KP Unpacked

Your Data Is the Product and You Already Agreed to It

• KP Reddy

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 42:21

The Facebook moment just hit enterprise AI. Did you miss the terms of service update?

In this episode of KP Unpacked, KP Reddy and Nick break down why the Alex Karp CNBC interview landed like a bomb in enterprise boardrooms but barely surprised anyone actually building with AI. Construction company CEOs were getting texts from board members within hours: "Did you see this? What are we doing?" The answer for most of them? Running Microsoft Copilot, banning Claude, and quietly dealing with ransomware attacks that have already put subcontractors out of business.

KP walks through the apple pie analogy: buying a pre-made pie (frontier models) is cheaper, faster, consistent. Making your own (open source) costs more, takes longer, outcome uncertain. But here's the real insight: the question isn't open source versus frontier models. It's what data should you never feed any model, period. Then a Bay Area contractor says something that cuts through all the noise: these YC kids have no construction experience, no relationships, no reputation. If they take our data and screw it up, they move on to their next startup. What do they have to lose? That's not a technology question. That's a trust question. And construction figured out the answer decades ago when they started vetting subcontractors.

Key questions answered:

  • Why did the Karp CNBC interview send board members texting their CEOs?
  • What does "we're training on your data" actually mean legally and technically?
  • Is the apple pie analogy the best way to explain open source versus frontier models?
  • Why is ransomware quietly killing subcontractors before AI even arrives?
  • What does a contractor's subcontractor vetting process teach us about evaluating AI startups?
  • If a YC startup takes your data and folds, what does the founder have to lose?
  • Why did Claude's updated terms of service change the conversation?
  • What's the Red Hat playbook and why is Palantir following it?
  • How does Zero RFI's opt-in audit trail architecture solve the data trust problem?
  • Why are most AEC firms staying on Microsoft Copilot and not moving anywhere fast?
  • Should startups building on frontier models be worried about their defensibility?
  • Why does Karp get away with saying things every other public company CEO won't?

If you're a construction company trying to figure out what to tell your board after the Karp interview, a startup wondering how to build trust with enterprise clients around data, or an executive who just realized you never actually read those terms of service, this episode will help you figure out what you actually agreed to and what to do next.

Listen now.

San Francisco Reset And Sleep Rhythms

SPEAKER_04

Yeah, it'll be interesting. We always see each other three.

SPEAKER_03

Hopefully uh it's like a little awkward to AC because we're sitting a little close. Probably very my coffee breath might be a little much to the smaller.

SPEAKER_04

But I think this is a preamp to put in a couple of weeks. We're gonna do a KPM patch live, like actually live, live studio audience.

SPEAKER_03

Yeah, I guess everyone seeing this and listening will hear the recorded version. This is being recorded, but yeah, there will be an in-person event now that I'm in San Francisco. So why are you here? Why are you in San Francisco? We're hanging out. Okay. Brought the brought the family for a month. You know, reasons to be here if you work in the venture in tech world. I've heard I've heard, I've heard there's a lot happening. So yeah, I got the bug to come out here and spend some time and some, you know, if you listen to our conversations around the house, like we might be test testing out a you know, potential living situation here.

SPEAKER_04

Yeah, it's a little bit cooler probably than Kentucky.

SPEAKER_03

It's like 50 degrees cooler currently. Yeah, we had a we checked the weather out the other night. Feels like 46 was the okay. Was the temperature in Cole Coal Valley? Yeah. Foggy, cold. But we came from yeah, 105 degrees last week in Kentucky. I mean, we were right in the middle of the heat wave, like everyone, and it was a reprieve, was very nice.

SPEAKER_04

Yeah. I mean, that's what we enjoy here. Like we're down on the coast and like 400 yards from the ocean, and it's not bad. You have to kind of wear sweater some days during the summer, it's not the worst thing in the world. So I just spent a week in Atlanta. That was like a thousand degrees in high humidity.

SPEAKER_03

So, but it's been do you have any new perspective on Atlanta when you go back to this?

SPEAKER_04

Yeah, I just I think there's two things that are happening with me. One in living in Half Moon Bay, my commute to my office in Half Moon Bay is like three and a half minutes. Yeah, I've become very much country mouse. Yeah, like I feel that way in Kentucky for what it's worth. Yeah, that's commuting into Atlanta. Yeah, I was like, the traffic, yeah, and the big buildings and the parking deck, yeah, and like this is such a pain. Like, how do people live here? Well, I lived there for 53 years, but it definitely you get yeah, you get you get stuck in your routines and you get comfortable. And you know, here I get up at 4 a.m. and you know, the rest of the world has already started, right? I start getting my views in and do you wake up at four because you're you feel behind? You wake up later. It's a bizarre thing. So I mean, I don't know what it is. I don't know if it's with the startup with zero RFI or just AI or whatever's going on. I just wake up every morning with like massive enthusiasm and FOMO.

SPEAKER_02

Yeah.

SPEAKER_04

Like ready to take on the day, ready to take on the world. Devin actually asked me this morning. He's like, he's like, you just seem like you're getting up earlier and like getting up much earlier than you're right. He's like, Were you up at three? I was like, maybe, you know, like maybe I was. He's like, What are you doing up at three? And I'm like, I was just excited. Like, I have a big day, a lot going on in the world. Yeah, and I'm just excited to go be part of it. And he's like, Well, I mean, he's like, I guess that's you know, he's because you know, the whole worry is like, are you getting enough sleep and all that? I was like, I was in bed at eight last night, feeling I go to bed early, you know.

SPEAKER_03

So yeah, yeah. It's interesting that you kind of kept your Eastern, you're almost like on your Eastern cadence. Yeah. I mean, like, I've thought a little bit about it. I mean, we've we're three hours ahead in Kentucky, we're on Eastern time, and since coming here, you know, we wanted we're here for a month, so we were like, all right, we're not gonna do the delayed, like, let's stay on the old time zone thing. Let's get acquainted to Pacific time, let's go to let's get the kids in bed at like normal hours. And so still working through the kinks of that. Yeah, I mean, like, I it's it's it's interesting that to me that you've kept more or less the exact same sleep pattern. You're I mean, you're a couple of you're more than a year in at this point.

SPEAKER_04

Yeah, almost a year and a half, yeah. It's kind of wild too. Like I wonder if there's a biological element to that. It could be, but I mean, honestly, I'm also enjoying my afternoons, right? If you think about yeah, you know, we have a board member in Dubai, so I check in with him every so often. The world starts on Eastern time zone. Yep. And then honestly, around three o'clock over here, it gets quiet. Yeah, it's kind of nice, right? It kind of gets quiet. I go get a coffee, I go for a drive, you know, I'm still available, but it's it's kind of nice having those like afternoon. Yeah, nice. But it's what's kind of funny is when my four-year-old hears me up at five in the morning and he comes down, he's like, make me eggs, make me. I was like, it's five o'clock, go back to bed. It's not it's not actual time. You need a couple couple more hours before you before you get up. So speaking of, we haven't really talked about topic, but we can't whether you want to or not, you don't get to ignore Alex Carp.

SPEAKER_03

Yeah, you're your guy. I got a text message from you last week that said, I might love Alex Carp. I'm not exaggerating when I say the word love. I have a cr I have a man crush on him.

SPEAKER_02

Yeah.

SPEAKER_03

So yeah, let's talk about Alex Carp.

SPEAKER_04

I think I think the you know, his what's what's interesting is I get up every morning, I have

Alex Karp Sparks The Data Debate

SPEAKER_04

CNBC on, which by the way, I don't know what happened with CNBC. I think after they spun out to Versant and are part of the big NBC machine anymore, they've gotten a lot better. Interesting. They have a show that comes on at like 5 a.m. Yep, like pre-market stuff, and they're just like digging in, they're having better guests, they're doing some longer form interviews, and like all the interviews I noticed that.

SPEAKER_03

I think so on CNBC, I think they've really leaned into their advantage, which is we have great interviewers, yeah. Like world-class interviewers, and you know, even and so you have like Joe Joe Kay, and a lot of these guys have been doing this forever. These guys, I mean, they're truly world-class interviewers, yeah. And they're fun and they're conversational. And I think you see what's happening on social media, and a lot of the the traditional talk shows have not performed as well compared to some of the podcasters and some of the the social media influences that are that are out there. So I think they took a strategic look and were like, all right, what's our advantage? Yeah. Oh, it's interviewing and it's the strength of our strength of our hosts. Yeah. How can we lean into that? It's longer form, it's it's it's spending time with really high quality guests that no one else can get for 10, 20, 30 minutes, an hour. But those long-form interviews are great. Like they you can't really get them everywhere.

SPEAKER_04

And I think they don't disagree, they don't agree on a lot of things, right? There's even like, I think it's interesting because if it was purely political, it gets annoying after a while, right? Becomes left and right and all that. But because it's kind of like around business, yeah, and one of them has more of a left view, and one has a more of a right view, but it's not just around the politics of things, it's like how does it affect business? It's like policy and and that kind of thing. So I I think that they've done a really good job. I used to always say, I mean, I still won't listen, I still won't watch them in the evening. Who's that guy? Like, I won't like the some of that consumer stock picking stuff. I'd stay away from Framer, yeah. Like I can't too much. But so anyway, I'm so I have it on in the morning just to kind of hear what everyone else is hearing, right? And Alex Carp was on. So I actually listened to it live, and I was like, that was a great interview. Yeah, I had no idea, like then five minutes later, it's all over X, it's all over everything blowing up. But I think I showed it to Barry, our CTO, and he was like, his reaction was like very was very like, duh, like we've been talking about this for two years. He's like, Why is this this super surprise thing when we've all known this? Like, this is not a new dynamic, right? And why is it that he gets up there? And I think I told you one thing that I can't stand about what Alex Carp does is he takes commonly known words and technology and somehow makes them his own. Like all of a sudden he owns ontology. Ontology. Yeah, it's like yo, like was building ontology in Al 20 years ago.

SPEAKER_03

Like, what's great, you know, taking words out of the out of the dictionary and associating them with your company company as his product name, sometimes it's wild, right?

SPEAKER_04

But just to recap what Alex said, right? He said, look, you know, the large language models, the foundation models, they're basically stealing all your data. Look at what that happened with Figma, which, if you don't know, the CMO for Anthropic was on the board of Figma, resigned three days before Claude Design was released. And so he was basically making the point that if you're training them on your data.

SPEAKER_03

Well, the other thing with Figma, too, to point out is they had so CMO was on the board, but they also had a deep partnership with Anthropic and were helping them train their models, right? So they, you know, the agreement was that Anthropic is going to be how Figma develops their AI, you know, generative AI design tools. Right. And instead they ended up competing with them. Right.

SPEAKER_04

Right. Exactly. And so I think that was kind of one of the points he made. And this idea of sovereign AI, which has been talked about now for a long time, and as well as like the open source models, right? Which, you know, it's it's interesting because you know, most of what I know about technology I learned during the internet days, right? So it's all everything I look at is packets, how I think about it, everything is routing packets. That's just how I think about things. And so I think it was really interesting when he was talking about open source, because you know, we I remember the first router I made from scratch was with Linux. I took a machine, put two network interface cards in them, ran a kernel of Linux to make my own router because he couldn't just like go out and buy a router on the internet, right? And I remember a lot of the infrastructure as we grew, we had a problem, which was we couldn't find people that knew Linux. So we ended up moving over to Cisco systems because there were Cisco engineers that knew Cisco's interface. We moved our web servers from Apache to Microsoft ISS because we could find a bunch of Microsoft engineers. Yeah. And so it became less about the pride of like, hey, this technology is better. It became as I scale, I need to have people that know what they're doing. What's interesting is that's where we came from in the internet days. And then corporate America all of a sudden thought Microsoft invented the internet. Like, that's how that world went, right? What's interesting is of course people are going to use anthropic and open AI models and you know build stuff on them. But now we're talking about going the reverse, right? Now we need to move to open source, and it's easier said than done. And you know, another podcast that I used to love, all in podcasts, which I kind of hate them now because they're just in the link. They had a little segment on the Alex Carr brand, right? And I think the host there said something like, Hey, if you're a YC company and you're building your tech on top of these models, like you're done. Like you're you don't have a path forward, which may or may not be true, right? Which may or may not be true. But I think the the it's interesting how it sent this shock wave around, oh, what's happening with our data? And I think it's fascinating that it's like, oh, you didn't know. It's like the people, the early adopters of Facebook, you remember, if you're not paying for if you're not paying for the service, you are the if you're not paying for the product, you are the product. Yeah, people are still on Facebook. That's always this, right? So the question will be is like how many companies really will say, oh, let's go do open source. Not trivial to do an open source model and train it yourself and you know choose choose your own weights, right? Choose your own weights. Yeah, and and guess what you have to hire someone that knows what they're doing with that. Yep. And I think your CIO at a construction company has no clue. They're a Microsoft shop, they're happy with Copilot.

SPEAKER_03

And what I mean, what's interesting about you know, I mean, CARP's segment on CNBC, just to be clear, was a big promotion for what they're selling and what they're doing and how they're working with clients as a as a consultant, which is hey, we're gonna do a lot, we're gonna do all that for you. If you want to be sovereign over your own data set, which his point is like that's that's where the value in your enterprise resides. Yeah. And I think a lot of people listening probably would also agree that's where you know your enterprise value resides.

SPEAKER_04

Yeah. So you know, like I have all these like weird observational things, right? Because I'm pretty much just beyond AI pill, right? So my wife's like, Fourth of July, we need we need to have apple pie, because that's the thing. I'm like, okay. So I go to the grocery store and I get a pre-baked apple pie. And she calls me and is like, hey, like, don't buy one. You need to get the ingredients. I'm gonna make one. And so I'd already bought the pre-made pie. So I go back into the store and I get like the crust and the apples and all the stuff, right? I go back to the counter and say, Hey, can you exchange

The Apple Pie Test For Open Source

SPEAKER_04

it? And so I they they credited me for the pre-made pie, and then I had to buy all the ingredients and I owed them 10 bucks. So I just like looked at it and said, I could have bought a pie that was already made, which I don't know where the ingredients come from, but it's probably a pretty good pie. But I returned that for less money, and then I bought all the ingredients to make a homemade pie, which was more money, ten dollars more. And then my wife has to go make it. And I'm unclear whether the outcome will be consistent with the bought pie.

SPEAKER_03

So it's more work, it's more expensive, and a lot of buttons. You do get the other ingredients for you know for future pies. That's that's something.

SPEAKER_04

No, it was like just for this pie.

SPEAKER_03

Oh, wow. Yeah, it was just for this pie.

SPEAKER_04

Okay, all right. So I paid ten dollars more. Yeah, had to wait on it. There's also like variability. I don't know. It was a little her pie was a little well done. Right.

SPEAKER_03

Not the best, not the best cooking, cooking job.

SPEAKER_04

So I have to do more work, I have to put the ingredients together, and I have to hope, I have to hope that the outcome is as good as the bot pie. That's to me like using anthropic versus an open source. I like the analogy. Okay.

SPEAKER_03

So you can bake your own pie. I'm not sure why you would. And Carp's saying, I'm gonna bake you a good pie. Yeah, that's he's basically like, hey, you can't do that on your own. Yeah, you're not gonna bake a good pie. Too much variability. So let me handle that for you. And I might be maybe still more expensive, or I'm not gonna be the cheapest option in town, but I'm gonna make you a great pie that you that you know effectively helped bake yourself, right? And you're in control of if the exact ingredients that went into it. Right.

SPEAKER_04

It's your unique pie. Yeah, right. You get to use organic cinnamon or organic apples, you get to know where everything comes from, and you get to know how it's all being used and how it all bakes together, and you get to decide how under or overcooked you want to bake it, right? But that's just not going to be for everyone. Yeah, my wife wants to bake a pie. I was kind of happy. I'm happy with one of those hostess brand hot pies. You know, like McDonald's pie is pretty good too. So I was damn it, a great apple, you know. So I think you know, this idea that there's right or wrong, which was what he was the must be clear, he's trying to get his stock price by talking to his buck. And you know, I think the reality is people have to make these decisions. The interesting thing is the next day outside of breakfast with the ENR 500 CEO and asked him about it. And he said it was like a bit like you know, word got out within the company. He got some emails and some text messages. Some of his board members said, Hey, what are we doing? Did you see this interview? What are we doing? And he's like, Man, I just spend like an hour explaining to people like what was going on and what we were doing. And it's like, the good news is like we we aren't doing any wild things, we are working with them, we're not running it across our enterprise. But the what's you know, the downstream effect of that is I had a lot of people call me about it as well. And the funny thing is for the AEC industry, most of them are running Microsoft Shops, and most of them are running Copilot. They're not doing a lot more than that on their own. Most of them are not running Cloud on their own or chat, you know, it's more like individual people have it, and many of them have just banned it altogether. But what it shifted their thinking to was the incumbent software companies and the startups they're working with. Like, what are they doing with our data? Yeah, right. How's you know, you know, ProCore came out with their owners, new owners package, and in the press release, it says, trained on millions of projects to your benefit. And it's like, where do those millions of projects come from, Procore? Right. And I think that's what really everyone's just started hit, kind of hit pause and said, What are all these people doing with our data? What is that startup doing with our data? I mean, I think we've talked about there's like what a hundred estimating startups out there. There's no way any of them are gonna survive. And most of the tech is like, meh, it's okay. But what are they doing with all your data? You know, are they are they really just gonna stay and be an estimating software company? Are they gonna become their own contract? You know, what are they gonna do with all that data? Because it's not rational that this much money has gone into these estimating software. I mean, there's a lot of money that's gone into them and unclear, right? And so a lot of the CEOs asked me, like, I guess we should be asking our these startups that we're working with what they're doing with this what they're doing with our data, because maybe that's the the the scarier thing for them. But the other thing that I thought was interesting with the CEOs I talked to was how prevalent cyber attacks are in our industry. I had no idea. Several of them said, you know, we're struggling with even understanding AI because we just got ransomware last year, and so we've really just tightened everything down. Yeah, and I had no idea. I didn't I didn't know our industry was, I mean, I've I know lots of industries are more susceptible to to hacks and all, but apparently, and then I talked to one of our mutual friends, and he's like, Yes, I know several subcontractors that have been put out of business because of cyber attacks, they got ransomware attacks. Yeah. And I think there is like, okay, well, if we can't even control that, yeah, how are we gonna do anything with AI? Yeah, and I don't know if they're connected, yeah, yeah, but I think that's just a general feeling of you know, cognitive load. Like, I'm too worried about getting hacked and think about upside.

SPEAKER_03

Yep. So to recap a little bit of the a little bit of the carp dialogue and how you're thinking about building your own sovereign suite of tools, or what you you know, objectively, what you think the best path is that these enterprises should take if they're considering if they're if they're listening to the CARP interview and they're saying, okay, what are we doing and what should we do? Yeah. Is is your recommendation that they index more to the sovereign side and explore open source, or is your recommendation

Cyber Attacks And Enterprise Anxiety

SPEAKER_03

that they should be in the phase where there's still just accurring the benefits of AI and they should try they should trust, they should trust the you know the clauses that the model company that they're working with, or whoever whatever's whatever service provider sits in the middle that there is some privacy on the on the data side, or should they not care at all?

SPEAKER_04

Like should they just I I think there's a couple of things to think about, you know, and I I just tend to like get back in my time machine to the 90s. So Linux was very rogue. You know, I remember meeting Linus, right? Uh Torralis. It was like that rogue. It was a cool experience. Yeah. I mean, I mean at the time it was just another dude. Yeah. Like we're at these internets.

SPEAKER_03

He's like very, you know, very, very, uh, very famous now. But at the time, I could imagine at the time.

SPEAKER_04

We're all just like a yeah, we're at internet, we're at networld interop, and we have these little card tables with the skirt around them and a handwritten, you know, our company name Monumboral

Sovereign AI Services And The Red Hat Playbook

SPEAKER_04

just hanging out there. Uh, that's how I met like Jeremy Lair, Cold Chief and right, like he's like big deal now with Circle or whatever. So, yeah, we all used to hang out and just doing stuff, right? And so I think what happened is Linux went from this open source rogue thing to then you had companies like Red Hat show up. And Red Hat said, look, we have the open source Linux kernel. We've put in the security measures, we've created our own kernel, and we're not gonna sell it to you because you can't do that when you're open source. I think later on they found some flaw that some way around to charge for software. But they basically said, here's the open source software, and we'll do services on it for free. So corporates had an option now. I think Red Hat's gonna buy IBM now, but a friend of mine used to Red Hat. Had got really big. A friend of mine actually was running it for a while. They went public and I think they ended up selling IBM. But the idea was like, okay, here's a pre-made open source software, and we're going to give it to you for free. But you're a big company. You can't recruit and hire the right people. You don't know how to support it. It's a whole different language, so to speak. You know, if you're a Microsoft stack, how do you drop open sources? So they basically said software is free, but we charge for services. So I think that's where we end up. And that's all Palantir is doing is saying, like, you know, oh, you should have your own open source model, which means you can't do it yourself. And we're gonna we're partner with NVIDIA, yep, and we'll come in and deploy, you know, and we'll do it for you. And I I think that's gonna be an option on that. I mean, we're doing it for some of our clients, right? Now we're starting to architect out, like, and and they don't know how to manage this and deal with it, they're worried about cyber archives, they're worried about everything else. So I think so. I think it'll be, you know, we're doing some stuff, we're not that worried, we're just gonna work with the anthropic. Hey, KP, can you help us run our own open source model and manage it for us? Yeah, sure. We've had people ask about on-prem. We actually have a project right now for facilities management that building like an on-prem open source.

SPEAKER_03

What's what's the what's the appeal to that for that client to do on-prem? Building systems.

SPEAKER_04

They don't want their building systems connected to the broad internet, yeah, just cloud, cloud. They don't want it on cloud in general, right? They want it local, and also there's a connectivity challenge that they have with these systems. So no, so I think they're I think the answer is some people, you know, will just stay the course and keep doing what they're doing if they're not worried. Some people, I think you might even see some on-prem stuff. I met a company the other day that I thought was fascinating. They said, How often is your prompt or query or whatever you want to call it to an LLM? How often is it really time sensitive? And if it's an MCP server, why does it have to happen in real time? And what they're doing is they're because of like because of power, power is cheaper at the middle of the night. Yeah, sure. Right? So they're like caching the prompts offloading them and offloading the middle of the night when it's cheaper. And so when you come back in the morning, you have all the answers.

SPEAKER_03

It's like inference, it's like inference hedging in a way.

SPEAKER_04

Yeah, yeah. I was like, that's really that's really cool.

SPEAKER_03

Like, yeah, but sometimes it would be you hey, I need an immediate answer, right? Other times it's not as important to get the prompt response back immediately. Because if you're not paying for tokens, you're paying for power, which is kind of their point.

SPEAKER_04

It's like even if you go to open source models, it's not free. You still have to power, you have to buy the hardware, possibly, and yeah, and run the model and run the compute. So but you know, the good news is everybody's thinking about this stuff. That's that's why when CARP gets on TV and it's like, hey, everyone's well, I mean, for what it's worth, I think there's like a couple threads here.

SPEAKER_03

The I think it's one, it's it's good to have the conversation. I think it's like what at the very least, he brought awareness to the choice, which like back to your Facebook analogy. You know, you're you know, you are you are the product if you're not paying for it. He brought awareness to saying, like, hey, we're all using the model the frontier models because they've been the best. They've been best in class, you know, they're faster. Open source models are behind frontier models. Like, there's a reason to use the frontier models. And, you know, we're all experiencing the AI adoption curve in real time. Like, we're all wanting to use the best in class model. Now, for specific things and you know, related to your sensitive data, we may not need to use the best in class model. And by the way, the open source models are not nearly as far behind as they used to be. We're catching up. Yeah. And so to me, uh, the the the great thing about his interview was like he brought broad attention to the architectural question that people should be asking. And it's just they need to they need to know, they need to know. They need to know they have the choice and they need to think through what the right architecture is for their for their company.

SPEAKER_04

So so the interesting thing from all this is I was talking to a construction company that's in the Bay Area, and they said they are getting hit up like every day by another YC company estimating this, some, you know, whatever it is. And I was like, you know, are you meeting with them? Like, no, we're not meeting with them. And I was like, but what if they have something interesting? And the person had a very like just cogent response. He said, Our industry is about trust, relationships, and reputation.

Trust, Startups, And Changing AI Terms

SPEAKER_04

These children have none of that, they have no construction experience, they don't have relationships. If they take our data and screw it up or mess up, there's no attachment to the industry. Yeah, and they're gonna move on and go do their next thing, right? Whatever the next thing is. And I thought it was like really interesting, you know, especially for a contractor, right? A contractor thinks about their subcontractor, right? People don't realize a lot of people say, oh, they just want to beat up their sub, beat up their subcontractors. It's actually no, they want someone they can trust that has mission alignment and has a reputation and it and cares about their reputation, right? If a project goes wrong, it's not gonna be about the lawyers and the and all of that. It's gonna be like, no, we need to make it right because our reputation matters. And I thought that was just such an interesting, like cogent point of view around how you think about evaluating technology. It's like if this company goes out of business, what does this founder or CEO have to lose? If my data gets leaked, what does this founder have to lose? And if it's nothing, then like why are we gonna do business with them? Which I thought was a very, I mean, yeah, in many ways, uh, you know, our friends in construction are like they may talk simple, yeah, but in many ways they're brilliant.

SPEAKER_02

Yeah, right.

SPEAKER_04

Just the experience of it all like, I mean, I'm not gonna work with a drywall sub that just shows up out of the bullet. Yeah, like I would never hire a 20-year-old drywall sub that has never done a job in your in their career, like, and that that's almost like less important than me putting my corporate systems on a platform. So I thought it was a very simple view of the world in a world where everyone's like trying to like get an edge. Oh, let's have an innovation, like we need to get an edge. Like, no, yeah, we're just fine.

SPEAKER_03

Yeah, actually, like uh it's it's gonna be interesting to see related to the the trust factor. If you're buying any sort of AI application product, how do the 20-year-old startups build trust with with enterprises? And I think this gets to a little bit of what CARP was alluding to. It's like they're all knowing that most startups are using the frontier models, they're just a pass-through their gateway to to spending tokens on those models. Well, those models are training the data that are that that's being fed to them. In fact, like I think one nuanced thing to point out about that that interview and the timing of it. So if you're using Claude Fable, which has been all over the news, the US government banned it, it's now back open. Security issues, right? Safety issues were the reason they they shut it down. Fable had different terms and conditions, which was that they're gonna train on your data, they have access to your data. Previously, Anthropic, if you use claud cloud models, yeah, they did not have that clause in there where it's saying, hey, we don't, like they're basically said, hey, we have we're we're keeping your data you know private, and who knows what that actually means, but they they did have a clause that says, hey, we're not, you know, we are not going to steal your data. But they changed that. And so related to the startups and how they build trust, I think this is an interesting thread because what is the right architecture, even for them? Should they be using if if if they're providing an AI application that's spending tokens with the models, should they also offer an open source option? How do they is that a trust building exercise? Yeah, does it even matter? Like if you don't, if you don't control everything in the stack, yeah, does the owner, you know, the decision maker, the purchaser of the product, do they even do they even care if they can can't control and see everything? Or is there a way for them to re-architect their product experience and build trust?

SPEAKER_04

No, I think there's a couple of things there. I mean, I think, you know, what does it mean? Like your data is yours and we're not taking your data. And are we training on your data? We don't even know what that means, right? Technically, what does that mean? Or am I training or not training on your data? And it comes back to I was talking to someone the other day about NDAs, right? You know, NDAs protect you from disclosing information that you disclose, it does not protect you from someone stealing your idea. I can, you know, you can go into a company, have an NDA, and say, oh wow, that's a really interesting business you're in. And let's say you don't do a deal, and you go, wow, that's a really cool business. There's nothing that keeps you, you're not in violation of any any NDA to borrow someone's idea. Now, if you take their customer list and start calling your customers, totally different thing, right? But much like you can't reproduce a book and sell it as your own, you're allowed to read a book, train on it, and then go start giving talks, yeah, and write blogs and do all the other things, right? You're allowed to do that. I mean, how many people have written blogs about every book out there, right? I mean, plenty of people do. And so I so I think part of it is just not knowing, but then ultimately is like, does it matter? Right. Does it matter? Because you know, much like the internet, the internet doesn't function with lots of people being on it. Yeah, if you're on the internet alone, it's not that helpful. Yeah. So I think that's part of it's gonna be like understanding like what am I giving away and what shouldn't I give away?

SPEAKER_03

And it's I mean, you you've had to make these choices with with Zero, Foundation Zero, which is your platform. I know you have an AI specific a project-specific AI stack that tries to solve some of these data privacy issues on you know, project by project. So just talk about like the I mean you're actively making these choices, like talk about that.

SPEAKER_04

Yeah, I mean, I think I I don't I wouldn't say like we had some great vision around it. We just we understand, you know, just being from industry, the sensitivity of sharing information. Right. Not not letter of the law type of thing, but let's talk about feelings, right? When BIM first came on,

Project-Level AI With Audit Trails

SPEAKER_04

architects would say, Hey, we'll share the BIM with you, the data, you need to sign this waiver that we aren't, you know. So there's always been sensitivity. I don't I don't know that architects were worried that someone was going to cut and paste a BIM and build the building, right? Like I'm gonna reproduce it and build a copycat of your building. But it was more liability-based. So I think when we thought about like this industry is all about projects, everything is about projects. Corporate overhead is funded by projects. That if you can make projects more successful, and our definition of project success is not that you know the owner wins and the contractor loses, it's that everybody has a great experience and everybody makes their fair share of money.

SPEAKER_01

Yep.

SPEAKER_04

And you reduce waste out of a project system, that that's that's the goal, right? So the way we thought about architecting is if I'm gonna actively contribute data to a project, and that project is gonna learn from that data using Foundation Zero, I need to tell you like what it's doing. You need to see an audit trail, you know, you need to understand everything that's happening within how it's being touched, and then the next piece of data. And and you need to opt into it. It's not reading everything. You have to either upload it or email it, or like you have to actively say, I am actively sharing this, which I think was a both a business and technical decision. Because I also believe, you know, for years we've talked about we need better standardized workflows and processes and construction. And I was like, yeah, that was before AI. With AI, it should be the creative unlock. No project should be the same. Every method can be like you don't need this standardization, kind of like we talked about the McDonald's hamburger, right? It's like it's a very high-quality hamburger, it doesn't mean it tastes good. I was talking to an architecture firm the other day that we're doing a proof of concept with, and I was like, maybe you guys can design nicer buildings now, or maybe less ugly buildings, because what happens is you do this creative thing, and then the client says, like, how much is that gonna cost? And you have no idea. You can't, and the reality is great design, shouldn't cost more. It doesn't have to cost more. It's the complexities of building it that end up. So if we use if we can use AI and robotics to reduce the complexities of building something unique, I think that's yeah, that's where the future is. Let me look at Monumental Labs, you know, like the we think about some of the companies we've invested in, even right? Yep. They're not trying to build houses like factories, they're trying to do unique things, right? So I so I think we just had an approach that, hey, we're from the industry, this is how this is how it works. The other thing is like how I would argue that if you have a model that's training on everything you do, it might start to think that a hotel is a hospital and a hospital is a hotel based on how it's laid out. You know, so I think project context, if we believe that every project is unique and the players are unique, and it's more like producing a movie than a factory, then context and relationships and understanding should reside at a project level. Very interesting.

SPEAKER_03

So let's talk about Alex Cart, why you love him. We we've talked, we've talked, we've talked about the interview, but what's the what's the crush of how I mean I think you know he's clearly neurodivergent.

SPEAKER_04

Uh I may have some of those same issues, right? It's kind of funny like how often the people around me that love me have to like go tell other people. Like it's not personal, like he just like he he cut you off short and walked out of the room. Like it wasn't personal, right? He's not making up he just moved on to the next thing. So I think like seeing someone that's I mean,

Why Alex Karp Cuts Through Noise

SPEAKER_04

I think he's maybe a little bit further divergent than I am, but that he can be successful and then he gets to say what he wants to say. And I think I have a good filter for like I don't care about the theatrics, but he's rarely wrong. Absolutely. You know, he's rarely wrong.

SPEAKER_03

There's always something the re like the if it was purely for entertainment and just his theatrics, I think it's a good point. He would not have the stature that he does, and people would not be as tuned in. But he always is speaking some sort of new truth, and he's usually just making a point that's correct, it's truthful, yeah, and that no one else is willing to say. Yeah, I think that's the key thing is like he's willing to say things out loud that might offend people, might he might have business business partnerships. You know, he talked about specifically Dario and Sam in the interview, and he's like, Look, I like hanging out with Dario, yeah, nothing against him. He's like people are concerned about how he's building anthropic. Same thing with Sam. I like Sam, but people are pissed.

SPEAKER_01

Yeah.

SPEAKER_03

And not every CEO is willing to go on TV and say that truthfully and emphatically. And it's a public company.

SPEAKER_04

That's the other dynamic, right? I think a lot of people out of CEOs, like, oh, we're a public company, I have to behave a certain way. I read this great article, I thought it was pretty interesting because there's been the rise of the Indian CEO, whether it's Adobe, formerly Starbucks had one, and there's all these Indian CEOs. And like, they're like, Why are they why are there so many Indian CEOs of public companies? And I was like, because they're not controversial. They do the work, they don't, they're okay, you know, most mostly immigrants, right? So not have had to adapt to a culture and society that's not their own. And running a public company, you kind of have to be vanilla. You can't be wild, you can't have strong opinions, you have to make everyone happy. And I think a lot of immigrants that come to this country, you know, I grew up in the like keep your head down and fit in, assimilate, you know, type of thing. If someone says, Oh, you're Indian, I'm part Cherokee, you just laugh along with them, don't get offended, you know. And so I think there is some of that that, like, being a public company CEU, you have to go along with stuff. You don't get to say what's actually on your mind, yeah, which is miserable. So I think that's that's what I also think is fascinating is it's running a public company. You know, everything he says can be construed as guidance. And every, you know, there's a lawyer in the wings telling you you can't say that. And so I think that that's really what's interesting. I think it's the same thing we, you know, people like to debate about, you know, Elon Musk. And it's like, you know, if you take away all the nonsense, let's just talk about the engineering, let's just talk about like making an impact. And but I think that's why I think part of what I like it about it is I feel like when I hear him talk, like I filter out all that noise and I actually listen to what he's saying. And I don't know that everyone has that ability to like filter it out. They just get too caught up in the noise. Yeah, so so sometimes I feel like maybe he's speaking directly to me.

SPEAKER_03

That's cool. It's a gifted communicator that can do that. Yeah. I mean, the other thing is like he is entertaining. No, he is very entertaining, and the way that he tells stories and speaks, he has so much energy, it's hard not to tune in when he's on a TV. Yeah, like in the background of people having a live podcast. Yeah, I mean it should be an easy, easy call. Easy call. I do have to cut this interview interview short. Maybe the shortest podcast we've recorded in a while, but I'm I'm a busy guy in San Francisco. I got meetings.

SPEAKER_04

I hear, I hear.

SPEAKER_03

So you know important stuff. Um so we'll do, I imagine the next couple weeks, we'll do it, do it live, and uh we'll have to do the uh the event later this month, and we'll share details with anyone who's local that wants to come by.

SPEAKER_04

If you're in uh Bay Area or want to fly in. I know we have our um I don't know when we're doing this. We have our Hackathon. Okay, Bay Area, but I'm gonna do

Live Recording Plans And Closing

SPEAKER_04

that recording.

SPEAKER_03

But in front of a live studio on a week, maybe we'll do uh a bridge version of the hackathon. Alright, it's a fair