KP Unpacked

Construction Is Now the World's Bottleneck

• KP Reddy

Share episode

This episode is a reality check for anyone who thinks construction is just catching up to tech. It's not. Construction is now leading it.

In this episode of KP Unpacked, KP Reddy and Nick make the case that design and construction have become the single most important constraint on technological progress. Data centers can't get built fast enough. Housing can't scale. Power generation is racing to keep pace. And for the first time in history, construction is facing technology-driven upgrade cycles, not aesthetic ones.

But this isn't just macro. KP walks through live experiments with Claude Cowork and Claude Code: automating LinkedIn grooming, generating $7K in Substack revenue, replacing million-dollar consulting contracts, and sending 1,000 personalized emails in under an hour. The breakthrough? AI agents don't need APIs anymore. They're reading screens and controlling desktop applications, which means on-screen takeoff, Revit, and legacy construction software are suddenly vulnerable.

Key topics covered:

  • Why on-screen automation could kill 50+ construction tech startups in the next year
  • How AI agents control your desktop by watching and clicking, not integrating via API
  • Real experiments: LinkedIn automation, competitive analysis, email campaigns, vibe modeling in Excel
  • Why construction is the bottleneck for AI infrastructure, housing supply, and energy distribution
  • The shift from trickle-funding to big bets: why seed rounds should be $15–25M for real problems
  • How to get surgical about ICP definition using AI-powered research
  • The 48-hour email delay hack: protecting your time when automation makes you too efficient
  • Why sales-oriented, variable-comp businesses are ideal for AI leverage right now

If you're a founder building in AEC, an investor trying to understand where capital should flow, or an operator wondering whether your software strategy is already obsolete, this episode will reframe how you think about the next five years.

Listen now.

BuildingWorks & Brookwood Sponsors

SPEAKER_01:

Hey Nick, how's it going?

SPEAKER_00:

KP, what's going on? Haven't seen you in hours.

SPEAKER_01:

Yeah, I know. I appreciate these our startups that recognize that some of us are on the West Coast. My 5 a.m. board meetings with our Israeli founder startups are a bit exhausting.

SPEAKER_00:

It's brutal.

SPEAKER_01:

It was it was refreshing to have a 9 a.m. my time board meeting.

SPEAKER_00:

Oh yeah.

SPEAKER_01:

That's pretty nice.

SPEAKER_00:

Yeah. Yeah. Must be nice to also have to avoid the snowmagadon that came through the East Coast. I've had foot on the ground and zero degree temps for a week. It sucks.

SPEAKER_01:

Yeah. You haven't been to our Half Moon Bay headquarters, but we have our we have our own roof catwalk, and I put some outdoor furniture up there, and I sit up there and uh do a lot of my calls, a lot of my Zoom calls from up there. It's the best. Love it. It's like 64 today. I had to put a sweater on.

SPEAKER_00:

So wow. Rub it in. Yeah, I mean, I'm excited to chat today. I think we've got a few good good topics. I did want to plug a by the time this this airs, the the fight will have officially ended, but I wanted to at least mention your your fight night promo that is happening with Dustin Devan, which is a debate on the value of pre-construction software.

SPEAKER_01:

Is that like the overall topic? Okay.

SPEAKER_00:

So Dustin Devan, uh former founder of Building Connected, his new company is called Edify. Is that is that right? Yep. Which is a a pre-construction, I think agentic focus pre-construction software. And and your your opposition to his point of view on the value of pre-construction is is what let's let's tell the audience what your argument is.

SPEAKER_01:

I mean, I I don't think it's unique to him, right? I mean, look, it it's gonna be a lot of fun February 3rd if this comes out in time, but it's gonna be a it's gonna be a lot of fun. Oh, like I said, like I think he and I both agree on a lot of things. It's not that we disagree on everything.

SPEAKER_00:

I think our pre-fight.

SPEAKER_01:

We want to but I think our approach is very different because he's very GC friendly and I'm very owner friendly, society friendly. Like, how is this helping us build more schools better and giving kids a better education? Is more my thing. And he's like, Well, no, it'll help. But I but I think there's a point that we both, and I think it's a good segue to some of the other conversations, is so what? Who cares about pre-construction? Right? And actually, Google Gemini just put out their new computer vision version yesterday, and there's a company that they backed, I don't think a lot, I think it was more just for press release purposes, that does fully automated plan review and quantity takeoff and estimating. Like it's what it's supposed to do. I don't think it does per se, right? But so do we believe a year from now, Google Gemini, you're gonna upload your drawings and get a an estimate? I mean, probably.

SPEAKER_00:

You think?

SPEAKER_01:

I think, and so that vibes.

SPEAKER_00:

That would kill a lot of a lot of startups in our space.

SPEAKER_01:

Oh, like 50 of them at least.

unknown:

Right.

SPEAKER_01:

And so I think the argument is really that some of these tools are a big so what, and you actually have to stop being like you know, a painkiller or a vitamin. You have to be like the cure. You've gotta really you've gotta really solve a real problem. And is there a problem in pre-construction? Probably, as a tool set, however, and you know, and I think unless you don't follow me, people know I've been like night and day with cloud cohort, right? And I have two machines running, pretty much 24-7 running cloud cowork. And and the reason why this is relevant, right? So cloud co work plugs into your desktop, it runs on your desktop, it also interfaces into software applications where it does where there is an interface, and then it logs into your web applications. Now, when it logs into your web applications, it's like reading the website, i.e., if you log into Bluebeam Cloud, it's looking at the screen just like you are, and it's reading the buttons just like you are, and it's clicking on those buttons. So I'll give you a good example. So, what I do is I get it running and then I watch the dialogue button. So, not meant to be a flex, it just is. I am out of LinkedIn connection, right? So every day I have to make room for 10 to 15 LinkedIn connections, otherwise, people can't connect to me. The people I want to connect to me cannot connect to me. So I every morning I go through and groom my LinkedIn and disconnect from people that maybe are no longer relevant. And so I said, Cloud, I told Cloud Code, this is Cloud Cowork, this is my problem. It said, consider disconnecting from retired people. I was like, that's a good idea. I was like, don't disconnect from all the retired people. So it logs into my LinkedIn, it prompts me for my username and password. It keeps it, by the way, a little bit of security risk. It's holding on to my username and password. It logs into LinkedIn and starts searching for all my connections that are retired, that have retired in their profile, some things up, right? And then it then it says, Would you like me to disconnect from all of them? I was like, No, I would not. I'd like you to disconnect from 10 as a test. So then it goes in and starts disconnecting. About the first one in, it says, Oops, literally says oops in the dialogue box. I accidentally recommended the connection versus disconnected it. I and it says something like it doesn't say I got confused, but I misunderstood. Let me fix it. Then it says, I now disconnected it. It appears as a second-order connection. So it must have worked. Would you like me to do the rest? I said sure. And it went through and did it, right? So it's it's reading the website, right? Whereas in the Excel, in the Excel plugin or some of the other tools where you plug in locally, it's actually running on the back end. It's less of the visual aspect. It's actually plugging in to kind of think about like the API, right? And so it went through and it's like, would you like me to do more? I'm like, sure, go ahead and do 50 of them. It's like on it, right? It went and disconnected 50 people that were retired, right? Which also means I can do the opposite. I could say, hey, I'm looking for 10 founders in the construction tech space, go find them on LinkedIn and connect me to them. Could absolutely do that, right? And I've got it connected to Clay, which is an email mining tool. If it if someone has their LinkedIn set up where you have to enter in your email address, it will go find the email address and enter it in and keep going. So good news, bad news, you're gonna get even more spam on LinkedIn. Because I'm not the only one that's doing this. So what does that mean? So what that means is if you're running AutoCAD locally, if you're running Revit locally and not in the cloud, in theory, it can take it over. So I haven't tested yet. I asked, you know, we had one of our mastermind group meetups this week, one of our Zoom meetings, and I asked some of the folks on there, if you have AutoCAD or Revit, see if you can tell it, like, hey, here's what a fire a fire station looks like. Design a fire station for me, map out all my pipes for this water treatment plant. Here's an example. I haven't tested that yet because I just I'm not allowed to have software like that on my computers. Maybe I should go buy Revit just to play with it, right? But but so that's kind of a big deal in the sense of like what I'm in a debate about, right? If I'm already running on-screen takeoff, why can't I just command and control it from cloud cowork?

SPEAKER_00:

Yeah, I mean, the only reason you wouldn't is if it didn't have the ability to integrate connect via API or you know, mcp with the platform that you're using, or it didn't understand drawings.

SPEAKER_01:

Right. But so what I'm saying is you don't have to have mcp. It's just gonna look at the screen. Literally, it's on-screen takeout.

SPEAKER_00:

Oh, I see. So if you have the application open, you're saying it's gonna just be no, it runs.

SPEAKER_01:

Yeah, it takes over. Yeah, yeah.

unknown:

Got it.

SPEAKER_01:

I crashed it the other day because it opened up 1500 tabs on my Chrome.

SPEAKER_00:

Yeah.

SPEAKER_01:

Just used up all my memory and said bye-bye.

SPEAKER_00:

It would be interesting to test out the quality of an application that is open, but it's not integrated with versus a pure integration via API or MCP. Like I'd be I'm curious if it's any different, the results.

SPEAKER_01:

So my argument might be I mean, uh, and if you're listening to this and have some technical capability, please go play and let us know. Right. I think this is important stuff. It's kind of the future of our industry stuff, right? Because one could argue the problem with APIs is what the publisher of the software only allows you to access certain functions and data sets via API, right? It's kind of limited. Whereas if it's like logging in and can control the screen and is reading the screen and getting better and better at reading the screen, I don't want to, I don't need an API.

SPEAKER_00:

I mean, related to a tool like Revit, I mean, my assumption is it's all about the quality of understanding. Like it doesn't need to understand how to use a Revit, it just needs some to be able to read what's on the screen, which the capabilities are increasingly better by the day, and they're already pretty good.

SPEAKER_01:

Well, and I I didn't realize too. The only thing that's a little bit of a negative is I'm spending a lot of time, so my cloud co-work is better than my colleagues. Because I asked someone to say, I actually asked Devin, I was like, hey, redo this PowerPoint, right? And he tried to do it, and it didn't come out as good as mine did because he hasn't taught it. So it keeps the learning very local, right? But think about this, then I'll start playing around with with Zapier, right? Zapier is not a hard tool per se, right? I said, hey, I want these two applications connected, and here's the data I want it to pass. And it went logged into Zapier and went and logged into all the things and created the workflow. So my point about the API stuff is MCPs rely heavily on APIs, but this is not, this is like maybe it's beyond APIs. And so like I said, I don't have access. Maybe I need to go buy these tools and take the work off and display.

SPEAKER_00:

I think if you're if you're looking for analysis, which like take off you know, quant quantification materials take off cost estimation, that's like an analysis-based thing when the when the when the drawing is done and you know the the the project plans are are complete. It's not like it's not going into your computer and actually doing the work for you. Like it's not doing the drawing itself. It's not doing we don't know. It could. I mean, in in the future it it it very well could, but it's not doing that in Revit today, right? Like it can't. I mean, I don't know. I would imagine like that and without an integration, it just can't. Yeah, like it can't.

SPEAKER_01:

Well, I think what I'm saying is with an integration, Autodesk isn't gonna let you anyway. Right, right. Right. So the interface is all the buttons and menus.

SPEAKER_00:

Yeah, and maybe Revit's a you know, Revit's just one example.

SPEAKER_01:

Like, so the cowork is watching you and says, Oh, here's how you name your files, file new. Here's another example, and we'll talk about ClaudeBot. Uh, what is it called? ClaudeBot. They got me Moltbot for a second in a minute. But I have this issue being diabetic, is the big insurance companies always want to deny me, and I have to go to my medical provider to say, can you send the letter to CBS? Right. So I end up in my my care provider's portal, I end up in CBS's portal, and I end up in United Healthcare's portal. And all I'm doing is getting them to do each other's jobs, right? Please send this over to CBS, but I automated that. So when I get rejected by CBS, I hit a button and it notifies it logs to my physician's portal and says CBS is requesting this and sends a message and then waits to hear back from CBS.

SPEAKER_00:

I mean, that's what the drug companies, the pharmacies should be doing, right?

SPEAKER_01:

Right.

SPEAKER_00:

I mean, yeah, I mean, like that's a workflow that should just exist for them. Like if you if you if the one Walgreens is out of stock, like go to the one down the block and make sure it's ordered. Right. Except they physically call them and yeah, 100%.

SPEAKER_01:

Right. So think, I mean, so my point is like if you if you start thinking about automating your computer, right? It starts to be really and to me the breakthrough is that it's on desktop applications as well as cloud applications. So all these people that are building software to AI, like we're like Bluebeam, but we're AI. We're like OST, but AI. If it's not happening today, I mean in six months, I think they're all done. I think they're all done. Because you're basically just saying I'm gonna run a little robot right on top of my screen. It's like in San Francisco, they have this robot that makes coffee called Cafe X. The robot's just an eight-axis robot, there's nothing special about it. The coffee machine, there's nothing special about it. They're using all existing equipment to make you coffee, right? There's nothing special other than the enclosure. So if you think about that applied to software where I'm taking I'm basically like taking stuff that exists and automating it, there's two things that happen. One, you already have a lot of that information. So the big thing about OST is you have all your cost data from all your previous jobs already in there. So I think it's a big breakthrough that is somewhat being overlooked by people. I say, I mean, honestly, maybe I have too much time on my hands. I'm probably spending two to three hours a day on cloud code work, but it's probably gaining me like so much time.

SPEAKER_00:

I mean, we talked about I think two episodes back, how while it's saving you so much time, you're just it's saving you time to do more and more things. It's like the workload actually increases. Like your work hours aren't necessarily being re reduced, but the tedious work is being completely automated. I for what it's worth, I haven't I haven't spent a ton of time in in co work, but I've done I've I'm spending like a lot of time in cloud code. To me, it's just like easier to choose one. And Cloud Code does everything that co-work can do, just in the in the terminal interface.

SPEAKER_01:

Yeah. So I think so. Here's a good, but but I think besides like saving me time, it's saving me a lot of frustration because I can tell Cloud Cowork something two or three times, and then it does it perfect. That's not how that works with people.

SPEAKER_02:

Yeah.

SPEAKER_01:

You know, so or it's like, I mean, think about this, you know, we've both worked in scaled organizations. You ask someone to do something and they say some nonsense to you, like, how would you like me to prioritize that with all the things I have going on?

SPEAKER_00:

I mean, I will say one of my one of my takeaways from how excited you are about cloud co-work and cloud code and the I'd I'd say maybe in the last, it feels like in the last quarter, like you got you got pretty jacked up about these tools.

SPEAKER_03:

Yeah.

SPEAKER_00:

And you know, my experience working with you is you're a fantastic delegator. Like you are like you very like you do not need to be told, like, hey, I need to offload this project, I need to get this work off my desk. Like you instinctively know that you need to do that. I think it's just like your your background having run many companies. Like if something piles up on your desk, like you're offloading it very quickly. And if you don't have the resources to do it, you're hiring someone to do it. You don't think twice about it. Yep. And so as someone who has that instinct naturally, you're you're now replacing that instinct with, I'm actually just gonna automate this and not have to spend money on another resource and not have to train someone else and educate them on my process and what I like and my workflows, et cetera. Like I've even seen you, your instinct of like all from an automation standpoint, previous is like, let's go build software to do that. Like we did that at we built software at Shadow, you know, Matt built a couple different tools for us that I think, you know, effectively, like, you know, if you were running running cloud code, like you you probably would be able to get, you know, a decent amount of the chunk of work that we we did previously. And so yeah, it's it's a the observation that I've had is you're excited to delegate work to this, you know, this code, this coding agent, this, this, this coworker. It's not painful to educate them or train them because you know they're gonna get it right once, you know, once you spend a little bit of time with it. And it compounds. Yeah. So like the more and more information you feed it, the more and more work you give it, the more it's actually gonna get, you know, incrementally better over time. And it's not just like an obsolete, you know, a tool that was up obsoleted, you know, when a better software came on, came along. It's just like let's quickly rebuild it. Let's like let's iterate the workflow with no marginal cost increase.

SPEAKER_01:

Yeah, so it kind of opens up this other conversation. If if you told your 20-something year old what business should you go get into, right? I would say anything that has variable comp. Like if I had the bandwidth right now, I would be an investment banker. Because that's variable comp, right? I had a friend of mine who you know who's a CMO of a very large company. He's paying one of the big consulting companies a million dollars a month to do a competitive analysis. He sent me an old one they did a couple years ago. He's like, hey, we like the format. I'm like, great, let me turn it on there. It went out and basically did an updated competitive analysis for them. As good if not better, right? Because it actually did a better job of citing it's where it found things because people are terrible about citing where they find things. This was actually better at it, and he was like, You just replaced a million dollars a month that I'm paying this firm. This was in the evening. I maybe spent an hour and a half or two with them. And he, you know what his biggest issue is like, dude, like, thank you so much. I'm like, Well, don't you do this yourself? He's like, Oh, my company won't let me buy a Mac and use the software. I was like, Go out and buy a Mac for God's sake. Because right now it's only running on Mac, right? I was like, for God's sake, go buy your own personal computer and do this stuff yourself. And I think I wrote an article about hey, IT might be your biggest obstacle to using these tools. One you can't have a Mac, you need to turn this off and turn that. Like, no, just go get your own Mac. It's your personal device, right? Just go do that. And but oh it's like super fast. So you start looking at the things where there's asymmetry in the work, and I mean, I was like, Chester, like a real estate, like a residential real estate business, because I can probably Predict people that are like selling their houses. I can I mean all the things, right? And I can send out automated. I had to do this one thing. I took four people on my team, I loaded their bios, I loaded a thousand active projects, construction projects, and I told it to find the contacts on these projects, match them with the right person, write a customized email from that person to that prospect about their project and their background and how they thought they could add value to the project, and told it to email it. Thousand emails in under an hour. Very custom, you know, and and and very much like, hey, I'm really interested in your project because I live in this market. And like, I mean, that correlation of like high personalization was massive. Now I didn't it it took me an hour because I wanted to approve, I wanted to read every email before it went out because it was a new experiment. So I was like just like little pattern recognition, hitting send send send. Yeah, but I I don't know, and and the and the team said like half of them came back was like bleed. Crazy, right? So I think that's the key, right? What are the businesses that I mean, because let's not try to invent a new business. What's the traditional business that you can go get that the rewards and the effort are it's very rewards-based, right? Real estate, investment banking.

SPEAKER_00:

Real estate's a great one, yeah. I mean, any sales-oriented activity, like large enterprise selling. By the way, another hilarious observation I have about your activity is your first instinct to go automate stuff is lead lead generation tools.

SPEAKER_01:

Yes.

SPEAKER_00:

Just like in your DNA.

SPEAKER_01:

Sales is everything. I will argue with people all day long that will want to argue like you've got to be able to deliver, like, you've got to be able to sell. Sales drives everything, and people that want to argue with that will get nowhere. Sales drives everything.

SPEAKER_00:

Yeah, and I also think like your your biggest when it comes to sales, what you know about how to drive momentum leads, and then ultimately, you know, conversions at the bottom line is through through your network and top top of funnel, and you're always kind of trying to activate that in yeah, in the most automated way possible. A lot of lot of lot of a lot of data to sift through there, but yeah, you're always tinkering with that.

SPEAKER_01:

Yeah, and I think so. The other experiment I did, it made me$7,000 yesterday. Do you want to hear about how it made seven thousand dollars yesterday?

SPEAKER_00:

I'd like to make seven thousand dollars today. Tell me.

SPEAKER_01:

Well, none of this stuff happens overnight, right? So a little bit building on the assets you have. I downloaded all my Substack users, all my subscribers. It's about a it's about 95,000. You pay, I think it's like 1,200 a year to subscribe to my Substack, something like that. So I had it review all the people that became subscribers and people that were potential subscribers and which posts they were most active with. And I asked it to rewrite up some topics that would push conversion of my um my free subscribers that are most active to pay it. And it said you should write about one of these three topics, and then I did seven grand.

SPEAKER_00:

Nice. Did uh and everyone got the same piece of content? It didn't it didn't differentiate the content based on what they're okay, got it. Cool, that's exciting.

SPEAKER_01:

So I mean that's I think I think this is the fun. I mean, the fun part of this for me is running the I mean, even if they don't work, yeah, for sure, running the experiment. But my point is like I think the people that are gonna follow down that are like, oh, I'm too busy. It's that I'm too busy thing, you know, oh I'm too busy, oh I don't have the right machine.

SPEAKER_00:

What's the number one thing if you're advising a C a CEO of an engineering firm or construction firm on how to on on where on where they should be spending their time with it with AI and like immediate practicality and their in their daily workflows and where they would drive value? So like they they may not care about you know activating their subsec audience, but they might be looking for ways to increase enterprise value for their companies. Like, how would you where would you tell them to spend time?

SPEAKER_01:

Well, I'm telling them to this has been in the conversation because all these people see my stuff and some of them reach out, like a lot of our LPs have reached out to me. Like, hey, explain to this, what are you doing, right? The first thing I'm recommending is free up your own time. And what we're learning is a lot of these VPs, they spend so much time managing. My CFO needs this, my CEO wants this. So I walk one of them through log into your dumb Dell Tech, sorry, Dell Tech, log into your dynamics, all these things that you would have normally used Power BI or Tableau or any of these tools, and just tell CloudWork to go get you what you need. Say, build a report. Here's the bio of my boss and their side program. You know, so we'll tell you a little bit about some of the other things we're doing. But I was just like, this is from my boss. My boss just asked me what our IT spend is compared to last year, and where should we be spending and where should we re be reducing? Perfect. Just let it run. As Claud worked good, logs into dynamics, looks at your budget, looks at your budget that's sitting in Excel. Boom, boom, boom, charts and graphs, charts and graphs. Here you go, boss. 15 minutes later. The the hack I try to tell all these people do not email it right away. Put it on a 48-hour delay because your boss needs to feel like he spent like what if you put it on a 48-hour delay, you just saved your you freed up 48 hours of time that your boss thinks you're working. Because if you send it too quickly, they're just gonna give you some other nonsense to go help me understand this, right? So you have to put your email on a time delay to free up the time.

SPEAKER_00:

The perverse incentives around completing work with an agent.

SPEAKER_01:

That's a golden hack though, right? I mean, I feel like I feel like that's a good one, right? They're like, oh yeah, I got to my boss in 15 minutes. I'm like, no, no, no, no. You're not gonna free enough time, you're not gonna free up time to tinker on things. I'm not saying, you know, if yeah, if you want to go play golf, whatever, right? But for me, it's more like okay, you free up more time to have space to keep tinkering with code. Right, right. That's why you put it on time delay.

SPEAKER_00:

You mentioned something that I think the it'd be good to just yeah, mention this for the audience and just another tool to go tinker with. So inside Claude Claude Cowork, you can now, there's now Claude for Excel. And there's a new phrase. We've talked about vibe coding on this podcast before. There's now a new phrase called vibe modeling.

SPEAKER_01:

Okay.

SPEAKER_00:

You can you can actually just tell tell Claude the things you want to model. You can do it directly in Excel or download the Claude desk desktop application and and actually just tell it exactly, you know, in natural language what you are trying to financially model out. So I played around with it a little bit myself, but I like if yeah, whenever I have a free uh free couple days, I'm gonna like truly try to get in the weeds of that because it looks fun.

SPEAKER_01:

Yeah. Another exercise for our startup friends out there is you know, we limit our thinking of an ICP to one or two, right? Because it's like, oh, well, you can't do more than one or two ICPs. I'm I'm not saying you you you be all over the place, right? But I think we, if I'm a founder, I want to start thinking more granular about ICP. In other words, it's pre-construction, pre-construction, a GCs under 500 million. The head of pre-construction is a in their mid 40s, likely male, drives a truck, like a deep, deep persona ICP on your on your actual target, right? Because now all of a sudden, because you know, so we we see the tam same sums from people all day long, right? And we say, Well, what's your first market? And they'll say something like data centers. I'm like, well, that means a lot, that means a lot, right? That can mean a lot of things. And so I think because of this, like what you're talking about, like vibe modeling, right? I think you can start, you know, being more of a sniper, except as a sniper, you've got, you know, 50 snipers out there, right? Versus trying to blow up the block, right? Versus like scorching the earth, right, to try to find customers. And I think that's a behavior that I think is really challenging for a startup because I think we challenge them all the time. Like, who's your target customer, general contractors? It's like, yeah, that doesn't tell me anything. That tells me nothing. Or the construction industry is a$64 trillion business.

SPEAKER_02:

Great.

SPEAKER_01:

What's the first$10 million? Where's the first$10 million coming from? Well, we we can't do that level of work. And as investors, we don't expect them to do like that Bane level copy of you know type of work because it's like expensive, it's time consuming. So I think we have to start as investors to start demanding because they can do more. And if they're not gonna do more, if they don't know how to use the tools, do we want to invest in a founder that's not using these tools to be deeply analytical? When you have such limited resources, don't we only want to back a founder that's being very analytical about where they want to expend resources and run experiments?

SPEAKER_00:

100%.

SPEAKER_01:

So I think we have to be more demanding. You know, it's like the thing that I always you know ask founders like, which two trade shows are you gonna go to next year? And they're like, oh, well, you can't just go to two, we got to go to a hundred. I'm like, and you're raising a billion dollars. Like, you can't do it well. What are the two? And how are you gonna spend money and do it well, right? Like that kind of stuff. So I think from that perspective, I kind of feel like founders are really busy with cloud code. Fantastic, right? Fantastic, right? But now things, but that's for the hackers, right? I mean, the whole hustlers and hackers, right? I think the hustlers have to get on the on board with cloud cohort and start really focusing on our markets, where are we going, all those things.

SPEAKER_00:

You said the magic word, which means a transition in this conversation. You said data centers. So to transition to the other topic here, you mentioned to me that you think the AEC industry in construction right now is the biggest bottleneck holding up massive technology change in society. Let's hear, let's hear the thesis expand on it.

SPEAKER_01:

So, what I'm giddy about, right? Which you know what giddy looks like on me. I smile. So, giddy to what I'm giddy about is having been through several technology cycles, there's never been a real reason to innovate construction, right? Oh, we need more offices. So what, right? Okay, it doesn't keep you from growing your business. We now have uh in the biggest technology wave, the biggest bottleneck for growth, which is data centers. Two components to that. I mean, there's there's the what goes in the data centers definitely a problem, right? Whether it's the HVAC systems, I mean, yes, chips and all that, but that's not what I'm talking about. Although there is a component to that. So we need to build more factories to make more chips that go into the data center. Perfect. We need to build more data centers. We need to build more H build factories that can build more HVAC systems that go into the data centers, right? And oh, by the way, oh, these data centers need energy. We now need to go build more power generation facilities and distribution facilities to feed these data centers. So the only people that are holding things up, because clearly capital is not a problem, is our people designing construction. We are the single biggest bottom for AI. That means we're very popular, Nick. We're the cool kids all of us.

SPEAKER_00:

Us construction kids are now. Let me just add some fuel to the fire here in terms of the importance of our industry. So so that's that rings true. I feel like that is completely accurate. Like the most important thing in the world of technology is growth data centers, and the limiting bottleneck to growth of data centers is construction, construction of them, permitting, integration with the grid, energy, et cetera. Right. The other biggest problem in society right now is housing. Yep. And where we where you see that bubbling up is in politics. And you know, the it, you know, don't need to be dead dead horse. I think anyone listening understands the you know, the problem with basically the lack of affordable housing in desirable locations. So, you know, if you want to live in San Francisco, if you want to live in New York, if you want to live in cities that are attractive to, you know, call it, call it young, young families in the millennial generation, it's like it's you know, it's really hard to afford a home. And so how do you build decrease, you know, decrease regulation, you know, basically increase the increase the supply and then also reduce the cost of construction, which from a thesis standpoint for us, that's like the game we've been playing since we started Shadow is like how do we decree, like we can control if if we can control the cost of construction and bring it down as much as possible. We can't control regulation. Like we can do, you know, we can lobby, we can, you know, be involved in government, but like at the end of the day, it's not going to be like that's not our game. Our game as investors is to support the founders that can then go and and and lobby and are actually building the solutions to you know bring the cost of construction down. And so those two narratives are arguably the the most important things to solve for in society. I'd say top five at the at the minimum over the next five years, maybe, maybe the deck, maybe the decade, forever, right? And those are literally the only things we're investing in.

SPEAKER_01:

And I'll add on to that, right, is now we have these new factory towns, right, for these AI factories that are being put in the middle of nowhere, because in the middle of nowhere is where there's power, right? Abilene, Texas. People, I mean, it may not be thousands of factory workers like Hershey's Pennsylvania, right? But it's still a lot of people, and they're not gonna live in trailers. This isn't fracking. These are people with families, and they need restaurants and they need mail salons, and they need little indoor play areas for their kids. Like it's so, yeah. I mean, yes, yes, right. So I think so. I I think we we are uh in violent agreement on this, right? But what does that mean to how we think about investing as as investors and adapting? And what I would say is we as construction/slash construction tech investors, I think we have to do two things. One, we have to work with founders to reinvent their business models because too many of them come from industry. I was at XYZ construction. This is what I observed. I built some software to solve the problem I was dealing with in my old company, right? The problem is that is just very rear rear window looking, right? It's rear, it's it's all hindsight. I think we have to shift those talented people and say, look, let's talk about the next 10 years. Let's not talk about what you did, let's talk about where it's going and shift them into bigger ideas. That also means feed rounds are not$500,000 and a million dollars. I think seed rounds are like$15 to$25 million, and series A rounds are like$100, like I think you have to do bigger things because the and I think the reason we sprinkle money in in our industry is because the cycles are so long, it takes so long to get anything done that it's like, well, let's just trickle them along, right? We're not going to bet big because they'll just waste all the money. Because the the absorption of this new technology historically has been very slow. But I'm telling you, if you come out and you're like, hey, I've got technology that allows me to build data centers faster, here's 100 million bucks, right? I mean, like, done. And so so I think there's that. I think the other thing we have to remember is the pace of technology is moving so quickly. The data centers we built last year are getting a renovation next year. I had a friend that back in the day, back in the 90s, they had a case company. They sold like white-labeled cases for you to build your own computers because that's how you had to do it back then. You know, not everybody could afford Dell servers. We had to build our own servers. Those guys made so much money because every time a new chipset came out, oh, need a different need, two fans, need a different heatsink, right? So there's a constant upgrade cycle. So I think the first time in our history as an industry, we're gonna benefit from an upside from an upgrade cycle. It's like my cousin, who I think you may have met, who is in the case business, runs a company called Casemake. Crash is it. Because every time there's a new form factor iPhone or AirPods or iPads, he gets a whole new inventory of SKUs that end up in the ATT store and Best Buy and everywhere else. Not because it's obsolete, but because it changed. Right. So now we're gonna we're gonna go through a cycle where, yeah, that data center we built last year needs a total new renovation. Like it's a year old. Yeah.

SPEAKER_00:

Yeah, it's about to accommodate NVIDIA's blackwell chips, GPUs. We need, you know, we didn't need an entirely new rack set up or you know, liquid cooling set up to you know to work with these new these new GPUs for sure.

SPEAKER_01:

And you know, you know, you know, Nick, usually I caveat everything I say with like while I'm a little bit biased. This time it's not about bias. This time I'm right.

SPEAKER_00:

Which part?

SPEAKER_01:

All of it. I am right on this. And you know, if we want to we should start doing debates on here, people can come on and debate with us. That'd be great. You'll be a balanced, you're you're a level-headed, right? But come on, let's let's debate it. I think we have reached an inflection in time where design and construction is the most important industry for our future for all these reasons. And our upgrade cycle, it's not gonna be because oh, that carpet in the lobby's out of date. Let's do a renovation. These are gonna be technology type upgrade cycles. The new chip comes out. Oh, turns out we need water, turns out we need some other solution. Turns out we need this, turns out, you know, it's just gonna be a constant upgrading of facilities to meet the demands of technology. And I don't think we've ever seen that.

SPEAKER_00:

I don't, I I certainly, I certainly, yeah, I certainly have it. The I think it's a good it's a good call out. I mean, I think essentially what you're saying is it's not it, like this isn't really up for up for debate, like it's actually just true.

unknown:

Yep.

SPEAKER_00:

And I think the only assumption that you're making that maybe some people disagree with is that AI is the most important thing in society. Like they may just not have that technology bias. But if you have that technology bias, Which but for for what it's worth, I think is silly. Technology drives all progress in society. However, if you don't have that bias, then like maybe you're not gonna agree that you know the bottleneck to that technological change is necessary. Yeah.

SPEAKER_01:

I don't so here's my thinking though is I don't think people have to agree with us on that. All we have to do is does anyone want to argue where the capital flows are going? I mean, the capital flows are still coming, they're gonna keep coming, right? I mean, so in many ways, like if the entire world is wrong on this, we're we're all gonna have a different problem. But I just I just don't see it. I mean, there's just too much, and and what's interesting is you've even seen now some of the thinking around headcount and hiring AI engineers is slowing down. And when you ask people why, they're like because we don't we don't have a data center, we don't have anywhere to push this stuff, right? We don't have anywhere to push it, and of course they're also getting better and better. I saw something, I don't know if it's true, I saw something on the internet where the reason why co work came out so quick was because Claude Code wrote it, which kind of makes sense.

SPEAKER_00:

Yeah, and which by the by the way, the I know I know that story. The the Claude Code, the inventors of Cloud Code, the two two anthropic engineers who who built it, they they didn't write any code for Claude Cowork. Like Cloud Code wrote the entire thing, 100% of it. And then and they did it in in less than two weeks.

SPEAKER_02:

Yeah.

SPEAKER_01:

Yeah, someone was asking me, how did Claude come out with a better plugin for Excel for Excel than Microsoft?

SPEAKER_00:

Really embarrassing. Did you did you did you see why though? Because they're they're working, they're creating a different, like the way the the way that Copilot within Excel was analyzing and working within spreadsheets, like it had to it, it it it was interacting directly with the cell, but what Claude for Excel is doing is it's creating a Python environment for it to analyze and then and then sp and then spit and basically do analysis outside of the work workbook.

SPEAKER_02:

Oh, interesting.

SPEAKER_00:

It creates its own work environment, so it's much faster and just can manipulate the data way better versus versus Excel is actually like the yeah, the the copilot version that that Microsoft has, it sent it essentially was trying to do everything within the sheet, which was super limiting. Interesting. Well, and then we so we're those we're we're we're important more important than we've ever than we've ever been. Let's and let's end the show on that on that note.

SPEAKER_01:

No, I think you and I've been talking about you know, next fund, next fund, what does next fund look like? And and I think the fact that we don't have a lot of clarity is probably a good thing. Yeah. Because I think things are moving very quickly. And you know, if we put a deck together and shipped it last week, it'd be wrong this week. And I don't know that, like I don't know that that's gonna change, but I do think we're in a spot to understand the patterns and say, look, the patterns are moving this direction. You know, and I think with the statement that design and construction is now the most important industry on the planet, control. And we have all the same issues, right? We're unable to we don't have enough talent, blah, blah, blah. So therefore, the scarcity that we have, the only choice is to build more stuff to create abundance of capability.

SPEAKER_00:

Yeah, totally, totally. Gotta stay, stay flexible, yep. Stay on the stay on the front lines of the tools. Bill Gurley had a had a quote I really liked. He's like, with all, you know, everything that's happening in AI, the only move is to run at it. Yeah. You just have to run at it at full speed. And I think like that's the only way to to yeah, to to know what the right choice is in any given moment in time because things are moving so fast.

SPEAKER_01:

Yeah, and I'll tell you, like, my family, I think on any given day thinks I'm insane. I'm up at 4 a.m., man. Me and Claude, we're up at 4 a.m. And I don't think I don't think you can really understand this stuff, just like you're doing Claude Code. I don't think you can really understand what's going on in the world unless you're in it. I don't think you can just sit there and read blog posts and read Twitter and you know, and that's one of the things I've been really coaching a lot of my executive friends is like, please go buy a map and carve out time in your day, weekends, whenever, and start getting to work. Because if you don't, you're just not gonna be, you're not gonna see the opportunities, right? Yeah, forget. I mean, I try not to be like, you're gonna be out of a job, you're gonna be homeless. Like, you know, that's not my job, right? My my what I'm saying is like you will not be in the know unless you're on the keyboard. There's just not enough you can read. You're reading and podcasts, including this one. Hopefully it's a motivator, but we didn't tell you anything that like is really that useful unless you go jump on a keyboard, right? You have to experience it. It's like I tell people all the time I can sit here and describe an amazing cheeseburger to you all day long. The how gooey the cheese is. I can describe everything you can describe about a hamburger to you right now, but until you bite the hamburger, you're not gonna know, you're not gonna get it. So go get a cheeseburger, everybody.

SPEAKER_00:

Good to attend on. I don't know where I'm gonna go from on. I haven't had McDonald's in a while.

SPEAKER_01:

I had in and out burger the other day. Okay, I can't eat like that.

SPEAKER_00:

Animal style. Huh? Yeah, yeah, I can't. Turn 55 next week. I bet a 55-year-old, you gotta stay out of in and out.