00:00:00:00 - 00:00:23:21
Unknown
Welcome to another episode of Energy Bites is an international episode. We only have a handful though, so far. So we've got, Paul Brown joining us from here in Scotland. Is that correct? Yeah. That's right. You. Thanks for. Thanks for having me. It's so funny because like, literally I can see you make a post on LinkedIn and we get this set up in like a day, and then we go some other people I've been trying to get for two years.
00:00:23:21 - 00:00:43:07
Unknown
And they say we have to talk to, you know, legal and all this and you know, it just takes a lot of red tape to get it to make it happen. So they're across the highway. We can't get them in the office, but appreciate you joining us across the pond. Yeah, but we were already having some great conversations before we got on here.
00:00:43:09 - 00:01:04:22
Unknown
But I know we were kind of leading in with, just the wild ride that is oil prices right now. Because when I checked last night, it was 113 and 112 and WTI was above Brant, and now we're down to $93. Just because Trump opened his mouth and, so, you know, you know, here we are.
00:01:04:22 - 00:01:25:07
Unknown
But, but yeah, I mean, I know, Paul, you were saying it's, you know, a lot of this has caused some, logistical issues out, your way. Yeah. We, we have outside the, the the petrol stations, the gas stations, like we have kids are like 15 minutes, you know, half an hour long, and you get there and and there's no petrol on the pumps.
00:01:25:07 - 00:01:50:01
Unknown
And like, at the is, I don't really know what's going on. So I don't know if it's the same there or what, but yeah, we're not seeing obviously that shortage like that. But I mean prices are definitely going up. But like I said, you know, I was wasn't thrilled to put $4, a gallon in my tank, but you guys are probably at least I think we are.
00:01:50:03 - 00:02:25:04
Unknown
15 or $20 a gallon would work out something like that. Like it's it's, I mean, I ride my bike like David all. We are. We are spoiled by cheap energy here, especially in Texas. But it's also, I think, cool and important to see what else is going on across the rest of the world and how things like, you know, them blocking the Strait of Hormuz, which has, I don't know what percentage 30, 40% of the world's oil goes through that strait.
00:02:25:06 - 00:02:52:13
Unknown
How that has trickle down effects. Not I mean, I wouldn't even call that trickle down. It hasn't been that long and it's already impacting supply chains across across Europe and Asia. And so that's, no, I think that's you know, like I said, in in America, we get so we're just so used to that. We've had we've gone two generations now since there was any kind of like oil embargo or price shock like that to where the supply chain was actually disrupted.
00:02:52:15 - 00:03:21:05
Unknown
And so we don't know. Right. Like we don't know what that's like even though our parents do. And so it's part of me worries about that because it's like, hey, people don't think about energy independence and how important that is or the effects of that, that, that, that has on your life as an individual. You know, I guess the closest thing we got to was that the the freeze a couple of years ago where we were, I was without power for like almost a week.
00:03:21:07 - 00:03:44:02
Unknown
And then you're like, oh, yeah, this is important. Heat is important. When it's snowing outside, we don't have like, we I don't remember any short. I'm sure there were some, but not that big. But we now have like something like this, the highest energy costs in the developed world. And and that's, three times higher than I think it's three times higher in the US than the US.
00:03:44:02 - 00:04:03:02
Unknown
It's because of all the, you know, like the push for renewables and stuff. There's abandoned drilling, been abandoned drilling for the last ten years by banning them, issuing new licenses and all sorts of legal stuff to prevent drilling. And then and it's resulted in like just the industrialization. But you can't compete in manufacturing, you can't compete in industry.
00:04:03:04 - 00:04:28:04
Unknown
There are no engineering jobs or a very few engineering jobs here. The oil industry is just, you know, just just shutting down. It's energy independence. Like you don't think about it until it starts to have an impact on your, you know, everything. No, absolutely. That's, Well, that's the thing. It's like energy is the one of the biggest inputs to all manufacturing costs, right?
00:04:28:04 - 00:04:51:10
Unknown
Like that's one of your biggest costs and anything that you manufacture. And so it literally trickles down all the way, all the way down to, you know, it being trucked to your house or to a warehouse or whatever. And so it's, yeah, you're right that like, you get kind of lulled into a false sense of security until it is knock knocking on your doorstep and you don't have power.
00:04:51:10 - 00:05:19:12
Unknown
You can't get gas. And you're like, oh, this is very fragile. Well, before we jump in, I'm going to shamelessly plug the podcast because we finally have merch. Bobby and I are repping the, the hoodies. Bobby, give us a little twirl so we can check out the back. So the back says data are pain. So everybody in the industry who has ever dealt with our data, I think, knows and appreciates that.
00:05:19:14 - 00:05:39:06
Unknown
But, those are available on the doghouse over on Callide. If you guys want to check that out, we're super excited to have some merch. I've also got we've got an energy bites hat that I'm currently wearing, and there's also a, short sleeve t shirt, because summer is already here in Houston, and these cities are probably being retired after this episode at least until next.
00:05:39:07 - 00:06:05:02
Unknown
Next, December. I'm pretty sure my wife's going to steal it. Yes, my mine have all been stolen already by my daughter and my wife. So, be aware. Anyway, let's jump into it. Paul, tell us. Tell us a little bit, about yourself. How did you, Yeah. How did you get started in either the energy or the tech side of the the industry?
00:06:05:04 - 00:06:25:07
Unknown
So I am, I studied mechanical engineering at university, and, Let's go. You're in good company, Paul. Yeah, yeah, I think so. Yeah, yeah, it's him, but on the side I like. It's about, you know, I worked in a couple of, I did quite a bit of software development for, like. And the engineering companies. I'm on projects.
00:06:25:09 - 00:06:53:23
Unknown
I went to drilling and I worked as a drilling engineer for, like, internationally, deepwater exploration and like, so, like, I worked in Mediterranean, North Africa, West Africa, South America, Caribbean, the UK, North Sea and so like drilling engineer West I engineer, night company man, all that sort of stuff. And then I took a break from the industry, like, just time for a change.
00:06:53:23 - 00:07:15:05
Unknown
And I started a small manufacturing business where, like, the goal was to take kind of low tech. Yeah, not necessarily low tech machinery, but kind of dumb machinery. And, but apply, you know, like engineer, engineer. The process, to, to make it more efficient. So I spent years doing that, spent like five, six, seven years doing that.
00:07:15:07 - 00:07:41:12
Unknown
But I spent most of my time developing software, developing automated machinery, and got to the point where the like, the administrative overhead was, was like nothing like next to nothing. The the, the place ran itself. And, and I eventually sold the business, and to focus on the data side of things. And so for the last few years I've been, I've been them to, to in work, data work and in drilling and upstream.
00:07:41:14 - 00:08:19:05
Unknown
So things like, taken, taken like an operator or I read contractors data, building a system to analyze it and then providing the analysis as a service. Right. So, like comparing rig versus rig performance or, or checking to see if less lessons learned had been implemented, regression or whatever. Building systems for, for like oilfield upstream businesses like business data systems, like reporting systems and that and but more recently, like I'm focused on two things and the, like the big one is taking legacy data, like taking these old bio files that are, that are everywhere.
00:08:19:06 - 00:08:35:22
Unknown
Taking them, analyze them in my systems and providing the, the, like the client with the just the results like so the prepared well montages and the prepare the prepared you know output so that they can actually use it. And that's where I've got the, the Wells schematic.com system that's used by, you know, that's why I developed that.
00:08:35:22 - 00:08:55:06
Unknown
I needed to I needed the, the use for my own my, my own work is connected via API and so I just, I just released it and it's used by quite a few people. So that's one big thing, the legacy data analysis. The other thing is, I'm doing workers do quite a bit of work is, like consultant.
00:08:55:09 - 00:09:15:13
Unknown
So you've got these oil field data projects or drilling data projects where, where the, the, the drilling guys know everything about, you know, drilling and wells and stuff and the data guys and the software guys and everything about data and software. But they don't like there's no they they just can't bring together all this demand with. Yeah. Yeah.
00:09:15:13 - 00:09:42:14
Unknown
So like so I just kind of work with them and do the kind of translation and just make sure they're talking the same, the same about the same thing. I would say now though, the do you even have to do that? I mean, I feel like they could go right to you. And now with like some of these coding tools and stuff, like, I feel like you could even cut out that person because you know, enough about software engineering, it sounds like, and the domain like you're the perfect person to just say, like, let me run with this plot or, you know, go to X, whatever.
00:09:42:14 - 00:10:06:22
Unknown
And just knock it out like I can. And I do that for companies, but that that's okay for small companies. But what they say, what they always say is like what happens? So you get hit by a bus and like in like, you know, you've got the deployment and all that sort of stuff and administration. And, and like, you know, it's not just oil companies where it's like software companies and, and software for, for, for drilling companies.
00:10:07:01 - 00:10:32:06
Unknown
So like, they've got the software guys and maybe they're just not quite connecting with their, their, their client customers. But the other thing is like when you have these big companies, you've got lots of data, you've got lots of software guys, lots of data people, and it's enterprise needs like, yeah, yeah. Like it's straightforward to, to, to build, a system that's like the drilling engineers would want to use and actually use and it'll meet the requirements.
00:10:32:08 - 00:10:52:06
Unknown
But and all the enterprise bells and whistles and all that, you know, like that you need to involve other people and they need to do their thing and, and like, yeah, there's a different level of rigor, you know, on a managed SoC. Two audits. Paul. Yeah, I know, no, I don't blame you. Yeah. I've got a question.
00:10:52:06 - 00:11:15:13
Unknown
When you were when you were drilling, what was what was your where was your favorite place to drill and what was the like scariest or sketchiest place you've ever drilled? The favorite place was, I don't really know. You want to reckon. It's like they're all kind of. Yeah, that's. It's fairly. Yeah. The sketchy this place was, West Africa like.
00:11:15:13 - 00:11:41:21
Unknown
That's different. That's a different level. It's a different place. Like, I was, you know, the, subducted a couple of times, you know, like, you get em. I mean, walking down the middle of a street and there's an the security guards everywhere of AK 47 is like the busiest street in the busiest setting is busiest region. And then, like a guy who just flagged down a taxi or two taxis, the intercept it and then try and, you know, like, try and take your stuff and bundle.
00:11:41:21 - 00:12:00:10
Unknown
Yeah. Like, it's, it's, it's a different, yeah. It's just, you know, in the morning reports, if you're on a wreck, you get more reports, like, weather reports, you get, you know, weather. Put acid at 611, 6 a.m. and in Africa you get it's like a weather report. It's got the heat map, and it's got like him.
00:12:00:12 - 00:12:25:03
Unknown
It looks like a heat map, but it's a heat. Heat map of piracy activity. So like what we were we were on the rig one day, and then there was a morning, there was a call and it was like, oh. By the way, the oil tankers, disappeared or like the oil tanker has been hijacked. And there's as it's like a maersk or one of these big company tankers and for refilling, you know, in the ship in line for refilling the vessels.
00:12:25:04 - 00:12:52:18
Unknown
Goodbye. And they just the phone then or. Sorry, the people just hijacked the tanker and just taken away and runs with it. And it's like, how do you how do you where do you hide the tanker? And I think, yeah, it's just like normal. It's just normal. Yeah. It's happened all the time now that's, I've heard a lot of stories about about Africa, like, you know, you're living on a compound.
00:12:52:18 - 00:13:20:03
Unknown
You're being driven with armed guards everywhere you go. Even, like, entering and leaving the country through customs and stuff can be very touch and go and like, it's very, normal to have to give some kind of bribe or whatever to get in or out without any kind of issues and all that fun stuff. Yeah. No, that's, never had to, never gone over there.
00:13:20:03 - 00:13:41:01
Unknown
But I've, I've definitely always heard, some sketchy, sketchy, sketchy stories pretty much always come out of that. But yeah, that's that's crazy. And that's how you got to go down to the Falkland Islands as well. Yeah, that was that 13 years ago when they were just exploring. So that was just when I was kind of maybe like 15 years ago.
00:13:41:03 - 00:14:04:20
Unknown
And they, they if they're just exploring. So you, you know, that's like an 18, 22 hour flight, something like that, you know, like, get you to Australia, you know, other side, other side of the world, but you just going south and then so it was just that all the supplies have to come, like, you know, at 10,000 miles over at night, one has and it's in, it's just this middle of nowhere.
00:14:04:22 - 00:14:32:04
Unknown
It's the it's just the middle of nowhere. As all good oil fields end up being. Yeah. They're never near anything. No, no, no. I wonder if that's part of people understand like, say, the eagle bird. I, I underestimated it to describe. My first asset that I worked in was like, I worked in an office, and you could drive two hours on a major freeway, and you were pretty much there be in the heart of the field.
00:14:32:04 - 00:14:57:03
Unknown
Yeah. But other. Oh, I've not even then it's like you drive down to Katrina or Carrizo and you're like, where the hell am I? You're just on these giant ranches and you never, you know, you can go miles without seeing people. It's a it's a weird. I mean, it's weird because it's like you have the infrastructure at least to get stuff there, but it's still in the middle of nowhere.
00:14:57:05 - 00:15:23:08
Unknown
Like we end up building infrastructure around the fields if they're good enough. Like that's Midland is the perfect example of that, right? I have no comprehension of like, you know, like the UK oil industry, a lot of the, a lot of international projects, Africa and Europe, however they're managed like Europe, sort of means that the UK but the UK is very much a bubble compared to like, you know, America, Texas or whatever.
00:15:23:08 - 00:15:38:20
Unknown
It's it's kind of like self-contained almost in the UK's bubble. And they don't really there's not a lot of an I feel like there's not a lot of interaction between them. Like so I have no, I, you know, I have no I see things in TV and stuff. I don't have any idea, but I think it's the one of the biggest differences in onshore and offshore too.
00:15:38:20 - 00:15:59:17
Unknown
Right. Like I've never been offshore, but I know it's a completely different animal than the shit we do onshore. It's just the scope is different. The costs are different. Like everything about it is different. We're doing the same things generally. The you'll have a much harder, a bigger problem because you're in the middle of the ocean most of the time.
00:15:59:19 - 00:16:23:00
Unknown
But, from a logistics perspective at least. But they are, you know, people think oil and gas and they just group it all together. But those are two very, very, very different animals at the at the end of the day, like, I can't even I don't comprehend the offshore timescales that you'll deal with. Right. Like planning for five years and just the uncertainty for some of you do all.
00:16:23:00 - 00:16:49:13
Unknown
Yeah. And yeah I get a dry hole. Some of these whales are running like 100. I think 100 million whale. And and they're just vertical whales and they're getting abandoned. They produce nothing. They're not designed to produce anything. They even and driven. Engineered and like, it's totally different. Like, you think it's just the same things. But as I understand it, like in America, you're optimizing your fracking teams with your, iPad making.
00:16:49:14 - 00:17:09:11
Unknown
I don't even know the words. Your pad making teams representing your. You don't do that at all. Yeah. It's much more of like a, almost like a manufacturing mindset, especially at this point now that we've kind of got a lot of it worked out. And now how do we optimize all these pieces? I mean, it's but it's not even close to the same as, like some of these manufacturing operations with like just in time.
00:17:09:11 - 00:17:30:05
Unknown
But they're trying to get there, right and reduce downtime. But it's like that's why we know we're going to get oil. You know, it's more about like how efficiently can we make operations. Is it going to pay out. You know, are we going to can we keep our costs low enough and get the IP high enough so that in 6 to 12 months we get our money back and then we can start making money on it.
00:17:30:05 - 00:17:53:13
Unknown
But it is a very different. Yeah. There's very little exploration going on in onshore shale these days. Right where it's like you would never do an offshore project without some kind of exploratory data. Seismic, like there's all the stuff that leads up to even the actual drilling of it. Whereas here most of the plays have been fairly well mapped out and identified.
00:17:53:13 - 00:18:18:00
Unknown
And you know what you're going to see when you get there. And then, yeah, it has just become an optimization of operations. Right? Like when I started 15 years ago, it would take about a month to drill, you know, a 10,000ft lateral, which that was a long lateral back then. Now they're drilling for mile laterals, you know, double the length and they're doing it in a week or two.
00:18:18:02 - 00:18:37:05
Unknown
You know, the used to you used to be able to use the rig count as a really good indicator of what oil price was doing, and you can't anymore because the rigs are so efficient, like they can drill a well in 5 to 7 days all day long. And then the frack crews got more efficient. You know, how many stages can we do in a day so that we're not spending a month?
00:18:37:05 - 00:18:57:02
Unknown
Because that's what it used to be. We used to plug and perf everything, and we didn't zipper. We would come in. We they drill a 30 stage two mile frack or well, we come in and we frack it. It might take 2 to 4 weeks to frack the well. So you're two months now, you know, a month to drill, a month to frack, your two months into spending.
00:18:57:02 - 00:19:24:09
Unknown
You know, all you've spent all of your AFI essentially on this. Well, between the drilling and completions and you have no production yet. And then it might take another 2 to 3 months to get it on a production. And so they thankfully they figured out that like, hey, this is a terrible idea. This is like, this would be the equivalent of manufacturing something, sitting it on a shelf for three, 3 to 6 months and then going to sell it instead of trying to get it online as quickly as possible.
00:19:24:11 - 00:19:59:14
Unknown
But it's yeah, it's all an optimization game at this point. From a, a field perspective, I think we're going to start seeing a shift because they've, they've they've squeezed as much juice out of that, that fruit as you possibly can get on the the operation side, at least with onshore. And so now they're going to, they're going to start looking more, they're going to have to start looking into the data to figure out, you know, how do we increase our recovery factors or, you know, they're going to also have to start doing some exploration because we've kind of proven out most of the shale.
00:19:59:16 - 00:20:19:12
Unknown
But, you know, we're only recovering 5 to 10% of the oil in place from the shale to begin with. So it's like there's a huge opportunity just there. But we'll see what we'll see what happens on all this stuff. Okay. That's I don't know. I don't know much about that, fracking band here in, in band in Europe is 005.
00:20:19:12 - 00:20:41:18
Unknown
I mean, for now, until you guys can't get any oil at the pump. Yeah, until the gas and oil gets shut off. Yeah, yeah. You know, as the guy Robert something Robert Bryce, he's got the whole iron law of energy or whatever that, you know, people will, you know, do whatever they can to keep keep the lights on like, like they'll pay whatever much and do whatever they have to do to keep the lights on.
00:20:41:18 - 00:21:07:14
Unknown
So hopefully they start doing that here at like something to keep the lights on here because. So let's jump in to let's talk about BTC data. And then I think, you know, the thing that kind of spurred this podcast was I saw your post about, well, schematic.com. So can you kind of talk about maybe whilst Malcolm first and like just kind of what that project looks like and what it does and and then we can get into what you, what you're doing on the BD side as well.
00:21:07:16 - 00:21:29:16
Unknown
Yeah. So basically when I, when I am starting this data stuff, it was like you can do you can can if, if you know the drilling domain and you know the software side of things, you can do anything, but you can't really do everything. So you have to focus and you know, and something so so I started like the whale schematic, dot com system and it's for drawing well schematics.
00:21:29:16 - 00:22:03:14
Unknown
Right. Like, where's Maddox are used by like everywhere. And these musical schematics show what's in the whale and, you just need to make a schematic. And so I made that system because the way people use every single drilling company that I've seen the the use Excel to draw the schematics, like the, the, the dragon, the drag, the the size of the sails, the rows, the columns to get the, the like, make boxes, the highlight the boxes and then the, the click like right border, left border.
00:22:03:15 - 00:22:26:01
Unknown
And that's how you draw well schematic. And like that's not very efficient if you have a thousand people worldwide in that any time. So it's almost like we were doing that in a tool that wasn't designed for. Well yeah. Yeah. I mean, you could say that I don't know that our industry loves to put as much random shit into an Excel file as we possibly can.
00:22:26:03 - 00:22:46:09
Unknown
At least that similar between the between, like the rest of the world and the medium. Yeah. So using this bit, so I made this very schematic system and that's been used by, it's used by quite a few people, people to draw these schematics like guys and rigs and stuff and and I connect to that, you know, I mentioned via the API for the, for the data analysis to the PDF data.
00:22:46:09 - 00:23:05:09
Unknown
So like the company send me their files. It's just me. So I can't really like, and I software as a service is not really realistic as in terms of like data analysis. And companies, I don't think they really want it. They don't really want the hassle of, of the contract and new services going through it. All that.
00:23:05:11 - 00:23:23:05
Unknown
So that connects to the warehouse schematic system. So there's this kind of mutual incentive to, to, to improve the schematic system. I know it's the case where I added and, you know, talking about the everyone using spreadsheets for things like case and design companies are still using your options are a spreadsheet that takes you like two weeks everywhere.
00:23:23:06 - 00:23:43:11
Unknown
You know, you may only do a case until then, every year, or maybe less. And these big whales. But you're spending like two, two weeks to to build your calculations and excel and they're just not good enough. There's no loops or anything like that. And you know, alternative is then there's other software that's like five grand a month minimum, three months like minimum.
00:23:43:11 - 00:24:04:23
Unknown
So it's like one way was 15 grand. Could be so there's obviously a bit of an opportunity there. So about the you know, the case design systems, calculation systems. So you can go in to contact them, you can draw a schematic and then you can just apply a casing loads directly there. And companies do that but it also connects via API to other systems.
00:24:05:01 - 00:24:36:13
Unknown
And I've been providing like technical. Well and you didn't. And parts of the process, the delivery process as a service. Right. So you imagine you've got you imagine you've got drilling departments that's under resourced right. Like, I love it's project management. They don't necessarily have the technical expertise or like it take them weeks to get back their head back end to to technical expertise in terms of like the calculations and stuff like that, etc., hiring new people instead of contracting new software, you just, you know, speak to me.
00:24:36:15 - 00:24:53:09
Unknown
And it's kind of like a turnkey service. You want your casing design done for hand calculations done. And it I come software, if you know what I mean. Like I use my software as a service and and yes, there's like a front end on the website Mediacom and in the case design, but it's all the calculations are all exposes end point.
00:24:53:09 - 00:25:14:11
Unknown
So I can just access them via, you know, the API or via MCP or whatever to, to make things faster. And I think it was like an, you know, the post you had, you felt like a lot of, like old, you know, books and stuff into it as well. Or you were able to, you know, like enhance it with some of like the engineering, you know, literature and everything.
00:25:14:13 - 00:25:34:16
Unknown
Yeah. So like, I mean, a system a few years ago, 2 or 3 years ago, so, so like when you're working on index, like, everybody knows, you'll get like a hard drive and it's people gave you files and they gave you, like, the PDFs and they gave you, standards and all that. And, and so a couple of years ago, I had like, like 50,000 documents.
00:25:34:18 - 00:25:58:12
Unknown
And then a couple of years ago, I made, like, a like system where, you know, ingested just those documents, so you can search over them. And, and I made, like, a kind of simple front end, interface, and, and I and I kind of released it. I mean, obviously, I figured filtered out for, for, like, you know, copyright stuff obviously affected it first, but I released it and then people started using it and as like, I don't want to want to get involved in this.
00:25:58:12 - 00:26:27:00
Unknown
You know what? I'm going to try deactivated the writing system. But but basically what? Yeah, what you can do is you can then like, let's say you're looking at it, you know, you want to find, relevant information on a particular very specific thing. And drilling. You did a rock search. I tell you do a search, it returns a, you know, obviously an AI summary or whatever, but also, the context that was ingested, like the chunks that you, embedded.
00:26:27:02 - 00:26:47:07
Unknown
And then you can click on it immediately go to the source document. So, so in terms of retrieving your data and retrieving like in like if you're not working something for a while, I don't know what relevant standard is for something. It becomes easier to do that. Like you're no longer searching the things for errors. You're just it's just, you know, just out, you're done a couple a minute.
00:26:47:09 - 00:27:05:08
Unknown
Nothing. You're speaking in John's language here. I think that's, Yeah. That's what that's what I've been working on for the last two years. Is our enterprise drag, offering. But no, I mean, that's that's the thing I don't I, I feel like it's starting to come around a little bit more. At least for the average person.
00:27:05:10 - 00:27:29:14
Unknown
But, like, if you've got a bunch of documents that you constantly have to reference, a rag is the perfect solution for that. And it's so easy now to go out. I mean, there's so many open source GitHub libraries where you can just set the whole thing up from pipeline to front end with a few commands. And it's, you know, I would like to be able to tell people that it's that easy.
00:27:29:14 - 00:28:00:01
Unknown
It's not. There's always complications with the data. Ingest is normally the the hard part on especially with oil and gas documents, depending on what you're trying to ingest in there. But the technology around that continues to get better. So that continues to get easier. And then yeah, like if you're constantly referencing API's, you know, API procedures or API standards or anything like that, a ragu will save you so much time.
00:28:00:03 - 00:28:29:18
Unknown
Walk me through so. The the well schematics of I building new wells in that. Or am I able to upload a schematic from a PDF and then have it kind of digitize that? You can build a entirely new schematic. It has a very basic I think it's just like structured output. It's the it's a is a very it's like, I had a landmark 2.0 or something just for speed and just I've not upgraded it yet.
00:28:29:20 - 00:28:51:05
Unknown
Because nobody's paying me, so. Yeah, you can, you can make a new schematic using the using the, the forms there. It can, you can connect the via MCP as well. So, so that you can connect it to like, you know, a dot cloud or something like that. Which means you can automatically generate schematics, but there are limitations in that, in that it, you know, these things, there's security limitations.
00:28:51:07 - 00:29:17:21
Unknown
So you can display, you get an SVG that directly comes from the, you know, the service. But in terms of, making a schematic from, from a PDF, it just depends. Like, I'm thinking maybe to release a system that automatically creates schematics, but, like, there should an issue where data is that you and you can.
00:29:17:21 - 00:29:38:04
Unknown
Absolutely. You know, if you connect the via MCP, you can just take a picture of a schematic, you draw on your phone and it will just it convert it straight away, you know, like it's straight forward. There were these well fails. Like the issue is that let's say you're analyzing old world files. You might have like, you drove the way on the 80s.
00:29:38:06 - 00:29:59:14
Unknown
You might have like a mud report. Daily drilling reports. You might have, you know, annual report. Don't know, report some better report all these different documents that relate to the, well, like that are talking about it from slightly different perspectives. And there's overlap. But how do you incorporate all those files, and all the data in these files into like a single source of truth.
00:29:59:16 - 00:30:30:04
Unknown
So, so you got all these different sources and then you've got that extrapolated across time. Right? So you drill in the 80s, you go and work over in the, in the 90s site tracker, you know, a couple of years later you get the same mesh as a data that's hard to implement and like, an automated system that just does everything and, and like, and it's a totally different issue than, like, ingesting, semi-structured IDC, like CSV files from, from a, from a regular reporting system or whatever.
00:30:30:05 - 00:30:52:14
Unknown
It's like there's different views. Legal docs are my favorite because they're just plain text. I have no issues getting anything out of a legal doc because it's just a plain text document. And there's struct, there's structure, there's headers, there's subheadings, there's all that stuff. And then you go over to a drilling report and it's like first page wellbore schematic outputted from an Excel spreadsheet.
00:30:52:14 - 00:31:33:06
Unknown
Second page the actual wellbore diagram, third page time log of, you know, the daily activities that are in a table or that were structured in a table, probably in Excel. But then outputted without any of the the the table bounds. So and Lem doesn't really know whether it's table or plane, like it's just there's so different. And I don't think people outside of like, you know, people who are doing this type of stuff on a daily basis truly appreciate how excuse me, how hard it is because of the variety of data types.
00:31:33:08 - 00:31:55:09
Unknown
And then even then, it's like, you know, one drillers report looks totally different than another drillers report. They all have the same basic information in them, but they're structured in different ways, and they're displayed in different ways, and they're formatted differently. And all these nuances that that make it a huge pain in the ass on the back end to get the data out.
00:31:55:11 - 00:32:27:20
Unknown
I think I think what we really need, Paul, is we need someone like you to come out and, build a, an AI standard for, for drilling documents or something like that because, well, but like, think about this. Right. Like, what do I need to know from a wellbore schematic? And it's, it's a handful of things. But how do you take that image on a piece of paper with all the lines and all the crazy shit in a wellbore diagram, and translate that into something that a rag search would actually be able to render accurately.
00:32:27:22 - 00:32:49:19
Unknown
Right. And it's like I don't need the whole diagram, I just need x, Y, and z things extracted from this diagram and tagged to it as metadata so that when I search, it will show up. And so it's I think there's there's a really interesting space there in the future for the energy industry because you know, it's like okay, raw text fine structured numbers and a table.
00:32:49:20 - 00:33:14:02
Unknown
Fine. But then you have these diagrams and like logs or maps, but you have these like it's their images and there's important information in them. But for a computer vision or an OCR model to be able to understand what the hell is going on in those, it's almost impossible to an extent like you can extract the data out of it, but it doesn't have any structure to it.
00:33:14:02 - 00:33:31:18
Unknown
It doesn't have any like, logic behind it. And so it's it's almost like, hey, why don't we just, you know, the data is already in there. Why aren't we just generating a second PDF page with all the metadata about the prior page that we care about? And then that's what gets fed into the model or into the index or something like that.
00:33:31:18 - 00:33:57:15
Unknown
I don't know, I'm just, ranting right now, but I do think that, like, there's there's an opportunity in that space because people think that data is always, you know, PDFs or, or tables essentially, or csvs. And it's like, hey, there's there's data in this subset of data that it doesn't really make sense in a text based world, which is all, all things.
00:33:57:17 - 00:34:26:02
Unknown
Well, these days that, you know, how do we how do we handle that? Yeah, that would be, I don't I mean, yeah, thing is that a lot of these old files are like, they already generated all the data. Oh, yeah. And it's like. Yeah, all the old data point forward. Yeah. Yeah, it's it's wild. Are you doing anything like that with, you know, someone sends you, hey, here's, here's this acquisition we had 20 years ago that we haven't touched, and here's all the data for it.
00:34:26:02 - 00:34:41:00
Unknown
Can you help me figure it out? Yeah, that's that's basically I'm doing so you've got like offset reviews. So you might as in your drawing in an area you've never tried before. You need to look at it all the way I was, I just finished some work for and, and then sort of and then you've got like I say, acquisitions.
00:34:41:00 - 00:34:59:18
Unknown
Then you've got a lot of people in Europe, and we need to. You've not been on a whale in 20 years, and you need to know what's there. So when you go and you don't, you know, you know what you're doing. But as far as acquisition ones, I think that's a big one, because I think a lot of companies or a lot of them investors buy fields based on geology alone, and they just kind of ignore the drilling.
00:34:59:20 - 00:35:16:10
Unknown
They do. They look at the reserves and they don't look at how how much it's going to cost to get are those reserves out the ground. And you can only do that by looking at the whales. But if your whales are like your your old whale, your whales are like, you've only drilled that field like 70 years ago.
00:35:16:12 - 00:35:40:15
Unknown
What, you don't you can't make out like, you don't know what it's going to cost you unless you do the analysis of the whales. But I just completed work for a company. A European company. And it was, the files were somewhere from the 70s, some 80. Some some like from ten years ago. Some were two pages, some were 650 pages long.
00:35:40:17 - 00:36:06:06
Unknown
French, Dutch, English. And like, I don't speak any, speak English. That's that that that's just about it. And the units were like, you know, you get oil through the normal units, feet in barrels, and then you've got like meters and all that, and then you can like French units, like, like, some I don't even know what they were, but you get that sort of thing in like and some of the whales are vertical, some are like three side tracks.
00:36:06:07 - 00:36:33:06
Unknown
I, that's the sort of stuff. Yeah, yeah, there's a whole I mean, it's I guess it's probably worse in the offshore world just because offshore has been around longer than the shale stuff. But onshore conventional, I feel like is very similar to the offshore world. And the paper aspect of like my father in law has filing cabinets full of logs and, well, files and, you know, production cards and all of this shit.
00:36:33:08 - 00:37:06:08
Unknown
And it's like he knows where it is. No one else in that company has any idea that that data even exists. Knows how to how to access it or how it's organized or anything like that. But walk me in with Scott Caso or, you know. Yeah. Right. I was going to say walk, walk us through, kind of how that, how that went because I've done a very similar project to that, last, last year with a big offshore operator, and they gave us a shit ton of data that was all scanned banker's boxes.
00:37:06:10 - 00:37:40:20
Unknown
Shout out to Scott for doing all the scanning over at Caso. Yeah, we're looking forward to getting that work. But it's one of those things where it's like, okay, hey, you know, I don't whatever percentage of this is what I would call like a modern data set where, you know, it might be a PDF, but it's a structured form or there's structure and stuff to it, and then you start going back into like the 80s and 70s and 60s and it's like, oh, here's just a blank page of paper where the engineer did all of the calculations by hand, and it's like, what am I, what?
00:37:41:01 - 00:38:05:23
Unknown
What am I? How am I supposed to give that to you in a search result like no one. It's handwritten. Two it's it's handwritten math. So it's not even words, it's just symbols. And then three it's like, does that even need to like in what world would would you want that as a result? Right. Like trying to reverse engineer how to get the user back to it.
00:38:06:01 - 00:38:24:09
Unknown
And so like, it's just such a, I want you to talk to it because it is such a gnarly problem. So like, walk people through, you know, the trials and tribulations of that for like, the, the, the kind of difficult stuff. So I don't really I don't really do the right stuff for, the vectorization for these sorts of files.
00:38:24:10 - 00:38:48:07
Unknown
Like they just want schematics, they just want like, I don't know, maybe they want to map this one various output, but it's, it's output that the Jalan team can use it for any licenses or anything like that. They just they want they want their Excel files. But I start off with, like, you know, it's, you got your schema, you got your structure, your warehouse structure, and then, and I just, you know, get wet, smell that sort of thing.
00:38:48:07 - 00:39:14:00
Unknown
But they're far too big. And, you know, I've cut it down to what? What I would want to see if, you know, the information is actually used 99.9 times out of 100 when you're, when you're drilling, I usually use. So you got these files and I use like OCR to, to get the text out and actually like I don't use the Bayesian model models or anything like that because what I find is you have to iterate on it and you have to do all sorts of stuff and go back and back and back, and it's just it's just them.
00:39:14:02 - 00:39:39:05
Unknown
It these documents are too big. But it's a case of like once you get the structure, once you get a data model, then everything just kind of falls into place. Like, you can just you can quite easily work out those calculations are relevant, are not if they conform to your the structure if if if if said is necessary for their particular goals.
00:39:39:05 - 00:40:03:18
Unknown
If you know what I mean. So like the worst thing is you've got the OCR data, you've got the structure that you want. You can use LMS to extract, to to extract the structure. But you get into all sorts of issues with, like, with, with, like the files are too big, the schemas to bag, you get nested, nested like, you know, yes, you're speaking my language now.
00:40:03:20 - 00:40:37:11
Unknown
It's getting tables out of a PDF is much harder than accurately. Let me rephrase that. Getting them perfectly replicated out of the PDF into a table. CSV, whatever structure you Json you want is such a under appreciated problem in my opinion, because it is so hard. Like it's such a boring right? It's so awful. It's so terrible. And it's like I can tell if I look at this, why can't you computer vision model what the hell is happening?
00:40:37:11 - 00:41:00:06
Unknown
It's a very basic table. Like I've. I've narrowed it down to this one page. In this one document, out of a thousand pages. And all I want you to do is extract this table perfectly, and you still can't do that. And, like I've tried, I mean, I, I went back to trying like og just table from PDF tools that existed before LMS were ever around.
00:41:00:08 - 00:41:32:18
Unknown
Some of those work okay on certain documents, some of like and that's part of the problem, right? Is it's like there isn't a single source solution for that right now really that's remotely accurate in a lot of ways that I'm I'm cautiously optimistic that that's going to keep getting better, just like the language models have. But, yeah, I think ultimately everything ends up running on vision models because if they can get that cheap enough and accurate enough, it solves a lot of those problems.
00:41:32:20 - 00:41:57:15
Unknown
That a text based model alone can't solve. If they can do that. But the the jury's going to be out on that for a while, I think. But I am the vision models like, I just, I just, I just avoid them. I use them open source models mostly, even though they're less powerful and they're less than exciting because I find, like the, you know, open source models get you to maybe maybe that's 80% accuracy.
00:41:57:15 - 00:42:22:07
Unknown
Like what I need to do is like the accuracy levels are they need to be equal to or better than the accuracy a driven engineer would get. You know, when they're looking at the data manually. And the open source models might get you like 80% accuracy or, and these big models, the Geminis in that there's obviously you wouldn't I can't use them because of data privacy concerns.
00:42:22:09 - 00:42:59:12
Unknown
They'll get you like 95%, you know, whatever. But but the line of work that takes you to get from like 80% to, to to the over the threshold that you need is the same amount of work to get you from 95% to like over the threshold. So so most of what I do is, is deterministic stuff. So like you mentioned those tables and that you, you assume they've extracted properly, but it's because I have kind of like a graphical interface and I can, I can and because I have like a warehouse schematic, you immediately see if something's wrong, immediately see if it's wrong, and then you can implement that.
00:42:59:12 - 00:43:18:20
Unknown
You can even implement like, add tests and checks and evaluations and all that sort of stuff to make sure it doesn't happen again. And so you just constantly evolving this model that like if you have a table of trajectory points, like maybe there's ten pages, these are just pure numbers and they're just a plain text. Right? Like you say.
00:43:18:22 - 00:43:47:15
Unknown
But then if you plot on a trajectory and it goes like upside down and cross over and stuff like, you know, it's wrong and you can put it in text now, I think that's a big part of it, right? Is being able to build in like agents are great when they have the correct tests and checks and boundaries and like, but just letting an agent go free rein on something is not going to work the way that you want it to, especially in in the data world, if things what are you using?
00:43:47:15 - 00:44:14:06
Unknown
You were talking about open source models. I, we, we've built a, we've got a fine tuned model for petroleum engineering. Now that is passing the AP exam, that we're super excited about, but, that's one of the reasons we fine tuned it is because we realize that these, these foundational models are not going to be good enough at these domain specific, applications and such.
00:44:14:06 - 00:44:31:02
Unknown
And so, on the data side of things, I try to use open source models as much as I possibly can when I'm doing any kind of extraction or building any kind of a genetic stuff, just because I know if I can get it to work on an open source model, it will absolutely work on a foundational model.
00:44:31:04 - 00:44:52:12
Unknown
But I think it's an interesting point, too. We've been talking about this at the office that like, no one is thinking about how expensive these are going like, all of the tokens currently are super subsidized. They're unnecessarily cheap because they're playing the market share game. But like at some point there becomes a winner and the prices start going up.
00:44:52:12 - 00:45:13:07
Unknown
And you've now built all these automations and workflows, and now you're very risk heavy into whatever company you chose. And, I don't think enough people are thinking about that. But let's talk about your open source stack. Are you using like, vellum or Allama or what are you using to, to host these models? I am so for like prototype work for it.
00:45:13:08 - 00:45:35:01
Unknown
So I would test the system is using like, publicly available data, like not client data. And for that I would do some fancy like, like grok for Q, which is like so, so I would do something like that because it's fast and it's dirt cheap. Like it's so cheap. It's so, so cheap. Yeah. Like it's it's they got put by, I don't know throws in Nvidia.
00:45:35:03 - 00:45:52:04
Unknown
I saw a couple of months ago. So I don't know what's going to happen to them, but is like that's what I'd prefer to use. But I notice that their terms and conditions are like it used to be any other data retained. And it. I think they changed it then. I think they changed it back. And it's like I don't I can't really rely on that.
00:45:52:04 - 00:46:11:04
Unknown
If you're saying to people no, no your data doesn't leave it, you know, like data security. So what I tend to do is, it's open source, but I run it and I run on, like, cloud infrastructure because clients pay. Like, I would have thought that running things locally would be like your client would want that, but they say, no, that's a red flag.
00:46:11:06 - 00:46:39:06
Unknown
It guys go, no, no, no, you can't do that. It's got to stay. And like ISO 27,001 compliant cloud infrastructure, it's like okay so what I had to use them. I use them like I can never remember the name of them. One of these, these, like, GPU, things on, cloud infrastructure. I run them up for like a few hours as long as I need them, and then just shut them down and use them in the I can't remember, you call them.
00:46:39:06 - 00:46:57:08
Unknown
There's digital DigitalOcean. Do a couple of good ones. They did once. Like you just press the button and it and install the model installed endpoints and stuff for you. And they do one. It's just GPU infrastructure. But see, because I'm paying the bills, I set the clients for a fixed cost, like turnkey, like here's how much it's going to cost.
00:46:57:10 - 00:47:28:18
Unknown
And so I need to keep it. I can keep one of these cloud droplets running for, for, for like a year. I can't afford that as, so so that's why that's how I manage it. Now, have you had any pushback on specific models like we've, we've seen, some, some enterprise that are that will they've come out and said in meetings like, we will never use a model that came out of this country or from this company or anything like that or not, not yet.
00:47:28:19 - 00:47:53:00
Unknown
Even even open source ones, specifically open source ones. Why? Why is why that even they're worried that, you know, deep has some unknown, malicious something in there or back out under the hood or something, even though it's like if it's in your infrastructure, you should be able to spot that immediately. If anything, it's like going away, like, right.
00:47:53:00 - 00:48:13:08
Unknown
Like, no, I haven't, I, I've had anything like that. I tend to use it's, I don't even, I don't even like, I don't understand the, the kind of business model for these a big biggie, like, you know, your anthropic in your, OpenAI. Because I don't even know what's model. I, I don't know which one. It's like, if one doesn't really work, you just switch out the endpoint first.
00:48:13:08 - 00:48:44:00
Unknown
For one thing, it's not it. I am, I think I, I use it's either llama so like matter or the OpenAI OS, we the open source OpenAI one, the GPT OS. Yeah. Yeah. That one. Yeah. So that's a good one. I just, I just downloaded GEMA for, earlier this week on my laptop, running on a llama, and I've been super impressed with it.
00:48:44:02 - 00:49:01:03
Unknown
It's it's less than, like, I think it's less than 20 gigs. The full model. And so it's and it can run on. I mean, it's basically built for laptops, or to be able to be able to run on a laptop. They've already got a bunch of quantized versions of it out there, which makes it run even better.
00:49:01:03 - 00:49:29:14
Unknown
But it's, it's fascinating because like on the foundational side, it's more tokens, more context. It's bigger models, right? More capabilities, more tooling, more logic, all that stuff. But then in the open source side, it's like this dichotomy of better models, but they're smaller or they're more efficient, or you can run them, you know, in places that you absolutely could never run, you know, a, a GPT five, or anything like that.
00:49:29:14 - 00:49:50:14
Unknown
And so I think it's fascinating. I'm very curious to see how this plays out because like, at the end of the day, I don't need a foundational model to go do some a genetic extraction on a specific page where I'm telling it exactly what it needs to it. Like I don't need a foundational model for that now for like routing for my queries into my my rag to figure out which tools or which path to get.
00:49:50:14 - 00:50:14:15
Unknown
Like, I think that people, you know, people see GPT or quad and they think, oh, it's just one model and it's one thing doing one thing, and it's like, no, this there's so much in between. When you hit enter on your prompt and when it starts returning an answer that the user never sees. We've talked about this before on the show, but it's like that fuzzy middle layer of like the tool calling the orchestration the agent routing.
00:50:14:15 - 00:50:37:00
Unknown
Like there's so much happening in there. And like you don't typically get a lot of that with the open source models. But if you're just, you know, trying to get some automation done, you don't need an open source model to do that, especially on the data side of things where you're just extracting or reformatting data or whatever. So that's that's really cool.
00:50:37:00 - 00:50:56:03
Unknown
If you haven't played with any of the local stuff like Obama or BLM or any of those, I would highly recommend, just messing around with it, because it's like being able to run an agent that does, you know, that's doing extraction or, you know, different things like that, just locally is it's fun to see. It's it's not any faster.
00:50:56:03 - 00:51:21:19
Unknown
Of course it's free. But it's it's pretty wild how good some of these small models are getting. Is the general one good? Is it like like it's usable? It's really good. No, it's, it's a all. I think the gem client is out of Google, so it's based off of the Gemini models. And the Gemini models, to my understanding, were originally trained for like, classification.
00:51:21:21 - 00:51:46:23
Unknown
And they're really, really good at that. And so like being able to feed it in like a data set and say classify these as, you know, X, Y or Z. Beautiful. I think the GEMA for one is now multimodal to you in some ways. Or it has that capability at least. And so yeah, the again, super lightweight model now has all this functionality that these bigger models used to have.
00:51:46:23 - 00:51:54:15
Unknown
And so I'm very curious to see kind of how that all that stuff evolves. But
00:51:54:17 - 00:52:18:19
Unknown
How do you like just with your experience, how do you kind of see AI feeding into the industry? Where do you see it kind of touching first, and where do you think it has the most kind of applications? At least this type of AI? Because, I'll, I'll preface everything with language models because we're AI is. Yeah, yeah, yeah, probably used, as CRM driven.
00:52:18:19 - 00:52:49:23
Unknown
I mean, churn is not very good. Like you say, everyone is using Excel spreadsheets and like, databases came out, you know, got 4 million other entities, 25, 30 years ago. You also have the 13 tops in them yet as in like and so I so our lines are. Where does it go first I think in these small cases, like people are, like, I'm like taking a picture in your phone of a weather schematic and it's drawing the picture.
00:52:49:23 - 00:53:16:07
Unknown
Okay. Like, but that's not a major. It's. I don't know, you say that, but the person who's having to build that in Excel by hand is probably like, this is the best thing ever. That's the thing. It's like there's there's a lot of low hanging fruit in this, you know, in this sense across the industry of like, why is this engineer who we pay that hundreds of thousands of dollars a year doing the stupid menial task.
00:53:16:09 - 00:53:39:08
Unknown
But the the standalone impact of that across a company doesn't isn't like sexy, right? Like it's not hugely impactful, but it's giving that engineer their time back, right. Like it's a weird trade off I go to miss. I get messages on my emails, schematics, system in the case design system, I've got these, feedback buttons and you click the button and you just put in feedback and it's easy to do.
00:53:39:08 - 00:53:54:17
Unknown
So people send me that, they give me feedback and I get these messages every so often. It's like, you know, I think, oh, maybe, maybe nobody's using this system. I've got analytics and people are using this system, but it's like, maybe people aren't actually using it and it's only point of view in it. I got a message yesterday.
00:53:54:17 - 00:54:21:00
Unknown
Like the guy who obviously it probably works and you know what? Someone in the there somewhere, all caps. It's like, that's it. This is this is excellent. Thanks very much for. Yeah, yeah, yeah. So now I mean, and and so people are people do want these tools, but I wonder if the driver comes from lower down, you know, for actual tools at work, it comes from like saving people downtime rather than, like from above.
00:54:21:02 - 00:54:42:17
Unknown
That's I think that like, that's the thing you get a rag or an automation or anything like that in front of the person who is currently having to do that. It is like the best thing since sliced bread, right? But then you try to communicate that ROI to management or to upper level management or whatever, and it's like, well, but that's what I pay them to do.
00:54:42:19 - 00:55:12:16
Unknown
And it's like, really? Are we I mean, but but maybe you don't have to have as many of them in the future because they can actually do the parts of their job that you hired them for not spending, you know, a week filling out permits or reports or, you know, like all this stuff. There's so much manual work in our industry on the back end, you know, and most of it is literally just moving data around or filling out forms of with data that already exists in a database.
00:55:12:16 - 00:55:22:20
Unknown
So it's like, why is a person having to do this? Yeah, yeah. It's it's it's yeah, I agree, it's,
00:55:22:22 - 00:55:43:21
Unknown
I don't know, I don't know, it's it's a strange industry and like this. Strange, that's for sure. Yeah. But I think you're right. I think, like, getting getting those tools that make the average, like, the end users life easier ends up being a much easier. Like it's sticky, then they keep using it, right? And then they like, then they want more, they want something else.
00:55:43:21 - 00:56:04:13
Unknown
Or it's a lot easier to start, you know, start small like that. That's what I'm so I was I was always of the opinion I was making these other systems before I had that system that just a service. We send the data, I send the results. But I was I'm focusing on trying to get like SaaS software and doing analysis as a service and all that sort of stuff.
00:56:04:15 - 00:56:26:09
Unknown
And my conclusion, and the same conclusion that I've had for the last 15 years was like, the drilling industry is really conservative and nobody really, you know, nobody really wants to adopt new technology when c when like assure them it's even like doors open slowly. Right. And it's a challenge to open doors and get these conversations going. But when are you in the system like how the system works?
00:56:26:12 - 00:56:58:20
Unknown
Here's the way I analyze your data. It's you know, you assume they're conservative people, but as soon as you see it and they go, okay, that's fine. It's it's like a five minute conversation. Be like, okay, we'll send you some data tomorrow. You can test it and like, okay, we'll use that as an they're not slow. They're slow to adopt a large part of the reason I, I wonder that people are still using Excel is because of this, this difference in this communication problem between, like, you know, data that a lot of the software isn't just isn't it's just not appropriate for for the drilling guys, it doesn't make life easier.
00:56:58:22 - 00:57:21:13
Unknown
But as soon as you show them something that does actually make the life easier, it's like it's sold. It just sells itself. And so, so yeah, well, that's the thing. Like especially on the op side, like making their life easier means they get home earlier or they don't miss, you know, a kid's sporting like it has tangible impact for that person.
00:57:21:15 - 00:57:40:19
Unknown
And so you're right. Like it immediately is like, oh yeah, I'm going to use this every day now. You're not having to like get a phone call two in the morning, some some from three. And you're not having to stand up justifying how you made the same mistake. Like you like in mistakes can be embarrassing. Or if. Yeah, it just makes it easier for.
00:57:40:22 - 00:58:02:16
Unknown
Yeah. If you were in real world benefit for the people in right. Yeah. No. And then, you know, again I struggle with like the fact that this is literally like very obviously valuable to the individual, but then trying to get that value sold to the company at a company level is a whole nother like it's a whole nother challenge.
00:58:02:18 - 00:58:21:00
Unknown
Yeah. And that's the thing is like the way I schematic systems and the case in design. Yeah. It's wonderful that you've got this medal, you know, software people can use, but drilling engineers, drilling guys, guys in the rig, they don't walk around with company credit cards in the pocket like they it has to pass for, procurement and all that sort of stuff.
00:58:21:00 - 00:58:49:07
Unknown
And you get ten layers of approval. It's, Yeah, it's something definitely to think about. And then these things, well, I'm back. Energy is important. He had a speaking of importance of it. Yeah, I guess I haven't had a beautiful day in Houston. Will probably have all year, but such as life, but, so yeah, I'm coming back into it.
00:58:49:09 - 00:59:06:04
Unknown
I mean, I know we're running up on, you know, probably time here shortly, but I, I do want to ask Paul, like, I mean, kind of it's all as I'm doing like to talk about the journey. Like, where do you start with, you know, getting into like, the coding, like what are what was like your first language? I saw, like me on LinkedIn that you started with, PHP, Laravel.
00:59:06:04 - 00:59:40:01
Unknown
So like that, like Lamp stack stuff and then how it evolved now with like cloud computing and all the other advancements. And yeah, it was Lamp stack, actually as well ago that I remember that. But yeah, that was like 20 years ago. You're in PHP and you and and it's like echoing out the HTML from the PHP and then, you know, to build these systems and then, you know, when I was in drilling out of it for a bit and then back into it and, and like, it's amazing how the, the drilling industry doesn't change at all, like you're still using things over, but the improvements in software and then these frameworks is like
00:59:40:01 - 01:00:01:13
Unknown
it's unbelievable what you could do. Like and I started using Laravel and even up until not quite recently, it was it was PHP and Laravel, and I did that for the data analysis work. So like I'm obviously not like a hardcore like data type person, but it's specific to, you know, it's only specific to these particular problems.
01:00:01:15 - 01:00:19:06
Unknown
But because it in, in like Laravel just in PHP, it's just so easy to get up and running like it's, to build a web app that works and it functions and they don't need to deal with any sort of boilerplate stuff. It's just you focus purely in the back end and like, fair enough. The back end stuff is not as efficient.
01:00:19:06 - 01:00:43:08
Unknown
There aren't as many frameworks and are like, packages for data analysis or engineering calculations, but then it kind of means that you're, you know, instead of instead of just typing a histogram, an important a histogram package or whatever, you're you have to do it manually so you get a good understanding of underlying, like what you're actually trying to do, rather than some sort of abstraction layer.
01:00:43:10 - 01:01:15:10
Unknown
Right. PHP and Laravel and and like, JavaScript in the front end and then like relatively recently switched to Python, just, just purely for the extensibility here. Well, the tools. Yeah. You can how can you you can't do that. You can't not do that. You can't. You have to switch to to to Python. I don't know how you if there's a better way than Python and pedantic and, and and those things you use machine learning or any sort of machine learning.
01:01:15:12 - 01:01:40:19
Unknown
So how did you get into coding though? Like was it from university or it was just it was like magic. Okay. It's yeah, I did Matlab projects for, for like, a couple of companies and my own project. I used to have these, web websites and I would make enough, you know, this was in the time of Google AdSense ad where whatever the advertising one was and you, you know, you got clicks on your site and you got advertisements.
01:01:40:19 - 01:02:01:05
Unknown
So I used to have multiple ones of these and I would make a bit of money, but not not much. And I just, I just like doing it. And then, and then that these various projects and then I don't know, like I like building things and you can not understand how to do any sort of programing if you if you like building things these days that you can you get.
01:02:01:07 - 01:02:22:02
Unknown
I feel like people see like people like us and you're like, how does a mechanical engineer get into data or coding or anything like that? But, I think one of the things that like at least I see is, you know, or very logically minded people generally, and code is just logic, right? Like it's just a set of instructions.
01:02:22:04 - 01:02:51:12
Unknown
What's some advice that you might have for current, you know, engineers, mechanical engineers, whoever that might be in school or coming out of university, around like being familiar with code or anything like that, like I think first off, like, don't be over rely on these, you know, the AI stuff. I think that's dangerous. And you cause too many problems yourself and it might look good, but when you try and tweak it and stuff it, you can run the problems.
01:02:51:14 - 01:03:19:01
Unknown
But I would definitely I would very much see like. Regardless of what your career is or whatever, like if you get into programing, if you're doing like pure programing, you're going to be limited to maybe you're going to be limited, like building software. But if you have like domain expertise and engineering or like accounting or anything like that, and you can program like you're, you're able to leverage computers to, to do things that most other people in your domain can't do.
01:03:19:01 - 01:03:38:08
Unknown
So, and it's also interesting, you can do a lot with it, like if you like coding things, where my daughter calls, she calls me a fixer, and it's like, we just like to fix stuff. And, I'm like, that's the best description. I'm, like, honored to have that as my, nickname in the house. But it's true.
01:03:38:08 - 01:04:01:21
Unknown
Like, we just like to make stuff and solve problems. And like, so many problems these days are rooted in data. And so, like, as an engineer, I almost feel like you're not doing your job to the fullest capability if you don't at least understand the data right. You don't have to understand how you got the data or how to generate it, but have that foundational understanding because like data is everywhere.
01:04:01:23 - 01:04:21:08
Unknown
And, but yeah, that's, I don't know. I just think people, people group them like, it's either you're coding or you're an engineer and it's like, I completely agree with you. The future, especially now with how easy it is for an average person to pick something up and try and vibe code and app to solve their problem, or a script to solve their problem.
01:04:21:08 - 01:04:48:03
Unknown
Like if you just have a basic level of understanding of code and, you know, kind of general architecture, what a database is, what an API is, how they work, you can go really far because it is like, is it prior to that, a few weeks. Like, I mean, he he used to screw around with AR and I remember like chat with him when he was still on on, on Twitter and stuff, but like he's got a full on like SaaS offering you that.
01:04:48:05 - 01:05:10:19
Unknown
He's like building on he's got the MCP and API and like just all these thing. It's just it's wild. So I listen to his a podcast. You did this with him? Well, I wasn't, I wasn't that and I saw someone he posted, yes, yesterday, the day before. But the MCP system that he's got and it's it's it's not your it's not your, it's not your standard.
01:05:10:19 - 01:05:31:12
Unknown
Like, it's like it understands the domain. Like he's an actor and that's and like I was think I was thinking about this morning to just like I mean I think my, one of my little taglines on LinkedIn has always been like solving business problems with technology. Like, like, don't put me or anyone us in a box is that you're a data engineer or you're a analytics engineer, you're a data scientist.
01:05:31:12 - 01:05:54:20
Unknown
No, I'm I'm what I need to be at the time. I might be an IoT engineer tomorrow because that that's what you need to solve this problem. Like but I am, I can I can love, figure out and leverage technology. And I know what those pieces can do. And then it's just a matter of. And now, like John said, like the the barrier to actually like diving into that from the technical side is so low now it's like, oh wait, hey, I need to connect to Azure IoT hub in blue, blah blah blah.
01:05:54:20 - 01:06:15:04
Unknown
Like it will lay it out for you, but you have to know what you're trying to solve and what those tools do. But as long as you understand what the tools, the right tool for the job is and the problem you're trying to solve, like implementing it is like nothing. I mean, like I got these PDF reports for a, for an operator, like, and I was like, use AI.
01:06:15:05 - 01:06:35:04
Unknown
So cloud could use auto and, you know, matplotlib and or plotly or whatever, you know, thing. And like, here's what I want about it. Freaking did it like an I under an hour I had like this thing that would have taken me days is there and nudge things here and, you know, create the data frame logic and all this stuff.
01:06:35:04 - 01:06:59:16
Unknown
And it just it's just wild. It's a crazy time. Do not speak dare not speak to kind of that point. And the previous thing is like what like what advice to like somebody in engineering who's starting out in program ever. And it's like the, the, the main thing is I don't even know what you call it, but it's like, you know, when you prepare an ERD like a, like a, like a data model, it's like the pros, I'm about backing me.
01:06:59:16 - 01:07:17:14
Unknown
I've got like all these sheets of paper of these IDs. And it's like, once you've done that, it doesn't matter if you're using Python, doesn't mean PHP as a matter of using the AI. What is the not? You plug it into whatever it is you're doing and the code writes itself. But but like structuring that, having that like underline.
01:07:17:16 - 01:07:38:02
Unknown
Yeah yeah yeah. Like you can go into that, you go into business and you can map their business map market department, you can map a calculation system or you can map a data structure. It doesn't matter. It's the same thing. So like whatever that is I don't even know what you call that whatever that is. But but again it's just turning the business into, you know, like you're just codifying the business essentially.
01:07:38:02 - 01:07:57:15
Unknown
Right? Like, yeah, I mean, but but that's you. But I think some people talk about it too, like, and I think who was it? There was a, some guy was a relatively bigger company. Oh no, not big big, but right. There's a, why is it called, you know, it was a thing. My son did it for a little while.
01:07:57:17 - 01:08:19:04
Unknown
It's like this little, exercising, but it's really a program. But I think this one of the founders was, like, the main coder behind it. He's like, I have shipped so much code so much faster. But he's like, I don't understand it nearly as well as I used to. And I, I don't, you know, understand I haven't done that work to take the business into that, like as thoughtfully as I used to.
01:08:19:04 - 01:08:40:18
Unknown
So he's like, you kind of worries him because he hasn't done some of that, you know, mental side of it as much. But, I mean, I think the pros outweigh the cons in some of those. But you have to, there's often some trade offs where it's like, if you were more thoughtful about how you were turning the business logic into like a data model or code, then, it's even more important or more valuable, I think.
01:08:40:19 - 01:09:02:13
Unknown
I think the crazy thing now is that he can just ask her to explain how his code base works, and it will. It's it's a very weird time because I have experienced the exact same thing where it's like, I've been working on this big a genetic extraction thing, and I run it and it and it works. And then I'm like my, my senior devs like, oh, well, what did you do?
01:09:02:14 - 01:09:18:16
Unknown
It's like, I'm not really sure to be honest. I just had, you know, a day long interview with Claude where we went back and forth and I did what I do and beat my head against it until I got the outputs that I wanted and now it works. And then it's like, okay, Claude, explain to me how this works.
01:09:18:16 - 01:09:38:18
Unknown
Generate a mermaid diagram for me. I do all of these things. And then that's how you. It's a weird, but it's weird not knowing, like what's in the thing that you built. It's a very weird place. But I think, Paul, one of the things you hit on that I don't think enough people appreciate is because we get this question all the time.
01:09:38:18 - 01:10:04:20
Unknown
Like, what language should I learn? It's like, it doesn't matter anymore. The code doesn't like the language. The code is written in generally does not really matter anymore. Because like we, we we migrated our front end from Ruby to react in two and a half weeks, which is crazy. Like that's a distraction project just a day long, you know, in the old world, right?
01:10:04:22 - 01:10:30:11
Unknown
I mean, we probably maxed out all of our token, use for that those two weeks, but it's like at the end of the day, we were able to do it, and it's like, it's just crazy to say out loud because, you know, there's there's so especially on the engineering side, there's so much industrial software that's still written in like Fortran or COBOL and like, the people who know how to code in this language are dying or already dead.
01:10:30:13 - 01:10:51:18
Unknown
And so, like, what happens, you know, like I had a professor who worked at Lockheed doing missile, you know, missile defense systems and they still all that's this was 20 years ago. But all that was done in Fortran. And I was like, why? And she said, well, that's what management and the C-suite know how to code in. And so they're comfortable coding in that.
01:10:51:18 - 01:11:10:13
Unknown
And that's what, you know, they want to code in. And so it's like some of those problems like that whole kind of knowledge gap of like, well, how do we transition, you know, the incoming people into this language that's been dead for 30 years. That kind of goes away, at least with some of these language models and stuff.
01:11:10:13 - 01:11:33:06
Unknown
But it's, it's it's just a wild time, around all of this. Now, I agree that m the where's that where schematics are for iris and data and, like, images. And then I changed it to react. And so it was all front end stuff and then I saw somebody share me if I do. And I was like, oh, you know, a fast computer obviously had too much rubbish on it.
01:11:33:06 - 01:11:56:01
Unknown
So I had to switch it to the back end for the, you know, for the integration, the with the like language models as well change the Python. So it's changed three times already and the people on the front end don't know. Right. But other than maybe it's better faster. Yeah. Right. As long as it works. Yeah. What, what what coding, tools are you using on the LM side?
01:11:56:01 - 01:12:21:18
Unknown
And then we can wrap up. I wanted to ask you that I use cursor. So I've tried, like, I know there's a lot of know you mentioned Claude. I've. I've tried called. I did not like the total detachment thing. I'm not, I like cursor. I like, if, I turn off the autocomplete now, like, I just, I turn it off a few weeks ago, and I prefer it like, short term, maybe slower, long term.
01:12:21:18 - 01:12:48:05
Unknown
It's faster. I only use the agent stuff for like boilerplate. A lot of front, front end stuff. Just just repetitive stuff. And and if I've got like C, I'm using like an abstract sort of thing and, and I've got the structure right, then I can use agent thing, but it's like cursor with a lot of features turned off and, and you know, like if I'm at if I'm stuck in something, you can ask a question like, you know, what's the right and you know, what can I do here?
01:12:48:05 - 01:13:20:15
Unknown
But just to know, that's cool. Our, I think our, our most cracked out devs use clod in cursor. But I, I agree with you because I use cursor pretty much all year last year and then in I think December we got our cloud code instances. And that's what I've been using mostly. But I don't like it is a it's a weird feeling being abstracted from any of the data or what it's doing or anything like that, because then you genuinely have no idea what the hell is happening.
01:13:20:15 - 01:13:38:17
Unknown
And so when you get yes, it might work, but you're at the end, you're like, what? What did it do? How does this even work? But having having the cursor, having clod in cursor at least lets you be able to see all the files you can click into the files, you can see what it updated and all your diffs and stuff.
01:13:38:17 - 01:13:58:18
Unknown
And so I, I agree with you especially, you know, I tell people I'm a mechanical engineer because I'm a very visual person. And so like I, I like to see especially on data projects. Right. Like I want to see the data output. I don't want to have to ask you to see the data output and then you can't render it in CLI because it's a Json or whatever.
01:13:58:19 - 01:14:23:13
Unknown
That's yeah. No. Highly recommend recommend for cert. Definitely. Definitely like it. All right. So all right. But I will speed round round. Okay. So I was like in the air. You're is it is air as I say it. Scotland. When I was looking because my Scotland geology is pretty poor. You're not really close to Edinburgh or Aberdeen.
01:14:23:15 - 01:14:41:21
Unknown
So, I mean, there's a, Is that where you're from, or is there a good oil and gas? You know, footprint there and not really. No, there's not the zero here. There's nothing here. The only oil and gas is in Edinburgh, which is like 4.5 hours away. Sorry. Aberdeen. Sorry, not Edinburgh. And a 4.5 hours away.
01:14:41:23 - 01:15:01:12
Unknown
Edinburgh is outside the country. That's where a lot of like, well, a lot of things on their air here, there's like tourism. Okay. No, sorry. That's so it sounds similar to I went to Ireland a couple summers ago. I'd like to like do and where? It's like, you know, very small, but, you know, pretty popular, you know, for some tourism and everything, but.
01:15:01:14 - 01:15:26:03
Unknown
Okay. Well, but if, if we were going to take a trip to Ayr, where where would we be and what would be our drinking has. Yeah. Where would you ask? I have young children. I don't go much. We and we do to. We feel we feel that. Yeah. The my preferred place for eating McDonald's because they can make a you know, they make a mess, but nobody cares.
01:15:26:06 - 01:15:53:04
Unknown
Other kids are screaming to like, the if you come to Ayr there's, there's all sorts of like traditional unnatural and there's also it's tourists go to Edinburgh. But if you come, come to that side of the country like I'm not getting political, but like Donald Trump's got his place to just there and it's like really fancy and and it's like, oh the there's a few whiskey places and, tradition is Robert Burns like famous poet.
01:15:53:04 - 01:16:12:15
Unknown
He's all around here. So it's traditional Scottish stuff. But to be honest, I think we just go to either McDonald's or Indian Indians. Like, Scotland's good for Indian takeaways and stuff. So what's something interesting on the menu at McDonald's that would probably surprise us, do you think? Yeah, I get a Big Mac, my kids get chicken nuggets.
01:16:12:15 - 01:16:30:21
Unknown
Happy meal. It's a there's nothing like there's nothing. So I know some places you go like different parts of the world or there's like a more some local spin on it. Is there any kind of local dishes or anything? No. No. They, they if they do spend, it's like a like, no they don't. It's the McRib.
01:16:30:23 - 01:16:52:01
Unknown
Yeah. I, I mean, I don't want to do that. Okay. Fair enough. No. There's nothing. Do you. What's your favorite whiskey? I like him, love Laphroaig. It's another good whiskey drink, right. You can drink too much. A bit too easy, but like I, I bet from a Laphroaig, like a different. It's like it kind of tastes. And I don't know much about whiskey.
01:16:52:01 - 01:17:05:22
Unknown
Yeah. I've got a bottle of Laphroaig for my birthday that I still have to crack. Like we were talking about this weekend with, with my dad. I was like, I've got a couple nice bottle of scotch. If I, if I actually like scotch. Yeah, yeah. Me too.
01:17:06:00 - 01:17:26:19
Unknown
Like, you have this. I don't know if it's been there for for ten years. I think half the box was drunk at young kids at the end. You don't get a chance. No. Well, then you feel like I randomly found a bottle of, it's. I'm a big bourbon, or I was. I don't really drink that much these days, but, used to be a big, like, bourbon fan, bourbon collector.
01:17:26:19 - 01:17:51:02
Unknown
And I found a really good bottle in this random, like, corner liquor store in New York before I got on a flight, and I almost left without buying it. But I went back because it's like you'd never see them for sale. No less. But it was like, I don't know, 300 bucks or something, 400 bucks and, so then I got home and I was like, well, when the hell am I going to drink this?
01:17:51:07 - 01:18:10:00
Unknown
Like, because it's like I start doing the mental math of like, okay, it's this many ounces. And then it's that much per drink. And I'm like, okay, well, that's a lot more expensive than my, you know, $30 bottle of, off the shelf bourbon. But it, yeah, I have a very similar story to that. I feel like.
01:18:10:02 - 01:18:35:17
Unknown
But we, we, I think using it on special occasions and celebrations and stuff is a good way to still drink it, but not just drinking it casually because I don't like buying things and not enjoying them. I'm a little more technical. On what's your favorite cloud platform I used to use? I don't like AWS. That is I mean, it's just it can do everything, but it's it's just me.
01:18:35:17 - 01:19:03:13
Unknown
It's like, is it too much faffing about? DigitalOcean is good. But but recently I moved to railway, and I think it's kind of like Heroku or some like, but nice. I think we've heard a couple that from a couple folks was at the country these that I think or I don't remember. We've got some of our actually our growth team is using cloud code and then pushing the deployments on railway to build internal apps and collide.
01:19:03:13 - 01:19:23:21
Unknown
So that's that's pretty cool. That was the first I had never if I'd heard it, I hadn't paid attention to it, but I had never heard of of railway until that point. But I've been helping some of them debug APIs and search on it, and it looks awesome. It seems incredibly easy to use. It's one of these things you're like, is it going to dirt cheap?
01:19:23:21 - 01:19:45:12
Unknown
But no. But like in a year's time when you're when you're stuck on, is it going to just 100 times the price? So I don't know. It's good. It's really good though. Paul, thank you so much for, spending your your afternoon with us. Yeah, yeah. Thank you so much for that. Some of these were able to get you on, you know, from post, production here pretty quickly.
01:19:45:12 - 01:20:04:22
Unknown
Another other time. This is not so easy. So thanks for, working with us. It's, It's been a pleasure. Thanks, guys. Again, new energy bites. Merch is is available over on the doghouse. Please go. Grab that if you want some. Also, please make sure you guys, subscribe and like and share and do all of those fun things.
01:20:04:22 - 01:20:10:15
Unknown
But, Paul, thanks again and we will see you guys next time. We appreciate it. All right. Thank you.
Unknown
Welcome to another episode of Energy Bites is an international episode. We only have a handful though, so far. So we've got, Paul Brown joining us from here in Scotland. Is that correct? Yeah. That's right. You. Thanks for. Thanks for having me. It's so funny because like, literally I can see you make a post on LinkedIn and we get this set up in like a day, and then we go some other people I've been trying to get for two years.
00:00:23:21 - 00:00:43:07
Unknown
And they say we have to talk to, you know, legal and all this and you know, it just takes a lot of red tape to get it to make it happen. So they're across the highway. We can't get them in the office, but appreciate you joining us across the pond. Yeah, but we were already having some great conversations before we got on here.
00:00:43:09 - 00:01:04:22
Unknown
But I know we were kind of leading in with, just the wild ride that is oil prices right now. Because when I checked last night, it was 113 and 112 and WTI was above Brant, and now we're down to $93. Just because Trump opened his mouth and, so, you know, you know, here we are.
00:01:04:22 - 00:01:25:07
Unknown
But, but yeah, I mean, I know, Paul, you were saying it's, you know, a lot of this has caused some, logistical issues out, your way. Yeah. We, we have outside the, the the petrol stations, the gas stations, like we have kids are like 15 minutes, you know, half an hour long, and you get there and and there's no petrol on the pumps.
00:01:25:07 - 00:01:50:01
Unknown
And like, at the is, I don't really know what's going on. So I don't know if it's the same there or what, but yeah, we're not seeing obviously that shortage like that. But I mean prices are definitely going up. But like I said, you know, I was wasn't thrilled to put $4, a gallon in my tank, but you guys are probably at least I think we are.
00:01:50:03 - 00:02:25:04
Unknown
15 or $20 a gallon would work out something like that. Like it's it's, I mean, I ride my bike like David all. We are. We are spoiled by cheap energy here, especially in Texas. But it's also, I think, cool and important to see what else is going on across the rest of the world and how things like, you know, them blocking the Strait of Hormuz, which has, I don't know what percentage 30, 40% of the world's oil goes through that strait.
00:02:25:06 - 00:02:52:13
Unknown
How that has trickle down effects. Not I mean, I wouldn't even call that trickle down. It hasn't been that long and it's already impacting supply chains across across Europe and Asia. And so that's, no, I think that's you know, like I said, in in America, we get so we're just so used to that. We've had we've gone two generations now since there was any kind of like oil embargo or price shock like that to where the supply chain was actually disrupted.
00:02:52:15 - 00:03:21:05
Unknown
And so we don't know. Right. Like we don't know what that's like even though our parents do. And so it's part of me worries about that because it's like, hey, people don't think about energy independence and how important that is or the effects of that, that, that, that has on your life as an individual. You know, I guess the closest thing we got to was that the the freeze a couple of years ago where we were, I was without power for like almost a week.
00:03:21:07 - 00:03:44:02
Unknown
And then you're like, oh, yeah, this is important. Heat is important. When it's snowing outside, we don't have like, we I don't remember any short. I'm sure there were some, but not that big. But we now have like something like this, the highest energy costs in the developed world. And and that's, three times higher than I think it's three times higher in the US than the US.
00:03:44:02 - 00:04:03:02
Unknown
It's because of all the, you know, like the push for renewables and stuff. There's abandoned drilling, been abandoned drilling for the last ten years by banning them, issuing new licenses and all sorts of legal stuff to prevent drilling. And then and it's resulted in like just the industrialization. But you can't compete in manufacturing, you can't compete in industry.
00:04:03:04 - 00:04:28:04
Unknown
There are no engineering jobs or a very few engineering jobs here. The oil industry is just, you know, just just shutting down. It's energy independence. Like you don't think about it until it starts to have an impact on your, you know, everything. No, absolutely. That's, Well, that's the thing. It's like energy is the one of the biggest inputs to all manufacturing costs, right?
00:04:28:04 - 00:04:51:10
Unknown
Like that's one of your biggest costs and anything that you manufacture. And so it literally trickles down all the way, all the way down to, you know, it being trucked to your house or to a warehouse or whatever. And so it's, yeah, you're right that like, you get kind of lulled into a false sense of security until it is knock knocking on your doorstep and you don't have power.
00:04:51:10 - 00:05:19:12
Unknown
You can't get gas. And you're like, oh, this is very fragile. Well, before we jump in, I'm going to shamelessly plug the podcast because we finally have merch. Bobby and I are repping the, the hoodies. Bobby, give us a little twirl so we can check out the back. So the back says data are pain. So everybody in the industry who has ever dealt with our data, I think, knows and appreciates that.
00:05:19:14 - 00:05:39:06
Unknown
But, those are available on the doghouse over on Callide. If you guys want to check that out, we're super excited to have some merch. I've also got we've got an energy bites hat that I'm currently wearing, and there's also a, short sleeve t shirt, because summer is already here in Houston, and these cities are probably being retired after this episode at least until next.
00:05:39:07 - 00:06:05:02
Unknown
Next, December. I'm pretty sure my wife's going to steal it. Yes, my mine have all been stolen already by my daughter and my wife. So, be aware. Anyway, let's jump into it. Paul, tell us. Tell us a little bit, about yourself. How did you, Yeah. How did you get started in either the energy or the tech side of the the industry?
00:06:05:04 - 00:06:25:07
Unknown
So I am, I studied mechanical engineering at university, and, Let's go. You're in good company, Paul. Yeah, yeah, I think so. Yeah, yeah, it's him, but on the side I like. It's about, you know, I worked in a couple of, I did quite a bit of software development for, like. And the engineering companies. I'm on projects.
00:06:25:09 - 00:06:53:23
Unknown
I went to drilling and I worked as a drilling engineer for, like, internationally, deepwater exploration and like, so, like, I worked in Mediterranean, North Africa, West Africa, South America, Caribbean, the UK, North Sea and so like drilling engineer West I engineer, night company man, all that sort of stuff. And then I took a break from the industry, like, just time for a change.
00:06:53:23 - 00:07:15:05
Unknown
And I started a small manufacturing business where, like, the goal was to take kind of low tech. Yeah, not necessarily low tech machinery, but kind of dumb machinery. And, but apply, you know, like engineer, engineer. The process, to, to make it more efficient. So I spent years doing that, spent like five, six, seven years doing that.
00:07:15:07 - 00:07:41:12
Unknown
But I spent most of my time developing software, developing automated machinery, and got to the point where the like, the administrative overhead was, was like nothing like next to nothing. The the, the place ran itself. And, and I eventually sold the business, and to focus on the data side of things. And so for the last few years I've been, I've been them to, to in work, data work and in drilling and upstream.
00:07:41:14 - 00:08:19:05
Unknown
So things like, taken, taken like an operator or I read contractors data, building a system to analyze it and then providing the analysis as a service. Right. So, like comparing rig versus rig performance or, or checking to see if less lessons learned had been implemented, regression or whatever. Building systems for, for like oilfield upstream businesses like business data systems, like reporting systems and that and but more recently, like I'm focused on two things and the, like the big one is taking legacy data, like taking these old bio files that are, that are everywhere.
00:08:19:06 - 00:08:35:22
Unknown
Taking them, analyze them in my systems and providing the, the, like the client with the just the results like so the prepared well montages and the prepare the prepared you know output so that they can actually use it. And that's where I've got the, the Wells schematic.com system that's used by, you know, that's why I developed that.
00:08:35:22 - 00:08:55:06
Unknown
I needed to I needed the, the use for my own my, my own work is connected via API and so I just, I just released it and it's used by quite a few people. So that's one big thing, the legacy data analysis. The other thing is, I'm doing workers do quite a bit of work is, like consultant.
00:08:55:09 - 00:09:15:13
Unknown
So you've got these oil field data projects or drilling data projects where, where the, the, the drilling guys know everything about, you know, drilling and wells and stuff and the data guys and the software guys and everything about data and software. But they don't like there's no they they just can't bring together all this demand with. Yeah. Yeah.
00:09:15:13 - 00:09:42:14
Unknown
So like so I just kind of work with them and do the kind of translation and just make sure they're talking the same, the same about the same thing. I would say now though, the do you even have to do that? I mean, I feel like they could go right to you. And now with like some of these coding tools and stuff, like, I feel like you could even cut out that person because you know, enough about software engineering, it sounds like, and the domain like you're the perfect person to just say, like, let me run with this plot or, you know, go to X, whatever.
00:09:42:14 - 00:10:06:22
Unknown
And just knock it out like I can. And I do that for companies, but that that's okay for small companies. But what they say, what they always say is like what happens? So you get hit by a bus and like in like, you know, you've got the deployment and all that sort of stuff and administration. And, and like, you know, it's not just oil companies where it's like software companies and, and software for, for, for drilling companies.
00:10:07:01 - 00:10:32:06
Unknown
So like, they've got the software guys and maybe they're just not quite connecting with their, their, their client customers. But the other thing is like when you have these big companies, you've got lots of data, you've got lots of software guys, lots of data people, and it's enterprise needs like, yeah, yeah. Like it's straightforward to, to, to build, a system that's like the drilling engineers would want to use and actually use and it'll meet the requirements.
00:10:32:08 - 00:10:52:06
Unknown
But and all the enterprise bells and whistles and all that, you know, like that you need to involve other people and they need to do their thing and, and like, yeah, there's a different level of rigor, you know, on a managed SoC. Two audits. Paul. Yeah, I know, no, I don't blame you. Yeah. I've got a question.
00:10:52:06 - 00:11:15:13
Unknown
When you were when you were drilling, what was what was your where was your favorite place to drill and what was the like scariest or sketchiest place you've ever drilled? The favorite place was, I don't really know. You want to reckon. It's like they're all kind of. Yeah, that's. It's fairly. Yeah. The sketchy this place was, West Africa like.
00:11:15:13 - 00:11:41:21
Unknown
That's different. That's a different level. It's a different place. Like, I was, you know, the, subducted a couple of times, you know, like, you get em. I mean, walking down the middle of a street and there's an the security guards everywhere of AK 47 is like the busiest street in the busiest setting is busiest region. And then, like a guy who just flagged down a taxi or two taxis, the intercept it and then try and, you know, like, try and take your stuff and bundle.
00:11:41:21 - 00:12:00:10
Unknown
Yeah. Like, it's, it's, it's a different, yeah. It's just, you know, in the morning reports, if you're on a wreck, you get more reports, like, weather reports, you get, you know, weather. Put acid at 611, 6 a.m. and in Africa you get it's like a weather report. It's got the heat map, and it's got like him.
00:12:00:12 - 00:12:25:03
Unknown
It looks like a heat map, but it's a heat. Heat map of piracy activity. So like what we were we were on the rig one day, and then there was a morning, there was a call and it was like, oh. By the way, the oil tankers, disappeared or like the oil tanker has been hijacked. And there's as it's like a maersk or one of these big company tankers and for refilling, you know, in the ship in line for refilling the vessels.
00:12:25:04 - 00:12:52:18
Unknown
Goodbye. And they just the phone then or. Sorry, the people just hijacked the tanker and just taken away and runs with it. And it's like, how do you how do you where do you hide the tanker? And I think, yeah, it's just like normal. It's just normal. Yeah. It's happened all the time now that's, I've heard a lot of stories about about Africa, like, you know, you're living on a compound.
00:12:52:18 - 00:13:20:03
Unknown
You're being driven with armed guards everywhere you go. Even, like, entering and leaving the country through customs and stuff can be very touch and go and like, it's very, normal to have to give some kind of bribe or whatever to get in or out without any kind of issues and all that fun stuff. Yeah. No, that's, never had to, never gone over there.
00:13:20:03 - 00:13:41:01
Unknown
But I've, I've definitely always heard, some sketchy, sketchy, sketchy stories pretty much always come out of that. But yeah, that's that's crazy. And that's how you got to go down to the Falkland Islands as well. Yeah, that was that 13 years ago when they were just exploring. So that was just when I was kind of maybe like 15 years ago.
00:13:41:03 - 00:14:04:20
Unknown
And they, they if they're just exploring. So you, you know, that's like an 18, 22 hour flight, something like that, you know, like, get you to Australia, you know, other side, other side of the world, but you just going south and then so it was just that all the supplies have to come, like, you know, at 10,000 miles over at night, one has and it's in, it's just this middle of nowhere.
00:14:04:22 - 00:14:32:04
Unknown
It's the it's just the middle of nowhere. As all good oil fields end up being. Yeah. They're never near anything. No, no, no. I wonder if that's part of people understand like, say, the eagle bird. I, I underestimated it to describe. My first asset that I worked in was like, I worked in an office, and you could drive two hours on a major freeway, and you were pretty much there be in the heart of the field.
00:14:32:04 - 00:14:57:03
Unknown
Yeah. But other. Oh, I've not even then it's like you drive down to Katrina or Carrizo and you're like, where the hell am I? You're just on these giant ranches and you never, you know, you can go miles without seeing people. It's a it's a weird. I mean, it's weird because it's like you have the infrastructure at least to get stuff there, but it's still in the middle of nowhere.
00:14:57:05 - 00:15:23:08
Unknown
Like we end up building infrastructure around the fields if they're good enough. Like that's Midland is the perfect example of that, right? I have no comprehension of like, you know, like the UK oil industry, a lot of the, a lot of international projects, Africa and Europe, however they're managed like Europe, sort of means that the UK but the UK is very much a bubble compared to like, you know, America, Texas or whatever.
00:15:23:08 - 00:15:38:20
Unknown
It's it's kind of like self-contained almost in the UK's bubble. And they don't really there's not a lot of an I feel like there's not a lot of interaction between them. Like so I have no, I, you know, I have no I see things in TV and stuff. I don't have any idea, but I think it's the one of the biggest differences in onshore and offshore too.
00:15:38:20 - 00:15:59:17
Unknown
Right. Like I've never been offshore, but I know it's a completely different animal than the shit we do onshore. It's just the scope is different. The costs are different. Like everything about it is different. We're doing the same things generally. The you'll have a much harder, a bigger problem because you're in the middle of the ocean most of the time.
00:15:59:19 - 00:16:23:00
Unknown
But, from a logistics perspective at least. But they are, you know, people think oil and gas and they just group it all together. But those are two very, very, very different animals at the at the end of the day, like, I can't even I don't comprehend the offshore timescales that you'll deal with. Right. Like planning for five years and just the uncertainty for some of you do all.
00:16:23:00 - 00:16:49:13
Unknown
Yeah. And yeah I get a dry hole. Some of these whales are running like 100. I think 100 million whale. And and they're just vertical whales and they're getting abandoned. They produce nothing. They're not designed to produce anything. They even and driven. Engineered and like, it's totally different. Like, you think it's just the same things. But as I understand it, like in America, you're optimizing your fracking teams with your, iPad making.
00:16:49:14 - 00:17:09:11
Unknown
I don't even know the words. Your pad making teams representing your. You don't do that at all. Yeah. It's much more of like a, almost like a manufacturing mindset, especially at this point now that we've kind of got a lot of it worked out. And now how do we optimize all these pieces? I mean, it's but it's not even close to the same as, like some of these manufacturing operations with like just in time.
00:17:09:11 - 00:17:30:05
Unknown
But they're trying to get there, right and reduce downtime. But it's like that's why we know we're going to get oil. You know, it's more about like how efficiently can we make operations. Is it going to pay out. You know, are we going to can we keep our costs low enough and get the IP high enough so that in 6 to 12 months we get our money back and then we can start making money on it.
00:17:30:05 - 00:17:53:13
Unknown
But it is a very different. Yeah. There's very little exploration going on in onshore shale these days. Right where it's like you would never do an offshore project without some kind of exploratory data. Seismic, like there's all the stuff that leads up to even the actual drilling of it. Whereas here most of the plays have been fairly well mapped out and identified.
00:17:53:13 - 00:18:18:00
Unknown
And you know what you're going to see when you get there. And then, yeah, it has just become an optimization of operations. Right? Like when I started 15 years ago, it would take about a month to drill, you know, a 10,000ft lateral, which that was a long lateral back then. Now they're drilling for mile laterals, you know, double the length and they're doing it in a week or two.
00:18:18:02 - 00:18:37:05
Unknown
You know, the used to you used to be able to use the rig count as a really good indicator of what oil price was doing, and you can't anymore because the rigs are so efficient, like they can drill a well in 5 to 7 days all day long. And then the frack crews got more efficient. You know, how many stages can we do in a day so that we're not spending a month?
00:18:37:05 - 00:18:57:02
Unknown
Because that's what it used to be. We used to plug and perf everything, and we didn't zipper. We would come in. We they drill a 30 stage two mile frack or well, we come in and we frack it. It might take 2 to 4 weeks to frack the well. So you're two months now, you know, a month to drill, a month to frack, your two months into spending.
00:18:57:02 - 00:19:24:09
Unknown
You know, all you've spent all of your AFI essentially on this. Well, between the drilling and completions and you have no production yet. And then it might take another 2 to 3 months to get it on a production. And so they thankfully they figured out that like, hey, this is a terrible idea. This is like, this would be the equivalent of manufacturing something, sitting it on a shelf for three, 3 to 6 months and then going to sell it instead of trying to get it online as quickly as possible.
00:19:24:11 - 00:19:59:14
Unknown
But it's yeah, it's all an optimization game at this point. From a, a field perspective, I think we're going to start seeing a shift because they've, they've they've squeezed as much juice out of that, that fruit as you possibly can get on the the operation side, at least with onshore. And so now they're going to, they're going to start looking more, they're going to have to start looking into the data to figure out, you know, how do we increase our recovery factors or, you know, they're going to also have to start doing some exploration because we've kind of proven out most of the shale.
00:19:59:16 - 00:20:19:12
Unknown
But, you know, we're only recovering 5 to 10% of the oil in place from the shale to begin with. So it's like there's a huge opportunity just there. But we'll see what we'll see what happens on all this stuff. Okay. That's I don't know. I don't know much about that, fracking band here in, in band in Europe is 005.
00:20:19:12 - 00:20:41:18
Unknown
I mean, for now, until you guys can't get any oil at the pump. Yeah, until the gas and oil gets shut off. Yeah, yeah. You know, as the guy Robert something Robert Bryce, he's got the whole iron law of energy or whatever that, you know, people will, you know, do whatever they can to keep keep the lights on like, like they'll pay whatever much and do whatever they have to do to keep the lights on.
00:20:41:18 - 00:21:07:14
Unknown
So hopefully they start doing that here at like something to keep the lights on here because. So let's jump in to let's talk about BTC data. And then I think, you know, the thing that kind of spurred this podcast was I saw your post about, well, schematic.com. So can you kind of talk about maybe whilst Malcolm first and like just kind of what that project looks like and what it does and and then we can get into what you, what you're doing on the BD side as well.
00:21:07:16 - 00:21:29:16
Unknown
Yeah. So basically when I, when I am starting this data stuff, it was like you can do you can can if, if you know the drilling domain and you know the software side of things, you can do anything, but you can't really do everything. So you have to focus and you know, and something so so I started like the whale schematic, dot com system and it's for drawing well schematics.
00:21:29:16 - 00:22:03:14
Unknown
Right. Like, where's Maddox are used by like everywhere. And these musical schematics show what's in the whale and, you just need to make a schematic. And so I made that system because the way people use every single drilling company that I've seen the the use Excel to draw the schematics, like the, the, the dragon, the drag, the the size of the sails, the rows, the columns to get the, the like, make boxes, the highlight the boxes and then the, the click like right border, left border.
00:22:03:15 - 00:22:26:01
Unknown
And that's how you draw well schematic. And like that's not very efficient if you have a thousand people worldwide in that any time. So it's almost like we were doing that in a tool that wasn't designed for. Well yeah. Yeah. I mean, you could say that I don't know that our industry loves to put as much random shit into an Excel file as we possibly can.
00:22:26:03 - 00:22:46:09
Unknown
At least that similar between the between, like the rest of the world and the medium. Yeah. So using this bit, so I made this very schematic system and that's been used by, it's used by quite a few people, people to draw these schematics like guys and rigs and stuff and and I connect to that, you know, I mentioned via the API for the, for the data analysis to the PDF data.
00:22:46:09 - 00:23:05:09
Unknown
So like the company send me their files. It's just me. So I can't really like, and I software as a service is not really realistic as in terms of like data analysis. And companies, I don't think they really want it. They don't really want the hassle of, of the contract and new services going through it. All that.
00:23:05:11 - 00:23:23:05
Unknown
So that connects to the warehouse schematic system. So there's this kind of mutual incentive to, to, to improve the schematic system. I know it's the case where I added and, you know, talking about the everyone using spreadsheets for things like case and design companies are still using your options are a spreadsheet that takes you like two weeks everywhere.
00:23:23:06 - 00:23:43:11
Unknown
You know, you may only do a case until then, every year, or maybe less. And these big whales. But you're spending like two, two weeks to to build your calculations and excel and they're just not good enough. There's no loops or anything like that. And you know, alternative is then there's other software that's like five grand a month minimum, three months like minimum.
00:23:43:11 - 00:24:04:23
Unknown
So it's like one way was 15 grand. Could be so there's obviously a bit of an opportunity there. So about the you know, the case design systems, calculation systems. So you can go in to contact them, you can draw a schematic and then you can just apply a casing loads directly there. And companies do that but it also connects via API to other systems.
00:24:05:01 - 00:24:36:13
Unknown
And I've been providing like technical. Well and you didn't. And parts of the process, the delivery process as a service. Right. So you imagine you've got you imagine you've got drilling departments that's under resourced right. Like, I love it's project management. They don't necessarily have the technical expertise or like it take them weeks to get back their head back end to to technical expertise in terms of like the calculations and stuff like that, etc., hiring new people instead of contracting new software, you just, you know, speak to me.
00:24:36:15 - 00:24:53:09
Unknown
And it's kind of like a turnkey service. You want your casing design done for hand calculations done. And it I come software, if you know what I mean. Like I use my software as a service and and yes, there's like a front end on the website Mediacom and in the case design, but it's all the calculations are all exposes end point.
00:24:53:09 - 00:25:14:11
Unknown
So I can just access them via, you know, the API or via MCP or whatever to, to make things faster. And I think it was like an, you know, the post you had, you felt like a lot of, like old, you know, books and stuff into it as well. Or you were able to, you know, like enhance it with some of like the engineering, you know, literature and everything.
00:25:14:13 - 00:25:34:16
Unknown
Yeah. So like, I mean, a system a few years ago, 2 or 3 years ago, so, so like when you're working on index, like, everybody knows, you'll get like a hard drive and it's people gave you files and they gave you, like, the PDFs and they gave you, standards and all that. And, and so a couple of years ago, I had like, like 50,000 documents.
00:25:34:18 - 00:25:58:12
Unknown
And then a couple of years ago, I made, like, a like system where, you know, ingested just those documents, so you can search over them. And, and I made, like, a kind of simple front end, interface, and, and I and I kind of released it. I mean, obviously, I figured filtered out for, for, like, you know, copyright stuff obviously affected it first, but I released it and then people started using it and as like, I don't want to want to get involved in this.
00:25:58:12 - 00:26:27:00
Unknown
You know what? I'm going to try deactivated the writing system. But but basically what? Yeah, what you can do is you can then like, let's say you're looking at it, you know, you want to find, relevant information on a particular very specific thing. And drilling. You did a rock search. I tell you do a search, it returns a, you know, obviously an AI summary or whatever, but also, the context that was ingested, like the chunks that you, embedded.
00:26:27:02 - 00:26:47:07
Unknown
And then you can click on it immediately go to the source document. So, so in terms of retrieving your data and retrieving like in like if you're not working something for a while, I don't know what relevant standard is for something. It becomes easier to do that. Like you're no longer searching the things for errors. You're just it's just, you know, just out, you're done a couple a minute.
00:26:47:09 - 00:27:05:08
Unknown
Nothing. You're speaking in John's language here. I think that's, Yeah. That's what that's what I've been working on for the last two years. Is our enterprise drag, offering. But no, I mean, that's that's the thing I don't I, I feel like it's starting to come around a little bit more. At least for the average person.
00:27:05:10 - 00:27:29:14
Unknown
But, like, if you've got a bunch of documents that you constantly have to reference, a rag is the perfect solution for that. And it's so easy now to go out. I mean, there's so many open source GitHub libraries where you can just set the whole thing up from pipeline to front end with a few commands. And it's, you know, I would like to be able to tell people that it's that easy.
00:27:29:14 - 00:28:00:01
Unknown
It's not. There's always complications with the data. Ingest is normally the the hard part on especially with oil and gas documents, depending on what you're trying to ingest in there. But the technology around that continues to get better. So that continues to get easier. And then yeah, like if you're constantly referencing API's, you know, API procedures or API standards or anything like that, a ragu will save you so much time.
00:28:00:03 - 00:28:29:18
Unknown
Walk me through so. The the well schematics of I building new wells in that. Or am I able to upload a schematic from a PDF and then have it kind of digitize that? You can build a entirely new schematic. It has a very basic I think it's just like structured output. It's the it's a is a very it's like, I had a landmark 2.0 or something just for speed and just I've not upgraded it yet.
00:28:29:20 - 00:28:51:05
Unknown
Because nobody's paying me, so. Yeah, you can, you can make a new schematic using the using the, the forms there. It can, you can connect the via MCP as well. So, so that you can connect it to like, you know, a dot cloud or something like that. Which means you can automatically generate schematics, but there are limitations in that, in that it, you know, these things, there's security limitations.
00:28:51:07 - 00:29:17:21
Unknown
So you can display, you get an SVG that directly comes from the, you know, the service. But in terms of, making a schematic from, from a PDF, it just depends. Like, I'm thinking maybe to release a system that automatically creates schematics, but, like, there should an issue where data is that you and you can.
00:29:17:21 - 00:29:38:04
Unknown
Absolutely. You know, if you connect the via MCP, you can just take a picture of a schematic, you draw on your phone and it will just it convert it straight away, you know, like it's straight forward. There were these well fails. Like the issue is that let's say you're analyzing old world files. You might have like, you drove the way on the 80s.
00:29:38:06 - 00:29:59:14
Unknown
You might have like a mud report. Daily drilling reports. You might have, you know, annual report. Don't know, report some better report all these different documents that relate to the, well, like that are talking about it from slightly different perspectives. And there's overlap. But how do you incorporate all those files, and all the data in these files into like a single source of truth.
00:29:59:16 - 00:30:30:04
Unknown
So, so you got all these different sources and then you've got that extrapolated across time. Right? So you drill in the 80s, you go and work over in the, in the 90s site tracker, you know, a couple of years later you get the same mesh as a data that's hard to implement and like, an automated system that just does everything and, and like, and it's a totally different issue than, like, ingesting, semi-structured IDC, like CSV files from, from a, from a regular reporting system or whatever.
00:30:30:05 - 00:30:52:14
Unknown
It's like there's different views. Legal docs are my favorite because they're just plain text. I have no issues getting anything out of a legal doc because it's just a plain text document. And there's struct, there's structure, there's headers, there's subheadings, there's all that stuff. And then you go over to a drilling report and it's like first page wellbore schematic outputted from an Excel spreadsheet.
00:30:52:14 - 00:31:33:06
Unknown
Second page the actual wellbore diagram, third page time log of, you know, the daily activities that are in a table or that were structured in a table, probably in Excel. But then outputted without any of the the the table bounds. So and Lem doesn't really know whether it's table or plane, like it's just there's so different. And I don't think people outside of like, you know, people who are doing this type of stuff on a daily basis truly appreciate how excuse me, how hard it is because of the variety of data types.
00:31:33:08 - 00:31:55:09
Unknown
And then even then, it's like, you know, one drillers report looks totally different than another drillers report. They all have the same basic information in them, but they're structured in different ways, and they're displayed in different ways, and they're formatted differently. And all these nuances that that make it a huge pain in the ass on the back end to get the data out.
00:31:55:11 - 00:32:27:20
Unknown
I think I think what we really need, Paul, is we need someone like you to come out and, build a, an AI standard for, for drilling documents or something like that because, well, but like, think about this. Right. Like, what do I need to know from a wellbore schematic? And it's, it's a handful of things. But how do you take that image on a piece of paper with all the lines and all the crazy shit in a wellbore diagram, and translate that into something that a rag search would actually be able to render accurately.
00:32:27:22 - 00:32:49:19
Unknown
Right. And it's like I don't need the whole diagram, I just need x, Y, and z things extracted from this diagram and tagged to it as metadata so that when I search, it will show up. And so it's I think there's there's a really interesting space there in the future for the energy industry because you know, it's like okay, raw text fine structured numbers and a table.
00:32:49:20 - 00:33:14:02
Unknown
Fine. But then you have these diagrams and like logs or maps, but you have these like it's their images and there's important information in them. But for a computer vision or an OCR model to be able to understand what the hell is going on in those, it's almost impossible to an extent like you can extract the data out of it, but it doesn't have any structure to it.
00:33:14:02 - 00:33:31:18
Unknown
It doesn't have any like, logic behind it. And so it's it's almost like, hey, why don't we just, you know, the data is already in there. Why aren't we just generating a second PDF page with all the metadata about the prior page that we care about? And then that's what gets fed into the model or into the index or something like that.
00:33:31:18 - 00:33:57:15
Unknown
I don't know, I'm just, ranting right now, but I do think that, like, there's there's an opportunity in that space because people think that data is always, you know, PDFs or, or tables essentially, or csvs. And it's like, hey, there's there's data in this subset of data that it doesn't really make sense in a text based world, which is all, all things.
00:33:57:17 - 00:34:26:02
Unknown
Well, these days that, you know, how do we how do we handle that? Yeah, that would be, I don't I mean, yeah, thing is that a lot of these old files are like, they already generated all the data. Oh, yeah. And it's like. Yeah, all the old data point forward. Yeah. Yeah, it's it's wild. Are you doing anything like that with, you know, someone sends you, hey, here's, here's this acquisition we had 20 years ago that we haven't touched, and here's all the data for it.
00:34:26:02 - 00:34:41:00
Unknown
Can you help me figure it out? Yeah, that's that's basically I'm doing so you've got like offset reviews. So you might as in your drawing in an area you've never tried before. You need to look at it all the way I was, I just finished some work for and, and then sort of and then you've got like I say, acquisitions.
00:34:41:00 - 00:34:59:18
Unknown
Then you've got a lot of people in Europe, and we need to. You've not been on a whale in 20 years, and you need to know what's there. So when you go and you don't, you know, you know what you're doing. But as far as acquisition ones, I think that's a big one, because I think a lot of companies or a lot of them investors buy fields based on geology alone, and they just kind of ignore the drilling.
00:34:59:20 - 00:35:16:10
Unknown
They do. They look at the reserves and they don't look at how how much it's going to cost to get are those reserves out the ground. And you can only do that by looking at the whales. But if your whales are like your your old whale, your whales are like, you've only drilled that field like 70 years ago.
00:35:16:12 - 00:35:40:15
Unknown
What, you don't you can't make out like, you don't know what it's going to cost you unless you do the analysis of the whales. But I just completed work for a company. A European company. And it was, the files were somewhere from the 70s, some 80. Some some like from ten years ago. Some were two pages, some were 650 pages long.
00:35:40:17 - 00:36:06:06
Unknown
French, Dutch, English. And like, I don't speak any, speak English. That's that that that's just about it. And the units were like, you know, you get oil through the normal units, feet in barrels, and then you've got like meters and all that, and then you can like French units, like, like, some I don't even know what they were, but you get that sort of thing in like and some of the whales are vertical, some are like three side tracks.
00:36:06:07 - 00:36:33:06
Unknown
I, that's the sort of stuff. Yeah, yeah, there's a whole I mean, it's I guess it's probably worse in the offshore world just because offshore has been around longer than the shale stuff. But onshore conventional, I feel like is very similar to the offshore world. And the paper aspect of like my father in law has filing cabinets full of logs and, well, files and, you know, production cards and all of this shit.
00:36:33:08 - 00:37:06:08
Unknown
And it's like he knows where it is. No one else in that company has any idea that that data even exists. Knows how to how to access it or how it's organized or anything like that. But walk me in with Scott Caso or, you know. Yeah. Right. I was going to say walk, walk us through, kind of how that, how that went because I've done a very similar project to that, last, last year with a big offshore operator, and they gave us a shit ton of data that was all scanned banker's boxes.
00:37:06:10 - 00:37:40:20
Unknown
Shout out to Scott for doing all the scanning over at Caso. Yeah, we're looking forward to getting that work. But it's one of those things where it's like, okay, hey, you know, I don't whatever percentage of this is what I would call like a modern data set where, you know, it might be a PDF, but it's a structured form or there's structure and stuff to it, and then you start going back into like the 80s and 70s and 60s and it's like, oh, here's just a blank page of paper where the engineer did all of the calculations by hand, and it's like, what am I, what?
00:37:41:01 - 00:38:05:23
Unknown
What am I? How am I supposed to give that to you in a search result like no one. It's handwritten. Two it's it's handwritten math. So it's not even words, it's just symbols. And then three it's like, does that even need to like in what world would would you want that as a result? Right. Like trying to reverse engineer how to get the user back to it.
00:38:06:01 - 00:38:24:09
Unknown
And so like, it's just such a, I want you to talk to it because it is such a gnarly problem. So like, walk people through, you know, the trials and tribulations of that for like, the, the, the kind of difficult stuff. So I don't really I don't really do the right stuff for, the vectorization for these sorts of files.
00:38:24:10 - 00:38:48:07
Unknown
Like they just want schematics, they just want like, I don't know, maybe they want to map this one various output, but it's, it's output that the Jalan team can use it for any licenses or anything like that. They just they want they want their Excel files. But I start off with, like, you know, it's, you got your schema, you got your structure, your warehouse structure, and then, and I just, you know, get wet, smell that sort of thing.
00:38:48:07 - 00:39:14:00
Unknown
But they're far too big. And, you know, I've cut it down to what? What I would want to see if, you know, the information is actually used 99.9 times out of 100 when you're, when you're drilling, I usually use. So you got these files and I use like OCR to, to get the text out and actually like I don't use the Bayesian model models or anything like that because what I find is you have to iterate on it and you have to do all sorts of stuff and go back and back and back, and it's just it's just them.
00:39:14:02 - 00:39:39:05
Unknown
It these documents are too big. But it's a case of like once you get the structure, once you get a data model, then everything just kind of falls into place. Like, you can just you can quite easily work out those calculations are relevant, are not if they conform to your the structure if if if if said is necessary for their particular goals.
00:39:39:05 - 00:40:03:18
Unknown
If you know what I mean. So like the worst thing is you've got the OCR data, you've got the structure that you want. You can use LMS to extract, to to extract the structure. But you get into all sorts of issues with, like, with, with, like the files are too big, the schemas to bag, you get nested, nested like, you know, yes, you're speaking my language now.
00:40:03:20 - 00:40:37:11
Unknown
It's getting tables out of a PDF is much harder than accurately. Let me rephrase that. Getting them perfectly replicated out of the PDF into a table. CSV, whatever structure you Json you want is such a under appreciated problem in my opinion, because it is so hard. Like it's such a boring right? It's so awful. It's so terrible. And it's like I can tell if I look at this, why can't you computer vision model what the hell is happening?
00:40:37:11 - 00:41:00:06
Unknown
It's a very basic table. Like I've. I've narrowed it down to this one page. In this one document, out of a thousand pages. And all I want you to do is extract this table perfectly, and you still can't do that. And, like I've tried, I mean, I, I went back to trying like og just table from PDF tools that existed before LMS were ever around.
00:41:00:08 - 00:41:32:18
Unknown
Some of those work okay on certain documents, some of like and that's part of the problem, right? Is it's like there isn't a single source solution for that right now really that's remotely accurate in a lot of ways that I'm I'm cautiously optimistic that that's going to keep getting better, just like the language models have. But, yeah, I think ultimately everything ends up running on vision models because if they can get that cheap enough and accurate enough, it solves a lot of those problems.
00:41:32:20 - 00:41:57:15
Unknown
That a text based model alone can't solve. If they can do that. But the the jury's going to be out on that for a while, I think. But I am the vision models like, I just, I just, I just avoid them. I use them open source models mostly, even though they're less powerful and they're less than exciting because I find, like the, you know, open source models get you to maybe maybe that's 80% accuracy.
00:41:57:15 - 00:42:22:07
Unknown
Like what I need to do is like the accuracy levels are they need to be equal to or better than the accuracy a driven engineer would get. You know, when they're looking at the data manually. And the open source models might get you like 80% accuracy or, and these big models, the Geminis in that there's obviously you wouldn't I can't use them because of data privacy concerns.
00:42:22:09 - 00:42:59:12
Unknown
They'll get you like 95%, you know, whatever. But but the line of work that takes you to get from like 80% to, to to the over the threshold that you need is the same amount of work to get you from 95% to like over the threshold. So so most of what I do is, is deterministic stuff. So like you mentioned those tables and that you, you assume they've extracted properly, but it's because I have kind of like a graphical interface and I can, I can and because I have like a warehouse schematic, you immediately see if something's wrong, immediately see if it's wrong, and then you can implement that.
00:42:59:12 - 00:43:18:20
Unknown
You can even implement like, add tests and checks and evaluations and all that sort of stuff to make sure it doesn't happen again. And so you just constantly evolving this model that like if you have a table of trajectory points, like maybe there's ten pages, these are just pure numbers and they're just a plain text. Right? Like you say.
00:43:18:22 - 00:43:47:15
Unknown
But then if you plot on a trajectory and it goes like upside down and cross over and stuff like, you know, it's wrong and you can put it in text now, I think that's a big part of it, right? Is being able to build in like agents are great when they have the correct tests and checks and boundaries and like, but just letting an agent go free rein on something is not going to work the way that you want it to, especially in in the data world, if things what are you using?
00:43:47:15 - 00:44:14:06
Unknown
You were talking about open source models. I, we, we've built a, we've got a fine tuned model for petroleum engineering. Now that is passing the AP exam, that we're super excited about, but, that's one of the reasons we fine tuned it is because we realize that these, these foundational models are not going to be good enough at these domain specific, applications and such.
00:44:14:06 - 00:44:31:02
Unknown
And so, on the data side of things, I try to use open source models as much as I possibly can when I'm doing any kind of extraction or building any kind of a genetic stuff, just because I know if I can get it to work on an open source model, it will absolutely work on a foundational model.
00:44:31:04 - 00:44:52:12
Unknown
But I think it's an interesting point, too. We've been talking about this at the office that like, no one is thinking about how expensive these are going like, all of the tokens currently are super subsidized. They're unnecessarily cheap because they're playing the market share game. But like at some point there becomes a winner and the prices start going up.
00:44:52:12 - 00:45:13:07
Unknown
And you've now built all these automations and workflows, and now you're very risk heavy into whatever company you chose. And, I don't think enough people are thinking about that. But let's talk about your open source stack. Are you using like, vellum or Allama or what are you using to, to host these models? I am so for like prototype work for it.
00:45:13:08 - 00:45:35:01
Unknown
So I would test the system is using like, publicly available data, like not client data. And for that I would do some fancy like, like grok for Q, which is like so, so I would do something like that because it's fast and it's dirt cheap. Like it's so cheap. It's so, so cheap. Yeah. Like it's it's they got put by, I don't know throws in Nvidia.
00:45:35:03 - 00:45:52:04
Unknown
I saw a couple of months ago. So I don't know what's going to happen to them, but is like that's what I'd prefer to use. But I notice that their terms and conditions are like it used to be any other data retained. And it. I think they changed it then. I think they changed it back. And it's like I don't I can't really rely on that.
00:45:52:04 - 00:46:11:04
Unknown
If you're saying to people no, no your data doesn't leave it, you know, like data security. So what I tend to do is, it's open source, but I run it and I run on, like, cloud infrastructure because clients pay. Like, I would have thought that running things locally would be like your client would want that, but they say, no, that's a red flag.
00:46:11:06 - 00:46:39:06
Unknown
It guys go, no, no, no, you can't do that. It's got to stay. And like ISO 27,001 compliant cloud infrastructure, it's like okay so what I had to use them. I use them like I can never remember the name of them. One of these, these, like, GPU, things on, cloud infrastructure. I run them up for like a few hours as long as I need them, and then just shut them down and use them in the I can't remember, you call them.
00:46:39:06 - 00:46:57:08
Unknown
There's digital DigitalOcean. Do a couple of good ones. They did once. Like you just press the button and it and install the model installed endpoints and stuff for you. And they do one. It's just GPU infrastructure. But see, because I'm paying the bills, I set the clients for a fixed cost, like turnkey, like here's how much it's going to cost.
00:46:57:10 - 00:47:28:18
Unknown
And so I need to keep it. I can keep one of these cloud droplets running for, for, for like a year. I can't afford that as, so so that's why that's how I manage it. Now, have you had any pushback on specific models like we've, we've seen, some, some enterprise that are that will they've come out and said in meetings like, we will never use a model that came out of this country or from this company or anything like that or not, not yet.
00:47:28:19 - 00:47:53:00
Unknown
Even even open source ones, specifically open source ones. Why? Why is why that even they're worried that, you know, deep has some unknown, malicious something in there or back out under the hood or something, even though it's like if it's in your infrastructure, you should be able to spot that immediately. If anything, it's like going away, like, right.
00:47:53:00 - 00:48:13:08
Unknown
Like, no, I haven't, I, I've had anything like that. I tend to use it's, I don't even, I don't even like, I don't understand the, the kind of business model for these a big biggie, like, you know, your anthropic in your, OpenAI. Because I don't even know what's model. I, I don't know which one. It's like, if one doesn't really work, you just switch out the endpoint first.
00:48:13:08 - 00:48:44:00
Unknown
For one thing, it's not it. I am, I think I, I use it's either llama so like matter or the OpenAI OS, we the open source OpenAI one, the GPT OS. Yeah. Yeah. That one. Yeah. So that's a good one. I just, I just downloaded GEMA for, earlier this week on my laptop, running on a llama, and I've been super impressed with it.
00:48:44:02 - 00:49:01:03
Unknown
It's it's less than, like, I think it's less than 20 gigs. The full model. And so it's and it can run on. I mean, it's basically built for laptops, or to be able to be able to run on a laptop. They've already got a bunch of quantized versions of it out there, which makes it run even better.
00:49:01:03 - 00:49:29:14
Unknown
But it's, it's fascinating because like on the foundational side, it's more tokens, more context. It's bigger models, right? More capabilities, more tooling, more logic, all that stuff. But then in the open source side, it's like this dichotomy of better models, but they're smaller or they're more efficient, or you can run them, you know, in places that you absolutely could never run, you know, a, a GPT five, or anything like that.
00:49:29:14 - 00:49:50:14
Unknown
And so I think it's fascinating. I'm very curious to see how this plays out because like, at the end of the day, I don't need a foundational model to go do some a genetic extraction on a specific page where I'm telling it exactly what it needs to it. Like I don't need a foundational model for that now for like routing for my queries into my my rag to figure out which tools or which path to get.
00:49:50:14 - 00:50:14:15
Unknown
Like, I think that people, you know, people see GPT or quad and they think, oh, it's just one model and it's one thing doing one thing, and it's like, no, this there's so much in between. When you hit enter on your prompt and when it starts returning an answer that the user never sees. We've talked about this before on the show, but it's like that fuzzy middle layer of like the tool calling the orchestration the agent routing.
00:50:14:15 - 00:50:37:00
Unknown
Like there's so much happening in there. And like you don't typically get a lot of that with the open source models. But if you're just, you know, trying to get some automation done, you don't need an open source model to do that, especially on the data side of things where you're just extracting or reformatting data or whatever. So that's that's really cool.
00:50:37:00 - 00:50:56:03
Unknown
If you haven't played with any of the local stuff like Obama or BLM or any of those, I would highly recommend, just messing around with it, because it's like being able to run an agent that does, you know, that's doing extraction or, you know, different things like that, just locally is it's fun to see. It's it's not any faster.
00:50:56:03 - 00:51:21:19
Unknown
Of course it's free. But it's it's pretty wild how good some of these small models are getting. Is the general one good? Is it like like it's usable? It's really good. No, it's, it's a all. I think the gem client is out of Google, so it's based off of the Gemini models. And the Gemini models, to my understanding, were originally trained for like, classification.
00:51:21:21 - 00:51:46:23
Unknown
And they're really, really good at that. And so like being able to feed it in like a data set and say classify these as, you know, X, Y or Z. Beautiful. I think the GEMA for one is now multimodal to you in some ways. Or it has that capability at least. And so yeah, the again, super lightweight model now has all this functionality that these bigger models used to have.
00:51:46:23 - 00:51:54:15
Unknown
And so I'm very curious to see kind of how that all that stuff evolves. But
00:51:54:17 - 00:52:18:19
Unknown
How do you like just with your experience, how do you kind of see AI feeding into the industry? Where do you see it kind of touching first, and where do you think it has the most kind of applications? At least this type of AI? Because, I'll, I'll preface everything with language models because we're AI is. Yeah, yeah, yeah, probably used, as CRM driven.
00:52:18:19 - 00:52:49:23
Unknown
I mean, churn is not very good. Like you say, everyone is using Excel spreadsheets and like, databases came out, you know, got 4 million other entities, 25, 30 years ago. You also have the 13 tops in them yet as in like and so I so our lines are. Where does it go first I think in these small cases, like people are, like, I'm like taking a picture in your phone of a weather schematic and it's drawing the picture.
00:52:49:23 - 00:53:16:07
Unknown
Okay. Like, but that's not a major. It's. I don't know, you say that, but the person who's having to build that in Excel by hand is probably like, this is the best thing ever. That's the thing. It's like there's there's a lot of low hanging fruit in this, you know, in this sense across the industry of like, why is this engineer who we pay that hundreds of thousands of dollars a year doing the stupid menial task.
00:53:16:09 - 00:53:39:08
Unknown
But the the standalone impact of that across a company doesn't isn't like sexy, right? Like it's not hugely impactful, but it's giving that engineer their time back, right. Like it's a weird trade off I go to miss. I get messages on my emails, schematics, system in the case design system, I've got these, feedback buttons and you click the button and you just put in feedback and it's easy to do.
00:53:39:08 - 00:53:54:17
Unknown
So people send me that, they give me feedback and I get these messages every so often. It's like, you know, I think, oh, maybe, maybe nobody's using this system. I've got analytics and people are using this system, but it's like, maybe people aren't actually using it and it's only point of view in it. I got a message yesterday.
00:53:54:17 - 00:54:21:00
Unknown
Like the guy who obviously it probably works and you know what? Someone in the there somewhere, all caps. It's like, that's it. This is this is excellent. Thanks very much for. Yeah, yeah, yeah. So now I mean, and and so people are people do want these tools, but I wonder if the driver comes from lower down, you know, for actual tools at work, it comes from like saving people downtime rather than, like from above.
00:54:21:02 - 00:54:42:17
Unknown
That's I think that like, that's the thing you get a rag or an automation or anything like that in front of the person who is currently having to do that. It is like the best thing since sliced bread, right? But then you try to communicate that ROI to management or to upper level management or whatever, and it's like, well, but that's what I pay them to do.
00:54:42:19 - 00:55:12:16
Unknown
And it's like, really? Are we I mean, but but maybe you don't have to have as many of them in the future because they can actually do the parts of their job that you hired them for not spending, you know, a week filling out permits or reports or, you know, like all this stuff. There's so much manual work in our industry on the back end, you know, and most of it is literally just moving data around or filling out forms of with data that already exists in a database.
00:55:12:16 - 00:55:22:20
Unknown
So it's like, why is a person having to do this? Yeah, yeah. It's it's it's yeah, I agree, it's,
00:55:22:22 - 00:55:43:21
Unknown
I don't know, I don't know, it's it's a strange industry and like this. Strange, that's for sure. Yeah. But I think you're right. I think, like, getting getting those tools that make the average, like, the end users life easier ends up being a much easier. Like it's sticky, then they keep using it, right? And then they like, then they want more, they want something else.
00:55:43:21 - 00:56:04:13
Unknown
Or it's a lot easier to start, you know, start small like that. That's what I'm so I was I was always of the opinion I was making these other systems before I had that system that just a service. We send the data, I send the results. But I was I'm focusing on trying to get like SaaS software and doing analysis as a service and all that sort of stuff.
00:56:04:15 - 00:56:26:09
Unknown
And my conclusion, and the same conclusion that I've had for the last 15 years was like, the drilling industry is really conservative and nobody really, you know, nobody really wants to adopt new technology when c when like assure them it's even like doors open slowly. Right. And it's a challenge to open doors and get these conversations going. But when are you in the system like how the system works?
00:56:26:12 - 00:56:58:20
Unknown
Here's the way I analyze your data. It's you know, you assume they're conservative people, but as soon as you see it and they go, okay, that's fine. It's it's like a five minute conversation. Be like, okay, we'll send you some data tomorrow. You can test it and like, okay, we'll use that as an they're not slow. They're slow to adopt a large part of the reason I, I wonder that people are still using Excel is because of this, this difference in this communication problem between, like, you know, data that a lot of the software isn't just isn't it's just not appropriate for for the drilling guys, it doesn't make life easier.
00:56:58:22 - 00:57:21:13
Unknown
But as soon as you show them something that does actually make the life easier, it's like it's sold. It just sells itself. And so, so yeah, well, that's the thing. Like especially on the op side, like making their life easier means they get home earlier or they don't miss, you know, a kid's sporting like it has tangible impact for that person.
00:57:21:15 - 00:57:40:19
Unknown
And so you're right. Like it immediately is like, oh yeah, I'm going to use this every day now. You're not having to like get a phone call two in the morning, some some from three. And you're not having to stand up justifying how you made the same mistake. Like you like in mistakes can be embarrassing. Or if. Yeah, it just makes it easier for.
00:57:40:22 - 00:58:02:16
Unknown
Yeah. If you were in real world benefit for the people in right. Yeah. No. And then, you know, again I struggle with like the fact that this is literally like very obviously valuable to the individual, but then trying to get that value sold to the company at a company level is a whole nother like it's a whole nother challenge.
00:58:02:18 - 00:58:21:00
Unknown
Yeah. And that's the thing is like the way I schematic systems and the case in design. Yeah. It's wonderful that you've got this medal, you know, software people can use, but drilling engineers, drilling guys, guys in the rig, they don't walk around with company credit cards in the pocket like they it has to pass for, procurement and all that sort of stuff.
00:58:21:00 - 00:58:49:07
Unknown
And you get ten layers of approval. It's, Yeah, it's something definitely to think about. And then these things, well, I'm back. Energy is important. He had a speaking of importance of it. Yeah, I guess I haven't had a beautiful day in Houston. Will probably have all year, but such as life, but, so yeah, I'm coming back into it.
00:58:49:09 - 00:59:06:04
Unknown
I mean, I know we're running up on, you know, probably time here shortly, but I, I do want to ask Paul, like, I mean, kind of it's all as I'm doing like to talk about the journey. Like, where do you start with, you know, getting into like, the coding, like what are what was like your first language? I saw, like me on LinkedIn that you started with, PHP, Laravel.
00:59:06:04 - 00:59:40:01
Unknown
So like that, like Lamp stack stuff and then how it evolved now with like cloud computing and all the other advancements. And yeah, it was Lamp stack, actually as well ago that I remember that. But yeah, that was like 20 years ago. You're in PHP and you and and it's like echoing out the HTML from the PHP and then, you know, to build these systems and then, you know, when I was in drilling out of it for a bit and then back into it and, and like, it's amazing how the, the drilling industry doesn't change at all, like you're still using things over, but the improvements in software and then these frameworks is like
00:59:40:01 - 01:00:01:13
Unknown
it's unbelievable what you could do. Like and I started using Laravel and even up until not quite recently, it was it was PHP and Laravel, and I did that for the data analysis work. So like I'm obviously not like a hardcore like data type person, but it's specific to, you know, it's only specific to these particular problems.
01:00:01:15 - 01:00:19:06
Unknown
But because it in, in like Laravel just in PHP, it's just so easy to get up and running like it's, to build a web app that works and it functions and they don't need to deal with any sort of boilerplate stuff. It's just you focus purely in the back end and like, fair enough. The back end stuff is not as efficient.
01:00:19:06 - 01:00:43:08
Unknown
There aren't as many frameworks and are like, packages for data analysis or engineering calculations, but then it kind of means that you're, you know, instead of instead of just typing a histogram, an important a histogram package or whatever, you're you have to do it manually so you get a good understanding of underlying, like what you're actually trying to do, rather than some sort of abstraction layer.
01:00:43:10 - 01:01:15:10
Unknown
Right. PHP and Laravel and and like, JavaScript in the front end and then like relatively recently switched to Python, just, just purely for the extensibility here. Well, the tools. Yeah. You can how can you you can't do that. You can't not do that. You can't. You have to switch to to to Python. I don't know how you if there's a better way than Python and pedantic and, and and those things you use machine learning or any sort of machine learning.
01:01:15:12 - 01:01:40:19
Unknown
So how did you get into coding though? Like was it from university or it was just it was like magic. Okay. It's yeah, I did Matlab projects for, for like, a couple of companies and my own project. I used to have these, web websites and I would make enough, you know, this was in the time of Google AdSense ad where whatever the advertising one was and you, you know, you got clicks on your site and you got advertisements.
01:01:40:19 - 01:02:01:05
Unknown
So I used to have multiple ones of these and I would make a bit of money, but not not much. And I just, I just like doing it. And then, and then that these various projects and then I don't know, like I like building things and you can not understand how to do any sort of programing if you if you like building things these days that you can you get.
01:02:01:07 - 01:02:22:02
Unknown
I feel like people see like people like us and you're like, how does a mechanical engineer get into data or coding or anything like that? But, I think one of the things that like at least I see is, you know, or very logically minded people generally, and code is just logic, right? Like it's just a set of instructions.
01:02:22:04 - 01:02:51:12
Unknown
What's some advice that you might have for current, you know, engineers, mechanical engineers, whoever that might be in school or coming out of university, around like being familiar with code or anything like that, like I think first off, like, don't be over rely on these, you know, the AI stuff. I think that's dangerous. And you cause too many problems yourself and it might look good, but when you try and tweak it and stuff it, you can run the problems.
01:02:51:14 - 01:03:19:01
Unknown
But I would definitely I would very much see like. Regardless of what your career is or whatever, like if you get into programing, if you're doing like pure programing, you're going to be limited to maybe you're going to be limited, like building software. But if you have like domain expertise and engineering or like accounting or anything like that, and you can program like you're, you're able to leverage computers to, to do things that most other people in your domain can't do.
01:03:19:01 - 01:03:38:08
Unknown
So, and it's also interesting, you can do a lot with it, like if you like coding things, where my daughter calls, she calls me a fixer, and it's like, we just like to fix stuff. And, I'm like, that's the best description. I'm, like, honored to have that as my, nickname in the house. But it's true.
01:03:38:08 - 01:04:01:21
Unknown
Like, we just like to make stuff and solve problems. And like, so many problems these days are rooted in data. And so, like, as an engineer, I almost feel like you're not doing your job to the fullest capability if you don't at least understand the data right. You don't have to understand how you got the data or how to generate it, but have that foundational understanding because like data is everywhere.
01:04:01:23 - 01:04:21:08
Unknown
And, but yeah, that's, I don't know. I just think people, people group them like, it's either you're coding or you're an engineer and it's like, I completely agree with you. The future, especially now with how easy it is for an average person to pick something up and try and vibe code and app to solve their problem, or a script to solve their problem.
01:04:21:08 - 01:04:48:03
Unknown
Like if you just have a basic level of understanding of code and, you know, kind of general architecture, what a database is, what an API is, how they work, you can go really far because it is like, is it prior to that, a few weeks. Like, I mean, he he used to screw around with AR and I remember like chat with him when he was still on on, on Twitter and stuff, but like he's got a full on like SaaS offering you that.
01:04:48:05 - 01:05:10:19
Unknown
He's like building on he's got the MCP and API and like just all these thing. It's just it's wild. So I listen to his a podcast. You did this with him? Well, I wasn't, I wasn't that and I saw someone he posted, yes, yesterday, the day before. But the MCP system that he's got and it's it's it's not your it's not your, it's not your standard.
01:05:10:19 - 01:05:31:12
Unknown
Like, it's like it understands the domain. Like he's an actor and that's and like I was think I was thinking about this morning to just like I mean I think my, one of my little taglines on LinkedIn has always been like solving business problems with technology. Like, like, don't put me or anyone us in a box is that you're a data engineer or you're a analytics engineer, you're a data scientist.
01:05:31:12 - 01:05:54:20
Unknown
No, I'm I'm what I need to be at the time. I might be an IoT engineer tomorrow because that that's what you need to solve this problem. Like but I am, I can I can love, figure out and leverage technology. And I know what those pieces can do. And then it's just a matter of. And now, like John said, like the the barrier to actually like diving into that from the technical side is so low now it's like, oh wait, hey, I need to connect to Azure IoT hub in blue, blah blah blah.
01:05:54:20 - 01:06:15:04
Unknown
Like it will lay it out for you, but you have to know what you're trying to solve and what those tools do. But as long as you understand what the tools, the right tool for the job is and the problem you're trying to solve, like implementing it is like nothing. I mean, like I got these PDF reports for a, for an operator, like, and I was like, use AI.
01:06:15:05 - 01:06:35:04
Unknown
So cloud could use auto and, you know, matplotlib and or plotly or whatever, you know, thing. And like, here's what I want about it. Freaking did it like an I under an hour I had like this thing that would have taken me days is there and nudge things here and, you know, create the data frame logic and all this stuff.
01:06:35:04 - 01:06:59:16
Unknown
And it just it's just wild. It's a crazy time. Do not speak dare not speak to kind of that point. And the previous thing is like what like what advice to like somebody in engineering who's starting out in program ever. And it's like the, the, the main thing is I don't even know what you call it, but it's like, you know, when you prepare an ERD like a, like a, like a data model, it's like the pros, I'm about backing me.
01:06:59:16 - 01:07:17:14
Unknown
I've got like all these sheets of paper of these IDs. And it's like, once you've done that, it doesn't matter if you're using Python, doesn't mean PHP as a matter of using the AI. What is the not? You plug it into whatever it is you're doing and the code writes itself. But but like structuring that, having that like underline.
01:07:17:16 - 01:07:38:02
Unknown
Yeah yeah yeah. Like you can go into that, you go into business and you can map their business map market department, you can map a calculation system or you can map a data structure. It doesn't matter. It's the same thing. So like whatever that is I don't even know what you call that whatever that is. But but again it's just turning the business into, you know, like you're just codifying the business essentially.
01:07:38:02 - 01:07:57:15
Unknown
Right? Like, yeah, I mean, but but that's you. But I think some people talk about it too, like, and I think who was it? There was a, some guy was a relatively bigger company. Oh no, not big big, but right. There's a, why is it called, you know, it was a thing. My son did it for a little while.
01:07:57:17 - 01:08:19:04
Unknown
It's like this little, exercising, but it's really a program. But I think this one of the founders was, like, the main coder behind it. He's like, I have shipped so much code so much faster. But he's like, I don't understand it nearly as well as I used to. And I, I don't, you know, understand I haven't done that work to take the business into that, like as thoughtfully as I used to.
01:08:19:04 - 01:08:40:18
Unknown
So he's like, you kind of worries him because he hasn't done some of that, you know, mental side of it as much. But, I mean, I think the pros outweigh the cons in some of those. But you have to, there's often some trade offs where it's like, if you were more thoughtful about how you were turning the business logic into like a data model or code, then, it's even more important or more valuable, I think.
01:08:40:19 - 01:09:02:13
Unknown
I think the crazy thing now is that he can just ask her to explain how his code base works, and it will. It's it's a very weird time because I have experienced the exact same thing where it's like, I've been working on this big a genetic extraction thing, and I run it and it and it works. And then I'm like my, my senior devs like, oh, well, what did you do?
01:09:02:14 - 01:09:18:16
Unknown
It's like, I'm not really sure to be honest. I just had, you know, a day long interview with Claude where we went back and forth and I did what I do and beat my head against it until I got the outputs that I wanted and now it works. And then it's like, okay, Claude, explain to me how this works.
01:09:18:16 - 01:09:38:18
Unknown
Generate a mermaid diagram for me. I do all of these things. And then that's how you. It's a weird, but it's weird not knowing, like what's in the thing that you built. It's a very weird place. But I think, Paul, one of the things you hit on that I don't think enough people appreciate is because we get this question all the time.
01:09:38:18 - 01:10:04:20
Unknown
Like, what language should I learn? It's like, it doesn't matter anymore. The code doesn't like the language. The code is written in generally does not really matter anymore. Because like we, we we migrated our front end from Ruby to react in two and a half weeks, which is crazy. Like that's a distraction project just a day long, you know, in the old world, right?
01:10:04:22 - 01:10:30:11
Unknown
I mean, we probably maxed out all of our token, use for that those two weeks, but it's like at the end of the day, we were able to do it, and it's like, it's just crazy to say out loud because, you know, there's there's so especially on the engineering side, there's so much industrial software that's still written in like Fortran or COBOL and like, the people who know how to code in this language are dying or already dead.
01:10:30:13 - 01:10:51:18
Unknown
And so, like, what happens, you know, like I had a professor who worked at Lockheed doing missile, you know, missile defense systems and they still all that's this was 20 years ago. But all that was done in Fortran. And I was like, why? And she said, well, that's what management and the C-suite know how to code in. And so they're comfortable coding in that.
01:10:51:18 - 01:11:10:13
Unknown
And that's what, you know, they want to code in. And so it's like some of those problems like that whole kind of knowledge gap of like, well, how do we transition, you know, the incoming people into this language that's been dead for 30 years. That kind of goes away, at least with some of these language models and stuff.
01:11:10:13 - 01:11:33:06
Unknown
But it's, it's it's just a wild time, around all of this. Now, I agree that m the where's that where schematics are for iris and data and, like, images. And then I changed it to react. And so it was all front end stuff and then I saw somebody share me if I do. And I was like, oh, you know, a fast computer obviously had too much rubbish on it.
01:11:33:06 - 01:11:56:01
Unknown
So I had to switch it to the back end for the, you know, for the integration, the with the like language models as well change the Python. So it's changed three times already and the people on the front end don't know. Right. But other than maybe it's better faster. Yeah. Right. As long as it works. Yeah. What, what what coding, tools are you using on the LM side?
01:11:56:01 - 01:12:21:18
Unknown
And then we can wrap up. I wanted to ask you that I use cursor. So I've tried, like, I know there's a lot of know you mentioned Claude. I've. I've tried called. I did not like the total detachment thing. I'm not, I like cursor. I like, if, I turn off the autocomplete now, like, I just, I turn it off a few weeks ago, and I prefer it like, short term, maybe slower, long term.
01:12:21:18 - 01:12:48:05
Unknown
It's faster. I only use the agent stuff for like boilerplate. A lot of front, front end stuff. Just just repetitive stuff. And and if I've got like C, I'm using like an abstract sort of thing and, and I've got the structure right, then I can use agent thing, but it's like cursor with a lot of features turned off and, and you know, like if I'm at if I'm stuck in something, you can ask a question like, you know, what's the right and you know, what can I do here?
01:12:48:05 - 01:13:20:15
Unknown
But just to know, that's cool. Our, I think our, our most cracked out devs use clod in cursor. But I, I agree with you because I use cursor pretty much all year last year and then in I think December we got our cloud code instances. And that's what I've been using mostly. But I don't like it is a it's a weird feeling being abstracted from any of the data or what it's doing or anything like that, because then you genuinely have no idea what the hell is happening.
01:13:20:15 - 01:13:38:17
Unknown
And so when you get yes, it might work, but you're at the end, you're like, what? What did it do? How does this even work? But having having the cursor, having clod in cursor at least lets you be able to see all the files you can click into the files, you can see what it updated and all your diffs and stuff.
01:13:38:17 - 01:13:58:18
Unknown
And so I, I agree with you especially, you know, I tell people I'm a mechanical engineer because I'm a very visual person. And so like I, I like to see especially on data projects. Right. Like I want to see the data output. I don't want to have to ask you to see the data output and then you can't render it in CLI because it's a Json or whatever.
01:13:58:19 - 01:14:23:13
Unknown
That's yeah. No. Highly recommend recommend for cert. Definitely. Definitely like it. All right. So all right. But I will speed round round. Okay. So I was like in the air. You're is it is air as I say it. Scotland. When I was looking because my Scotland geology is pretty poor. You're not really close to Edinburgh or Aberdeen.
01:14:23:15 - 01:14:41:21
Unknown
So, I mean, there's a, Is that where you're from, or is there a good oil and gas? You know, footprint there and not really. No, there's not the zero here. There's nothing here. The only oil and gas is in Edinburgh, which is like 4.5 hours away. Sorry. Aberdeen. Sorry, not Edinburgh. And a 4.5 hours away.
01:14:41:23 - 01:15:01:12
Unknown
Edinburgh is outside the country. That's where a lot of like, well, a lot of things on their air here, there's like tourism. Okay. No, sorry. That's so it sounds similar to I went to Ireland a couple summers ago. I'd like to like do and where? It's like, you know, very small, but, you know, pretty popular, you know, for some tourism and everything, but.
01:15:01:14 - 01:15:26:03
Unknown
Okay. Well, but if, if we were going to take a trip to Ayr, where where would we be and what would be our drinking has. Yeah. Where would you ask? I have young children. I don't go much. We and we do to. We feel we feel that. Yeah. The my preferred place for eating McDonald's because they can make a you know, they make a mess, but nobody cares.
01:15:26:06 - 01:15:53:04
Unknown
Other kids are screaming to like, the if you come to Ayr there's, there's all sorts of like traditional unnatural and there's also it's tourists go to Edinburgh. But if you come, come to that side of the country like I'm not getting political, but like Donald Trump's got his place to just there and it's like really fancy and and it's like, oh the there's a few whiskey places and, tradition is Robert Burns like famous poet.
01:15:53:04 - 01:16:12:15
Unknown
He's all around here. So it's traditional Scottish stuff. But to be honest, I think we just go to either McDonald's or Indian Indians. Like, Scotland's good for Indian takeaways and stuff. So what's something interesting on the menu at McDonald's that would probably surprise us, do you think? Yeah, I get a Big Mac, my kids get chicken nuggets.
01:16:12:15 - 01:16:30:21
Unknown
Happy meal. It's a there's nothing like there's nothing. So I know some places you go like different parts of the world or there's like a more some local spin on it. Is there any kind of local dishes or anything? No. No. They, they if they do spend, it's like a like, no they don't. It's the McRib.
01:16:30:23 - 01:16:52:01
Unknown
Yeah. I, I mean, I don't want to do that. Okay. Fair enough. No. There's nothing. Do you. What's your favorite whiskey? I like him, love Laphroaig. It's another good whiskey drink, right. You can drink too much. A bit too easy, but like I, I bet from a Laphroaig, like a different. It's like it kind of tastes. And I don't know much about whiskey.
01:16:52:01 - 01:17:05:22
Unknown
Yeah. I've got a bottle of Laphroaig for my birthday that I still have to crack. Like we were talking about this weekend with, with my dad. I was like, I've got a couple nice bottle of scotch. If I, if I actually like scotch. Yeah, yeah. Me too.
01:17:06:00 - 01:17:26:19
Unknown
Like, you have this. I don't know if it's been there for for ten years. I think half the box was drunk at young kids at the end. You don't get a chance. No. Well, then you feel like I randomly found a bottle of, it's. I'm a big bourbon, or I was. I don't really drink that much these days, but, used to be a big, like, bourbon fan, bourbon collector.
01:17:26:19 - 01:17:51:02
Unknown
And I found a really good bottle in this random, like, corner liquor store in New York before I got on a flight, and I almost left without buying it. But I went back because it's like you'd never see them for sale. No less. But it was like, I don't know, 300 bucks or something, 400 bucks and, so then I got home and I was like, well, when the hell am I going to drink this?
01:17:51:07 - 01:18:10:00
Unknown
Like, because it's like I start doing the mental math of like, okay, it's this many ounces. And then it's that much per drink. And I'm like, okay, well, that's a lot more expensive than my, you know, $30 bottle of, off the shelf bourbon. But it, yeah, I have a very similar story to that. I feel like.
01:18:10:02 - 01:18:35:17
Unknown
But we, we, I think using it on special occasions and celebrations and stuff is a good way to still drink it, but not just drinking it casually because I don't like buying things and not enjoying them. I'm a little more technical. On what's your favorite cloud platform I used to use? I don't like AWS. That is I mean, it's just it can do everything, but it's it's just me.
01:18:35:17 - 01:19:03:13
Unknown
It's like, is it too much faffing about? DigitalOcean is good. But but recently I moved to railway, and I think it's kind of like Heroku or some like, but nice. I think we've heard a couple that from a couple folks was at the country these that I think or I don't remember. We've got some of our actually our growth team is using cloud code and then pushing the deployments on railway to build internal apps and collide.
01:19:03:13 - 01:19:23:21
Unknown
So that's that's pretty cool. That was the first I had never if I'd heard it, I hadn't paid attention to it, but I had never heard of of railway until that point. But I've been helping some of them debug APIs and search on it, and it looks awesome. It seems incredibly easy to use. It's one of these things you're like, is it going to dirt cheap?
01:19:23:21 - 01:19:45:12
Unknown
But no. But like in a year's time when you're when you're stuck on, is it going to just 100 times the price? So I don't know. It's good. It's really good though. Paul, thank you so much for, spending your your afternoon with us. Yeah, yeah. Thank you so much for that. Some of these were able to get you on, you know, from post, production here pretty quickly.
01:19:45:12 - 01:20:04:22
Unknown
Another other time. This is not so easy. So thanks for, working with us. It's, It's been a pleasure. Thanks, guys. Again, new energy bites. Merch is is available over on the doghouse. Please go. Grab that if you want some. Also, please make sure you guys, subscribe and like and share and do all of those fun things.
01:20:04:22 - 01:20:10:15
Unknown
But, Paul, thanks again and we will see you guys next time. We appreciate it. All right. Thank you.
