Energy Bytes

Applied Science Manager at Verdad Resources, joins hosts Bobby and John to tackle the pressing issue of cybersecurity in the oil and gas industry. With his deep expertise, Kurt delves into the costly mistakes companies are making, such as ransomware attacks like those on Halliburton, which have exposed vulnerabilities across the sector. He discusses the growing cyber threats, the challenges of maintaining operational efficiency while enhancing security, and how data and AI can help build resilient defenses. Kurt also offers unique insights into unconventional strategies and the future role of technology in protecting the industry's future.

Join the conversation shaping the future of energy.
Collide is the community where oil & gas professionals connect, share insights, and solve real-world problems together. No noise. No fluff. Just the discussions that move our industry forward.
Apply today at collide.io

00:00 - Intro
00:48 - Halliburton Cyber Attack, Cybersecurity, Data Breach
05:36 - Default Mode of Operation for a Complex System is Failure, Systems Theory, Risk Management
08:00 - Testing Your Security, Security Assessment, Vulnerability Testing
10:18 - Different industries, Industry Applications, Sector Comparison
10:52 - Feedback loops, System Feedback, Continuous Improvement
17:50 - Different teams, different goals, Team Dynamics, Collaboration
20:02 - Useful statistics, Data Analysis, Key Metrics
23:54 - Trade-offs and models, Decision Making, Modeling Techniques
31:43 - How to encourage more experimentation, Innovation Strategies, Experimentation Culture
34:07 - What are we looking for in an experiment, Experiment Design, Hypothesis Testing
35:31 - Learning from failures, Failure Analysis, Lessons Learned
41:55 - Engineers vs scientists, Roles in Science, Engineering vs Science
45:20 - Applying science to data, Data Science, Scientific Methods
51:10 - Communicating complex data relationships, Data Visualization, Communication Strategies
53:18 - How to validate models, Model Validation, Accuracy Assessment
57:30 - Charlie Munger, Investment Strategies, Decision Making
58:56 - Overcomplicating things as engineers, Simplification, Engineering Challenges
1:05:40 - Python scripts, Programming, Automation
1:06:40 - Shazam for logs, Log Analysis, Data Recognition
1:09:48 - Speed Round, Quickfire Questions, Rapid Responses

https://twitter.com/collide_ai
https://www.tiktok.com/@collide.io
https://www.facebook.com/collide.io
https://www.instagram.com/collide.io
https://www.youtube.com/@collide_io
https://www.linkedin.com/company/collideai

What is Energy Bytes?

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 to another episode of Energy Bites. I'm Bobby Nilan. We've got the Rad Dad, John Calfan, here with us. How's it going? How's it going? And uh excited to have Kurt Rucker in from uh Colorado, uh the Applied Science Manager at Verdad Resources. Is that right? I'm happy to be here. Well, welcome [snorts] welcome down to beautiful Houston, Texas. It's Yeah, it's awesome. It's nice and cool and ready to go. you picked a good good time. It's been the most uh comfortable end of August that I think we've had in a very long time. Yeah, and it didn't seem like that about 3 weeks ago. No, yeah. We went from the the hottest week of the year to some of the most comfortable Houston. Very few people are complaining about the weather down here. Yeah. Just wait wait a day. That's not the perk. Now, are you from this area? From Dallas? I grew up in Dallas. Okay. I cuz I saw the 512 phone numbers and I was like, "Okay, he's not from Colorado originally." And then uh you know, I was trying to figure out, you know, where the connection was, but um Well, cool. Um but I know, you know, we were chatting a little beforehand, but I mean kind of actually actually remarking just how not nearly as many people in oil and gas were aware of Halliburton's cyberattack. Right. As you would you would imagine. Yeah. Um especially people that are tied in. Um but just, you know, we can talk a little bit. I mean, it looks like it seems like some ransomware, um which is becoming more and more uh frequent and uh just kind of get interested to get you guys'thoughts on on that and uh just that potential risk for oil and gas in general. Yeah, so it says a uh they believe it's a group called I'm reading, but um [snorts] And of course, Halliburton came out today as of this recording at least and said that they don't believe the hack is going to affect their bottom line, which all [snorts] good PR firms would say, but Yeah, they're they're doing their job right now, clearly. Yes. Yeah, I mean, the fact that most people don't know about it, I think shows that they're doing a very good job and probably deserve a little bonus. Well, at least they're doing a good job of keeping it [snorts] That's what I'm saying, like managed. Yeah, yeah. The real work is on the back end of keeping it getting a solution worked out. But yeah, it just says, you know, the uh, they're still evaluating the nature and the scope of the information that was stolen, and they're not commenting on what or how or or any information about it; but brings us back to I mean, we talk about hacks, I feel like, every other month. Yeah. [laughter] on here and I'm I think it'll be I don't know if we'll ever hear any of the details [clears throat] about it, but I do think it'll be very interesting with how much especially [ __ ] we talk about IT being barriers for tech [clears throat] innovation. It will be interesting to see if this was a, you know, outside penetration or if this was just like a phishing email or something generic like that. Password And that's really scary, too, cuz I mean like this is a big fortune whatever 500 whatever company you know, publicly traded company and like again who conceivably you think you would think has their IT [ __ ] together, you know, a lot more literally has to with socks compliance and everything like that. I mean, to be at a mid-size company like us and you know, I mean you all are mid smaller size. I mean, we're I don't want to speak for everyone, but since I mean, we do a good job of securing ourselves, but I mean that not nearly to the degree probably that of a publicly traded company does or you would hope that they do. Um that you know, anyone's at risk and I've even heard of smaller players getting hit even, you know, in the last month or two. So, it's a it's definitely an exposure that people need to think of and I mean even like do you have cyber security insurance? You know, you know, you know, your CFO I don't want to pay that. Well, do you want to pay the ransom [laughter]? Um but uh Yeah, get get your thoughts. And for oil and gas it's funny cuz we um value our we have expensive engineers and we think that the data that they're using is highly correlated or real closely connected to the decisions the expensive decisions they're doing. So, your insurance what are you really insuring? Cuz if you actually lost the data, the value of it, what do you how do you how do you handle that? That's that gets pretty complicated quickly. Yeah. And then you've got thousands or hundreds depending on the size of your company of employees and each one of them is a point of failure. And this I think that's a when you're designing cyber security or just your your data governance across your organization, being able to think about not just one one failure is almost guaranteed over a organization large enough. So, then how do you ad-accept that that's going to be part of it and design around it as opposed to trying to defend against one eventuality because all you one attack has to do one attack gets in and Right. Yeah. I mean it's a matter of I mean obviously the to your point like there's acceptable risk. You have to you if you get rid of all the risk, then people can't do their work. Right. Yeah. Right. Um I've heard of companies removing all the USB drives from [laughter] Yeah. Yeah, I mean I goes back to like I remember someone saying like if I if I want to hit a really straight golf shot, I'll use my putter. But like I can't use that [laughter] the whole way down the course, right? You know, so I sometimes have to decide is it worth trying to pull the driver out or you know, my five iron whatever it is. Um It depends on the game. If somebody's going to say you lose a child if you go off the course, you might putt the whole way. It could be worth it. No, that's a very good way to to look at it. It's man, I that's one of those things that's like as an engineer who just wants to solve problems that like the fact that there's essentially always a problem there is fascinating to me but also terrifying at the same time, right? It's like you have to design so that with the expectation almost that there will be penetration of some sort whether it's fishing or password, whatever. And so, it's like, that's just such a mind, like, hard for me to wrap my head around, like, okay, I know it's going to happen. But if it is going to happen, we just need to make sure it's designed so that we can contain it, identify it, separate it, back it up, and keep things moving. There's a there's a book called The Systems Bible and uh uh the one of the it's full of a bunch of kind of one-liners or quotes and comments about large systems and design. It's I think from the'70s. Uh but the one of the key points um that the author makes whose name I'm drawing a blank on is that the default mode of operation for a complex system is failure, not wild [clears throat] success. And as engineers, we like to think of I'll say I'm scientist, but the engineering mindset is to assume that we've designed this delicate wonderful system and it's going to work just how I designed it as opposed to it's going to be breaking all the time. What does it look like when it's breaking? And if it's unmanageable, if it means my data is locked away or gone or public, is that if that can't work, then your system probably isn't really well suited for what you're doing. And as organizations, I think we've moved from an industry that was more straightforward to one that's much more difficult to manage and to understand, harder to perceive and less intuitive. Sure. And that this is I think this Haliburton discussion is an example of one of those places where it's becoming so hard to manage, but we are you we've built our structure around assuming it's simpler. And how companies make that shift is I think that's the problem that all of us are dealing with right now. Yeah, and to your point I mean and because again like again they're big and they should have and I mean I'm sure they do have their stuff together. I don't want to make this seem like I'm just jumping all over them. But I mean but again but yeah, but the bigger that the bigger they are like again you start to see talk about points of failure. You have thousands of employees, you have however many IoT devices, cameras, like thing. all these different points of entry into a system like and you'd like to think you could think of all of them and the possible ways people can penetrate those, but I mean it's it's difficult. I mean like I mean it's almost, I I would say it's impossible, but I don't know, that's not really my world, so I can't really say that. But it's when there's people that have been hacked... there was one recently, and I can't remember I feel like it was a State Department or something but it coming in through a thermostat. Yeah. What? I didn't even know that was possible. And to imagine so then you talk about IoT in the field trying to protect against that sounds Yeah. awfully difficult. Yeah, and I mean Jeff Bezos will tell you in most places you can walk out on the field and there's probably like the admin password on the inside of the of the box you know depending on the on the on the company. Um Yeah, and I it's funny you were saying that cuz I think it was like the AWS CTO I mean he may still be, but he was back then, but I mean his whole motto was everything breaks all the time. Yeah. Like, I mean, you have to design with that in mind. Like, and I think AWS and I think Netflix, I mean they, they put their infrastructure through the ringer constantly and are going through like drills and stuff like that, because it's just, if you want to have that high availability that you say you want, then you have to really know how to anticipate and get ahead of the curve. Saying you have it versus testing it and actually proving that it can do that is very different. All right, we're secure. All right, let's go let's go do some pen testing. Yeah. No, like give me a couple months to fix this couple of things that [snorts] Imagine like a white hat black hat oil and gas conference of like go out into the field it's in Midland and we're going to just like let's see what we can do. Let's go play. I think that would rock [snorts]. Let's go to Joel Cutters IT [laughter]. I mean we're totally down to host that. Good luck getting operators [snorts] to sign up for that. You guys both in? Is that what I'm hearing? I don't know. Maybe [laughter] get it done in Saudi Arabia. I'm sure you can find some way to pull something off cuz that would be amazing. Grayson Miller Energy and There we go. [laughter] 4 weeks. Bobby's volunteered. No, that's uh yeah. I mean, that's the thing. Like, I guarantee you, if somebody wanted to, they could just go out to a number of locations in West Texas right now anywhere in the in the oil field right now and walk up to a SCADA box and, you know, in most situations be able to have a a field day. But that's that's the thing; it's like, I don't, you know, anyone that's ever built any kind of software, you quickly understand that it's more about fixing the broken [ __ ] than it is about making it work. Yeah. Yeah, and and so I mean, that's just... it's kind of part of it at these, at this, or, you know, table stakes with with that kind of stuff. But it is just one of those like very unenviable positions, I feel like, if you're in the IT group, cuz you're being pulled you're literally being pulled in complete opposite directions, right? one of those things where it's hard to get credit. Oh yeah. Well yeah. Especially in our industry where anything IT or data is seen as a cost center and you know do more with less like that's If nobody hears about you it's good and if they hear about you it's a bad thing so you never get credit for not being heard about [laughter]. It It's It It's interesting to look at how um So, I I find like in oil and in in all industries uh the heavy industries, the capital-intensive industries are really different than the tech industries. And yet, we keep looking at tech for how to manage data, but they're such different companies cuz they don't have If you ran the wrong ad campaign, Yeah. you're fine. Yeah. If we run the order the wrong equipment or install the wrong pipe or blow up a well, like it's it can be game-ending. And so, you know, as we think the risks for us and the or the risks for if you're Caterpillar or GE are just completely different. And looking at how, when any of those organizations really embrace technology, and how they're able to connect it to the decisions of their organizations, and how they when things are working, cuz we all kind of go through eps and flows of when it falls apart those are the companies I look at for trying to understand what can we change about our mindset in oil and gas. Sure. to connect technology to the activities that we're involved with when the risks are so high. Yeah, and then we talk about I mean, there's all all the time. I mean, again, it's like oh, Spotify was wrong, let's go to the next song. So, what? Um, but I mean, I but I think, you know, John John's talked a a lot about this, too, cuz he came from a mechanical background, and he did some of that more like just-in-time type, you know, lean manufacturing, and it's like... And again, like there's that's the hard thing about our world, is that the the facilities and the, you know, everything moves all the time. But at the same time, that doesn't mean we can't pull things from there. But, you know, those are probably a better, [snorts] um, analog for us to go off of than looking at, "Well, Google does it this way, why don't we do it that way?" [laughter] Yeah, well, that's the that's the big thing that, you know, I've looked at this a lot and thought about it a lot and you know, Google can you it is everything is on their platform for the most part, right? So, it's like they control all the data inputs, all the data sources for them I know that that's not true, but generically. I was thinking that's what [snorts] they want you to believe. Right, but it's like so, for example, CAPTCHA is such a great example and use case. Like most people don't realize that that's just you training their computer vision model. And it's like the way that they implement that is so like or well done in my opinion, right? It It's facade It's not really a facade for security, but it is a security element. But in reality, you're training element You've been having the world train [laughter] train your models for you for free for I don't know how 10 years what however long CAPTCHA has been around. And so it's like those are the things where it's like, "Hey, how do we learn from that and implement these feedback loops into AI for oil and gas or things like that?" We can't control all the data sources cuz we know that's [laughter] not how the oil field works. The The service companies have all the data and they share it with the operator, you know, there's that weird data relationship where it's like operator owns the data, but they don't generate a mid large majority of it, right? It's generated from Haliburton or whoever. And so then that's where everything starts to kind of really fall apart with the structures and the [snorts] the field names and just normalizing. We don't have any standards at least outside of drilling around any of that stuff and that's where things really start breaking down. Whereas if I'm Google [snorts], I have my data standard and my data schema and anytime someone fills out a form, it's going to be in that schema and standard. Yeah, but they're they're their structure isn't constant, right? Like Google or Amazon are good examples of companies that have been evolving in how they approach what they're doing. And the CAPTCHAs are an interesting one cuz I think of that as being like a network effect where they've and they were able to connect it where they had a need and they were able to build basically an internal network that built out of a positive feedback loop for them on their systems. But another part of that is they got really clear about what's the data we have and what's the thing we need to learn about it. And I think that's where oil and gas has the biggest problems is we have loads of data and we are afraid to share it because we're worried somebody else will figure out something before we do. Even if it's on a something you've like a well you've part You're a partners on. If you're one of your partners figured out something, you would still be your next in line to learn once they figured it out. And I think that we're so worried that we're proud of ourselves for being super smart um and having the the smartest employees, the smartest teams, and then we're afraid to share a lot of data cuz we're afraid somebody else might be smarter. Yeah. And I think that's something we can learn from the the true digital companies that they are getting really clear about what's the data that I care about. What's Where's the value to my business in this massive info of data? Where's the information that I can pull from it? And what's the What do I need to do with that? And the captures are a like a concise example of Google doing that. And I think figuring out ways, what would be the feedback loop that you'd want in oil and gas? You know, how many times have we It took a long time for combo curve to come up with a marketable decline curve tool that took off like it did. And yet we still don't have They probably have something, but a feedback loop on decline curves and engineering work. That seems like something that should be manageable or petrophysics or And then the decline curves are a huge one where it's just like and to your point like all these companies have some really smart people. And then I guarantee you at every major company and even even the number of hours are devoted to the same job. Yeah, there's still some kid that came out of college as a petroleum engineer and he's writing a DCA thing in Excel or Spotfire or whatever else. And it's like, "No, this has been solved for a long time." You know, why can someone from Shell just share their damn you know, Python script already? Like the Well, I think if So, if people weren't solving it the same way every time, Yeah. I think it'd be more interesting. it was if it was if [laughter] it was a if it was if it was a novel approach, Right. Right. But everybody's got rate, time. [snorts] Done. All right. Somebody got fancy and did rate [ __ ] or like it But there's not if you were coming at it from a different way and Yeah, almost like what like kind of Biota was doing at one point, right? Where they were using, you know, the declines from molecular stuff to you know, like that's a that's a novel approach. And then yes, maybe that is a competitive advantage and maybe you want to retain that. Or maybe you should share and see how it can be enhanced, right? Cuz you get more smart people on it. Like Yeah. Now, the perfect use case for that, as a frac guy, is uh FracPro was and probably still is the industry standard for frac acquisition and in-van uh frac work, right? Everything in the van, for the most part, goes through FracPro. All your stages, all your volumes, like it does all the calculations, it does all the math for you. As an engineer, one of my jobs was marking breakdown, you know, uh if we were dropping balls or whatever, flush, ISIP, you're marking things in the chart. I did that for and I know people are still doing that. Why wasn't any of that data stored and then put it fed directly? Like you had thousands of people using this for decades and then they still had to develop a new way of pick you know, like they should have been the first ones to come out with automated stage tracking. or ways [snorts] I mean I I think about completion, that's a challenge I'm always uh excited when I meet somebody that's working on something or has a a service or is kind of playing with the idea of uh having data around completions, uh especially if right after a frac, you could score if it was a good job, which is different than did you do what you thought you were going to do. But did you do a good one? Yeah. And the amount of knowledge the lack of knowledge that we have on that is amazing for one of the most expensive parts of our Yeah. of our assets. Well, then it comes down to what teams value though, too. Like is it, you know, speed [snorts]? Is it, you know, how well you fracture the rock? And Yeah. The completion engineer is going to have a very different bonus and and comp structure than the reservoir engineer in most situations, where it's like the completion engineer is focused on AFE and number of wells and the reservoir engineer is focused on production and, you know, it's it's that in and of itself is part of the problem, in a lot of my opinion, right? Is like you have teams that have completely different comp structures that are incentivized very differently. Therefore, they don't work like that's exactly not to go full conspiracy theory, but like that's exactly what the government would do if they didn't want people sharing information, right? Like they would silo them and they wouldn't let them share and they would incentivize them differently. It's But even like I mean to go even deeper on a more theoretical level like spacing, right? Spacing is not necessarily a geologic, you know, or engineering thing. It's also an economic thing. Like if oil's, you know, $120 a barrel, I'm going to pack those things in nice and tight, you know, cuz I can afford to. But you know, if it's less than, I'm going to go more and I, you know, I can get...but I...you might be destroying value later, but it's about keeping the lights on now. I mean, there's there's all these different levers that you can pull or that decisions are made on and it's not a one-size-fits-all. No, I think that's the craziest part about the completion side is just there are so many knobs to turn, right? Like stage spacing, I mean, your perfs, the fluid rate, the fluid chemistry, the amount of sand, the loading of the sand. Like there's just so many and that's just the well, right? And then you have all the rock and [ __ ] of course on top now you have like simulfrac and stuff which is like a huge deal like, you know, you can zipper twice and like Yeah, that's thinking of it in a vacuum, too, right? Not with any existing production, with any pressure sinks, with any depletion and all that stuff. And then yeah, that but that's the real world that we live in where it's like, oh yeah, there's this natural fault that runs right across and guess what? It takes the frack and runs it over to this well earlier, right? Like I mean, like you can design all these perfect things in a pretty little vacuum Yeah, see that's not that's not there's not a textbook for that. Yeah, and then the real world comes in and says, "No." Yeah, [ __ ] you. I'm sorry. Like So I I'm going to come back to the completions, but I think there's like two parts of it. There's one which is what are you chasing? And so what are the KPIs? What are the metrics that are actually useful? Uh there's a an investor Michael Mauboussin and he's an amazing person to read for a number of reasons um and a really good synthesizer of a lot of complicated concepts and thoughtful ideas. Um but listening and reading him, I've learned of the the term a useful statistic. So, you can have a term that is you can think of 10-year return invested capital. It's sort of like that's the thing we want. Um and so I want something that's predictive of that feature. Um but I also want it to be uh persistent, which is that as I measure it today and again today and again today and then tomorrow, I want it to be sort of the measure today's measurement to be predictive of tomorrow's measurement so that it's fairly consistent. It's not bouncing all over the place. Um and so there's some statistics you could use to like on time series analysis to figure out to say how much noise is there. Uh what's the rate of mean reversion as well as how correlated is it to the metric I care about. And that you think of a met a matrix of predictive and persistent and that's uh like saber metrics for baseball. John's trying to do his thing for his fantasy draft. He's trying to predict how consistent a player is week to week. Would I rather have the guy who gets 12 points every week or someone that gets 30 and then six? So, there's a yeah. We're going to come back to completions. Uh but so on that one, so if the person that's swinging, they're uh there's like their mean performance. And if somebody is, uh, swinging and has, um, is able to get away from that mean of the group or their mean and say sustainably above that mean, or below, for a longer period of time, you can attribute skill more to their performance. And the faster the rate of mean reversion, whether that's because of it they're involved in a chaotic process (so that's oil and gas) or they're just a more erratic slugger, that's skill may be less of an explanation for what they're doing. And so you can look at the rate of mean reversion as a way to assess how skillful their activities or how much their results are attributable to skill versus luck, which is I mean it's how you create it's it's a the construction of your fantasy You know, it's a derivation of uh the the mean and the standard deviation away from the mean. And so it's like it's the same thing like distance. It's the variance as well. long it's taking them to come back. So, if you're able to hold a small standard deviation above the mean, but you're able to hold that over a longer year, that's saying that that there's something that was happening there where you could should quantitatively at least attribute more skill to that game you're Right. Right. Something you were doing is responsible for that. Yeah. Uh, but I think like in oil and gas, how amazing would it be to be able to do the same thing to our engineers or our completions and start giving people using that kind of concept one to chase better KPIs and to think about corporate strategy, and how do I feed that back down the the chain of my my organization, so people are chasing things that are actually relevant as opposed to EUR or pounds pumped into the ground, um, or cost. And if you can turn that into a feedback loop, that's a positive thing. Or if you can use it to find the right metrics, so that's that persistence predictive, finding the right metrics to be giving people. So, you can give engineers feedback on did you do a good job or did you just get lucky? And if you can tell them and you are you in service of things that we actually care about or is it just a familiar metric? Yeah, or the company may not really bad. I mean like we've we talked we were we were trying to do this back in the day but I mean like if we can provide them like a template, this is what a perfect or good frac looks like. I mean like all right, are [snorts] you too aggressive, not aggressive enough, getting up to rate, you know, are you are getting to break down? I mean, you know, how long like certain things, but again then you get a cookie cutter thing and is that even what you're giving them? Is it even representative of the rock or the geology? how sure are you that that's even what you want? Yeah. So, I think that's the place where I start to break from the engineers. I feel like I'm often the geologist that's in the friendly camp with engineers. And then once we start getting into over engineering or overconfidence in our engineered path through something like completions, I think I go more to the data. So, we're talking about completions and spacing um and perf design and kind of how we design how we design the development uh development intensity of a square mile, say. Um there's a lot of fun stuff I find in the data to go and check. As you said, there's some tradeoffs between completions and um and economics. But we because that's complicated, we kind of assume it's sort of one-to-one. Half my completions doubles my economic or you know, I can double my well spacing or something. But in every basin, you can check and just kind of get a sense of actually what data what's how what's the slope of that relationship. And in our heads, it's usually we think it's one-to-one and it's I've not yet seen one-to-one. not take like lateral length. I mean, like all right, we're doing 3-mi laterals. Like those fracks are not as efficient at the toe as they are at the That's such a good point. Yeah. Well, or that we even when you look at the production, right? You don't when you 2-mi producer at year two does not have the same volume as two 1-milers. [laughter] So, we all know that. So, what happens then when you start having more wells in a unit or further and how you think about the trade and production trade-offs and with completions? But the economic trade-offs are huge, you know, like you know, on the especially on the drilling side, you know. But but the question then is how do you be getting your organization or the your subsurface teams uh just get calibrated on what those trade-offs are cuz the decisions are all about trade-offs. And so, turning things into trade-offs instead of models. And models can be useful for figuring that all out, but I find that's a place where all the excitement about machine learning people are getting a little bit lost and forgetting what the point is. And I think you can do a lot more you can do a lot of damage with simple models focused on things you actually care about like cash flow. And I don't care what the curve looks like on my production. I just want to see what are the wells that we've pulled and what's been their correlation to outcomes that I care about. Yeah. And you can often learn an off an a lot um like in DJ, we see uh a rem for a large area of the field, there's a remarkably s- almost insensitivity to uh frack size. So, wells have doubled their frack size for similar spacing and uh they seem to have no impact on there's no on interference. Yeah. And so, you well, that's not the model I had in my head. Yeah. [laughter] I think that that's something in each basin. I've looked at it in some others and it's and you start seeing more relationships and in different basins and looking understanding that and being able to to get calibrated and sort of instead of just saying I I have a belief that such and such ought to be the case. Yeah. That's great, but then there's the real world. got to prove it. so much data now. You just kind of finding creative ways to bring it together to I I call it feature engineering, but figuring out what are the metrics that you can use that help tell us help you build a story in your head about what's happening and then trying to figure out what are what do you believe that's wrong that we can change how we assume our field should be developed. And I think that's one thing I would kind of always go back to is people lose the heart of like the words data science. Like when we we all learn scientific method. My daughter just learned it in third grade, you know, but you have a hypothesis. But are you trying to prove your hypothesis, you know, correct, you know, or are you actually going to test it and be willing to accept if your hypothesis is wrong? It's I like I mean I was really disappointed or just sad when Daniel Kahneman died this year. And I, from him and his work and sort of people that followed on from him, I've become much more appreciative of those, the kind of the problems in our, in the human operating system, sort of the way our brain works where, when faced with a challenge, we kind of go straight to validating our priors as opposed to proving ourselves wrong. Or Philip Tetlock, he did the wrote the book Superforecasters and he ran these forecasting tournaments and I think he had a line that um beliefs are hypotheses to be tested, not treasures to be guarded. And I think we in oil and gas, I think people nod their heads and believe that, but then as organizations you could look at all the white papers. All of those are guarded treasures put on a pedestal as opposed to hey, here's something I believed, it's wrong. And here's how I showed it was wrong. I learned something. We don't talk about learning, we talk about pride and how much we how right we always were all along. That's I mean that's part of the problem, right? Is that that's that's essentially become a uh [snorts] You know, it's a it's an award, right? It's a prestigious thing to be whatever of the year or author or And I'm not trying to like knock on like I think those papers are super useful, but as a metric we or a thing we all we don't have like in academic publishing there is um they're annoying sometimes, but there's an incentive for you can build a career by just tearing down other people's papers, but that create there's an incentive to tear things apart and if you put something out and some whole bunch of people it's a lot easier to make a career tearing down old ideas than coming up with new ones, which is fine and it's helpful cuz it helps us hone in on where the problems are. We don't really have that as much in the technical papers. We kind of look much more for saying how great we were. It's like oh, I had this idea about frac hits and I [clears throat] then I tested it out and guess what? It worked perfectly because my data shows exactly what I wanted it to, right? Yeah, it's like if you were that great in 2012 then why am I going back and re-fracing these [laughter] Kirsten Wallace, a guy I worked with who was uh uh a lot of fun and very intelligent, wrote a paper and they just went back looking to see how often was the scientific method like followed in a bunch of papers, and what they basic conclusion for was uh very rarely, and that usually what happens is somebody collects some data, comes up with a hypothesis from looking at data, and then tests that hypothesis against the data that they got the theory from in the first place and found, oh my god, it's there. [laughter] Like that's that's not a you haven't really there's nothing robust about that process. You really ideated something and then walked and then walked away and we just stop at this I came up with a concept as opposed to getting rigorous about saying what does it mean? Is it a good explanation? Does it predict something broader than the system I'm just looking at? Does it apply to a broader situation than just this one ad hoc case? might work in one area of your basin and it does not generalize at all or else, you know. I think this is where, like, the science engineering side, kind of, there's a balance. And looking at how we approach machine learning and just the amount of data we have now, this is a funny thing about where there is a little bit of a difference, cuz the engineering world is much more ad hoc, solve problems, and keep practical. And then the science side is much more come up with a theory. Um so it works in theory or it works in practice but doesn't work in theory. And that's a good caricature, but as the systems get more complicated, there is some value in trying to keep grounded in what beliefs do we have that we can update and change by evolving our theory. Yeah. Um and I think that's where the amount of data we have and what we use it for and how we connect it to where the real purpose of our business is, what the value is that we generate, where the cash flow comes from. To me, that's the exciting part of what we can do with it. Yeah, and then there's also are you willing to push those limits? I mean, you talk about like say a a Pareto curve or something like that. Like you you have to ex- you know, go to the extremes to understand you what that's working and it probably falls in the middle, but you have to push the boundaries to to understand. Right. You've got to [snorts] establish boundaries in order to know what they are. Uh or you've got to be willing to wait for someone else to do that cuz you don't want to spend the money to do it. But then you're going to take the risk of being behind the curve Yeah. on some of that. I've got a question. So as a non-operator guy, how do operators since both of them are on the operations side, how do operators encourage more

tinkering experimental experiments? Let's just call them experiments, right? How do operators encourage it more experiments in a controlled way? Scientific method, right? Cuz I've also seen that where it's like Shell did that. And then you go for it. It was like we did a pad in every single every well design was different, every frac design was different on each of those wells. It was like, what are you all learning here? You're learning that if you change [ __ ] up, it's going to be different. But you don't you didn't control like there wasn't a constant across any of them. Like that's a terrible experiment. Uh-huh. Yeah. Yeah, that's what most of them all did. Yeah, I mean, I mean, Conoco did similar when I was there and and on the fringe of when I left. Um but I think they were trying to push the limits of like what you know, the water you know, proppant ratios and just playing with all those things and and rates and then but then but then like did you flow them all back the same way? Yeah. You know. No, absolutely. Or should you cuz if I did, you know, high sand water ratio, I should probably flow it back a lot harder than if it was a lot of water, you know, but but again there's so many different things that impact the quality of of a well. Yeah. Well, so but that's what that's what I'm getting at, right? Is that like there's all these variables to play with as an engineer. I want to go twist and turn knobs and play with things and change [ __ ] up, but then there's also the business, right? So like we have to make money, we have to keep lights on, we have to hedge, we have to do all these things. It has to be educated guesses. I mean like I mean you you've got a bunch of experts. I mean like we know we've got parts, you know, the geology. We know this is a thinner zone, so let's try to do something that we think is going to you know, be a better strategy here, but understand that it's not going to generalize. We can't just paint with a broad brush, but you probably have to be just be more, you know, let's try this in a specific place or where it's a less if it's a fringier piece of our acreage. We already know it it might suck anyways. So let's try something else here and that may that's where you actually get the value that no one else is willing to to go after. Yeah, I think that there's a little bit of a a met

a challenge of getting I think I just lost my train of thought completely. Um Sorry.

Have a pause in your podcast. We can we can Yeah, I mean I just completely lost my Yeah, I will edit them out. It's uh We were talking about the the challenges of experimentation Oh, so when you do the experiment at some point you have to you can do lots of tests. And so when we do a test, what are we looking for? We're usually looking for or my experience is what we what I tend to find teams are look are expecting to see is something either an EUR difference or a some production signal. But as you pointed out, you have your completions team, you have where you where your geologist put your your wells, you have your development intensity, you have your frac details of your frac, details of your flow back, details of your drawdown, your facilities, did you have any pipeline issues? And so now you have your volume on surface constraints. [snorts] That's too much. Yeah. If you're unless you did a colossal test, you said, "Oh, what happens if I don't produce this well? Do I get anything? What happens [laughter] if I don't frac it?" Like you need to do the the knob you have to pull turn is so far. The lever you have to pull so hard to be able to resolve a signal through all of that. And we're trying to optimize on much more minute signals. Yeah. And until you have something that you're chasing where there's you're coming out at with some sort of theory about what's the change you expect to see that will validate or invalidate the theory you came in with. If you don't do that, I think you're in trouble [clears throat]. And you're talking about like the scale, that's like design of experiment type things, which few companies do, lots of people talk about. Um and I think this is a place where the like lean manufacturing concepts come in, which is I mean design of experiment comes in there, too. But the how you how you can allow tinkering with purpose, and that's I think Toyota, and um, kind of Amazon has their own approach; but there are companies that, the companies that have been able to make this work, it seems like that's a template that we can learn at least elements from for how to do it and actually get something, as opposed to, you know, your completions team try something but doesn't tell anybody, and then three other people do experiments, and you have them all come together, and you never learn anything. Yeah, no, I think that's I would be very curious to know how often that's happening, right? Where it's the I mean we're mostly an industry of wildcatters and if you put an engineer in a box for too long, he'll figure out a way to go around the box and still solve his problem. So, I would be very curious to know how many forgiveness instead of permission projects are out there A load. currently running at the same time without Well, they don't even have to be forgiveness. They could also be um um you know, a manager walks in and says, "Hey, I want you to try X." And at some some somebody says, "You know what? I'm I can't fight you." Like here we go. I got other I got too many other things to deal with. Sure, we'll do that. And that I I would say how much value is destroyed in bad experiments that are just all you're doing is adding noise to your production. the manager coming in and like, oh yeah, we've got that that bonus thing that we need to check off the list with AI this this year. Let's go ahead and do something with it so we all get the same company said they'd give us, you know, come on another thousand more pounds. Let's see what happens if I throw that in there. Sure. Sure. [laughter] I'm curious. Uh, I think the other thing that you also pointed out is it's like we go in trying to prove our basically prove a hypothesis that we already have. But I think the other part of that is, like, everyone should accept when it's not right and publish that, and like, that's where learning actually happens, right? I I I say it a lot. I think it'd be such a fun conference. No, I mean, well, because like an engineering failures conference. That's a great idea. Well, I mean, I mean, I've seen those in in the data world. Like they have like the the happy hours where people just come and say it, this is how I really [ __ ] up. This is how not to do this. This is how I shut people out of prod for two hours, you know. Yeah, but like you actually learn from that. Well, like I think about that [clears throat] downhole long-term downhole gauge job that we had where both of the gauges read the same exact thing and we had them recalibrated and everything and the operator was like, well, there's no way that's right. So, we had two independent sources verify this at the same time. They didn't agree with what it showed, so then we sent the gauges off to the manufacturer to make sure that they were calibrated and certified and everything. They were fine. But the gauges were wrong, not their hypothesis was wrong and this was actually telling them something valuable and useful that they just didn't think about. My well's not in a vacuum. What are you talking about? Yeah, right. Well's not going in a vacuum. I think it's fair. This is where I why I think [snorts] like Kahneman is so important is we I think a just too few people understand kind of what he was the work that he's done to help us understand how our brains really work and how human it is to make that mistake. Right. And it's not like there's bad people. It's that the expectation is unreasonable that their default case is going to be not is going to be to pursue truth. And I think that's where like leadership really comes in and in all the all these industries as they get more and more complicated um multi-faceted with lots of swarming us with data. Um learning to uh organize your teams and to reward people and incentivize and sort of have your raw raw go team moments around uh discoveries and disabusing yourself of falsehoods that you just believed. And every time you discover something new to be like, yes, we have just stepped one step away from the world the model of the world that we had that was wrong and moving towards more truth truthful understanding. And I think that's um cuz if you don't make that uh something that you can rally around, then it's it feels bad. You say, "Oh, I was wrong." Right. I don't like that feeling. I don't want to do I don't want to tell anybody I was wrong. But if you say, "I discovered something. I I now know more than I did yesterday." And how small of a thing can I learn from? And I think building organizations around that, and looking for organizations that do that um, whether it's in tech, like the tech industries, or in oil and gas, or anywhere in manufacturing that incentive is seems to be one of the biggest, the most rare things to see that's critical to getting anything out of AI, machine learning, or data. Well, then a failure at an operator and at least from my uh friends and feedback that I've gotten essentially results in well, you're not going to get a budget for anything ever again. Right? So, like people are terrified of failing at their you know, working at whatever company they're working at. And so, it's like well, that's a horrible culture. Right? Like that doesn't work very well if everybody is just incentivized on agreeing with the same thing. And it's like the older I get, the more I realize that you know, just the hubris of humans is just ridiculous that we think we know so much and it's like we don't know [ __ ] about [ __ ] that we use every day, right? Like you can be get simpler. You can just say 10 years ago uh looking back today at what we were doing we didn't understand a whole lot of things. Mhm. Uh we couldn't achieve the rates that we wanted. We couldn't We didn't understand how many wells to put in. We barely understood tar Like we had Yeah. very little idea 10 years ago we were telling how much better we gotten in 4 years. Yeah, we were we were so proud of ourselves then. Yep. And then 10 years on we are at a completely different place now. Mhm. I bet in 10 years we're going to be in a There's a whole bunch of things we believe today that are just hairbrained. And in 10 years we're going to look back and say, "Well, we were just had no idea. We were idiots." Except it's us. Right. And so you can choose today like how do you how do you get your people? How do you how fast can you make that evolution and and move up that learning curve as fast as you can? And whether that's collaborating with others and sharing data or incentivizing your team or chasing the right KPIs. I think those are all things that we we get proud about and then sit on our laurels and that's the slowest way to get there. Yeah. Yeah. Yeah, I think one thing I want to kind of bring up I have kind of two thoughts on this and then I would do want to pivot into like more like the nitty-gritty kind of more technical side like with Python and all that [ __ ] But um I think one thing here too though like we're talking about a lot about science. And people I think kind of forget in a way engineers are not scientists except for Bill Nye the Science Guy. All right, but like an engineer applies the science, right? Like a chemical engineer is not a chemist necessarily. I mean like they they understand similar domains, but like chemical engineer's job is to productionize the science that was probably done in in a lab essentially, right? Yeah. So they're not probably the people to be performing the experience or the experiment rather. Um and with that I mean I think you kind of hit a little bit like some of the biggest discoveries we've had whether it's medical or otherwise have been accidental. You know, we found penicillin cuz he left it in the sink too long, right? I mean like there's stuff like that. I mean to say even like refracts and like the different kind of refracts that we have where they're you know, defensive or offensive. I remember Conoco we're like, "Huh, this well did better." and it was next to a refract. You know, we we didn't do it intentionally at first. But then we we then we saw the data and we're like, "Wait, why is this one doing so good?" Well, we put a refract here and now [ __ ] there's not a pressure sink and you know, and like maybe we should start doing this. We should start pairing this with our new development. Well, I know where where So, one of like on the medical connection, um you know, statins? Mhm. Yeah. So, stat-the I believe it's statins. There's a Nobel Prize for the woman that discovered statins and they come from uh I believe this is right. Um uh and she basically found it in a Chinese ancient Chinese uh herbal medicine Okay. book. And was able to figure out that she doesn't have any kind of medical training. She's not a doctor or anything and she has [snorts] a Nobel Prize for this drug that we all that gets used all the time. And it's this [snorts] What could I learn from something that was old old information that was recorded and they didn't have a full understanding of what they were dealing with. But the hints are there and if I can unpack it and connect it to my problems today, Yeah. [clears throat] huge value. How many of us spend much time looking at conventional fields anymore? Brian McDowell. Not many of us. Brian McDowell. Yeah, right. So, there's a lot of, there's a lot of stuff that we've kind of, there's, there's skills and, uh, information sources or just, um, knowledge about oil and gas we've kind of, we've been [clears throat] moving away from. And what we can go back to and take? What have we learned in the past that we can apply to the problems of today? And there's so much to be gleaned from that Yeah, we always say the best place to find oil oil is where it's been found before, right? Yeah. But we won't use the learnings Right. Yeah, we got to we got to go over the learning curve from start from the scratch and it's Perfect flush volume is a perfect example of that, right? Like we when I first started fracking, we over flushed every well after the stage and then I went to work for a different frac company and we mostly did conventional fracs and we you never ever flush a conventional frac because it's got a lot more permeability. And then I'm sitting there like, "Well, huh, I literally only knew to overflush a well and then I go into the conventional world and we you will get your ass chewed out if you overflush by half a barrel. And it's like well, [ __ ] I wonder if we're overflushing all these unconventionals. Like they have even less perm. Are we just moving all the perm even further away?

But no one, you know, it's just like these yeah, no one learns lessons from the past and rules rules forward, I think. Yeah. So you applied science manager at Everad. What what does that mean and as far as like especially on like the technical side cuz like you know, I think the one thing, you know, that John set and I set out for this podcast was to share whether it's different tools or processes that people are using. So just kind of curious like what does your data stack look like and then what does it enable, you know, at Everad? So there's been well, one I can kind of connect it to to give you a bridge. The you know, you're talking about the engineers not being scientists and we have all this information. I think applying sort of scientific thinking to the the wealth of data that's out there. And a lot of what I've done has been pulling using public data and trying to build features and useful information out of that massive data. And so little things which are, you know, anybody can calculate on things like well spacing or and well characteristics or production characteristics and trying to build sort of turn things into a vector and then figuring out how does what's the matrix look like that converts that to explaining six-month cumings, 12-month cumings and trying to then understand trade-offs. But I think the biggest part of all of that is you can spend a lot of time and you could build complicated models and I'm not a generally a fan of complex models. But you've got to find a way to communicate it and connect it to the decision and that's really where all of my attention goes. So things like mining surveys and getting

and uh clustering pads and finding ways to kind of automatically do a lot of that, and automatically doing taking care of regularly pretty simple problems like petrophysical mapping. Um, so that when those things are automated, it becomes you can do more more involved discussions and get into more involved discussions about the trade-offs and how it relates with the goals of your of your asset. Um, but for tools, uh, I mean, we almost all most of what I'm doing is Spotfire and Python scripts. Okay. Um, and data functions within there or just Python scripts outside to kind of move data around and then Yeah. So, I we're on like a really old version of Spotfire. Uh, and so I'm most of it's doing this the Python and SQL type stuff are you on if you don't mind me asking? Uh, I don't even remember. Uh, I What are we at now on on 14? Yeah, so we might be on 11. Okay. Um, We were I mean, we were on 11 or 12 not Yeah, we went I think we went from 11 to 14 or something. I think we just we're kind of on a a really basic version. Um, but it works for us and there's some reasons for why I know you're about like 7 and 1/2 That was the one I ever I would hated going from like 7 to 10 No, we're I think we're at a little uh, I think we're past that. Okay. Um, and we don't really as a company we're not like a we don't have a There's a couple of people that are um, Spotfire users. Um, and we're How big are y'all? I don't I don't even have much Um, we're for people I I we're around 100, I think. Okay. Um,

so we're a lot of it is like using it, you know, everybody lives by PowerPoint. So, it's building PowerPoints and trying [snorts] to show people data relationships and building them into sort of as we go through and do like pad reviews and planning our development, trying to connect how what our expectations are based on like So, well spacing comes up a lot. Um, how what kind of development intensity do do we want? Um, try One that came off that was um, in the last couple of months was uh, we had done actually a little bit what you're talking about. We for reasons we'd kind of changed some of our um perf design. But we were getting some suppr-Sorry, we just get season production results that we didn't fully understand and weren't really expecting. And trying to figure out to what extent does do the perf or other things explain uh is that expl-does that explain what kind of correlation is there between that and the results that we're Mhm. And I find one tool that's been really interesting that I've used a couple of times has been to build like really basic models like a random forest and regressing a suite of parameters and spending a little time trying to figure out which of these are non-correlated. How can I build some non-correlated features, which I think is something we we don't spend enough time on. Um and once you get your feature set, looking at run it through a random forest to forecast the outcome you care about and then applying sort of a Shapley analysis to understand the relationship of how What is a a change of some amount? What kind of effect is it is the model saying that will manifest in the outcome? And so it gives you a a mechanism to start saying,"Oh, if you increase your um your stage length by 50 ft, that I all things being equal I expect that to have a impact of dropping your 12-month [ __ ] per foot by uh 2 and 1/2 barrels per foot."And you could even say for different lateral lengths maybe that that What's that What does that mean? And it gives you a it's a language where you don't have to spend a lot of time talking about the the model. You're just kind of it's letting you speak about the entire strength non-linear relationships. And with some charts and diagrams that are not too that they're fairly intuitive. Um because the real That's one of the challenges I think every the really advanced like data folks run into or the modeling people will run into is how do you communicate and build intuition about something that's really Yeah. Something that's really that's really complicated and non-linear when our brains at the end of the day, we are we think about we understand linear. Once you go multivariate non-linear, we just most people say black box. Yeah. Just because we can't keep up. Yeah. Um But you know, like random forest is nice cuz a lot of times you can break it out and you can see the decision tree that Yeah, you could definitely do that. Um Even just basic stuff there, right? Like I mean, as an engineer, especially with how bad public data has gotten these days, it's like I want to see a root of the data and how we got to whatever you're showing me. Mhm. Because I just need to know, you know, like as [snorts] an engineer, like so I it and then I get frustrated at the same time when Colin's like,"Okay, hey, tell me show me how you came to that in this Power BI plot."I'm like,"Damn it." [laughter] It's like,"Okay." Yeah. So [snorts] it's tricky, right? Cuz it's you want to give it cuz it mostly engineers, we want to know the why, regardless of what it is. Yeah. But then it's like, okay, there's a lot of complicated [ __ ] going on here. And sometimes we don't even know the why, we just know this is how it worked and this is what popped up. Well, that's what I find um I came So my master's was tectonic modeling and structural geology stuff. Um so I love that world, but when we start talking about sort of completions modeling and talking about the details of what's going on 50 ft outside of the well bore. How are you going to validate that? Like that's really tough a tough space to live in. No, I completely agree. I'm uh I love seeing a lot of people what people are doing, but my skepticism goes through the roof and and I get concerned that we start then saying, "Well, what's uh have what kind of fracture intensity have I generated out here?"And actually nobody cares. Yeah. What you care about is what kind of like what's my return on investment? And I think you can go to simpler metrics and that's what I find is useful and you and communicates well. And you could have a technical team that could really understand a lot of those details to to get to the why. Why is this affecting the the realized returns that I'm getting on my on my frac, my investment. Um but the the models that give you high detail there it's hard to validate. I don't really know how that's large companies that probably has more use for. For us, if where we are cash flow focused and we're that's just not I try and help people see um sort of the the thinking of the tornado chart of how important are different features. And to be able to see, oh, petrophysics is really low on the list. Frac size is really high. I love petrophysics. I love geochemistry. And then I'm coming in and saying, these are really low order like in DJ petrophysics is a low uh at least as I've seen it fairly consistently as a low impact as far as explaining the variance in production. And one reason for that is not because petrophysics doesn't matter. Yeah. It's because the variance in the petrophysical properties is low compared to variance in all the other stuff. And I think that's um that's the place where I'm constantly trying to bring I would say bring calibration so you guys know what's the role of the applied science and engineering. a statistical property, right? It's just a property. And and but trying to keep focused so as a company we don't get up lost in, you know, hiring more and more complex answers to what's the petrophysical characterization of what we're looking at unless you can come up with a completely different angle to approach it than what we're doing. It's not really worth spending the time on. Now, you want to talk about completions? Lateral length or things like that, those those matter a lot. And we can spend a lot of Those are things that deserve more attention or proppant intensity. And there's uh fluid is a really big one, but being able to put the attention into the stuff that's actually going to give you some value and helping the the company keep focused on that especially when, you know, cuz we all have friends and somebody whispers in your ear and suddenly it's it, you know, you find that there is a uh a trend of the day that everybody's kind of wanting to spend a lot of effort on to. And if it's something that you can keep showing in your data is not the best use of your attention. To keep that focus on the stuff that matters is that's probably the That's a lot of what I end up trying, or I'm often That's the war I'm constantly trying to fight. No, sure. And so much of it comes back to that 80/20 rule, right? I mean, like It's spot-on. I mean, 80%of the production is coming from 20%of the you know, the things, but if you now you there's a lot of world to dig into in those that 20%and that's worth way more of your time. Yeah, I mean, like I said, you've got so many knobs to to turn. Just taking those from 100 down to 20 makes it a little bit easier to start figuring out game plans and how to do an actual scientific method. done the analysis, so I can't say this is true. Maybe that says how how my brain works, but I suspect it's not that the paint color on your pipes probably doesn't matter. Yeah. [laughter] Nobody would. No team's going to spend a whole lot of time analyzing the the density of the paint on their on their pipes or on their tanks, but uh, and I think that there's a number of things like I say that as cuz it's a joke and nobody ever would agree but if you find that there's other things that are of a similar importance to that that are familiar metrics to chase, those are the kinds of things where I find it really fun to be able to say,"This is like, this is where we're hitting a threshold of things like below this level,"or, "Here's the tier of things to not pay attention to." And in fact, I think of a one thing that I, one thing that I've found useful, um, instead of trying to tell people, "Here's the optimum that we should go for "cuz actually that's a really hard answer to give is [snorts] much more to say,"Here is the the operating envelope to stay out of." Yeah. And inside of this play to your heart's content cuz I don't know what the right answer is and this is where you should play and this is where your completions engineer should bring their expertise to work. This is where um my production engineers can really actually affect some changes. That's where engineers live, right? Absolutely. All my education is around figuring out formulas with bounding conditions, right? Like whether it's physics or drilling like algorithms can really help you or data analysis can help you figure out where what to avoid and just kind of put some areas that if you were going to walk into this space, you better have a damn good reason. Because everything we've seen so far says don't This is where the dragons are and just stay away from the dragons. And the return on your investment of time is very low. Like Or the expected value starts getting negative cuz you're walking into something where the downside tail starts getting colossal. Yeah. Uh another person who just died uh uh uh Charlie Munger had a lot has a line about him and uh uh at Berkshire Hathaway and saying how much money It's amazing how much money or how successful we've been uh just by looking around and seeing what doesn't work and then not doing that. Yeah. And I think there's a lot to be taken from that in oil and gas. We really want to chase and we're using algorithms to optimize. And optimize is hard. Yeah. But figuring out what doesn't work is so much easier. And so [snorts] to go and look at your data sets and say,"All right, what's what's the terrible ways that people have operated?"You have a loads of examples to work with cuz we as we're talking about 10 years ago, 20 years ago, we made loads of mistakes. So And they were educated mistakes. We we went into them with intending for them to work well. We thought we were doing a good job, but given technology available to us today, we can do better. And to be be able to use that to rule out areas or also to say come in and you start working with your engineers and saying here's an area where something a data space that performed really poorly. And they say,"Oh yeah, it's cuz they were running gel with all of that while they did it." Yeah. Uh if you remove the gel, maybe everything else works and you can start coming up with theories of things that are worth testing. That's how we came up with today's modern practice. No one's running gel anymore. Yeah, right. But No, I mean like we were running 30-lb, 20-lb crossing jobs in the Fayetteville. Like why? Why? Sliding sleeves and everything else. Packers Plus, man. It's No, I think that that's where my next question was going to be is like engineers have the I am I'm an engineer, so I take full responsibility for this. We have this ability to get really deep in the weeds on [ __ ] that probably doesn't matter most of the time. And so I think that's another thing where it's like looking using data to help drive your focus because a lot of the times it's like frac hits for example, right? Like people were doing all these [ __ ] networks and these meshes and importing every single natural fracture from the frac law. Like they were going to insane detail importing that into the patrol models and the grids and all that. Right? Like [laughter] yeah, and then you had SRVs and all this stuff and then it's like and then So what? Right? So it's like nothing really came about that and then you had the Devon FDI stuff where it's like put a pressure gauge on offset well and you can see the communication and then you can literally calculate the travel time and all of this stuff and it's like, hey the simplest answer actually was one of the most effective ones potentially, but engineers, myself included, don't necessarily always look at that. Oh, we have the simplest answer, right? We want the complicated one. So there's a number of people, I mean, I've mentioned uh like Berkshire Hathaway and some of these other people. There's a lot of the investing world that I find really interesting because they deal with you have a trading floor of humans dealing with this flow of information that's assaulting them. Worse, every time you figure something out about the market, it starts to undermine and it that that that the market starts changing. A nice thing about reservoirs is they don't change once you learn something about them. So our job is actually a little bit easier. Or we can model all the pressure drops and all that stuff. Sure, but like if you know what the pressure drop is, it's not like the physics of fluids ex-fluid expansion doesn't change just because you studied it. Well, yeah, but we also were I mean, it took Dave Kramer's paper, I don't even know when that was, early mid-2000s, and that got rejected by SPE for whatever reason. Uh where he was talking about this case study they did in the Barnett. They went in existing field, right? They had producers on either side. They could not frac the well between them. That was a new infill well. So then they started looking at it. They're like,"Hey, what if we look at this as a pressure system like a reservoir is? And we see that oh, these have been on production for a year. They have a our sink. We're adding new pressure here, so all that fluid is probably going to those pressures. Let's shut them in. Let's let the pressures build up, and then they came back and they had no problems with it. And it's like the fact that we weren't even in you know, this is a number of issue reasons behind this, but like we weren't even looking at the fields as a whole, right? Like the engineers were looking at the well that they were working on and not anything around that well and whether that was going to have any impact on it or not. It's like it's such a simple thing, right? Like when we were talking about the FDI stuff, I I vividly remember distributed load from my statics and dynamics class in college where it's like I push a force on this table. If I know the dimensions and I can read the force from any point on that table, I can back out what that force was and I know where it originated from. It's like that's literally what we're doing with a frac in two in an offset well. You've got a pressure reader and then you've got a force being applied to the rock from over here and then it's just a distance and it's velocity, right? Like And so I was like, "Well, [ __ ] How do we not But in order to do that, though, you have to make some assumptions about You've got to make a lot of assumptions about the rock and that's like what you're saying like there's certain parts of the petrophysics side that are we just have to assume are constant enough, right? And then like the other things are the things that we can actually impact. permeability, right? Which you've also got I think you're like another part of what you're describing cuz we're talking about um good good explanation in understanding these complex systems is you're talking about a situation where somebody looked at it and said,"I have a problem."And then started to think about,"What's a theory that can explain the problem that we're seeing, and how do I condense it down to something that is, uh simple, as simple as possible but no simpler and has like broader application that's not just to the one situation I'm looking at, but actually is telling me something about the way the world works?" Yeah. Uh and that that sounds I mean, it sounds awesome. Like I I don't actually haven't I don't think I've read that paper, but it's that's the right a healthy approach to how to study something that we don't do enough of or forget. I think we get lost in like trying to keep our execution hat on and making sure we keep moving forward. That's a an awesome example of actually trying to learn something. Yeah. No, we we just don't do it enough. That's the the tricky like it it's how do you how do you incentivize that while also still being a profitable business, right? Cuz you can really screw that up really quickly. mention the investing side. So there is there are a number of myths about markets and our we have a fetish for complexity. It's kind of once we say I have my beliefs of what I expect the world to work like and as soon as I see something's not working how I expected we start adding layers on top of it so that we can say I was right all along. There was this new additional feature. Right, it's going to be in a vacuum. Yada yada yada you're now sacrificing virgins to make the sun come up because you don't want to release your like priors, this belief system that you've built. And I think getting building organizations and understanding how humans are that we we have trouble giving up on ideas and that we're just going to add new layers of complexity which will and your smartest people are the best at convincing themselves and convincing others that they've got the right answer with more and more complexity. And that's I think something as an industry we just don't have we have a lot of pride and we don't have actually a lot of awareness of how how human that problem is. And as the industry gets more and more sort of non-linear so you could think of it used to be I mean literally out in West sorry East Texas people would drill bumps, right? Like you go find a hill and you put a you sink a hole in the ground. That that's what you did. That's that's pretty straightforward. Hill, hole, oil, done. That's that works. Okay, so now then you're in unconventionals and you've got parent-child interactions [snorts], frac, drawdown rate, there's just so much going on. Horseshoe well, don't forget the horseshoe. Yeah, oh that's right. Well, and then all all the LCM you had to pump in while you were drilling. Like all sorts of stuff. It's now getting to a point where it's not linear and so we are just adding we don't have healthy systems in our organizations to um be able to be aware of whether we're adding layers of complexity that are helpful or if they just make us feel better.

This is not where I thought this conversation would go, but I'm really glad it did. It's been a really good conversation. Yeah, it's been really good. You know, we didn't get nearly all the technical stuff, but it's not like I mean I mean [snorts] outside of like cuz the only other thing I was curious about you talk about Python scripts, like how are you orchestrating those? Yeah, so I um I'm super basic [clears throat] uh on that and I'm just writing writing in idle uh for the most part and writing like one-off little tools um that work for me. Emacs or Vim or something like And yeah, I should. And then I and then I've been using GitHub a lot more. So that was a couple of years ago I started really trying to move stuff into GitHub. Um just because I was running into version chaos. Yeah. Uh and so then and also searching around at GitHub and getting ideas of what other people have done or trying to find scripts that people have written cuz there's nothing special about Python. No. Except that the libraries are amazing. libraries. Except that someone else has probably already done it. It's so amazing uh and what's out there. And so to go and find people that are not in oil and gas like uh Stumpy is one I want to play with a lot more for time series analysis. But going and looking at what other people have done elsewhere and trying to and How can I apply that looking for what are other applications sort of new angles on old problems that you can bring to oil and gas. Um I heard there's a a paper some guys I saw them present it in in Bell in a Denver and then they presented it uh UrsaTech this year on automated top picking using the same technology that's behind Shazam. So you know Shazam? It figures out your like what you're listening to. The finger the way it connect collects a a fingerprint of the audio and then matches that with uh a bash What's it called? Uh I Uh there's a special array and I can't remember what it's called. A hash table that and how it's using it to to figure out like what it's listening to and to quickly match that just to a different Uh it's basically taking the the spectra of time series and using to match. So they said well that well that works for logs. So let's treat log as a separate channel as an in the audio file and then start throwing that at this Wow. at at the the same technology and they built a tool to start doing that and they're kind of playing with it. And I love it. I think like the concept is remarkable and it's one of those it's a to me it's one of my favorite examples that I've come across recently where somebody's taking a totally different approach. And that's where the technology to me is so much fun is you can go see what's worked for somebody else in something that I can bring over and I have a completely different use for. I mean I don't want to get into too much specifics but company we used to work for they had a solution where could use you know taking sonar shots but then we're using a software that was developed like Cornell or something for like bird identifying birds. That's awesome. But again like to pick the I mean the what the hydrostatic head or the reservoir scale detection and all that. But I think I think that's where um the creative fun of where the oil and gas industry is right now. There's An, a wealth of technology, and there's loads of stuff that's just being thrown off, like at it's aging out or it's not useful for other places, that's we're going to start bringing into our industry and it's going to...we're going to be able to use and finding sort of these novel uses for old technology or things we can pull from other industries or other. I mean some, Pythons the code is so fun cuz you've got some random guy out in India who just wrote something for fun. Yeah. Bird call identification and suddenly you're use you've got a use case Mhm. You just have the risk of creating that XKCD car cartoon where it's everything's leaning on the work of some random guy out in in Nebraska. But aside from that it's it it's amazing what you can how fast you can get something off the ground and start just playing around. Yep. And you start when you start unlocking and realizing that there's a correlation that's connected to the value proposition of your business it's it's really fun. I think the Shazam thing is kind of funny that too cuz I think there's a story and I I not confirmed but I'm pretty sure it was oil and gas and I think it's like CEO or executive called someone in and like, "Hey, I've got an idea for an app." And it was like, you know, if my phone could hear a song and it could tell me who sings it. Like, yeah, it's called Shazam. Yeah. So, like and I'm pretty sure if that was in oil and gas, it's just funny that it's come full circle and now it's like we're utilizing like, you know, this Shazam algorithm to actually do something useful in oil and gas [laughter], but Yeah, totally. That's a good way to to wrap this up. You want to do a speed round real quick? Yeah, just couple couple uh [snorts] some some books. What are What are some good books that folks out there on the data and tech side should read that you'd recommend? Cuz I know you read a lot. So, uh Team of Teams was recommended to me by Alex Fleming and it is a a look at the uh US war against Al-Qaeda in Iraq. And the connection to oil and gas is an understanding of how to work in dynamic and complex environments. So, where the data flow is so fast and the environment is moving so quick that a top-down sort of decision structure will slow you down and stops you from being able to to operate. And that's how uh Al-Qaeda was able to basically be defeating the US in Iraq and yet they were wildly under-resourced compared to the US military. And once you switch, the same story actually works for the the German blitzkrieg in World War II, but it's the ability to give as a management uh concept to push uh decisions. So, strategy can come from the top, but decisions should come from the bottom. Yeah. And trying to give you want your pumpers go out to the wellhead and be have the ability to make decisions that are aligned with your strategy without having to with as much authority and to do what they need to be what needs to be done at the site of action as fast as as much as possible. So, I like that book a lot. Uh another fun one that I have two kids and uh Range uh by David Epstein. Yeah, the um generalist rule. Yeah. So, that's that really gets One of my favorite concepts from that is the uh that the world and we're not playing the world is not a chess is not a metaphor for the world. Yeah. Um it's just too structured. What we're actually doing is we're playing Martian tennis. Where you can see that there is a there's a court and there's some balls and rackets, but you don't know what the rules are and they can change at any moment without notice. And the the kind of concept of the book is about the role of generalists sort of as a counter to the Malcolm Gladwell's 10,000 hours. yeah. Um and for kids it's really gives you a lot of license to calm down and not worry about you're not a you're specializing right now because it's not going to end their life [snorts] if they don't. Um Yeah, that is a very good book. And then, I've mentioned Michael Mauboussin and so his book, Success Equation, which is kind of a rough I don't like the title because it sounds like it's some self-help book but it [snorts] really gets into sort of how saber metrics could work and how it works in from an investing standpoint and business and thinking about what is what are useful, useful metrics. Um, and how how quantitatively, sort of the math behind how you do that, and then as a as he's approaching it from an investment standpoint, using it to design portfolios, which is much the same thing as saber metrics constructing a team and I think all of those apply really easily to oil and gas and are concepts that we're not really familiar with. Um I just been more time with. Awesome. Some good good recommendations there. Um I will still last one favorite social media follows. Jim O'Shaughnessy on Twitter is like the happiest guy on Twitter, and he is Yeah, he's I like him a lot, and he...I find him, he's just gleeful and has good gifts, but also really good at thinking about investing and connecting with people that are in in that world who are thinking creatively about building businesses and using technology without being too proud about their own knowledge. Um He's my favorite. I'm not huge on social media. Um And I've gotten of it more and more of late. Kids are related to that. Yeah. Uh so, I mean I got to stop with the It's all pretty good. It's, you know, go touch grass with your kids. It's probably better. Yeah. Yeah. Better than just doom scrolling and adjusting just toxic Yeah. [snorts] It happens. content, but um well, awesome. Well, really glad we got you in. I know, you know, you're not always in Houston, so it was No, this is awesome. It was really good to come in and see you guys. Yeah, absolutely. Thanks for having me in. Thanks for coming, man. If you guys is uh like what we're doing, please make sure to give us a like, go subscribe to the channel somewhere around here. And uh we'll see y'all next time. Appreciate