Subscribe
Share
Share
Embed
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.
0:00 And we're back for another kickass episode of injury bites. I like the energy. I'm Bobby Neelen. I got my co-host, the rad dad, John Calfan here with us. How are we doing? How are we doing?
0:12 And happy to have Steven Barrow from Petronas entry, the principal consultant. So let's go. Thanks for joining us all the way from Bentonville. That's right. That's right. Arkansas, Arkansas,
0:23 maybe. I'm getting more excited. Hey, it's opening, opening weekend football. Yeah. Talk to him on Monday. We'll see how we're playing We're playing some AM school that I've never heard of. So
0:34 I think it's Alabama AM, so I'm happy about that one. I'm glad we're not kicking out. I did say a great meme from SEC Barstool about it was Taylor Swift and Travis Kelsey with her marriage or
0:45 engagement photo, but it was AM and eight and four.
0:50 Get ready, boys. It's coming again.
0:53 Here we go. I don't know which is worse being pissed off that you're eight and four or just being in the the doldrum slums of a Hofstra's gonna be undefeated this year. That's all I know so I'll
1:04 just laugh at all of you who care.
1:07 Oh Man, I'm excited. I'm excited for football We have a draft on Monday night for Fantasy League 2 and I have done exactly no research Yeah, so ever going into yeah, I would do that the other
1:17 night Like I'm in one through my wife's Some of her co-workers and I lay in there with mine put my son to bed at like 7 55 and I got the notification from ESPN So you're draft stars in five minutes
1:28 until my wife was like, oh crap. I don't remember my login. We were scrambling
1:34 It's all good. No, we we actually we had Steven on was to talk about it. Yes. Yeah
1:41 We're all old and married now, so we were even flirting with the idea of just auto draft everyone auto-drafting this year Just not have to worry about it. Oh Oh, yeah. See who's good on the waiver
1:51 wire? I mean, that's where that's where leagues are one. Let's be honest, beautiful. All right, well, Steven, I guess just
2:01 who are you? What do you do?
2:04 Yeah, that's a great question. These days I say I'm an AI developer in the oil and gas industry. That's usually my intro. And then if someone cares no more about that, they're like, you know,
2:15 we kind of go from there, but that's what I do. That's awesome Or Kimmy, is that right, it'd be why you? That'd be why you chemical engineering degree from there. I'm a fourth generation oil
2:29 field worker. Like my great-grandpa was building the refinery in Sumatra when Japan took it over during World War II. Oh, wow, that's wild. And then just since then, Texas raised, you know,
2:43 oil is kind of in the family and the blood. Great, Thanksgiving dinner conversations, you know.
2:49 and you know it. I went, got my engineering degree at BYU. And then my grandpa was, I was fixing to graduate, had some internships already with BP and Dow Chemical. And I was looking for some
3:02 career advice, so I asked him. And he said, well, you know, he started the old school oil field away, just as a, as a pumper pusher and kind of up that way. And he said, hey, these days
3:14 everyone wants to go to Houston and Denver and they never see the field, you know, so you've got to start out in the field. So that's what I did I just found the crustiest hiring manager at
3:25 Anadarko and went out and worked for him in Southwest Wyoming and saw where the rubber meets the road, cut my teeth in the field. Then from there, I went to Kingfisher Midstream. I don't know if
3:37 anyone on the podcast remembers that.
3:40 It was a great idea. I was tied to Alta Mesa resources and then I went bankrupt in 2018. So it was a fantastic story. Great people bummed out that it didn't work out. but we built a bunch of stuff
3:54 and then helped it through bankruptcy and then on the other side of bankruptcy is when I decided to leave there and start patronus energy. What was kind of your biggest motivation there to leave and
4:06 start it? Man, I just, at the time we had three kids and I was wanting to spend, we just had more control over what projects I was taking on and what I was learning as well as just more
4:19 flexibility and like how to, yeah, up until then I was just chasing the drill bit, you know, I was just always gone long hours and things like that and I just wanted to be with my family. We had
4:32 this,
4:34 we always talked about wanting to be nomads and travel and so I needed to find some sort of a gig that would let me do that and being a freelance engineer was the answer at the time. Yeah, sometimes
4:48 we got a great round gig, right? That's right. Yeah, sure. It's crazy on gig one. great advice from grandpa too. I think I genuinely don't think enough engineers get in the field.
5:01 It's just so much harder to understand the context of everything until you have experienced it firsthand and understand that, you know, it's like, yeah, this works perfectly on my laptop in an
5:11 office with perfect internet and, you know, it's a brand new laptop. Yeah, but stick on a map and then you get other like, oh, this goes down that far and out that far. Yeah, yeah. Yeah,
5:19 everything becomes a lot more real The stress, the, you know, constant bullshit that is always, there's always something. And so it's, you know, it's a good advice from grandpa. I think that's
5:32 just still today, not enough engineers end up in the field. It's probably more now than it used to be, but like right after I got out of the field, for example, prices, you know, we're just
5:42 running. And so people were just hiring engineers left and right, but like most of those engineers were just doing office stuff because they needed more people because they had so many rigs. They
5:53 were just pushing paper and it's like, I'm selling frack services to someone who's never been on a frack site before and doesn't understand what frack really is outside of a textbook. And then like,
6:03 this is dumb, yeah, it's crazy. But, so, I mean, because we kind of blew through, yeah, probably high school through, you know, starting your own thing, but like, you know, fourth
6:15 generation, so, 'cause I know a lot of people even if it was just, my dad was not, my dad was not, but if that person's, father was the one who I feel like, I'm never going in the oil, like,
6:24 but was that always, like, for you, you're like, I wanna do this, like, wanted to get into it? Yeah, a good question. Yeah, I don't know I, guess,
6:32 I didn't really know what I wanted to do when I was in high school. I had a knack for, I remember - I was thinking equations. Yeah, yeah, I just like, I was, you know, differential equations,
6:43 no. I think I had a knack for, like, taking stuff apart and putting it back together, like, I don't know, anytime like our lawnmower would be. die. And instead of throwing it out, I'd be like,
6:55 can I have that, you know, like, take it down to the crankshaft and then like, see if I can put it back together, you know? And so I graduated high school, I still didn't know. I mean, I
7:06 actually had a degree, or I got into BYU as a geologist, mostly because like, at the time, my GPA wasn't good enough to get into like BYU. But I found this obscure scholarship for geology. They
7:22 like really wanted to build the program. And so that was my angle.
7:26 So I applied in and got it. And so then they, you know, sympathy let me in. And then two years in, I love geology. But the thing, I just love to nerd out about like earthquakes and volcanoes.
7:41 And I didn't, I did an internship with a geophysical company. And I was like, yeah, I don't think I want to do what these guys do, you know No, it was like, ah, it was awful. So I love
7:53 geology. I just like, I hate the idea of what it takes to make money doing geology. Sure, yeah. No, that's a very valid statement. Yeah, if I go out and, you know, look at rocks and like
8:04 different formations in a while, that'd be cool. But like, yeah, and I still do like my, my kids are into rock climbing. So when we go, I'm always like, guys, guys, look at this For TRV, I
8:15 do the, I do the same rock. Yeah, like driving down I 10. Look at that out. That's a rock. Yeah, they're just like,
8:23 yeah, exactly. But they know the kids know they just got here for me. And, you know, for sure, make me feel like, you know, I'm saying something interesting. Yeah. Our rock advent calendar
8:35 did not go over very well with my kids this last year Where's the candy? What is this? They make that. They did. Nat Geo makes one. That's really, yeah.
8:47 She was just like, It's another rock. Yeah, 25 days of rocks, maybe. Yeah, here it comes. Get ready. That's how this works. So, yeah, two years in, I was like, yeah, I don't want to make
8:59 money as a geologist or geophysicist, so I switched to engineering. Nice. Chemical engineering, and then - I saw you were on the, like our president of the SPE at BYU. I didn't know there was an
9:10 SPE at BYU. There wasn't until I should go. Oh, okay, there you go Founder, founder, president, we started it.
9:19 I was like wanting to, I don't know, again, I didn't have the greatest GPA to get into BYU at the time. There wasn't that many oil and gas companies recruiting at BYU, it was Exxon. And Exxon
9:31 only wanted you if you were like a three, eight and above or something. So yeah, they weren't looking at me, you know. Getting here with tier lowly 30 GPA. That's right Yeah, we let this kid in,
9:44 you know. So, I started SB there because I saw it as an angle to like attract more companies, get something on the resume. That year was, you know, ATCE was in the Netherlands. And so, we,
10:01 yeah, we went, so we, there was petrol bull, I don't know if they still do it. Do they do it? Okay. We, we just like threw a team together and we competed in the regionals and won. We're like,
10:12 well, so all these like Utah, like, it's, isn't there a Utah mines college? Yeah. That was probably in there. Yeah. Exactly. I mean, like, so using for CTA is like way more resource heavy
10:24 and BYU historically hasn't been like energy in just, well, they're like coal, but not like oiling gas. And anyway, so we have just these like five BYU kids like don't really know only gas except
10:38 I'm the fourth generation guy who's like recruiting all these buddies and like, come I'm just like, study this. And so we, we just throw our names in the hats, we compete and we win. And then we
10:47 get invited to go. And so that's awesome. The Netherlands. That's cool. So it was rad. It was super fun. I loved it. Another slubbed on thing about our industry is just the, the opportunity
10:58 that is there to travel if you really want to, you know, whether it's through that or in the field, right? Like, you know, you did time in Wyoming. You were talking about it earlier. Your wife
11:09 asked if you got scammed when you pulled up to the field office. Cause it was just a tiny office in the middle of nowhere Exactly. Yeah, I was just like that story. It's like I was, she was used
11:18 to me working at, you know, like Dow Chemical. Yeah, big corporate. Yeah. She, I mean, if no one's been down to free port, it's just a, it's almost as big as, it's bigger than any Wyoming
11:29 city for sure. And it's all pipe. It's just like a big old city of plant. Yeah. And Baytown's the same way. And so she's used to me working at, you know, spots like that. And then we go get
11:40 this job offer in Wyoming we show up at this feet. Green River, Wyoming, field office. And it's just this like apartment building basically. And
11:50 she's like, babe, did we get scammed? Did you send yourself a security number already? 'Cause it's like,
11:56 are they already duplicating our identity? Yeah, yeah, yeah. Like, no, this is real, it's happening. So
12:04 then we live there for three years, three and a half years. And the other thing about Wyoming is like, great to start in the oil industry is, it's like, I know recently there's some hype in the
12:17 powder river basin, but like Southwest Wyoming, nobody cares, nobody's investing in that. And so it's old decrepit assets that you have zero money to operate and fix. And then every year you have
12:29 this winter where stuff gets down to like 40 below, zero is breaking and freezing. I remember, I went to my supervisor at the time. I had my field track and I was getting stuck all the time,
12:42 trying to go out to a compressor station. Said, Hey, is there room in the budget? Can I buy some snow tires? Said, Come back tomorrow, I'll have something for you. I was like, Cool. I come
12:51 back to the office and he gives me a snow shovel.
12:54 Oh boy. Thanks.
12:59 I thought for sure it was gonna be chains at least. Yeah, but a just snow shovel. It's crazy. We didn't have the budget for chains. Just a snow shovel and man. That's crazy. I was like, I was
13:09 eating 4, 000 calories a day just like hold my same body weight 'cause water below zero, even with like multiple layers of thermals. Yes, this just cuts right through you. Yeah. The wind just
13:20 blows and shh. Well, that's a whole 'nother ball game when you start getting into insulated everything and the pipe has to be diff, like everything has to be different, right? When I first broke
13:30 out in the Fayetteville, the first quilt tubing job I ever went on, it had just, it had like rained and sleeted the night before, then frozen overnight, and started snowing. And so you get out
13:42 the location and it's literally just a sheet of ice, like to the point that we're all cabbed up and stuff and they call over the radio, like don't walk across location if you don't have to right now,
13:52 because it's literally just a safety hazard. And so it's a, but like, you know, none of the Arkansas equipment was winterized at all. And so we're out there with warm burners following all the
14:03 fucking manifolds and all the pipe and shit. And it's just like, man, it sucks You know, and like that was, and it was maybe 10 degrees out. It's like it wasn't cold in relative scheme of things.
14:15 So it's just, I can't even fathom what that's like. And Colin has crazy stories about being up in Alaska and stuff and just like, you have to pre fall. You have like a fall area where you just put
14:25 equipment that just came in to fall out before you use it and all this crazy stuff. But I do want to talk about it because I, my first experience in the oil field was also in a very old and was in
14:35 the smackover down in South Arkansas outside of El Dorado and those wells were drilled in like the 30s and 40s and stuff. And, you know, all stripper wells, nobody has any budget, you're just
14:45 trying to squeeze another barrel or half barrel or, you know, you go from one to five barrels a day on a well and you're genius, right? Like, but you're spot on with it. Like that is by far the
14:57 best way to learn and those old guy, like the knowledge that those guys in the field have that run those operations is just like bar none, right? Like they figure shit out and they figure out how
15:07 to get it done because in most of those cases do, there's not a lot of, can't just run down to Granger and pick something up. You know, it's not like the supplies even are limited in most of those
15:18 areas. It's not well supplied with a bunch of oil field vendors and stuff, it's just a different world. Yeah, you can't like, it's different if, if you're going to go straight into like a design
15:31 office in Houston or Denver, I mean, you know, like go to your lunch and learn So it's not, I'm not knocking on them, but dude, go out in locations. 30 below zero, that's your lunch and learn,
15:42 man. Yeah, you'll learn way more easily. You're going to need two lunches. Yeah, not like, so we haven't even Talk about software or data, anything else. I mean, I did, is this something,
15:54 did you go for two around with it, say, high school or before, like, you know, with any of that? Or is this something like necessity and you've figured out in the last five years? Like, what
16:02 does that look like? Yeah, good question. So, like when I started Patronus Energy, I took on, I was just, my impetus was I just wanted to my own time schedule and do something. So I was like,
16:14 all right, I'm a, I've got a basket of random skills and experiences. Like, what do people care about? And everyone seemed to not want to do process safety. And so that was, you know, we
16:23 became a PSM company. We did audits and hasops and all the other random stuff with PSM and just kind of focused on that niche. But then when I always wanted, I loved tinkering with software and did
16:37 some coding and things, but I was never like a. software engineer or something like that, just like a closet, tinkerer, you know. When the tinkering start, when did you start messing with that?
16:50 Probably, when I was in a darco, like right after, well, I remember in college, we took a class that was like supposed to be a primer on like modeling softwares and things. And - Mad lab. Yeah,
17:03 like mad lab. I was just gonna say, yeah. And ISIS and whatever. Yeah, there's, I don't know You can only spend, I mean, there's so much offer out there. Like, what can you meaning, clearly?
17:14 Did you like to do a bunch of like CFD and fluid dynamic stuff with all the facilities in man? Yeah. That's heavy stuff too. That's not just like, oh yeah, put some noise in it. It's like, it's
17:24 funny 'cause like for years I've actually gone back and sat on the educational board at BYU for engineers. And it's always this topic of like, well, like engineers could go into like all these
17:34 different things. Like what are we supposed to do in this like 12 week class? So I don't like fault them. It's always this ongoing conversation about what kind of exposure do we give. But at the
17:44 time, there's like a one week section of that class that was all about a visual basic. And I remember sitting in that class and he was like, blue in my mind that I could like think of a thing that
17:57 I wanted to do. And with like lines of code, it would do it. And I remember being like, I need, I want to like dive into this And I just, I don't know, career and moving around, it like never
18:11 gave me time to do it. I was just chasing projects. And when I started experimenting with it a bit when I started Petronas, just because I was a freelancer, I wanted to leverage my time. Yeah,
18:24 more efficient. And it's like more efficient. But then when like Chi Chi PT came out, beginning of 23,
18:31 no. 21. 21, yeah. I became an early adopter of that. Again, tinkering with that and then we came out with the API access. I was like, this is amazing. It was just like kind of accelerated
18:45 what I was doing to, to learn and develop. And then, um, I mean, everything was kind of project based, project driven. I was, I was like, what do I wish existed that doesn't exist. And I
18:58 would just try to build that thing. And then once I did that, then I tried to build another thing and another thing. Yeah So that's awesome. What, uh, when you were at Amdarko, what, what
19:08 languages were you playing around with it? Was it like Python or R? Python. Yeah. Like, uh, some are, but it was mostly Python. And I remember trying to, we were in the midstream system.
19:22 There's pigging, right? We're trying to like pick lines and I was trying to come up with a machine learning. So like, uh, using Python and Python libraries to do a machine learning algorithm on
19:33 like when to pig and that was the whole. thing I was trying to do. Smartpigging, basically. Yeah, no, that's wild, 'cause there's a bunch of people working on that. Like with Smartpigs, we've
19:44 had to have all kinds of sensors and shit in 'em and all that stuff now, yeah. I geek out about it. Yeah, so how did that go? Well, you know, the company wasn't like, I was definitely the, I
19:56 don't know, I was driving it from my interest. The company as a whole wasn't, and it wasn't ready to adopt it, it just wasn't their priority So when, just after I left, it died on the vine.
20:09 Yeah, it's like most things. Like that's why most oil and gas companies shouldn't ride their own software. That's right. And there's that study last week, right? There's about like AI
20:20 initiatives and how many of them like don't achieve positive ROI. You know, I think a lot of that is if you have, it's just a different business model to be a product like a software provider.
20:32 It's totally different Do you want to invest money in your drill bit or do you want to? invest money in software and if you need to have support. And, you know, yeah, there's a lot of swan bug
20:41 fixes. Yeah, exactly. There's a swath of things that come along with that that if you're not aware of will be very surprising very quickly to you. But I mean, interestingly enough, though, to
20:51 that end, I would say Antarctica was in the EOG, like there's a couple that actually do a decent job with some of that stuff, but it's in my mind, it's because it's structurally like it's
21:04 intentional, right? Like it's, it's not like, I feel, you know, not gonna knock anybody specifically, but I do feel like some of the companies are just like, hey, let's go. Yeah, that's
21:14 Antarctica where EOG is doing it. So let's go do it, which, you know, they do that for everything else. So why wouldn't they do that for organizational stuff? But it's like, you know, EOG
21:23 hires a ton of software people and like, they work all the time and there's all kinds of stories that come about it, right? Like it's, it's not just software developers, you have to to understand
21:32 how software works and what support people they need, and okay. life cycle of a software and you need your CICD and you know how are you going to implement improvements on the software and who's
21:42 going to and all this shit and it's like yeah that doesn't work very well when you have the industry that has all the booms and busts that we do and you lay off half the people that wrote the software
21:51 and or you get a I mean you got plenty of rogue engineers who are super intelligent love to tinker and mess with shit and they're like oh python I can process and I can solve a problem with this and
22:00 like so you do something really cool but then like how do you recycle that and institutionalize it if it works or if you are if you're ready for it even like yeah and there's a skill set too to like
22:09 because that's like a level of hey I'm writing a python script that's going to run on my machine locally and I'm like you know manually adding a spreadsheet to it and then there's no way running it
22:19 yeah manually clicking a run but a script not software right exactly yeah and that's like one level but then you think all right how would I take this to run on our like APIs you're in how do I like
22:31 access live data yeah but still like it's local on our mic. computer servers. And there's a whole nother level of, all right, now I'm going to take my script and make it a software that like, I
22:44 can share with other employees, other coworkers, or it's like accessible in the cloud. And then there's security, then there's privacy. Yes. And the
22:54 point you're talking about machine learning ML ops is one that's always been a
22:57 pain. They ask like, I mean, like, you know, I think they've sort of figured out DevOps, right? Like, you know, publishing software to that. But like novice machine learning thing that needs
23:05 to learn from itself and once it gets to a certain efficacy, then a new version of the model gets published and like all that kind of stuff. It's way different. And I think I've seen it with a lot
23:16 of those multivariate things. Like people forgot all the cleansing steps it took to get the data into that model. And now you need to deploy those as data engineering, you know, steps to feed into
23:27 that model before you can get there. So. Yeah, a hundred percent. So, yeah, there's, those are some challenges to like operating companies wanting to do. You got an AI, I think, in some ways,
23:38 maybe makes that problem bigger because it makes some of the development more accessible to in-house tinkerers and stuff. We did, I was on a project with an operating company that wanted to optimize
23:52 their call outs. We're calling it a water holler, a oil holler, a pumper, managed by exception They invested a million dollars to build their own in-house thing and then the company, right as
24:07 they finished it, before they could achieve like, see value from it, the company got acquired and the acquiring company was like, We don't have this, we don't have this. That's a good point too,
24:19 even just on the MA side, because I mean, whether it was what we did at Grayson Mill or, you know, I know like, Ensign sold to Marathon and like, Ensign had some really cool, you know, data
24:30 stuff and after talking to Andrew and Sean from there they're like, No, we don't want it. We just want it to be your acreage, like, yeah, yeah, just, so that's kind of like, do you want to be
24:43 a software product company or do you want to be, you know, an operations company? And it's like, in some ways, you got to stay, stick to your lane, you know, and I have the trust to go find
24:53 good, you know, software providers and partner with them. And but we're even like some of like, you know, like Devin, and I think it was pretty as public now, I actually saw it on the website,
25:04 but like they developed this thing called Leafcutter in house, which is basically a way to stream OSI Pi data into Snowflake. They developed it, but now they actually work with this company, Radix,
25:13 to take it to market. Cool. So I think they've gotten some royalties from it, but like the Radix company is handling actual software and marketing and all that kind of side of it. When I think
25:23 that makes a ton of sense too, right? Like there's so many software companies or contractors, developers out there that
25:31 would be willing to try and do that, like need the data to build that. And so you have to have some kind of partnerships with the operator in some form or facet. And so being able to have that like
25:44 jointly developed internally and then also you can go sell it, makes it a lot more attractive in the long run and alleviates all the typical headaches of having a software company. But that's rad.
25:59 So again, I mean, Tinker up and then, you know, you've got Patronus and there was a few years of Patronus just doing the process of safety management stuff. And then GPT stuff came along and this
26:09 is where you're like, inflection point of like, I can build things that people need. Yeah, I mean, it was fascinating that like GPT, as a large language model, was solving certain problems
26:21 really quickly that normally would take me hours or something, you know? So I was experimenting with it to like help with those tasks, right? some of it's like reviewing documents or. preparing
26:34 other documents, something that's like an LLM is really strong with. But I don't know, maybe my creativity kind of ran away. I have in my mind, Tony Stark, right? I'm like, how he can just
26:48 interact with Jarvis and you know, hey, I wanna redesign that. And it's like, whoa, it all kind of happens. And who knows if that'll ever like be the reality? But I thought, what's the next
26:57 step? And this is a cool like LLM thing and it has its place But what's the next step to actually doing replacing engineers in certain workflows? And so I started just tinkering and building apps
27:13 that replace myself in our workflows.
27:18 And I set a goal probably - I got serious maybe a year and a half ago. Once a quarter, I'm going to release an app. I'm going to have a full stack production build that is solving a real problem
27:32 that I care about. maybe someone else cares about it too. I don't know, like that's not my motivation, but
27:38 it was mostly for my development and my learning and it's gonna be, you know, production ready. Like anyone can go, it's on the cloud. It's not, you know, it's like, it's got all the security
27:49 and everything. All the authorization, authorization. Yeah, all the authorization, passwords, and the authentication.
27:55 And it involves AI replacing some part of, like, engineers in the workflow. What's your favorite one that you built and what's your least favorite one that you've built so far? So my favorite one,
28:08 the last one that was released, my favorite one. And you can take a set of PID PDFs. You send it, like, you create an account, you upload the PDFs, and then it automatically identifies all the
28:20 components on the PID set, and it gives it back to you in Excel format. So, you know, if there's an EPC company that's trying to, you know, put together an RFQ,
28:33 really quickly, sometimes that takes days if someone's manually building a valve list or an IO list or something like that. And now it just takes five minutes and it's done. And so that's my
28:45 favorite one. Probably, first, I think the most meaningful problem so far that I've solved. But second, it's also the most technologically hard from an AI standpoint, 'cause you have computer
28:58 vision different from an LLM. That's identifying schematics And then you have,
29:06 there's multiple AI agents that are working together and collaborating, double checking each other's
29:15 work, following MCP. And so that creates a product that, if you just try to feed a PD into chat GPTs. I gotta work good though. Yeah, that can work.
29:24 So I think from the advancement, that's exciting to me because now my brain starts moving of like, all right, what other work. those require multiple disciplinary like collaboration and how can we
29:38 just create multiple agents that talk to each other to do that same thing. Yeah. So what's your least favorite one? My least favorite. It's funny. So we I did the hackathon that I did back in
29:52 September. And the only reason my least favorite is because I shrunk this what I normally give myself a quarter to do to the three days or whatever
30:04 And it was the, uh, you just take a, you know, it's basically repository where you can dump a bunch of documents related to a well and then it uses an LLM to chat with them. That's what I would
30:16 think. So that's my least favorite, just because, uh, you know, it's like, it was rushed, but. I mean, it was still, it was super fun to do. No, that's the double-edged sword of a
30:27 hackathon, right? Is it's like, I'm not proud of this, but no one is proud of it because it was done in 48 hours with no sleep by a bunch of developers randomly, right? I like, yeah. That's
30:37 awesome. What are you most excited about with computer vision? 'Cause I feel like we could geek out over this for a while. It's so cool. It's just at the end of the day, it's just cool. I mean,
30:50 I could nerd out about computer vision, like first, I tried computer vision before, like even before, kind of like the Chachie BT, like era, you know, or you're taking TensorFlow and training
31:02 Lambda models. And it's like, it takes a lot of effort to train a computer to see something. But now you have tools like Roboflow,
31:14 makes it so fast, like so easy. And there's a bunch of libraries and stuff out there too now? like our right, pre-trained models that are trained on billions images.
31:26 I do think it's hilarious that the general public still doesn't know that we've been doing computer vision training all along for Google with CAPTCHA.
31:36 It's like no one thinks about that, but it's like, yeah, they disguised it as security, but you're literally training their computer vision model, which is why it's so good. That's
31:47 a bridge. We could. I think that there's so many applications for computer vision that are just starting, you know, it's like a blue ocean, if you will. For sure. No, the PNID stuff is also a
31:59 really good one because it spans multiple industries, multiple personos, so to speak, and stuff. And so it's a, I've been asked about PNIDs in the past as well. And so it's, it's something that
32:15 people are very interested in, so I don't, hopefully it does continue to get better, or at least what to do with it gets better, because Facebook ticked down a picture of or three years ago, it
32:24 was a selfie of us in front of like Petco Park or whatever. And it said like, yeah, it was offensive or something like that. Like,
32:34 yeah, you know, that's when the Armenia, as I was clean-shaving and everything.
32:41 When all the - Do you have a shirt on or what? It was North Face, which was offensive to people on the street, but that's why it was - That's why it was - That's exactly why, right? No, that's
32:50 when the Armenia, Azerbaijan stuff was going on a couple of years ago. I posted just a side-by-side picture of the, from World War I with some refugees from the bombings or the attacks that
33:03 happened recently, and they leaked in, flagged it for offensive or like misinformation or whatever, like it's literally a fucking picture that someone took. Yeah, like that was just
33:13 misinformation. Like, anyway, that's the scary side of language models, in my opinion, It's just like, if it's not a. fully, even if it is open source, you don't know the biases that went
33:25 into it. Like, you can test it and stuff, obviously, but with the, the made, most of the major models now all come baked in with like guardrails and stuff that I don't know how I feel about like
33:37 from a product perspective, I'm glad that I don't have to build that, right? And prevent people from prompt hacking or stuff or, you know, trying to do other stuff. But on the other side, it's
33:47 like, what are you blocking and or sometimes it will block random stuff that you're just like, how did that get blocked? I don't know. But yeah, I don't know. Interesting that. And I don't want
33:60 to go too far or go and keep this on him. But like, I'm the first computer vision thing I think you and I were thinking about trying to solve was like identifying the parts of the fracksages. And I
34:08 think people have been able to do that. And I was able to do it decently with, you know, rhythmically with like neural networks and stuff like that. But like, at the end of the day, it's like,
34:17 I can look at this and point that this is that and this is that, you know,
34:22 It's fascinating how simple that seems as a human, how incredibly difficult that is to bring a computer to do. Yeah. But even then, it's still incredible, a majority of the time. We're doing
34:36 this big project with an operator and it's over 75, 000 scanned pages. It was all paper files that they digitized. And if you can, generally speaking, if you can read it looking at the digital
34:49 PDF file, we can transcribe it, which is handwriting, all this crazy stuff, forms and checkboxes and all this crap. But then it's like, I'm looking at some of these documents and it's just
35:00 putting out gibberish on the extraction and then I'm looking at the document. I'm like, I can't literally read. It's like cursive in a sharpie on
35:10 crappy paper. And so it's just like lines. And there's like, well, I guess I'm going to throw that one out But yeah, I mean, it's just wild that, you know. OCR was so promising for so long,
35:24 and then it has been completely supplanted by computer vision, in my opinion, today. It's so much better, it's so much more flexible. Yeah, yeah, absolutely, it was just good. I mean, that's
35:34 what, like, 'cause even the old approach of computer vision, it's like still hard to take unstructured. It would need to follow a form, like a somewhat rigid thing for it to kind of get it
35:47 outside of those boundaries It's kind of garbage, but today it's like, you can have different form types and different, I don't know, handwriting types and different errors, and it kind of just,
35:58 it's way easier. Yeah, just go another thing. I computer vision too. So when the war in Ukraine, since the war in Ukraine changed like a lot of warfare, like drone development, in terms of use
36:12 in sabotage, people in the United States are now think, so like process safety company, That's what Petronas started as.
36:21 We look at a risk matrix like three years ago, nobody was talking about like drone attacks. Now people are talking about it. Yeah, I mean, thank you. Thinking about, I've always found it like
36:32 remarkable how close midstream facilities and pipelines and downstream facilities are to just like public roads, no less like if you have a drone. Like that's a real easy target. Oh yeah. I never
36:46 even thought about that, but. So like the ability for computer vision, we're using like layered technologies, right? Like, 'cause it's still, the weird thing is it's still, like it's a federal
36:56 crime to knock a drone out of the sky. Just 'cause it falls into the FAA rules, you can't do it. So the next best thing is just to detect and not disable the drone, but to respond, like get your
37:08 people out or shut the facility down or something like that. And so having a layered approach, where like you have a machine learning algorithm that's on a mic and it's listening And so it's trained
37:18 on what the. of a drone sounds like and it kicks on a camera and now you have computer vision that can scan this guy real quick and it detects the drone and trajectory maybe you have RF signal so
37:29 it's like AI is interpreting you know radio frequencies and so you have this three-layered approach that you can put on a simple device and now you have a response you have a drone response plan so
37:41 yeah that's pretty dope yeah someone would radar be in that or is that too expensive so radar I mean you have active radar which is at the airport like spinning um that's too expensive so you have
37:54 passive radar systems and that can be put into a small package and um so you can do that and AI incorporates in that AI is gonna you're basically having two radio frequencies for more has that be at
38:07 least two and use AI to migrate the radio frequency into meaningful information like hey it's moving this fast, hey it's this many miles away or or whatever it is, but yeah.
38:21 Yeah, that's wild, 'cause one of my favorite random things to do is watch those videos of the drone defense contests that they randomly do, like the military will host, where it's like whoever can
38:34 knock the drone out of the sky given these parameters and they're super dead. You know, there's people shooting nets and shotgun type stuff. And then I think the most like prominent and successful
38:45 ones are actually like, they just block the signal and the drone just falls out of the sky or it just hovers. They just interrupt the signal. So that's another thing that you could just layer into
38:55 that where it's like, if it gets too close. Yeah, so like jamming. Jamming into that, AI kind of unlocks that. It used to be, 'cause jamming is not, jamming is around for pre-World War II,
39:07 but you know, with AI's help, you can now put it on an edge device and it can turn on a jammer only when you need it to automatically and target and so. Um, again, I, I geek out about this cause
39:21 we're a process safety company. And now like people are all of a sudden asking us about this. And so,
39:28 you know, my God, I got an, I got an, it's awesome. What are like a couple other major challenges that you think AI will be able to help with in that process safety space, PSV like I'm So.
39:38 question Good. yeah,, Oh?
39:41 I mean, site inspections, all kinds of site inspection type things, like a PSV inspection or mechanical integrity inspection. If you can put, again, it's kind of computer vision. If you can
39:53 scan an area with a camera and it's trained to find anomalies, then, you know, it can speed up identifying something that's going wrong. You can see a fire, like, I remember I was working in
40:08 Wyoming for Anadarko and it was like
40:12 one morning we were sitting in the control room and And there's a trucker that's on the loading rack, and, like, look away, look back at the screen. Now it's just like a big ball of fire. And we
40:22 clear the control room and like run out there. I was grabbing like fire extinguishers and trying to put this thing out. And
40:29 we finally got it out. And first, we thought that like all we could see is this crusty cab that's like smoking. We thought there's going to be a guy inside. Thankfully he had left there and was
40:40 like in the porta potty. It had no idea that it stuck us on fire, you know? You imagine. Yeah So you're a long drive, you get there, all you want to do is take a look and then you get back out
40:51 of your truck so on fire and everyone's like, Dude, we thought you were dead. Gosh. That's wild. So after that, I thought, Man, there's got to be a faster. I mean, we could only respond that
41:04 quick because we were all like staring at this camera, but it could have been, I don't know, during some time where we just weren't watching and it would have been way worse if we had been like 30
41:15 seconds too late It could have easily lit up. his truck or lit up the 90, 000 gallon NGL tank, right?
41:24 So yeah, computer vision can help with that sort of thing 'cause it can just immediately detect anomalies, sound alarms, trigger like daily use systems, things like that.
41:35 Yeah, that's, I think there's so much room for that in our industry, right? Like what are you seeing, I've got lots of questions but going back to the survey piece, are you seeing that being
41:48 done by people or more and more people moving to drones to go like schedule, it comes out once a week, flies a pattern around the facility, does it survey and then goes back and dumps the data and
41:59 then analyzes it, is that like a real thing that's happening today? Or is that just something that I've seen it trade shows that no one's actually. No, that's great. I mean like, so the
42:08 technology is there 100. It's like
42:12 the regulators are still squeamish to turn the keys over, let people actually do that. And because the whole idea is we want to be able to do it faster, more reliable, remove the human error piece.
42:22 But where it's at now is human stuff, the babysit drones. But if you ever really want to see, I mean, come up to Bentonville, Arkansas, the headquarters of Walmart, if you want to see drone
42:31 delivery happening on demand, it's happening there, you know, that that's their RD facility. So they have like the drone hive, people ordering Walmart stuff. And it's just like delivering right
42:40 to your house. Um, and they're, what they're doing, it's kind of, I mean, it's pertinent to the oil field because their, their research and they're like, you know, they're the big gorilla
42:51 that can pull out a string. So they're able to do that sort of like pilot study for years and then get the FAA on board to like let that happen across the U S. And once that happens, then you can
43:03 get all these technologies that trade shows that can actually, you know, on a routine cycle, go up and do inspections and dump the data and all that kind of stuff. Like the technology is there.
43:14 For sure.
43:16 the regulations. Regulation, can we feel okay with drones flying? Yeah. Now that's a big pro of living in Northwest Arkansas, as you get to see all the cool shit Walmart's doing. Yeah. There's
43:28 one store in Fayette. One of the stores, the one on the north side of town in Fayetteville was one of their like test stores. And so I remember in college, they had just like updated it. And it
43:37 was very similar to what like targets look like now, but it was like a complete, was a shark. It was very different It was bright and happy and like, you know, my typical experience in '25. Yeah,
43:49 my typical experience in Wal-Marts are, you know, in the middle of nowhere when I'm driving to location and it's probably 2 am. and it's super scary. I usually go to Wal-Mart to remember why I
44:00 don't. Yeah,
44:02 it's different. I mean, it's like, everyone goes to Wal-Mart 'cause it's like, this is the gold standard, you know? For sure, no, that's, I don't think my wife knew what target was before We
44:14 got to favor.
44:17 Yeah, it's awesome that they're doing that. And even like the deliveries, right? Like that's fascinating in and of itself. And then you're like, oh, well, you could see how that would
44:25 translate into a lot of oil field stuff, especially when you think about the whole like, kind of the way that we spread our kind of network down our employees and the companies and stuff, right?
44:38 You've got your office, but then you've got all these field locations. And it's like the field locations have the parts typically that you need to use in the field And so instead of you driving two
44:47 hours down back roads, the drone can be there in 30 minutes with whatever part you need and shit like that. It's going to be wild or into another guy that he wasn't pitching it for oil field
44:59 applications. But the minute they they remove those FAA requirements, I'm going to be calling him backup, but it's drone operated lights. Oh, right. So like construction construction. Large
45:14 manufacturing, heavy equipment, just outdoor stuff. But really fascinating element of that is because it's like directly overhead and, you know, go very high, like you don't have the safety side.
45:27 There's no shadows, there's no like dark spots or anything like that on the location. Whereas, you know, the light plants are a direct beam and a direct direction. And so there's all these
45:37 different things and stuff. But the whole negative of that is you can't, you have to have a person operate them because of the stupid regulation That's right. They're tethered and you still have to
45:46 have a person operate them. Yeah. What's it going to do? Like, I mean, I guess there are bad things that they could do. But it's - Yeah. Or like, so comms is always a big thing in the field,
45:55 right? It's like, there's so much money you have to dump into, like, radio towers and, like, internet of things and whatnot. But some of that is just so you can get all your oil field data from
46:06 in devices back to your servers and stuff. But imagine how much money would save if you just have a drone that sits out there once an hour, once a day, it flies up, you know, 400 feet. And now
46:18 it's like a beacon and it's like received everything and disperses and then it comes back down to recharges. You know, and now you don't have to spend the millions of dollars to put a radio tower in.
46:30 You know, again, like that would, it's only possible once you don't need a guy standing there to just sit there and watch the drone as it's doing that. But, you know, it's coming. Yeah, it's
46:40 gonna be wild. What things look like in the future Yeah, so I've got one more kind of technical question and then probably getting the speed around.
46:49 So, you're building these SaaS apps and it's all pretty neat. Like, what's a typical stack for that? I mean, I'm assuming Python's baked in there at some level 'cause you're doing a lot of the AI
46:56 stuff or what is like a full stack like for you, typically. It kind of just depends. So, first, like, I like Nextjs as far as like a user interface. So, Nextjs web apps, easy to deploy that
47:09 to Vercel It's so easy to do like CICD.
47:13 with it. So that's the front end. Back end, you kind of pick your flavor of secure servers. Again, there's these companies, because data security is a big thing in the oil field. So there's
47:25 companies that just specialize in like back end as a service, right? And you think like Firebase by Google or like convex or these, you can kind of, there's a bunch of them. But so you have some
47:37 sort of back end for data and storage.
47:40 Clerick is amazing for developers for authentication. They just handle everything. They'll integrate right into Microsoft. So like your enterprise client, all use Microsoft stuff, like Outlook,
47:50 whatever it's called now, it's not intro. Yeah, which if you're like developing, you know, a SaaS product that someone's going to use, it's like a headache. If that company, your enterprise
48:02 client is like, people are getting hired, fired, they change positions or whatever. And so you, you Yeah, yeah, I you don't want to be managing their I went to that, I remember we were doing
48:14 it more traditional often, like, you got to have some kind of policy where, you know, people have to change their password every month or two, you know, just so that way, someone doesn't have
48:24 access for too long when they shouldn't. Yeah, then you have to worry about where the password is. We've brought an admin panel for them to manage it and they still will probably forget. Oh, yeah.
48:31 Yeah.
48:34 100. And so those are kind of like the key pieces that seem to pop up on every build, but then you'll have your AI, which sometimes is an LLM So it's like your cloud or your chat should be to your
48:44 Gemini or whatever it is in the specific model. Robo flow, maybe for computer vision or some other open CV or Python thing. Are you using any like Langchain or any of the frameworks out there?
48:56 Yeah, Langchain, Pinecone, like the RAG stuff. Are you using Pydantic at all? No. No, but I don't know, maybe. That's another one. That's the only reason I ask. It's another one. There's a
49:09 bunch of them. Um, so this kind of like the main things, um, I've done, if there's going to be a lot of processing and you're going to need Python, so you build like a flask web app, and then
49:23 that is generally tied to the front end via like a API or something like that, a web hook or something. Um, I, I, if I don't already have a server provider, then I like, um, oh, what's it
49:38 called? Uh, Linode, it used to be Linode, now it's, okay, I've heard of it, but they got bought a couple of years ago by another company, but you know, it's just like infrastructure as a
49:46 service type company. And so yeah, that's, that's the common tech stack. Nice. That was a really cool thing. Uh, something that I needed to dive into in there. So, you know, that's, those
49:56 are kind of things for this podcast. We like people to walk out like, Oh, I need to check that out. Yeah. That's cool. Yeah. What are some of your other favorite kind of slept on tools or
50:04 frameworks that services that you've used that maybe other people don't know about? I just went to this. Like, there was a meetup yesterday here in Collide with AI. It was pretty cool, pretty rad
50:17 to meet some other people, other builders.
50:20 I shared with some people. I like VS Code as an IDE, like a development platform. It's been around for so long. It's got a great community and plugins and stuff. Cloud Code used inside of the IDE
50:33 is my favorite flavor. Cloud Code runs in the terminal, right? Yeah. And I've tried all kinds of things And that's just my favorite. For one, it's in your project. And so you don't have to copy
50:45 anything over, which some of the guys I was talking to yesterday are still doing that. They're like, you know, build a script. And Cloud, they're copying it to their project. And it's just like,
50:54 so much time. And then you run out of chat tokens. And now I got to start a new chat and do the whole thing over again. But if you do it in the IDE, it's just boom, right there. That's all your
51:04 code. That is really cool. Yeah. Yeah, I've heard great things. I need to just bite the bullet and do it Carve out the half a day or a day and just like. I'm gonna do this. Yeah. Have you used
51:13 cursor, wincer, for client or any of those? Yeah. So I've used them,
51:19 like they're good tools. I just, everyone kind of finds their preference in their flavor. No, that's why I asked, there's just so many, right? It's like literally without trying them all. You,
51:29 that's the only way you know, right? Yeah. And a lot of them now, it's just so nuanced as to what the differences are. Like they all have their pros and they all have their cons in some aspect of
51:39 it, right? But you know, cloud code definitely seems to be the incumbent. Yeah, people seem to really love it, so. Yeah, what I'd say to you for like other thinkers or builders, like some
51:51 people I talked to yesterday are in recent conversations, they're wondering like, how do I get my start? And like, is there a course to take or whatnot? And I get, because of who I am on
52:03 LinkedIn, I get hit up all the time by those like, you know, Wharton Harvard, like AI, 12-week course offers. I just think that those, not that you can't learn something from them, to me,
52:14 they're kind of worthless. That's the money grab. That's the money grab. And if I was, you know, and I do this, if ever, you know, we hire people, or we're trying to look for people to work
52:24 with or partner with, I'm never gonna look at your piece of paper behind your name. I'm gonna say like, what'd you do? What'd you do? Yeah, what'd you do? Where'd you get that repo with you?
52:32 Exactly, yeah. What's your GitHub repo? Let me see what you've done. Like, okay, I can see you know this tech stack and that tech stack. Like, I don't care at all if you have some, I don't
52:41 know, degree or certificate on AI. Like, I just think it's kind of. I think that's one of the most like fascinating things about this show is that it took like, I don't know, 50-something
52:51 episodes before we had our first degree developer on it. Yeah, yeah. It was like it was almost a full year of episodes before we had a dev, like a train dev on her. I don't want to say it about
53:03 Talal's. That's true. But just, you know. In general, like, I think the point is just most people don't come up with a traditional background. It's like that fine way to solve problems. Most
53:14 people were not software devs or computer scientists that have been on the show, which is pretty awesome. But I think that's, I genuinely think that's what the future of work ends up looking like,
53:24 right? Is it's like everyone is AI enabled, and if you're not, you're not working, right? Because it's that it just accelerates your output and your decisions and all that stuff. And so it's,
53:37 yeah Yeah, I mean, I think that general principle applies just to everything, you know, every endeavor in life is like, you know, I've partnered on like real estate deals, right? And if
53:50 someone wants to partner or wants to do something together, I never once ask, you know, where'd you get your real estate treat, right? Yeah, I'm just like, what have you done, man? Yeah. You
54:01 know, and if I think it, you know, if you've done something similar, it makes sense. Let's go, you know? And I grew up. in an era where like there was this heavy emphasis on college degrees.
54:13 And like
54:16 I'm grateful for my college experience. I'm grateful for my college degree, but like I'm not convinced that I'm gonna push that as hard for my kids to go to college because I care about competencies.
54:27 I care about skills and for them to be useful individuals. I'm not convinced that there's something magic about
54:32 like
54:36 a piece of paper that makes them competent or ready to be a contributor. It's the mortgage of death that they have coming out that makes them the competent adult, right? Yeah, I'm right there with
54:48 you. I don't know, I mean, I understand why engineers need to know like Diff EQ. and stuff like that, but I'm not sure that that has ever been used by me ever since I left campus. And so it's,
54:59 yes, I completely agree with that. Or like bouncing equations. Have you done that in a while? It was less than you did that I don't think I ever got a college year coming from a junior. Yeah, I
55:11 just do heist, heist does that for me. Yeah, one quick thing and then we'll move on. Are you using like a Vovable or Replet or any of those tools for like POCs and prototypes or anything like that?
55:22 Do you have any favorites there?
55:25 Yeah, so I've used a bunch of them. You know, I think Replet is probably my favorite for like prototypes and
55:33 it kind of just at this point because I've done enough builds I think
55:39 I find myself more and more bypassing the prototyping phase. Like I don't need - Because you can probably prototypes really fast. Yeah,
55:47 as you were just talking about doing it. It's crazy how easy it is to prototype with prod stuff now.
55:52 Yeah, and I would never take the output of a Replet, you know, build and that's like my production app or something, you know.
56:01 So yeah, I kind of more and more I'm skipping that phase and just prototyping live, you know Yeah. one piece of advice you would give to someone that's trying to go from scripting to prod type
56:15 development. I know that's a really weighted question. That is a very big jump. But just like what are some things to make sure they pay attention to? Security is obviously one of them in my
56:25 opinion. But yeah, good question. Maybe like the first baby step, if you're scripting, let's say in Python, then go to streamlet. I don't know if you're going to streamlet, but use it a lot
56:36 Yeah, I think I recommended John. He did. Yeah. More beautifully. Yeah. It's like the next next baby step. And there's no authentication. I mean, like, that's definitely not like full stack
56:47 pride, but it's the baby step from going. If you're going to ask stuff like that, you can deploy a snowflake, snowflake boss streamlet, right? And now that you can build streamlet within
56:57 snowflake, that's like deploy them right in snowflake and let that handle all your off and all that stuff. I haven't played with that since they bought it, so that's awesome. Yeah, I didn't know
57:06 that either. Yeah. There's a little. Might have you be setting that up for me too, buddy? It's already set up in there for you. You should be in here, but yeah, we need you.
57:16 So yeah, that would be like, I would say, hey, go to Streamlet, if you're at that, like I'm a Python scriptor kind of guy and mess around with Streamlet, it's snowflake. And then from there,
57:27 it's kind of like the next question. It's natural for you to say, all right, how would I get authentication next? Or how would I, these are great for, you know, doing something and maybe I'm
57:38 tying like an open database to it, but how do I get a secure database behind that? And so you start learning about secure databases. Yeah, working and like, yeah, all that stuff. So I think
57:48 that's maybe - You don't just put zero, zero, zero on your iPhone.
57:52 One, two, three, yeah. Yeah. No, that's a good answer. I think, like I said, it's, the average person just does not understand the difference in like what I can code and what Bobby can code
58:05 They just think we can both code and it's like no. None of my code touches prod. It will never go to prod. Yeah, and I do like data engineering level code, but like I'm deploying a SaaS app. But
58:15 again, the barrier to that has come way down and you can get there. Well, I mean, you can literally just ask GPT how to do it and it'll walk you through the steps on generally how to get that shit
58:25 set up depending on what infrastructure you want to use. But if you find something services too, like anything with authentication, I'd much rather have some kind of service that can stay current
58:32 with all these 'cause like even it just. Except for all three Too important, if they can kick rocks. Yeah, well another thing I'd suggest is like get on dev hack, which is, you know, they
58:44 aggregate hackathons and you can hop on there, find beginner hackathons and pick one for like the beginners. Because the reason that matters is it will likely give you access. It'll give you the
58:57 exposure to GitHub, like continuous development, continuous deployment. That's a skill where, you know, for those that don't know, like kind of get hubby, it can be like. the Google Docs of
59:09 coding. It lets like multiple people in different parts of the world like contribute to, you know, a code set. And that's a skill to be able to like, collab with someone else. Cause if you were,
59:22 if you're getting serious and you want to build a big, big project, like no one's going to build Facebook on their own. Like you're - And a Python script. Yeah, and a Python script. It's like a
59:32 verb is going to have a word with you. Yeah. Or some PHP and my SQL on the query is, there's no E and V file, no security. Yeah. Yeah, so it's a skill to like be able to, how are you housing,
59:46 if you have multiple API calls that you have all your API keys in the spot. Listen, go on, when you leave here, go on GitHub and just search API underscore key. And it is terrifying. Or search
59:59 actually search SK - Okay. Which is the prefix for all the open AI keys It is incredible. How many open AI keys are published on GitHub right now, or not just open AI, but all of them? That might
1:00:12 be the best knowledge drop. It's some free tokens. I was going to say, do without what you wish. I am not promoting that you go use other people's API keys, but also don't post your API keys on
1:00:26 GitHub. Yeah, put it in your git ignore. Private repos or something, but well, I'm on a bit of time crunch. So I think we'll give him a reprieve on the speed round today Do one quick one for him.
1:00:37 All right, if we're, I'm going to bend the lark and saw where we're going to eat. Okay, so you should go to the big orange. Okay. It's like the best burger. And right next door to it is, I
1:00:49 would do this. I would go to get the burger from the big orange. And right next door, you go to get tallow fries. Oh, nice. That's right. That's fried in tallow. Man, they're awesome. And
1:00:57 then go to JJ's for a beer. Watch the hogs.
1:01:02 We'll end it on that. We'll dig soon Whooping baby. Here we go. cool me. Well, thanks for joining us. I'm glad you're in sound like you're in town fairly often, but glad you come be with us in
1:01:11 person. Absolutely. How can people get in touch if they want to, want to chat? Find me on LinkedIn. Yeah, you look me up there. Steven Barrow. Awesome. Yeah, enough. Appreciate it, man.
1:01:22 Thanks, everybody. Thanks y'all.