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. Um, we're here another episode of Energy Bites. First one of the uh college football season. So, I'm rocking my my razor back shirt or hat. Go Buffs in the big resurgence last week. So, love that. That made me so He's How do you not want to run through a brick wall after after hearing him talk, right? It's it's pretty awesome. But here today with with Bobby Nean, my my wonderful co-host and and David Thuel, co-founder or founder of Guma. Yeah. Yeah, absolutely. Thanks for having me. Appreciate it. Appreciate you coming. Thanks for being here. So, we'll kick it off just like we kind of do with most people, but kind of tell us how you got into both energy and tech. Uh whether you know starting out, you know. Yeah. Were you a kid tinkering with stuff? Yeah. Were you tearing [ __ ] apart like I was and breaking computers because you were uh downloading viruses or how did you get into it? No. Uh I have a really different background. Um I listened to Talal's episode too. So like I'm the opposite of Talal. Um so I'm a geologist and geochemist by background. Um I ended up as a geologist just like every geologist by failing out of engineering school. Um and and I ended up kind of in this consulting and big data space. Uh, in a way I'm kind of like Brian McDall who was on maybe one of the last episodes and we have a lot of overlap in our history. Um, so I was always into like doing things, building things, but I was never a computer guy. Um, I actually I I talk a lot of trash on my dad. Um, because my dad and my grandpa both worked at Cray Research, Cray Supercomputers. Um, but like on the finance side, like they but you know, they didn't clue me in to like the fact that computers were a big deal. Oh yeah. And and so we actually, you know, we we we had very modest means growing up, but we had one of the only computers on the block. My my dad got an Apple because he was a numbers guy and doing accounting and things like that. So he had it. So played math blasters on the math, you know, stuff like that, right? Math blaster. Oh man. But I'm I'm I'm not a gamer. I'm not a And now I'm like a tech nerd, I guess. Like I'm into hardware. I like I like the advancement of it, but I didn't didn't grow up with it. Not a computer guy by background at all. I don't play video games. Yeah, I wasn't really either. I mean, yeah, I didn't really get into it after till after teaching coaching. Like once I got into oil and gas like, "Oh, this is cool." And really had no desire before that. Yeah. I think that's the coolest part about it, right? Is like there isn't a singular path that everybody kind of follows to end up into into that sector. There's just so many different I mean I'm a mechanical engineer because I hate theoretical intangible things and and here we are talking about data and technology which most of it is intangible right yeah totally but I mean you talk about big data I mean geology in a way was one of the original kind of big data you got all the seismic and stuff I mean oil and gas has been handling some really big data sets for a while even before it became kind of cool thing to do let's unpack that because I think a lot of people don't especially just the average energy person doesn't really understand how much data there is on the geo side. Yeah, it's it's fascinating. I think there's a lot of there's like three or four things I want to cover and and so one of them is that, you know, oil and gas was super early on supercomputing technology and things like that. You know, they had one of the biggest and best, you know, systems at Exxon for supercomputing and they still do. They partner with HPE for that. It's crazy impressive. Um, you know, but then there was kind of this like winter of advancement. I mean, and we we weren't really quite sure where to go and and I think as you know, a lot of things hadn't caught up yet. So, like seismic processing was really high-end, but we had no way to deal with manual records and paper records and and things stored in the salt mine in Kansas that are the permanent record of history, right? And and so like we kind of went through this winter of like how do you possibly ever catch up? And that's kind of part of what Brian tries to deal with and um from Sabana. And it's like, you know, yeah, he he has one approach to it. I happen to disagree with his approach, but I love that it works for him and we have a great partnership on things that we do. But like, you know, we're trying to really push those models to change that aspect of it and really catch our business up. Um, so I think that's interesting. Uh, I also like in oil and gas technology that like Exxon made their own typewriters, okay, and their own and their own computers. And, uh, actually like as an artifact of that, their fonts are different than everybody else's. And that's a fascinating problem. And like that's how diverse like the technology approach in oil and gas is that it's everything from like we're going to make a typewriter. Yeah. Like this is like true builder. Yeah. Exactly. Exactly. This is fascinating. And you know to this day there's there's not a lot of like Exxon Mobile holdovers that remember that they used to make their own typewriter. But I'i've actually run into a few and you know they're like yeah well it was the only one we could bring in country so that's what we used. Fascinating. Yeah. No, I mean very typical, you know, super major style, right? Like, hey, come show me how this thing works and then we'll try and back engineer it ourselves and do it on our own, right? And yeah, I think I think you know the data problem in in subsurface is if you're if you're oil and gas adjacent or if you just maybe know someone that does it like almost everyone's familiar with seismic data, right? Sound waves through the earth, they reflect off the rock, they give a picture of the subsurface, like everyone kind of gets it. But then there's well logs, and there's dozens and dozens of different types and iterations of tools, and there's the cores we pull out of the ground, and there's the description of those cores that we do as geologists, and there's taking plugs out of them and chips off of them and the pulverized rock from the ground and analyzing all of that, and there's outcrops, and there's satellite imagery. and then there's all of the things we make as geologists. We make a lot of visual collateral. We make graphs and charts and maps and thin section pictures and annotate this and business graphs, things that show P&Ls and like there's just so much visual information that that that goes in and out of companies. I mean, even the most modest oil company has a 40 or 50 terabyte server. Yeah. Like that's like, you know, we got three guys in a basement and they have 40 terabytes, right? And then, you know, super majors have hundreds and hundreds and hundreds of pabytes. Sometimes they have 100 pabyte scale single business units. Yeah. With the history they have and everything. I mean, you know, perian what goes back 100 plus years now. Yeah. And and that doesn't even include all the stuff that hasn't been digitized yet. That all has value, right? Um, so it's a it's a wild data space. And I think it's interesting because I like I overlap with like Silicon Valley tech guys a lot because I talk to VCs and you know we try to explain what we're doing and they don't they don't quite get it. And and the reason is because if I'm in the city of Houston or the city of Austin, I'll get more data from cell phone location pings Yeah. in the course of an hour, then I'll have from the entire Powder River basin triple combo logs on every average 8,000 foot well across 50,000 well sites. So So it's like if I didn't capture the data today or even this hour, I'll catch it next hour, right? It's no big deal. But but like at oil and gas, if I drilled the well and I didn't analyze it, capture it, catalog it, or preserve it, not drilling that spot again. Yep. Yeah. like what's you know what's the closest you twin a well maybe you know maybe a few hundred feet on really rare occasions but usually it's it's quarter mile half mile and a lot changes in the world yeah you know in those sorts of distances so I think it's a it's a challenge that that is interesting from the from the subsurface side For sure, and I think that's one of the big things too that, like, I always after working at at RDS is doing the gauges and well tests and stuff like that you know, there are so many people you talk to that are like, "Man, I wish we would have done this on the well or whatever." Uh, well test, PTA, a certain type of log, etc. And you can't you can't go back and do it in most situations. And so it's or it's incredibly it's way more expensive to go back after the fact, run like a production log than it would be to do something on the front end. And so that's always a struggle I feel like in our industry, right, is you're it is a business so businesses have to make money, but you can make a lot better decisions if you have more information than if you don't typically especially if they're cheap ways of doing it, too. Correct. Yeah. They don't have to be the sexiest, latest, you know, interpreted log, you know, or the synthetic log stuff that they've gotten into and all that stuff. It just have good data across most of your Yeah. It's only version rock that time, you know. Yeah. And consistent data, right? like you know don't change your gathers or your collections every every two wells like it doesn't create a comprehensive data set that you can actually analyze in any sort of way. But to that end though too you get on the other side of like people capture a lot of data and they don't use it at all for sure you know or goes somewhere to die or the analysis ends up in a PowerPoint and no one's using it. Yeah. Yeah. There's a there's a big problem um in our business on data utilization. And I was telling my co-founder Cali this yesterday and I said it's because we went from the notepad generation to the iPad generation and we skipped the ThinkPad generation. No, you know, Lenovo, if you want to sponsor me, hit me up um for that plug. But like I I love I love me some Lenovo. But I I think I think like there's a very real like like gap in data literacy and utilization literacy and and part of it is the business process and speed have changed a lot right so you know one of my other co-founders Steve uh he's been in the business for 40 years he's been boots on the ground in 75 countries and you know over the course of his career it went from I make a list of things I want on a map and I send it to the drafting department and Then I get to sit around and think about it for a week while it gets drafted. Yeah. To, you know, I'm making things on the fly from the data room on the airplane away from the bid round. Right. Right. And and and it's like that has changed the data velocity a lot even in the subservice world where you know you think oh well you know we'll drill a well we'll collect some data we can interpret it and it's like no the the velocity of that is is you know daily weekly sort of things now. and and that's that's a that's a burden. Yeah. But even the fact that you can process through those data sets, you know, to to do that, you know, within a week or I mean because it used to be those supercomputers at the time could run analysis in a week or something like Yeah. Now it's like now you can spin up a computer on AWS and do it in Yeah, it's awesome. Yeah, pretty awesome. Yeah, I'd say that's one of the big like I don't know subtle kind of advancements over the last decade plus is just the not only the amount of compute power that now exists and the accessibility to it, but also even just at the software level, right? M I worked for a company that did basin modeling and reservoir simulation and stuff and their big kind of differentiator from Eclipse was they could run on as many CPUs as you wanted it to and Eclipse was limited to however many it was. This was a while ago, but um you know, and it's like well going from four or eight CPUs, whatever the limit was to 20, 30, 40 cuts down that model time from weeks to days or hours. And like that's a huge deal, you know, especially in in the oil field. Now, I feel like that's another kind of thing that's compounded it is everything is so much more operational focused and we all realize very much so that time is money. If you don't get your [ __ ] done in a quick time, you're the decision is going to be made whether you have the information or not. They'd prefer that you had as much information as you could, but it's going to have to get made because you're holding up production and operations, right? It's a it's a tricky it's a tricky game for sure. Yeah. Let's talk Let's talk a little bit about kind of how you got into Geo Luminina. Sure. Yeah. So, um, like I said, I I come from kind of a weird background. So I started the business as a consultant doing petroleum energy chemistry and basin modeling work for a consulting shop out of Denver at the time. Principal client was was EOG and then it kind of expanded out to kind of everyone doing unconventional stuff. So we were were early in a lot of North American plays. We did a lot of delineation work for these companies. It was super interesting time 2008 9 10 11 12 and things were going 100 miles an hour. It was like every time we turned around there was a new discovery and it was crazy. It was crazy. And so, you know, I did a bunch of that, and then I ended up at a research group at the University of Utah, um, where I ran the petroleum systems and geochemistry research group. did, uh, you know, exploration and production sort of research for super majors, IOC's, you know, 45 different companies around the world on different projects. I worked 50 basin in 15 countries in four years on a $7 million research program. It was it was massive. And that was where I really kind of got going and the sort of automation and work orchestration side. So my co-founder Cali uh when I worked there, she worked in my organic geochemistry lab and basically orchestrated everything from sample tracking to instrument analysis to outbound data that came to me for me to interpret. Um because we couldn't really solve the interpretation problem. It was like we couldn't we couldn't make the computer do it. The data sets weren't big enough. The technology wasn't advanced enough. And so it was about as much as you could do to automate a life is is what we did because I was on the road 60 70%of the year in countries all over the world. Um after that I jumped to Austin to join a a private equitybacked company. Um did exploration in the Rockies and then I founded a private equity backed exploration company. We raised $75 million, went out at least 100,000 acres, drilled some wells. you know, it was a it was a really exciting time and things were were interesting there. Um, and then, you know, we kind of started working on projects on the technology side. We just weren't really sure what to do next. And so, uh, started building things and hacking and I learned more programming and started prototyping stuff and we landed on this like let's use machine learning to replicate geological insight because uh, so my co-founder Cali, she made a bird recognition app for her photography. Okay. And uh it just blew my mind because I I still remember she showed me this like California quail that's in this juniper bush. And it is super hard to see a quail even even as a human. But the computer vision model had no doubt. And I was like my god like that's just what geologists do. We look at a lot of stuff and make and so we kind of thought oh this computer vision is like the way to go. Yeah. And so we started building product and we landed in a um an incubator that was funded by Shell called Six Lab. It's in Austin. And uh so we were in the first cohort of that uh really set to launch at South by Southwest 2020 and that's in March. Oops. And everything fell apart. But you know we got to launch product and start showing it to Shell and we got really interesting feedback. It was like one we don't know what to feed these really cool apps. like where do we find this data like and it's like well your core photos you know where those are and they're like no kind of but you know basement somewhere well I mean you know they're in the they're in the server but like different people had different knowledge of where things sat and internal search wasn't great and then you know they also kind of said well this is cool also do I really have to learn another tool I don't really want to can we do something else and I said okay like automate it right so so that's what we kind of transitioned our product to is this automated level of interpretation that does different geoscientific and engineering things. Um, from there we caught the attention of Techstar's uh, accelerator program. They have an energy specific one that's funded by Ecuador, Kongsburg, and Cap Gemini. So, we did that in 2021 and that was like right as we were transitioning our product. And, uh, out of that from our demo day pitch, we landed uh, PC with Ecuador and that went really well. Um, you know, they're kind of loving that product as uh, it's called Envision. So it's a it's an autonomous system of agents that does interpretation and search and and lots of really powerful stuff for users. Um and so yeah, I mean they're they're a great company for that. I mean like that you know they've had their issues you know obviously onore you know this and that but like as far as just like data and you know data open source and access like they're they're a really great company from that standpoint. Absolutely. Yeah. They're really pushing the technology and they do a great job of maintaining open source packages and Python and and stuff like that. And I think they have a really tech forward standpoint. Um, that's really good for the industry because they they can kind of show that, hey, this is this is value accreative and we know it is and it's it costs money. Absolutely. But but I definitely, you know, they definitely get their money back many many fold over, right? And and that so that makes it really compelling and and easy to deal with them. Yeah. Absolutely. Yeah. So, you know, maybe stepping back a little bit. So, sound like Cali did a lot of, you know, programming, but then you end up getting in on that side, too. And like, what was your first kind of foray or dipping your toes into that? I've been doing terrible programming since like 2011. Yeah. Um, just like the absolute bare minimum, you know, stuff in Excel, you know, whatever I could do with a tiny Python script, lots of SQL, lots of server stuff. So, I did a lot of database and and that sort of maintenance. I didn't really get deep into Python until maybe 2017 2018. I really like sat down and started learning. Um I wasn't great at learning from a book. And so like prior to the proliferation of like web videos and and guided kind of course content. I didn't do a great job learning it. Um and it's hard especially when you know a little bit. Like that's where I struggle with it. Right. But as far as like py python though, like the mid2010s was an awful time to learn Python. I mean like I got in because I was getting into it and I was, you know, I even when I got in, they were like, "Oh, you're going to be the Spotfire guy." And I was like, "All right, well, this is enough." But I always kept pushing that. I don't need to learn R to code or whatever, but like I started trying to learn Python. It's like because that was in the time when they were switching from Python 2 to Python 3. So like and then there was Anaconda and PIP and I was, you know, when you don't I look back now like I I get it, but like when you're new and learning, you're just overwhelmed. need to learn the syntax. And it was like, you know, I'd buy a Udemy course and it was 2.7. Another one was Anaconda 3 point whatever. And I was like, I already installed this package. Why isn't it not showing up? You know, like and it just didn't know how to where to start. Yeah. It was it was a terrible time for it. And I luckily I just kind of got to ignore it long enough cuz I was more geo and less data. Um, Callie, bless her heart, she learned it hardcore. Then she was part of a National Science Foundation funded like research organization doing biology stuff. Nice. Um and so she got into it by just managing like huge swaths of Yeah. uh nationwide incoming data streams and so like that's how she really got the power. And so she was like my surrogate coder. I always joke that together like we're one really competent scientist, you know. Um but but I mean I think to your point maybe back originally was like you can't really learn pro maybe you could but learn programming from a book like you've got to try to solve problems that are relevant to you and then then it it comes together. Yeah. Absolutely. Like even just having a little toy data set that you get to play with that I think like really makes that possible. And actually that's a really weird hard part in geology. There there are kind of not great toy data sets that that people can play with and and like they're getting better. And I know like Software Underground, they have like a a data wiki that's getting better and like you can find stuff there and you can always go download your own thing. But that's a whole, like, data engineering problem in and of itself. is like, if you don't know what to do with it to put it in a format to start working with it, then you're you're also dead in the water, or you get it and then...I mean, I've done this many times. you get your data set and then you start going through it and you're like, "Well, why is this throwing an error?" And then you get into it and you're like, "Oh, well [ __ ] there's data missing here or like, you know, you've got that whole ETL, the transformation piece before you can start analyzing it that you then have to learn and figure out to to do it. I wanted to ask you how how did you go about really kind of learning it and what would your kind of advice be to someone trying to learn it today from the more, you know, science geo engineering background side of things?" Yeah. Um, if the goal is to learn Python, I I really love free codecamp.org. Um, they they have great exercises. So, I I still do free code camp every day. Yeah, I try to do 20 minutes in the morning when like just to keep my skills up because it can give me like some Python exercises just help me remember language or I'm learning JavaScript, god forbid, you know, I can I can do a little bit of that. So um like I think I think like finding something that builds data structure is really helpful. I also did the entire CS50 uh Harvard computer science class and that really clarified a lot of like the interconnection for me. Even if you don't do the exercises but you just watch every lecture you can really get a lot of information there. And then like on the geoscience side, there are a couple guys on LinkedIn that that do a lot of like analyze a log with Python and um stuff like that that I think is is really good. Um is a good it's like a light introduction because L files although they're not perfect, they're relatively standard. There are packages for reading them. You don't have to do a lot of things to it to get going and you're probably already doing log analysis anyways. Yeah. So like that's like a lightweight place to live. I feel like a lot of those like regardless of where where it is or what the application is, those little pre-built kind of libraries or packages is like the gateway drug for a lot of people into a lot of into Python or R or whatever it may be. I think we were talking about that, I think with Andrew, but just about how like Python it almost feels like can feel low code at some point. Because like, you get this package and it's like, you know, one line or one method. You know, there's a lot of crap going on behind the scenes that maybe you don't, you know, know unless you want to dig into the source code. But like, you can download this thing and just say, you know, "read,l" and there it is. Yeah. Yeah. Exa exactly and like packages like that like lazio segio um all of those things are super valuable. um scikitlearn, sci like if you want to do statistics, those are a great place to do it. If you're only doing statistics, learn R. It's probably better. Um if you want to do all types of other things, Python's awesome. Um I think you know it it's all about like a strategy. Like if you're a petroleum geochemist and you're listening to this, all three of you on the planet, uh learn sci-fi because you can do hierarchical clustering, you can do K means clustering, you can do all the stuff you need for your workflow. Same if you're a petrophysicist, like that's a great place to live. Yeah. Um I don't know what geoysicists do. I mean like I I always try to pretend but well I mean well so we had Andrew on Muno's um have you met him or not? Yeah. And then what I think he actually did well he talked more broadly about geology with energy 101 but um I think he got into a little bit about what a geo physicist does but yeah I just mean like how do I how do I point you in a place to get started? I mean there are packages for loading segue files and stuff like that but in terms of processing an entire 3D volume like that start starts getting hard. Um but you know there are plenty of people on software underground that do it every day and talk about it all the time. Sure. So um like it's it's pretty accessible. Well and one thing talking about learning right now I mean I think you was it yesterday you posted or two days ago about chatbt on LinkedIn like you know I almost felt that was almost like is this the new uh let me Google that for you. Totally. Like people ask you questions like I'm just gonna go Google what you asked me or I'm I'm just going to look it up on chat dbt. And I do that all the time, right? Like I I get asked questions all the time. I get tons of messages every week from different geologists, engineers, data science people, people that are just interested in our business, like all types of things. And I find myself a lot googling things for them. And and like I you know I think I have a pretty good breadth and depth of knowledge in the business but like I can't always remember every fact and fix. So it's like you know yeah like there's this big gap in in self self-discovery isn't right but like yeah you know adjacent interest discovery is is a real skill and and I think you know Generation iPad doesn't know it necessarily and, and some are very good, some are very bad. But like, we have to, we have to keep encouraging people to learn. So like, I hope people will chat GBT answers, but also maybe I'll chat GBT answers and we'll come to conclusions and I'll send them to them; and yeah, they can learn from me and I can learn from them. I like to keep my finger on the pulse of what people want to do. So Yeah, I mean I was so I did a bachelor's in mathematics, and like it was just a ton of proof base, and I think that that actually taught me how to Google really well, cuz I'm like, "How the heck," you know? But I learned really well how to put in like what I needed to find to get, you know, good answers to help me kind of unlock stuff for me. And like that's probably one of the biggest skills I've used now what I'm doing. It's like it's not like yeah, I'm not that smart honestly. I mean, I just I know how to find the answers or find or put the pieces together that work to solve a problem. Like Oh, I I agree with you. I think there's, like, there's a sweet spot, just generationwise, where we grew up with the internet for the most part, but we didn't have a lot of, you know we were asking Jeeves and, literally, you know, I mean, Yahoo and all those, you know, all those searches from back in the day where, like, you really had to dig through a lot of [ __ ] to find some good stuff. Yeah. You had to do all of them too because like you couldn't just go to Google and get everything on cut page one, right? Or automatically with the GPT uh or Bard prompt now. But I was just reading a really fascinating uh article about how like on the top tier of Google results now almost everything is generated. So, so we're actually ending up in a world where the top returns are actually like generated marketing garbage. And so now you have to go to like page two to get the good stuff. And and I think I was listening to another podcast with the CEO of OpenAI and and you know he was being asked about like are we going to run out of data to train on? Are we going to do whatever? And I think the problem is actually much more insidious than that. It's it's that the training data is going to become generated and you're going to end up with the model drift that's that's unexpected for these people who work in Silicon Valley and and are used to access on deep sets of data that are available to them and their resources. But like it's going to it's going to drift their models in ways I think I I'm sure someone in the organization is thinking about it, but like I didn't hear it mentioned in the podcast and I think it's really the most pressing problem for large language models is going to be learning from itself. Learning from itself and and and mostly learning nonsense from itself. Like it already Yeah. Well, because it's very uh confident. Well, it is. And most of the language it learns is casual social language and interaction. It doesn't know stuff, right? The sliver of geology that's on the internet for large language models to learn from is so approximately useless to a subject matter expert in geology that it's it's really painful. And um we did some experiments with with chat GPT early on and and quickly concluded well this is awesome. Let's build on it and make you know geology GPT and really focus on that directionality of technical um tuning I guess and and and to really put emphasis there because like it can no doubt create language. It creates nearly perfect language and inspired and interesting language but what it doesn't know is is technicalities in context in context and so that was like a really powerful direction for us to learn. Sure. So, one thing I'm want to kind of expand on then is like, so, so yall are kind of going down a lot of the AI LLM kind of side of it, or I mean, because you, you've been doing this for a few years and it's changed rapidly. You know, you know, even a couple years ago it was deep learning, but I mean, can you kind of even just dive into the difference between machine learning, deep learning, and then some of this cognitive stuff that we're in now and just kind of like the overlap but the differences or? and where that you know kind of how that applies with you guys and where y'all are using it and stuff. I'll I'll do my best to delineate our space. Um I think this is like a really complicated question and really smart computer scientists argue about what's what every day. So, you know, if people ask what I do, I say we're an AI company. I used to say that we were a computer vision company because pretty much everything we did was in the visual sense. So a lot like what they do in medical imaging with uh like identifying tumors and looking for skin cancer like that's what we were doing in geology that was the last two and a half three years and we've gotten pretty advanced on that. We've got models that cover a lot of the a lot of the disciplines and are always building more and um you know in the last 12 months we've really started leaning into the language side um actually not to do things that most people think it's it's not to be like what is the paracity of the such and such or tell me about the um KPIs of Exxon Mobile like you can do that that's that's all available to you as as part of the product, but that's not really the intent. The intent is to obfuscate away work. So, right now, if you're a geologist, about 10%of your time goes into imparting your knowledge on the system, and like 80%of the time goes into just clicking buttons to get things done. Yeah. And then 10%more of your time is like actually deciding if what you made was worth a damn. And and and like that's that's crazy, right? like you know we have we have Petra, we have Patrell, we have Kingdom, we have all these softwares and like I have to click and click click and click and click and click and click and and that's my job, right? Uh that's a waste. Um so our our goal is to orchestrate that work. So uh instead you know our language model and our kind of cognitive model knows oh Dave his user persona is organic geochemist um and explorationist. So kind of make those models available to him and uh what I might type in is something like uh create an end of well report for these five wells and contrast them to this one. while you're at it, go ahead and analyze this thin section, that core, and this thing, and wrap it all up and send it to me. And then my job is to walk away and sit in six hours of meetings because that's what we do at big companies. Yeah. And then to come back to my desk at the end of the day and review that information and go, okay, does this align with my worldview? Does this align with what I know? Is this is this in accordance with my heristics? Yes or no? If yes, okay, do I believe in it enough to move forward? Yes. Okay, let's bring in the other experts and have them vet it. If no, okay, I need to do remedial work. Maybe I didn't have access to the data I needed. Maybe something was missing from my command and my prompt in my workflow. But like our view is that work orchestration is the future of work. It's it's it's not button clicking. It's not going through stuff. And and so like we've been focusing a lot of our language model development on orchestration and and so it's kind of like a a fascinating world. So like how does all this stuff exist? Well, like I don't know. I don't know that we'll ever get to artificial general intelligence. Like, yeah, it sometimes it feels really close, right? And and I just saw an article about oh, you know, one of these experimental models at one of these companies ran into a capture, so it went out to the internet, Fiverr, and had someone solve the capture, right? It's like, well, and even then, the the person that they messaged said, "How do I know you're not a robot?" And it I forgot what the response was, but it was like it was pretty intelligent as far as like how it right. It over overcame the obstacle. And it's like I don't know if that's general intelligence or if it's just copying the internet, right? Like who knows? And and maybe it doesn't matter. But, you know, I think there's there's kind of a this aspirational world of general intelligence, but like the oil and gas business is pretty far behind in adopting any type of machine learning, AI, computer vision in in different like they're really good at it in certain spaces, right? But but like any business, they're really bad at it on the like broad deployment sense. And so there's so much, you know, to be gained from from putting this in the hands of people that that I think like we don't we don't need to worry about like where where open AI is going. Like good as soon as they make the next coolest model, I'm going to be piggybacking off of it and training it for my expertise. And and I think that that's a really um fruitful place to be. So you know we do so when I think of like the world that that we're talking about like when I think of machine learning I think of like numerical approaches on digital data sets. So like that's what's in my mind. It's like seismic interpretation identifying faults. We don't do a lot of that at at GLU. We do a little bit we have some log analysis modules. We have some things that that dovetail with all of the unstructured things that we analyze. And so most of our models like I said are in computer vision which their entire purpose is to take legacy information make it structured indexed and searchable with no action by a person. Yeah. And and so like we we do that we deploy that regularly. So we have one client that gave us a couple terabyte drive and and said you know show us what you can do. Right. Sure. And so, you know, over the course of 12 weeks or whatever it was, um, we let the system process through and create that structure and and build information behind it. And so, at the end of the day, you can go into the Envision viewer, which is, uh, the tool that's client facing. And you can basically turn on and off tags and get absolutely everything you need from that completion report to the thing that says there was an oil show to um anything that has handwriting on it because maybe that note is really valuable. Yeah. Like you can find it at the click of a button and it lights up and is in front of you instantaneously rather than having to go to like Azure Cognitive Search and type in a string and burn like $100,000 in search fees every month. which is what this client was doing prior to our implementation. They were burning a couple hundred thousand a month. That's insane. In search because why? All their language data was unstructured, right? And it would oneoff search it every time you tried to hit it. And it's like, well, that's insanity, right? And and so not only did our searches become way faster, but they become way less costly. Sure. All right. Right. So I mean this is not I mean geo is in the name obviously because that's your background but this is not specifically geo work necessarily like no it can it can kind of understand and do anything if you're a finance person it can find the stuff you're looking that sounds similar to a project we did at at Grayson Mill recently where we um and if you want to get into the stack that you use or if you want to keep it more you know close to the vest that's fine but like we used like the tesseract um deal and ran some like leases through it because there was a clause that our land team wanted to find. Sure. And um you know our data engineer Sean you know ran OCR over over those and then dumped all that information into elastic search and we were able to index it and you know search that stuff and you know right how good was your OCR? It went pretty well. I think he found like a handful of files that just weren't like formatted really. But he was able to take most of them and like align them and work. But and some of them they almost had like shadowy stuff over the letters. So it was they weren't great. But I mean that was a small fraction and those were ones that someone individually could go Yeah. look at if they needed to. But yeah. Cool. And and so when you scale that those those few dozens or hundreds of records 4,000 we did. Yeah. Yeah. They they become tens of thousands of messed up pages. Oh yeah. Right. And and so like we've spent a ton of time building models that completely cope with that. Things that are not in full page width text like written leases, right? Things like drilling reports that have multiple columns and are weirdly assigned and you know like the hardest part is getting the language out in any way that is some cohesive thing that can be searched. And and so like we've spent a ton of time building models that do that across different formats and layouts and old typewriter fonts and handwritten things and and like those are the gaps in Tessact, right? Like Tessact's great if it's a relatively modern font and it's even something that came from a computer or a typewriter that is a fairly common one. Yeah. But like if it get starts getting weird, man, you're in a lot of trouble and it takes a lot of effort to to kind of get that data validity up there. And so um yeah, we we take a a similar approach but with different tools. So we have our own we have our own OCR engine. We have our own uh search uh kind of structure and engine that's built off kind of a different technology than elastic search. So yeah. Nice. No, I think that's you know going back to how much data we have especially how much data we have that is not even remotely close to being or it might be digital but it's not structured you know it's a PDF. Yeah. It's not digital. I've got that's computer paper, right? I've got a decade worth of uh frack, you know, post job reports and PDFs. Fantastic. What do I do with that? It's just garbage, right? And so I think that's one of the coolest things. I mean, when I saw y'all's demo and stuff at Tech Night, it's like, man, just being able to like say, "Hey, I'm looking for, you know, completions." And then you can get into how how y'all have been able to structure everything and make everything discoverable regardless of, you know, what division of the company it's in or what what uh region or what basin or any of that stuff. It's it's all there and it's all linked in these really nice clean ways. It's it's pretty awesome. Yeah. Thank you. Yeah, we're we're we're really happy with it. I mean, again, for something we didn't think that we were making, like it's it's, really turned out to be cool because the. the beauty is once you, once you read a...and we say "pages" a lot, but like, it might be a slide from a PowerPoint, or it might be an actual PDF page, or it might be an Excel table from a report. Once you've read it and recognized the system has recognized what it is, it takes action. So if it's a table of parasium permeability data, it converts that to a parasium permeability graph and where it can pull in like the lithology data either from that table or from some other comparable table. It will then color the images by that procility data because it's all connected, right? And so like it's taking action like in a lot of ways that you don't even expect um as the casual user like a lot of people are just like, "Oh, I'm going to use this as a search engine. And it's going to make finding things better, and it does 100%. It makes finding things way, way better, but it also makes finding the derivative product so seamless, you don't even know that it was created by the system in the background. Yeah. Well, and that's I mean, you know, we talk about silos in the industry so much. And that's the biggest issue, right? Is it's like, oh, well, so and so's got, you know, a year's worth of these reports on his my docs folder in his computer. And then, you know, someone else in a different basin actually has them structured and in a database. It's like, well, how do you unify all this stuff? because that's what a lot I mean so many people do, right? Whether it's on the geo side where it's like, okay, we've got all these logs, they're individual logs, we're going to put them in the software, we're going to line them up, we're going to start matching the the zones and stuff and figuring out what this actually looks like in a threedimensional space. And then, okay, we're also going to throw drilling on top of that in completions and reservoir. There's just so much data from all these different places that we have so much of it, but it's going unused, as Bobby mentioned, in a lot of ways because the amount of human effort and resources that would have to go into that to get it all structured and synchronized and and playing nicely is tremendous, right? Yeah. It's it's really multivariate and it's and it's very um integrator like it's it's and it's like you know the principal job of a lot of people is to like try and put information together but if you can't find it you can't put it together right and and then like if you can find it but it's a pain in the neck like you don't you don't actually get the job done. So we had we had one little pilot for a company where um their problem is their end of well reports. Uh basically every time they drill a new pad, their ops team is supposed to evaluate those end of well reports from however many wells are within x distance of that new one. Right. Well, that could be two or it could be 200. Yeah. Right. And and so what they're doing is they're going into the server, double click, and then they're going to the well name that they know that or the pad name that they know they need to review, double click. Then they're going to uh geo steering, double click. Then they're going to uh reports, double click. Then they're going to end of well report. Double click. And then the PDF is there. Double click. Right now it's open. Okay, that took, you know, not that many minutes. maybe a minute or two if you're good at finding stuff. But uh you know now you got to now you got to do that 50 times, right? And so, in the Envision system, you turn on the End of Well Report tag and you can put in the latl/long that you're working from of the new pad, or it can populate it automatically, or you can search for a well pad name that you want; and then all of those end End of Well Report slides are there as individual entities, and you can say, "Oh, I actually only want the Inzone Percentage slide."give me that. And now they're all there in a grid. And you can just scan them real quick and go,"Okay, uh the last 10 wells, they were 80%in zone. There's one fault."And uh we think this zone is really bad for sticking the bit. Yeah. You know, and and when we when we pitched them the pilot, we said,"Well, wouldn't wouldn't you rather just say, "Write me a synopsis of the end of well reports for everything within one mile of this new location." They're like, "No, we don't really want that." I said, "Okay, cool. That's fine. You I won't give it to you yet." And and so instead we showed them this super powerful search of click three tags and you have everything right in front of you. You don't even have to open them, right? Like they're they're they're complete pictures on the screen. So like you don't ever have to open the PDF. Okay. Yeah. Well, that does a lot of cool things for you. One, if they're in the cloud, it saves you opening them fees, which everybody gets hammered with, right? Yeah. And if they're on your personal server, it just saves you the time of like opening it and it's 600 pages and it takes five minutes for it to come open, whatever. So you know but what did we make as a result this ability to orchestrate write me a summary of the end of well reports within a quarter mile of this you know if it's got this amount of redundancy only do 10%of them whatever it is right and so now you can orchestrate your work that way and then you go to your ops meeting and then you come back and it's on your desktop. Yeah. Is that so is the idea of being able to essentially prompt it something that y'all either currently have or are working on in the Yeah. So we've we've been building the background models for a while. There's functionality in the system like demo style. It's mostly limited by the actual data we have in the back end right now because like our demo data sets aren't very deep. Yeah. So like I can tell it to give me the core core summary of the 10 sleep and it will access five cores and write that into a report and I can download it as a word document. It's pretty cool. Yes. Yeah. It's really good. No, I mean that's Well, because that's what ultimately that's what a lot of people are using on a daily basis, right? Is it's like morning drilling meeting, morning, you know, it and they want to know, they don't care about the detail. They want to know a summary of where are we at on all of the active rigs. How many how much time have we been there? How much time is left? Are there any issues that are going on? Is it going to be delayed? Is it going to be ahead of schedule? Is it and then that's goes into, you know, how that affects the the production and completion stuff downstream of that and upstream of that. And so it's uh, yeah, I, I think that's and it's also the way that people, you know, especially now with all the LLM stuff, people almost think that way now where it's like, "Okay, um, give me this." And in this, in this space, it's, you know, that extended Google search almost where you're adding those very specific things that you're looking for. I like I my my my own selfish goal is to eliminate daily reporting from anyone's task list in the next 5 years. Yeah. I don't care what kind of daily reporting it is. My my goal is to completely automate that away. And and it's not because we don't want to do it. It's not because there's not value in it. It's because it negatively impacts the revenue per employee of each of these companies. And and I don't know if enough people use that as their most important KPI, but it really should be. Yeah. Right. Because if I can take a team of five geologists that spend 60%of their time reporting on the next thing that's going to happen and I can I can eliminate that. They can still know all of those report aspects, but they can be reallocated to finding behind pipe opportunity or stepout opportunity or completions collaboration optimization opportunity with the engineering team. That that has a return on investment many tens of times higher than accurately summarizing the last 20 wells. Yeah. Right. And and like I think we really lose sight of that as processoriented companies, right? Because because it's incredibly important to summarize the last 20 wells so that the engineers know what they're going to run into. It's not incredibly important that a person does it. Yeah. The same. That's what computers are really good at, right? The same repetitive process over and over again. And humans really don't like doing that [ __ ] most of the time anyway. So it's a it's a win-win. I'm kind of like what Matt Harman's talking anyway, making work suck less. I mean like people spending way too much time. I mean, right, let them they can be productive and have lives and like they shouldn't be fighting right bad software to do it. Exactly. And and like we have a we have an incredible talent problem in in oil and gas, right? Especially in subservice like engineering is one thing and and you know engineering enrollments have gone down quite a bit and they've started to come back a little bit, but they were never as low as geoscience positions. Like entire petroleum geology departments are now gone from dozens of universities. and that is a really critical risk because
you know drilling horizontals in the lower 48 is not long for this world. Like I've I've worked 20 North American basins. think I've seen every horizontal play and I think I know the cut point at which each one of them is economic and we have to still go up quite a bit for any of the marginal stuff to come back to being a reality. Yeah. And and so that means like we're way behind on the resource discovery curve. We're way behind on the supply curve and we're way behind on the personnel curve because so I'm 39. I just had my birthday on Monday. Happy birthday. Thank you. and and I'm about the oldest of the young geologists. I came into the I came into the business right at the 2008 crash and and so after me everyone no one else stuck it out. So they all they all they all failed out of geology and went back to engineering or something. Um you know but so like there's not a lot of people you know behind me for a couple of years and then there's a blip coming up but then we had the 2014 and 2016 problems and so it's like there's just not a big trunch of people there. Well, really how many of them have even done exploration work because it hasn't almost zero. Yeah. Lots of lots of well sitting and there's lots of stuff to be learned there. Sure. But when they were sitting wells, if the computer was sitting wells for them, they would have been able to learn exploration and and step out development opportunities. And so like there's a problem there. And so then Steve, my co-founder Steve, he's like I said, he's been in it for 40 years. So he's on the younger end of the baby boomer retirement wave. So, like all of his friends are out or on their way out of the business in the next year or two. And like he's he's explored everywhere there is to explore. He's been on the ground in 70 countries. Like it's pretty hard to find someone in our business that's been on the ground in 70 countries. You're going to have to bang on a lot of doors. Yeah. Right. He's done Libya and all of Africa and uh both coasts of the Atlantic and like you know he's been there and done it. Yeah. And like I think I've covered a lot of basins. done 50 basins in 15 countries, but like that's nothing compared to SNE, right? And and so it's like we have but you're going to even be in a top what are percentile people, you know, right? We have a we have of experiment of experience, right? Well, I think I mean that's another, you know, the people that are pretty anti-AII stuff. I think that's one of the biggest things as an industry that we we need, right? Like institutionalize some of that knowledge. Yeah. because they're either retiring. They've been forced into retirement. They have retired or they're dead. I mean, to be honest, like there's a lot of like my my father-in-law's in his 70s. He started in the 70s back in the the day he was a petroleum engineer and stuff. And he's going it's, you know, his friends are slowly getting older and all this stuff. And that knowledge that they have is a lot of people don't think it's that relevant because of all the shell stuff. And it's like how much more production of like I don't know what the split is from shale versus conventional but I I know conventionals outpace from a production global production perspective outpace shale by quite a bit and so yes conventionals are the new or unconventionals are the new the new thing but uncon or conventionals aren't going anywhere and they produce so much and then I mean The only reason I know anything about conventionals is because of my internship ship and stuff that I did early on. And then I went into shale, and I've always I've been in shale ever since. And it's like, unless you're work in the US at least, unless you're working Gulf of Mexico or some legacy, you know, perian stripper law type stuff. yeah there's not much room for that and then you go outside of the US and it's like it's all conventional stuff for the most part right and so being able to have the knowledge and train models that impart the 30 40 plus years of knowledge from the last generation
especially in
all these secondary and and tertiary kind of tiered acreages aren't very economic even at $85 a barrel today you got to get over a hundred and we we talk about all the time about making a digital Steve so like you know like that's a very real goal right and it's it's because I think People always used to live on in their publication record. Yeah. Right. And and so like there is a lot of knowledge that would get summarized by people at the end of their career in their 60s and 70s as APG members. I this is what happened when we discovered this field and this is what happened when we did this and that's really fallen off. And so there's there like unless you knew the person like you don't know how it got there and how it happened and what went on and and the insight that was gained. And so, but like if you can if you you can do that in your company, you can in very real ways never ever lose an employee again, right? Yeah. And and I think that that's when I talk to the most CTOs in the oil and gas business, that's the biggest thing they're missing in their vision of the industry, right? Which is that they have the opportunity to never again lose an expert. Yeah. and and that's not core to many of their strategies. And it's because they're overwhelmed by the SAP and Oracle and the accounting systems and all of this other stuff that they think is where money gets made. I'm sorry there there no money is made in the accounting system. You can't squeeze a barrel of oil out of it. The money is made by having the resource in the ground that's producible at a value at a cut point, right? And and like those are the people you need to be really really worried about not having access to anymore because we're going to have to relearn it. Yep. Right. Like like as an industry, we're going to have to relearn how to explore for oil and resources and whether it's uh carbon sequestration reservoir or a geothermal reservoir or an oil or a gas reservoir. Like we're in very real danger of of going through the pain of having to relearn all those things. No, it's it's you're spot on with that because I agree with you. A lot of the IT stuff, they they see it as a well, this is where a lot of money is, therefore that's where the money comes from, right? But it's like if you're an IT department looking at an accounting group, the way you're going to benefit the company is by cutting costs there. And they're not generating revenue, right? If anything, they're a cost center. Cost they definitely are. It's and we have we have a lot of great people back there, but it's like I never understood until I was at this company where I had the view of it like how big the back office is. It's huge. Well, a lot of it is because we still do so much [ __ ] on paper and we put in all these processes like field ticketing. Don't even get me going on that [ __ ] uh you know the fact that you have to get a ticket stamped or stickered and then submit it and then just so an accounting person can manually code it when they get the PDF of it. It's like this is literally exactly what computers were made for, right? Like code it all up, bring it all together in a back end with a few tables and it will just do itself, right? Like you don't need humans touching every single invoice that comes through the company. Yeah, we've got this has been awesome. time has some more. There were things I want to get into, but I don't I can come back in another few weeks and we can nerd out more. At some point, we're going to do just like AI panel and we would love to have you back on that just discussion. It' be great. Okay. It was funny you brought up the the digital Steve though cuz I heard on the way in that at the uh Cowboys Stadium this year, they're going to have an AI Jerry Jones that people can go talk to. Oh my god. That's And they were saying on on the radio like this going to be interesting like if they censored things that they can ask him, you know, someone's going to ask them about, you know, improprieties or this or that. And like it's just be interesting to see what the thing comes back with. Do you guys know the Tiger oil memos? No. Oh my god. Okay. When we get out of here, Google Tiger Oil memos. Um, but they are a treasure trove of ornery, offensive, outrageous, uh, internal memos from an oil company that existed, you know, 70s and 80s and I think even into the 90s, but they're like a fascinating glimpse of, uh, you know, how some people run companies, but I would love to make a Tiger Oil GPT. Um cuz it's just like you know like there are oneline memos about uh don't you dare leave for lunch before 12:01 like uh oral the spoons are mine stop taking the spoons from the break room uh I mean there just like all range of hilarity mostly not at all pertaining to to oil and gas but just like inter office operations and this guy's view of what you were allowed and not allowed to do in his company. I'm going to check it out. Yeah, that's there's so many of those. I might have to I might have to scrape that and and make Tiger Oil GPT or just like uh Smarter Child back on AM and just ask it questions just to get the ridiculous answers from Yeah, exactly. So, at the at the end of our episodes uh we'll do we do like a speed round where we just kind of ask you some questions and pepper off some short answers for it. The first one I'll ask is uh what's your you know favorite um favorite book or kind of most influential book or top you books. I really like uh Travels with Charlie in search of America. Uh it's a Steinbeck book. Uh it's a story of his own travels across America and like kind of things that he's seen. It's fascinating. I also really love uh Sand County Sand County Almanac. Um that's like a I don't know that was like firstwave environmentalism. Um it was like a really popular book for that. Uh it's basically a chronicle of a guy living in San County and uh all of the comingings and goings of wildlife. Super interesting. Yeah. You got uh kind of two piece. So cloud or onrem and if cloud what platform? Yes. Yes. That's a good answer. No. Yeah. Because we didn't really get into that. I was going to ask how stuff was deployed. We we I made a joke at the AAPG and I got chastised because uh he was not from America. Um so he doesn't like it, but I said, you know, like we're American Express, right? We're everywhere you want to be. Um and like that's unfortunately the reality of of private, you know, AI services is we have to be on prem, we have to be in the cloud, we have to be hybrid, we have to be private, we have to be whatever. So we unfortunately meet everyone everywhere and uh it's the bane of my existence. So I mean it sounds like you you had to build it modularly so that it could be deployed roughly anywhere. Not not like cloud native or anything but yeah I mean like it it is served internally on AWS and that's where it lives but like only because they provide startup credits and that was a great place to build it. Um it doesn't have any particular advantages over anybody else. So, cool. Um, what are some of your favorite git repos or just some git repos even in general that you know geos may or may not know about that are that would be super helpful? That's a great question. Um, or libraries. It doesn't have to be a repo specific. Yeah, I really love um there are a lot of different repos that like chronicle advancements in different AI models and so basically like you can find them and they're like great at chronicling the difference from one thing to the next. I don't I can't think of any names off the top of my head, but I really like those. They're like great summary places to start playing. Um a couple AI companies make some really cool like open-source like H2O AI they make. Yeah, like the ability to make chat bots pretty easily. Um, there's like uh I don't even remember what it's called. It's like uh open chat UI or something like that. It's it's basically like chat GPT front end, but you never have to build it. It's it's just a repo that you can just plug in and sweet. You can use it as your front end. That's really convenient. Um stuff like that is good. Like I like I like things that are like kind of generalized and and like you can plug into some other usability. So those are my Cool. I I don't code every day, so I don't have a lot of great insight. Oh, it's right there with you. Yeah. Um, so you're a really good follow on LinkedIn. Who do you like to follow? Or is there is there a certain follow that you really enjoy? Um,
man, I feel like I thumbs up a lot of stuff I see on LinkedIn because I really enjoy it. I like I like a lot of this Texas focused energy and grid discussion. maybe David Blackman and a few other people like they talk about this stuff a lot and I think it's really fascinating. Um I I Snoop Dogg I just started following Snoop Dogg. I haven't seen him post anything. Yeah, he just he just joined. Yeah. No. Uh Snoop Dogg. So everybody everybody follows Snoop Dogg. Um I think it's a fascinating like brand pivot for him. So I I haven't seen like anything come up in my ve from Snoop Dogg yet, but I'm hoping. Fingers crossed. Um, oh, I I really like the uh who's the CEO of Liberty? Uh, Chris Rice. Yeah, I love Yeah, I love following Chris. He's always got something good to say. No. Yeah, he does. And he's got a good one, he's got a great background for it, right? Like he's if anybody should be talking about that stuff. He is one of those people. Yeah. Uh but he also just has a great perspective on it, right? It's not one way or the other. He he is very straightforward and factual about things and also has a really good diverse background and experience to kind of back up that that opinion. I absolutely agree with you on that. Yeah. Um
we'll do one more and then we'll wrap it up. Favorite video or board game?
I don't play either and my wife hates me for it. Um I really Okay. there. Have you have Have you played the game? I think it's called Pandemic. Have you played it? It's It's like uh you know, diseases spread around the world and you have to fight a video game or No, it's a board game. I think I played played the app version of that for a while. It's pretty fun. If I if I have to play a board game, I'll do it. I'm not great at remembering all the rules. I I will play some D and D. I I like I like that a little bit. Um but like maybe once a year is enough. Yeah. Uh so yeah, I'm not a big gamer. I spend a lot of my time being a nerd and thinking about geology. So yeah, I'm not far off with you. Like I haven't played video games for probably since I was younger and it was always sports games, but like for me like people like to do puzzles, like if I'm going to solve puzzles, if I'm going to use my brain that way, I'm going to do it, you know, like on the computer. Yeah. Yeah. Something something financially beneficial. Totally. Uh I'm a big outdoors guy, so like if if the choice is play a board game, play a video game, or go ride a mountain bike, I'm going to ride a mountain bike every time. For sure. Yeah. Cool. The last question we normally ask is what's what's some advice you would give to uh to someone you know geologist, engineer, etc. getting into or trying to get into kind of energy tech space. Start building. Yeah. I mean I mean and and validate often. Um and also don't give a [ __ ] about if people don't like your idea. Um Absolutely. I I I think I've been told 2,000 times since I've started this company, uh, it's never going anywhere. It's never going to be adopted. It'll never work. Oil and gas is dirty. You can never do this. There's no money for that. And and like if I had listened to any of them, we wouldn't we wouldn't be where we are today. And and yeah, it wasn't a perfect journey. It's not like linear. But I think I think like if you want to do it, get after it. And if you know nothing about starting a business, find a find a mentor, find someone that's that's built something and and just let them be your cheerleader. They'll probably also say no too many times, but it's because they're they're pushing their own experience on you and and so just like like hear it, internalize it, and then decide if you want to pay attention to it or not. Like it's it's totally valid to ignore advice. So feel free to do a lot of that. I think like if you follow like how it's the how it's made podcast or how we built this how we built this like over and over and over again those stories are people that just like said no I'm doing it and and made it happen and and I think that's how advancement goes. Yeah. No, absolutely. You've got to be if if you're not being rejected, it's probably not that innovative of a of a thing, right? It's an iteration, not a step change almost, right? Yeah. Well, where can people find you? Where can people find you aluminina? Yeah, you can find me on LinkedIn, David Thu. Follow me, uh, connect with me. I don't care which one it is. Actually, I prefer connect. So, do that. Uh, you can find Geo Luminina at www.golumina.ai.
Um, we've got a we've got a website that kind of talks about the product. Uh, I don't spend a lot of time anywhere else. Most of my time is on LinkedIn. I I have a I have a uh Instagram, but I stopped posting on it because I only post on LinkedIn now. So, that's where you can find me. Cool. and and oh, we're in Brian and College Station. So, if you ever come through there, uh, hit me up on LinkedIn and I'll take you out to the best restaurant I know and the best cup of coffee or the best beer. Um, and that goes if you're a student or an executive or in oil and gas only tangentially or a Crusty Field person. It doesn't matter to me. Um, you know, I I love the community up there. So, if you come through, uh, make sure you let me know. Awesome. That'd be awesome. Yeah, appreciate it, man. Yeah. Thank you so much for having me. has been great.