0:00 All right, well, welcome to another insightful episode of Energy Bites. I'm Bobby Neil and I got the Custitro host at Trusty Co-host. I think there was a Freudian slip in there. Yes, I also I
0:14 was just making fun of a lady on Funniest Videos. Five seconds ago, I am the one that records a lot, so check. But John Kalfand is here with us. How are we doing? Awesome. I'm exhilarated to
0:26 keep the adjectives coming for each new episode I've got to get a bank of a little bit, I'm sure Chad TBT can help with that. But excited to have Usan Raman with us from now, the AI agency. Yeah,
0:40 that's right. Appreciate, you know, excited to be here. Yeah, man, the long time coming, I think. Yeah, for sure, for sure. It's been a while. I've, I kept, you know, pushing off,
0:48 scheduling this muknag. I just got to do it. It's time to rip the band-aid off. It's time. Let's do this. I know you got, you know, a little one on the way. It can be much less time for that.
0:56 Yeah, exactly. Go and knock out all the things. So for everything before October, we got to get done. But at the same time, people have seen you. I mean, you're a collide superstar, so. I
1:04 try. I try. But, yes. I need to check the leaderboard and say, who's the top of the leaderboard right now? I think I might. Well, now you guys added that monthly kind of view, which is
1:15 actually very clever. I'm very at the bottom of this. I am behind the game of mine. I'm glad you noticed the cleverness of it. No, yeah, I'm gone. No, I didn't My part, I had my time.
1:26 Somebody else is now running the show. I forgot who's top now, but I got to look back. Yeah. Now, the monthly thing was a good idea from our team. I was just like, oh, yeah, that does make
1:37 sense. No, I think it's really smart, too. It can incentivize people to kind of continuously engage with the community. Yeah, ever since you and Reed broke the system and ran up, y'all found
1:46 the glitch in the - It wasn't a glitch. It was calculated. figuring out the algorithm, how it works and went all in on it. No, I'm not. There's no, no shade being thrown here. I am absolutely
1:58 the, the person that would also have done the exact same thing. Once I recognize it, but props to you guys for recognizing that. But yeah, no, it's a, I'm glad to have you on. I'm excited to
2:08 talk AI and all the fun stuff and have, it's always fun having, you know, just people that you're actually friends with in real life come on the show and talk. You know, so not looking forward to
2:19 it Not for sure. And I know you've been at it a little bit longer than me, but I know we're on a similar journey. And I've been able to share some, uh, compare some notes for sure. Yeah. Yeah.
2:25 But, um, so,
2:28 I mean, like, let's just talk, I jump into like, what are you up to? Like, what are you working on these days or what is AI, you know, all the, all the things? Yeah. So AI is a consulting
2:37 agency that I started towards the end of last year. Um, I kind of told you guys both of the story, but essentially, you know, I've been in oil and gas for probably going on the 13th year now One
2:47 of the corporate ladder have you know, been grateful to have a lot of great experiences and kind of and see different things. But it was just time for something a little bit different. We had our
2:57 daughter at the time and we needed a little bit more flexibility for the family. I was I was away quite a bit. It was really crazy times back then. And it just so happens, you know, sometimes
3:06 like really mysterious things happen in life. And they kind of just push you in a direction. And those are the things that started happening. I have three brothers and they started co-locating back
3:14 to Dallas. We found out somebody, you know, one of them was having a kid and a number of other different things And it was time of like, I think it's time to make, you know, a big move. So we
3:24 decided to go ahead and, you know, quit the job, sell the house and move it to Dallas. And what I did is essentially set up a consulting agency, specifically for tech and oil and gas. So the
3:33 whole premise is, you know, we're just a team of petroleum engineers that are tech-forward and we have oil and gas teams automate, streamline and optimize. So that's everything from automating
3:41 your Excel workflow sheets to building dashboards to ML and AI solutions, kind of where really would make sense. So that's kind of what we're up to. I mean, does AAI stand for anything in
3:51 particular? Or, I mean, is it AI? Obviously it's a part of it. So, is it a yellow pages hack? No, I wish back in the day, this would have killed. That would have been perfect. Everyone did
4:00 A's in the video, and this is double AA, you know? So actually back, whenever like data science became a really big thing, seven, eight years ago on machine learning, et cetera. Well, as the
4:09 time I was working at Hess, and I was trying to figure out a way, like how do we explain what we do? 'Cause it's like, we were trying to figure out, it was at the time, it's like, do you have
4:15 like a central analytics team? Do you embed people in the assets, et cetera? Well, we kind of do a lot. So I came up with this idea of like, well, analytics, automation, innovation is kind of
4:23 what we do. So I kind of just tagged it in as AI. And then you have a little bit of AI in there and between. And I kind of just kept that. And that's what I actually found the company way back
4:32 when, but I never really did anything specifically with it until it was very recently. It just turns out, it's great because it goes with the Gentic AI as well. So I think I'll cover my bases
4:40 there. But from an SEO perspective, it's terrible because everyone searches AI. The only way you get out of top if somebody misclicks an A essentially, right? But that's what it stands for. Okay,
4:50 cool, yeah. You were a visionary. You knew that the Gen-Ticki - I knew exactly what was coming eight years ago before anybody else. Like a keyword hacking game. So I think, I mean, something I
5:00 want to cover, and I think this kind of flows into it, 'cause it sounds like you said it's not just AI, right? And like, I think you're seeing it now when you step in with different operators
5:08 with Collide or whatever, and I think people want to use AI as that, you know, just hammer, you know, that's the wrench, that's the sledgehammer. Yeah, and like, I mean, how often are you
5:18 stepping in like people maybe bring you in for what they think is an AI, you know, solution problem. And you're getting like, no, you just your process sucks. Or, you know, this can be done
5:28 with, you know, heuristics or whatever it is. I mean, like, I think that's a problem a lot of people running into an hour. They just want to bludgeon everything to death with a AI or USB neural
5:37 networks or whatever. And it's like, they don't need that. Yeah, it's the same cycle as it was seven, eight years ago, right? Yeah, it's the exact same thing. It's the same thing, you know,
5:45 AI back then, you know, it's fun the way that we define that, I guess, from like an outside perspective in. AI back then was machine learning. AI now was kind of these LLMs, right? And
5:53 everybody thinks like once they see these demos, it can do everything, right? And now that I can chat with it, oh, it can optimize my production profiles and do this. And so that to answer your
6:02 question all the time. So we go in and it's like, oh, can it do this? It's like, actually, you don't need any of that. What you need is just like a solid data engineering pipeline and a nice
6:09 dashboard. And you'll get 85 of the answers there. And what we do actually, like I much rather start simple. People, once you start introducing these systems, BMSU learning models, LMs,
6:18 whatever it is, has its own other infrastructure you have to worry about. Version, you can control this and that, right? So I'm like, always start simple and let's start from there and let's see
6:27 if that works. It'll get you almost all the answers you're actually looking for. We can always add in some additional features to the process, right? But I always try to dumb it down. So maybe
6:36 almost overwhelming majority of the requests that come in is literally just kind of scoping and figuring out it's like, now you don't need this, you just need this. There's even one case of like,
6:45 you know, Man, I want to build this air pipeline. You know, it was a good use case, like you're scraping data from PDFs, et cetera. I'm like, you know what, actually you need to do just up
6:51 the subscription service, and so you get the raw data from them instead of you having to scrape it. And that's a lot cheaper for you to do. I think that's one of the things that the guy I referred
6:57 to you. Yeah, yeah, it was a great talk. It would be a lot easier, too. It's like, you don't want to have this headache specifically about this, just up the subscription can get the data. I
7:05 think for that particular service, I think there's people hidden in Flux Point where it's too much, but like, at that scale, that's, yeah, just pull down the Excel or CSV It's a great business
7:14 model to store that database and just force people to get PDFs out of it, kind of as a whole different talk, but anyways, yeah. If we can get it and imagine what company that would be. There's
7:24 not just one, as well, I mean - Not trying to get sued on the podcast, but yeah. There's lots of them. There's lots of cloud-based softwares in the industry that for whatever reason, moving to
7:37 the cloud allowed the user to have less access and less control of their data than they did before. And it's absolutely crazy in my mind. a year 2025 when Colin's car drives him to work by itself
7:51 that we can't get basic API access to, you know, a software that a client is probably paying tens of, if not hundreds of thousands of dollars to have, which is, it's just like this crazy
8:02 conundrum. Let all the dollars do that, yeah. But that plus, if you get an API, it probably is terrible. More than likely, yeah, it's not documented for sure. Like one of them, one of our
8:12 devs was doing research on the API and the only thing he could find was that there were API updates mentioned in the release notes, but there was no documentation on it actually. Yeah, I had the
8:21 same thing. So we were working on a small project. We're trying to tie into essentially a model and we're looking at the documentation of the specific model and they had all these endpoints and
8:30 these features, et cetera. And we're being, why doesn't this Python package work? Like why? And then we contact the sports like, Oh, we never built it out, actually. Just in the documentation,
8:38 we were gonna get to it at some point, right? Oh, yeah, no, I mean, thanks for leaving the documentation. Yeah, it's like, why? It's just like, don't publish it until it's ready, right?
8:44 And we're like, why is this not working? Yeah, well, I mean, I had when I spent a couple days working on integration with this API and realized like, I can't get numbers to tie out to what I see
8:52 on the front end and talk to them like, oh, well, if like a vendor is rolled off the platform, like, we still have the data. But like, you can see it on the front end, but like, you can't
8:60 actually get it from the API and like, I wish you to tell me that, like two days ago, because now this is useless and it was a bizarre API to work mess with anyways. And like, you've had some,
9:09 like, where it's like a production app, a well production app, but they don't actually have a well header or any kind of like Yeah, it's all, it's all route in actual job driven. And it's like,
9:21 so there's no main table of your assets that you're doing the jobs on. There's just a table of jobs. And then within those jobs, there might be assets. There might be a bunch of other things.
9:31 Interesting. So it's not like not linked or anything like that. There, there, there's a lot of relationships going on in that, that API, but there isn't just like a master table, which is makes
9:41 it. So you have to almost like a burst engineer Yeah, it's annoying. but yeah, we see that all the time. Poorly documented, no documentation or it's like, or it is documented and there's
9:52 nothing there because - I think in the year, I can go spin up what a fast API with Python and it generates swagger documentation. It's like that. Like for you, that's just part of the package.
10:02 Yeah, yeah, yeah, exactly. But meanwhile, like the largest, one of the largest vendors out there, like I just got PDF documentation of an XML API that I'm not gonna use. That's my favorite is
10:14 when it is natively stored in a digital format and then they convert it to a PDF just to put friction. It's a business model. I see it. Why do we keep buying into this shit, guys? Like honestly,
10:28 it's crazy. Getting the data in isn't the problem. It's always getting the data out. Everyone listening, please focus on that, when evaluating your softwares, because that is where all your
10:38 problems will ultimately come from. Like so many of the things all of us are working on are not necessarily the most complicated tasks in the world, The challenge comes from accessing the data,
10:49 which is crazy because the operator owns the data at the end of the day. 100. Yeah. So I mean, how much are you seeing right now? Like that, I mean, obviously you got the AI in your name and
10:58 you've got, you know, you're doing the work on the AI front and I guess we can get into Petri here shortly as well. But how much of that, like, are you getting a lot of interest on the AI front?
11:08 And then, you know, it's funneling in there or people just want to still just need a lot of help with data. I mean, like, you know, honestly, where I think people, a lot of people are window
11:16 shopping right now in AI. Everyone is very interested and very intrigued, but they don't know exactly how to execute against it, right? So the conversations usually start around that of, okay,
11:25 you know, I see you're probably in the space doing something like this, you know, just curious what others are doing. Everyone's so interested in what everybody else is doing, right? But usually
11:31 when you get to the problem, you know, again, talking through the problems, it's a data engineering problem at the end of the day. Almost always it starts there, right? And then from there, so
11:39 like, you know, the way I like to think about is, you know, streamline, automate, optimize is kind of like the three phases that I go through and streamlining is just like. Can you first even
11:47 see your data in a dashboard, right? Can you actually do you have that? And if you don't, that's probably where we should work on. You build up your pipelines, you did engineering flows, if
11:54 database, if you don't have one, et cetera, and let's get you there. And then from there, then you get to automate where like, okay, now you can see stuff, now you want to do stuff on your
12:01 data, right? Then you can start automating all these repetitive tasks. You can sprinkle a little bit of AI if you want, you know, in some automation, but really what it comes into is kind of the
12:08 optimization portion of any touch data science machine learning. But it's really a lot of it, like, you know, is just people coming in asking and being really curious, talking about their
12:16 problems. Like, yeah, you don't really need this. You probably are. You fit in most likely in phase one. And it's focused there. I think a lot of people, technology is the tech companies are
12:23 really good at, like, making shit seem so easy. And most people don't, you know, we saw this with the ML stuff five, seven years ago, where it's like, oh, yeah, ML is going to change
12:35 everything. And then it took five to seven years for us to realize, like, or before it started having real impact in the industry and I credit a lot of that. to the fact that we had a bunch of
12:46 shitty data and it took us five years to go back, re-architect, you know, apply all of the good QA, QC methods to those data pipelines to get the data in a place where we could then feed it to a
12:59 machine learning model and not get a bunch of shit out on the backside. Because I think the biggest thing that people don't generally understand is machine learning models, large language models are
13:09 predictive models. They are based off of what you give them. And so if you give them a bunch of crap, they're gonna get output a bunch of crap. And so the data that goes into these things is
13:17 incredibly important. And if you're not focused on that first, and you wanna jump straight into machine learning models or LLMs, you're gonna have bad results and you're not gonna be happy with it,
13:27 but it's not because the tech doesn't work. It's because it wasn't architected the right way from the jump. 100, I mean, they say like, you know, you've got ML, traditional ML, it's like 80 of
13:36 the workload is just data pipe, right? Just making sure that it gets in the future engineering portion, et cetera. The smallest amount of time is actually training it. A little bit different for
13:43 LLMs, but in that specific world, it's essentially your contextual information, that context that's critical, right? And that's kind of where the majority is. You have to have that for these
13:50 systems to actually perform well in your specific vertical space or else they're not going to do well. Yeah, I mean, RAG is the same way, right? RAG is a little different than a traditional
13:58 foundational model where you're fine tuning, but it's the same thing, right? It's got the only context it has is what you give it. And so if the context in there sucks, like we were doing this
14:08 project where we got 75, 000 scan pages from bankers boxes that were in an operator's office. And it's like, if you can read the handwriting on the PDF, we can extract it, but there's a lot of
14:20 handwriting in there that's like done in a Sharpie on rice paper 40 years ago, in cursive, that guess what? The AI model extracts a bunch of junk. One of them explicitly says butter and pain,
14:32 which I'll never, I wanna make a shirt for a client that just says butter and pain because it's so funny. But it's like, I'm not gonna put that in the index because butter and pain have no. no
14:42 context in the oil industry, right? Like it's junk data. So that doesn't need to go in there. But if you're not paying attention to that and it does, and then you get bad answers, you're like,
14:51 oh, well, the technology sucks. It's like, no, 80 of all of this shit is data. Like you got to get the data right. Yeah, I mean, I think where they
15:01 all can shine right now is maybe on this, like where OCR was falling short and those kind of things. Computer vision. It's me, Ridge and stuff. And it's like, all right, now this is just a
15:09 better alternative And that's, this is a spot where I can rip and rip place. Right, it's still a tool. That, but like around that still needs really good. I went from a manual screwdriver to a
15:19 drill. Like that is the different, like that's exactly how people should be thinking about these. All of these things are tools in a toolbox. Just because you have the tool doesn't mean you should
15:27 use it for every single job, right? Like you're not gonna hang a nail for a picture with a sledgehammer. You're gonna use a smaller hammer to do that, right? But I think that's where a lot of
15:38 people, you know, they get these promises or they see the marketing of like, oh, it can do all this stuff. And it's like, you, I can only imagine how much time and effort went into just the GPT
15:49 demos that they did at their demo day, the other day. And even then the graphs and stuff that it was pulling up were wrong. Oh, I know that it's classic. I mean, that would be the scaling and
15:57 stuff. Yeah. And everyone saw it. And it's like, if you would have, I gave it, I took the screenshot that I took on my phone and I gave it to chat GPT and I asked it what was wrong. And it's
16:05 like, it identified it And it's like, you guys didn't even use your own tool to review your deck. I remember seeing this, right? I was like, I kept looking and I'm like, am I missing a second
16:13 axis here? Am I not seeing it? Like, what am I just stupid? Am I not seeing it? But to play on the other side of it, so like the use case that I have found for this, because it is absolutely a
16:23 tool. And I think it's what people understand. This is not like a bandage for everything. It's literally just a tool when you have to use it for the right job. I think what I found for the
16:30 specifically, the L. Amuse case is transferring something unstructured to something that's structured and that's that's where it shines.
16:37 or extracting from an unstructured into a structure. Correct. So like, you know, use cases like you had a company that bought out another one and they had a bunch of operational notes. And this
16:46 new company called Company A, right? Had these downtime codes and these equipment ideas. And what they need to do is digest these years worth of operational notes and map it. That's a phenomenal
16:54 use case for that, right? Something like that or like in the classic example of being able to extract text from a PDF document and then translate it into something else, that's phenomenal. But
17:03 it's not gonna optimize your production out of the box. That's not what it's gonna do We get that question so often. When we go into clients' offices, they're like, Oh, I want this GPT thingto
17:11 optimize my artificial lift. Like, that's not at all. Like, that's not at all what it's there for, right? Like machine learning was built because we as humans can't understand all of the
17:23 numerical data that we've got. Language models are the exact same thing, probably built by some federal agency at some point to spy on all of us. But ultimately, but like, yeah, ultimately, it
17:34 now trickles down into the public sector. And it's the same exact thing. We have so much text data that we can't comprehend all of it. We can't find everything. And so we can layer this language
17:45 model on top to help us with that. But it's not this magic, like that's ultimately, to me, it all boils back to educating the users on like, what they are, what they're capable of, what they're
17:56 not capable of, what you shouldn't try to use them for vice versa. It's a lot of it Again, it's just breaking that psychological piece of like, well, I've seen this or I've read this and it's
18:10 like, yeah, but you know, oh, well, it's the best on this benchmark. And it's like anybody that's training or going after a benchmark is training just for that benchmark. They're not training
18:20 for anything else. It doesn't mean anything. There's a lot of controversy around that. But yeah, it's yours, but I think people just see like these demos and they work, but like to put a little
18:28 bit like behind the scenes, since you know, you guys know this, It's just a lot of tooling behind it, right? what these tools are doing is taking that unstructured prompt, turning it to a
18:37 structured tier, the steps, and for each one of those steps, you have X number of tools that's going to call anyways, right? So that optimization is actually not happening, like for production,
18:44 let's just say, actually happening in the model, it's more so a call to an optimizer, and we'll take those results and give it back to you, right? It's just a level of orchestration, essentially.
18:53 Yeah, now that's a good point. There's a ton of orchestration and tool calling and stuff happening under all of the foundational model LLMs that when you're using them, you don't even realize it.
19:04 Like the average person doesn't. But like anytime it's writing, they all have this little canvas piece now where it's like, oh, you need to do math. Okay, well, math belongs to Python. So I'm
19:13 gonna go write the Python code that does the math and then give the answer back to the language model, right? Like it's, but yeah, that's a very good point because a lot of people don't think
19:21 about these as like layered, it's like a cake, it really is. Like you just layer things on top. Well, even, and we talk about this all the time, but we actually, well, you're an engineer too,
19:30 but we have another engineer here. Explainability also is a huge thing. It's like, if you don't understand what's going on, it gives you an answer and even if it's right, but like you want to
19:40 understand why it's right. Or, you know, especially you deny your clients or people that we're working with are, you know, really intelligent engineers, geologists, like scientists, like they
19:48 want to understand why it is working. How did I get to this answer? And can I get to that same level of efficacy again? For sure. I mean, like what, so when I was back in the engineer and back
19:58 in the Hestays, we're working on essentially building a machine learning model to be able to predict early time performance of these wells. This is before plugging and perf, et cetera, up in the
20:06 Bakken. So we're trying to leverage public data to be able to understand what we do. The biggest challenge, once you move away from a traditional hyperbolic decline, type curve methodology,
20:14 something more machine learning, is how do you explain it to management? And how can you show it that actually like this is, it's within the ballpark, right? And, you know, we tried all of
20:23 these models from linear regression, all the way to the fancy stuff. And we ended up leveraging essentially like a random force model just because that's kind of like the medium between - Yeah,
20:32 because you can see the whole decision tree and you kind of claw it back and we built all these fancy dashboards showing how it works and like you click on a prediction and they'll show up on the map
20:41 of exactly the use cases that it used when the decision tree to make it to help explain the ability. But whenever it comes down to like actually making an investment, that's a big, big thing. If
20:50 it's making decisions that are worth hundreds of millions of dollars, you want to know what that thing is doing. And with these things, with the LLM models, it's kind of obscures it a little bit
20:56 because you have to have a massive amount of traceability and even then it's still, it's not a deterministic system, right? Whenever you do machine learning, at least you have the parameters and
21:04 they're kind of put in place. The same exact input in is the same exact input out. Whereas with these, depending, you know, you might not get the same exact thing, even if you put the same thing
21:11 again. No, that's, I think that's another big piece of like the gap between, you know, a script that I write for myself versus, you know, a prod piece of software is like, especially with
21:25 language models, like test the repeatability and of them because, you know, And there's ways to change that and adjust that. There's different parameters you can play with, but like out of the
21:37 box, the temperature is set way too high. Like there's a lot of things that you as the user should understand about them and how to modify them in order to increase the reliability of your answers
21:50 because you're absolutely right. You can ask the exact same question to the exact same model and get a different answer. They might be in the same ballpark, but there's different answers. So then
21:59 you go back to your boss and he's like, What the hell? Yeah, exactly. And then how do you explain that right? Exactly. And also these LLMs too, we're talking about it. It's like each one of
22:08 these parameters you can fine tune, even the prompt itself is a hyperparameter. And just like in machine learning, right? You can fine tune that specifically. And as soon as you change your
22:16 foundational model that you're using, that's a whole different level optimizations, right? So you kind of like have to be careful the number of knobs that you twist and what you get, right? It's
22:24 a whole process in of itself. Yeah, I
22:27 always relate it back to like frack jobs. Like honestly, 'cause there are just so many fucking knobs to turn and you turn one of them and it waterfalls down into everything else. And you're like,
22:36 Oh, this is fun. Yeah. But in all honesty, that's why we've put our chips behind the rag side of things because you can trace how it ended up at an answer. And you can start turning those knobs,
22:52 at least to a degree. But the other big thing that I've said before,
22:59 but it's foundational, large language models are trained to give an answer, hard stop. Not the right answer, not the best answer, and answer. And so that's why when you tell it that it made a
23:09 mistake and it's like, oh, you're right, you're so right. I'm such an idiot, you're a genius, and it does all this stuff. But that's why. And so with a rag, you have way more control around
23:22 what it's the goal of what it's trying to do And so like another big benefit of rag, why we went. with that is like, you know, ours is set up so that if you ask it a question, like right now, if
23:34 you wouldn't ask it who the president was, it wouldn't, it would tell you it didn't know because it doesn't have access to that information which is what most people, that's some of the best
23:43 feedback we've actually gotten about our, our app is that it says I don't know when it doesn't know. And it's like, it's kind of counterintuitive that, you know, your product is not giving an
23:53 answer which is a good thing, but that is a good, that's what people want. But you'd rather have that than having a fake answer and then go down this rabbit hole, figure out if it's actually right
24:01 or wrong. Same thing you want from people that you work with too. Right, exactly. Yeah, I'd rather you just tell me, look, man, I don't know, I'm not qualified to answer that then I can just
24:09 sit here and espouse and sound like a smart person in the room and give you a bunch of BS and then you tell me I'm wrong and then I tell you, oh, you're right, I am wrong. Here's a different idea
24:18 that's also wrong. You're gonna buy the cap yet? They just started making merchandise of all of it, I'm surprised the other one had that. I haven't seen that yet. I need to send one of your way.
24:26 Yes Oh my god, you're so right. So before we get into you. kind of your plugin, we were talking about going from unstructured to structured, I think I tagged you in one of those posts recently,
24:35 but you know, Jacob Mattson, who we had on, and I don't know if the duck deep evil or made this plugin or not, but he's been showing it off like it's called CSV everything. I saw that on LinkedIn.
24:44 Yeah. Yeah. And like literally like you can like grab a chart, like a picture of image of a chart and it generates tabular data from it. It's awesome. Yeah. I saw one like he'd even have a
24:54 y-axis or x-axis, but he was able to figure out essentially the trends and then he just plot it That's great. It's awesome. Yeah. I was thinking like you could probably take that and apply it to
25:02 say to investor decks or or like you know investor calls and stuff like that and just you could create a history of all seven and you actually maybe go back and see if they were even like right about
25:11 what they projected or you know numbers you know because it's not just stuck in a PDF anymore. But sure.
25:18 But I think that's something like a chrome plug-in. But wanted to get on to so you've got a product now called Petri. That's right. A lot of people say Petri, so thank you for saying Petri.
25:28 Petrile, I'm engineering, I'm assuming. Bam, there you go, you got it. Yeah, so we have two tools, actually. One is Petri, and one we call the business toolkit. The idea behind Petri is,
25:40 once you start messing around with these systems, you find out, and to your point, is context is everything, right? Especially in a specific enterprise. And context, when it comes to
25:47 engineering, happens in these day-to-day conversations, as you do your analysis, et cetera. So I've seen it back previously before, and I'll give you a story. I was an operation production
25:56 engineer out in the pyramid, and I'm responsible for a few hundred wells. So I come in, and the first thing that I do is I open up a production profile, and I see the dips and swings and all this
26:03 stuff. So the first thing I wanted to do, of course, is analyze it to be able to figure out the context of the well, 'cause I need to figure out how to maintain this and optimize it. So I go and
26:10 dig through the well files, this and that, this and that, and then by the time I have this done, it's like, oh yeah, the previous engineer did the same thing, and now you're rotating out. And
26:18 then the next engineer comes in, they do the same thing, right? So the whole idea behind it is, Given these systems right now, people are very interested in using it. They are, they can be very
26:26 powerful for discovery, but they're only good as the data that you give them. So the idea behind Petri is we're calling kind of the copilot for only on gas. It just sits right next to you as you
26:33 work and allows you to be able to capture that context from your analysis, from your note to whatever it is. So the whole idea is that you just dump it as you find things. So it's an Excel Outlook
26:41 and web and Chrome at the moment. So if you find something in Excel, just type in a free form text editor, right? And we'll automatically take that, map it to the right assets, and that way
26:50 whenever you open up that Aster card again, you have everything you need to know by your assets right where you work. So the Outlook integration, you know, there's a ton of conversations happening
26:58 from field ops, especially late at night, something went haywire in the field. All those comers are really critical, right? And the problem is, is a lot of people have tried doing this knowledge,
27:08 or solving the knowledge problem, but what tends to happen is that once you force engineers, anybody out to a whole different platform, so just another task. And then you're forced to fill out a
27:17 bunch of fields, right? Whereas we're like, you know what? Microsoft Copilot did a phenomenal idea, and I love it, if it just works for our next you, what if we do the same thing and just drag
27:24 and drop? So now in Outlook, you just take an entire email thread, drag and drop it or process it, and now you have those insights back on those assets. So that's the idea behind Petri. Yeah, I
27:33 think we're huge on that. You know, we've talked a lot about that, whether it's knowledge transfer or institutionalization of knowledge, or it's just like, 'cause I even think now with those,
27:41 and then maybe like transcripts from Teams calls, like the morning operational meetings, and now you could actually have contacts and they could tie it back to the - Exactly. In Bob5H well, but
27:49 like now when someone comes back to it, like that work isn't lost Like, again, I hate seeing how much work is replicated throughout the industry, and we still see it now like five smart people at
28:00 Conico are doing the same thing and five smart people at Hess are working on. You know, that's a whole different problem, but then like if the work's already been done within the organization, why
28:08 aren't you leveraging it? 100, and I think it's even more so like I see this a lot in Offshore too. Offshore, you have so much less wealth, right? But the same analysis tends to get repeated
28:17 because the best you can do right now is maybe put it in like an Excel file and a bunch of lines and no one is really gonna go draw years worth of that, right? So there's a lot of power in being
28:25 able to capture that context, and then whenever you open up a well or whatever it is, you get the summary of everything that you need to know. And then again, where these systems come really
28:33 powerful is discovery. And that's why we use chat TPT a lot, right? So if you have that context, you start discovering, Hey, what did happen three years ago, a year ago, last week, et cetera?
28:41 That's when you start unlocking a lot of it, but you can only do that if you actually have the right context And so, is this able to tap into whether it's a well-view or, you know, your production
28:52 system, like where you've got some of these field notes as well, like that have been documented or downtime codes and reasons? Like, are you able to wire these things into Petri or? For sure. So
29:02 like, what minimum we need is kind of production data, and then whatever other data you give us access to, we would be glad to ingest that, right? So, absolutely. Now, what we're not trying to
29:12 do is kind of be the app for everything. There's many other tools that do really good jobs. We're trying to be more of a compliment to everything. And that's the idea behind working in the browser.
29:20 There's a lot of tools that are web-based web-based their for phenomenal production allocation systems, et cetera. We're just trying to just be your co-pilot, right? As you find things in whatever
29:28 systems, you spot for our dashboards, Power BI, just drop it in and we'll go ahead and mine that for you and save it for you. So that's what we're doing. So we can't tie into it, we love to tie
29:37 into it, but just wanna make sure like, you know, we're just a compliment to things we're not trying to replace everything. Yeah, well, I think that's in my mind, that's kind of the way of the
29:44 future, right? It's like we've got all these incumbent softwares in the industry and getting people off of those is ripping teeth Sure. Even though they can't access half their data. And so, but
29:58 I do think the future ends up being where you've got, and maybe they're just databases at this point, 10 years down the road or whatever, but you've got underlying applications and databases of
30:11 your data, and then you have a handful of tools that tie into all of that, that allow you to kind of centralize it and democratize it across the company, right? And so even book, you know.
30:24 ahead of that. I just love that you guys meet the user where they're used to working and I don't have to have another tab. Yeah, that's exactly the whole lemonade, the tab fatigue, exactly, or
30:34 another pop-up window, whatever it may be. So we're just trying to meet where users work the most and excel is still alive and well. That's for sure. And then you got Outlook and the web. We're
30:43 thinking desktop, we're not sure yet, but I think we covered our decent bases there. Yeah, no, I mean, when a majority of the industry still uses Excel, I think that's a hell of a place to be
30:53 Yeah, sure. So let's say I'm the director of IT and some young control engineer comes up and says, Hey, who sounds got this cool thing? I want to use it. And then he's like, Well, where the
31:04 heck's this data going? I mean, they can feed in
31:09 their notes and all this information. So this is something that they host or they can go install it today. And then where does this data live and how is it like? Yeah, great question. So it's all,
31:18 it's self-hosted within the cloud infrastructure of the client. It's all and this is sensitive stuff right this is their data So we we have in your entire package or we're just going to play
31:28 internally and the same thing with kind of the excel Addins and stuff you can go ahead and publish add-ins your organization And that's kind of how you distribute it. So it lives within the cloud
31:36 doesn't leave anybody's firewall whatsoever. It just stays there Yeah, okay, so it's not like it just be people can't just go know that tree from the market place We haven't give it to Palantir
31:44 anything like that. Yeah, no, we're not doing that. We're not in that business I saw they just bought what was it ora the ora rink? Yeah, I saw that. Did they? Yeah, there was a bearded there
31:54 was a bad post about Throwing my werea ring in the trash so that Palantir can't mind my data anymore, and I was like, it's too late
32:03 But also that's terrifying. Yeah, that's wild. I mean, there's some other some pretty big data engineering News yesterday. I think it was five train acquired on to Biko. Who's like sequel mesh.
32:14 Oh, I didn't see that. Yeah, really Yeah, it's actually end up being not really surprised a lot of people because I already knew like things as a George Frager or whatever, the five trans CEO
32:26 was an early seed investor in SQL Mesh. And then I think they had started using SQL Mesh internally, but then like Ben Rogajan sent us up my last night and it was, I was thinking the same thing as
32:37 like, all right, so here's the model. You invest your own money and then you get your, the big company or the CEO to go buy that. And now you profit on both sides of it. It's a great investment
32:48 strategy. I don't know what to tell you. Like someone else came in and said that too, it was David Yaffe who we'd had on from estuary, would be a competitor to them. But it was like, if his
32:57 company had to buy that company that it probably wasn't a very big, 'cause the only way to really make money is like a seed investor or whatever, is that that company gets big. Right, yeah, yeah.
33:05 I mean, I'm sure he made some money, but it wasn't like, he was creating earth, life-changing money for himself. Yeah, sure, sure. But sorry, I didn't mean to totally derail that. No, I
33:15 didn't see that. I don't wanna check that out after this thing but yeah, I can see what's called meshes. It was supposed to be the
33:23 one that ended DBT's reign, but I think once DBT brought in, they brought another SDF for it, I forget exactly, but it was basically a rust engine that they were able to plop in that allowed them
33:34 to parse everything faster. Like, see, we'll measure this kind of string, we'll hold on to being a better engine, kind of fell off by the waysides of it. Yeah, DBT is great. I love it. Yeah,
33:43 it makes things so much easier. I wish we had a way back when. Yeah. Before we were like doing a bunch of store procedures and all this stuff in SQL and just in the blind, no version control
33:52 whatsoever. Like I used it for the first time a few years ago. I'm like, Man, this is awesome. Yeah, it just turns everything inside out. Yeah. Like in just, because I was even showing a new
33:60 guy, the client who he'll probably be the one that maybe inherits what I've built and I'm gonna help get him stood up. But, you know, he's plenty capable. You know, we're kind of a reservoir
34:09 engineering guy, but does a lot of SQL. And I was like, dude, you can play in this, you know, when we set this framework up, if you can write SQL, we can do this. Yeah, yeah, for sure It
34:17 doesn't take any crazy knowledge at that point. Just got to get it set up the right way. It's a, no, it's a great framework. But her great things about SQL Mesh too, but it'll be interesting to
34:25 see kind of now where five train goes with that. Like someone was saying that five trains can be like the next Informatica.
34:33 I don't know. 'Cause now that you've got the full end-to-end Right, yeah, yeah. And transformation inside. But I guess my subscription's probably going up now because they had to buy that. But
34:42 do you expect the price hack in the next bill? So someone's out of either go bankrupt or you live long enough to see yourself become like
34:50 a cheap
34:52 software or whatever That's pretty good. Yeah. Sure, that's pretty good. Let's back up a little bit. How did you get into the spot? Like what's your, how did you get into the, what was your
35:02 gateway drug into like the data side of things? Right? 'Cause you're petroleum by degree? Yeah, I am petroleum by degree, AM, proud. I stumbled my way in. So Nick, the other Aggie here is
35:14 outside. So y'all can do your handshades. We'll do, I think we were probably did it your secret dance. Yeah. No, but what I started, so I started back with Hess, and I was out as a operations
35:25 engineer early on. And I was tasked to essentially what we're trying to do is forecast the number of rigs we needed to go for maintenance, assuming the number of sucker rods you needed, ESPs, et
35:36 cetera. And they had given me this Excel sheet, and they're like, yeah, you just, one iteration is gonna take you over a week or something. But there has to be a better way than this. What is
35:44 this, right? So I go and I found out VBA, that was my gateway drug I'm like, wow, you can actually code in Excel and automate all this. So I went and bought the biggest book I could find in
35:54 Excel. It was literally this big and a nice green cover and just started reading it. This is before chat GPT days. Now it's much easier, right? And started with that. So then I started building
36:04 like a little app and I'm like, wow, that feels awesome to like take this concept of having your mind and actually make it real. And that was my drug. So I started building that out, worked with
36:12 a phenomenal engineer, Chris Cardero. It was just awesome And we had put this together and presented it, One iteration now went down to like under 30 seconds. And not only can do that, but now
36:24 you can forecast the rig maintenance. You can see when you had gaps in your schedule, you could put those maintenance rigs on special projects, et cetera, right? And that was my start. I'm like,
36:32 wow, you can do this. And then luckily enough, I hadn't met. So Brock Mario, who you guys had on before. So actually, he's the one that got me into has. So the story is, and this is important.
36:42 So I study internationally at AM and Qatar. I transferred it over there from Call of Station. And it did a number of years there And it was an awesome time. But whenever I had missed the hiring
36:53 time to - that usually happens in the fall or something like that. And I was overseas. So whenever I graduated within 24 hours, I went ahead and just took a flight back to Dallas. And I was trying
37:01 to figure out where I'd do. I had no job, nothing. And I'm living in an out hotel. And I'm just spamming LinkedIn messages left and right. And I'm just trying to figure something out. And it's
37:09 suddenly this guy named Brock messages me. He's like, hey, I might have something for you in the summer. I'm like, sure, let's do it. And he's like, you know what, I'm sorry, I don't. I can
37:18 get you an internship. I know you're looking for full-time Mary Kay with that. I'm like, Yeah, let's do it. So he got me in, just got a rental and went to Houston, found a guy on Craigslist who
37:27 had it, or primed it up and took it essentially. Then my internship, and then from there, I got hired on shortly after Hess. I say that, so Rock is kind of the one that helped me stumble in.
37:36 And then afterwards, he was a mentor to me early on in the data visualization space, learned a ton from him and Spotify, et cetera, and really started taking off from there So I went from VBA,
37:46 then R, that's where I fell in love with TIDR and all of the packages, then Python, and then went to software development and the rest of the history. That's awesome, shout out to Brock. Shout
37:55 out to Brock. It's a pretty small world. It's a very small world. Because Brock had died with the high school, well, it went to the same high school, so he was in my brother's grade, but yeah,
38:03 too behind me. But yeah, I grew up around the family and everything, and then he's randomly reaching out to you on LinkedIn. Oh, you just like the web of people, yes. Yeah, yeah, it was crazy,
38:11 I still remember it, so I remember those days. That's awesome I mean, it's a funny time I'm on Brock though, because like. And when I hear what you're doing with Petri, I feel like if you were
38:19 able to plug that in with like Wiserock and maybe all of those conversations, like I think it'd be super powerful because it's like he's building that one stop shop for all that stuff already joined
38:29 in with the conversations and all that stuff. So it'd be pretty wild. It's the definitely in the back of my mind, let's put it that way. There were combos before I think it was early on. I didn't
38:37 have the idea concrete enough, but now that we have something we'll see, but you're on to something maybe. Yeah, also, anyone who says millennials aren't like efficient and resourceful It's a
38:47 full of shit because that's a hell of a story right there. Like, I just remember I had two summer internships in Little Rock for the same company, but it's like, yeah, I went on Craigslist and
38:59 found this shitty basement apartment that was the cheapest thing I could find that was decent enough. And I was just thinking about how like foreign. of a concept it is today to think about finding
39:12 a place to live on corrects. I found it and it was it was actually pretty decent. I just took it. I'm like, I need it. I start like tomorrow. You did it back then. There was no like that's it.
39:21 And we just made it work. You know, and it kind of all the pieces kind of fell into place. I felt like it was definitely fate for sure. But it all worked out. Very sure. That's awesome. And
39:30 tell us have a Mac one more time. So we talked about Petrie, but you said there was another tool also was there is so so Petrie's kind of like a data mining for for context. There's a ton of stuff
39:41 that happens Excel and what engineers tend to do is they rebuild the same exact excel sheets over and over, be it worker over economics, economics in general, forecasting you name it. So I've seen
39:50 it the entire my career, and I'm sure you guys have to. So the whole idea behind this is I found out how powerful these Excel items can be literally software, you can build whatever you want in
39:58 them. So I'm like, why not also just centralized your key workflows into this one spot. And instead of you rebuilding these models and may potentially make mistakes, etc. You just have access to
40:07 them. So you work the way that you want to work and so you want to run automated forecast, basic example, right? You just highlight your input data, you just tell it, this is my inputs, click
40:16 run, and you tell it where you want your output so you can just put it right back out. Maybe you want to publish a forecast to an internal database, right? Now I see companies like, they download
40:23 an Excel sheet, got to send it to a guy, a guy sends another guy and pushes it in, well, what if you could just highlight your data and just say run and push, right? So it's just like a, we
40:30 call it like a business toolkit, essentially just a central way, a way to be able to centralize all your key workflows in one spot and give it to all your engineers and they just run it whenever
40:37 they need to It's a really smart, it's the big problem with Excel is that it's not distributed in general. I think you, so Excel is phenomenal what it does, it's a great UI, it's low barrier to
40:48 entry, right? You customize it the way that you want, there's a reason why people love it, right? But the idea behind this toolkit is maybe we can take a little bit of those downsides out, right?
40:55 Instead of having to have the business core logic there, what if you just strip it out, let people do what they want to do, but give them access to it as they need it. And that way then too, if
41:03 you're running some really heavy workloads, you just strip sit on the backend, right? If you want to do forecasting for your hundred wells in your Excel sheet, you don't have to do it in your
41:09 Excel sheet. You just send it off and it comes right back in a few seconds and you're done, right? So that's the idea behind that specific toolkit is just being able to centralize those key
41:17 workflows and then give it to our old engineers. Yeah, it's funny to bring up like work over economics, but even just any of these, like, and I was saying earlier, you've got smart people at 10,
41:26 20, 100 different companies building the same shit that they need to rebuild Exactly. Every company I've been in, like, the production engineers have their little, whether it's Excel sheet or a
41:35 spot fire thing that does their back of the napkin and work over economics because they don't want to send it to reservoir to run a full blown, you know, areas or peep or whatever, you know, kind
41:44 of like economics on it. But it's like, but then you a new company starts and someone rebuilds it again or, you know, it's like, why are we doing this again? Like, it's not about, like, I was
41:54 going to say, that's the other big thing Like, from my perspective is, like, if I'm in management, it's like Excel is kind of dangerous in the sense that it is the wild west of, like, do you
42:03 want to use this formula versus that one? you can, and if you're not experienced enough, you might not know why you should use one or the other, and that sort of thing. So being able to have like,
42:14 this is our company standard place for all of these things to live, and this is the standard that the company uses, makes a shit out of sense. Yeah, that's the idea behind it. Yeah, no, that's
42:27 awesome. And then you can also, you know, there's a plane to that as well, where we're talking to a few consultants who are, you know, aging and ready to go, but they have a phenomenal amount
42:35 of work. And they've done some really interesting things. They found out exactly how to do economics correctly, artificial lift optimization, et cetera. But they don't want to go and develop a
42:43 whole software suite. They don't want to go into that game, right? So this is kind of like almost a bridge to that. Like, you know what, we can also offer your solutions here and just distribute
42:51 it, right? We'll forgot the revenue, model cost model, et cetera. But at least you're capturing that domain knowledge before it goes. And they're actually, you know, winning off of it as well.
43:00 So the whole idea is also to bring in some of those standardized workflows learnings from these experts. and be able to provide it in a kind of a central way. That's awesome. Yeah, 'cause there,
43:08 I mean, there is so much industry standard stuff. Yeah. That's like, why are we doing this again? Yeah, exactly. A log, log plot is a log, log plot. Oh yeah.
43:20 Okay, so we started working, you know, from back forward a little bit. So we got VBA and then like spot fire and all this stuff. So I mean, like what's been kind of the evolution, you know,
43:29 you know, from there? I mean, like, whether it was with Hess and then ultimately to Waterbridge from Hess, like what does that kind of progression look like? Yeah, like a career standpoint. So
43:39 I was, so I did operations, production, reservoir at Hess. I did about six and a half years there. It started my career, it was awesome. Kind of bittersweet, honestly, to see it get sold, to
43:48 be honest, you know, the Hess name. Oh, they spun off some really cool, like there's some really smart people and smart things coming out there and like innovation. Yeah, end of an era. So at
43:57 the time, you know, I was ready to kind of jump. You know, I realized that I was going to probably stay the same position I was for a number of years and I want to see something a little bit
44:04 different. So I got acquainted with the Waterbridge guys, and at that time it was right before they were about to go on this massive exponential growth curve. And the beauty about that was, is
44:15 that they had nothing, and that just excited me. I'm like, this is perfect. Blank slate. Blank slate, and you know, the executives were entrepreneurs, and they're just like, go figure it out,
44:24 you guys got it, right? And that was awesome. So I came in and kind of grew with the company, became a director of analytics there What we did was in Waterbridge, we had built some phenomenal
44:34 systems, and we had built a tool called Wave, which I believe is still using. Essentially, it was a way to be able to, like, a collaborative planning platform, specifically was for the
44:42 midstream space was a bookable to upstream. So we built V1, myself, manager, another individual named Howe, so hidden Howe, shout out to them, and built V1 and Spotfire actually, and that
44:52 completely transformed the way we did things because for the first time we had visibility into operations, right? we were like, I think we got this right,? And that's kind of scary as a mystery
45:01 company. Um, but now you want to put something like that to kind of a planning tool where you can see any executive can see this is the amount of water coming in, um, and this is our capacity of
45:10 our system. Here's our forecast. That was, and we can go to clients and go and say, Hey, with certainty, even a buffer air, you know, a buffer range that we can take your water, right? So
45:20 then we took that and then we got some, they approved some executives and we started building a commercial software out of it So we had the opportunity to lead like a team of software engineers and
45:29 designers, and it was an awesome time. So we built something really, really cool. At that time, the idea was for us to spin us out, but at the end of the day, it kind of just stayed as a really
45:38 good internal planning tool. So I did my time at Waterbridge there for close to five years. And then, yeah, kind of mentioned in the beginning, you know, progressed that in my career. I've been
45:47 really grateful to have some great mentors and, you know, experiences, but, you know, we just had our daughter and it was just, it was time for something a little bit different And, you know,
45:54 I think early in your career, you really value being able to go up that chain. But at some point in time, you realize, all right, maybe there's nothing specific about when I want to go on, right?
46:01 Usually they say people go on like two peaks. The first peak you get up there, you're like, Oh, this is probably not it. You go through a value depression, definitely went through that. And
46:07 that started to climb a little bit back out of it, right? So I'm like, this is, I want to structure life a little bit differently. And that's when we kind of made the move to Dallas. And we said,
46:17 we want to go back closer to family and kind of focus on that. And that's how this kind of started. That's awesome.
46:23 That's, yeah. No, I think that's been a big thing for me ever since I started working at startups specifically is like, it's very like, I wanna do things that are fulfilling to me and building
46:33 shit is very fulfilling to me, generally speaking. And so once I got into that, it was
46:40 like, I'll never go back. Like it's so much, you know, just like the self-fulfillment, but then on top of that, the startup startups are very like, you know, painstaking and emotionally
46:53 emotional terrorists and all of this other things. But, you know, the benefit of being in a startup having your own company that you guys both realizes, you know, just the, the work-life
47:04 integration. I won't say balance because I don't believe that there's a true basis of that, right? It's like work-life integration. Yeah. Yeah, harmony or whatever. Right. Well, and, but
47:13 that's the beauty of being at a startup or as a self entrepreneur is like you're in control of all of that. Now, does that mean that, you know, you might miss dinner because you're on a call or
47:23 you're doing work? Yes, but that also means that you got to see the play at one o'clock in the afternoon that you wouldn't have been able to see So it's like, again, it's not, to me, it's not so
47:31 much a balance because it shifts all. It's constantly shifting both ways, but it's you're integrating them. So you have, you might be working more, but you actually have more time doing the
47:42 things that you're actually wanting to do. A hundred percent. That's exactly what I was telling my wife. I feel like I'm even though I'm working probably more and not, like not to put a glossy
47:49 picture, it is stressful as hell. Let's put it that way, right? Is a lot of stress going from your W two to something that is, is nothing, right? It's on your shoulders to kind of figure it out.
47:58 So the flexibility that it has is phenomenal, but you also deal with that. But to your point, it's like, even you may be working more at times. You feel like you're a lot more present. Um, so,
48:06 you know, I was just telling y'all, you know, so we're, we're expecting a baby boy in the middle of October. Um, so my, my office now is directly in my daughter's playroom. Um, so maybe boy
48:14 took the office. So I, as I'm working, I have her running around. It's just awesome that we're having that time before she goes to school, right? Cause once she's in school, she's in school.
48:22 Um, so yeah, that's, that's cool. Um, so one thing you brought up in there, I think how you guys kind of kicked off at Waterbridge and I think it's a fairly typical story, an oil and gas. But
48:32 I've always said, like, I love Excel and spot fire, just even from, like, prototyping standpoint, like, and just being on the build up, almost a live wire frame, uh, ultimately, um, and
48:42 help multiple people have actually sold companies or built companies on top of spot fire, you know, like, for sure. Um, well, we did back then is you realize something, you can't replace Excel
48:53 unless you rebuild Excel period. So might as well just go to Excel. if that's where they're used to. So we did it for V1 specifically. A lot of our planning was happening in Excel. Great. What
49:03 we did was we made that a module in Excel and we just put some VBA code behind the scenes so that whenever you're ready to export your latest like well schedule, you just highlight in your go, right?
49:11 And then with Spotfire, we just tie us into that database and you're good to go or the other specific Excel sheets, right? So prototyping Excel and tying it to Spotfire, definitely the way to go
49:21 for before you touch any software. It probably will get you what you want or at least guide you and maybe correct your assumptions of what you think you need. I think that's a big thing that not
49:30 enough people, I feel like prototype and or even just like quick, it doesn't even have to be an MVP and it can be like a beta or an alpha point zero, but like just going through that exercise, you
49:42 will find out things that you didn't think about when you first sat down thinking, you know, to go out and execute this thing. And so, but the beauty of that is you didn't go invest, all this
49:53 time and money into building out a specific architecture and going through all these different services and only to find out that, oh, well, shit, we don't have access to this database because
50:02 it's in the cloud and, you know, all that fun stuff. But I don't think enough people, like, put enough attention to that because it's like, if I can quickly prototype something and then
50:13 immediately show to somebody and they can immediately see the value in it, it's much easier for them to write a check. It's much easier for me to go and to know what I'm going to be doing and to
50:21 scope out and to estimate versus you and I sitting across from a table, just talking through an idea, trying to scope it out and then see you in two months when you deliver it to me, right? Like,
50:31 that's a, it's such a different experience. And so like, especially now with all these AI tools, like we, I talked to somebody the other day, no, it was the thing. It was one of the Umbridge
50:42 guys was on the panel and he was talking about like, you know, using whatever transcription software from their meeting with the client, feeds into. you know, gets processed through chat GPT or
50:54 one of them and then gets fed over here and creates a PRD. And then he pushes that PRD into a prototyping tool to rapidly develop a prototype just based off of that. And they have like a mockup at
51:07 minimum at best, a working, you know, general pseudo prototype within a day for the client without ever touching anything. And it's like, that's crazy, especially like, software development is
51:20 so iterative that like, if you try to boil the ocean, it's almost always going to fail. For sure. But I don't know where I was going at that point. I wasn't saying it to your point, it's, you
51:31 know, what I've also realized too, is that sometimes people can't picture what you have in your mind, unless you actually show them, even if you try drawing it out. So having these MVPs, like it
51:41 be on the spot fire side, not software, of course, you can buy code and things like that. Just to be able to show something a bit more tangible facilitates a much more valuable discussion 'Cause
51:51 some people that really need to see that and that's how they learn, they might see, yeah, I think that sounds good, right? But they'll give you a whole different set of feedback once you actually
51:59 show them it. So that's kind of what I've learned a little from that process. Yeah, even just from the dev side, right? Like having something, I've dealt with this when we had our remote devs
52:09 and stuff, and it's like, oh, well, here's the feature. It's a button that does this, right? And it's like, okay, well, they made a button that did this. It's in the wrong spot, it's the
52:17 wrong color, it's the wrong size, you know, like all these things And then you keep going, you as the user get frustrated. So like even if you're giving someone, you know, working with a
52:27 consultant or whoever to build an app, you should prototype it as well or wireframe it at minimum so that your expectations are as clear as they possibly can be on both sides of it, right? It makes
52:37 their job easier 'cause they're not guessing. They don't have to come back to you and say, well, did you mean this or did you want it here? Or, you know, there's so much behind that. But yeah,
52:47 I think we've got a few more minutes, but like one thing I wanna ask you to, you know. because I think we're both doing the solopreneur thing and I don't think there's ever been a better time to do
52:54 it now 'cause now with all these tools at our disposal that can make us more efficient, just curious like, what are some tools that are within your stack, the day-to-day that's like helping you do
53:04 things more efficiently as a one-man shop. I know you've got a handful of folks maybe helping you on the software side. Do we have, it's a team like two or three sub-contracts at the moment kind of
53:12 just depending on the workload. But for tools, I guess on the dev side, definitely cursor, I'm a cursor guy. I haven't tried Cloud Code yet, but it's in the bucket list Chatch EPT is kind of my
53:22 go-to specifically at the moment. Anytime I need to do some really deep research or analysis, I'll spin up both Chatch EPT agent as well as
53:33 Manus. Okay, fine. And Manus is phenomenal, you gotta check it out. Okay. And I'll spin up those two really quickly to go do that research and they'll go and get back to me whenever I need
53:39 something. If I need to develop a kind of like a pinging system for me, I'll use NAN to develop a drag and drop workflow. New contact came in, go and send me a message you're out there stuff on
53:52 LinkedIn and let me know. Yeah, it's kind of like Zapier, but it's
53:58 primarily framed at least around the AI tools, but it does a lot of data stuff too. It's great, you have like a workflow and you just need to kick it off. Like simple workflow, I set up like a
54:08 few of those and it just helps to streamline things that I always do a little time. We've used that for some prototype stuff very, and it's great. It's great, honestly, it's great. Call a web
54:17 hook to trigger something. Yeah, get your output Email, just forward it, or whatever it is, it's really flexible, and it's very powerful. A ton of integrations, too. The best thing is open
54:25 source, and that's why there's so many integrations built for, and I think that's why I've probably taken off recently. They even just released an agent feature, agent feature, inside of NAN to
54:33 build the workflows that you want for you. So it'll go ahead and stitch all the workflows for you. But anyways, I'm using granola to take notes and meetings. That's been the best one I've found so
54:43 far. We use granola too, it's awesome. The new Google Vision models to be able to create some AI videos for a different project, I think. Not banana, a VO3. Okay. I haven't tried banana yet.
54:55 I haven't seen some crazy stuff. I've been getting all kinds of stuff on Twitter about it. I haven't had time to mess with it yet, but. What else? And then I use Claude as well for coding things.
55:03 So that's kind of like the general stack. What do you use in cursor? Are you defaulting to anything specifically? Do you let it pick or? So if it's simple, I let it pick. But more Jordan at the
55:13 time, I'll go Claude. And then if I get really, really stuck and then I'll get Gemini to jump on board and that's kind of how I progress through things For my work lately, Gemini's actually
55:21 knocked it out of the park when I need to work Claude's kind of stumbled on itself. Like Gemini's been really good. They both keep like supplanting each other. I feel like on the code side, GPT is
55:31 kind of, I feel like falling off on the code side. And I don't know if that's for a specific reason or if just because like people started using Claude or hearing that other people were using Claude
55:41 more for code and then more people just started using it. But it's fascinating to see what I always ask when come across another cursor user just to see because on the more on the more technical side,
55:56 I lean heavy postgres where I can. So PGAI extension from Timescale DB is what I use. So it's great for vector stores and queries and things like that. And they're preferred cloud. If
56:10 I can choose AWS, but I'm forced to use Azure because oil and gas is a Microsoft shop. So that that's my take. You don't you don't love all the individual permissions you have to set up in Azure or
56:20 that you can't rename shit. No, I nor do I like nor do I like how I started. I've had so much headache with deploying dock containers in Azure and things failing and you have absolutely no logs,
56:31 even though you enabled it and all this stuff. It's just a headache, man. You're using AKS for that? Like, no, just deploying dock containers. So like we'll like build like a Python back end
56:39 and we'll go ahead and deploy it as a container. So that way it's kind of just an endpoint service in the cloud. Well, you haven't tried Azure Kubernetes service. You speak Azure container service.
56:45 No, it makes makes it pretty easy. Is it? Yeah, I There we go. But yeah, that's kind of the stack. Or I told John about that when he was working at Hivesell. I was like, you realize that the
56:57 Azure has a, basically does that. The exact same thing, like, oh shit. There we go, done, I'm doing that after this. No, Azure is like, but yeah, it's not perfect. It's just got so many
57:08 nuances that I feel like the other ones don't. And I say that as someone who has done very little and either of the other clouds, actually the most cloud experience I probably have outside of Azure
57:18 now is Google. But,
57:22 you know, RAWS and Google guys working in Azure are just like, they come across the right, like you can't rename stuff. In Google? No, in Azure. Like you spin up a, you know, a blob storage
57:33 or a container, and you're like, oh, I meant to, I should probably name it this. And it's like, nope, sorry. You gotta recreate it all over again. We create the whole damn thing. Or I want
57:41 to copy something from one to another. Oh, well, there's a button here that says copy, but to actually copy it, it's a whole process. It's just all this stuff. My favorite is what Azure SQL is
57:52 basically a SQL server in the cloud, but you can't get a BAK file like out of it. Oh, that's good. I see one of those is the backpack file, which is really obscure. I haven't had that, but I
58:03 won't use that now. Yeah, it's pretty cool service. It's serverless, but it's something in the world, but you're so used to it in the SQL Server world that I get a BAK file and I load it or
58:14 unload it and then just like, you can't get a BAK file. Like I can upload it. I can upload it That's how I got it in there, but good to know that I won't be getting that data back out. It's there,
58:25 at least you know it's there. It just needs to be there for me to query. So that's all it needs to be there for, but were there any other tools? I'm sure there is, but I can't think of it at the
58:35 moment. Yeah. Oh, what about Arc? I saw you were using Arc. Oh yeah, Arc is my daily driver. Okay, so like from that tooling perspective, I'm a big Arc fan. I'll have you guys used it for?
58:45 So Arc Devs all use it. I'm about to, oh, I think I'm gonna hear I think ArcGIS. No, Arc is like kind of like a different take on a browser. So it's called a company called the browser company,
58:54 which actually just announced you're getting acquisition today by who they didn't say, but they just turned like a merger today. It's going to be Google and they're going to ruin it. So I knew
59:03 they're going to get acquired at some point because there's some. What was the company like a week or two ago that said they were going to buy Chrome or made a perplexity perplexity. They said they
59:11 want to buy like Chrome for 34 billion, but the court order came out saying that it's not a monopoly that Google can actually keep Chrome That's crazy, I love how all the tech companies that pose as
59:24 non-monopolistic platforms because 1 of their revenue comes from an Android software or what a non, it's like, okay, sure. Sure Chrome doesn't have a complete monopoly on the internet and all of
59:37 the data that you're looking at on said internet. Yeah, but ARK is just like a different browser. Okay. Just, it's reimagined in a way. Yeah, it's kind of AI enabled, Also just like The
59:47 structure of how you organize your stuff is different. I, they, what my devs love it. Interesting. I'll check it out. You got to check it out man. That was a whole one of those podcast was,
59:58 you know, sharing things people never heard of and I need to check that out. And superhuman is the last one I'll say. Are you superhuman for all my email? Okay. You have to use it. I think Brian
1:00:05 Becker made it. Somebody has brought that up before. I can't remember. It is life changing. For managing a bunch of Gmail accounts and things like that, it is awesome. That's awesome. So
1:00:14 essentially like it's just a way to be able to get to inbox zero much quicker And they have a lot of really interesting like AI and automations inside of it. So automatically figure out if
1:00:22 something's important for you and it'll ping you or other marketing stuff. And I'll send it away. And the beauty about it is you can have these custom split inboxes. So you can see if I get an
1:00:30 email message that looks like this, put it into like inbound or outbound or marketing. It's just great. It's awesome. Check it out. That's cool. That's good to know. No. I think we should
1:00:40 probably just add the what's in your stack question as a recurring one on here Because 100, that would be a good one. Well, shit, there's an hour. I just looked over at the time. That was crazy.
1:00:51 Wow, I don't wanna look fast. It always does. It always does, man. That's crazy. When you get good people on it. Yes, it definitely helped having people. Appreciate that. Good folks like
1:00:59 yourself. So I guess we'll jump into the speed round. Do a couple. No, here we go. Where did you say you, were you in Dubai you said for your honor? No, Qatar. Qatar, okay. What is
1:01:11 something that someone that visits Qatar should absolutely go do that's never been there before? Perhaps there's so many things. Also, is it Qatar or Qatar or? I actually don't even know. So I
1:01:22 think in airbase it's Qatar. We'll say Qatar. Right. We'll say it in Qatar. So it's a different pronunciation. It's almost like Qatar, but like - You mean it's not Qatar? It's not Qatar, not
1:01:31 Qatar, sir. But it's funny 'cause like for Qatar airways they have like, you know, they ask, it's like, welcome to Qatar airways. You know, it's like a British accent, right? So you say it
1:01:38 whatever way you want. It doesn't matter. It doesn't matter. One thing that I do there is you have to have something called Karak. Karak is like a very sweet milk tea. It is top notch and you
1:01:48 need to go to the sand dunes out in the winter. Go during winter time 'cause that's when people set up tents and they'll just invite you in. They do a bunch of cookouts and stuff. It's awesome.
1:01:56 And then you're on ATVs and then I, it's on the dunes. You're not gonna do that. That sounds cool. I think I've seen them do that on the real housewives one time. And then of course that's my
1:02:04 cultural plug-ins. Bravo, you should. I gotta take you. I gotta take you out. No, that is on my short list. I want to do, once we make some money on Clyde and we sell this off, we will be
1:02:17 going to the Met for a while. When Chuck gets his jet back by and just - Yeah, just remember me, you know? Yeah.
1:02:25 All right, so I mean, you brought up tea, but I know you're pretty big coffee, you guys about like, so, and you're back in Houston. Like, so, you know, what's your favorite coffee place in
1:02:32 Houston in that second country? Siphon. Siphon. Siphon downtown near downtown area is my top. I love that spot. Great coffee. So I just bought like a few bags actually. I need to go back and
1:02:42 get some more because they don't sell them online But if you haven't tried siphon, try siphon. And the breakfast burritos are top notch too. All right, sounds like a winner. Yeah. What is your
1:02:50 daily driver for language models? If you've just got like, not a specific coding or like a very specific application, but like, I just need to ask it a question or like, you know. I use chat TBT
1:03:03 still, chat TBT is my go to. So do I, which is so weird, because I use Cloud for pretty much every time. Yeah, exactly. No, I mean, and Gemma is really good to be able to just, I don't like
1:03:12 my brain, I'd almost created like, almost like the Xerox or the Band-Aid branding, where it's like, I need to get these chat GPTs. I just use chat GPT for so many things, like work, personal,
1:03:20 I guess it's like a habit now. I trust it. I just use it actually like in my Uber driver, where we're looking for like a new car for the family, and it was like a Toyota Grand Highlander, and
1:03:28 it's one of the ones that we're thinking, and it's like, well, it's a great car, but he didn't speak English, so I just started talking to chat GPT, you know what, just translating Spanish, I
1:03:34 need to learn Spanish, it's on my bucket list, and we're just going back and forth for a few minutes, and it's great. That's wild, yeah. Spanish has gotta be easier than Arabic, So I think that
1:03:41 should be a good idea. I just need to do it. I would love to learn Spanish. Yeah, it's like I learned Arabic when he was at a very young age and - Yeah, my parents taught us Arabic when we were
1:03:52 really young and made sure that's the only language we speak for a while and then, so no, of course we go both. Yeah, no, that's exactly what I'm doing with Mike. Well, not quite as strict
1:04:02 because my wife, nor I speak Spanish, but my kids went to Spanish school when they were little with that concept, at least in mind, but it's a, no, it's a great one down here, right? It's just,
1:04:16 but even just traveling, right? Like I told my kids, they don't like, they get shy about speaking Spanish in front of people. I don't know why, but I was like, if you guys wanna come on these
1:04:26 vacations with us to Mexico, guess what your prime ticket is, is translating. So you gotta do that, but
1:04:35 what's one more? Let's see, or I just asked you the LLM1. I don't want to steal one from you. Yeah. been written lately or a favorite book, you know, whether it's data related or not. And you
1:04:46 can put me on the spot, man. I am trying to get back into reading. I am bad. What about podcasts other than this one, obviously. So how you built this is a favorite one. I love how I built this.
1:04:57 And then also the one by Lex, not Lex Friedman. It's something with Fireside. It sounds just like a tech podcast, very, very popular PM world. I'm forgetting, I'm blanking on the name. On the
1:05:10 books, I'm trying to read some more parenting books There's one, I forgot the title of it right now that I'm reading, but it's more so like, you know, staying or spending a lot of time with
1:05:17 specifically about your kids, especially in the early years. So it's a really well-written book that I'm probably going through. Nice. Yes. Parenting us are very good. Especially for my wife,
1:05:29 like last weekend, something with the kids. That's like the hardest thing I've ever done in my life.
1:05:36 It is nothing prepares me. You love it, but it's like, it is hard Your spot on meth, there is no way to prep. That's what I tell people, I'm like, I don't know if I'm ready. I was like, you
1:05:44 will never be ready in all honesty. Like maybe get, you know, financial wise and stuff. But like beyond that, there's nothing that can truly prepare you for that. Yeah. Even a dog, you can
1:05:53 leave it home alone for a while and it'll be fine. Like generally speaking, yeah. At least it's socially acceptable. Yeah. I'll get out of there a while in a cage. Yeah. Not slightly frowned
1:06:04 upon with children. Awesome. I mean, thanks for, uh, you have finally catching up with those. I mean, this is down here I appreciate the opportunity. I'm not even here to. I mean, I know,
1:06:13 you know, usually when we can get people in person, it's so much better. No, I much rather do in person. So I was going to come down even just for this. Well, how do people get in touch with
1:06:20 you if they want? Yeah. The need your services, um, contact me on LinkedIn, uh, who saw him or Juan. You can search you up there. Uh, you won't find many of us, I think, um, and then, or
1:06:29 you go to just aiagency and I'll take you to our dark to the site. Perfect man. Beautiful. Yeah, appreciate it. All right. Appreciate y'all.