Bản chép lời đầy đủ của Will Vibe Coding Kill Software Engineering?

which is like, I don't think software is dying. It's just very different than it was. And the software company... is very different than the process of writing software or having software or using it. And the giant enterprise software company is going to be changing. How many people work there, what they can offer for benefits, what that does to like, is that going to be the pinnacle role that everybody aspires to working at a fan company? Well, that might change, but we're still going to need software. We need applications that can do things, but they can be more bespoke. It's much easier to customize them. Pete Reilly: Yeah. Stew: Yeah, think, yeah. I mean, this is as maybe the only economics major. I don't know if we have any other economics majors or both of you guys. Yeah. Okay. Everyone's an economics major. Right. There was this great, you know, there was this British economist Jevon who, you know, is most famous for coming and kind of popularizing the concept of diminishing marginal utility, which is a different topic. But he also had this Jevons paradox, which, you know, in in the 1800s, it became dramatically cheaper to run steam engines. You needed a lot less coal to drive the same amount of power output. And what happened was coal usage went through the roof, which would be exactly what you would think the opposite would happen, because you could get the same amount of work done for dramatically less coal. But as the cost of what it took to run a steam engine dropped. You just got a lot more steam engines to do a lot more things where the higher costs wouldn't have made sense, but it got down to a level where we're like, ⁓ now I could be using a steam engine to do this type of work. It's than, you know, the person I was paying to do it before or whatever. So that's, ⁓ think going happen with software, software is not going away. We'll probably have a thousand X more software than we've ever had in our life. And we'll have Pete Reilly: Thank you. Stew: you know, it'll make way, even people I think that have gotten into vibe coding understand that, that they're now, ⁓ know, creating new software to do things that you never would have thought to do before. Cause just the time and cost and, ⁓ know, PETA factor of trying to do that would have been way too high before, but now you can do it very quickly. so, you know, I'm kind of just reinforcing everyone else's comment. mean, I think there's going to be more software than ever how it's. Pete Reilly: Right. Stew: Developed is going to radically change, which is going to reshape the way that the what it means to be a software developer is going to reshape what it means even to be software how software looks but But will be Software will everywhere now more so than ever possible because it's no longer as constrained to create it Pete Reilly: Yeah, I couldn't agree more. I've personally created more applications over the last two months than I have in the last four decades. it's, for me, ⁓ taking the place of where I would normally use a spreadsheet ⁓ do something and track something and track the progress of something. It's just as easy now to build an app to do that. So why not? And now I can share it with other people and it's actually a much richer experience. So I think I'm with all you guys. think you're just gonna see a lot more software, but as you point out, and I'm actually curious to ask you Mike, Mike, you're a software developer, you know, this is sort of your bailiwick. This does change what it means to develop software. And I could imagine some people, It is such a big change that that's a tough thing for folks to sort of adjust to. What's your perspective on that and how do you view that? Mike: Yeah, that was a great question. Look, so one metric that's interesting is to say, we all got used to getting software for a dollar when our iPhones could download an app, right, for a buck. And the interesting thing is now you could create a brand new from scratch application that is custom for you for about a dollar, right? It's not one of the ones that's already out there. It's brand new for you. Now, then the question is, is everybody going to go create their own apps for a dollar instead of buying one that's already out there? And the answer is no, of course not. You're going to buy Pete Reilly: So. Mike: applications that somebody has spent a lot of human time thinking about and getting right make it right and so to make the thing that you want right. So in this in context of saying, you know, ⁓ does it mean to the people that make software? Well, yeah, it means that a lot of customized applications are going to get done by the end user a lot of customization by the end user. So great. I don't have to go down the last mile for every customer out there. It also means that a lot of fairly simple utilitarian things are going to be ⁓ cheap and easy and available and everybody will have them and what that's going to leave is the interesting middle, right? That means that now if I'm going to make an interesting software product that I'm going to offer out, it has to include all of that reliability and ⁓ and all the things that the good design that have always been there ⁓ it has to include the ability for ⁓ user to jump in and say, by the way, I don't like this menu. I want to make these five changes to it. So every software application is going to be a vibe coding platform for itself. Right. So instead of something being configurable, like in the old, you know, low code days where if you knew what you were doing, you could escape three times and start doing slash commands to do JavaScript or whatever. Now you're just going to say to the, to a prompt somewhere in that application, you know, by the way, when, ⁓ know, I run, I don't want to see whatever the water in blue, I want to see the water with actual ocean waves on it, right? And it'll go figure out a way to make the ocean wavy for you on your running app, right? And

that's just the kind of thing that people will be able to do again for just not too much money inside that app. Does that make them vibe coders? I would argue that it does. Now, if you never look at the code, are you a coder or are you just a vibe user at that point, right? So it's sort of a... maybe a without a difference kind of a problem where we're just saying that like every advancement in technology over time, it's how you incorporate it into the useful applications, not how it replaces something that came before it. Pete Reilly: In my, just so I can sort of repeat, I think some of the key things you're saying, I think what you're saying is, listen, in the past, if you got Salesforce you, would spend, I mean, companies spend hundreds of thousands of dollars configuring Salesforce to match how they do business and how they report on things and how they forecast their pipeline. I think what you're saying is the future of applications is maybe ⁓ layer of vibe coding on top of the applications as opposed to sort of the processes you might go through. Is that what you're saying? Mike: Yeah, that's right. That's right. We used to joke that there were only two commands in software. One was set bugs off. And the second one is do what I'm thinking, right? ⁓ And so, you know, of course, that's still a joke today. It's still, it doesn't work that way right now, even with vibe coding, right? You can't just turn bugs off. In fact, you got more bugs than ever to go find, and you can't just say, do what I'm thinking. You actually have to translate it into meaningful, differentiated, ⁓ interesting stuff, right? Because these applications, vibe coding models are trained on millions of examples of code, mean just like the way that they write language is same way they write code. It's very generic, right? It's going to be, so it's essentially an average of what's on the internet, a very high quality average. It's Oxford English perfect, but it's not creative. It's not different differentiated. It's not going to have your uniqueness in it. Right. ⁓ and so, so that's really kind of the core differentiator here for, know, how these tools are used versus what's already out there. Pete Reilly: So do you guys think that world, obviously you need the source code to these things. So does all software become source available, open source? that? you Stew: I have a slightly different way of maybe saying a very similar thing, but maybe it's also different. don't know. think it's much more around these main system of records that you interact with or main systems that we're used to working with on a computer. I think those systems are now ⁓ for the agents. and meaning like ⁓ as software developers you need to think about how do I make those systems friendly for agents to interact with and then I think the human experience is going to be Over time going much more to we interact with agents who then work with these systems on our behalf the same way that a a, ⁓ you know, if

we just had an assistant that we talked to all the time that just said, you know, like, which used to be the case, right? You know, like, Hey, you had an, if right, if you were a rich CEO or whatever, you had an assistant that was going through your email and doing all of these things for you. And you would just kind of talk to them quickly and tell them what you want and they would know you well enough to do it. think the world's turning more in that way as part of doing that. I think these agents are going to have to create. like almost like custom UIs to, you know, communicate what they need from you in order for them to get their job done. And I think you'll be building a lot of these differently. ⁓ don't think that necessarily means that all software becomes open source and all that things. I think it just means that interfaces have to be very clearly defined, right? Like the back of, there's a lot of stuff in the software that doesn't need to be open source. But I think you need to have... agent friendly interfaces. don't know if that's. Pete Reilly: Your vision is more that ⁓ may ⁓ my own software, but that software ⁓ connecting to Gmail and Salesforce and QuickBooks. And so those systems of record continue to exist, but I may very well just create my own Shanti Greene: Yeah. Pete Reilly: ⁓ experience for utilizing them. Stew: I think it'll be, ⁓ it'll be, some that will happen. There'll definitely be a period of that. I think it's gonna be much more that ⁓ we're all wind up using a handful of super agents, you know. Claude, OpenAI, something those lines. hopefully there'll be open source versions of those agents, by the way, which I think a lot of people would want. And those agents will be interfacing into these systems, vibe coding, the user experience we need to support the very specific thing we're trying to achieve. Pete Reilly: With the LLMs themselves, yeah. Stew: But that's kind of like, it'll be Jarvis doing it. And we're just talking to Jarvis, if that makes sense. But maybe that's a little far out. Shanti Greene: Something that I look for as I'm evaluating new applications, new tools, wearables, platforms is all around integration points and open API specs. Because I don't, to your point, Pete, we don't necessarily need open source. What we need are open documented integrations. How do I get data in? How do I pull data out? And even better, will that place collecting data push it out via web? If somebody's willing to publish a webhook and an API spec, I'm ready to use their product. And if they're not, I'm very hesitant because I need my data living in the place I want it to live because I can build any UI I want on top of that and combine it with all of the other things. But if you're going to try to lock that away and build a mode around my data, that's a big no. Pete Reilly: Hmm. Stew: By the way, I think you're seeing something. There was something released this week. That's a little more mains. We talked last time, I think

around like Claude bot or whatever. Uh, but it opened Claude, whatever its name of the week is. But I think the, um, the perplexity released something this week, uh, it's called perplexity computer. Now it's a high toll to get into it. I think you have to be on the light $200 a month plan or something, but. Shanti Greene: you Stew: But I was at the end of my subscription, so I could sign up for the $200 ⁓ and only have to pay the prorated last three days or something of it. I it and got it. ⁓ And it's like you give it something to do you as your Jarvis. And it will go out and build a custom UI around that for you, a custom app. Shanti Greene: Yeah Stew: for you that's built just for that. I using it yesterday to some very interesting kind deep research on a topic ⁓ for one our customers. But I've seen other people building like Bloomberg. ⁓ The famous online now is someone basically ⁓ had it a Bloomberg terminal just to look at the Nvidia release announcement or something. So I think coming fast. Shanti Greene: I will say the comments on that Bloomberg terminal specific like extra are pretty hilarious. Also Stew without talking you and I did the same thing which is signed up for that. I built three apps in like 24 hours on perplexity computer and including like a little stock screener, a due diligence screener for a specific investment and one other thing. I was like, how much of this can I use before my subscription expires? I don't wanna do two months on it. Stew: Yeah. Yeah. Shanti Greene: And how similar is it to what OpenClaw is letting me do? Because I run five instances of OpenClaw in different places. It's not bad so far. It's not my favorite. I actually think wallet code is still a better tool for building things. But it's pretty good and easier to use. Mike: Well, and what you guys are pointing out is how every time some new thing comes out, we find more ways to use it, right? It's just this flywheel is just going faster and faster and bigger and bigger. Stew: Yeah, yeah, yeah, go ahead. All right. Mike: you know, the real constraint here, that is, that's going to sort of slow it down and put the brakes on is there's just not enough chips out there. There's not enough power out there to make all this stuff keep working. We're, you know, sure, in some number of years, we'll catch up to all that demand and, you know, we'll be data centers in space now, Google and SpaceX are both doing that. So that's definitely coming and there will be that that era of plenty for AI sometime, you know, call it three to five to 10 years out. But there will be a period in the short term where billion people on the planet are all trying to build something that some deep, unique thing for their needs. And we're going to have some scarcity of those resources. Pete Reilly: Yeah, I think you're right. but regardless of all that, should every CEO or CXO be vibe coding? Shanti Greene: think everybody needs to try it because you'll see these opposite perspectives.

So people are like, I vibe coded these 20 apps, they're perfect. And then people will be like, everything you buy code is trash, it's all slop. ⁓ the reality like always is somewhere in the middle. It's closer probably to the, built 20 apps and they're all good. It's just that a lot of those apps don't have robust integrations where they really work the way you want them to. They mostly work the way you want them to. Pete Reilly: Yeah. Shanti Greene: getting one to connect through like Google OAuth so you can log in with your Gmail address. Okay, that's not actually easy. But if you make the right connections, like supabase will handle that for you. Shout out supabase, you've got to sponsor us, that's great. Pete Reilly: It's a pain. if you're persistent enough, you just keep asking the model, like, how do I do this? And eventually, ⁓ getting there. ⁓ Stew: I, yeah. Shanti Greene: Yeah. Stew: Yeah, I think the interesting thing that was less true even a month or two ago than it is today is I think there's a much easier place to get started that'll start to give you a flavor of the possible. And if you look at just over the last month or two, know, Claude, co-work out, which is kind of offers a lot of the power of these vibe coding tools, but is, but it's less for ⁓ ⁓ I'm creating ⁓ reusable software. Then it's like, it's, I'm going to ask an agent to do something and it'll create code, but I'm asking it to do something very specific for me, like create this Excel document, create this PowerPoint document, go through all this stuff, you know? And it connects to a ton of systems, you know, which gives you a power of how it does that. And, know, the perplexity computers of the same ilk. Right? They have pros and cons. So I think there's these intermediate things now that don't require you to have to wrap your head around all the intricacies of ⁓ coding deploying that code and all of that other stuff that can give you a real ⁓ window into coming power of these agents. Mike: Yeah, look, I mean every CEO has to a defensible position ⁓ a market that people want to be in, right? I mean, that's what they that's what their job is now So could you say that they must learn to vibe code to get that job done? ⁓ you I've said from the beginning of all this I've said that the that the new program is actually examples the ultimate software, the ultimate program to run your business is examples of how to do things right. Yeah, that is your defensible edge. That is how you are going to continue to win time after time. if you can continue to provide those ways to make good decisions in your space and apply those through technology, through people, through process, through SAP for all I care, then going to do well. ⁓ I would that... You know, if a CEO came in and said, no, no, we don't have electricity at my house. We don't use that. Therefore this business shouldn't have electricity. Fine. You'd kind of scratch your head and say, I'm not sure that that's the right person. But the CEO, ⁓ you know, came in and said, you know, I used Sora and I used, you know, Claude cowork, and, doesn't understand exactly how, you know, vibe coding, Python works. I think he's probably okay. Or she. Sure. Yeah. Pete Reilly: I think that if you, ⁓ a leader of a company of any size, you don't at least attempt this and maybe you don't have to download cloud code, but you know, on replet or bold or lovable or ⁓ Because here's the thing, until you ride the bike, you have no idea what this is really. You can read about it all day long and listen to Andrej Karpathy or whatever. But until you actually sit down and start working with it, just not going to get it. So I would say, yeah, you actually, yeah, you have to, because otherwise you are just going to get your doors blown off by the people that do get it. Mike: Yeah. So, so even if you don't, if you don't love electricity, you should still go to a place that uses it so you can see what it's all about. Right. So you can understand it. Yeah. Yeah. Yeah. Shanti Greene: Well. Yeah. Pete Reilly: ⁓ Whoa! Light and heat! Those are good things! Shanti Greene: Yeah, exactly. Well, and the barrier to entry is so low. It's like, okay, are you going to write four sentences today? Great. Write those four sentences and see if it creates a thing for you. The other thing it just tells me about is like, are you generally curious? Because you really need to be to get through today's world where things are moving this quickly. You've got to want to know what's out there and what those things are capable of. Stew: you Mike: Yeah. Pete Reilly: Yeah. Yeah. I agree, it's just changing too fast. Mike: Yeah. Yeah, well, and it's remarkable to me. It's still, can't believe this, you three years after GPT launched, we're still seeing a lot of people that are skeptical, even engineers, people that are in technical roles, people that, you know, do this stuff for a living are saying, well, it's just a next token predictor or whatever. I'm going get a t-shirt that says, I'm just an oxygen breather. You know, that's all I am. You know, the Stew: Mm. Shanti Greene: Ha ha ha ha ha Mike: You know, because it is like you're saying Pete, it's very emergent and at the same time, ⁓ know, people sort of are in denial or refusal about, you know, ⁓ to use it. And I think that, you know, the winners in this are going to be the people that adapt and update how they see the world by including AI in it, not the people who, you know, resist AI because they could do better or it's slop or whatever. ⁓ You all that said, I do have a ⁓ simple that that If you're going to send me a document, want you to have spent more time on it. Then it's going to take me to read it. Right. In other and, so, you know, when somebody, somebody in a that's got 10,000 lines and, and they've got a ⁓ 30 second loom video showing me that it works. ⁓ That's, nonsense. Right. Like, like I want to, I to, I want to know that, that you put some automation

testing in it. I want to know that it followed the architecture. I want to know that, that, Pete Reilly: Yeah. Stew: Okay. you Mike: that it's compound engineering, right? This term compound engineering means that everything you do next is actually easier because of the thing you did before. And the slop world, ⁓ know, the whole problem is that as if somebody's just writing a sloppy document, the longer it gets, the less useful it is, right? And it ⁓ true in code. ⁓ sloppier you build, the harder it is to add one more thing. The whole thing becomes brittle as you add onto it. So if you can vibe code in a way that's compound, meaning Pete Reilly: Yeah. Mike: you're developing new rules for how your application works. You're reusing common platform elements. You're refactoring when there's a chance to reuse two things. are unit test cases, right? All of these things build that compound engineering. And is the way that that Vibe coding scales up to large large teams, and ultimately over time, crushes the cost and quintuples utility of Pete Reilly: Well, the slop palette is true not just for code, right? It's true for a that I made or a document that I made. think someone's going to come up with a word for that. It's going to be considered to be very bad manners to ⁓ have reviewed whatever it is that I'm sending you. Now you're just putting the work on me, which is, is ridiculous. Mike: Right. Yeah. Yeah. Stew: Well, I like, I guess because I had a background for a while and other types of content. Creation video games videos things along those lines. I think you know, we've had pretty similar Things happen over the last ⁓ years in media like look at videos as an example, right the cost of video production basically went from something that to do something serious and professional right needed a lot of infrastructure to almost nothing and The result of that is always, it is like, look at YouTube. Are there a lot of slop videos on YouTube? Are there 100 pieces of crap content, if not 1,000 pieces of crap content for every one bit of good content? Absolutely. But is there also some of the most incredible and helpful content? mankind has ever seen and is it and can you you know now they can a motivated person anywhere in the world as long as they have access to YouTube basically learn anything about anything and you know, yes, right So I think you get this it's just what happens ⁓ when when you drive something like this I think you're gonna have a lot of slop you're gonna have a lot you're going to, there's lots of ways to avoid the slop. And Mike's points are awesome. But you know, what, you're also the same thing that puts this long tail of really sloppy content is going to create remarkable content that'll change the world. And I think that's just what happens in this case ⁓ when make it ⁓ so easier to do something. ⁓ Pete Reilly: Mm-hmm. Yeah. Stew: A lot of it's going to be bad. Most of it will probably be bad. Mike: But there'll be some outliers that are remarkably better than anything you could have done. Stew: Yeah, but yeah, you'll still be better off than we were before it. Shanti Greene: Well, in the legacy mindset, you like once something was built, then it became maintain it became this whole process. Well, you never want to start over. Now I start over very frequently because anything I built last year, whatever principles I was applying, even if they were pretty good then, maybe better if I just started from scratch now. And that entire process is so fast. It's like, well, how much have I really lost in time? An hour? So tokens? Stew: you Mike: Mm-hmm. Yep. Shanti Greene: If you get too attached to the thing you built previously and aren't willing to re-architect with modern standards, you're never going to advance. So you can build with foundations that let you do that, modular components. Let me just rebuild one component instead of having to rebuild the entire application and take advantage of those new things. Pete Reilly: Okay. Mike: Yeah, absolutely. And the fact that you can to a model, hey, look at the thing that I did last year, bring the boilerplate, but let's switch what language we're in. Let's go to a new database. I want a different security protocol. ⁓ And let me know when you're ready, because we're going to invent some new functionality. And so in 30 words, I am now I have stood up and this is your point Pete about lovable or or you know others out there that in 30 seconds I've stood up a thing that is now a safe place to build new functionality. Now I get to kick in. Now I get to be human. I get to be creative. I get to be differentiated. I get to go over and above right and that is truly the fantastic part of this ⁓ the crazy thing is it's not just in software right? It's in it's in you know the the someone who is buying a new car and doesn't has never done that before and doesn't have anybody helping them negotiate. It's in, you know, somebody that's trying to figure out a, you know, a traffic ticket and needs some legal help. It's all of these important things that, know, have been ⁓ folks with resources and established infrastructure. They've been straightforward to do. I'm not going to say easy, but they've been bureaucratically straightforward to do. And now all sorts of people that didn't have the reach to be able to have access to those capabilities. That's going to improve their quality of life. They're going to have access to a whole different level of, you know, of integration with the social economic system, right? That's going to lead to more people writing fantastic code and fantastic documents and leading companies and starting companies, right? so, ⁓ know, kind of like you're the bell curve and analogy, Stew, where, when you know, because there's so much more code, there's going to be a lot more bad code, but there's going to be some stuff that's outliers compared to the past. That's going to happen across as well. I'm not so worried that, you

know, that, that, ⁓ we have to retrain some really smart people who did a thing that we can do in an automated way. Now that's, that is a problem. Individuals are affected. and it's a tough thing for a lot of people, but so many more people are going to come into that. It's going to basically sharpen, the, point where, you know, where, people that didn't have access to careers, didn't have access to resources in the past are going to, for less than a dollar. be able to access any resource in the world contextually for the problem they have at hand. to me outweighs a lot of the ⁓ that we see from this. And it's a super positive thing that I don't see integrated enough in dialogue about AI. ⁓ Pete Reilly: Yeah. Stew: I think your comments kind of reminded me. I think it was Karpathy, maybe someone else, but was kind of pontificating on this point. one of the big AI researchers has been thinking about this a lot longer than us. His his point was the thing that's going to drive a lot more value now agency versus intelligence, you know, and you know, kind of the distinction here, you know, is agency kind of being like, you know, I have strong opinions about, you how I think something should be done. And I have. Mike: Hmm. Stew: strong instincts to go out and make it happen. And I think people that have high agency now ⁓ will better tools, more capabilities, whatever else than ever before. think ⁓ point though, but if ⁓ you were of just thriving on being super smart, but had low agency, you might be in trouble. fact that you spent, ⁓ you know, decades honing, ⁓ a certain intelligence skill along those lines, the, the AI is ⁓ not going care about that, you know, ⁓ and it's to be able to do some of those things, really fast, ⁓ well, and error prone that value of just being intelligent. without agency is going to go down. I think, you know, it takes a lot of intelligence to put agency to work in a way. ⁓ So don't the value of intelligence is going to go down, but I think the value of intelligence without agency is going to go down. And that's something that we probably all have to start thinking about. How do we get our kids to think about that? How do we get ourselves to think about that? How do we get our workforces to think about that? Mike: Right, right, right. Pete Reilly: Yeah. Mike: Right. Yep. Pete Reilly: So we're getting into the sort of maybe the final topic here, which is what do we think ultimately happens to the workforce? Jack Dorsey at Square just announced that was essentially cutting the company in half and they were doing well actually. He's leaning into this. Now, I would that Elon Musk over Twitter and laid off 80 % of the workforce and didn't skip a beat. So ⁓ maybe this just Jack realizing ⁓ he needs get more efficient, but. It does seem to be a little bit of a canary, or at least people are treating it as a canary in the coal mine. What do you guys think happens to them? Stew: By the way, there was a counter study this week, which

was interesting, which is, think it was Indeed or ⁓ of these job posting things actually posted a study this week. with a chart showing a relatively clear trend where software engineering jobs have been on a steady increase overall in the market. ⁓ you know, ⁓ some kind of points and counterpoints. I think they're actually internally consistent with each other in that, you know, it's, I there's a lot of jobs now to try to build these tools and make and drive these efficiencies and run these agents. that are people are ⁓ get enough people to do those well. And there's a million more projects ⁓ people are willing to take on now because the cost of doing that project is lower. And back to the Jevons paradox conversation from earlier. ⁓ a, ⁓ a lot of people have had a job. Mike: Right. Stew: doing things that the AI now does really well. Right. And so there's this retooling effort that has to happen. You know, I think in the same organization, right, I think overall it may kind of come out OK, which is what has happened in past technological revolutions. But but it was disruptive because, you know, those jobs that aren't Mike: Yep. Yep. Yep. Pete Reilly: Okay. you Stew: You know that, that the AI can do really well. They're probably going to go away and there's going to be this new class of job to do things we can't even imagine yet. But you know, or, we would, or in our head, we're like, why would you do that? That's way too expensive to try to do that for the value it would create. Well, if it's no longer expensive, maybe it's worth doing. And so it's worth having a job for. So I think that'll happen. I also think by the way, all of our jobs are kind of stupid. Mike: Yeah. Right. Stew: to people 150 years ago, you know what mean? I spent a bunch of years making video games. Could you someone's job was to make pixels on the screen slightly more visually appealing than the other ones so someone ⁓ could be marginally more entertained by it? ⁓ think ⁓ humans are really good coming up with... Pete Reilly: Right, we were all farming. ⁓ Mike: Yeah. ⁓ that guy going to survive, right? Yeah. ⁓ Pete Reilly: or expectation of services. Shanti Greene: You Stew: new ways to be competitive with each other, new ways to do things that drive social status. think we'll wind up, things will wind up being created. ⁓ basic things that keep us fed and sheltered and all of the lower ends of Maslow's hierarchy will keep being driven more and more efficient, but we'll figure something out. Mike: Yeah. Yeah. Right. Yeah, yeah. Yeah, I mean, look, it's undeniable. you you obviously if you go one person at a time, everybody's got a story and it can be tragic. But if you come from the top down and you say, all right, the the the knowledge worker economy, let's say half of the knowledge workers out there lost their jobs because the work got reorganized, AI is doing part of it, half the knowledge work that that turns out to be about 10 % of the population. ⁓ it's actually some of the more switched on highly educated 10 % of the population. So now you're telling me that as society where we don't have enough doctors, we don't have enough we don't have enough, ⁓ know, of a million different teachers, right? We don't have enough really important jobs in society. You tell me we can't absorb 10 % more into fields. ⁓ That crazy to me. Now, the switching costs are high and everybody has to make their individual decisions and it's who am I to impose a million things like that. But from a macro perspective, ⁓ the story doesn't play. Pete Reilly: Thank you. Shanti Greene: Yeah, I think the doomsday story is just it's slightly different, right? Intelligence work as it exists today, this exact way is going to change, right? It's going to be disrupted. It's not going to be the same. If you are a traditional CPA who enters people's taxes, that might change because an AI can do that really well. It can understand the rules. It can figure out all of the different forms it needs to fill out. Yeah, it can do a lot of those things. So that's going to change. But Mike: The big beautiful bill. Right. Shanti Greene: Pivoting will have to happen. So it's all about the perspective on the job as they exist exactly will change. Pete Reilly: Yeah, yeah. It was just like we weren't farming 150 years ago. just as Stew was talking about. A couple of things I just sort of throw into the mix. One is, we're just talking about software developers, I think about Andrej Karpathy, who you mentioned earlier, Stew. Mike: Yeah. Pete Reilly: talked about how he felt like AGI was still 10 years away. And the work that needed to be done was a lot of scaffolding around it to get value from it in the short term. So I think there's going to be a ton of software engine. And my guess is that's a lot of what that software engineer chart that you talked about is sort of driving. There's a ton of work. And we know because we do a lot of that work. Number one. Number two, I could see companies almost using this as a forcing function, not unlike what Klarna did. probably maybe last year where they just said, all right, we're just going to move on for support to AI. And then in a little while later, they're like, ⁓ not a great idea. need to dial that back. So I could see some folks sort of pushing the envelope there. And I suspect that's some of what's maybe going on in a couple of companies. I agree. There is going to be disruption. I don't think anybody's saying there's not going be disruption. I think part of the outcome of that disruption, I think, will likely be the fact that Mike: A little pioneered, yeah. Yeah. You Pete Reilly: Creating and starting a company now has probably decreased in cost by an order of magnitude. So I would be willing to bet that you're going to see an explosion in the entrepreneurial class in this country and in the world for that reason. A, the 10 % of the folks, Mike, that you mentioned, they got maybe laid off. Well, now they got a six

month runway, which in the past, wow, six months might not be, you feel like a lot of time to get something up and running and build a product. Well, guess what? Stew: Okay. Mike: Mm-hmm. Pete Reilly: You know, you can be in market like inside of weeks and then the hard work begins in terms of marketing and getting customers and so on, which may be even harder because of there's so many things. Right. And so I think, and then the final thing is, look, there are humans involved in all this stuff and we don't like change. Generally we sort of a little bit slowly and you, especially if you're running a big company, you got to bring people with you. Right. And so I think it's going to be some elements, some governor. Mike: Yeah, right. create differentiated value that you can sustain right that's Pete Reilly: all this change, it's just the nature of change and how to bring everybody ⁓ I think it'd be interesting to maybe go around and do whatever side of the argument you want to take in terms of the pro or the con or the optimist or the pessimist. And maybe you go around the horn. Stew, where are you on all this? Are you on of the more the AI pessimist side or optimist side or somewhere in between? What's your perspective? Stew: I think it's gonna be, ⁓ I'll probably get way too, you know, philosophical. I long-term it'll probably be really good if we make it through the short term. ⁓ But I think it's gonna be very disruptive. I think there's a lot of other things happening in the world, you know, where kind of the probably the, Pete Reilly: Thank Stew: the old way of doing things, the old world order is breaking down and will have to be replaced by something new, which is something that throughout history seems to happen every 80 to a hundred years. so I think it's you know, it's a scary time, right? And, you know, it's, it's ⁓ a of are likely to change. I look, you know, the last time we went through something this large, was in the early part of the 1900s. And the end result was you got to the 1950s, you know, and 60s, 70s, 80s, 90s, you've got unprecedented wealth and prosperity and peace in the world for a long period of time. But, you know, the world kind of went through a rough spot there for, you know, the transition period. And that's more what I worry about, I guess, is how all of this plays out. I said, answering a bigger question than you asked. But I think that's what I worry about is just of what is that transition going to look like at a macro level. Pete Reilly: Shanti, how about it? Mike: Yeah. Shanti Greene: Yeah, would say, simply, I think there's going to be some pain, and especially in the next two years or so, where we see a lot of these things hit people adjusting quickly, not knowing exactly what to do. will probably be job losses in many different sectors as companies reorganize, get more efficient. And when you look at a lot of large enterprises, there is a lot of redundancy. And you have to wonder, well, if... All you did was keep half of the number of people because they're actually three or four times more efficient. that okay? Shareholders public companies would be happy, but it won't be good immediately. And you have to see what happens. Like what's that transition going to look like to Stew's point? How long will it take for people to make changes and reintegrate and try new things? And what are they going to do next? And I don't think those questions have easy or straightforward answers. Pete Reilly: Mike, bring us home. Mike: All right, yeah, so look, I think that I'm going to draw two sort of, you know, humps here for you. So let's start. Let's do like a socioeconomic hump first, right? I think the folks at the bottom end of the socioeconomic scale are going to benefit off the charts. It's going to be an amazingly important for education, for access, for You know, I've got a doctor, I've got a lawyer, I've got a counselor in my pocket, right? Huge change in their life for less than a dollar. Not going to rip it out of their cold dead hands. It's be that important. High end of the spectrum, they're the people that own the shares in the companies. And again, they're going to benefit a whole lot. We're talking about the middle squeeze here. The middle squeeze ⁓ is the % of the population, the half of the knowledge workforce that has to learn psychology because that's how you program now. Pete Reilly: Thank Mike: Right? mean, an LLM is programmed by convincing it to do the thing that you want, right? Not by programming it correctly, right? So that's going to be the squeeze. Some of those folks are going to change jobs. Some of them are to retire. Some of them are going to become artists. And I hope all of them get fulfilled. other hump that I would draw is the timeline one, right? So think right now we're living in this sort of curious abundance period where, you know, look at everything I could do with Claude Code yesterday. I think in 10 years, that will be the case. It'll be widespread and those resources will be very, very accessible. But that bump in the middle is going to be a massive period of development for things like robotics, that use this stuff, all of the ⁓ supply we have with silicon, with power, ⁓ with centers. So there's going be lot of contention for these resources in that middle period. But again, we hit that period of abundance at the end of it. So that's the way that I'm seeing it, the way that I'm advising young people that are in school, the way that I'm advising my own family. Pete Reilly: Awesome. Stew: I think by the way, there's going to be an interesting, and I was talking to my daughter about this last night. I think there's going be an interesting side effect of all of this stuff, which is, local community in human interaction are going to grow dramatically in their importance. ⁓ I think, you know, ⁓ talked about slop and code. Mike: Thank you. Shanti Greene: Yep. Pete Reilly: Thank you. Stew: The real slop is in video and all these, you know, we're like, what do you trust? Right? You can't, you know, you could make us say anything very convincingly with video now. So I think people are people will really be able to trust in their day to day lives will be, you know, what they can see and feel and interact with locally. And I think actually like. it's going to be an interesting swing back, I think, to heavy importance on local communities and then probably trusted online communities, Like groups of people that somehow you know and are trusted and can verify will be an interesting kind of side effect of a lot of this. Pete Reilly: Yeah. Mike: Yeah. Pete Reilly: Wouldn't that be sort of amazing thing if this is the technology that gets us to put the technology down and really spend a lot more time together. All right. Well, guys, to see you all. Have an awesome weekend. Looking forward to doing it again next week. Mike: Thank Shanti Greene: Absolutely. See you. Bye. Mike: See you guys, bye. Stew: See you guys. Yep.