The Exchanges

Q a question about that because one of the things that I've always thought about when it comes to this tension between academia and industry, you might hear about needs and requirements from industry, but when you're in research environment, you have unlimited open-ended, you know, vision and an algorithm and a research paper is very different than something in production at industry scale. So how do you guys navigate that?

A First of all, students, they do internships, they understand It's a problem. They develop the first solutions. And then with that understanding, you can start more principled designs to build these systems. But the truth is that in every successful project, we had at least one or two partners. For instance, uh, for, um, we work very closely initially with Facebook. When we started working with Facebook, Facebook, you know, has an entire, uh, cluster, big cluster for big data. It was 80 nodes. And their big data team was like three people. Then, for instance, in the case of Mesos, we work very closely with Twitter. And Hinman went to Twitter and worked very closely with Twitter engineer to deploy Mesos in production. And actually, the feedback from Twitter has a big impact on the Mesos evolution. From just supporting big data, cluster computing frameworks like Hadoop, it went to support These long running services.

AI assessment note: “worked very closely with Twitter engineer to deploy Mesos in production.”

Q What are some of the differences in, in sort of like really specific differences that you don't see in US apps? I'm thinking of the Constellation, for example, like I think that's a really interesting trend.

A Yeah. In the US, a lot of internet companies will have multiple apps and will choose to keep them separate standalone apps as opposed to in China. Alipay, for example, would very unlikely create a separate app. For Sesame credit or for some of its other products. The idea that everything lives in one app is actually seen as a good thing in China because it's less friction at the end of the day. So one very, very small detail to point out in the Alipay app is the app when you open it looks like a grid, right? And each grid has a logo and text underneath and clicking into it takes you to a bunch of new features. You can completely tailor and customize just like your home screen on your iPhone, which Items you want on the first row, which items you want on the second row.

AI assessment note: “In the US, a lot of internet companies will have multiple apps and will choose to keep them separate”

Q was in Uganda and they were kicked out During the seventies with Idi Amin, my mom's sister went to Sweden and my grandparents went to the UK. And I think that's the other side of this discussion. You mentioned the impact on receiving countries and communities. Can we talk a little bit about that? Like what some of the fears and concerns are and what's top of mind for folks?

A If you visit Lebanon, Jordan, Turkey, these are the countries that have received the most Syrian refugees because, as Lina said, they're, they're the neighboring countries. And put this in perspective, in Lebanon today, somewhere between a quarter and a third of the entire population is a Syrian refugee. There are more Syrian children in Lebanese public schools than there are Lebanese children in these schools. Imagine the burden that this puts on these countries. Transpose it to the United States. The number of refugees in Lebanon, if you transpose that to the United States, it would be the equivalent of us taking seventy million people. And you know the debates we're having about taking 10,000 Syrian refugees. So these countries are bearing a very significant burden. And Lena's exactly right. Ultimately, the way to end this challenge and problem is to end the civil war in Syria and to stop the bombing. And we're working 25 hours a day Trying to do that, even just getting a cessation of hostilities to take that pressure off and get humanitarian assistance flowing. But as that's happening, what's driving them when they get to Jordan, Lebanon, and Turkey are two things. One is access to education. They want their kids to be in school. And second is the ability to work. What you're seeing is so many people and so many middle-class people who've left take their savings with them. …

AI assessment note: “So these countries are bearing a very significant burden.”

Q It's kind of frightening actually to think about because you can now pivot a company three times and not even blink an eye. And here your product is this thing that has so much planning and development that goes into it. Just to break it down at the technological level, why is animation so hard?

A An animated feature film has about a 110,000 frames in it, and each one of those frames has an enormous amount of data. At that time, just to render a single frame could take a day. So imagine how much computing power it would take to render an entire film. And then you had to have the software to actually manipulate the characters, the three-dimensional image of Woody or of Buzz or something like that. How do you then manipulate that image to make it seem as though it's alive, to give it emotion? Because emotion is very subtle movements of the eyes, very subtle movements of the mouth. Pixar had to invent the technology in order to do that.

AI assessment note: “just to render a single frame could take a day”

Q It's kind of frightening actually to think about because you can now pivot a company three times and not even blink an eye. And here your product is this thing that has so much planning and development that goes into it. Just to break it down at the technological level, why is animation so hard?

A An animated feature film has about a 110,000 frames in it, and each one of those frames has an enormous amount of data. At that time, just to render a single frame could take a day. So imagine how much computing power it would take to render an entire film. And then you had to have the software to actually manipulate the characters, the three-dimensional image of Woody or of Buzz or something like that. How do you then manipulate that image to make it seem as though it's alive, to give it emotion? Because emotion is very subtle movements of the eyes, very subtle movements of the mouth. Pixar had to invent the technology in order to do that.

AI assessment note: “An animated feature film has about a 110,000 frames in it”

Q original network effect. So there is a certain density there. Well, speaking of the trend side of things, we talked about, The sharing economy. So let's talk about examples like ride sharing and crowdfunding, and those are new technologies. I mean, they haven't been around before, or maybe they've happened, and now smartphones are here to sort of give us new behaviors around them. Do those trends have network effects?

A Yeah, so I think I would combine, you know, I would call ride sharing, um, even food delivery, all of those, I would put them, like the on-demand category in general. I think it's a new phenomenon after mobile, which has made it really easy for consumers. So the way to think of it is, um, so if you look at ride sharing services, right, it's, you know, 10 times better than taking a taxi cab. The product is way better, and consumers use it, but it's a city by city rollout, to a large extent, like OpenTable. However, the difference, I would say, in a ride sharing service, it's a point to point service. In OpenTable, you offered the tools to the restaurants, and then, you know, that sort of, you know, the restaurants used it for the tool, and then you brought in the network, which is the diners, and You know, you match sort of the two sides. In ride sharing, it's really point to point. So I would say the general ride sharing, we think it's more supply side economies of scale. And what that means is the more drivers you have on the platform, you know, you can make sure that you get a good quality driver within five minutes, but that's where it stands, right? So it's like, it's almost like the Amazon first party, if you wanted to think of it, which is like, it's still a good business. We just call it as different, which is You know, supply set economies of scale, which is scale econo…

AI assessment note: “which is scale economy versus a network effect.”

Q were specifically focusing on. I think it'd be helpful to us if you honed in on as if each of those four men or women, actually, I hate the fact that it's all blind men and that anecdote are touching the elephant and then sort of your big picture view of the elephant. How would you define that? The blockchain from each of those vantage points as well as overall.

A So I, I have 10 characteristics, uh, 10 properties that the blockchain has. And the ones that people really think about first is the cryptocurrency aspect. Uh, that's very fairly easy to understand. And the second one is the distributed ledger property of the blockchain. And that's what the enterprises have really honed in on. Uh, because it's something that's easy for them to understand. It's like an accounting thing, um, and it touches, uh, their processes. But in reality, it's a bit more than that. It's a computing infrastructure. Uh, so think of it, uh, like another cloud in the cloud. Uh, it's also a transaction platform. It's a place where you can have your identity. Uh, it's a decentralized database as well. So it's a great software development platform. Uh, in the same way that Java, uh, Uh, was so important for developing web applications. Uh, there's a whole new segment of, uh, the smart contract languages and blockchain specific languages that are just coming out now that, uh, that are going to allow us to, to write blockchain applications. So, uh, there's a virtual machine, um, element, uh, of, of it, similar to the Java virtual machine. It's an open source platform, the blockchain. So this means a lot of developers can contribute. So it makes the innovation widely open. It's a financial services marketplace as well. And then the last two that are very important, an…

AI assessment note: “I have 10 characteristics, uh, 10 properties that the blockchain has.”

Q and technology can break down these barriers. And I think we'd like to take a step away from that and, and really dive into what's been happening. Like what's the evolution that we're talking about? The supply chain, for example, like I think one of the most interesting ideas you put forth is that this historical evolution of the supply chain. Can you talk a little bit more about that?

A Sure. And there's no reason not to be grandiose. I write books with neologisms and word terms like global civilization. So I'm hardly one to shy away. I aim at the hundred year perspective, but also there is lots of granularity in the argument. I think, you know, it is true that though, when you're in the Valley and you're talking about connectivity, of course, your first instinct is to think about digital connectivity and the internet and your mobile phone. And all of that is true, but the internet is actually the only the newest layer of global infrastructural connectivity. And, you know, the fundamental argument is there is the most, the most powerful force of this century is connectivity. But that connectivity falls into three categories. There's transportation, energy, and communications. And that evolution that you asked about begins in the late 19th century with intercontinental and cross-border highways, And then, of course, railways built by empires such as the British. Even the Ottomans had railways that spanned across the Middle East. And of course, the Russians did, too. That's just transportation. Then came energy. We didn't have international oil and gas pipelines a hundred years ago. Today, we have about two million kilometers of them, right? Not to mention the four million kilometers of railways. And so, on top of transportation, on top of energy, you have commu…

AI assessment note: “that evolution that you asked about begins in the late 19th century”

Q back and talk about what that news is, but also more broadly, like, why are bots in the conversation? And also just this, I mean, do you guys even agree with this notion that there's like, it's like, it was web, and now it's apps, and now it's bots. And by the way, we should probably start by talking about what the hell a bot is in the first place.

A So yeah, there's, there's different ways you can come at this question. So what Facebook has announced is Is a platform that lets a brand talk to you inside Facebook Messenger. Um, it sends a message, you reply, it can see the message and produce a response. And the messages from the brand can, are a bit more sophisticated than just a bit of text in that they can be a card, they can have a couple of calls to action, they can put three or four cards up that you can swipe back and forth. Um, which means that you don't have to know what you can ask in the way that you would otherwise. You don't, you're not faced with just kind of a blank command line the way you would with like an SMS service that you might have been using 15 or 20 years ago. Um, you get a little bit more of kind of an interactive model. Um, and so they've launched, there's a payment more plan in there. There's a few other bits and pieces. Very oddly, there's no sort of social or viral mechanic in it, which we can talk about a bit later. Um, and this fits into, like, two or three big industry preoccupations. Like, there is one preoccupation, which is AI and natural language processing, and the sense that you can sort of So you can talk to the bot and it understands what you said and send something back. There's a second preoccupation, which is, um, finding like another runtime after the web and mobile apps, prefer…

AI assessment note: “So what Facebook has announced is Is a platform that lets a brand talk to you”

Q When you said supply and creativity, I don't ever put those two words together. How does that actually play out?

A Xiaomi has created this sweet spot, which no one previously thought existed in the market, where that is, is a well-designed product for cheap enough to be competitive with luxury products in its same domain. What it lets Xiaomi do is do these flash sales where they say, Look, we've got a 100,000 of the, of the, of the Hongmi, or we've got a 100,000 of the Me Too. We're going to sell those now. You have to sign up for the right to get a ticket to have a chance to buy the phone, which means Xiaomi has now used e-commerce as a way of signaling massive unmet demand. They can then take that down to Shenzhen and say, we know how many people want this phone, and we want you to give us these components at the million unit price. So they can actually squeeze costs out of the manufacturing system even before they make the phone, and they can use that money to invest in design without driving costs to the roof.

AI assessment note: “they can use that money to invest in design without driving costs”

Q fascinated in, because you brought this up a number of times in the book, is power differentials, and that's power differentials between people, like, speaking truth to power, people who have less powers, people who aren't in management, but who are contributing original thinking, even power differentials with people who are not represented, like, whether you're underrepresented in the organization. How does those play out in this, in this scenario?

A Unfortunately, the evidence suggests that lots of people who come from non-dominant groups are less likely to get heard when they speak up. So Sheryl Sandberg and I wrote an op-ed last year called Speaking While Female, where we covered a lot of the evidence that when a woman and a man make the same point, the man gets a big pat on the back and people start to rally around him, and the woman is either not heard or punished for being too aggressive. And you see the, the same effect with different kinds of minority groups, because this is not fundamentally a gender effect, it's a power effect. Right. So the, the groups that are perceived as, you know, as not occupying high status positions in society get stereotypes as, you know, sort of needing to find their place. And, you know, they're often perceived as, uh, as stepping out of bounds when they're just trying to make suggestions or get their opinions heard. You know, I think from an organizational standpoint, we need to be especially careful to welcome dissenting voices when they come from people who don't look like everyone else and who don't come from The most common backgrounds in the organization, but I think from an individual perspective, one of the opportunities to overcome these biases is to make sure that you earn status before you exercise power.

AI assessment note: “people who come from non-dominant groups are less likely to get heard when they speak up”

Q What about the cultural differences in the talent markets? What do you see happening in China, Connie? Are there cultural differences in the talent market in China? Um, and, and down to the mindsets or other aspects of how they think about, you know, working at startups versus big companies.

A I'd say the best students are still open to working at startups, just as they are working at an Alibaba or a Tencent or a Baidu. Um, and in addition to that, they're very excited to work for domestic companies. It's not just this halo effect of a Microsoft or Google in China, um, that those students want to go work at. A lot of them are very excited to work for domestic ones. And part of it is because they've seen, you know, other students work for domestic companies and become very successful and monetize that opportunity extremely well. And so I think it's a misconception that some people have that you're always naturally going to get the best talent just because you're a global brand. That's definitely not the case in China. You still have to find that unique person who is very passionate, ideally fluent in Chinese, um, who is very excited to work at a global brand versus the domestic one. Um, it's unclear now that you have a huge advantage of just being a global brand. I'd say another cultural difference is, um, At least something that I noticed at Alibaba, which has been extremely successful for that company, at least in its early years, is a really strong sense of building company loyalty. Um, and I think that paid off for Alibaba tremendously, um, versus the eBay China team in the early years, because Alibaba was able to retain really good talent. They were extremely har…

AI assessment note: “I'd say the best students are still open to working at startups”

Q What was the limitation of, like, why wasn't what was in place MapReduce, I guess you were describing, enough?

A So MapReduce was, was actually great for running sort of large batch jobs that kind of scan through the whole data and, and summarize all of it and give you an answer. But it was designed mainly, you know, for jobs that take tens of minutes to hours. MapReduce initially came out of Google, where it was used for web indexing, and the whole point was, I will run this giant job every night, and in the morning, it's built a new index of the web. But what Facebook wanted to do was different. They had a lot of questions that they wanted to ask almost interactively, and there's a person sitting there who launches a, you know, a question at the cluster and needs to get an answer back, and it wasn't very well suited for that.

AI assessment note: “it was designed mainly, you know, for jobs that take tens of minutes to hours.”

Q What was the limitation of, like, why wasn't what was in place MapReduce, I guess you were describing, enough?

A So MapReduce was, was actually great for running sort of large batch jobs that kind of scan through the whole data and, and summarize all of it and give you an answer. But it was designed mainly, you know, for jobs that take tens of minutes to hours. MapReduce initially came out of Google, where it was used for web indexing, and the whole point was, I will run this giant job every night, and in the morning, it's built a new index of the web. But what Facebook wanted to do was different. They had a lot of questions that they wanted to ask almost interactively, and there's a person sitting there who launches a, you know, a question at the cluster and needs to get an answer back, and it wasn't very well suited for that.

AI assessment note: “it was designed mainly, you know, for jobs that take tens of minutes to hours.”

Q So, um, Nathaniel, speaking of Argentina, though, there are plenty of places in the world that have hyperinflation. Um, what about the environment there made Bitcoin take off? I mean, there could have been other alternatives. I mean, it's not like it's the only option.

A It's not the only option, and in fact, people have lots of other ways of, of getting around the local currency and getting around the local financial system, and frankly, the, the most popular way to, to get around the peso is the dollar, so there are already, there's already a whole infrastructure and world of money changers in Argentina who are changing between dollars and pesos, um, and that's, That is something that sort of set the stage for Bitcoin. I think that what Argentina has that a lot of other places don't have with dysfunctional economies is a very well-educated, literate, uh, tech-savvy population that, that knew how to, um, embrace this technology and that had the skills to do that and, and, and a comfort with, The idea of a digital currency. I don't know that that's the same conditions exist in a place like Venezuela or Zimbabwe, so Argentina is somewhat unusual in that respect.

AI assessment note: “what Argentina has that a lot of other places don't have... is a very well-educated, literate, tech-savvy population”

Q Your system is able to deploy in three months, whereas a traditional ground station deploys in three years. What enables you guys to have that level of suite up?

A It really comes down to the vertical integration, and what we mean by that is kind of like what I, what I mentioned of being able to stitch together all of the different pieces in the end-to-end system so that they inform each other. So, for example, when we say three years before, what did that timeline look like? It looked like an antenna vendor who, you know, gets a phone call. After they get the phone call, they press the order on the supply chain. They gotta wait X number of months to actually get the parts to arrive. Then they assemble them all. It's a bespoke manufacturing line because it's, you know, it's a bespoke system. They're just, you know, they're doing it, uh, one off. And then they ship it, and they're shipping a giant piece of equipment, so it probably has to go over, you know, uh, ocean, uh, shipping methods. And then it arrives at site, and you have to put together a giant construction project because the size of, Many of these antennas, they're like multi-story buildings. So you have to do permitting, you have to lay a concrete foundation, you have to have whole construction teams in order to put them together. That adds up into a three-year timeline. Whereas for us, because we're thinking of all of those things at the same time as we're doing that development, we get to say, we need our antennas to fit in a standard shipping container that can go on a comm…

AI assessment note: “It really comes down to the vertical integration”

Q Your system is able to deploy in three months, whereas a traditional ground station deploys in three years. What enables you guys to have that level of suite up?

A It really comes down to the vertical integration, and what we mean by that is kind of like what I, what I mentioned of being able to stitch together all of the different pieces in the end-to-end system so that they inform each other. So, for example, when we say three years before, what did that timeline look like? It looked like an antenna vendor who, you know, gets a phone call. After they get the phone call, they press the order on the supply chain. They gotta wait X number of months to actually get the parts to arrive. Then they assemble them all. It's a bespoke manufacturing line because it's, you know, it's a bespoke system. They're just, you know, they're doing it, uh, one off. And then they ship it, and they're shipping a giant piece of equipment, so it probably has to go over, you know, uh, ocean, uh, shipping methods. And then it arrives at site, and you have to put together a giant construction project because the size of, Many of these antennas, they're like multi-story buildings. So you have to do permitting, you have to lay a concrete foundation, you have to have whole construction teams in order to put them together. That adds up into a three-year timeline. Whereas for us, because we're thinking of all of those things at the same time as we're doing that development, we get to say, we need our antennas to fit in a standard shipping container that can go on a comm…

AI assessment note: “It really comes down to the vertical integration”

Q What are the levers that could, uh, make a dent in, in reversing some of these trends as it relates to the, to the food system?

A Yeah, from a food system standpoint, I think the biggest single lever that we could pull, which would be extraordinarily politically challenging to pull, is fixing our, like, crop subsidy system. So if you look, the U.S. government over the last, uh, I think it's the last decade, has spent close to, like, a hundred billion dollars on crop subsidies. They're basically subsidizing corn, soy, and wheat, uh, so that American farmers can grow corn, soy, and wheat. Because it's subsidized, it's artificially cheap, so we grow a lot of it. Because it's subsidized, it's artificially cheap, and so these big food companies use it in everything. And so these big food companies have basically gone on like a 30 year journey to replace sugar with high fructose corn syrup or to replace, um, olive oil or something like that with soybean oil, which is highly processed and inflammatory, causes obesity and all sorts of like rat and human models. Um, and, and I think that this sort of wholesale swapping of like, let's take the, the worst ingredients in their most highly processed form and replace them with an ingredient, um, that, that humans traditionally ate Is one of the root causes of chronic disease. Like today, the average American gets almost 20% of their caloric intake from soybean oil. Like this is historically anomalous. And it's not because anyone wants soybean oil. Like you're probably …

AI assessment note: “the biggest single lever that we could pull... is fixing our, like, crop subsidy system.”

Q Yeah. Are there any other really big levers that you would focus on if, if one of your main goals was to target the mental health crisis?

A I think that we are hugely underestimating the impact of, uh, biological health and how that is tied to mental health. Like there's some really interesting studies out there around taking a functional medicine approach where, you know, you come in and say you're depressed and rather than like doing talk therapy and things like this, that statistics show like sort of don't work very well for many people. Uh, we focus just on like reducing your inflammation. And fixing your gut. And there have been studies that show just by focusing and taking a functional medicine approach, you know, focusing on sleep, focusing on gut health, focusing on lowering your inflammation through diet and exercise, that combination of therapies works better than, you know, just talk therapy and things like that. So I think that we massively under underestimate the degree to which we are, you know, human beings who are biological organisms and like our mental health is very tightly coupled To our emotional, um, you know, sorry, to our physical health. And so I would have much, I think that we should fund much, much more of research along these lines of what's called metabolic psychiatry, which is basically this idea that many diseases that we put in the mental health bucket, things like epilepsy, depression, schizophrenia, many of these things have causes that are actually metabolic and root. And like, f…

AI assessment note: “we should fund much, much more of research along these lines of what's called metabolic psychiatry”

Q not just in terms of its, uh, you know, health impacts, but they thought, they thought they were sort of like getting involved in all sorts of foreign wars and like influencing our foreign policy and stuff. And I just say that to say, Could there be competitors to some of these, like, mega, like, could startups even emerge that would begin to, or is it just, like, structurally unsound?

A I think it's really hard structurally, and I think that that doesn't mean that it's not going to happen, and hopefully it will, but one of the things that you have seen is with Monsanto, for example, like, the, the pesticide lobby and the, the chemical and ag lobby is one of the strongest in America, and a big part of what they're doing is, like, For example, a couple months ago, uh, Monsanto was lobbying hard, spent tens of millions of dollars to get a, uh, a rider and a bill that would basically make sure that no one could ever sue them for the potential health impacts of one of their most popular products, glyphosate. Uh, that sort of thing is like, these companies get big, they provide a bunch of chemicals. Monsanto specifically has done a lot of things that, uh, you know, I think are incredibly, like, Morally dubious. Um, and, and then once they get big enough, once they like start selling billions and billions of dollars of these products, they spend a lot of that money on lobbying and making sure that they are immune from any of the harms and health impacts, uh, that come from some of the, some of these like toxic products that they spray everywhere. Uh, like glyphosate is the number one pesticide sprayed in the U S and the largest maker of it is spending as much as they possibly can to make sure that Glyphosate, which, you know, there's been fourteen billion dollars of …

AI assessment note: “I think it's really hard structurally, and I think that that doesn't mean”

Q Are, are some of these ingredients that we're, we're discussing or, or foods or, or, you know, soda, like so danger, so addicting or so dangerous or so appeasant that they should be outlawed? Or is it like, how do we think about that?

A No, I, I actually don't think that many of these things should be outlawed. Like, I, I do actually think that consumer choice is a very important thing. Um, you know, but we have got, like specifically with kids, we have guidelines And we already don't think that kids should have free choice. Like we don't let them drink. So I think that there is a place for soda in this country. I just also think that if you want to drink soda all the time, that probably we shouldn't be subsidizing it by subsidizing high fructose corn syrup. Probably we should not have Coca-Cola spending one hundred forty million dollars to influence our, our nutrition guidelines like they have over the last, you know, 15 years. And I think that it's things like that where A lot of times the big food companies and people will sort of fall back on this idea of consumer choice. I think it's very important, but we also are like currently grappling with an incredibly important chronic disease crisis. And like, if we don't take effort to uniquely like address this thing, I think that the U.S. is like in an incredibly tough spot, 10, 2030 years from now.

AI assessment note: “No, I, I actually don't think that many of these things should be outlawed.”

Q Um, that's exciting. So what else can you share about plugs for the company or things to watch out for, or for people who, you know, want to get assistance?

A Yeah, look, uh, I just say, if you don't have an assistant, you are the assistant. Step one, if you're starting a business, you should hire an assistant, and if you've got 20 bucks a month, use ChatGPT. If you've got, you know, 10 bucks an hour, go on to Upwork and hire someone yourself. If you hire someone yourself, I would say my main recommendation is you need to interview a lot of people. You know, Athena, we have 50,000 assistants apply per month, and we put them through a huge battery of tests, and we hire like one in 300. So if you're gonna hire on yourself, make sure you interview Not 10, but probably like 50 people. Uh, if you have the resources for 3000 dollars a month, then you should work with a company like Athena, who can recruit, train, manage, and teach you how to delegate. And then when you have the resources for an in-person assistant, uh, hundred, 150 K a year, you should do that as well. Um, and then ultimately, you know, when you get later stage, you have Achieva staff, in-person assistant, and, you know, a fleet of Athena assistants in the cloud.

AI assessment note: “if you have the resources for 3000 dollars a month, then you should work with a company like Athena”

Q Yeah, that's very interesting. What, um, going back to other sort of founder principles, what is your stance on, um, Transparency with the company. Like, what's information that's helpful, maybe going back to Thumbtack days, what's helpful for the, the company to, uh, employees to know, or like, what, what are principles, like, that guide, you know, how much you share?

A Yeah, I mean, default transparency is obviously the best, but there's obviously some lines. Like, there's, uh, you know, at Thumbtack, there was a moment, you may remember this, a decade ago, where I woke up, And I had a message from someone on our team in the Philippines, and it said, Thumbtack does not exist on the internet. I was like, what? And I went to Google, you type in Thumbtack, and nothing shows up, and we basically received the death penalty from Google. Uh, there was a misunderstanding, they thought we'd done something wrong, but they, like, eliminated us from the internet, and so our traffic went to zero, uh, and our revenue went to zero, and I remember waking up, and like, I go to the office, and there's a class of 25 new Thumbtackers, who it's their first day, I'm supposed to do the onboarding with him. There's a journalist, like, walking around outside trying to get a quote, and, you know, that's one of those times where you can't share everything, and so, uh, you want to ultimately be transparent, but if I just walked into a new class and say, hey, the business might be dead, uh, it's not gonna work, and so you have to kind of get a handle of the situation, uh, you know, the people who need to know it, know it, and then once there's a plan and a solution in place, then the whole company, uh, is told.

AI assessment note: “once there's a plan and a solution in place, then the whole company is told”

Q What are other principles, like, you know, I love this delegate by algorithm, not by task. What are other principles or frameworks that are really important or non-obvious to, you know, get the most out of an assistant?

A Yeah, the, the cardinal sin of delegation is that it will be faster or better to do it myself. That's the number one blocker for most people, and the reason it's a blocker is because it's true. It will actually be faster or better if you do it yourself that first time, and it takes more effort to delegate, to teach someone how to do it. It might not be as fast or as good as you'd like it, but the only way you get leverage is by going through that work. So you have to overcome that activation energy. The second mistake people make is they don't compound for the long term, and they hire someone, and then switch an assistant every six, 12 months, but it's the compounding that creates incredible leverage. I've been working with Marnie for a decade, like a little sister now, and she knows everything about me. Um, so that's just kind of like common principles. Um, in terms of the way to delegate, I think the, this is just tactical, the methods. You know, the most common way people delegate is with your thumbs and your phone. That's the worst way. Very slow. Um, better use all 10 fingers at a computer, but that's actually pretty slow as well. The best way to really delay is using your voice. And so voice, you can talk two to three times faster. You can do it on the go. You can do it on a date, between lunch, you know, in the Uber. Um, you know, at Thumbtack during hyperscale times, I …

AI assessment note: “the cardinal sin of delegation is that it will be faster or better to do it myself.”

Q I'm curious for your interpretation of meetings. Some people are like, ah, you know, I, I want to have as few meetings as possible. Meeting's not the most efficient. I'd rather do everything async. What is your kind of sort of thinking between async and synchronous?

A I mean, Naval has a good line that's like, ah, a phone call's better than a meeting. A voice note is better than a call. A text is better than a voice note. Uh, I think in general that's true. And, uh, There's this, it's almost like a generational thing where I feel like you're in this a little bit of like, there's a reluctance to call people on the phone, and it's just, we do everything, text and email, but calling people is super effective. It's really high bandwidth. You can figure shit out in five minutes. And so, uh, I personally try to minimize meetings and then do more picking up the phone and talking to someone for five minutes. Um, when I'm in more chairman mode, Then I prioritize as few meetings as possible and voice notes, and, um, I, uh, I, what I request is voice notes out and text message back, because listening to voice notes is slow.

AI assessment note: “I personally try to minimize meetings and then do more picking up the phone”

Q One thing that you've been phenomenal at is, whether it's some tech or, is, is hiring executives. W, what do you think that you do differently, or what, what are the principles or frameworks that really guide your, sort of, exec hiring process?

A I mean, one thing that is obvious as you hire execs is the more senior they go, the better they are interviewing. Almost by definition, and so when you get to C-Suite, they all are really good at interviewing, and so you actually have to discard the interview more the more senior they go, and so I rev, uh, uh, I, I use references much more the higher you go, and then the other thing that I think is interesting that people should do more of is I just ask people for their three six year reviews. Like, at their last company, 10 people wrote reviews of them that are honest about their strengths and weaknesses, and I just asked to see it, and I say, hey, I'll share you, share mine, you show me yours. I actually think this would be a cool kind of reference pro startup, um, where you could like ping the LM about someone versus pinging their references. So I try to get to more ground truth on the person versus the UX of an interview is just gonna be pretty good when you get to the senior level.

AI assessment note: “I use references much more the higher you go, and then... ask people for their three six year reviews”

Q Yeah, that's very interesting. What, um, going back to other sort of founder principles, what is your stance on, um, Transparency with the company. Like, what's information that's helpful, maybe going back to Thumbtack days, what's helpful for the, the company to, uh, employees to know, or like, what, what are principles, like, that guide, you know, how much you share?

A Yeah, I mean, default transparency is obviously the best, but there's obviously some lines. Like, there's, uh, you know, at Thumbtack, there was a moment, you may remember this, a decade ago, where I woke up, And I had a message from someone on our team in the Philippines, and it said, Thumbtack does not exist on the internet. I was like, what? And I went to Google, you type in Thumbtack, and nothing shows up, and we basically received the death penalty from Google. Uh, there was a misunderstanding, they thought we'd done something wrong, but they, like, eliminated us from the internet, and so our traffic went to zero, uh, and our revenue went to zero, and I remember waking up, and like, I go to the office, and there's a class of 25 new Thumbtackers, who it's their first day, I'm supposed to do the onboarding with him. There's a journalist, like, walking around outside trying to get a quote, and, you know, that's one of those times where you can't share everything, and so, uh, you want to ultimately be transparent, but if I just walked into a new class and say, hey, the business might be dead, uh, it's not gonna work, and so you have to kind of get a handle of the situation, uh, you know, the people who need to know it, know it, and then once there's a plan and a solution in place, then the whole company, uh, is told.

AI assessment note: “default transparency is obviously the best, but there's obviously some lines.”

Q Maybe talk about what's different about defensibility for even the, the bigger players today in the AI era than it was in, let's say, the web two era. Are the companies today more defensible, less defensible, or how should we think about sort of the strength?

A I don't know. I, I think the, the less defensible part, I mean, this is, this is why a lot of enterprise software has gotten beaten up in the public markets. It's kind of two reasons. Number one is that if you're doing per seat pricing, like, how do you come up with a pricing model that people feel is fair? And a lot of it is just psychology, and for whatever reason, for the last 20 years, it's like per seat per month, with like, uh, you know, you've, you've heard my joke, the, the tall grande venti model of, like, software, uh, charging. It's like somehow that felt fair. And whether that is fair or not, like, I don't know, but like people are like, oh yeah, it's like 85 dollars a seat, you know, per month. Yeah, ok, that sounds reasonable. Whereas if you, if you propose that pricing 40 years ago, you would have been laughed out of town. So this just became the norm. Um, and the reason why, as I was saying, public software companies have been beaten up a little bit is like, uh oh, maybe you won't sell as many seats. Like, is Adobe gonna sell as many seats if now you don't have to hire as many graphics designers? Or is Zendesk going to sell as many seats if the, Software just answers all the queries. Like, the answer is no. It doesn't mean that the companies are toast. They might actually quintuple their revenue because now they charge per outcomes as opposed to charging proceed…

AI assessment note: “I think the, the less defensible part, I mean, this is, this is why”

Q like Uber and Airbnb, et cetera, that, uh, you know, the incumbents, you know, some of them became trillion dollar companies, you know, I've got to buy mobile. When we look at the, you know, business impact of, of the AI era, um, what's your mental model for thinking about sort of the incumbent or startup or kind of net new company in terms of value, you know, value capture?

A I, I think a lot of it is the same, like, unless you really screw up the, the pricing model, or like, you know, you're all per seat pricing, it's very, very hard to just get the market to adopt something that is just violently different, and you're operating in the public eye, and your technology team is bad, there, there are a lot of ands that need to happen. I have a hard time believing that incumbents will really suffer. Um, I mean, there probably are some things, like, you know, take, like, one example of, and this kind of goes back to distribution versus technology, Like, all of these business process outsourcing companies, these BPOs, they're the largest employers on the planet. So, like, Tata, Wipro, Infosys. So, if I'm JP Morgan, and I say, I need a call center, and this call center needs to have access to, like, customer records, and it needs to be safe, and everybody needs to be trained, like, and I need to have, like, a 100,000 people that can answer the phone, you know who can do that for you? Infosys, right? Or Tata. Um, Tata has already done the integration with JP Morgan in this case. They might just add AI, and now they don't need a 100,000 people, and they maintain that JP Morgan contract, and they operate in the area of the Goldilocks zone, where it's like, they're gonna make, like, a hundred times more money. That, that's one case. That's the bull case for Ta…

AI assessment note: “I have a hard time believing that incumbents will really suffer.”

Q will, you know, your companies will, you know, there'll be a vibe-coded version of it, but we've seen none of that so, so far. Is your mental model is, we'll, we'll see it to the, in examples where the, the cost is significantly high, or in which there's sort of Greenfield, you know, opportunities, or what is sort of your mental model for the types of software that we'll replace?

A Yeah, I mean, I think the Greenfield one is always true, but when you look at Greenfield opportunities, you need two things to be true. You need the entrepreneur to be very, very patient and say, I'm not going to try to sell to everybody who's, if I'm, if I'm starting a net new payroll company, I'm not going to try to sell to GE because I recognize that they are, they are hostages to ADP and that's never going to change. So one is that patience of entrepreneur, and the other one is you just need a high enough rate of new company creation To really make it work, which is why, um, like to pick on one space of electronic health records or electronic medical records, how many new hospital systems are created every day? I mean, it rounds to zero. So if I'm trying to go build a new EHR system to go compete with Epic or Cerner, I can do that. Um, there are a lot of edge cases there, but it's like, and I might have patience as an entrepreneur, but wait a minute, like, I need to sell five million dollar deals to big hospital systems. Every single hospital on earth is currently using an EHR system. Going to be really, really hard to make that work. So I think, I think both of those need to be true. Like the right type of entrepreneur who's willing to be patient because it's, it's often a very lonely game of it's like, I built this great product. Wait a minute. I don't have any customers …

AI assessment note: “I think the Greenfield one is always true, but when you look at Greenfield opportunities”