Keith Rabois on AI's Startup Shakeup, Ben Smith on the Paramount Merger & More | Aug 7, 2025

0:13Welcome, everyone, to the Informations TI TV. My name is Akash Pasricha. It is Thursday, August 7th.

0:30And we're also going to get to Duolingo. We've got some big guests coming on the show today. We've got Keith Raboy from Coastal Adventures, who is going to be with us in just a few minutes. Ben Smith from Semaphore is going to help us break down the Skydance and Paramount merger that closed this morning. I'm also excited about a chat that we're going to have with our electric vehicles correspondent, Steve Levine, who's going to help tell us about a new startup that a Robinhood co-founder has launched. And finally, we've got an AI drug discovery funding round out of Menlo Ventures that I'm excited to get to.

1:01Before we get going, the big story this morning is President Donald Trump's plan to impose 100 % import tariffs on semiconductor imports. But I want to highlight to you some reporting that our Asia Bureau published this morning. The fact of the matter is that big AI chip makers like TSMC and Samsung aren't really going to be affected here because of the fact that they have some kind of a manufacturing presence in the United States. But what will be affected are companies making older chips that you can find in things like home appliances and cars. This is obviously a developing story, but I will link our analysis in the show notes for you to read, and we will keep you posted on how things evolve.

1:40Let's get through some of the earnings from last night. A reminder, I'm going to lay out one number from each print that you need to be paying attention to. For DoorDash, the most important number last night was 23%. That was the rate at which the company's marketplace gross order volume grew in the quarter. And crucially, this is the fastest that that figure has grown in seven quarters, going all the way back to 2023. Shares of DoorDash were up this morning. For Airbnb, the most important number was 9%, which is about how fast the company expects revenue to grow in the current quarter. Now, 9 % would be a deceleration from what the company saw in Q2.

2:19And the company's reasoning was we did so good last year that it's going to be really hard to live up to that. Airbnb shares were down this morning. For Lyft, the most important number was 11%. That was the company's top line revenue growth. That was also a deceleration from last quarter. And that was probably why the stock was down this morning, even though the company did say that it is expecting gross bookings to grow faster this year. And finally, for Duolingo, this surging company that doesn't get enough attention, the most important number was 41%. That was the astonishing rate at which the company is growing its top line.

2:55That was even faster than last quarter. And that is also roughly the same growth rate that daily active users grew on the platform. Frankly, for a$15 billion company is pretty impressive. Okay, we've got an awesome first guest for you today. Keith Raboi is the managing partner at Coastal Ventures, and he is one of the most prolific investors in Silicon Valley. He is also a sharp critic of many aspects of tech culture. And today, we want to talk to him about OpenAI's big news and also the latest on startups and venture capital. I want to bring on our editor-in-chief, Jessica Lesson, for that special conversation.

3:31Jessica, over to you. Thank you, Akash. And Keith, it's nice to see you. Thanks for joining TITV. Pleasure to be with you. So there's a lot to unpack in AI, of course, but I want to start with something I read on X that you wrote last week, because I think we cover AI startups so much, but don't pay enough attention to how AI is changing what a startup is, how startups get built. and you identified this as I think you wrote a topic or the primary topic at five out of six of your board meetings last week. So A, that's a lot of board meetings and B, bring us inside this conversation. What are founders and their boards talking about as it relates to AI and startups?

4:20Yeah, so I observed that the primary topic at most of the board meetings I've attended in the last week or so was about how AI is not just transforming the product in the user experience and the value proposition, but actually transforming the culture and the way of building companies. And there's a couple of things that are obvious in the public domain. And then there's a lot more important subtle things. Let's talk about the obvious ones first. Compensation, you know, everybody's paying attention from the president on down to these grants, you know, that Ben and other people are offering AI grade researchers.

4:50The president quipped about it. Everybody had centered pop culture. They're being paid more than, you know, professional sports athletes. Now the downside to that, or the flip side, is meta, pretty coherent strategy, meta means money. Startups, basically by definition, do not. Virtually no startup actually is profitable, including some of the highest profile startups in AI, open AI, Anthropic, et cetera. So if you're losing money and you want to compete for AI-grade researchers, if you're going to spend that money, and you can debate whether you should and how and when and how many of those people, but if you're going to spend any of that money, you need to substitute from somewhere else.

5:28So either you have to hire a hell of a lot less people to offset that cost, or you've got to decrease the compensation to engineers, which is a topic that no one really has been talking about for 30, 40 years. So you basically have a salary cap. Like in sports, you have a salary cap. Every startup has a budget that they can afford to spend on people. And if AI researchers consume tens of millions of dollars, you have to figure out where that money's coming from. So that's in the public domain. And which way are companies leaning? I mean, are they leaning toward, where are they taking it from? Well, honestly, I don't think many of them have been super thoughtful about this is when you have a strategy, it needs to match your resources and your competitive advantages.

6:10That's why, you know, meta doing this since it means more money than most, paying more money than most makes a lot of sense. But it doesn't make sense. It would be incoherent for a startup to compete on the basis of cash compensation to AI researchers. So that's one topic. The other obvious in the public domain topic is the laws of physics of growth have been completely repealed. You look at the top 25%, just the top quartile, of revenue growth in the first year from launch. And six months in, it's$1 million of AR. In the year from launch, it's$5.4 million. These are unprecedented 40, 50 years.

6:49And then you see things like companies launching and getting$50 million of revenue in two years or three years. It's also basically unprecedented in the history of technology. So if these are setting the norms, every company that's raising venture capital has to compete with that level of growth. Because venture capitalists don't sit around and say, oh, we've got a 10 % budget for AI companies and 80 % for this and that. So if you can't grow a traditional startup at the same rate as an AI startup, all the money of venture capital, all the risk capital is going to swing into the highest potential companies, which is usually measured by revenue growth.

7:26Now, not all these companies have durable revenue. Not all these companies make profit or contribution margin on the revenue. Some of them could be eaten alive by the next ChatGPT launch. Well, that's a third point. So now you're getting into the more subtle ones, which is, do any of these companies or which of these companies have a future? Because the qualities and characteristics of the general purpose foundation models are increasing their performance. And you have to isolate, let's say you're a venture capitalist, but an entrepreneur. Where is your differentiation? What's durable? What's sustainable?

8:01And with the rapidly emerging tech landscape, it's not so obvious. Usually when we see a technology breakthrough, there's a step function, sort of in evolution. There's a step function, and then things are smooth for a while. With AI, that's not the case. If anything, the pace of change is accelerating, not stable. So this leads to another subtlety of startup building, company building, which is I think you need, you've always had to build a company for speed. Speed compounds the best. Executing companies tend to do the best, the fastest they execute. It definitely seems to predict success. But now you need speed in a different dimension, which is adaptability.

8:41If the platform underneath you is rapidly emerging and things that were impossible are now possible, the companies that can adapt fastest are probably best situated to take advantage of that. And when you build for adaptability, you have to think about decision-making, hierarchy, even things like PM and product management become obsolete. Lee, I saw this interview that Peter Fenton did where he talked about the notion of product management makes no sense in the age of AI, and I tend to subscribe to that. Why is that? Well, because think about what a traditional PM does. You kind of talk to your customers and you develop a set of features, and then you sequence your features and create a roadmap over the next year that's logical.

9:23Well, if the technology is changing every quarter or every month and enabling new features and actually enabling things that were virtually impossible, A, your customers have no idea what's coming and what's possible. B, you need to take advantage and seize on the opportunity as soon as something is actually possible. And so you have to rip up your roadmap like every week or every month or maybe every other month. So the idea of this orchestration over a year makes no sense anymore. So these are just some of the topics. But I think what's more interesting is that it doesn't just apply to AI companies.

9:57So for example, I'm involved in the company Ramp, which is about probably the best you could build a company in the last 40 years from the ground up on the traditional technology playbook, which I've been practicing for 25 years. Ramp is literally perfect. However, if all these trends and forces are different, even a company as good and as excellent as ramp needs to think about which of the things we've taken for granted for 25 years are no longer going to be true. So for example, if you're growing so fast, can you even hire people even if you wanted to at that rate? Can you hire customer success people?

10:35Can you hire customer support people? No, if you're going from 15 million to 50 million, that's actually a pretty good startup. It's not even the best. In one year, you can't hire people at that rate. So you have to build your company in a way that doesn't require you to hire that people, which requires you to rethink everything. So all of these are in floss, top down. In enterprise world, going to market first to enterprises, Fortune 500 companies, Fortune 1000 companies, almost never has worked. Right. But now, because of the pressure of AI on boards and CEOs, many AI companies are getting inbound interest from Fortune 500 companies.

11:12So even the sales go to market motion of which segment you enter and why and what sequence is being inverted. So all of this is intellectually incredibly interesting. It makes me feel a little obsolete. But rethinking this all from the ground up is pretty cool. And the companies that figure this out first are likely to create competitive advantages. But you're right that the fundamental issue is ultimately AI, the general purpose AI, is becoming so proficient that whether a standalone company in customer support or legal or accounting or sales or CRMing makes any sense? Not clear. There are some that probably do and will, but the CEO has to navigate to the future and build sustainable accumulating advantages.

12:03And some are thinking that way, but many are not. Yeah. Well, it's fascinating. I love how you started with comp because I think that is the hidden engine behind so many reasons startups get built the way they do and how they're able to compete with the big companies. And that's shifting. I mean, isn't the answer just that startups are going to get smaller too? I mean, how do you think about sort of the employment within the startup sector? Probably, yes. I think like the key metrics that people are going to be looking for, investors are going to be looking for, public investors, private investors, are basically revenue per employee, contribution profit per employee, which have traditionally been strong for tech companies, which earn tech companies significant multiples compared to traditional companies.

12:47But I think every company is going to get evaluated that way. And so you're going to have to be disciplined about the leverage on every single employee and every dollar you spend on compensation. Yeah. Okay, I'm going to go through two topics quickly because I want to pick your brain on them. So today, OpenAI is expected to be debuting, sort of as we speak, the latest version of their model, GPT-5. We've reported a lot in the information about the sort of leapfrog and coding capabilities of this model. But you've invested in OpenAI. I could count through multiple firms now. You obviously know the company and its growth well.

13:25What are you excited about and looking for in their announcement today? Well, I have no insider information on what's going to be in the product. However, given that they're batching releases to be really impressive and not just doing consistent releases every week or every month, I'd expect it has significant improvement and really does plant the flag in the ground that OpenAI is the leading company on the planet, which is true. But eventually people argue on eight of the 50 benchmarks we can compete. I don't think there's going to be anybody competing with OpenAI for a long time. I think this is a monopoly business.

14:03Wow. And you think OpenAI can outcompete Google, Meta, Microsoft. Why? Oh, easily. I think that fundamentally, first of all, ChatGPT is the fastest growing consumer product of all time. That alone is a couple trillion dollar business if they just keep sustaining it. The brand, you know, normal people use ChatGPT now. And that's not changing. Like, just like Google search wasn't that much better than other search engines, but 70 to 80 % market share was Google search. It was the verb. ChatGPT is the way you describe using... I think the difference between Google and Yahoo back in the day, but we can debate that.

14:40But yes, but clearly... ChatGPT is the way consumers access. AI. And I don't think that's going to change in the short term. There are some things that other companies could do that would make life more complicated. But right now, you can see 70%, 80%, 85 % market share on a super side. Enterprise is more competitive. We'll see if that continues. Obviously, Anthropik's done well with its coding, cloud coding. That's why I think OpenAI planning a coding flag down, if it's impressive, will really deprive Anthropik of a lot of oxygen because I think that's where all the oxygen is coming from. I think Gemini has classic innovator's dilemma kind of issues.

15:24We'll see if they can get their act together. What about the... I think OpenAI, I mean, what a story to watch and it's been incredible. But the question I have, I mean, you started by talking about Meta having an endless cash machine. Google obviously has one for now. OpenAI doesn't. Losing a tremendous amount of money. So I guess my question for you is, OpenAI IPO 2026 or 2027? No idea. Be happy either way. My personal bias always is technology companies should IPO early. I've been on this crusade for 15 plus years. I'm glad that some of my critics have admitted I'm right. I think it's always better to be a public company, but we'll see.

16:05They do have an infinite thirst for capital, no doubt about it, but consumers are willing to pay. Like at the end of the day, one of the things we talked about actually at your launch party many years ago. Many, many years ago. Many, many years ago. 12, but it was counting. Yeah, literally at that party, you and Dick Costello were chatting about this. The original sin of the first internet was that consumers won't pay for things. And that is actually wrong. And the best thing about the information and the best thing about OpenAI is they're proving that if you produce a quality product, consumers will pay for it just like they do everything else in life.

16:40Love it. So I think OpenAI can be profitable just on charging consumers for their usage. I actually, and then of course the cost, you know, tokens and energy is decreasing and all that stuff. But consumers really value and crave what OpenAI and ChatGPT can do for you. So I think ultimately long-term OpenAI will be a very profitable company. And I actually do wonder whether advertising has a future, Google-based advertising, whether you should be shorting that. Yeah. Well, Keith, we're going to end it there for now, but I'm going to be back on soon because there's so much to discuss and you gave us a lot to think about.

17:14So thanks for joining us today. Thanks. That was Keith Raboy with Jessica Lesson. Before we get to our next guest, I want to tell you about a story that we published this morning on the margins of one of the fastest growing AI coding companies. My colleagues at The Information reported that Replit has seen its annualized revenue surge from$2 million last August to$144 million in July 2025. But even as revenue has picked up, so too have the costs of paying for the models and the models that power all of its AI coding tools. In fact, according to our reporting, gross margins this year for Replit have ranged from negative 14 % to 36%, obviously much lower than some of the other software gross margins that we see with other companies.

18:01And it all raises the question around whether or not these AI coding companies can actually reach profitability. I mean, it's still quite early for many of these businesses, but it really highlights all the costs that come with serving customers. You've got fees to access all the other companies' LLMs. You've got cloud computing costs. There is a lot to take into consideration, and we will link that story in the show notes. Okay, we talk a lot about tech on this show, but we also want to make sure that you have the insider scoop on something else that has been making headlines. This morning, the long-awaited Skydance and Paramount merger closed.

18:34It is the latest in Hollywood's game of survival musical chairs, and it puts a brand new spotlight on Larry Ellison's son, David Ellison. To tell you more about that and everything you need to know, I want to bring Jessica back on for a special conversation with expert media analyzer and Semaphore co-founder, Ben Smith. Jessica, over to you. Thanks, Akash. And hi, expert media analyzer, Ben Smith. I think I actually wrote that, so I take credit for it. But it's nice to see you. Thanks for being on. Nice to see you. I do like to think of myself as an expert media analyzer. You know, I think you are.

19:09I hope you can put that in the chiropractor. You're the editor-in-chief of Semaphore. I'm personally an angel investor in Semaphore, so I want to get that out of the way. You're indeed angelic. And actually look at how business is doing, but we can come back to that. So this deal is done. There has been much, much angst about it, much regulatory angst, much fighting between various bidders. Why does this matter? Why should, you know, close watchers of technology and media be paying attention to this deal? And does it matter? I mean, I think in a way that's the question, right? Like, this is a company that, you know, has, you know, both, you know, owns a bunch of huge sort of movie franchises and a set of, you know, once great, once incredibly relevant television brands like MTV that they really just drove straight into the ground over the last, you know, 15 years, 20, I don't know.

20:06It's been a long time since the Sumner Redstone Empire was doing anything other than just milking the company for cash and driving it into the ground. And that's obviously true of the CBS brands, too. And so I think the question is, can one of the world's richest families bail this thing out? Can they build either a streaming service or can they consolidate their way into a big competitive streaming service? you know, what are they going to do in the news space is a very interesting question. Of course, looming over all of it is the fact that, like, in another context, we would basically be saying that, like, and how would you say it?

20:46Like, the regime forced the transfer of a media asset to a favorite oligarch. Like, that's how you would describe the transaction. That's some bold language there. And I think you're obviously referring to— And that's kind of neutral. Like, that's—I mean, I don't mean— That's neutral. Oligarch is neutral these days. Is it? No, but I just mean if we were talking about Russia or Hungary, That's how you would sort of describe how the deal went down. So there's that. And I think that leads us sort of. Thinking there would be and also got major concessions. I mean, the big controversy was was the canceling of the Colbert show tied to the final approval of this deal.

21:22What do you think? Right. Right. Was, you know, I don't know. It hasn't been reported. And the company claims, you know, accurately that that kind of late night TV was on the way out. sure did cancel it right before the deal. But I think more broadly, Larry Ellison is a billionaire who's close to Trump, and there is a sense that you can only get regulatory approval for deals that the president favors these days. And so I just think that's sort of looming over this in a really big way despite the details and the fact that the day after it closed, South Park, which is one of their, you know, remaining really, really successful shows, just kind of savagely lit into that particular.

22:02I mean, just because hilarious. Great satire, great whatever your political opinions are. I'm not sure everybody would see it that way, Jessica. Yeah, no, that's true. Well, I enjoy South Park. But let's unpack the business a little bit. Has Ellison given any signs of what he wants to do with the traditional TV business? I mean, is it going to be selling off some of these TV assets, Nickelodeon, Comedy Central? We don't know yet. We don't know yet, but I think yes. I mean, I think when you read that he has an open letter to shareholders today, and when you read it, there's a lot of talk about, like, combining the tech stack, putting Pluto TV and Paramount Plus into the same stack.

22:41I think there's a big question of are they going to try to buy Warner Brothers Discovery? because I think there is, even though Paramount has, in particular, Yellowstone, you know, it has, I don't, it's hard for me to see a world where Paramount Plus reaches the kind of, you know, exit velocity of Disney or Netflix and that Warner and that Max or whatever they call it now is really struggling to get with a much bigger portfolio. So, I don't know. I mean, I think that's a real challenge. Although, of course, that whole landscape continues to change. And what about news? I mean, there's been reports that they were looking to potentially acquire Barry Weiss's Free Press, bring her in to help with the news business.

23:24There's been more chatter of reaching out to well-known folks on that, too. I mean, CBS News certainly still has some prestige. What do you think about what their plans will be on the news front? Yeah, I think CBS is the last broadcast news show that has 60 Minutes in particular has this kind of, or had until recently, this kind of brand of being this powerful independent voice. Although, frankly, broadcast news has been in decline for our whole careers. I can't remember a time when even people of our generation like me, I never watched this stuff, except when somebody tweets it. So it's not like, I don't think it's been particularly relevant for a long time, but there still is something about 60 Minutes that gets you the big sit down with the president.

24:10And they produce some great stuff. CFES Sunday Morning remains probably like the only, I don't know, a quite successful broadcast show. I should say, like, I am told. I don't think a lot of people of our generation actually watch it. But, you know, and I think with these news, these broadcast news brands, it is really existential. Like, what is this for? Do we need this? Is there going to continue to be a CBS News? And if so, where does it live and why? Like, I don't think those are just commercially or editorially. Like, there's not an obvious answer to that question. The same is true for ABC News in particular.

24:44For CNN, I think there's this question just like, what are these things for? That I think are challenging. They're for reporting. Oh, I mean, I'm into that. Like, that's, but it's pretty. We have to make sure people know that. No, and they have a history of doing amazing journalism. I think there is this question of, is there a tacit understanding that this is going to become a right-wing channel and a pro-Trump channel? Yeah. I don't know. And I think Barry Wise isn't that, right? But it would certainly be a sort of a move to kind of move the channel to the right and away from what I think, you know, President Trump feels is unfair criticism of President Trump.

25:18Okay, one more question before you go. Well, just give us another try. You've got a weekly media newsletter with Max Tanney at Semaphore. You're always watching the tea leaves. What's something else that's sort of intriguing to you at the moment about these big media companies, Hollywood, news? I mean, I think the most interesting thing that's happening in this space right now is that, like, you know, we've just been through this moment of massive fragmentation. The election was played out on these fragmented creator voices and platforms. And I think, you know, there's an old line that, you know, that media only does two things, it bundles and it unbundles.