Welcome to Energy Bytes with John Kalfayan and Bobby Neelon. Your essential guide to the intersection of data and energy. This podcast dives deep into the world of energy, shedding light on how data, AI, and technology are revolutionizing this sector. Each episode equips listeners with insights into the most efficient tools and resources, paving the way for a data-driven future in energy. From technical nuances to broader industry trends, Energy Bytes offers an unparalleled perspective on the evolution of the energy industry. Join us as we decode the algorithms of energy, one byte at a time.
Welcome to another episode. We got JD Long here from Renaissance Re. Um, you know, does all things data for those guys. And, uh, no, I mean, and I I was telling before jumped on, but it's had a, you know, a lot of his tweets or, you know, stuff on LinkedIn and everything is, you know, had a pretty big impact on how I kind of view data or how I try to have us do data, um, at at Grayson Mill. So, uh, really excited to have you on and and pick your brain and just kind of hear some good stories. I know you're you're quite the character. So, yeah. I'm excited to be here. Thank god I I started and like I don't know why I'm here. I don't do that much energy. [laughter] So, okay. But I do a lot of data. So, happy to Yeah. Know I mean because I mean we're always trying to you know figure out what what are people using to solve problems? What are different tools? And I think sometimes you get that cross functional knowledge. I mean there's obviously things that you can you've done or can glean you know you we can glean insights from. So, uh, you know, we try to stick to energy in general, but again, I think you're just a good, you know, generalist and, you know, have experience across the spectrum. So, you bet. This would be a cross-pollination episode. I'm down with [clears throat] it. Exactly. I mean, we had someone what, two weeks ago, John, that works for like a tractor tech company. Yeah. He works in like they're doing Oh, fantastic. Tech. So, um, again, I think it all comes across. I mean, like again, you need energy for everything. It all, you know, it all comes comes back around and everyone has data that they need to do something with or they think. We do something with it. Plenty of it too much depending depending on how you look at it. But uh um but I mean with that I mean we you know a lot of new technology out there and obviously AI is kind of the new hotness. JD I saw you reposted this maybe yesterday but um about the Baltimore High School athletic director who uh used AI to basically what clone his principal's you know voice and and basically slander him by making it sound like he said racist things. And he actually got was what fired for a few months. Yeah. They put him on leave for a few months and did an investigation. The FBI and CIA got involved and then they figured out that the audio was fake and they [laughter] now turned around and hired that the coach. [snorts] I mean it's it's pretty insane that the I mean on the one hand it it's a it's awful. It's abysmal and it's immoral. On the other hand like the technology's gotten good enough that the uh that the coaches can figure out how to do it. That's kind of uh mass market if there ever was mass market, right? So, a lot of the% tooling have have I think gotten surprisingly uh user accessible. Yeah, you can go on 11 Labs right now and clone my voice very very quickly. Although, have you guys done that yet? Have you all tried it? I have not personally, but how how'd you find it, John? Uh, this was this was probably six months ago, so I haven't tested it since, but it was the thing I noticed about the the audio is it does not do a good job of like cadence and tone as well. Um, I have noticed recently that they've started getting better about putting in like ums and like pauses and trying to get more natural kind of cadence or whatever. But, um, that being said, like the audio when I was talking generally absolut like if my parents answered a phone and heard that, they probably would think it was me, right? Like it's super scary to think about, especially Yeah. Have you guys have you established like a safe word with your family so that someone doesn't hijack you? Yeah. And then do the whole I'm in jail. You got to send, you know, money to whatever ad. Target target gift cards. Yeah. Target gift cards. You know, it's no longer going to be the the Nigerian prince sending you a text. It's, you know, they're going to have your voice and Yeah. Yeah. It's, you know, it's pretty scary. Who was it? I was talking to so this guy Talal was one of our he was our first guest actually and I had lunch with him um this week and he was telling me about another thing where I guess someone used uh Zoom and was able to replicate like and his his boss this was like a bank and it was to like wire like maybe seven figures or something like that and like so he got the initial message from his boss and and they said hey uh you know I need you to wire this money and he's like I don't know if this is true so he sent like well let's jump on a Zoom call and he so he jumped on a Zoom call This person was able to replicate a bunch of his co-workers being in the room also. Whoa. Like, so it was all he felt like he was there with his boss and like and it the guy wired the money like Yeah. I um I I was talking to an executive with an insurance company who they had a a situation that was close to this and they got on the phone, got on the video call and we're going to do like, okay, I'm going to text you. So then you got the video call, you got live, and then you've got text connection between each person and they figured out they were being scammed and it all just went away. But it's like we need to rethink our trust network of how I mean, how do you literally do like, you know, transfer of large amounts of money as a business anymore. Yeah. Yeah. You essentially need a 2FA for that with within your company that's not That's exactly right. The technical 2FA, right? Well, and right I think without saying too much because it's his idea Tal is thinking about something along those lines too. It's you know um but I think even now he's like if if you get a call from me because he's got a small company so like if one of his co-workers you know basically when a new employee comes on someone's already sending him hey this is to I'll go to Target and get you a bunch of Apple gift cards but he's like if you get any even call that sounds like it's me and you don't think it's legit say hey I'm going to hang up and call you back because I think people can spoof the number one way but they can't you can't spoof it the other way right? Yeah. And uh so but no, I mean the AI stuff is going to be crazy, but it's it's one of those like you know some people like we got to shut this down. It's going to be too dangerous. Well, guess what? I mean the nefarious people are going to have it. Yeah. And like I mean China, Russia, it doesn't matter. It's going to be weaponized. So I mean I think the only thing that's going to beat AI is probably AI. Like [laughter] Yeah. No, I mean it's it's just like you know they they have forensic forensics on media already, right? be able to tell like geoloc if it was photoshopped or any of that stuff like there that will happen at some point or there will be mandates where you have to put you know something in the code or like even with GPT I saw an article a while ago uh where they were like there was a specific word um and they were talking about you know kids plagiarizing using GPT or whatever and they were like there was one specific word that if GPT exceeded so many characters it almost always included and so they could literally
filter, right? Or I also think the teachers are getting good at it as well where they put in, you know, they hide text within their uh summary or whatever the assignment so that when the the student copy and pastes it in there, you know, it shows up white on the document, but they copy and paste it in there and hit enter, then it shows up. It's like, make sure to include this specific term somewhere in this document every time. So it always ends up in the output and then that's a that's the tell. Yeah. And that's the tell. Yeah. Well, we we we got a bunch of decisions we got to make, right? Collectively as a culture and individually, you know, my my sis uh my sorry, my um daughter is in uh junior year of high school now and I guess I can't remember if this was end of last year or beginning of this year. She wanted to talk about ethical AI and I was like what the hell? like I didn't see that coming. She's like, I got an assignment and I didn't understand what they meant when they asked like I want you to write an essay that contrasts this this to this and I didn't understand what that type of assignment meant. Can I put that in AI and ask it to give me an example of that type? I'm I'm not going to copy and paste it. I want to use it to show me the type of writing structure. And I'm like, you know, she said, "Our school has a strict no AI policy. What does that mean? I'm not going to copy and paste it. I just want to I want it to show me a structure, an example of how you'd answer this type of question. I'm like, this is a really good point because it's not clear. Yeah. Well, and that's one way in my opinion and I use it with code all the time where that is such a great tool for let the GPT the language model whichever flavor you like explain and teach you things in a simpler way or like I would be so much better at engineering if I had this when I was in college and I could have it walk me through thermodynamics instead of just getting the you know here's the 200 pages you have to read this week and 10 assignments and you just you're drinking from a water hose. can't comprehend half of the material that you actually get. And so having a tool like this where it's like you can do it on your own time. It can give you examples. It can break things down. Like it's really fascinating. But to your point, it's like how do you use it for good but not in an in an ethical way. And then there's this ethical and then ethical in the way that your boss or your school or whatever approves of. Right. Because I could argue about ethics in the abstract, but my daughter's school may not agree with me. Right. Yeah. Right. And that's tricky. So, you know, one of the skills from preAI right now, I think um Stack Overflow is really u catching a bunch of heat because they've licensed with OpenAI to to include this and some folks have tried to take their answers down from Stack Overflow because they don't want it used for training these models and and Stack Overflow is not letting them basically putting it back up and blocking the users or locking their accounts. And so folks are folks are feeling intense about this. The the irony I I see or I guess the the um coincidence I see is actually what makes me good with Chad GPT with coding is the same skill of why I have a really high ranking on Stack Overflow and it's this idea of a reproducible small example because the single biggest mistake when folks come to these language models they give them too much crap. They'll try to put like whole block you know hundreds of lines of code in. I'm like, "No, no, [clears throat] no." What you want to do is try to figure out how to solve the little piece and you want to explain that piece and say, "I'm, you know, I'm trying to do this and then build that up." So, it's not, don't think of it as building your house for you. Think of it as building the Lego blocks and then you got to stack them together or you got to link them together. It's got to be iterative. It works so much better when you iterate through it. Start with the smallest, like if you're specifically for coding, that's mostly what I use it for, but like start with the smallest thing. Okay, I want to call this API. All right, then get it to write the script to call the API. Okay, now I want to put that API into a data frame. Right. Right. Then you get it to do that and you just build on it instead of like and you build on it. That's exactly right. Hey, I want call this, put it in a data frame, do this, like instead of doing it all straight down like it they're still not that great at at the reasoning and thought process. And and I'm sure JD, are you using uh G getup copilot at all or no? I I have used it just briefly and I like the conversational thing that John referenced so much that I actually after using it for a little while, I kind of went back to Chad GPT because it just there was an impedance mismatch between how I wanted to use the tool and how Copilot was built. Um and I've watched some of our developers kind of feel the opposite. they just want to put comments in and get like spicy autocomplete and I'm like I kind of want to do something a little more conversational, more iterative. Uh because I feel like the singleshot sort of approach it's not as good at and maybe that's just I'm not a very good explainer, right? So I need to I need to iterate with it. Uh so I like the more iterative conversational thing. Now, I'm working on a book about using these models for data science, and it's it's really about this, you know, fancy prompt engineering. And one of the things my co-author and I have struggled at is when we're doing that thing I we you and I just discussed, what do we call that? That's a great, you know, it's not conversation. Or is it or is that too much anthropomorphizing? Like we've sort of leaned towards iterative inquiry, but that almost feels too cute. Yeah. Or something like I can't figure out what the hell do we call it? What are we doing? No, that's that's very true. Uh you should definitely check out V uh cursor.sh. Cursor.sh. That's what I've been using uh here recently. And I think it's built on VS Code in some form or fashion, but it's got they have their own they're funded by or they have funding from OpenAI, but they have their own model built in, but you can also link Claude and GPT into it, but it's your VS code window and then the BR the chat right next to it and you can reference code d and it pulls it directly into the chat. Nice. It will generate code and then it'll ask you do you want to implement it and it will go through your actual code line by line change it and then you get to tell it yes or no like it's it's exactly what I was doing with GPT but it's faster because it's all right there more tightly coupled right it links them right there together yeah that I need to look at that sounds like a good idea so uh it's cool total geek question here have you ever used a foot pedal with your computer I have not not for anything other than uh gaming than video Yeah. Well, I I kind of got this idea off of I think the hardcore text editor folks, Emacs folks or or or Vim folks will use the uh you know the foot pedal for certain keystrokes so you don't have to stretch your fingers, right? Things that you use a lot. And I went online and found a used transcriptionist foot pedal and uh it didn't have drivers for Macs. I had to write enough Python to like sniff the USB ports. It went down a yak, dirty yak shaving hole. But in the end, I got it where I can press and hold and start voice to text for dictation and then when I let up it stops. So I can just be typing and then just press my foot and switch to doing it by voice and then let off. And one of the things I've been experimenting with is doing like brain dumps. like you know the conversation you have with yourself in the shower about a topic kind of dumping that and then giving that to like Chad GPT with some I have a very long prompt of what to do with this but basically don't change the intent but try to re reword so that it and it so it sounds like me and has my style but basically see if I can go from brain dump babbling to something that actually reads thought through but you still got to go back and read it because you still got to go back and and then I've edited heavily because there is that whole working out your thought as you capture it and and I think folks sometimes forget that that's like you know the whole thoughts detangle themselves on the tip of a pen and the combination of saying it and then writing it you're working it out and what I want to do is just work it out faster but I don't think I'm going to be able to just like you know skip a bunch of steps because that is the process right [laughter] Yeah. Yeah. Now, Colin, our my our CEO, he he'll they've got a big kind of walking trail and stuff, but in his neighborhood, and he will go on walks and just talk to GPT and have it just via, you know, AirPods and have it record everything and then summarize it, bullet it, outline it, whatever he needs to do. It's uh yeah, there's so many there's so that's the coolest part I think of this tech is like people overhype what it's capable of doing, but they sleep they sleep on all of these like I'm sure we'll talk about data science applications just like everyday applications or like just what we were talking about with your daughter like having it teach you how to do things like it's like YouTube on steroids, right? Like [snorts] Well, the the thing I have seen as a learner is often I don't know enough to ask a good question. Yes. And so there's a bootstrapping problem to learning because if you kind of know the idea, you at least know what to Google for, what you're looking for. You have a mental, you have a framework in your head. When I'm totally starting something new, I often like don't have the vocabulary. I don't have like a framework of how to talk about it. I don't feel confident enough to bother an expert with like, tell me how to think about this. They're what the hell are you even asking me? And I'm like, well, I don't know enough to ask a better question, but JPT will do it. Like, if I'm thinking about this topic, what are the big ideas I should be thinking about? Yep. And just bootstrap me. And that's like shockingly useful. And I think that's the sort of stuff, John, like you're saying, that people are sleeping on. Yeah. And maybe it's because we, you know, us and people like us and people listening to this podcast are learning a lot and learning faster than probably your average adult in the US. uh would be my guess because we're very much doing stuff that changes rapidly and you got to ingest lots of new information and learn new things. I mean that's the the thing that I where I see coders saying like oh this this isn't useful because if you know your libraries and everything you're faster and I I just chuckle because I'm like most developers are not using libraries and syntax that they're absolutely completely fluent in all aspects. I am sure there are a few folks who have been writing C++ for 40 years and they think in it. Y the rest of us don't. No. Or even just like the nuances of them or how they've ch like they change all the time, right? Like and it's just like yes, you know the base of it, but that doesn't mean you're not going to run into an issue where you know whatever the encoding is messed up or whatever it may be, right? like and or you're using AWS and there's like 15 new tools [laughter] that you've never heard of that will solve the problem you have, right? So, so you need some sort of back and forth. This idea that oh no, if you had this static knowledge, you can just fluently do this, I think, is is really malformed. Yeah. I mean, I I mean, I used GPT to walk a directory the other day and delete all the PDFs in it because they were taking up too much space. But, I mean, have I done that in a year and a half? No. But, I mean, like, I asked GPT and it was done in 10 seconds. Yeah, exactly. I could dillydally around and write that on my own and only Google 15 20 times. Uh but but why because it's easily explained and you know it's training data has a bunch of examples of something like that. You know similarly like the recursive delete on S3 buckets where you got to recursively delete everything so you can because you can't empty something if it if it has stuff and you can't delete a directory structure if it has something in. Anyway, it's just like that's really annoying to do and you know because I do it so rarely, you know, but to to write a little code. Yeah. And then then I those contribute to my utilities library of my personal junk that I keep stacked up. But at this point, a bunch of that gets written by like Chad GPT, you know, and there's no like corporate [snorts] private information in traverse this structure in S3 and delete files, right? Or ETL processes or Yeah. any of the fun stuff. Yeah, exactly. That's so with that I mean you were uh can shift gears a little bit but not 100% but um you know you were talking a little bit earlier about Vim and Emacs which are obviously old school kind of code editors but um I know you you've had a lot of thoughts on like say guey you know versus you know code and you know um I think if correct me if I'm wrong but you're generally a fan of you know stuff that can be stored as plain text so it's a little more reproducible and going that way but um just what's been your experience I mean like obviously there's more and more low code tools tools and even you know Tableau whatever else you know PowerBI out there but then you know why should people use that or when should they use that versus when should they you know write code um and you know tech debt kind of just a little bit of you know of all that oh yeah you're going to I may have to take my uh take my fleece off things are getting about to get spicy let's do it because [laughter] yeah I' I've done a ton of no code stuff and I have very like mixed opinions on it all right I yeah I'm interested to hear hear your experience I'll I'll share my bias and then you can push back on on what you see differently. So I see no code as a or low code as an abstraction and and I think Joel Spolski said, you know, any any um reasonably sophisticated abstraction leaks and the way no code leaks in my experience has always been you can do somewhere between 75 and 95% of what you want to do in the low code. That's great. And then you want to do something that you can't do with. So you end up with like this code block that ends up having a whole bunch of stuff in it and at first it's like oh well yeah it's just five or 10%. And then it becomes 25%. And then at some point you're like, I'm doing less than half of it in these noode tools and I end up pushing all my logic into this code block which isn't inversion control is locked up inside this tool. And it begins to feel like um not an enabling technology but a capturing technology. And if they didn't leak or what I wanted to do was trivial, it's kind of like the using AI. It can bootstrap you up to something really fast. The problem I have is after the thing becomes non-trivial, I feel like they're very problematic. Now, if the the way I would change my tune on this, and there was actually a tool for R that does this. Uh it was it was built out in Japan. I can't remember the name of it. I'd looked it up, but I can't uh remember it off the top of my head. But the idea is you use a guey type lowode tool and it writes the R alongside. And was it exploratory? Yeah. Um, there's exploratory.io which I've used before which is like I can't remember. It may have been that brings about I think it's a well then it's sorry Bobby it's a guey for data analysis but it's all built on top of R. And so like you can you can connect to databases you can pull in files CSV etc. And then it pulls you know your main dashboard your kind of data management dashboard just like Power Query or whatever. It shows you each column and you know the formats and all that stuff. But any if if you do any kind of uh ETL on it, you know, you can do it through menus, but then it will pull up the window with the the R statement, the R code in it. So if you want to modify it at all, then you can. It's actually really powerful for for that people like me who read code. I'm not a full-time developer, so I don't write a ton of code, but I understand the data side of it. And so it's like I know how powerful and good R is at data. I want to use R, but I don't want to sit and burn a bunch of time coding just to visualize a data set. Um it's really really good. Well, I always talk when I when I teach like data analysis like I always start with like none of us want to code. We want to solve some freaking problem. Right. Right. And that's part of the difference between I think data analysis folks and uh maybe folks who become software developers because some of the software developers I've worked with in the past. I mean they just want to code. Yeah. Um they really kind of like to not think about the business problem. They just want to code. Like they want to think about the coding problem. And so if you're in the data analysis world the mindset's very different. And so I I when I teach this or talk about I'm often like first of all we don't want to do this but it's how we're going to be effective. We want to get something done. All right. Well if our goal is to get something done let's drop purity test. I want to cheat. I want to do I want to steal. I want to borrow. I'm stop short of lying but I'll lie to myself and think I'm smart when I'm really just you know using Stack Overflow. I'm using every tool at my dis. I'm using AI. I want to use all these tools so I can solve problems faster. And you know, one of the tools that I encourage folks when they're just getting started with ggplot, which is plotting library in R, is this tool called ESQUIS, which it's a French word French word that means like to scribble or sketch, I think. Uh I don't speak French, so I'm sure I'm mispronouncing it. [clears throat] Um neither do we. It's okay. Yeah. Yeah, that was that was the one I was going to bring up, too. Um you know, it's pretty cool. Yeah, because it actually writes the ggplot code. Like you graphically design a graph and then it gives you the code and then you copy and paste that into your code or it's got a button that says drop it in there and then you can change it, right? You know, because it's there in code. Now that it becomes an easy on-ramp for coding and I like that that's different than being the trap, which is the first scenario I described where it's like you're stuck in this low code tool writing more and more code in your one text box for free form code. Now, how is that different than your experience, Joe? Yeah. No, I mean, I I think you're honestly spot on. I don't think we really disagree much um around that stuff. To me, the low code, no code stuff is really good um for prototyping, like if you're in product or tech, if you're trying to just get something out there to see if people like it, to see how it works, to see how it functions, whatever, to test, whatever it may be, it really is a good tool. But you are absolutely hit the nail on head. We actually just yesterday rolled our uh our platform over from one platform was on Ruby, one was a third party white label and the other was on a noode platform bubble. We rolled all of that over to Ruby and formally launched it yesterday. And it's for that very reason, right? Like, hey, we wanted to start doing more sophisticated stuff. It's like, oh, well, it's got a uh, you know, an OOTH integration in bubble to to do single sign on for your users, but then that goes into like there's no actual integration with octa or the actual SSO providers. And it's like, oh, well, then how the hell do we do that? or like oh I need my for for example another fun problem uh the API call to all the data tables works great but if there's a blank field and that's at the top of the pageionated return it doesn't return the field even though there's data in rows below it um just all kinds of that feels like a bug fun [laughter] garbage like that that you have to you have to deal with um or like the problem with them is they don't they fall into the trap of like we want to try and ultimately lock people into this. And that's not what people want. They want to use it to develop quickly, spin something up, test, prototype, PC, whatever, and then go do it at scale in some capacity. And you can't do that. For example, on this one, I couldn't. There was all this custom code that you needed to do in order to back it up to SQL. There was no way to back up your tables out of the platform, which is [ __ ] It's my data. Why can't I move it? And there's an API, right? But it's like the there they've got all these plugins for everything to ingest APIs or ingest data from a SQL table, but I can't push it back. Like why, [snorts] you know? Yeah. So, it's as soon as I smell a moat being artificially constructed, I'm either going to blow it up or I'm gonna go around it. Yeah. because I just the the vendors, you know, they want a moat, something to keep, you know, me from leaving or other tools from coming in and they better add tremendous value. Yeah. Cuz otherwise I'm going to see the moat coming and I'm like, I either want to go through it or under it or around it or ditch this tool and so our interests become disaligned because they want to exploit me and I want to not be exploited. So yeah. No, I think the the best versions have kind of a dual function or the best platforms have that dual functionality where it's like if you want to do it via guey, you I'm a firm believer of that with any kind of software, you should have at least two two ways to do everything, right? Like two menu but like a dropown menu and then an icon or a right click and an icon or code and an icon like whatever it is, right? And so like if I want to pull in your plugin and use it, great. But I also want to be able to code that if right I need to or right, you know, stuff like that. Bobby and I used uh QuickBase for a long time at the company me and him worked together and they did a very good job of building it. It kind of like a Salesforce in a way like it's a ERP system but they have lots of integrations. It has it's fully codable in Java under the hood. Like it made it dynamic enough that we could do really interesting custom stuff without having this huge rip and rebuild kind of situation. I also agree with you too wholeheartedly that the best things for these for like low code stuff is just transforming an existing code and then spitting it out, right? Like right ggplot that's a great example. I wish I knew about that library when I was using R back when Bobby started teaching me about it because I loved like I could transform the data and everything in R but then I did not want to spend the [laughter] [ __ ] time building the stupid code. Yeah. Yeah. Yeah. And and you know, I find when I use ggplot a lot, I kind of think in it. Yeah. But even if it's, you know, and I've written I'm co-author of the art cookbook, second edition. I mean, I've written books on this stuff. If I hadn't used it for a while, I I I can't remember, right? You know, I'm just too old for that. And so I have to, you know, use a tool or use, you know, now maybe Chad GPT uh and just Oh, yeah. Yeah, that's right. I remember now. And then I'm off to the races. And yeah, the other thing I want to mention, we kind of blew past it, but the idea of the low coding tools not being able to uh put your code in version control uh getter, whatever you want to use, uh that's a big one uh for me against those tools. you know, even VS uh sorry, um PowerBI, you know, has this problem of yeah, your code's in there, but it's inside of this thing that's really just a zip file and you can open it up and there's some weird hacks of getting in there, but then it's XML when you get it out. So, it's like if you change something, you know, it's it's in an XML block. It's as ugly as, you know, Jupyter notebooks are in version control, which is pretty ugly. And it's like the these are just that's an impedance mismatch with how I think I want to work or how I want a team of people trying trying to work together. Yeah. Yeah. And I think some of it's like there's a speed thing too. And it's like I think so sometimes you can build things faster in a BI tool. And sometimes that's necessary for people to get an answer out quick or to do something, but then there's that whole maintenance side of it too. like, you know, well, you know, actually, if I built a shiny dashboard or a template or something like that, like it may take some time that first time I build it, but then I can build off of it and um and then I can reuse that and do things more quickly and more sustainably going forward. What this is really one of the I think the hard problems in kind of the stuff people like us do is asking the question, what shape is my problem? Yeah, that's shockingly hard. And it sounds kind of dumb when I say it out loud. Like, what do you mean? That should be easy. But there's so many stakeholders involved though, like normally, right? Like my CEO wants everything done yesterday and he wants it done as cheap as possible and he wants the answer now and it's like, well, if if you want me to do this at scale 100 times, it's going to take a little bit for me to set it up so we can do all that, right? But if you just need a quick and dirty answer, I'll export it and pull it into PowerBI and pull it up for you, right? like I'll just pivot it in Excel or something, right? Like Yeah. And it's this this idea of a oneoff versus building a machine that will answer this question, you know, in into the future. That that's what a bunch of us face is that design pattern. The other thing I was thinking about like the shape of the problem is I have seen people approach any problem where there's numbers being presented as that's a PowerBI problem or graphs. And I tell you what, you can write I thought I'd seen some monstrosities written in Excel where people are basically building applications, but Excel is their only hammer. So they drive every nail with it. It's like, oh sweet Jesus, this Yeah. Just VBA out the wazoo and you know all kinds of modal forms and everything. And you can do the same thing with PowerBI. Yep. And it doesn't just because you can doesn't mean you should. and they end up with huge number of you know DAX transformations when you look at and you're like okay if you just were take a step back I could write a query for this and then we could write a query for this piece decouple them from all the PowerBI data model and just have you know fstring substitution in Python or glue in R substitute whatever you're selected into the query run that populate a grid yeah a bunch of these are that and the hoops that PowerBI will developers will jump through in order to do some of these is just astounding because that's their hammer, right? So that's a that's the issue of what shape is is the problem. And they'll look at it and go, "Oh, I recognize that shape. I should do it with this tool." And I'm like, "Let's think about this just a minute. I'm not saying you shouldn't, but let's make sure we really understand what shape it is before we assume that's the right tool for doing it." Yeah. the uh what was I going to say the uh the the whole the problem is really you got to define it right like is it do I need an answer now are we building something longterm for the future that's going to be replicable all of this stuff because it really does change I I am personally very guilty of something a simpler example of what you were just talking about where it's like yeah I should probably do this in the database and I should do this transformation in the database But I'm in PowerBI right now, so I'll just do it in PowerB. Right. Like, right. Yeah. And then it's like, oh well, no, we should just go implement this in the database so we're not doing all this extra [ __ ] Yeah. Do do you have a road map of how you think from moving from sort of a piece of it like analyst analysis to, for lack of a better term, we'll call it production, but I don't want to get in a big debate of what production is because I don't know. [laughter] I I do want to talk about that because that that's probably one of my favorite data tweets is that you have us about the whole Excel being in production thing. But uh but yeah, I mean again I again like I said I think I take a lot of way I think about data and how I manage it now like at our company like you know kind of from stuff that you've talked about but um I mean like we've got analysts in each of our different kind of functional groups but we have like a weekly meeting where we're understanding what they're working on so making sure we're not reinventing the wheel or that that that they know we're already exposing that. Um, but again like John said, like I view Excel, Spotfire, PowerBI, these are great prototyping tools and I need those people and and if that's their tool that they can do their work in, that's great. But once I realize that they're making business decisions on it and this is something that someone else needs to leverage, like I need to take that logic and pull it back into the data warehouse and we're going to calculate it there because it's going to be faster. I can materialize that answer as like a table or an output and they can pull it in much more quickly and efficiently. [snorts] Um you also have that element QAQC control too, right? Like that's a big thing that well yeah in that it's institutionalizing the knowledge right like I mean and again it all comes full circle you know we're using DBT now we it's in source control you know so you know any of these key metrics or calculations like people are getting it from the same spot I I you know I've kind of gotten the point where I mean we use a lot of typical spotfire you know PowerBI and people are going to use that on the edges it's inevitable Excel but I I I've been saying it like I want people to be drinking out of the same faucet but I don't care if they use a tumbler or a glass or a coffee mug like I mean like whatever but like you're going to get the answer from the same spot and it's going to be calculated the same way. Let me state this a different way and tell me if we're talking about the same thing. I'm pretty sure we are, but you basically want like the business logic being done in an ETL process that's version controlled and dumping into data base. And then you want the display to have very minimal logic and is basically like pivoting, transforming just the shape of it, but not doing quote unquote business logic. Absolutely. Yeah. That's that's the religion I've come around to after years of doing this. And one of the things that's really uh that I, you know, kind of fight a little bit against is that's a good idea. And as soon as a user opens PowerBI, the thing they're prompted to do is start transforming their data. And [laughter] that's very true. And I have this debate with with like a buddy of mine who works at another company and and his argument is like, "Yeah, but that's a huge improvement over the crap they were doing in Excel." And I'm like, "Fair, that's fair, right?" So he's like, "At least it's reproducible." And I'm like, "True." Uh, and you all are on a different point in the maturity uh of your data organization than we are. And for the folks on my team, I don't want them doing that. I want them writing Python usually orchestrated by Airflow in our shop and dumped into a database and then have the whatever reporting tool they're using and the hell the reporting tools sometimes it's best is a pivot table in Excel but I don't want any logic there I just want it presenting data uh you know or they may be writing an SSRS report or they may be u you know a web app of some kind but to your point Bobby it's all pulling from the same data so it's like you know one source of truth And you can see what transforms went into that data by looking at the code that should be readable by everyone in the organization who might want to. Exactly. Yeah. And that's beautiful thing about you know dbt and I'm sure I know there's been a lot of you know hate on the modern data stack this and that but I mean like as far as like I'm concerned like SQL is that lingua frana right that you know a lot of people can read and you know people have access to our data our lineage and they can go see the code that transformed what they're using. So there's some you know um you know transparency there. Um, yeah, I like I like the literate programming kind of thing. Well, literate programming. I like that you use the same source to build the data and then it creates the documentation. Yeah, I don't use dbt and anger much. I've just looked at it, but I literally have not built a single thing that's like in our system using dbt because we already had a bunch of stuff that worked to solve that problem, right? We already had we already had Python code. We already had Airflow. And so it was like, well, we didn't really have the problem that gives people the incentive to initially grab it. Uh, but we've done some spikes with it. And I think the generation of the documentation is some of the neatest bit of kit in there that I think a lot of folks sort of uh don't pay as much attention to because they're so excited that they can do so much with just SQL. And I think those of us who have maybe more tools in our toolkit, that bit's a little less exciting and some of the other bits are maybe more exciting. Yeah. No, 100%. I mean because that I think that is like people sleep on the full ecosystem of that you get the documentation you get the lineage and now they have column level lineage I mean and it's just be able to trace those things up and down and you know even and to test the data which again you can write tests obviously in Python um but again how many that's something people are definitely not doing in their uh in their BI is charged yeah [laughter] no I think I think what Bobby has kind of described is looping this back to the the low code stuff right like that to me is the like optimal pipeline Right. I the last company I worked for was a edge computing startup. And in my journey on in the sales side to to frame our our presentations and our pitches, the best analogy for what we were doing at that time was like this is basically like low code, no code, but for hardware. We had these Kubernetes clusters that we could stand up on the edge that they were literally like Legos. You could stack three on top and now you had a full cluster. Um, but you literally push button deployment, right? Like I I was deploying Kubernetes. I don't know [ __ ] about configuring Kubernetes. I have no business touching it, but I could stand up a cube cluster very easily. But like with so anyway, the the data that I researched on all that stuff talks about the the reason people like low code is taking off is uh you know the the business theme whoever that may be business analyst etc. It gives them the ability because they know the problem. They know what they're trying to figure out. Whereas a developer or just a pure data scientist that doesn't know the context or the industry per se doesn't. Um it it allows them to go in and actually build ways to solve their problems or find the answer whatever it may be. And then it's like so doing that with PowerBI where it's like you guys serve the data up to the user. They can build out the dashboard to a certain extent how they want it. Then you guys take that and basically turn it into code and productionize it so that it's just it's this nice feedback loop, right? And that's how it should work, right? Give them the ability to solve the problems, then figure out what they're doing, then take it and productionize it so that it's across a company almost. Let's poke on that just a little bit because this is a classic what's the shape of the problem. Like I think the problem we're solving is communication problem. Yes. And you know, it's like the the folks who understand the business and understand what the stuff means are over here and they often lack the technical tooling to do what this group of people do over here on the technical side. And I really feel like the single biggest value of the data science movement of the last 12 years is putting a whole bunch of us in the middle where it's okay to not be completely this and not be completely that. I mean, when I was first doing stuff that looked like data science, I had to fight, lie, and steal to try to get tooling on my machine because we had been part of I was with a company that had been part of GE and they had a huge amount of rules around what you could have on your computer and I wasn't allowed to have programming language, right? Right. So, I had I had to do VBA cuz that's Excel, right? That's what that's what people in the business are allowed to have. You can't have R or Python, right? And that's just insanity now. And because of data science, folks are coming out of college now having had that and seeing the usefulness of these tools and companies everywhere, you know, don't see this as a problem. That's just, you know, many of them have hosted environments for doing this. That's a huge structural change versus a number of years ago. And it's like what we're trying to solve is we now have people who can write some code and they're close to the business, right? So, you know, it's funny you described the data scientist maybe being more like a developer and they don't understand the business problem. I have a really strong religious belief that your your data scientists should be in the business, not in any other group. And they should be domain experts 100% because I way think that more than they should be code first. They should be able to get things done with code, but I really want them, you know, and maybe I'm just biased because that's kind of who I am. Uh, you know, I came into it all from the business side and then learn just enough coding to be effective and get things done. That's where I'm at, too. So, [laughter] I might be a little biased as well, but So, we may we may all be having confirmation bias right here. Right. [laughter] Well, no, but I mean I think that's a good segue too because I want to kind of get into how you kind of got into the space, but I I know you've talked a lot and we like to talk about people's journey, you know, and maybe a good path for people like into data or into, you know, the uh tech world or whatever, but um was you know, you've talked about I think before like you came in and you just had to be useful and I that's kind of how I felt too like I came in like I I was a high school teacher and a coach and I got lucky enough that someone said you should be in oil and gas and I came in and they're like you're going which happens to high school teachers in Texas in Texas only. [laughter] Yeah, pretty much. Yeah. And I showed up, you know, my my elevator pitch is like, I didn't know what a VLOOKUP was. I walked in and they're like, "Oh, yeah, just do a VLOOKUP." Like, "Wait, you got to slow down. I don't, you know, but you're going to be a data guy." But, you know, um but I didn't know data science was a thing. And then all of a sudden, that was 2014, right? The whole Time magazine, right? You know, and I was like, "Well, that's cool." You know, let me lean let me lean into that. Um and you just you find But I think the biggest thing is you got to find ways to help the business solve problems. Like I've I've talked to a couple other people trying to follow a similar path that I did. They commanded teaching. I'm like look it's great that you got this Python certification from Google whatever like do not go get a data job. Get into a business and find a way to make improve their outcomes like you know how can they report more frequently. How can they you know get more visibility and or automate things? And if you start solving those problems you're going to make yourself useful and you're [snorts] gonna you know that's your niche sticky. Yeah. It's gonna be very hard for them to get rid of you. That's uh Oh, it almost feels like a career moat. We've come full circle. [laughter] Yeah, it's true. Yeah, you you are you do become sticky because you're just so dang useful. Um I I got got some really like crazy good advice from my dad when I was like a junior or senior in high school. So, I graduated high school in ' 91. So, this would been like 89 or 90. Uh my dad I grew up on a farm, but my dad had been in the Air Force. He was a pilot and sort of his second career was farming and I we always had computers around the house. Dad was very into computers and uh did a bunch of computer work at the Air Force when as a reservist and you know I was like oh I want to go into computer science. My dad was like no you don't. I watch you you want to play like Larry Bird versus Dr. J on your Commodore Amigga. You don't want to code. You're just like you're not you've showed no desire to figure out coding. You've done like a the equivalent of bash script on my Commodore amigga. I can't even remember what it was called. But anyway, he's like you've done a little bit of that, but mainly because you wanted it to do something cute. You don't really live and breathe coding. What you should think about doing is solving technical problems like from within a business or industry that you find interesting and then use that use computers to do that. Yeah. And I'm like, [ __ ] in retrospect that's like data science right [laughter] you know without any structure around it just at its at its core and um I started college as an agricultural economist sorry as an agricultural engineer and I quickly learned that my rural school system in my class of 51 I was woefully underprepared in math and science [laughter] from rural West Kentucky I imagine you guys find that shocking but [laughter] I just about dragged found in, you know, physics and calc and I sort of pulled back and said, "What the hell am I doing?" And ended up in uh an agricultural, I think it was like a rural sociology class and I was like, I didn't know anything about social sciences. Um, and I'm like, this is actually really interesting. And ended up an agricultural economist. And then I took econometrics as part of that, which for folks who don't know, linear regression. Um, and I was like, this is really interesting, right? And I did my senior uh not senior my my keystone or capstone project in econometrics on um what are the impact if the factors that drive commodity prices and you know discovered it's interest rates uh you know because it drives import exports and it drives commodity price up and down and you know as a 19-year-old I didn't know that or maybe I was 20 or something I didn't know that and I was like holy [ __ ] I just used this tool to actually learn something about something I care about and that was really like the switch going off that this stuff is really useful. Now obviously there's theory and other people knew all this but I didn't and I was able to figure it out with this tool and that really you know I learned a little SAS and then I learned some Excel and then um you know ended up like a summer project helping migrate some data at the college of agriculture off of SAS tapes I think it was using like the data control language and into a Unix mainframe sorry coming off of an IBM mainframe moving into a Unix cluster like some kind of HP cluster. And that was kind of the start. And then one of the things I noticed is my major professor I did my master's work under, he was one of the few who wrote his own code. A lot of them worked with a they called them I can't remember their programmers or I think they were called programmers, but they would have an assistant who implemented what they wrote down. And the folks doing the analysis didn't actually code it. And what I watched Dr. Jerry skis at University of Kentucky do is he was doing it all himself and he would discover things he wasn't looking for. And the other faculty I felt like didn't have those epiphies or those kismmet moments because they weren't like exploring the data themselves. they would be like here's a theory here's a chunk of data I want you to code this up and test this hypothesis and it was like either fail or you know not fail to reject or whatever and and they didn't get any learning and so what I later would learn is Jerry would do a bunch of stuff that was effectively EDA right exploratory data analysis and he would figure things out or he'd figure out oh I got data quality issues meanwhile a lot of these other folks didn't and their programmers didn't know to look for these things what you know and it was more like what we're talking about in business where the you know software developers and the business people had a chasm and they were passing notes back and forth. Well, Jerry was both of those and I'm like he's terribly effective. Yep. Yeah. And I I was like that's just what you do because that's effective. And you know then the other piece of of very preient advice my dad gave me was get useful quick. Yeah. When I went into the workplace. And so I found that having handful of these tools I could be really useful really quick because we didn't have to pull in a software developer. Now we weren't writing software. We were often doing stuff that could be done in in Excel or in in you know Access database. I mean this was you know by the time I hit the the workforce I worked at the university for a while and uh at that to help pay for graduate school and so I ended up spending as long getting a master's as some of my friends did to get a PhD. uh but I only have a master's degree because I because I worked and then you know so about 2001 I hit the the workforce and I just thought this is how you do it right this is how you get things done and then later found out some companies don't like that right they don't expect [laughter] they don't want you uh you know writing something looks like code that's not your lane you're supposed to be in your lane over here and somebody else supposed to be in that lane I'm like no no but that's less effective because there's all this communication friction right but I'm also solving a problem right which is at [clears throat] the root what you want me to right? Like that's why [laughter] you're paying me a salary. Yeah. No, that's Yeah, I think this is a good Well, the that's one a couple things there. I I think you've talked to I think because didn't access at one time or there so it's like there's kind of that guey to generate SQL, right? Technically, I mean, and that was the way a lot of people were learning SQL back. Totally. How I learned to do SQL was Microsoft Access had like a drag and drop. You'd see your table with your list of fields and you drag a field and drop it on the other one and it would make a join based on that field and you would see down below it would write the SQL. That's literally how I learned to write SQL. I can't believe you remembered that. [laughter] Yeah. Um [snorts] and then I think some other things there too. I think we're starting to butt up on like a shadow IT conversation. Um, and I think, you know, it's almost like you talk about Moes and I think they try to put their moat around themselves, but like the thing is when you have problem solvers like yourself or me or, you know, we I work with a ton of engineers and John's an engineer, you know, you know, he's done a lot of sales, but he's an engineer by degree and mindset like if you put something in their way, they're going to find a way around it. Yeah. It's like the internet the internet if there's blockage, it'll treat it as an obstruction and wrap route around it. Yeah. Yeah. That's also the fastest way for me to be motivated to go get around something too is [laughter] you putting something up where I don't think it should be. Like my tendency to blow moes up or go over them or Yeah. A little bit of a anarchist built into a lot of us where it's just Yeah. Because it's like, okay, this is an unnecessary problem. I'm trying to solve a problem and you're just throwing [ __ ] in front of me that is slowing it down. It it's Yeah. Let let's talk about shadow it real real quick. And I got to start this with a caveat. I actually got myself in a little trouble with that comment on Facebook because some of the people I work with now. Uh there was some potential for hurt feelings because they thought my engineers I work with now thought I was criticizing. I'm like no we don't have shadow ID. You know what are you talking about? And I'm like so I'm not talking about Renaissance 3 when I have commented publicly about shadow IT. I'm really talking about some prior experiences and the experiences I see a lot of my peers having. So the the shadow IT concept is basically when the people in the business can't get what they need from their software team. So they start doing things that the IT side of the house thinks is their job. And it almost always stems out of someone on the business side can't get the things they need or can't get them fast enough. And instead of seeing it as a like failure of it to be responsive to the business needs, the IT often sees it as like a rogue security threat and approach it as such. And sometimes it is, but it's a little like, you know, I know some people who express uh hurt feelings comes across like anger. Yeah. You know, and it's a little bit like that. It's like you're seeing something and yeah, there's some security issues there, but what you're really seeing is your own failure. Mhm. Your own being the IT side of the house to give the business the tools they need and they have often put put in so much friction in getting work done uh that they made it impossible for folks to get work done. So, as as John said, the business routes around them. Yeah. Because they aren't they aren't helping, they're hurting. And this actually the the way we have solved this at Renry is is I don't want to speak for other folks. My experience has been I don't want to speak for Renry. The way I have observed this being effective is partnership. Like I meet every two weeks with our head of platform operations. I'm in the business. Yeah. Why? So we can get stuff done. Yeah. [laughter] And we always have something to talk about. And if I was having to interface with her through submitting tickets, I'd pull my [ __ ] hair out. Yeah. Yeah. No, that's And well, I [snorts] think people get that people get that like uh people talk a lot about it being, you know, a cost center and all this stuff and they're always in the way and it's because the structure almost forces that, right? Like they're in their IT silo and your directive is security or whatever whatever it may be, right? So it's like, okay, I'm gonna lock this [ __ ] down. It's like, well, that that doesn't fit with the in what the other person then you're going to do something even more risky to to get the job done, right? So, so this is a misunderstanding of optionality, right? And I see legal teams do this same thing. They get no benefit from the upside, but they feel like the downside is completely their risk. They feel like their career is selling puts. Yeah. So, they don't want any exposure on the downside. Yep. because they but but they don't give a [ __ ] if you know if it makes your life really hard because that's not their job. Yeah. As long as my risk it's all about the risk, right? As long as long as we completely risk. Yeah. Right. Yeah. And and once again, that's not my problem at Renaissance 3 where I currently am at all. But I've seen it up close and I've seen lawyers who just I think said no to everything. I was at an insurance company and the lawyers would say no to anything. Yeah. And it's like they're like, "Oh, but there's risk." It's like if do you know what we do for a living? Yeah, we literally value risk and sell protection risk. Like that's what we do. Why can we not treat this risk, you know, understand the size of it and talk about the tradeoffs, you know, is it worth the risk that we're taking? And I was really surprised later in my career when I met attorneys who think like that. Yeah. Who are like, well, this is an aggressive stance, but if this benefit, you know, outweighs this potential downside, I think we should go for it or whatever. And it's like these people think right, you know, they're thinking like, you know, risk and distributions of risk and thinking about the the the benefits and and and the cost of that. Um, and and our IT folks often do as well. They see just the risk of, well, there's there's a potential security risk, right? Well, I can eliminate all potential security risk, but it involve taking all these machines and running them through a shredder and piling the bits up in the corner of the office. Uh, but otherwise, there's going to be some risk. So now that we have we all agree that there's going to be some risk, let's talk about what risks are acceptable and what value we need to see, right? So yeah, that's a that's a big thing. It's I'm sure it's I know it's not unique because I'm just a big proponent that most business problems are in unique regardless of the industry, but our our industry as a whole from top down, everything is approached based almost exclusively off of risk, right? is and it it took me a a while to figure this out. Someone once told me they were like, you know, Facebook's algorithm going down doesn't really affect your daily life, but you know, a well blowing out and killing people. That's a real thing, right? Like that's a huge issue. That's a legit risk. And so our industry is very hesitant about technology implementation. I won't say adoption because we do look we invent a lot of really cool [ __ ] and we we uh we test a lot of stuff early but we're going to be a slow adopter because in most cases there's this inherent risk component right and so it's it's but yeah it's the life is risky right like you can't it's the whole yeah it just we all we all get in a car which is like the most risky thing we'll do all year yeah you know the people that are afraid of flying and it's like oh well that's a hell of a lot safer than driving [laughter] like go drive in Houston traffic for long enough. Yeah, you'll see. We could get into a whole world of like cognitive bias discussion here, right? But kind of backing up a little bit, I think this idea of uh thinking about tradeoffs is is really powerful and making sure that everyone involved in the decision has some ownership in both sides of it. Yeah. You got to get rid of the silos like Yeah. Because in the in the silo, everybody's just responding to the risk, right? Yeah. And it should be it is an enabler. They're not a a moat. Oh, and one and one of the things that we did is I think I think is really great and this was completely done by my IT side of the shop is they literally have a group called cloud enablement that's trying to take the friction out. Yeah. And it's like amen to that. I had no involvement in it. Right. Uh in the creation of it. I have involvement with them but I didn't you know create that or anything. I think it's a great idea, right? Like we should have more enablement, you know, that's that's the upside that is teams trying to help other, you know, the business get the business done. That's that's that's what we need more of. Yeah. It should build a moat around the company, right? So people aren't trying to get in, but within the company. They shouldn't build a moat around everything. They should enable the company because that's what they're there to do. And Bobby and I, we had a security guy on a couple weeks weeks. I don't know how long it's been, but um [snorts] you know, he was talking about like, okay, you you put up all these external blockers. It's like all almost all of the risk comes from internal, right? It's like someone gets a [ __ ] fishing email and then they have access to the account and then they click the button. Yeah. They click the button and now the next thing you know, they're getting emails from the CEO asking them to buy Target gift cards or [laughter] transfer money to this new account, right? And it's like that's way riskier in a lot of situations than, you know, letting an engineer have access to a data table that he needs access to. Like, [snorts] so so let me let me back up in our conversation and and scratch on uh one thing that we kind of kind of blew by and that is like how how do we move stuff from an analyst has figured out how to solve a problem, but it's bailing twine, duct tape, and bubble gum, right? and Excel centric and they're copying they run a query over here and copy and paste the results into the Excel spreadsheet. How do we go from there to something that I don't know if it's production but it's at least well controlled. Yeah. And more behave. Do you guys have a design pattern you like? [snorts] I mean again for us I mean like we're we try to fit most of it like into into DBT you know. So we'll try to take whatever they built you know the logic that they built and you know back that back out again into a source controlled spot. And then again we serve out you know our production tables through like a certain schema. So we have like kind of a gold you know if you we're going to call it a medallion architecture be like your gold layer but okay you know but pretty much most of the company has access to those that are kind of um you know those tables or views that have been vetted and have you know have gone through the ringer a little bit and then that way it can be served out. But it's again you've really got to iterate and understand what people have built and what they're using. [snorts] Um because you quickly get the temporary but permanent solution. Um which you know and again then it comes down to how do we define production which [gasps and snorts] John do you guys do this? Do you guys move it from that or do you just basically try to hide so no one knows that you have this [clears throat] this gubbins held together with balen wire? You're looking at the DW uh data engineer analyst and IT support person. Yeah we have 12 people. So my my pipeline is my boss comes over and says I want to know this metric so I update the dashboard with that metric or I need this report or whatever. It's uh it's much easier [laughter] at this size. But yeah, when you're a oneman shop it's a whole different beast. But now I mean even in the last couple months I've brought in like three different people like on my team or people like in the business and it's a whole different beast once now you've got a bunch more users and you know hands in the in the pot and everything just you know you really got to start you know tightening things up. But I it's it's always that push and pull like you're saying those trade-offs of like um you know how much access do I give these people but also I I don't want to be the one keeping them from getting their job done right um and you knowve allowed more people have access to the data warehouse but with that I've had to put you know I put together like a pull request template where they have to show that they've done certain things to oh that's um to make sure that they're doing the proper things before they push it through and then just and then it's got to go be reviewed by me and or one of our um our senior developers um just to you ensure a certain quality of code. Again, we're not super rigorous, but it's you need something, you know, but you know, it's it's it's hard to you know, I think we're going through that mediumsiz company thing and John talked about it like you you know, when you're small like John, you your team can move fast, you can break things and it's fine. And then you get to this intermediate, you're like, man, I don't want to become like Fortune 500 bureaucratic and everything's locked down, but at the same time, we need to have some some rigor. Yeah. Um around everything make sure that there's a standard that we hold our company and and the developers too even on it my at a company of our size like I said 12 or 13 people whatever it is right like there it's very easy for any most of us are also in the same office for someone to be like hey can you like you know update this on the website or update this in HubSpot or update whatever whatever it is and it's like yes can I do that right now though? No. So now we've got to figure out a system so that your requests can get in a queue and I can actually get to it at some point because it's not a high priority. It's just an idea that you have. But the idea is great, right? Like yes, this makes a lot of sense to have that metric somewhere at this field or whatever. And so it's like yeah, we I've implemented the same thing. We use ClickUp for all of our it's kind of like Jira. Um we use ClickUp for all of our um development stuff, all of our project management stuff. they now have like a basically a full-blown notion replica in there as well for documentation and all this. It's really powerful because it's it's a it's a pseudo noode type platform. Um, but [snorts] it sounds like it's scratching a bunch of really useful itches, right? Well, that's the thing, right? It's like, okay, well, if I'm documenting and doing projects, I also need form generation. Oh, they've got forms. I also need documentation. Oh, they've got documentation. I need a mind map. Oh, they've got that built into, right? Like, so it's got all these things. But even as small as a company we are like there's a bunch of pin links to different forms for different requests in our Slack channels because that's the only way to as you scale even as a small company right like you shit's coming in all the time from every different direction and so it's like yes we can move fast but there's always something to do there's always something broken or needs to be fixed or added or whatever and so even yeah at small scale just being able to document because in my experience mostly working on the startup side like So much of that [ __ ] just gets lost. Like you have a conversation about it at a meeting or at lunch or wherever and then it's just like and it just get lost or they throw it in a Slack channel. It's like okay well great, right? What is that going to be? Which is lost. Yeah. Exactly. [laughter] But I repeat myself. Yeah. No, absolutely. You know, we've we've gotten just big enough that we we think about this and it's it's really consistent kind of with both I think with John what you and Bob both said. We're our organizations are at different points. I mean kind of John on the small side, Bobby on the medium and we're large financially but medium by headcount. I mean the whole organization's a thousand people but um the thing we've seen is is we think about a continuum and you know maybe it starts with like an analyst figuring something out in Excel and that's very iterative very high flexibility very low control and then maybe the next layer of rigor is the group I'm in solutions who's going to do we're going to have our stuff in version control we're going to have it's going to be more code less copy paste. It's going to be automatable. Most likely it'll be run by airflow uh but less robust than our gold standard which is when our IT devs build a proper product out of it. So we we understand that's a continuum and what we've broken ourselves from is this group looking downstream and saying your solutions are illegitimate. Yeah, they're not illegitimate. They're making different sets of trade-offs for different reasons, right? And we all know what those are and we talk about them openly. And so, you know, if we're having problems here because we don't have enough controls, we need to go here. Or some solutions will never be used until they're here. So, like if you think about what feeds your GL needs to not be built down here, [laughter] probably doesn't even need to be built here. Like that shit's got to be right. and you know but however like a metric on an internal dashboard might be fine to just be created here in the middle uh you know it's not like we're going to report financial earnings wrong if if it goes down for a day or something right so we have different levels of reliability we have different levels of validation we have different levels of uh you know error checking and error handling as we as we go up that continuum and and nobody is looking at each other saying they're illegitimate We're just trying to get things in the right place. Now, we we a couple things that started kind of as a joke that have become like things we we just do. One is we do this thing in the middle where it's like it's better than the limpalong solution, but it's not the true production and we call it a productoype [laughter] and it's that well it's a little more than a prototype and we've been using it for a couple of quarters. I mean, and sometimes producto types are generating stuff that goes to the board of directors. Sure. you know, I mean it's but it hasn't been put into a formalized IT controlled full process, but it's been reviewed and you know, we have one of the things we've done is we've adopted the language of our business. So in you know we're a risk trading organization so we do a thing with deals called a second set eyes and we often just abbreviated SSE. So we do SSE on all of our deals. So we do that in our like solutions group as well. So this would be instead of saying, "Oh, we're going to do maybe full code review." Well, we're going to get a second set of eyes, and that means we're going to sit down with somebody else on our team and explain what we're doing, show them the code, and they're going to say, "Oh, okay, that makes sense. I see what you're doing." They're not signing off that I have independently validated tests on every calculation, but they giving you a sanity check, right? And that's better than not having that. You know, it's like a souped-up rubber duck debugging. You're having to force yourself to at least explain it to another human and them say, "That makes sense. that isn't crazy. There isn't anything obviously nuts here. So, doing the the SSE and second set of eyes, uh, it moves us from like the typical analyst to this thing that's kind of in in the middle, right? And that's a little bit, Bobby, like you guys are talking about doing adding a little bit of controls around that. Bobby is my SSE. I wish I had one internally. [laughter] You're right. I know, but you're right. Like I feel like that's almost human nature, right? like I think this is right, but can [clears throat] can just someone that I trust give me some kind of validation, help poke holes in it? Yeah. And especially as you're learning to do a new tool or a new library or something, it's like, is this the way to do it or should I be thinking about this problem differently and it's also a good time good opportunity for mentorship, right? Because you have your more senior people doing SSSE with your more junior people and they'll see, oh, well, let's talk about how to do this more efficiently, right? or hey there's actually this design pattern was worked out over here. Let me show you in this other repo a design pattern that gets you the same thing easier or something you know like and so it becomes like real time learning. Yeah. Uh and mentoring which you know is how we build people. No for sure. And talking about John reaching out to me it kind of brings up the whole uh low code thing too. was this past week, John, you reached out and he had done the same exact thing in PowerBI and in Looker and like the table the grand totals were different, but the tables looked exactly the same. I went through and hand calculated it multiple fortunately it was only a month's worth of data and it [snorts] was just daily data, right? So, like it was but I was in bed like in my bed 9:30, my laptop open, my phone out, literally calculating each row of one table [snorts] and then doing the exact same thing. I handed the phone to my wife and I called the [laughter] she called the numbers to me, I called them to her because I was just like, "What is happening?" Yeah. This is stuff that gets to hashy daddy drinks, right? [laughter] Yeah. And it's like, so what ended up being the cause? you can't bring me here and not resolve this this tension. So, so I had just like the two data tables next to each other and then I you know the sums at the bottom, right? And then I was like, oh well, let me see what other transformations are available in Looker for this data set. Oh, there's a running total transformation. So, I I copied the table and changed that one to running total and looked at it. The last number matches the number from the PowerBI data set, but doesn't match the sum the grand total the grand total of that same data set. it is bigger than the sum in that to data set and I'm like well that just shows it that it's wrong like that's how I ultimately figured it out with in conjunction with the calculator but it was just like what is happening and it's something to do with first of all uh only use Google products that lots of people use because things like looker are horrible for a reason um and so but it was something with how the SQL underlooker you know understands time or like calculate sums over time. I don't I don't know why or how whatever. I think it was almost like their version of DAX I I want to say like but where like if it doesn't do the filter on the organ it does like this the total on the full data set not not without without the filter not the filter data set. Yeah. I just wanted this range of data and I want the sum of that data from this range and that's exactly what it was. This is an example of you had an abstraction and it leaked all over your foot. Oh yeah. Well, like [laughter] unfortunately I was using PowerBI as my main dashboard and stuff anyway, but my one of my counterparts who was on the like marketing operations side was using this and we were both pulling from it was Google Analytics data. We're both pulling the same data set from the same source. She was using Looker and I was using this one and they just didn't match. I was like, what is happening? And that but that's scary too because if someone doesn't know what they're using, I mean like they're making you know assumptions or business you know, decisions off of that. Like this is the rise of the of the metric store right now. In this case, I'm not sure a metric store would have solved this exact problem, but the idea what we often see is you know a loss ratio is something that we calculate in insurance and reinsurance all the time. It's um you know losses divided by premium. So you want it less than one, right? To make money. Uh [laughter] if it's if it's bigger than one, you're paying out more in losses than you're uh than you're bringing in in premium. So, premium divided by um sorry, losses divided by premium. Yeah. Um so, anyway, there's often something like, okay, well, what if you do a no claims bonus? Does that come out of the the numerator or the denominator, right? Like, what do you do you add in one or do you subtract it from the other? And if you don't have some common place where a bunch of this is calculated, everybody will calculate it differently or there'll be there'll be you'll know of two ways to calculate it and the users will figure out two more. So there'll be four different ways of calculating this floating around your organization and you'll spend like I've been in executive meetings where folks were like spent half of it trying to figure out why two numbers don't match, right? That's the last place you want to be, right? You're not making any business decisions. You're just debugging your reports live at $10,000 an hour. Yeah. Well, I [laughter] mean to that point though, like that's the value of having that second set of eyes because it really like had I not been here, she would have built that dashboard and it that's what they would have used and then at some point they would been like what is happening like something would have would not have matched from the raw set and then but yeah it's uh it's always good to have someone curious about the data looking at the data anytime that you're dealing with the data [laughter] instead of just I tell my analyst I'm like anytime you have that feeling that's kind of usually manifest itself my inside my head as huh that's weird. Yep. Yeah. The your job is to resolve that. Yeah. You know, I tell them and and my job, right? It's like every bug I've ever found, every screw- up I've ever made, every one I even ignored and later blew up on me, there was some point where I had a huh like I would not have expected this quarter that number to go down. Oh well that no you just you just presented something that was wrong and trust but verified that's the yeah exactly if you if you get and well this is the value of seniority and experience right is you have a more refined heristic in your head that goes huh or you know immediately let me validate that number against this over here okay those don't match they should and a young analyst or a new person they don't have the intuition so they look at it it doesn't feel wrong to them because they don't feel anything, right? It just it's a bunch of numbers. There's some on the left of the period and there's some on the right. I don't know. Um, and they also don't know where they can quickly validate it and they've got no intuition around it. Right. So, this is part of why, you know, we get paid more as we get older because we build a bunch of this intuition. Yeah. Well, and luckily like I had similar experience but ended up I had done the calculation right but I I did something but it actually I kept checking and it worked but it seemed too easy and [snorts] I was like I [laughter] no cuz I I was just adding a condition to like a case statement but there was a lot of kind of nested I was like there's no way it was that there's no way it was that simple like I need you to that's because you've had a lot of trauma in your life and you should talk [laughter] about that with your therapist because that's that's the trauma speaking. It's like oh this didn't hurt. I must be doing it wrong. [laughter] It's true that but like that's you know that's people always they talk about failing fast and stuff and one of our last guests was talking about it's like we need to completely get rid of that failure word right and it's it's learning or iterate it's iteration is really truly what yeah it's back to iterating right it's not failure you if as long as you learn from it and it goes into your experience bank and you're like oh the next time that happens like Bobby it's like oh well maybe I should look at this or like it throws that flag off then it it helped, right? It didn't hurt you. I I think this iterating we're back to our conversation we started with with, you know, there's people who know business and people who know the development and and then they, you know, are passing notes back and forth to each other. And the the problem with that and I think the problem with maybe even agile development as it's currently being presented. This isn't how it originally started. But it's adding so much formality to that that it almost acts like discovery in the process of development doesn't happen. You work all that out ahead and then you just execute it. And like that's completely malformed based in my life experience. I have never worked on a project that worked like that in any like everything from working on my old Jeep to, you know, building stuff in, you know, inside of a of a major corporation. None of it works like that. You you you've got all this uncertainty you're trying to figure out. You think you understand kind of how something works and then it punches you in the face and you realize it doesn't work like that. So, you have to change how you're think. And if your if your people who understand the business are over here and are just writing their request on a node or helping create a user story and then a developer is supposed to go off and do it, you're just missing that iteration that fast high uh I guess what is it low uh friction high fidelity high-speed problem solving. You know, I I did a had a retrospective with um with one of my peers at one point about a project and we got it done. It was okay. He said, "You know, if I had to do that again, I would have taken one developer and myself and we would have just blocked a conference room and moved in for two weeks and we could have gotten done the two of us in two weeks what, you know, it took a whole team of, you know, three or four to do in a couple months." And he said just because there was there was not fast enough iteration, not high enough bandwidth of communication. And that was I'm not meaning any insult to anybody. Everybody had best intentions. It's just the idea of talk to me and then go away for even a couple days and then come back talk to me isn't if you haven't if you haven't gotten rid of the technical uncertainty that's too slow. Yeah, you need I like that you use the term fidelity and I completely like that's a great way to describe that, right? Is it's like there's way more fidelity than just yeah a request in a form or a user story, right? Like I I've become I've gotten into the practice of anytime we do any kind of development that has any kind of guey related to it, there's a mockup with it because you cannot put a mockup into words where every single person that reads that will interpret it the same way. Impossible, right? Like I I think the UI folks call it like a wireframe. Yeah, a wireframe, a mockup, whatever it is. Yeah, mockup. Yeah, if I'm making a request that there's a new feature that requis
your point, a lot of developers want the coding challenge, not the end use challenge of figuring out whatever the the big problem is. They want to figure out how do I code it and check the box off of my list so I can go code the next thing. And so it's like, well, yeah, the function's there, but it's unusable from a normal person because you buried it in a menu or you put it somewhere weird or whatever. Seven clicks deep. Yeah, exactly. And so it's uh Yeah, that's Bobby and I have run into this. We we built a a web app using third party, you know, external developers, and we got to the point in the company where it's like, okay, we want to move faster. Well, do we hire more external developers? Do we bring someone in-house where it's like, "Hey, when we want that problem fixed or that thing tweaked or that color changed, I can just go over to his desk and say, "Hey, hold up real quick. I want to show you something and change this for me." And then it's done. And it's not like, "Oh, submit a ticket. Well, where did you want the like there's all these qualifying questions, right? It's like, no, just sit with the person and do it and knock it down." Yeah. I I think there's like um there there's something with if you have commodity type work um and an example versus like bespoke work. So commodity type work is we want this stuff that's written in three different frameworks literally ported to a different framework. Yeah, that's what we just did, right? Here's my my no code app. Build it in Ruby like and you can test everything because it should give the same answers as the existed one, right? So, it's like there's not a need for that fidelity of communication, but if you're doing something new or something ambiguous or something with technical uncertainty or discovery around it, you can't just hand it off. For sure. It's uh Yeah. Well, because you'll keep you just keep iterating on it anyway. And that's the like that's where the frustration comes is it's like, "No, I meant it like this." and then they go back and they change it and then two days later they come back with it because it was down in their queue and it's like oh well that's not exactly how I meant it. Well, [laughter] even like you mean like I think the last one and then we'll probably have to you want to jump into our uh you know speed round. Yeah, we did it with the uh we had the developer and we were using time series data coming from sensors and they ended up building the time series plot off of the updated time stamp in the data because it would come in batches but it would be one second data but come in like five 10 minute batches. They plotted it. Oh, this is what you wanted. They used the updated time stamp and not the actual sensor time stamp. But right, you know, they didn't know. But but then that was even something. Yeah. Yeah. Yeah. And you've got to have that that iteration, right? And I think to this to this point, one of the things I want to fall all over myself to point out is in all these examples, everyone is working from good faith. Yeah. and they're trying their best to do the right thing, but we have friction in our communication or we have lag in our communication or whatever. So, it's not the people that are the problem. It's like, oh, we just need a different way of working together. We need a different way of structuring our projects so that everybody can be the best version of themselves, which is what everybody at the table wants to be. Uh I I honestly well most of the time I've worked at some places where I'm not sure that was the case, but absolutely where I am now. You know, there's no nefarious intent. It's just sometimes we don't set ourselves up for success. We set ourselves up for too much friction. Yeah. Yeah. 100%. All right. Y'all got a lightning round me? Yeah. I'm gonna go first because I I think uh this is a good one. Favorite video or board game? Oh, so video game or board game? Or both. Either one. Or both. All right. So, I don't play board games as much as I would like to because board games require organizing people in real life. And I'm [laughter] 51 and I can't get enough people organized around me to play board games. But I like the idea of them. I play Call of Duty's zombies on the treadmill walking at over three miles an hour. Great idea. That's my That's my exercise routine is uh zone out, shoot zombies in the face, and walk on the treadmill. That's such a good idea. I need to do that. I need you. You can do the 12% and really 12% incline. You really work up a sweat at 3 miles an hour and 12%. But I tell you what, if you go on mic on the game, people are going to be like, "Are you sitting on a washing machine? [laughter] What the hell is that noise?" Like, "I'm on a treadmill."
No, but that's a it's it's much harder for me to convince my wife that I need to play video games than it is that I need to go walk on the treadmill. So, that's a correct. It's time for me to get my exercise, right? So, yeah, everybody can get on board with that. And um yeah, you can't drink bourbon while you do it. That's the only downside. I mean, you could, but it's not very Yeah. So, I guess with that, I was thinking I that was gonna be my my question. What's your favorite whiskey or bourbon? So, you know, it's a little bit like our discussion of continuums. Whiskey and bourbon has a continuum, right? There's one there's one bourbon because you're from Kentucky and Yeah, I am from So, even bourbon, right? You got you got a whole whole continuum. Uh I was just discussing with Scotch is gross. Let's be honest. Scotch is gross, right? Yeah. That's taken me a while to get used. I had to go to Europe a bunch to, you know, learn to appreciate scotch. And really you you don't like Pete. It's not that you don't like scotch. It's 1,000%. It's the Pete. Yeah, I know exactly what that face means. There's And there's a bunch of scotches that aren't super pey. So anyway, on the on the um the whiskey side, uh the pepper has some uh rye. That's not a bourbon, obviously. It's rye. Rye whiskey. Um the the the pepper rise have been uh a punch above their weight in my opinion. And I'm also of the opinion that that everybody needs three bottles of bourbon. They need a house bourbon and that's like a well bourbon. That's something like when somebody comes over the house and they want to pour it on Coke, you don't cry. Yeah. Hydrogen beam. Yeah. Jimact right. Yeah. And and actually mine is Jim Bean Black. Yeah. Okay. Um you know it's like 28 bucks a fifth or something in Virginia. and I'll drink it on the rocks or you can uh you know if somebody wants to pour it over Coke, I don't care. Uh you know, but if somebody poured Papy Van Winkle over Coke, I'd punch him in the face. [laughter] I've got a fun story about that. Uh I had a had a buddy help me move. I I randomly stumbled across the only bottle of Papy I've ever seen in a liquor store outside of a bar. Um right when I was I've never seen it in New York. I had never seen it either. I almost walked out without buying it and I walked to the door and I was like, I've literally never seen this. I'm an idiot if I don't buy this. And so, uh, anyway, I had I' I've only drank it my both my kids were born when I got my masters and then I've drank it on one or two other occasions. I had a buddy helped me move to the house I was telling you about. And uh, you know, that was like one of the last things we moved was the liquor cabinet, right? So, and it was in the car because it's delicate, all that stuff. So, we were loading it up and I was pulled out the bottle of Peppy and I was like, "Oh, well, here we'll we'll have a drink now that we're done, you know, as a thank you or whatever." And I poured already. Well, I just poured them into, you know, the highball glasses. I didn't have a fridge at that point, so that it was just warm whiskey. So, he just threw it back. I was like, "No." Did it like a shot? Oh, I thought he was going to I thought he was going to pour Coke alas, but he just shot it. I was like, "That was like $50 or whatever it was." Yeah. And so that I was just like, man. But he was like, it's the best shot I've ever had of whiskey. I was like, like, oh, that tasted really nice. You know, you wouldn't know. You just shot it. So, yeah. [laughter] So, I think you need a you need a house like a a well a well bourbon. And then you need something for special occasions. And then you need something with a story. Yeah. And your special occasion may be a story, but like my uh I I really like stories. So, my my bourbon uh with a story is Green River, and it's because my ancestors came down the Green River from East Kentucky into West Kentucky and uh with all their possessions on flatbo uh coming down the Green River. And they got down there and they had family or friends or something, but they didn't have a house yet. So, their belongings went in a barn. And as it turned out, is 1860s. And some folks didn't like their neighbors uh during that time. Uh and somebody's barn ended up getting torched and half their belongings burned. And my parents still have a press piece of f wood furniture in the house that has charring on discoloration on one side and it's because it was in that barn that came down the Green River and was in a barn and got charred uh during the Civil War. So the Green River is and and because of John Prime's song Paradise that talks about going, you know, won't you take me um back to Paradise City down on the Green River where Paradise lay and the first time I cried during COVID was the night John Prine died. Yeah. And I just all of the stress of like the COVID situation, I just sat in my floor and melted into a puddle the night John Prime died. And so that bourbon reminds me of my ancestors and John Prime and it that's really cool. Yeah. So then and have something else top shelf and all you need to do is pick top shelf is just pay till it hurts and then you [laughter] That's easy. The one with the story is hard. Yeah. Well then it then it becomes a whole another ball can of worms where it's like well I've got this really expensive bottle. When am I going to drink it? Like you feel bad at times if you like consider drinking it because you're like, "Well, is this special enough?" And you're like, "Well, wait a minute. I bought this to enjoy it, so who gives a shit?" Have you seen the movie Sideways where they talk about this with wine? Not in a long time. But at the end of it, he basically has a styrofoam cup and he's got the bottle of special wine with him and he's in like the the the fast food restaurant. He's pouring wine into his, you know, styrofoam cup. And And you know, I'm kind of of that mindset of I don't keep stuff like that too precious too long. Yeah. Because it's the save it up a little bit enough so it's special, but then have somebody over and celebrate whenever something's appropriate. Find a reason to celebrate. You want to wrap us up, Bobby? One more. Yeah, let's do uh one more. I'm not good at I'm not good at rapid. No, it's okay. [laughter] No, it's all good. It's good stuff. Um, all right. You know, you're a pretty big open source guy. What's your favorite open source project? Favorite opensource project? Oh gosh. Uh, there's so there's so many to choose from, right? Because I I spent all so much time in the R ecosystem. Yeah. And I and I am very much in the I like like the what what Hadley did with the Tidyiverse and R that whole ecosystem, but I also get why some of like the folks who are more core R find it frustrating or don't like his approach or whatever. So I see both things going on there. But I think the um the the dlier libraries within the library within the tidyiverse did a whole bunch of stuff with data manipulation. I mean the idea behind it is it was going to be a a grammar of data manipulation in the same way that ggplot is a grammar of graphics. And I would say Headley and folks involved in that like like succeeded. And I still, you know, I use probably I use Python absolutely as much as I use R anymore. And I constantly miss Dlier. Oh yeah. I'd say if I get a chance to use R tidyverse, but to your point mainly the dlier, you know, it's it's such a better it's so much better than pandas. I'm sorry. It's just well the interface is so intuitive to me and a a bunch of stupid stuff that I shouldn't care about but like I get worn out typing my quotes in um that stuff matters when you have to use it a lot. It's just like in Python just you know quote comma quote comma and it just wears me out and and I just love how fast I can go in dlier because I just type fewer quotes and you know a number and I and I also have the advantage of I used it long enough that I I almost think in it so it's very intuitive and I find myself sometimes writing SQL like why the hell is that an error and realize oh yeah you you got to put the wear condition relative to the group by [ __ ] I forget about because it doesn't matter in direly right you just you just as you think about it. You put it in as long as your group buys above your aggregate, nobody cares. And SQL is a little more finicky. [laughter] I think that's a perfect perfect way to wrap it up, JD. Thank you so much. Fantastic, y'all. It's been uh it's been a pleasure. I enjoyed. No, likewise. I mean, so glad I could have you on or we could have you on. We'll definitely [snorts] have you on again and get get even deeper into the the our verse because there's just so many people in the industry that use it, right? And it's it's so powerful. Exactly. Thanks so much for your time, man. We really appreciate it. You bet, guys. Thank you. Yeah.