In an era where competition among rivals can quickly veer towards consolidation into giants, founders and entrepreneurs can struggle to align their vision with strategy while also scaling for growth.
Join this dialogue to explore what allows a new venture to thrive and what can hold it back.
At Davos, founders Steven Bartlett and Bret Taylor argued that building in the AI era requires a mix of focus, humility, and relentless experimentation. Taylor, CEO of Sierra, said he started the company as large language models “digitized the last remaining analog channel, which is the phone,” making high-quality customer service economically feasible. He expects AI advantage to accrue to vertical, problem-specific agents rather than generic tooling: “There’s probably too many software companies working on abstract agent building… helping companies solve their business problems is valuable.” He also warned of overheated markets, predicting consolidation once late-stage investors demand more than “hopes and dreams.”
Bartlett, who is building a creator-infrastructure business, described using AI to open new growth curves quickly—most notably translation. After 18 months of experimentation, “28% of my audience is Spanish,” turning access to “the other 90% of the world” into a compounding advantage. Both emphasized staying “unromantic” about methods and actively trying to “kill” current approaches before competitors do. Bartlett even tested fully AI-generated podcast episodes where “you would not be able to tell… which one was AI and which one was human.” Yet both stressed that thinking remains the moat: writing is “a proxy for thinking,” and leadership becomes “the orchestra conductor of these agents.”
Afternoon, everyone. I'm Jessica Lessen, the founder of The Information, and I'm so happy to be here with two amazing entrepreneurs for our what it takes to build session. So, before I introduce our speakers for this conversation will be, really getting behind and into, what's on the mind of founders and entrepreneurs as we kind of stare down this next AI era? I'm thrilled to be here with Bret Taylor, the co-founder and CEO of Sierra, about two years in. Sierra is a leading AI native enterprise company. With, I think you've said, 100 million in IRR.
As of November.
As of November. So more than that, with your customer service solutions and more. So thanks for joining us, Bret. Prior to Sierra, Bret was the president of Salesforce, the CTO of meta, the founder of Quip and Friendfeed. Did I miss anything?
No, that's about it.
It's been a decade in the Valley. And Steven Bartlett, the CEO of the media marketing agency Flight Story, and of course, the well-known, host of the podcast diary of a CEO. So, thank you both for being here.
Thank you.
Steven, maybe I'll start with you. So what what gave you the entrepreneurial itch to build your business? Especially? I'm in the media business. It's a little bit of a tumultuous time to be building a media company. We might say. What's been your inspiration and what are you setting out to do?
I think originally, just a lack of alternatives. I wasn't hireable, so that left me with very few options. So, you know, I wasn't good at going to school. They kicked me out of school. Then they expelled me because, as Mrs. Sprinkle said, the head of Key Stage five, I made them a lot of money. So that was like an inclination I had as a young age, like I could make money. And I liked business and psychology. Those are the only two lessons that I attended. And then I went to university for one lecture, and and that was the last lecture I ever went to. So. And then I was an entrepreneur at 18 years old. And ever since then, I've traveled around the world, built companies, taking those companies to the public markets, worked in Silicon Valley, and more recently, in the last five years, started focusing on building a media company in the creator economy. And I think I always knew that I was going to do that, because when I started making content myself, it became my ikigai. It was the thing that I thought I could be really good at. Having spent 15 years building audiences on the internet, it felt like it was helpful for other people. It was remunerative, and I enjoyed it. So five years ago I said my last dance, and five years ago I was, what, 27? So I just left my company at that point, I said my last dance would be building a media company. And so that's what we've set out to do now with my holding company, Steven Dotcom. And Flight Story is one of our, one of our companies within Steven Comm. And, yeah, I think it's I think it's a really good time to be building a creative media company in a world where that which is irreplaceable human is going to increase in value. So, yeah.
I'm sure I'm sure we'll dig deeper, but let's go a little bit deeper into how you're enabling creators. What is it that you see an opportunity to build around them?
I think I think so creators have disproportionate impact relative to infrastructure. So like the diary of a CEO will do, 70 million people will download the show every month. And, theoretically, we could run the show with like 5 or 6 people, which is and you compare that to the reach of like traditional media. It's quite it's quite different. And this is what we see with these creators. They have tremendous impact, but they haven't yet built the infrastructure, whether it's commercial infrastructure, whether it's the, the operational infrastructure to build and operate a team, whether it's new business lines like investing or product revenue or whatever it might be. And so what we're providing them with is that infrastructure. We we co-own the IP and we provide them with the infrastructure to grow their commercial business and their audience. And that could be technology that we've built in-house that allows them to a B test their their creatives at scale, or to have 1000 people watch their content using cameras before they publish it, so they get the retention graphs back, or it's everything down to the amount of CO2 levels in the room when they record their content, because that has an impact on cognitive performance. Like we're extreme about killing the guest.
Measure the CO2 level. Wow. Okay. How are we doing in this room? Are we okay?
It's actually not bad, but it'll get worse. Like, the way I calculate it is room size to the amount of people in the room. So if you imagine I sit in a room for three hours interviewing someone and it's a small room, and if the air is not going out, it will start at 400, 400 parts per million, and it'll get to about 2000 parts per million and over 1000 parts per million. They say it's like having a pint of alcohol. And so I'm very conscious of it now. I'm like very dialed in to the CO2 levels in the room because of this. So yeah.
That's amazing. That's not where I thought we would go at this moment, but that we're there. Well, we'll come back to the techie stuff. So. So Brett, what about you? What? You know, I mean, what made you think this was the moment to get in the trench again and build something?
So, my last job, I was a public company CEO, and so leaving it as.
Oh, I forgot you were you were president, then you were co-CEO.
That's correct. And so when you're leaving that job, you it's a bit of a ordeal. So I announced it on an earnings call. And just by coincidence, a couple of weeks later, ChatGPT was announced and I knew I was wanted to start another company just because I, I just like creating technology. And, you know, I don't necessarily believe in fate, but, you know, when you're an entrepreneur like I am, I, I would say like the phrase I live my life by. I credit to Alan Kay of Xerox Park, which is the best way to predict the future is to invent it. And I just can't help myself when there's a new technology, I want to help shape it. And, so I ended up sort of like navigating myself out of my, my previous job and, really thinking from first principles about how large language models would impact the economy and where there might be opportunities. And I think just like the invention of the internet, I think there's a lot of opportunity. I think if you go back to 1996, 95 and you look at like the cover of wired magazine, most of the predictions were actually pretty good. You know, commerce search, all the things we thought would be meaningful. We didn't know the companies that would do it. In fact, Google wasn't founded until 1998, but I think it was almost self-evident the impact it would have. If you pull that thread, there will be some things that surprise us and some things that are obvious. And customer experience, customer service was one of them. You know, it's interesting. I would like I really think that AI and large language models have digitized the last remaining analog channel, which is the phone. And, you know, companies with hundreds of millions of consumers couldn't afford to have conversations with them. I joke it's probably easier for me to call Sundar than it is to get Google customer service on the phone, and not just because I'm me. And it's partly because of your average revenue per customer is, say, maybe $10 a month. A phone call might cost $10. So you literally can't afford to actually provide the customer experience you want. Until now. So it felt like one of those technical moments where you're actually growing the pie, you're actually creating opportunity where one didn't exist previously. And it's been a really tremendous run over the past couple of years.
One thing that I think is interesting with your approach is we've seen a lot of AI companies that have gone horizontal or they've gone broad. Obviously, we've seen that on the consumer side. And you picked a very I mean, it's not a small vertical but very specific vertical. What say more about that decision. And if you also anticipate sort of reaching other verticals.
Yeah, I actually really believe in AI and particularly AI agents for specific verticals. Going back to the Wayback Machine, I found an article recently from like I think 97 about banks making websites, and they spent tens of millions of dollars and failed to put in a login form, and they were spending tens of millions of dollars to consultants and literally couldn't do it. Now, any one of us could code that in about 20 minutes.
You guys could vibe code. I'd be I'm working.
On it, I promise you.
You could.
And we could just just upload the recording of this, and it will do it for you. You know, it's that easy now, I actually think right now there's probably too many software companies working on abstract agent building. I my theory is that will be a commodity in the future. Just like making websites is sort of a commodity now. But Shopify is not a commodity, right? Shopify is helping a retailer put up a storefront. So I think right now we're just at the bottom rung of Maslow's hierarchy of needs of agents, and it's hard to make them work. I know that will be solved, and I'm not sure there's like a company there. There may be, but I just don't think there is. But actually helping companies solve their business problems is valuable. Helping us. You can replace your IVR system and have an agent pick up the phone is really valuable. Having an agent that can write software is really valuable. Having an AI agent that can help review a contract is really valuable, and I actually am really excited for the entrepreneurs who maybe have expertise in a business problem, like auditing your financials, something that might be sound mundane, but companies spend like hundreds of millions of dollars on. I want to see the entrepreneur who's making an agent for that, because I actually think that's really where the value in AI is. It's fundamentally about automating what we're doing manually today. So I don't really intend on broadening. Maybe we will. The market we are serving, as you know, hundreds of billions of dollars of spend. So that's sufficient for me. And but I'm actually I think, verticals are underrated and actually a long answer, but just to say, like there's this company called Harvey that does legal tech. I don't know if the information ever covered a legal tech company prior to Harvey. It's because it was not a good market. But now that AI is actually doing the work of lawyers, it's a very interesting market. I think that's true of a lot of different markets.
When I ask entrepreneurs how AI has changed their workflows, the most common thing is I call my lawyers less so. I don't know if that's good or bad, actually. But so, Steven, when you think about building out your businesses and maybe you'll talk about some new ones you're thinking about too, what role is AI playing in doing that, helping you do that?
I mean, it's just.
Everywhere and everything, and we really are trying to push on the frontier, especially if I think about the media business to we're basically betting on any area where we think that any rate of improvement in the underlying models is going to give us a unique advantage, or create some kind of blue ocean, as they call it, for a period of time. So one such example would be two years ago. We started the experiment to see if we could translate all of our long form creator shows into other, other languages. The thinking being, if we're just in English, we're reaching like 10% of the world. Can we? Is AI at a point where it can take a three hour conversation? And again, this sounds super easy, but it's not because in Spanish, three hour conversation becomes three hours, ten minutes because the words are longer and then the videos out of sync. So it's like a quite tricky challenge. And if it's slightly jarring in that language, people won't listen. And that can actually hurt you. 18 months. In 18 months of that experiment, we had liftoff, and it went from being this, like, expensive experiment that was taking place in the corner to six months ago. None of my audience were listening in Spanish. This month, 28% of my audience is Spanish.
Wow.
So if I show you I was showing Daniel, like, the graph the other day, it's it's like it's like this for two years, and then it's like this. There's nothing more important that we've done as a business for growth than translations, period. Because it's unlocking this other 90% of the world. So in my company, I have the main business. It's kind of going to the whole point of the innovator's dilemma. I have the main business which is pursuing what we do now, and then I have the innovation team that report directly to me that are trying to use AI to kill what we do now. And so one of the really interesting things I'm testing at the moment is I'm trying to see if I can take a long form piece of content like this conversation. Now put it into a large language model like, ChatGPT or Gemini and it tell me before I edit where people are going to fall off. Now I can test this, obviously, because I have thousands of videos I've made and I have the data, so I can see how predictive it is. And if I can do that, then I can be the world's best content creator, because I can be the world's most retentive content creator. And also I can create an agent that edits for me. And so this is what I'm doing right now. And it's, it's very it's very, what's the word early signs of very, very appealing. It's like 80% predictive.
And so you're rewriting your scripts based on what, what advice it's giving you.
Yeah. Like you can if, if, if a large language model is going to get to the point where it's as good as a human being of predicting where someone's going to drop off in the article or where someone's going to drop off in the video, then, the person creating it can make better stuff. And that's never been possible with video before. But Gemini in particular now allowed me to this week upload. My innovation team are working on this, but I like to work on it myself. Upload a 30 minute video which I just released to YouTube. I released the video on Sunday, and I could put in that 30 minute video into Gemini and say, where do you think people are going to drop off? And I have the data because the video came out on Sunday and it predicted it perfectly. It said, when this happens at minute 18 and this guy comes up on screen and he says this, you're probably going to lose people. This is a horror show for retention. I would edit that out. Unfortunately, we didn't because we'd already published it. But but it's about 80% predictive. And I think again, going back to my thesis, which is just bet on areas where a rate of improvement is going to give you a unique edge. If we start building that platform now, we're going to have probably 12 months significant advantage.
That brings up my next question, which is how much of this is of AI and content in general is kind of ultimately a race to the bottom in the sense that everyone will do it in a sense. Right. So maybe it's a question of having that edge, but, I think of it more even in the content creation side. Right? We talk about, oh, you can now make 12 versions of your video instead of two, but now everyone can make 12 versions. So how do you think as entrepreneurs and startups you need to find like durable advantages, right.
Yeah. And I ask myself that question a lot as well. So I think there is value in finding short term edges that you like windows, you know, are going to close, but actually getting there first, like the whole point of like Spanish have never been able to hear from the world's top podcast guests ever. They've never been able to hear from Michelle Obama or whoever it might be. So if I'm the first podcast to break in, I'm going to start compounding in that region before everybody else. And the reason Joe Rogan's biggest in the world is because he started compounding 15 years ago. So we've all had to play catch up. So there is there is advantage in compounding first. And then the other point about durable notes. I think in a world of AI where, yeah, you're right, kid, in Bangladesh or wherever or Botswana where I'm from can start, can make content again, that's going to get better. What do we bet on? And we bet on that, which is, I think I said earlier like irreplaceable human. So there's lots of the content that I make that I think is going to be commoditized, because actually, when you think when you bet, when you believe that anything that's better, faster, cheaper, easier is going to win out over the long term. You know, if someone's coming to my podcast to learn about the gut microbiome, and I've hidden that information one hour, 26 minutes into an episode, like the thing they want, I'm probably gonna lose to a large language model that can just give it to them for free and quick. Yeah, but the human stuff that I do, the stories that we tell, the very, very human stories of what Michelle Obama went through in the white House and how they didn't pay for anything for her and her family in the white, like these things you can't really get from an LLM. So how do we double down on that, which is human? And also in a world where institutions are losing trust more than ever, how do we connect our audience to people and not to like logos and institutions? And, and I think from a marlovian perspective, like, that's not going out of fashion, like connection, community, all these things aren't going out of fashion. So how do we double down on that part of our content? Yeah.
Some lessons for the news business in there too. But, Brett, how about you and how you work with AI? How has it changed being a founder and CEO for you?
You know, I probably self-identify as an engineer first, and it probably is the profession most impacted by AI. We've gone from authoring code to operating code generating machines, and it's just a very different profession. I think it's very humbling just because, I think many of us identify with our jobs and you get really good at something, and all of a sudden there's a machine that just does it objectively better than you, and you're never going to type faster than, you know, the best large language model. It's just not a differentiated skill anymore. And, so I use AI extensively. I use it to critique strategy. We use it to write code. We use it to analyze our agents. Our clients use it to analyze the performance of their agents. It's really impacting everything that we do. But the reason I brought up self-identify as an engineer. I think that's I think one of the most important things changing with AI right now is so many information workers. What they do in their job is changing in real time before their eyes. And I think it's a very easy to talk about in the abstract. But I think a lot of people's identities are intertwined with those tools. If you're very good at writing code or very using, good at using Microsoft Excel or, you know, very good at answering the phone for customer service, that is a part of who you are. And I think so much of this transition right now, I really appreciate the beginner's mind you have and like creating your content. And when that AI says you should edit out that part of the show you loved, that's like a very vulnerable moment in a lot of ways, I imagine, and you might just overrule it. But that's the world that we're all going to live in. And so I actually think that, like a big part of, I think truly being you called us sort of AI native, I think that's actually a good phrase, is actually letting go of that a little bit and letting AI do what it's great at recognizing that it doesn't remove our humanity. And in fact, being the orchestra conductor of these agents is actually the new job. And it's pretty awesome if you accept it. But it is a very different job than I had the previous two times I started companies.
Let's speak a little bit about the transformation in coding, because I think that it has been incredible. And I think some of us who aren't as technical probably don't appreciate it as much. And we're we're seeing it now hit the non-engineers. Right. We've got anthropic came out with sort of a more user friendly version. And I'm sure OpenAI will do the same. So if we fast forward, like, what are we missing about what some of the impacts of these tools coming out.
So I'll start with just what's going on for people who are maybe outside the industry. So ancient history three and a half years ago, software, which. feels like ancient history. Yeah. You know, software is written in programming languages, which is a structured language more structured than something like English or French, but structured. And they're essentially instructions for a computer if this, then that. And computers are very good at executing those rules really quickly. Just like AI models can generate English or Spanish, they can also generate programming languages. So one of the first things people did was rather than, you know, chat with it was instructed to write computer code. It turns out that there's a lot of things that are actually easier about generating computer code than English, because you can compile it with a compiler and see if there's errors. You can actually run the program and see if it does what you want, which means that you can have an AI agent that's essentially doing that in an autonomous way, trying to write a program and tell it works, which is largely what computer programmers were doing over the past 3 to 6 months. And certainly the past three, there has been sort of an inflection point in the quality of the models from all the frontier labs. As the chairman of OpenAI, I'll talk about Codex, just being biased. They've essentially passed the point where they're really quite capable of long running tasks. There's a really wonderful company called cursor, and they posted a video of essentially letting an AI agent write a web browser from scratch completely autonomously. And I don't know how much that video is embellished. I don't necessarily I don't think it was actually, it's kind of mind boggling if you just look about the engineering resources to write Google Chrome or Safari, and obviously it wasn't production grade, but it was an example of we are truly in a new world. What does this mean for business? I think if you look at most businesses, whether it's a new media business or a software business, the Scarcest resource, we're software engineers. That was always the highest paid scarcest resource. There was coding bootcamps. Everyone should learn how to code. And now all of a sudden we've made it something that is something a computer can do. So I think the essentially, the marginal cost of software has gone, has gone down by more than 100 fold. So what does that mean? What does it mean for industries that write code for a living? It's why a lot of people are wondering, like, can you just vibe code all the SaaS apps you use every day? Which is why I think a lot of people. on that. I think that's simplistic. I think most businesses don't want to build and maintain software, they just want solutions to their problems. But I think the thing that's interesting with AI agents is the way we'll engage with software is very different. If you think about something like a marketing automation tool, the tool is not the product. The tool is maybe the audience that you generate for your podcast or the leads you generate for your sales team. If an AI agent is doing that job on your behalf, the forms and fields in the web browser of that marketing automation tool are essentially worthless at that point. So the question is, will that marketing automation company make the AI agent that replaces it, or will startup do it and replace the company? And that is essentially what's going on in the market. And I think the reason why software stocks are down 30 to 40% year over year is not that an individual indictment on any one of those companies, but the market is saying, we're not sure which one of you are going to survive. And we're waiting and seeing, because when you have a technology that's disruptive, you've just inverted, like the cost structure of companies, when they go through a build versus buy decision, I think it's exciting. Given this is a panel on entrepreneurship, there are just very few moments in business where there's like a before and after. And certainly the smartphone was that. But it took a while for smartphone adoption to ramp up, so it wasn't quite so stark. I think we're seeing this in real time. And as a consequence incumbents advantages become disadvantages. Startups agility becomes an advantage. There's business model disruption as well as technology disruption. So there's probably been no more interesting time to create a company in technology, in my opinion, just because all the moats have been whatever the metaphor is, there's bridges laid down or there's no water in them anymore. I mean, there's just everything has changed. And that's a very interesting time to start a company.
Yeah. I mean, it's fascinating. And I think software is such a I mean, it could be very profound, right. As it ripples across. What about media, Stephen? What, what could disrupt Stephen Dotcom down the line?
Lots of things. And, I'm quite atypical, in the sense that I actively remain unromantic about what I do. And I actually just wrote a letter to my my team in slack when I had a minute off then about the importance of being unromantic, about what we do, of how we do it, but not like what we do. Like, I know what I'm trying to do for my audience, but I am. I'm unromantic about how I do it, and that means that I can adjust. So, like, you know, it's very easy for me as a podcaster to be really romantic about the fact I'm a podcaster and people listen and that's like, amazing. They come up to me and they say hello, but it's that very romance that history proves will be the end of me. Because as the tides change and times change, if I'm still clinging on to how I'm doing it as the innovator, the book The Innovator's Dilemma talks about I'll be hit by the bulldozer. And so, I have an experimentation team. We have a head of experimentation and failure. And as I say, I'm trying. I'm trying to find ways to kill myself.
That's a great LinkedIn title.
It is. That's her. That's her. That's her job as head of experimentation and failure. And one experiment we did about a year and a half ago was could we get an AI to synthesize my voice and write a whole podcast and publish it and make the display.
Picture about this.
Without without human involvement? And it's funny. Well, this is this is the crazy part. Like when you say that again, we worked on it for about six, seven months to get it to be right. But after about seven months, if I showed you the three retention graphs, which are the graphs that show how many people listened up to the point of an hour, you would not be able to tell, you would not be able to tell me which one was AI and which one was human. Between my show, one of our creators called Paul and the AI one, because the three graphs look exactly the same. And so again, if my audience want me in AI, then fine, because I'm still going to be delivering the same thing. I'm. You have to be unromantic about the the way you're doing it. Because especially in a world where people like Robert, Kurzweil predict the rate of change is only going to accelerate. I think, like the the most important skill for a media company, any real company is kind of what you alluded to then where you said the speed. For me, it's like your rate of experimentation because the correct answer is going to change faster than ever before. The correct way to make a podcast, the correct platform to put it on the correct way to edit it, all the correct answers are going to change. So how do we find the correct answer before insert competitor? Well, the obvious way is we're going to have a higher rate of trying things, and that means we're going to get more feedback than they are. You know, Amazon's Jeff Bezos says in that 2015 shareholder letter, we have to be the best place on Earth to fail. That is the DNA of my company. And when I say DNA, I mean, no, like the DNA, like these are our slack channels. It's like failure and experimentation. The failure team will teach you how to run a scientifically rigorous experiment where everything is controlled by one variable. They'll teach you how to create the hypothesis, post it to the whole company, then test it and measure it to make sure it was true. And then they'll make everybody in the business hurry the up like do it way faster. Run, you know? And so, it's interesting because like, yeah, the skill of the, the, the next generation I think is going to be humility, lack of romance, a growth mindset, dealing with change in a way that's productive, where your identity doesn't get in the way of the correct answer, which is like most of what's happening at the moment. Like.
Yeah, you guys are bragging about disrupting yourselves. This is very you know, it's not I've interviewed many founders and, you know, it's not I think the typical founder vibe, to be honest, but it's a very 2026 founder humility. I think.
I disrupted someone and there's going to be someone who is naive and unencumbered by either success or convention, or a big team, or clients who are asking for it to be a certain way. Who's in their bedroom right now? Who's about to kill me? So what I'm doing is I'm still in the bedroom, and this is what I mean by I have two separate teams. I have the flight team or the now called Studio Steve, who are trying to kill the main business. And then I have our main business who are serving our clients and doing what we've always done. So you have the sustaining innovation and then the disruptive innovation, like, of course someone's going to kill me, they're gonna kill me quick, you know?
So do you think is advertising going to be the business model through all phases of this or. Because, you know, when you this concept of like if others can get to an audience really fast, if audiences can move really quickly that that shakes the the ad business a bit.
Yeah, I think advertising is one part of it. I was actually just speaking to the CEO of LinkedIn about about the future business models. And I think we're in a really interesting moment in media where creators like me, 15 years ago, we all flocked to Facebook because the organic reach was high and we thought that we owned the audience. And what's actually transpired is I watched every year, year over year, me lad Bible, all of our competitors, we lost 50% of our organic reach every year. And you play this out and go, oh, okay, I don't own this audience. I'm renting them. And actually, in a world of interest algorithms, which is what we're seeing now with TikTok, even LinkedIn, Instagram and YouTube, it really doesn't matter how much of an audience you have. It doesn't matter that I have 50 million subscribers. It matters. What I posted today is the best thing that was posted today. For the people that are interested in that thing. And in such a world, what happens? Well, there becomes a greater need for sovereignty and creators don't haven't started thinking about this yet, but like, I actually need to own those 15 million people in a space where I can reach them without an algorithm. So this is why Substack is getting really interesting for a lot of people, because you can own your data. So I've hired a chief data officer whose sole remit over the next 100 days is to build the data, warehousing the enrichment process so that I can own those 15 million. I can pull those 15 million people into an environment where there's no algorithm in the way. And that's priority number one now, because that's clearly the big existential risk in the interest algorithm era.
Yeah, yeah. When I started the information 12 years ago, people thought we were crazy for building a subscription community. And it turns out to have worked out. So, I think we'll be going to questions shortly. You can't talk about startups without. Well, maybe you can, but I don't like talking about startups without talking about exits or talking about sort of, you know, the the state of the funding markets. And just like, what is the oxygen for these companies going forward? Brett, I know it's early, but when you think about just kind of this moment in the Valley, and obviously there's really abundant capital now that that could change. What do you see when you look out at funding sources and eventual paths to liquidity?
I think we're almost objectively in a bubble of some sort. And what I mean by that is there's so much capital that there's not really much filtering on who gets financing and who doesn't. Which is leading to I think this is going to sound funny, but almost too many competitors in each market. So you have too many people building foundation models, too many people building software engineering agents, too many customer service agents, startups. What will probably end up happening? Hopefully not a big correction, but as capital, sort of like, tapers, you'll start to see a lot of consolidation. And I think, I think it's actually really healthy, though I actually, you know, I've joked that your perspective on the.com bubble is very different. If you went all in on Buy.com versus Amazon.com and, you know, I think we're going to see both in this market, we're going to see. But what's the thing that's really interesting about the valuations? I think for the companies that do end up owning their category and winning in their category, I think they're actually justified. I the revenue levels of just take cursor as an example, truly unprecedented as far as I know. And, you know, and we reached 100 million in IRR in seven quarters, which if you had told me that when I started, I mean, it's great, I just wouldn't have believed you. It's partly because the demand for this and the value that it's providing is huge. And if you look at like, what is the addressable market of, say, a codex or a cursor or a clod, what is the market for generating software? I don't even know how to calculate that. I don't know if we've done that. So I think we are sort of in a period of excess. And I think because of that, you're not seeing consolidation yet because essentially the incumbent buyers, the valuations are too high to execute and people don't have to sell because there's always another financing round. At some point the merry go round will stop and or the musical chairs.
Changes that because we could have been here six months ago and you probably would have said the same thing and three months before that and the same thing, I mean, we sort of keep waiting for a change. I mean, is it a big macro event or.
If there's not a big macro event? I think, you know, at some point you'll really start to see the winners, like, as they get to later stage rounds, just because, you know, there are certain, like a certain stage of investment, hopes and dreams are not no longer a part of the investment memo. And so that that may happen just as time progresses. You have to keep in mind that most applied AI companies and, are less than two years old or around our age, so it's just hasn't been that long yet. I think similarly, if, you know, if for like things like inflation goes up and interest rates go up, you know, that will correct the market just like it did after the pandemic. And or if there's just sort of like, you know, there's a lot of uncertainty in the market right now because of geopolitical factors. Any one of those can cascade through. And macroeconomic factors are much, just a much bigger deal than any individual company. And so that is the most likely thing. And I think there will be a correction. I think it will be healthy. I actually think that it's good that we have a lot of innovation right now, and we need probably to consolidate and put more capital behind, the winners. But you don't want to rush that. This is the way the free market works. But I think it's inevitable. And and it will be fine. Everyone will be fine. It's just like it will be fine.
Over here.
Yeah, but I think we're just at that stage. We're probably. I don't I don't know exactly what the parallels are to the.com bubble. We might be in like 1997 or something.
Goodness. We're getting there.
But I and I have no, you know, I don't know, but we'll end up with that kind of choline. I think it's a natural ebb and flow of Silicon Valley.
What about you, Steven? Where have you raised capital from? What do you see down the line?
Yeah, we've raised we've raised a little bit of capital, but the business has been profitable. So our objective is to stay profitable all the way. And at some point in the future we might go to the public markets. It's quite interesting because I guess the nearest comparable is someone like MrBeast and what he's doing with Beast Holdings at the moment. I'm an investor in Beast Holdings.
You're going to be bought by an Ethereum holding company, is what you're telling me? I think that's what happened if I'm getting it right or not.
But but he's he's most certainly heading to the public markets probably in the next 18 months. I think that will be an interesting case study for how the public markets treats a creator led media company. And it's quite interesting because beast, I think he said to me that one in every 33 people in the world watches every one of his videos outside of if you exclude China. So there's obviously gonna be quite an interesting retail investor audience there as well for that kind of and, you know, yeah, these creator companies have huge in-built audiences that come with with them. So it'll be interesting to see what happens. And I'll let him go over the mountain first, take the arrows and then I'll Yeah, we'll see what happens.
That sounds smart.
But we're just gonna stay profitable. And that's our goal.
That is the goal. All right, let's go to some questions. Over here. I think we'll bring Mike's by. So if you want to wait for those.
Hi, I'm Nathaniel. I run an education company in Australia, and we empower the next generation of entrepreneurs and innovators. So I had a question around this. So we're seeing the rise of solo entrepreneurship, where a person can do what ten people can do with the help of AI. What I'm really curious on is, you know, you mentioned this idea of humility and having to almost, be willing to kill yourself in order to innovate and to know that there's going to be another play in the market. What are some of the other things that you predict are going to be the most or the biggest determinants for success for the next wave of young entrepreneurs?
So I do think it's really interesting how exciting it is to start a company right now. Just for context on, I've started three companies. The first company, we literally had someone building our servers and installing them in a co-location facility. The next company, Amazon Web Services, existed, so we didn't have to do that. And now, as you joke, can it just be, you know, someone in the proverbial garage, you know, using software engineering agents? I think that's great. I actually think that one of the most interesting things I think you epitomize this in a lot of ways is removing gatekeepers is really important for innovation. You know, if you have to raise a lot of financing for an unpopular idea, that might actually be great, you'll never actually manifest it. And I actually think that's really exciting for the world. I always imagine Christopher Nolan, the director, and I wonder, you know, my understanding is he had the idea for a lot of his movies very early on, but he needed to become famous first, because you can't make a $200 million movie out of film school. But now, could you make a movie? Not quite. But could you eventually make a movie like inception because you had the idea and the costs of the visual effects no longer are an impediment? I think that's extremely exciting. The only caution, I would say, and I don't know how big your team is, I think there's value in teams. I think there's value in camaraderie. I think there's value in people challenging your assumptions. And you could say, well, maybe an AI agent could do that. You know, I think there's a moment to leave a garage in. Most great things were created by groups of people motivated to change the world together. And so I say like, it's, I'm a huge believer, actually, when I give advice to founders in Silicon Valley, I often will advise people to find a co-founder, partly because it's extremely lonely. The default is failure. And, you know, it's lonely to be in the fetal position of your own living room. But if you're on a couch with a co-founder wallowing in despair together, you can sometimes navigate your path out. And I just can't emphasize enough, just like how challenging it is to start a company. And like, generating the code was never the hard part. You know, finding product market fit was, and having a partner to do that, I think is incredibly valuable. So that's my my counterintuitive take.
Yeah.
I was trying really hard to give you something that's not obvious. And obviously a lot of the things we're testing for in our culture test system, which is these 33 scenario based questions you take when you join the company, are testing for these core attributes that we want in our culture. So we call them like pushing on paper walls, which is believing that constraints might not be constraints, but you only know if you go up and challenge things. So we scenario test everyone that joins the company to see how they behave in given scenarios. So it's Christmas Eve client texts you they're logged out of their account. Person that has the password is on their honeymoon. What do you do? We believe that the best predictor of cultural alignment is behavior, not like words you write on the off site whiteboard about ambition. And so that's what we do. And so there's all those things I can tell you about hard work, blah, blah, blah agency, all those things, the unobvious thing that I'd say that I've been thinking a lot about is earlier on, I did a panel, and the CEO that was asking the question said he was disappointed because his son has given up computer science and is now wants to do creative writing. And I was thinking, actually, creative writing is a proxy for thinking. And even though everyone in my company has the same tools, they're not getting the same outcomes because they they it doesn't mean you have good ideas. Like even the thing I said earlier on about, can we put a 30 minute episode into a Gemini and ask it when we're going to, when our viewership are going to fall off? That is an idea that has come from 7 or 8 different reference points that I've collected over the last 15 years, about retention being the most important thing, about editing, about all these things. And so I'm like, how do you how do you in a world where most of us are going to delegate the responsibility of thinking to an AI, how do you keep that? And actually writing is like a really remarkable way, like reading and writing from a broad range of reference points is a really incredible way. And it's one of the things I'm trying to like really hard, because you almost live in a bit of a dichotomy, like use AI, don't use it for that, but use it for everything. But think for yourselves. Like I'm living in that dichotomy of knowing which tasks I should actually go through, the pain of thinking for, and then which tasks I should just make me a job description.
As an example of this, we write a letter to our board instead of doing slides, and I never use AI to write it because the purpose of it is exactly what you described, which is clarifying my own thoughts on the business and delegating that to AI would actually defeat the entire purpose. Like the documents, the byproduct of clarity I love I love that, that tension.
And that's probably one of the most important things I did in my entire life at 21 years old is I said, every single day at 7 p.m., I'm going to post a tweet 140 characters. And what it meant is, I went through a day like this and I knew it. By 6 p.m., my girlfriend would say, go on then. What she means is go away for an hour and think of something to tweet. Now that sounds like it's like an inconsequential thing. But Richard Feynman, the physicist, says, if you want to understand anything, you learn it. You then distill it. So a ten year old could understand it. 140 characters. Yeah, you post it to the world and you get the feedback. So for 6 or 7 years of my life, every single day, I would have this conversation. Then at 7 p.m., I have to find some truth in this conversation that I could distill to 140 characters of all the things I did in my life that accelerated my ability to communicate ideas to my team, to understand myself, my own trauma, why I avoided women, and all these things. It was that process of writing and distilling and sharing. It also meant that I had a it gave me a million followers, but that's actually the least beneficial thing it did. And so I think writing in an age where writing is going to become, Less appealing because it's easier to not write might be a great hedge, you know, and I think I'm, I'm hedging my bets there.
So pro writing this is what I'm taking away is good.
I agree with that strongly.
Because it's a proxy of thinking. And thinking is the you know, I can't prompt anything if I can't think. .
And there's more time to write. I mean, so if I had the pleasure of interviewing Andy Jassy this morning, and that was the kind of thing where, you know, there are parts of that where I can really prepare, I can really do my research. I have my perspective of 20 years of tech coverage on the questions I want to ask. And then there are parts of it where AI can take the transcript, pick the clips, suggest tweets, you know, and just take all of that away so I can go get ready for this conversation or the next one. Or I can do what I'm writing now, which is my reflections on how this interview differed from the one I did with him two years ago. And just some of that is like the feeling in the room that he had then and that he has now, and it actually think is a lot more confident, which says a lot about Amazon's AI journey. So, I, we can just all plus one agree on that. So, I think we probably have time for one more question. Let's do two quickly. Okay.
Oh, you're you're so very kind. Okay. Well, my question is, is there a piece of, conventional advice that you find, contradicting, the beliefs that you've held, in building your companies empirically? Because, you know, 90% of companies fail. And I'm just also wondering, is it because people are following a piece of advice that is flawed or misconstrued?
And then let's get the other question quickly, because we'll be right on the timer and we can.
Answer the other question really quickly is if you weren't building what you're building now as the proverbial kid in the bedroom, as Steven would put it, what would you be building in?
So advice that's wrong and what would else would you be building?
Steven? So on the first question, the the advice that a lot of entrepreneurs are giving is to like, never give up and you like it was like on the gym wall this morning. And I think it's terrible advice like you learn, you know, you're going to have bad ideas and that's like, that's a given. So it's like, how much do you overstay your welcome in a bad idea? I think cost people most of their life, whether it's a relationship, a marriage or a business idea. And even like when I was saying earlier, like went to school for one lecture, did this startup quit out the blue run, another one that became a public company? And then, you know, ten days before the IPO roadshow to uplift onto a bigger stock exchange, quit. I think quitting quickly is a remarkable skill. And I haven't interviewed hundreds of founders and entrepreneurs because that's like my job now. They also have those moments in their life where they made a really bizarre, objectively like ridiculous decision to quit something. And then on the second question, I'd say it would be something I'd be doing IRL community events, because that when I think about my ikigai would hit the same chords. And I think in a world of AI, I think there's going to be a certain fatigue, which we're seeing. I think we're seeing the early signs of where people there needs to be IRL are going to come in hot. So like buying football clubs, buying festivals, doing things IRL, I think is really, really in vogue.
Right. I'll just take the last question to keep it short, because I think we're out of time. I think there's not enough entrepreneurs working on building AI agents for boring business processes that are really important. I think there's a lot of people making AI agents agent building tools, but I'll just use an example I used earlier because I think it's interesting, every public company in the world, when their quarter ends, there's usually a period of call it three weeks before they have their earnings call, and in that period they pay auditors to audit their financials. They pay attorneys to look at every contract. And it's essentially auditing. So when you go and your CFO signs your, your earnings report and you're on your, on the phone with your investors, you're not committing fraud. And which has been a problem in the past for some public companies, almost all of that could be done by agents, you know, and I'm not sure if there's anyone working on that, because the domain, the Venn diagram of understanding revenue recognition rules and contracts and understanding how to build AI agents, there's not anyone in the intersection. But I think that's actually where a lot of the value will be over the next decade with AI agents is taking these processes that actually might be better done by a computer, but right now we're just in such the early innings where the people excited about AI and the people who understand those processes just aren't in the same room together. So I would just I hope there's folks out there who understand these esoteric business rules who, like you said, can learn how to like vibe code. Like, I think there's I think we're in the early innings of applied AI, and I think a lot of the value will be unlocked over the next decade.
Wonderful. Well, thank you both for this wonderful conversation.
Thank you, thank you, thank you.
Thank you.