kind of weird offshoots and backend. Anyway, uh .NET gets this like weird, you know, corporate rap that people don't want to jump into, but you know, if you play with net core. I mean, you can do you can script as easily as you can in Python using core and spin it up in in the same ways. And so anyway, I I don't it's funny. I still think it gets a bad rap. Um unnecessarily. Yeah. I mean, Microsoft in general, I mean, even like it was like, oh, they're so anti- ant open source and then they bought GitHub, but when they bought GitHub, they were the largest contributor to open source, you know, it's like which again they've always had that whole embrace, extend, extinguish thing, you know, in their history. So, you kind of wonder sometimes about their intentions, but again, they're going to be here for the long haul and I think you can build off of them and uh it's not just a flavor of them, flavor of the month. It's going to be supported and there's going to be stuff behind you that I think a lot of that just stems from like our how we grew up thinking of Microsoft and Apple like our generation's perspective of those softwares, right? And like that 90s early 2000s version of Microsoft is very different in my opinion than what they are now. But it took a long time for people to like trust that they weren't going to go back to that that it wasn't that like you know we're just going to buy them to kill off the competitor and not do anything with it kind of scenario. [snorts] Yeah. Even net they have an image problem because a lot of theirs their image even though it's an open source you know framework and everything it's still tied to the older framework. So a lot of people don't realize that you know is open source runs on Linux runs anywhere you need it to and like it's just a whole different ballgame. So, but that let's talk about that because I mean I think net core has been a big deal. Like has it always been that way or was that I feel like net core was a more what in the last five 10 years or has it always been Yeah. Um No. Yeah. Last five years. That sounds about right. They started with standard. They had a whole weird naming thing that's been confusing and now it's standardized back to .NET 8. Okay. Standard then standard then core then now it's mainline. Right. Yep. So how did that change? did like so I mean so when you started was it kind of more of a monolithic architecture um or I guess because again you I guess there's two sides of this too there's the whole running all your jobs to aggregate the data right and that's kind of a separate service in a way and then you've got to host your web app right I mean can you talk about you maybe the way that was architected and then yeah I guess did like say the advent of containers and uh net core and some of that did that change how did you you know do some make some efforts to break that out just kind of curious on that yeah I think starting kind of like high level architecture where you all started and then going through the evolution and things you had to change and both pros and cons things. [snorts] Yeah. So the original uh automation system was written in net framework. I think it even started on like 35 or something like that which is right when link was brought around which is kind of a big thing for fornet but uh yeah so it was written in three net framework full and it used app domain so it would dynamically load sub assemblies and uh basically run these like John mentioned earlier each of our sources as we call them gets their own codebase and uh they can be you know run independently and so we just basically had this thing that we call the zombie that goes out and gets these things runs these jobs and then delivers the uh the uh results back and then processes them and you know so it's kind of like you could spin up multiple instances of it but it did all run it was it was kind of a mess and so now that whole system has moved to core and we're even still making changes on it now it's definitely more um it's more broken out and each source is we're moving it to its own micros service so that they can just process the jobs as they go they don't have to dynamically load these assemblies and all that kind of thing which is somewhat wasteful And we can horizontally scale like say Texas, you know, when we get their production, we need to run just a ton of jobs. And today it's kind of we can scale the the zombies all horizontally together, but we can't scale Texas individually, for example. And where we're heading will make it so that we can scale Texas as resources independently from say like Montana that hasn't updated. But it's crazy though because the first iteration of our zombie was a windforms and you would hop on the machine and you would spawn up 15 instances of it and um you know you could direct it to pull jobs from one source or another but still you would open up the remote in remote end to one of our you know just workstations that we had and there would just be like this cascade of windows there running all the jobs and um when you think about that like how that used to look and then now they're just containers and Kubernetes and we just go change it to say we want 50 instances of it and then they all just you know spin up in Kubernetes and we have them all run in our uh in our cluster and um so yeah the it's a pretty massive difference when you think about that um codewise it's not I mean there was a lot that went into it but it wasn't that bad um no changing app domain stuff around to get it into and and you know just kind of losing the libraries that weren't core compatible and then once you do that you're good to go in Ubuntu and then once you're in Ubuntu you can spin up containers and Kubernetes No problem. Yeah, today it works largely the same. It's it's communicating instead of loading app assemblies over app domains, it just downloads the package and creates a gRPC communication channel between the zombie and whatnot, runs its work and then delivers the work back, you know, the result back. So, it's similar but different. So, you know, there's one other side of this I want to I want to touch on though is that the database side is interesting because I want to get on that because the way we do this um Oh, yeah. Well, just a high level, the database, every state or every source gets their own database. But then the the kind of collapsing into the instance that runs the site used to be kind of batch job situation where we would kind of every night we would try to pull these things together and then push them over, run all of our rules and normalization, aliasing, and all that stuff on like a batch job. Now everything's using messaging and pushing. And so like today, if you go to like North Dakota where the NDIC publishes out their permit file at like 5:00 every night or something like that, our jobs are are automated and scheduled to go look for that job and and look for that file. Once it finds the file, it processes, pushes the messages to the front that those go directly to our site. They're live on our site literally two to three minutes after they're made available from the uh NDIC. That's wild. And that's running through all of our rules and everything, you know. So that's right. and it reduced locking and everything going that way. It helped us make it a lot more scalable. Sweet. Yeah. I was gonna ask kind of just quickly about your journey andor experience to containers and then Kubernetes and then definitely also want to talk about the the database side as well because I think that's another one. And then at some point we're also going to talk about self-hosting your your hardware because that's yet another one. Well, I think this I think that that may play into the whole Kubernetes thing, too, because I think most people think of Kubernetes being on the cloud, but I mean, um, you know, the implications of that, you know, running on prem and everything. Yeah, let's let's focus. It's pretty funny, right? Because they call it cloud cloud, what do you call it? Cloud compatible, cloud ready, cloud whatever. All it means it could be cloudnative orchestrated and independently run like a micros service in Docker, you know, and um so anyway, yeah, it's funny because people do think Kubernetes, they just think that that's an I mean a Google thing or an AWS, you just spin up a Kubernetes, but like it's really just Docker orchestration and you know, setting it up yourself. Don't get me wrong, I'm not recommending it for the faint of heart. It's there's a lot that goes into that and it doesn't stop once you get your c your cluster spun up but uh it's I mean it's all very doable. Yeah, there there are a lot of knobs to tweak in Kubernetes and I don't think people really realize that and like even me as a non-developer I know nothing about architecture like when I was at hivesell you our push button deployment of it was great to deploy it but you could still go into I don't even remember the name of the management tool we were using uh you know you could go in the back end you guys using rancher it wasn't not for the the Kubernetes monitoring. Um, it was something else, but you go into it and you you could see your instance, you could see the the three clusters and you could but there is just like an infinite number of settings in this guey. And I'm like, I can only imagine what's underneath this with like the YAML and all the other config things that you can do. But it's uh yeah I mean it's for those people who you know are unfamiliar or whatever like I think a lot of people have either recently migrated to containers or are finally containerized whatever but then there's there's this like weird oh well you know Kubernetes is just like the orchestration of containers. It shouldn't be that big of a jump, but it's not easy and there's not a ton of people that are experts in Kubernetes across the world even as I found out when I was at Hivesell when we were trying to hire people. Um, and so like that's uh, you know, that's a big kind of it's not a risk, but it's a decision a lot of companies are looking at or have to make and stuff. And so I'm very curious on y'all's because it's it's one of those things that it's definitely not necessary for every application, right? I'm Bobby and I are in the same camp that I don't know that Kubernetes is the best option for edge as it is today and things like that, right? But like just give people your your kind of experience, thoughts, feedback on on that considerations [snorts] to think about. Yeah. Oh well, like John said, it's not for the faint of heart. And we made the jump all at once, right? So we were running we went from basically that Windforms application or a slight version of it into containers and running everything in Kubernetes. So it's kind of like let's just go all in. He didn't tow dip. It wasn't like oh we're going to try some containers. It was like no we're going to do containers with Kubernetes and just do it. [laughter] I think that was basically the move. And so u it was it was painful at first. I'm not going to lie. Um you know and we didn't really know what we didn't know. It's one of those things that we got in we used rancher to get started that made it easier, you know, because it kind of I don't know if you ever seen that there's Kubernetes the what was it? Kubernetes the hard way and you go read through that about setting it all up manually and you're like wow no thanks. [laughter] Not for not for two people. This is absurd. Yeah. And so yeah, we went into it with Rancher. It made it easing into that part easier. And you know we just kind of started using pieces here for like the uh for like the backend systems you know deploying the zombies there uh and letting it orchestrate as we added nodes in. Same thing with the site. The site is mostly like the main portal um is ma basically a monolith but we we kind of branch out to services as needed like we have an export system there. We have like a print system. Basically, anything that would slow down the main site, you know, like a hot path, we'll put it over there so we can get it. But all of those cluster and I do think we've we've done a good job of of being open-minded uh trying to always use the best tool for the job. And so what you said is is a good point. There are things that work well in Kubernetes and things that don't. And we've tested a lot of things and we've actually come to the conclusion that by and large stateful sets don't really do it's just it's obvious that the the K8s or anyway that was it was tacked on later. It's not well thought out. Um it works when it works and then it doesn't. and between the networking stuff that we had to learn that goes on inside of Kubernetes and then um and then yeah the the kind of run-ins with stateful sets um you know we've we've definitely come to conclude that there's aspects of it you know I mean the thing was built to run in memory that's what you have to keep in mind if your stuff can run in memory then then it probably goes well there if you need to write to disk then it's something to consider um small discs probably work fine large discs are a bit of a nightmare. And so at this point in time, we actually do run all of our databases outside of the cluster. Um, and then there's a few stateful things inside the cluster, but they are ones that that that are only stateful for the uh uh to to to hold on on reboot like reddest stuff that we can re reload and things, but they live in it just uses a pure cache in that case. And then even like our printing system, you know, like the job gets lost. I don't care. It has an empty disc. it gets mounted with, you know, 100 gigs of space and it gets wiped under. I don't really care about that. Damn. Yeah. Yeah. But I would say that anyone who's who's diving into there is a right tool for the job and and you don't just, you know, blanket, you know, lay a blanket statement across any of these, um, you really have to kind of look and see what the tools were meant to do. And if you're doing something it wasn't meant to do, maybe you should look around and see if there's another tool that is indeed meant to do what you're looking for. Yeah. No, that was that was the interesting thing conceptually about Kubernetes at the edge, right? Is it's like, oh, robust resiliency like failover [snorts] makes a lot of sense if you're trying to automate like an expensive or risky dangerous application, right? Like you're not going to be able to do that unless all all the code is at the source without no, you know, without having to connect to the internet. The flip side of that is Kubernetes was designed for cloud [laughter] architecture or you know just mainframes and just a steady single location single network computer setup and guess what when you are on the edge especially in the oil field and you're moving networks every two to three weeks it's not very [snorts and laughter] good Kubernetes is not like being set up on one network and then being moved to another Yeah, it's a No, no, no. The cluster itself can jump between networks, but don't move the nodes of the cluster. Yeah. Anyway, no, it's a stuff. No, there is though. I mean, again, it's I I saw something on on the LinkedIn which I question its values. Everyone knows, but uh the uh you know, people talking about, you know, not needing to, you know, make some big solution. You know, I get what they were coming getting at. I forget who it was and Josh you might remember more but it's just like people talking about use Excel if it's if it gets the job done and like okay there's a tool for the job been saying like if if you want to use Excel to get a proof of concept up cool do it uh but like we talked about earlier each of those steps is a debt that you're going to have to deal with on the technical side so you know just keep track of them and if you just keep working in Excel and keep racking up that debt then you are either building a house of cards or you you're going to hope to sell before it crashes on you. That's all I can waste a lot of time. I mean, same with Spotfire. Again, it's a great way to get started. Go fast, but I mean, like, at some point, you run into limitations or, you know, just again, you let enough citizen developers develop in it and it's way too slow or, you know, it's taking a lot longer than it took when they had less data two years ago. And um, yeah, just is what it is. It's just it's definitely a diff like I mean there's so many things graphana, Spotify, PowerBI like there's all those things that are just like great tools to help Zapier like all those little things that can really speed up just like you know okay instead of spending a day coding this and then maybe six months in the future I've got to spend another day recoding it because they changed their off you know uh version or whatever. like, hey, it literally takes you 10 minutes to set it up and then as long as you document it, like you survive to live to see another day and work on the more important stuff and hopefully you have enough money at some point, whether it's through funding or revenue that, hey, we can now hire a bunch of people to start taking care of this stuff. It's it's a very happy medium, very fine line because again, like it's I guess it's stuff same with Excel or Spotfire or some of these like but you know, they end up in production, right? Um they're being used to make decisions and so on. But I mean I personally from my lens now more as like a data architect, data engineer look at it more of like these are great prototyping tools. Yeah. Like I mean like if if I can give someone a live wireframe like this is what I want it to do but I needed you know to be done in a web app or something you know more robust then yeah. No that's I tell everybody that I talk to that is like working on an app idea or in the middle of development or in product like if you're talking to developers and they don't have like a deep understanding of exactly what you're doing or trying to solve, build a mock build some kind of mockup demo. Uh, I I used uh it's called TL Draw. So, it's like TLDDR, but TL Draw. It's a a new AI thing the other day to try and show our developers what I wanted our menus to look like on this next iteration of Collide. And it took maybe 10 minutes and a few iterations of me just drawing on this thing and adding text to it. And I had a live demoable thing of exactly how I wanted our text fields to work that I could it was a live website. I shared it with them via a link. Like it was that is awesome. like it saves so much time. So, let's chat about the database. No, I agree. You see you No, I'm sorry. Sorry. I know this thing developers and it makes it harder. I know so many of these people that don't even start with the developer side, you know, they they build out these things and and then like they get to developers later and once they do, they never had it documented properly and now all of a sudden they're trying to ex I mean and if you guys talk to developers like Josh and I are developers so does we we can say what we want about them like developers are challenging to get your point across too cuz like they are just problem solvers. They they're you know everything's a nail kind of situation. They're hammers. So, like you can't you can't just throw things at them. You have to line them out really nicely. And if you didn't document along the way, then uh having developers, you know, jump on what you said about, you know, doing it to send to developers, that's perfect. But if you don't have developers, then that's where you get in that black hole and you'll get developers one day. But yeah. Yeah. No, that's uh yeah, it's if you've ever if you've ever given a a a developer a uh you know a a backlog of a checklist like that's that's how I think about it. Anytime I I build a backlog or features or anything like that, it's like they are thinking of this as a checklist of [ __ ] that they need to do and the like as long as it meets the requirement that you put here that is checked off the box. So you need to make sure you are [laughter] incredibly accurate and detailed with that requirement. And another way to do that is just to use a mockup. Like so much easier to use a mockup or [clears throat] a template or something pictures worth a thousand words, right? Exactly. But even like that reminds me detailed spec. Like John and I were working on the web app for RDS and we had a developer and again we were capturing time series data like timestamped from the uh um from the from the sensor but the developer used the inserted timestamp from the database. So like everything was stacked on the same time stamps whenever it was actually like and we're like no we need the one from the sensor like you know but like again things you just assume that they knew but you don't realize how explicit you have to be sometimes. That's true. the the the phrase assumptions are deadly is never truer than in software [laughter] development. Never assume and then that was you know however much money in developer time that you paid for you know that then have to be you know doubling up to rework it. Yeah. In the future. Yeah. Um so yeah let's jump in on the on the database side because like you know I think we're getting downstream. You got these jobs whatever but you're you're dealing with plenty of rectangular data but you're also dealing with a lot of like unstructured data too. So curious how you maybe stage that. are those in different types of databases or layers and then does it all end up I mean you said Microsoft shop so I'm assuming probably SQL Server um but just kind of curious about that side of the stack.
Yeah. So yeah Josh sure yeah yeah so we started obviously with SQL server we're currently migrating um to Postgress kind of a long thing obviously we have stuff in snowflake as well but again with longstanding cost we haven't wanted to go all in on Postgress but um so yeah like our unstructured stuff ends up in Postgress today and it ends up so it basically we get files in and they'll go into a queue to get processed through a variety of different ways to um to pull out all that data and then we have the jobs go in and try catalog the data that was in the files. Right. Is that like dropping it into like a JSON blob kind of thing and then you know Right. Yeah. Yeah. Exactly. So we JSON blob and we'll go through and we'll have jobs that know how to kind of look for the data and the different file types and that kind we have you know a lot of these files that we get from the state like was it the uh Texas file for example you get it's like 200 pages with no bookmarks and that kind of stuff. [snorts] So you know we have to start way out of you know classifying these documents as we break them apart. So we have to start at real high level and then get really specific as as we can. Yeah, I was going to ask like right and that's the thing about No, go ahead. Sorry. [laughter] The unstruct no the unstructured data is an interesting one though because um because you know there are standardized forms and standardized forms are nice um and you know there's a lot of things uh as time has moved on uh for dynamically recognizing key value pairs out of structured forms for breaking out tables um but even those we then have to kind of cross reference against our data model to try to understand what this data is um you know of course there's ways the the revision numbers and document numbers in the corners of these things. But um at the end of the day like there I think we had an idea once that we could potentially make some generic uh readers that could handle you know once we classify the types of documents to what they were that we would use generic readers to extract the data. Um and then actually kind of even in the classification understand what the documents document is by cross referencing our data model to understand which properties are being addressed in that document then dynamically pulling out the key value pairs and then writing them automated into each of the data pieces. um which is not I mean it has worked somewhat um like North Dakota because is the opposite of what Josh said uh has a massive PDF but has lots of bookmarks in it and pulling out a form six out of the NDIC file is not hard and in that file um there's a handful of things that are really helpful things like frack stages and pressures and things like that um but yeah structurally wise it's it is still challenging because no matter what we talk about storage um you know What do we have Josh? 60 80 million documents in our system. Each of which could be 10 20,000 pages. It's all over the place. And convert those to images. The cost processing power. Yeah. Yeah. They're insane. It's I mean, if you want to use an AWS text, you're and you want to do the key value pair and table extractions, you're going to talk about six cents a page. Um until you hit scale. And even at that, then that scale means like you're spending a million dollars a year on that. And again, yeah, just the crazy thing, Ghost Script and some of these [laughter] other Python libraries. I was actually looking at that earlier this week. Yeah. Well, and there's and then so many of these things um you know, we we do have individual parsers that address specific types of documents just because they are I mean even that NDIC permit file I talked about earlier, it's it has this like crazy variable number of comma delimited fields and you have to basically inspect each one of them and try to determine which one it is and they'll just change it on you for fun. They'll add in bottom hole, section, township, arrange and the eighth space for no good reason and and not even really tell you what it is. And so you just have to like write I mean I think there's got to be a hundred different test cases for the file type in our system. Um yeah, was that the one like that one with the activity or there was I was trying to parse some activity. Yeah, I was Yeah, I could I felt your pain on that. [laughter] Yeah, it's a daily activity one that has the permit part at the top is cool. Yeah. Um but then it goes through all these other different activity types and um but the permit part at the top was the one I was talking about. We have 100 cases for all the other types. A lot of them we we do catalog what the movement was and then go back to the state page for that well and get the details from that. But yeah, um sorry we got off on a tangent there because the data files though that goes to Postgress and I think it is important that we talk about the um we've always used OMS in our code. I think because we always had the idea that we need to be flexible. SQL server was a great way to get started. Um but for horizontal scaling and cost we're going to be moving to Postgress. Um And so the fact that we're on, you know, EF Core means that we can we can make that change without an extreme amount of work. Um, and so I think that's important for flexibility. I mean, databases, especially the relational database side, um, you know, there's they're they're not all apples, I should say, but they're more like apples and pairs. they're similar enough that uh that you should you know when you architect your systems being able to be flexible on that backend database I think is an important thing part of the migration strategy is with those messages now you know before we were relying on replication directly in SQL to get this stuff done as part of batch jobs now these messages can write to both and so we can have this happening migration seamless to the customers right and then we just go all in on Postgress once we ensure that we have the performance and everything that we need to on the site. Can can you step back real quick and explain to people because I don't think we've talked about OMS yet on the podcast. Can like I mean for the people who may not understand, can you you know just give a real high level what is an OM and you know why is it why is it useful? You going down or me? Josh, that's you. No [laughter] you. Nice. Yeah. So you know obviously we store you have to store uh data in databases and so they have their specific data types that are in there and then when you get into languages you have you know different data types in the language and you have this mismatch between them and uh you can create an ORM lets you create a mapping basically from the database into your language of choice and it makes it seamless and most of them have built-in providers for different languages or for different database servers right so you can you don't actually write your SQL by hand you would write say I want this and it would go ahead and translate it and you know you're sending it to SQL server you're sending it to Postgress it'll translate that query to to your target database and you just get your data back and it doesn't really matter where doesn't care yeah doesn't care you know that's the ideal world you obviously have these providers for the different databases and inevitably have differences in how they generate the queries so you have a little bit of a leaky abstraction there but the idea is that it should be that Yeah. No, I've got two other questions. One, we haven't talked a lot about OCR with anybody yet. So, I want to kind of just get some some info from y'all there. And then two, I want to kind of roll that into probably our last kind of topic before we do our our uh speed round. But like how kind of once you answer that, how has kind of some of the new AI computer vision stuff, have y'all looked at any of that at all or like do you see that as kind of the next iteration of the the OCR, RPA type? So I say OCR, I really am talking about RPA and the grand scheme of things, but that's obviously a big, you know, buzzword uh consultant consultancy term that's been floating around these past few years. And I do think that, you know, like a lot of what y'all are doing, a lot of accounting departments, like there's so much opportunity for RPA where it's literally just removing human data entry from, you know, reading a form to putting it in another form. There's so much room for improvement on in the energy space for that. So, I'm curious kind of what your experience and stuff has been like with OCR. Um, and then kind of what do you see in the future of that space as well?
Yeah, I think we can both chime in on some of this. the the my my frustration with some of this is uh so like even Amazon Textract has had its machine learning capabilities built in and they work at scale and so building that model um works it it's it's good to do um the the open- source tools that we've played around with on OCR stuff um they have capabilities to be trained um but the scale in which you have to work on to train those models is sometimes challenging to deal with especially when you've got a data set like ours where we've got, you know, whatever it is 26 30 different sources of documents and none of them have more than, you know, that Texas would have um, you know, at least, you know, millions of the documents. But then these other states like North Dakota, 30,000 and you basically barely have enough to train a full model um to to do this part for you. I mean, the size of the data is is challenging. But if you use like an AWS to do it, they've got a lot of scale to work with, but they'll charge you for it. Um, and so like it's it's a little bit of a chicken and egg thing as far as there's great tools out there and it's and honestly they're getting better and better. Um, and even Open AI, Josh can talk to a little bit more on what we've done on the vision side there. Um but uh these big scale these these big companies that that have the scale have a lot of opportunity there but they also have a lot of expense there that it costs for people like us to work through. So there's a number of tools um but you know and and that's the other thing Josh has worked on some of the document identification stuff and some of these especially old documents right they just require so much massaging before you can actually run them through anything. Yeah. No, I mean that's that's got to be one of the biggest like my first roommate when I moved to Houston, he had just uh graduated. He was studying for the bar, but he was working as a land man. And I remember the first day he came back. He was like, "Yeah, you know," we were talking and I was like, "How'd it go?" You know, he's like, "Yeah, well, I went from whatever 18 whatever or 19 whatever to present." I was like, "That's a really random number." It was like 189 something, whatever. It's like, "Why? Why that?" He's like, "Oh, well, that's when they started using the typewriter." And I was like, "Oh." I was like, "So [laughter] you've got like 200 years of of title documents that are like script literally." And I'm like, [clears throat] "So land is just so screwed as far as like the digitizing part." Oh, yeah. I've got a guest that we need to bring on that. I just saw a platform last week. Was pretty sweet. Yeah. But yeah, I think that's just the big thing that you know, the average person that you know might be listening or whatever that isn't involved in kind of the regulatory space or the data space doesn't realize that every literally every single state has their own data sets, their own data types, their own data formats, their own structures, whether they have structured forms or unstructured forms, whether they consistently put the data in the same spot on that form or not. Multiple revisions of that same form [laughter] over the years forms changed slightly. Yeah. Yeah. And none of it was ever architected to be uh you know ingested by computers essentially on the back end, you know, like they all have nice little black boxes around them. You're like, "Yeah, just give me this region of it." And it's like, "No, it's usually you skewed real bad." And so yeah, when we started it, we started what with testact I think way back. And we thought we would just do the basics, you know, grab a file, throw it in there, and give us all this information back. And it turns out that's not at all the case. you give it a whole if you get a whole document scrambled gibberish paragraphs and it does give you text back but you have like no real recognition of what's going on how it lines up with the dog without actually looking at it you know yeah I mean we did some stuff with that and like but it basically just took the dump of the tester act and put it into like elastic search and at least made it searchable but like when you guys are trying to take specific pieces and put them into like a you know structured data set that people can use like it's a way different story well and No, that's it. And that's exactly what you see it working well for like solar and what elastic search ingestion. That's like what it does. And so you can find the documents in that case. But yeah, when we need like what was the depth or who's the operator on this filing or whatever, you know, it's like that's where John was talking about. We actually started playing with the newer chatgpt vision stuff where you know, we obviously had to build these kind of model and classification so we can identify the document and then go pull these areas out using test track, right? But like Chat GPT's new vision stuff, you can like give it a document and say, "Who's the operator on this thing?" And it gives you the answer back. You're like, "This is impressive." And you be like, "What's the official name of the of this company?" It gives you even extra information as long as you know it's a an older company, but it's super impressive. So stuff we started at least, you know, wouldn't say we could trust it 100% yet because it does still have some it'll grab random stuff, but you know, if we have these multiple data points, we can kind of get an idea of yeah, that that's right. Netflix will like the um chat GPT vision stuff will that give you like a um confidence rating because like I mean say with text I mess with it like it'll give you a confidence rating of a certain piece of information. Right. Right. Yeah. You can have it do that. Okay. Yeah. No, I think that's the thing. Well, that's the beauty about our systems and Yeah. We combat some of that too because we have a database of all the operators. We have a database of all the fields, all of the leases. is all I mean we have all this information for this existing data set and so it helps us to cross reference our existing data set um to help ensure that the data pieces line up nicely as expected. Um and then we have written um and this is unfortunate but we did write a uh kind of a key value parser out for standardized forms and it really is funny to do because you're like you know I'm looking for this keyword and then I'm looking you know left up beneath what's my next keyword what's my spacing to look for okay now I've got my box now give me the text out of that box now cross reference that box to my other data to make sure this looks correct and anyway it's it's so there's cool things and I mean we wrote that part um but you It's kind of thing that you'll get out of a a text track, you know, if you want to pay for it. So, that's always I mean, money comes into play in all this. Of course. For sure. Yeah. All all problems are solvable, right? If you've got a big enough budget, [laughter] the uh No, I think that's true. That's really the big thing that I've even just in starting to mess around with some of that stuff earlier this week. I've taken away is that like with RPA and OCR, if you've got a stand, if you've got an invoice and it's from the same company and it's the same structure and format and everything is always in the same places or in the same general areas, much easier to do than you've got 100 different data sources and none of them are the same and none of them some of them conform and will always be in the same way and others won't. Like it's a very different different story and of course it all boils back to the data which is yeah always well and you go look at existing solutions out there like you have Butler and a whole bunch of companies that are making these uh real nice OCR systems but they're more targeted at stuff like that like scanning in receipts not these big invoices and that kind of stuff and they even have templates driver's license all that kind of stuff and it's like okay well it's not going to work. Yeah. [clears throat] Yeah. Check is easy to read. Receipt's easy to read. [laughter] Even an invoice is almost I mean again line items, totals, sums. There's a few key words that I mean but when you were given like we talked about you know each state might have 15 different forms and 15 different revisions of those 15 forms and that's times 30 different. I mean it's a whole different then you get into like you're talking about title opinion. We were talking about with that company the other day and like we have some that are 500 to a,000 pages long. I mean like and hey and you talk about money like you you you have to spend enough money with open AI even for them to unlock that many pages you know but I mean it's it's not trivial [clears throat] um and then I mean I mentioned textract and like even AWS has part of the service where you can say if it's below this confidence interval we'll send it to some people in India and they'll look at it you know to do that last you know mile for you. Yeah. Yep. No, that's Well, and I'll say funny story about OpenAI. We fed some data into it um just back in December and I found it pretty comical. I was we were just testing out some things and um we started asking it questions about some data and some permitting documents that we sent um and it started returning back fields from one of our competitors and it just it said their name underscore lease or something like that. And I'm like, "Where did you get this?" And it's like, "Oh, sorry. We just pulled this." And like, "What do you mean, oh, sorry?" like where did this come from? We pulled it from just like we [laughter] Yeah, we just mixed up some data. Ignore this. And you're like, I can't just Anyway, sorry. That was pretty weird. Like it blew my mind. That's a that's an interesting conversation and my our even my personal and our thoughts as a company on what what that structure looks like in the future as far as like corporates and like how do you how do you provide people with general information from you know a GPT model but also allow them access to their private information within their own models at the same time so that the contextualized model can give you the detail and the highle general model can give you kind of the general stuff around it. Well, but think about these, you know, the the um you know, like Saudi or some of these countries where they have the federated like where you have to keep things within their bounds, but maybe they want to utilize that, but they you have to run everything within no data can leave, you know, the country. Y um no, it's it's going to be interesting to see path forward. I know we've said multiple times we need to pull together like a little panel discussion on like AI like just implications and all that. I mean, yeah. No, there's I mean [ __ ] I read a paper two or three days ago and it was uh there's a couple new methodologies where they're merging AI models. They're merging the answers of the AI models in this new way where it provides much more context, but also interestingly enough allows the models to essentially one of them to train the other one as far as it based off the responses. So if its response is better than your model's response, it'll learn that and continuously improve. And so it's like, golly, this is this sounds like Skynet. Yeah, this is going to get crazy really fast. And then I saw the video yesterday on reels of the I don't even remember the name of it. I'll put it in the show notes or something, but this new robot that they came out with and it's just like mechanical arms on a dolly and you train it physically how to do it. But it was unbelievable. Like it would go, it was making coffee with, you know, the cure egg pods. It was washing dishes. It was opening doors. It was doing all this stuff. And it's just like these arms that you put on. You do it yourself and then it learns it and then that's it. Wow. And I'm like that makes sense to me. Crazy. Well, it'll definitely fold closer [laughter] than me and my wife will be happy with. Oh god. Or just Well, I like stuff like that better than this. We've got this hyperfocus on these language models and and they're neat, don't get me wrong, but like they're very just like they kind of a to me they're being used um for more than they're intended for. And so it ends up in some weird state. So the idea of learning that kind of activity is actually way more interesting to me than these another long, you know, large language model chatbot goofiness. Well, I've always thought that that was fascinating, right? Like all of us grew up believing that robots would be butlers and like driving us places and doing all of these like low-level jobs. And in reality, the Jetson, right? In reality, it's now replacing, you know, creative jobs [laughter] or like what, you know, it's it's this weird dichotomy of like where are the robot janitors? Those don't exist, but now we have robot artists and, you know, musicians and all this crazy stuff. [laughter] Um, well, we're Yeah, it's odd, but I mean, it's true. It's easier to learn. Yeah. Y we're out of time, so we got to jump into this speed round because we got another one coming up. Yeah. I'm gonna I'm gonna ask you guys a gaming question and I'll try and keep it in a general time frame. Um, Sonic, NBA Jam, Golden Eye, or
Zelda. Which one? Pick pick first. Pick Pick one. Golden Eye.
See, cuz I love Tears of the Kingdom and Breath of the Wild, but [laughter] I almost I'm gonna go with Golden Eye, too. That first person like I played Doom and that but like [snorts] Golden Eye was awesome. So, yeah, Golden Eye wins. It's the only game I've ever been in a competition with. I Golden Eye is so It's so hard to play now. A mall gaming competition I won Mist. That's awesome. Another game. Yeah, I remember Mist. [laughter]
Uh, who's your favorite social media follow? Oh, geez.
It doesn't have to be energy related or anything like that. It could be tech content. Yeah, whatever. Blog, news, could be Snoop Dog. I don't care. [laughter] John's [snorts] off the grid in Montana. So, that uh Yeah, I don't This is where it shows my age a little bit. We were just complaining. I would say yesterday. Yeah, I'm very torn on it because it's a very good source of interesting information if you follow the right people, but then it's a lot of lot a source of a lot of just dumb [ __ ] as well. [laughter] Yeah, I follow a lot of just like business people on Twitter and I think those have been the most valuable to me like even not technical business even, you know, you have like was it Nick Huber startup and and so you have a few guys like that that are have a lot of great content they put out there. Gersley Yeah, Gerley is great too. Every time they do though, they they all think that they're smarter than everybody else and they they get so arrogant sometimes it's hard to like to to hear. I appreciate though. Yeah, that's right. There's a line. I I also having now kind of been in that on the media side of it. I also am much more confident in my in like people that have big online presences. I do feel like a lot of them got there because they know how to kind of like play that game. And there is a certain online persona that got them there. And that person as a person is probably not the same as that persona at all. But the persona is what the internet wants. So they keep feeding it. Did you see Nick Cuber? Like he's got a media team. He spent $750,000 last year on his media presence. I mean, yeah. Wow. That's impressive. Or or ridiculous. I don't know. Let's do one more. Uh favorite database.
That's a hard one. Say, do we have to have a favorite? My SQL. My SQL is so you can [laughter] put it anywhere. It's so nice. I'd say SQL light for that reason. Yeah, you can put it anywhere. That's true. When you bring up SQLite, I got to bring up. Have you guys messed with duct DB at all? I haven't. Bobby is still filling. [laughter] Come on. Someone's got to It's the SQL light for uh OLAP. Obby just really wants someone to use it so that he knows whether he needs to start playing around with it or not. Or have you started messing with it? I've messed with it here because you can you can make it right. No, it's supposed to be super fast. I mean, you can run it anywhere in the browser, you know, computer cloud, whatever. But it's a check it out if you haven't. Yeah. No, it will for sure. No, it will. No, I will say Postgress though. I I just like Postgress and and how the community has rallied around it with all of the different plugins and everything that exists. The hor the scalability. It's just it it's to me SQL server is the Windows and uh Postgress is the abundu you know SQL server does a lot of nice things for you out of the box. It makes things easy to get going and it and but like once you start scaling it gets in your way left and right and Postgress is the opposite. It does very little for you out of the box but then once you get it optimized then the sky's is the limit. Yeah. Yeah. I mean, what one of my uh coding boot camp instructors told me and it's kind of stuck with me stuck with me is like if you can't do it in Postgress, you probably shouldn't be doing it. Yeah, [laughter] it's pretty accurate. You know, it's a good uh exactly good tidbit to to end on. Guys, where can people find you, find the company, [snorts] find you on LinkedIn? Well, maybe not. Well, yeah, we're everywhere, but welldatabase.com is is what you want. We obviously showed our social media prowess here, but uh anyway, the we are on the LinkedIn and the like and um but yeah uh welldatabase.com and you can uh find all of our information and hit us up directly. Ask any questions you want. Yeah, they've got a great I'm on Twitter too more than anywhere else. What's your handle, Jeff? What's Twitter? I thought it was X. Sorry, X. Yeah, you're right. See, my age there. Uh Jholt456. I just realized that's going to be another thing that like our kids in like 10 years are What are you talking about? What is Twitter? Thanks. [laughter] Twitter.com still works. Yeah. Yeah. I mean, everyone's call it Twitter for a long time. It does. Um Oh, yeah. Hey, for what it's worth, my kids won't call it X either. They keep calling it Twitter. Yeah. Yeah. No, go check it out. You guys have I mean free trials and free like features and stuff out of the box. Yeah. Go play around with it. It's a great tool. Um thank you guys. Absolutely. Uh yeah, we we are a real SAS. You can sign up and you can actually get access without giving your freaking blood type or something for free, you know. Yeah, how about that? And it's not one of the uh you get a 7-day trial as long as you put in your credit card so that I can automatically charge you when you forget to cancel your trial scenarios, too. It's the they're great, guys. So, I appreciate you guys coming on. Thanks for having us. Hey, not a problem. It's pleasure.