Low pain tolerance: Engineering efficiency & accounting's future
Download MP3You're listening to the Breakeven Brothers podcast, a show by two brothers shooting the breeze about business, finance, technology, and everything in between. All views and opinions express are solely our own and not affiliated with any third party. Hope you enjoy the episode. If you do, leave us a review and drop a like. See you there. What's up everybody? Welcome back to the Break Even Brothers podcast, this is episode 16. Joining me, as always, is Bradley Bernard. Brad, how's everything going? Pretty darn good. Got a renewed kind of energy and focus going into February. So I think we had talked about it in the last pod, but let's just say lots of things are in progress, making good progress, firing on all fronts. Yeah, how about you? Yeah, good. Yeah. busy with with work and you know we've had some travel out of town um had some guests over including yourself um and so I've been super busy with just travel and work so I think this week is kind of getting back work is still busy but kind of getting back and just like being in the house getting settled again and kind of like not having any travel any real travel planned until June maybe that might change and maybe we'll get a little restless but like nothing planned which is kind of nice to be honest because there's been, you know, with the holidays and there was, you know, this travel coming out here, coming to California last week. So it's just kind of nice to settle down a little bit and lock in, focus, you know. It's good to get some stuff done. Everybody's been asking me. And so I'll update the podcast. So I went out to visit Ben. We had kind of talked big shop on Larerville in our last podcast to make that happen. sadly reporting back that we did not get the time and energy to set up his windows, a Laravel experience like I had dreamed of. I probably only flew out there just to set that up and it didn't happen. So I apologize to all the viewers who are asked, but I do have one success story. So we went out and visited Ben and my dad also came along for Super Bowl weekend. Parents came along. And at that time, my dad had kind of picked up Ph.P. picked up Larval. He's accustomed with the tooling and the infrastructure. But all this AI coding stuff has happened recently, and I just wanted to get him onboarded onto the new era of coding. And a little tough, you know, old-timer set in his ways, but I can report back with a kind of an anecdote of the experience. So I kind of took over his computer, set up, you know, cursor, the AI IDE, warp, another AI terminal, and then installed a phone. And then installed a fresh Larravel knew on his MacBook and then installed Orbstack, the replacement for Docker. So basically just went through each of his tools, upgraded them, installed Larravel and got him going. Deployed on Larval Forge, great software, super awesome. At the end of the day, I wanted to get our dad set up such that he could actually kind of participate in the vibe coding. So we talked about it last podcast, but this was essentially describing what changes he wanted, letting cursor code it for you. And I'm on tech Twitter. There's so many people that say, oh, this is the future. But not every day do I get to kind of show it to somebody when I already know what it's like and kind of see that spark of excitement. And so I showed it to him. I said, hey, you can just describe what you want. Like you want to make a layer of off with this? Literally just type it in, click enter and go. And I did it right in front of them. I know a simple laraville query, asked it to build something. Cursor knocked it out of the park. I think it was integrating with Twilio was what he wanted initially. And it suggested to install packages, hooked up routes, hooked up controllers, models, migrations. All the Larravel kind of its bits and pieces all hooked up. And then I just kind of looked at his reaction. He was shocked. I was like, that is powerful. And he was really excited about it. So I got them all set up on the tools. I got his Larval app using push to deploy. So he can commit on Git, push it to GitHub. Lareval Forge will pick that up and deploy it. And pretty much the best kind of feedback loop for building fast. And I had texted him probably the day after he had left saying, you know, how's it going? Like, are you enjoying it? And he's like, yes, it's the best thing ever. And so just a reporting back on my Leraville expeditions at Ben's place, we had one failure and one success story. And we'll look past the failure because the success story was eye-opening and kind of a success story for Curse or for Lerville. Larval and for the old timeers picking up new tools. Wow. Yeah. And if you click the link, you'll get a little discount on cursor from Brad. Yeah. No, you guys, yeah, it's cool that you guys all set up. I will say, just for the record, too, you guys were back there setting up for a long time. You know, so that was always my apprehension with Larval. It's a lot of setup. It's a lot of learning what you got to do and not do you guys are there for a little bit. But, you know, it sounds like it was enjoyable. and that's good. Good to see, you know, it's always nice when you show someone something or you introduce someone to something and they enjoy it. Very nice. Yeah, Larval. I'll get you next time. Don't worry. I mean, we'll put a plug out there for Curseur for Ford. You know, link in the show notes of my referral code. But to be completely honest, it was really nice to give him the power that I feel like I've been kind of leveraging over these past few months. And, yeah, his reaction was like really, really positive. And I was like, yeah, once you've been coding for many, many years and you see a tool that does kind of the boring stuff for you, like just kind of a light switch I could tell. And he was like, all right, I'm going to use this. And I think it's a little bit easier for newer projects. So, you know, kind of the best case scenario, but still powerful nonetheless. Yeah, yeah. And just all that, like, the low-hanging fruit. You know, I've used it for like Django applications and just like getting the URL route set up and like getting the templates, the HTML templates. set up when you're just like making an MVP just like give me a basic profile page like give me a basic you know like dashboard page you know like those things that are common across like many web apps you know you don't have to spend all the time of like here's my nav bar here's you know and kind of like defining all that it kind of just gives it to you so yeah it's pretty nice and I haven't used it nearly as much as I would like to but in the bits and pieces you know we talked a lot on the show about it it and it is pretty sweet just especially to see it work you know you kind of see it code out there for you. It's not like you say it and it goes behind a screen then it gives you it. Like you can see it almost like right as you as you're typing or if you were, you know, doing the microphone or the voice approach, you know, kind of types it as you speak. It's just pretty cool. Yeah. So I'm officially one and one for my Layerville experience. We'll make that a two and one in the future when I get Benny out out there over to my place. We'll figure it out. I'll box them in and then we'll get it done. I'll report back. Yeah. Okay. Okay. Yeah. I think next time we'll have to come up to up to the Bay Area since you guys just came out here. Yeah, it would be fun. Yeah, I don't know. I don't know what to make a plan for that. Cool. Well, so I guess just to say what you guys were here, or partly why you were here, besides just learning or getting everyone caught up on Laraville, was to watch Super Bowl. And always during the Super Bowl, it's a contentious time for the accounting industry because TurboTax is always advertising hard during the Super Bowl. and the reason why that's contentious, and I'm kind of curious, you know, I'll be curious to see if there's anything in your world that kind of is similar, but TurboTax has pushed a lot of people's buttons in the past because, you know, they will, they have a product called QuickBooks Online, which is like the gold standard for like, you know, like general ledger systems for like small businesses. You know, if you have a small business, you want to get set up in your accounting system. QuickBooks Online is like what everyone uses. it's very, you know, there's a learning curve, but overall it's, it's pretty simple once you kind of get started. They've done a good job with that product. But, you know, they kind of rely on, like, you know, advisors or people that are experts in that product to, like, you know, push that out to their clients. So, like, that would be accountants. That'd be CPAs, right? You know, there's CPAs and a lot accountants out there that will only work with you if you're on QuickBooks because they don't want to go learn, you know, another one called zero or they don't want to go learn, you know, NetSuite. And so, you know, the turbotax into it, I guess I should stop saying turbotax into it, which kind of has both of those products. They make this product that's, you know, kind of for accountants and helps with all the accounting. But on the same token, they keep, especially with TurboTax, they keep kind of bad-mouthing accountants. And they say, like, hey, we'll beat your last preparer's return price. You know, you don't need to pay your CPA, like whatever, $500. You can do yours for $80 here and talk to a. you know, like a professional if you need help kind of thing. And so they keep kind of undercutting on price pretty hard. And I think it was so bad where I think this is, this wasn't during the Super Bowl, but this was probably a month or two ago, maybe three months ago. I think they ran a program or a marketing campaign of like, fire your tax preparer or something like that, or fire your CPA or something. Really? Wow. Yeah, it was pretty aggressive. That's aggressive, yeah. It even went up to like, there was like the American Association of like tax professionals. something like that. I don't know, I don't remember the exact association, but there was some association that came out and was like, we do not like this. I don't know. I don't know if they actually had like any like ammo behind that, but, you know, they came out and said, like made a statement that they didn't like it. So I saw on LinkedIn, there's a couple of people posting, like, you know, for all the accounts out there, when you see the turbo tax ads during Super Bowl, like, don't get triggered. Like, you know, kind of like everyone, everyone knows it's coming. Because, you know, it's obviously great timing for TurboTax because everyone has to do their taxes, right. But yeah, so I did see the Super Bowl ads for TurboTax. I don't remember them being too bad. I think the one like fire or CPA was definitely the worst I've seen. I think they got some good pushback on that. But yeah, it is interesting to see a company that has a product that kind of relies on like experts or pro advisors. They mean, they call people pro advisors if you pass a certain number of qualifications like exams to use the system. But then they, you know, in the same breath kind of, talk bad about, you know, a certain faction of, of, you know, tax repairs. So it's interesting, you know, and always, you know, people talk about switching to a different one, you know, switching to zero, switching to NetSuite. You know, our dad, he's using Odo, which I'm actually going to hit on a little bit later. But, you know, QuickBooks is right now, it is the best one, in my experience. And so it's hard. You kind of just deal with their shenanigans a little bit and you just, you know. Yeah. I tried them a little bit when I was doing Bucksbyte full time, and I think I backed out pretty quick. But in general, I think the TurboTax or I guess Intuit software collection of products, I for one, I'm very ready for any newcomers to come in and take that space. I think usually what I do now, I just got my W2 in. I usually try to run it through TurboTax, see what's, you know, how much pain and suffering I'm going to have. then take it to Credit Karma who does free filing. So no fees, where TurboTax, you know, lets you get all the way to the end. And then you have to E-file for a fee. And I don't know, I double check it with TurboTax because I'm just used to their software. But I am more than ready for someone to come disrupt that space and just make it way, way easier. I don't know, AIify it, but it just feels like that whole software space is so lobbied and hard to get anything off the ground. So, yeah, I think for the Super Bowl ads, I don't remember what theirs was or if I saw it, but I did see Open AI's ad, which I thought it had this kind of 50-50 reaction. People were like, well, this is so good. And then I think the other half, which was probably what I was leaning towards was I saw their logo appear in the beginning. And I thought, oh, this is an open AI ad. But then the next 45 seconds, I was confused and thought, oh, maybe this isn't because I don't know what's going on here. And then at the end, it kind of wrapped up with like, you can do anything with AI. And in my back of my head, I was like, just show me what I can do, or not this. this kind of pixel art parade of 45 seconds. So, yeah, some interesting ads. There was also like a tongue one with food or something that was pretty funny. Yeah, yeah. And pretty odd, you know, for sure, but funny as well. So a nice addition to the Super Bowl is being plugged into those, which, you know, is always quite entertaining. Yeah. You know, TurboTax is a fascinating product because, you know, it's basically like a, they do a really good job of like putting like a certain like marketing or certain like I guess aesthetic for lack of a better word on what's really just like a really complicated checklist you know for the most part. I guess just you know I'm sure you've used or like experience with like type form where it's like you can create like funnels where you can click here then you go to this funnel if you click over here then you go to this funnel and so it's basically I think you know if you look under the hood I think it's basically it's like a giant you know click-through thing, but they ask good questions. They have nice kind of, you're moving along through the process. So, you know, I think it's, you know, there's so many great products out there, but, you know, a lot of times they're not doing anything super amazing. They're just, you know, they make it easy, you know. And I think that's kind of what the Super Bowl ad was. I can't remember the actresses name. There was an actress that, I think, was like kind of, you know, hosting it or featured in that ad. And I think it was about, like, doing things the hard way and TurboTax It doesn't, you know, it's easier, you know, which is, which I think is very fair because I think they, you know, for maybe other shortcomings that the program and the product does have, it is pretty simple. Like if you got a simple W2, like, you know, and not too much else, like I would do turbotax. You know, I mean, like that's, that's all you got. And that and not to disparage that, but just, you know, if it's a simple return, then just use something like turbotax. Because it's pretty quick, you can get done in and out pretty fast. And yeah, it's pretty self-explanet. but I think if you got a more complicated return, you know, maybe you have some, you know, kind of like real estate investments or you got like another side business, something like that, I wouldn't use turbo tax. You know, and they even have like, like, um, like you can like add on like audit protection and stuff like that. You know, good luck with that. Yeah. I don't, I turn all those off. I'm like skip skip, you know, whatever. I can't imagine. I think you are better off just, you know, if the IRS comes at you and you say, yo, we'll have auto protection on. I don't, I've never heard anyone be like, oh, that helped me get out of a bind with IRS. I don't think it's just from my own anecdotes. Yeah, it's almost like insurance. Like they just kind of sell you that for an extra 50 or so dollars. But yeah, I don't know. Yeah. So we'll see. I got to do my taxes, but I thought I'd plug the TurboTax ad because it's always a big moment for the accounting industry is kind of seeing what they'll come up with next, you know. So it's cool though. And then one thing I realized near the end of the Super Bowl was the Chipotle tweets. thing was happening again. So that was kind of them tweeting out a code when something extra. I think that's what they had categorized it as something extra happens in the Super Bowl. And so my wife reminded me, she said, oh, like, don't you have this Chipotle bot that will, like, scan their Twitter and do all that fun stuff? And if you know me, automating things is like my, my passion. I could do that for days and days. And I completely blanked. And so I think it was maybe third quarter, I pulled out my computer, refreshed the Mac out. that I'd written to essentially scrape the Twitter account, take whatever the Chipotle tweets, Twitter account tweets, and throw it to AI and ask AI to extract the kind of text code to text their number. So you text their short code, SMS number, whatever appears in their Twitter, and you get free Chipotle. And I started running it. It completely failed because I had hard coded it to the, what was it, March Madness, I think, section of games. And their code started with like, free for free as a prefix. So I like hardcoded my program to search for that and do a bunch of extra checks to make that work. So when I started running it for the Super Bowl, I thought, oh, I could just turn it on and it'll work. And the first time an event comes in that they tweet out a code, I'm like, all right, I'm going to sit back and get for each full a whole way. Nope. It breaks. It doesn't work. And it's crap. Like, let me go fix this. And then there's probably five or six spots that I thought I fixed it. I run it. It detects an event. I go try to get the code fails again. And so then I got really frustrated and kind of gave up. But then I was like, you know, I have time. It's not the end of the game yet. I'm sure usually Chipotle does like more codes at the end for whatever reason just to keep people around. And so I finally got to polish stuff and working. And then I realized I'm in an AT&T dead zone. So this was heartbreak. So I probably spent 15 minutes tinkering with it like stress coding, trying to get it done, half stress, half fun. once I had got it working. So, you know, Chipoli tweets, I take that tweet, throw it to AI, thanks. Google Gemini Flash, you're great. Basically extracts out the data that I need, pass it to the MacOS messages app, and basically auto-texts. But when I send the text, it just sits there with a loading bar at the top. And I'm like, is this my code? What's going on? I worked on my computer at home. And I asked Ben, I said, you know, what the heck? Like, what's going on here? And what's the reality? Yeah, AT&T is horrible in my quadrant of the neighborhood. Yeah. And I didn't even think about it because we switched to Verizon for that reason. Like literally, we moved into where we are now. Yeah, it was so bad. Even if we have Wi-Fi calling on, it doesn't matter. It was horrible. I called them and they were like oh yeah we will reset your network cache I like that Like this will work now And they did that and it was still horrible And I was just like you guys I just don think you have the tower to hit this I don't know. You know, and I live in a very normal area. Like, I'm not out in the boondocks, you know. I just eventually said enough for switching to Verizon. It's been good ever since. So when, you know, you, I think you and, you and your wife are sitting on the couch. And I think you guys were like, I don't know how you said it. But once you kind of like, because I saw you, you were working on it. I didn't think twice about you working on it. I was just like, oh, yeah, you'll just send the text and go through. And then once you were like, oh, does AT&T like not go here? I was like, oh, that's right. Yeah, yeah. No, it does. It is a battle of milliseconds in the automation game for these tweets. And I was like, once I saw the loading bar, I was like, my code is messed up. Then I thought my code is doing everything it needs to do. Sending a text message is not my responsibility for the code. That's just like Apple ecosystem plus, you know, cellular network. And then that's when I asked Ben. He told me the sad news, and I kind of just closed my laptop and said, all right, there's no chance I get a free Chololi burrito or whatever they're offering in this cell network. So, yeah, hopefully I'll get it next time. The program now is a little bit more generalized. So whatever Chocolay comes up with next, I won't have hard-coded to free for free as much. So that's the good news. Bad news is wasted effort, learn AT&T sucks out there. For others who suffer, it sounds like switching carriers is the solution. solution if you're in a dead zone. Yeah. You know, I went into Costco not that long ago. And sometimes they have like, you know, a phone booth there, like an AT&T booth or a T-Mobile booth. And it was AT&T. And I had just switched to Verizon. And I didn't say, I wasn't going to be an A-hole. But they were like, oh, like, what care you do you have? Just as I was walking by. And I was like, I ain't going back to you guys. Hell no. Hell no. I, we stuck it out. Like, anyone who, like, when we first moved here, anyone who I talked on the phone would like definitely tell you that, like, It was just impossible. I sound like a robot, like that you couldn't hear me for long periods of time. And it just became unsafe because it was like, what if, you know, my wife like needs me? I literally cannot be reached on this island I'm on. So, yeah, it's like that. But, you know, to be fair, you'll have March Bandis coming up here soon. So you'll be able to get. That's true. Yeah, that's true. Not far away. You'll be able to dust it off again. It kind of reminds me, like, when I'm moving into new spots, I'll definitely check the internet provider and see if AT&T fiber or other high-quality, the internet service providers exist because I don't think there is kind of an AT&T coverage map that's probably accurate. I'm sure if you pull up your address on one of these maps, it probably says it's fine. Then when you get there, it's just messed up from one reason or another. So, yeah, maybe they should have a real kind of map that shows like the dead zones because having internet is critical, but cellular is definitely the second most critical service that one should assume to be working, especially for a company like AT&T. So. I feel your pain that stuff. Yeah. Yeah. I feel like the cell companies would like sue to like take down the coverage map, like the true coverage maps, you know, like no, no, we're good there. Get out of here. We're totally strong there. One quick thing before we move on. I have your guys as pottery and I will mail it to you because we didn't get to bring it to mom and dad's, but I'll mail it to you guys. Actually, I haven't looked at yours, but mine looks wonderful. Good enough. Just to say that, yeah. But yeah, when you get yours, when I mail it, And you get it, we'll feature it on the pod. We'll do both bars, you know. Yeah, we'll do a reveal. Yeah, yeah. Yeah, when they were out last, last two weekends ago, we did some pottery painting with the kids. So it was fun. So, yeah, we'll mail that out to you. Yeah, we talked about merch. This might be our first merch and it might be unique. So we won't discuss price, but let's just say it's up there. Yeah, it's got to be. I mean, it's one of one. So, yeah. You know, yeah, it's like an NFT, you know. It is. Anyways, on Twitter, I was scrolling like I usually do. And honestly, I've been kind of off Twitter, but been trying to find things probably every few days. One thing I saw recently was from the Pragmatic Engineer, so probably a leading software engineering. I don't know if you call it journal editor. Essentially, it's one guy behind it, but he kind of goes behind the Pragmatic Engineer. Anyways, lots of VC companies are popping up for AI-funded software solutions. And he had posted a tweet outlining what back-end languages are kind of hot for these VC companies that they're looking to hire for. And so kind of looking at his tweet, the top one, typescript. Pretty, pretty common. You know, JavaScript is very, very common. TypeScript just types on top of it. Can't really hate it. Second, Python, which to me is a little bit surprising. I find Python in like the ML space, but not really the web or backend space. Third one is Go. I think it's created by Google. It's very efficient kind of back-end server language. I think it's compiled to. I haven't really ever touched Go, but I've heard lots of good things about it. Fourth is Rust. This one's heavily, heavily used in the crypto space. I do know Solana uses Rust. I think Ethereum potentially uses Rust, not too sure. So a specific sector of companies use that. And then the last one is the list is Ruby, which, We love PHP. Super sad not to see it on the list. I think, to be completely honest, the vibes are changing, so I expect to see it next year. But we can take solace that Ruby is there because Ruby on Rails is a distant cousin of Larravel. They're very similar. Batteries included framework. And so we can almost include PHP as number six, seeing that Ruby is number five. So pretty interesting to me. I think I imagine JavaScript Java. I'm surprised not to see Java on there. I guess, but seeing things like Ruby, Rust, and Go, like you can really tell that Rust and Go are rising in popularity given its open source packages, performance. I feel like people are spending a lot more time to rewrite old technologies in Rust to make it like, oh, this is the better version, the more modern version. Like we saw that with Pip turning into UV, a bunch of other tools. And so pretty interesting. I don't know if anyone's in the job market right now, but if you're looking to join kind of the VC funded companies, I would probably focus on this list, you know, top three TypeScript, Python, and Go. Just because there's a decent amount of jobs out there for engineering. And as a back in engineer, you don't have to touch UI, kind of a nice spot to be. And yeah, these are the most popular. Yeah, UI sucks. I hate working with UI. It's just because it's hard. You could obsess over like how rounded the button should be and like how much margin should be in the button. You know what I mean? Like just stuff like that. It never ends. No, yeah. I'm curious about, because it says in the Python one I'm looking at the tweet, it says usually with another language like TypeScript or Go, I wonder like how that works with Go. Because TypeScript is like JavaScript, right? It's just like a flavor of JavaScript essentially. Yeah, yeah. So like that makes sense. Yeah, I'm just, I guess, thinking out of I'm just curious how you'd pair Python and Go like where that comes into play. Yeah, I don't know. My guess is maybe TypeScript generates types. Pipes and pipe. I don't know. Actually, that doesn't make any sense. So I'll stop there. But, yeah. Well, I mean, I certainly believe the source. But yeah, I'm just, I'm curious. I think, yeah, and Go is a compiled language. I'm pretty sure. It's funny. I remember seeing years and maybe like this is probably 2017, 2016, maybe 2018. There was like the rise of like, I don't know, I'll say rise. It sounds dramatic. But like more and more popularity of like, like, like, quants, like, you know, the financial quant engineers, you know. And where I worked a while back, we had, like, a department that, like, actually, like, was that. Like, they did that. They were trading mortgage-backed securities and they were using, like, code. It was actually pretty cool. Some of my first, like, experience with Python, like, in a working setting. But I remember, like, that being just that, like, that job becoming more and more common, like, reading more about that. And I remember, like, Java was, like, I think in the mix there. for a while. I think I don't remember the exact quotes. It's been so long, but I feel like I remember reading like a Goldman Sachs or an article about like Goldman Sachs engineers like using Java to like automate their trading. And it was, you know, I didn't think anything twice about at the time. I was like, oh yeah, Java that's very, very popular. When I was in school, I took a couple of classes in computer science and Java was what we used, I think in the majority of the classes. But yeah, now I don't ever hear about it. So it's interesting, and, you know, I'm not in that space, but like, it's interesting to see on this list that they said they don't get any requests for Java, essentially. Yeah, I think all these VC companies are trying to hire, like, the sexy languages to get people in. I feel like Java is not sexy, although I do know it has good market share for enterprise software. So a few companies that I've been at have used Java for back end and boring and verbose, but I think it has a solid system around it. given any large company, try to migrate from one language to another when things are working, good luck with that. I think, yeah, I totally hear that point. I think the QuickBooks, speaking of Intuit and QuickBooks, I think their SDK is Java. You know what I mean? Like, I'm trying to find it really fast. I can't find it. I'll move on. But I remember there being like, because there's like the API obviously, but then when they had, desktop they had like an actual SDK and now I can't find it but um two two more call-outs from this article while I will I have it open and then we can move on the AI startups in this DC article are embracing AI and looking for devs who are almost exclusively using copilot cursor etc so study back end and get used to cursor and then the second note which I thought was quite interesting as someone who's in the mobile space is the demand for native mobile engineers has plummeted, which is pretty interesting. You know, writing a Swift UI app or Apple, like quote unquote native app and Swift or Objective C, we're not seeing that much at all for VC hiring. We're seeing React Native be the kind of end-all be-all. I think it kind of ties back to TypeScript being a very useful language, front-end and back end-end, React Native giving you one codebase for iOS and Android and even web these days. So it all makes sense to have less people and do more with less people given use AI, use React Native. But I do think it'll come to a point where native mobile engineering requires native at scale to have the best performance. While React Native has gotten a lot better, it's not perfect. So if you're a mobile engineer like me, don't be too scared if you see a headline like that that, you know, zero engineer, zero demand for native engineers just because companies are still hiring iOS and Android specific platform engineers and you're not going anywhere. But it is nice to work with things that, you know, VC companies are at least hiring for. So if you haven't touched React Native, maybe just try a little demo app and cursor because it gets pretty far. Yeah. And at the end of the day, you know, it's probably when you have so many applicants for jobs, which just seems like that's the case now, there's way more applicants than jobs available. So, I mean, I get why, like, having proficiency in one language would kind of make one candidate stand out above others that don't have that proficiency. But, you know, at the end of the day, like, so that I don't want to, like, neglect how important that is because you're just battling against another thousand applications, really, especially at these, like, fang or mango, whatever they're called now, companies. You know, you're battling against the numbers. And so, like, any single leg up you can get is certainly in your favor. but like, you know, once you're there, like, if you had the opportunity to really, like, work on something, like, it's not about, like, specific language, you know, it's like, what can you do? You know what I mean? Like, what can you do? Is that valuable? You know, I'm sure if you are a really competent engineer, you can pick up some of the nuance of typescript, easy, you know, it's not going to be the end of the world to move around. So, yeah, I mean, obviously, like, Russ is way different than TypeScript. I have no really work in either of them, but, like, just from what I, you know, understand. and no. So there's limitations to that. But in the day, like, if you're a backend engineer, you know, I'm sure some of the same concepts that you, you know, deal with, you can deal with node or you can deal with them in, you know, in Python or whatever, you know, it's not going to be. I see it as like, kind of like you mentioned, an accountant wanting to work with QuickBooks and not other stacks. It's like for the VC companies, they're looking for, you know, a Python, jango engineer, not because other people can't learn it. It's just kind of like, you're off for the races. And that exists for React. We want you to have experience with React, and that might be all you know, but you know it damn well. And that is like a sharp tool in the arsenal for this company. So that's probably the way I would read it. Yeah, totally. Yeah, there is a company in the mortgage space, I think it's called OptaFunder, and they post on LinkedIn occasionally, like job postings. And I think they are a Django house. So when I see that, I'm just like, that's right. That's right. I do know, I've seen Zillow job postings for Leravel, which is quite unique. I think, I think even it was posting the Leravel community maybe a few weeks ago saying, oh my God, Zillow hires Leravel engineers for a healthy salary and benefits. So if you're looking for a Leravel role outside of the Leraville core team, Zillow might be a good spot. Yeah, yeah, check it out. Plug. Yeah, thank you. Cool. That's exciting. One thing I want to talk about is technology switching. And this is near and dear because I'm doing this in two different places right now. And so for the first one, for my regular day job, we are, without getting so in the weeds that I lose everybody, you know, we have an annual process that we do. And we used to just do this with more or less like Microsoft Word and or like Google Docs, basically, where everyone's reviewing a doc, there's multiple edits back and forth. It's easy to lose things in translation when that happens or to kind of have version control issues. It's hard to like tie out or like, you know, notate, like where certain numbers are coming from. So all around just wasn't an efficient system. And, you know, everyone kind of knew that, of course. But sometimes it's hard to step out of the hamster wheel and like fix something, you know. You got to kind of have that time and you kind of know you're going to have that upfront pain to work with that. and get through that. But, you know, we are in the process of doing that. And, you know, it's already a lot better. Like, it's, you know, it's one of those things where, like, you've got to get over the learning curve and you've got to kind of embrace. That's going to be, you know, clunky at first. And if you let that distract or, you know, discourage you, you're just, you're not going to get anywhere with anything new, to be honest. So you got to kind of get through those bumps. But, you know, I'm already kind of starting to see, like, it's not even done yet. I'm kind of starting to see that at the end of the tunnel. And it's like, oh, my God, I can't believe we did it at the old. way, you know, just already, I'm like, I can't move how we did it before. It was so much more ineffective. We got the job done, but it was way more ineffective. So, it would be nice to kind of estimated time savings on any processes or tooling switches. You know, I would have to go back and do the math. I'd want to be really that precise on that. Maybe it's the account in me, but like an actual ROI because it doesn't just involve my time. It involves like other people's times as well. Like, you know, at end of the day, um, I think this is a great, a great segue. It's what I like to think about when I'm working. And this is just true for, um, you know, anything that I'm doing is like, if you can impact not just like your own like desk, but like, you know, your boss's desk or your team's desk, like how, how their work is done or if you make their life better, that is such a game changer. You know, I always, you know, make sure people on like my team or people that I work with, like, when they do something that improves their job or makes them more efficient, that's great. Like absolutely, that's amazing. Go for it. And hopefully, you know, you have more time to do like, you know, bigger and better things or, you know, get experience and new things. But when they do something that goes beyond, like, just that, like, it impacts, you know, downstream teams or impacts, you know, me and my boss or, you know, when you do something that goes beyond just your immediate scope, that's like where I feel like people will be like, oh, that person's so valuable. or like, you know, that work is so appreciated, you know, and it's hard to do that sometimes because, you know, you kind of have to think about what their problems are, not just what your own are. And you don't always know that. Like, you don't always know what their problems might be until you kind of go start asking or, you know, kind of get in conversation with them and you hear what their pain points are. So, yeah, this is one where, you know, I think there's multiple teams involved. I mean, so I'm happy to be part of it. We're not done. Again, we're still in the process. So I'm not, I don't want to get too far ahead of ourselves. But yeah, just it feels good to make other people's, you know, job or other people's, like, work easier. You know what I mean? I think that's just a good philosophy to go by, generally speaking. I'm a big fan of that, too. I think when I onboard onto companies, I'm always trying to make engineering processes more efficient, where that's like documentation, internal dev tooling, for mobile, for back-end, like, literally whatever is a pain in my life. it's probably a pain for the 100 other plus engineers working on the same project. And so half the battle is finding that time to make a solution. The other half of the battle is getting, you know, internal adoption and convincing, you know, key stakeholders to get it to work. But part of that is getting enough data to understand you know that ROI like how much time are you saving for one person times 100 engineers you know times one month Like kind of a cool number to say oh I save this much time But I absolutely for that too And I think at times I have to like, like, pull myself back because I'm like, oh, while I'm doing something, I found this to really suck. And I want to go kind of fix that. But then I'm like, oh, that's going to take a while to get to the right solution. and I think people have coined the term online a low pain tolerance for like engineering inefficiencies. I think I adopt to that mantra too where I know things can be better and I'm very subscribed in making them better. And if there's anything that I feel kind of, I wouldn't even call it like picky, but it more just like forward thinking of, oh, this sucks. Like how do we make it better? It's really easy to complain about things. It's hard to be the person that complains about it and then finds a solution and then creates it and then spreads it. That's like the, you know, the multiplier for organizations that people wish they had more of. So I'm definitely subscribed for that. And I wonder on the accounting side, like we have developer productivity folks at all the companies I've worked at, which are, you know, focused on how engineers get their job done, how to make them more efficient, building internal tools for that. Is there something like accounting productivity to you? I mean, maybe it sounds weird from your end, but for me, I imagine that could be a lever to just make people more productive, and that's the full team's focus. Yeah. Well, I see that in like two different venues. So, like, I would say like at like regular large, you know, public-traded companies or just large private companies, um, there is a newer role that has become more, more popular. Um, and it's like financial, gosh, I can't remember what they're called exactly. like finance transformations manager or like financial operations manager something that affects where it's not like they're doing accounting but they're in the finance org and they help make things better more streamlined is the idea so that is cropped up a bit and I think generally speaking they're doing just that where they're looking at like oh you guys are using Excel to like manage your clothes like a secure calendar like let's stop doing that let's get you on a modern tool like Mm-hmm. Flowcast, which is like a tool that helps you manage your deadlines or like SharePoint, even, even that I don't like SharePoint. But like, so like getting, usually I'm always kind of getting into like a new technology, you know, stop using Excel, start using things like alterics for like data processing, you know. So that's, I think, which I think is very needed. I think there's a lot of companies out there that still just keep things running the same, like a lot of accounting departments keep things running the same because they feel like they can't step out of the hamster wheel to fix it. or there's a lot of reluctance to change anything. I think accountants in general are very, like, risk averse. It's why we picked accounting. You know, it's pretty stable. It's, you know, the job market's pretty healthy. So when something, like when someone proposes a change, you know, to how they do something, it's not always welcome. I don't subscribe by that, but, you know, that's just my experience. And so you've got to kind of fight some of that institutional holding on the things because that's why we've always done it. But I think it's a good role. And hopefully, you know, people that are in those rules, have success because it's dearly needed in our industry. The other kind of point or other kind of area I see that is like in public accounting firms. So like firms that are doing, you know, accounting or taxes for others. You know, there's, and I realize that you're not in the accounting space. So there's public accounting, which is like PWC, Deloitte. Those can also be really small firms like, you know, Bernard CPA that just does some taxes. And so they are serving clients. Whereas like, you know, if I work at the accounting department at Toyota, like, I'm. in private accounting. So a small distinction there, but basically, you know, I see people talk in the public accounting space, like having to really need or some of the best hires they've ever done is like a C-O or like a director of operations kind of thing, where because that person is coming in and making sure people are being efficient, you know, they're monitoring people's, like their resources, make sure that they're being fully utilized. Because I hear a lot about like, you know, CPA firms that get started out. and like the CEO is there and he's responsible, he or she is responsible for like sales and getting clients in the door, but then like they neglects like how the work gets done. And that can kind of build up to be a problem later on. And, you know, I've heard some success stories that people say, oh, when I hired, you know, that COO or that director of operations, like took us to the next level, you know, or that really unblocked us. So, um, yeah, there's a bit of that in the accounting world for sure. It's the working in the business versus working on the business. It sounds like to me, And yeah, it sounds like a tough role. I mean, I think for a developer productivity, there's similar challenges that the teams face where even if there is a great solution, getting people to change their habits over the past four years at the company. You know, good luck with that. It's a slow burn on that front. But sometimes the advantage, like you mentioned, if you have a ton of pain up front, but it'll be worth in the end, that is kind of what you sometimes have to cling on to when you're proposing some of these changes. But yeah, it sounds interesting. I always wanted to like apply my technology background to other things. I feel like I get super jazzed about technology in general. And I'm so unfamiliar with other sectors that I'm sure have software that given my low pain tolerance would probably make me want to poke my eyes out and build something in that space. But I just, I don't know, sometimes it's just hard to get, you know, experience with that or even understand what sucks, you know, in other places. is like there's so many sectors, I'm sure, could be revolutionized by better software, but it might be like, quote, unquote, too boring, which I see a lot of YC companies come through and just focus on a, like a boring sector, automating, you know, large trucks like shipping or, you know, just something random like that where if you asked 100 software engineers what they want to do, maybe two would say yes to that idea and, you know, the rest would say no. And it sounds like some of the things that you describe as, you know, older systems are using Excel or I'm sure exists in other sectors. And I wish there is a way in general to like get an insight into that without being an employee at the company or like, you know, I don't know if there is any research methodology or understanding really to just say like, I want to take a look at what software doctors use, lawyers use, accountants use. Because I'm sure that I probably could be jazz about building some cool product in that space that. that solves the need and ties five of their systems together in a better way than they do manually. So if you're listening, you're out there and you have some idea to solve this need, please let us know. I think it'd be really cool to get some sort of face value experience to other tools, workflows across different job professions. But where I sit right now, not going to happen in the sad reality of things. Yeah. One thing that you mentioned there, and then we could probably move on from that topic, but, you know, the, I think you said YC, why C or VC, isn't it VC? YC or Y Combinator. Yeah, yeah, yeah. about boring businesses, like, I would want to build a product that people are, like, excited to, like, pay for. You know what I mean? Like, versus, like, oh, like, I have to pay this bill. Like, I don't want to. So I guess I'll pick this person. That's like an auditor. Like, I used to be an auditor. And, like, we're there because we have to be. And clients, you know, they're always, they're always great. But, like, you know, they weren't, like, excited to see us, right? And I'd rather be in a product space. Even if you can make money doing, like, certainly audit partners make a lot of money. but like that's just not that's not where I want to be I'd rather be like on the exciting part of like this thing saves me so much time or this like gets me so many new clients or this saves me so much money I cannot wait to get this product that that's like where the I'd rather be personally but you know there's lots of things that are boring or neat updating and what I was getting to is I remember reading this I pulled it up as you were talking you know you know what Cobol is right C-O-B-O-L yeah the old language used in banks or something. Yeah, yeah. It's still, like, it's still, like, underpinning, like, 95% of, like, banking, like, systems, you know? And I don't know enough about banking and coding and all that, but, like, I think Cobal was, like, made in, like, the fees or fuse. I don't know the exact, uh, okay, 1959. Okay, I'm looking at this article. So, like, it just seems like, you know, like, you hear about, like, rest, like, you know, revolutionizing or revolutionizing. Is that even a word? I can't talk with that word is. Um, revolutionizing. Is that not a word? That's a word. Probably not. Okay, whatever. I'm not going to edit that out. I'm going to sit with that one. But, you know, that's replacing C, you know, and it's like, I don't know what, like, the Cobra replacement is. Yeah, yeah. So, anyway, someone will come through and do it one day, but there's probably a lot of things that are built on COBOL that they feel like they can't change or unwilling to change. So, yeah, I just thought it was interesting. But, yeah, we'll see. Yeah, we'll have to talk about it in a different podcast, but talk. Talking about legacy technologies, one thing kind of brought this to mind is that I think most modern databases store timestamps as like the Unix epoch time. So essentially from 1970 and onwards, we're counting today's date as the number of seconds since January 1st, 1970. And as we get farther and farther away from that date, the storage bucket for holding that integer, that number amount only can hold so much. So I think it's two to the 32. And if you do two to the 32 on your calculator on Google, that's $4 billion. And the closer we, the farther we get from January 1st, 1970, the higher this value gets. But if it goes over to the 32, then we can't store it anymore. And so I think someone did the math and said, I think it was year 2038 was when we crossover for it to be too big. I'm not sure if it's 38 or 32. do sometime in the 2030s that we can't store this anymore. And then all hell is going to break loose because there's going to be old systems that aren't updated that, you know, people wrote in the year 2000 and it made it somehow to the year 2038 and then it just breaks. Again, probably talking about a different episode in depth, but that Cobalt, like there's so much legacy software everywhere. And hopefully maybe AI can play a part in modernizing it. but it's hard to say. I would not learn Cobalt today. And I don't know anyone, you know, as a new grad coming out, they don't teach it. No one wants to learn it, but it's in demand. And so, hey, if you're struggling to find a job, maybe add that to the hiring list, although you're not going to find a VC company with that background. Yeah, there's a company in the article called the Cobal Cowboys. And they're consultants for the companies that are stuck on. Yeah. Very smart. They probably do well. Yeah, no, that's pretty cool. We'll see. We'll see where that goes. Yeah. One thing I saw on Twitter recently is two CEOs of, I would say, rather famous companies have gone out over the past two weeks announcing that one CEO from Gumroad mentioned that they're not going to hire engineers anymore. The second CEO of Coinbase, Brian, mentioned that all the existing software engineering work is distilled. into tickets. So you can imagine a description of what needs to change. And AI is going to take a first crack at making the code change to fulfill that ticket. That could be a product requirement. That could be a bug fix, et cetera. But I think he quantified it as being essentially a low complexity fix. And I think Coinbase's stance on it sounds pretty reasonable, but I think Gumroad stance on it, if I remember correctly, let me pull up the tweet is pretty extreme. Like, I think the headline, no longer hiring junior or even mid-level software engineers. And this is Sahil from Gumroad and his justification, which I find to be interesting, is that each of their code bases fit within the AI context window. So the AI context window essentially is how much characters you can put in the chat and the AI can act on that. And I think Google leads with one million or maybe two million in their context window, two million tokens. I think four characters is roughly one token on estimate. So you can imagine they have a ton of code. I think he's outlining his code bases. Gumroad is 2 million tokens, so it barely fits within the context window of Google. And his justification is since our code can be understood by AI in its entirety, which is hard for any one engineer to do, there's probably tons of moving parts, we're not going to hire any engineers because AI is going to essentially take over work on all of our code bases. And, you know, one part of me says, this guy loves Twitter attention and will pose anything to get eyes. And I'll be honest, it caught my eye. Two, I think he's definitely incorrect in the fact that when you process a code base that's that large two million in terms of context window, like the maximum that can be processed, it really, really falls apart, the quality of code written and the instructions that you provide, if you say, change this one file, paste in, you know, million tokens, it's going to forget and it doesn't really work as you'd expect. There's a chart that, you know, if you're not watching the video or, you know, I'm not going to put it on screen, but you can imagine this chart that essentially tracks the recall based on the token length, I guess. And so if you paste in a code base, that's maybe 100,000 tokens long and you ask it to do something, the AI model can look at all that code and do a pretty good job. If you go up to 500,000 tokens, it drops off a bit, maybe 10% decrease in performance. You go up to a million, it drops off by like 30%, 2 million by like 60%. And so, yes, you can fit the code base within the model and it doesn't quote unquote break, but it really sucks on performance. It's not a one-for-one match. You can't just throw it in there, get it for free and it does a wonderful job. So I think his take on, oh, our code base like fits barely. We don't need to hire anybody. I think that's complete nonsense as of today. Maybe AI models, you know, get a lot better in the next few months, which they have been. So he could be right. That's why I kind of say half right, half wrong, a little bit for the Twitter drama and attention and a little bit for, you know, just an interesting take. So on my end, as someone who's employed by a software company, Seeing something like that can feel scary, but the way things are today, I'm not too concerned. And, you know, Coinbase's approach where we have a small, simple kind of task, an AI can take a stab at it. That sounds reasonable. It doesn't have a clickbaity headline, but still important nonetheless. Yeah. I think it would be curious because you certainly know this from your own experience to split my expenses, but also just working in tech companies. Like, I think from the outside, you know, how. I guess what I'm thinking about Gumroad, for example, like, how much does it take to maintain something like Gumroad? And, like, we don't know his stats and all that. But, like, you know, I think it's a pretty small company. I'm sure they do great numbers and revenue. But I think I don't think they do too much. I think it's, you know, for those I don't know, gumroad is a place where you can kind of host and sell, like, digital products for lack of better terms. And, you know, so, like, how much do you think, like, he's, they're actually coding, like, day to day. for something like that. Take out if they're adding any new features. Let's just say they're keeping things, keeping the wheels on the car, keeping it in motion. How much code does that take to maintain? I'd be interested for people to know. Yeah, I'm honestly not too sure. When I've interacted with their product, I've literally logged in, or not even logged in, purchased a product, whether that's paid or free, takes me to a download portal. I download it and I leave. So to me, it doesn't seem super complicated. I think if they're doing their own like in-house billing and not moving out to Stripe, it can be really complicated. But the discovery engine for me on Gumroad is I see a creator. They link it. I go directly to the product page. I don't use any of their homepage, search, whatever. So it's literally like an inventory or a directory of products and you buy it and you store it. To me, probably not that complicated, although, you know, you can think of something like Dropbox where you just store files, not that crazy, right? Yeah. It can get really complicated at scale. sinking all that. This doesn't have any, not all the pieces of that, but it does have file storage in some regard. It does have payments, billing, coupons, team pricing. So I would say it has the bells and whistles of a, you know, enterprise SaaS, but I'm not really sure how much they're adding and I don't visit it often enough to know. But my guess is they probably have a team of, I don't know, I don't want to underpin it, but maybe 50 engineers, 40 engineers. I would love to see their stats, but enough people to crack on bugs to solve their infrastructure and architecture to make sure they can build features in the future and they're not bogged down by tech debt. I think the app or the website's been around for a while. But I'd say there's a decent hidden cost in keeping the app up and running and making sure it works. And people don't always understand that bit. And that's just how software works. You know, you create gum road. you depend on 10 different open source packages. The next month you go to upgrade that package, but it breaks. You gotta go do a bunch of manual effort that you never estimated to happen. And that just spends engineering hours and no one even sees anything at the end of the day So there a ton of stuff that happens under the hood My guess is they have a relatively lightweight team but given their feature set billing storage maybe 50 engineers, maybe full-time. I'm not sure. I know they have a really weird company set up, so it could be some odd setup. But I'd say more than you think in general, and a lot of it is upkeep and, like, routine maintenance. Yeah. LinkedIn is estimating between 11 to 50 employees. So you're going to think you're right on the money. Okay. Yeah. My guess is they're probably pretty engineering focus with a decent support team. So if I see that, my guess is maybe 30 engineers and, you know, 10 support, 10 sales, something like that. Yeah. You know, I think we've touched them a little bit, so I won't go too deep on it. But I think accounting is in a similar boat. I think especially for especially for that private accounting. talk about public accounting and then private accounting. You know, I work in private accounting. I used to be in public accounting. But I think private accounting in particular is really ripe for, you know, an AI-enabled kind of, like, low-level task completion. You know what I mean? Like, especially with like a partnered with like a kind of technology first, like general ledger or financial system. You know, things that are more advanced than QuickBooks, obviously. There's workday. There's SAP. There's lots of kind of big enterprise tech that I think will have like that kind of capability. And I think companies should and will start to kind of, you know, kind of more similar to what the Coinbase, you know, tweet was about. Like, hey, for this low level task, like we're just going to have this get automated, you know, and just that'll be going on. But then for the higher level stuff, like we will need people there still, at least for the time being, you know, to manage all that. So, yeah, it's interesting. Yeah, I agree. And then some quick AI updates. so we're almost wrapping the episode. But last time we talked, I had purchased deep research. Well, I guess not exactly deep research, but I pay open AI $200 a month to give the podcast the inside scoop. And so I used deep research probably five or ten times, various topics, coding, life, travel planning. And I thought it was pretty good. One thing I would like to caveat is deep seek came out. And, you know, the big thing about deep seek was that had good reason. It was open source. But as I kind of compare deep research, which if you're not familiar with it, it essentially is OpenAI's method to go scrape the internet for a topic of your choice, understand what each of these internet sources is saying, piece together a puzzle that answers your original question, and produces a roughly 10,000 word document that goes very in depth, given the name deep research, on whatever you are asking for. And the power there is its deep reasoning model, so it's able to actually understand text at a very great level. The second part is web scraping. So, you know, you can imagine if you were Google today, you probably could have a pretty good deep research competitor, giving you've already looked at all of the internet. And so I think Open AI combines the power of internet searching, top reasoning, and their best in class AI models to make a pretty compelling product. And I say product because this doesn't exist in the API. You can't really use this as a developer. This is something they've developed in-house and have completely packaged into a product to kind of get more money for their company. They were a nonprofit. They're trying to move to a for-profit company. All this Elon Sam drama. But TLDR, I tried deep research. I think it's pretty darn good. But it really depends on the topic that you're going for. For example, I gave it, can you create me a travel itinerary for Japan and, Korea. And how deep research works is you ask an initial question. The second question that it asks you in response is probably five or six follow-up clarifying questions. So it's kind of like if I came to you and said, you know, let's audit someone's, you know, company. And you say, well, you know, what exactly are we auditing or give me more information, kind of the first base layer to make sure I don't go spend time on something that we were not aligned on. So that first response that it sends back to you after you ask for something, is that kind of level setting, understanding what you want. So you ask initially. It tries to clarify. You then respond to that. Then it's off of the races and roughly takes, from my experience, five minutes as the fastest has been, up to, I think, 20 minutes on some of my queries. And it has access to roughly the entire internet, as you could probably conceive. And the biggest thing that it's useful for, and, you know, as you can imagine for its name, deep research is that it really focuses on complicated problems. I found with my travel search query, it was good. A little bit tough to read because 10,000 characters is a lot. Like, don't get me wrong if I was jazzed about travel in a way that that was my end-all-ve-all. I think I would read it in a heartbeat. And I did throw it through some coding topics that I was really interested in and I was glued to that 10,000 words. But for the travel planning, I just want to see what it could do. And it did a good job at outline it, itinerary, dates, times, location. reasonings, like asked me initially in the clarifying questions what activities I want to do. So when you imagine you tell a travel planner, you know, plan me a two-week trip in Japan, they're going to ask you those clarifying questions. And when it asked me what activities I want to do, I thought, oh, duh. Like that's something I should have probably provided. But once it asked me, it made sense. And so it tailored the whole plan to what I had asked. I probably read half of it. And then once I got through it, I thought, oh, you know, I kind of get what it's saying at the end of the day. The coding ones, on the other hand, were really good. because sometimes when you're faced with a challenging technical problem, there's not that many resources out there. And what you have to do is kind of open up a tab and stack overflow, like developer forums. Each of these pieces has a little nugget of information. And that nugget of information needs to be applied to your kind of situation and altogether to make a really strong conclusion. And I think that is the power I see of deep research, is doing that manual effort of searching and applying much deep, reasoning. When I asked it for travel plans, you know, I'm not saying travel planning isn't hard, but it's not as complicated as piecing together, you know, deeply technical systems. I'm sure that would probably exist in the accounting world. If you had a problem that was complicated, it'd probably be a better fit than travel planning to begin with. And you get, for $200 a month, you get roughly three queries a day. So do with it what you might. I have not even hit close to $200 a month, but, or sorry, three a day, but, you know, I try to use them when I find fit. I would say given Deep Seek's abilities to also search and other companies coming out with deep research, I think it's really useful for select topics, medical, deep technical, research-e-esque topics. And so I'd be very careful on people listening to the podcast to immediately dive into it because I would say take a step back and make sure that the things that you're looking at do fit the category. Else I would say it's completely not worth the money. and for me personally I would say I probably will do one more month after that I'm not too sure so I'll report back but that's that's kind of my my pulse on deep research so far okay you know there's another one perplexity right that came out yeah okay yeah and so perplexity is kind of the same thing it doesn't do as much deep research although they very recently added it probably like a week or two ago using their existence existing AI search. I think previously they used Claude or some other off-the-shelf model. Now they use deep seek and I think they fine-tuned it to their use case of, you know, kind of deep research-esque. Okay, I got you. So it's kind of similar. I think perplexity's strength is that it's fast. You search Google and it does reasoning and it's done. Deep research is like, hey, I want to go spend 10 minutes and get you the best that I can get and then you can read my distilled information where perplexity is like, you know, you search, you're done. Not super complicated, although it's like faster than searching yourself. And then the second update that OpenAI put out is a mini. So this is their next model on top of a one mini. And it's their reasoning model, but it's the mini version. So it doesn't take a long time. It's 03 mini and 03 mini high. None of these, I think, are available for the API yet. But the naming sucks. And it's, I haven't tested enough to really tell you if it's good or not. I'm trying in cursor. It seems decent, but it's not. crazy. So, yeah, we have deep research. We have O3 and O3 Mini. I think open AI overall has a fire under their butt given all the deep seek stuff. So they're really shipping a lot more. And I'm excited to see some real competition in the space where they can't just hold back everything that they have. Yeah. One quick thing, because I know we're going to wrap this up here in a second, I just saw, this is a pretty live in my perspective, that Sam Altman was tweeting. He was trying to get feedback about what they should do next. Did you see that? I think that was so. Yeah, yeah. And he said, and I'm confused, so this is genuinely a question. He said something about, let me find the tweet really fast for our next open source project. And I was confused because I was like, they're not open source at all, right? Am I making that up? I think they started open source and then they stopped. I think that's their evolution because they started as a nonprofit. Again, I have not fact check any of this, but maybe I will by the next time. I think they started open source, then made oodles and oodles of money, and now we're trying to convert. But along that process, I think they release open source for really old things. So for example, people were asking them to open source like GPT3, which isn't even served with their API and is never used. There's three, three point-unquote generations behind. And people were asking to open source that because it was state of the art. I don't think they do that much anymore, but I think given the impetus for open sourcing and deep seek, I do think there's an urge to do that. I think he was asking for a phone model or, I don't remember what you know. It was O3 Mini or a phone size model, and this was yesterday. Yeah, and like if he open source 3 Mini, that's like their latest and greatest. There's no chance that happens. I don't imagine. Yeah. Okay. I just saw it and I was like open source. You haven't been open, you're not open source at all. I think he does little theatrics on Twitter. I hate to say it, but he does have a little bit of tips and tricks on Twitter with his persona, reach, and kind of financial shenanigans. Hey, we're talking about him. It's working, I guess. I'm sure once he found out that we were talking about him, he's like, got him. Yeah. We'll get him on the podcast soon. Yeah, yeah, yeah. I think he reached out last week. I'll have to respond back to him. Yeah. Cool. Awesome. I'll let you go first on yours. Yeah, mine's pretty quick, and I think these come up rather often, so I won't kind of coin it as something super new, but as cursor becomes state-of-the-art, more kind of implemented and allowed at various companies and suggested at a high level. There is a tweet that I'll link, which is Ted Worbel, and Ted talks about a prompt to get cursors, kind of composer mode, to really work well, And the composer mode is the mode that will take your query, run it, produce code, and then kind of look back on that code and kind of keep iterating. And his prompt, again, I'm not going to read it out. It's pretty long. It's just a long prompt of making sure that the AI plans its changes first and then executes the changes. So if you as a human were looking to do something similar and someone gave you a large problem, you'd probably want to plan your changes before you dive into making the change. and so this at a high level is called the quote unquote planner mode and cursor and other AI models need to be told how to think in a correct way and once they do that they perform better. Kind of stupid, but it works. So I'll link in the show notes. If you're a cursor user, check it out. I can report back how it worked for me. Yeah. Okay. That's cool. I have two. One of them is really fast. Well, they'll both be fast, but one will be really fast. so I saw that Humane is being acquired by HP. H.P.'s buying the assets. Yeah. And for those that don't know, Humane is like, or the pin was like a little device that you wear on your shirt. And I don't know all the details behind it other than what's his name, Marquise. I don't even know his Twitter handle. Yeah, what's his name? MKB.K.B.C. Marquis Brownlee. Yeah, yeah, okay. I think that he reviewed it and said it was like one of the worst products he ever, like, tested or something like that. And I remember when he did that, it was pretty, like, sensational. But people were like, wow, you must really hate that. And so anyway, so it was funny to see, I mean, that wasn't that long ago. I think that was 10 months ago. I just looked it up. Yeah. Yeah. Yeah. So. He's got the car as well. Well, yeah. But, I mean, they spent, I think, like the marketing was, was snazzy. they had the product. They had all like the right kind of landing pages and all that. And yeah, it doesn't seem like it. It went well. Like HP is buying it and then killing it, it sounds like. Yeah, and I don't think HP knows what to do with that. And so I'm not really sure how that ended up, but good for them. Hopefully they didn't come out totally, you know, lost money. I think buying for a fire sale is not horrible given their situation. I have an HP printer and it sucks. So I don't know. I'm just putting that out there. It's horrible. I do know. one at HP, so shout out to them if they're listening. But yeah, hopefully you guys do great things with Humane. Yeah, and have better printer drivers. I don't know. Cool. But the last one I had was actually a pretty cool one. So it's about open AI and like data security. So I'll make this fast. I know we'll wrap it up on time. A lot of the natural concern with using an L.M, especially for like client work. So like if you are a CPA and you want to use AI, but like you feel like you can't get the most out of it because you're afraid that you can't share client specific specific information. And so Jason Stats, he had a post a couple weeks ago that kind of walked through what to do if you're like afraid of like using client information in like any kind of AI assistance. And he focused on open AI as kind of the benchmark because and he ran through this was actually super helpful because I didn't know this until I saw his post. But I guess with open AI, you know, you can have your free plus. pro and then there's team and then there's enterprise so there's five different categories and with team the most important thing i think is that team they do not train on your data like so if you are putting in information open a is not going to look at the data and train their model on it whereas if you have your free or your plus or your pro they can do that like they're within their rights to do that you can opt out but you know with team it's it's no it's a hard fast no So that's kind of cool. And, you know, it has like SOC II compliance, which I won't get into the details, but on my accountants will know what that is. And kind of like it's controls, like statement of controls. Like we're like, okay, that's a legit system, has good controls. And so, yeah, and he gets other tips as well. So I won't go through all the tips he gives, but the team one is a good call out with a SOC2 compliance. And then the last point he makes that I want to mention is, you know, if you're still on the fence about, you know, like is this safe to use? for like my client's data. He's like, pull out all the other logins that you have that you're using with your clients. Like see which ones have like multi-factor authentication, which ones have like a SOC 2. They probably don't all that. They probably all don't have that. And so like, you know, realistically, like, why are you sitting off, you know, sides on this? When like you should be getting in, you can do it in a responsible way. And honestly, it's probably not any less safe than other things that you're using. It's probably things that you're using that are more at risk, you know, that don't have multi-factor authentication or don't have SOC2. So it was good because I'm definitely someone who tends to be on the cautious side about, like, you know, using sensitive data in something like that. But, again, I got a lot out of that post and, you know, I thought it was helpful. So, yeah, honestly, very helpful. Every time Jason's links make it on the pod, I do got to say his value he packs in the amount of text he writes is extremely high. easy to understand. I really should take a look at more of his post because easy to scan, easy to read, exactly, no fluff, like just makes sense, highlights things. I'm like, dang, if you're going to write tips or LinkedIn posts or anything, you know, maybe he's a good person to just follow to get an understanding of succinct and useful information density. Yeah, I don't know when the guy sleeps because he also has, like, phenomenal YouTube videos constantly coming out. So the guy's a machine. Goals. I'll be, I'm thankful for it, yeah, because he's got good nuggets in there for sure. Yeah, he's cool. Cool. Awesome, man. Well, I am dead tired. I think we'll call it there. But, yeah, good stuff. Yeah. Catch it next time. Sounds good. Sounds good. See ya. Cool. See yeah. episode of the Breakheavenbrothers podcast. To find more of episodes, you can go to Breakheavenbrothers.com or find us on YouTube, Spotify, Apple Podcasts, or Transistor FM. To connect with Bradd, you can go to Bradley Bernard.com or at Bradley Bernard on X. And to connect with me, you can go to Ben atbernard.com or at the real tech CPA on X. You can also find us both on LinkedIn. Thanks. See you next time.
Creators and Guests


