Skip to content
The Twenty Minute VCThe Twenty Minute VC

How Export Controls Helped Not Hurt China & Power is the Bottleneck to AI | Perplexity CEO

Aravind Srinivas is the Founder and CEO of Perplexity, one of the fastest-growing AI companies in the world. Since the start of the year, Perplexity has tripled revenue to well over $500M in ARR. Aravind has raised over $1BN for the company with reported valuations reaching $20BN. ----------------------------------------------- Timestamps: 00:00 Intro 01:25 — From Lower-Middle-Class India to a $20B AI Company 03:04 — “Attack, Attack, Attack”: Aravind’s Founder Mentality 04:06 — Why Perplexity Forced Google to Change Search Forever 08:27 — OpenAI, Agents & Where the Money Actually Is 12:05 — “The Model Is Not the Product” 18:42 — AI Agents Will Generate More Revenue Than Google Ads? 27:03 — The Future of 24/7 AI Agents 32:50 — Why Perplexity Thinks It Can Become the Ultimate AI Orchestrator 34:57 — The Biggest AI Bottleneck Nobody Can Ignore: Power 43:18 — Can Inference Companies Become the Next $100B Giants? 48:57 — The Next Massive AI Bottleneck (and Why It’s Not Models) 54:04 — Did U.S. Export Controls Accidentally Make China Stronger? 58:27 — The AI Jobs Narrative Is All Wrong 01:01:06 — Why Future Unicorns Will Need Far Fewer Employees 01:13:07 — Wealth Inequality, AI & The New American Dream 01:19:44 — Why Perplexity Is Training Its Own Models 01:21:41 — “Perplexity Was Voted Most Likely to Fail” 01:23:15 — Turning Perplexity Into an AGI-Powered Company 01:25:38 — SpaceX vs OpenAI vs Anthropic: The Best 10-Year Bet 01:32:44 — Elon, Jensen & Why You Should Never Retire ---------------------------------------------------------------------------------------------- Subscribe on Spotify: https://open.spotify.com/show/3j2KMcZTtgTNBKwtZBMHvl?si=85bc9196860e4466 Subscribe on Apple Podcasts: https://podcasts.apple.com/us/podcast/the-twenty-minute-vc-20vc-venture-capital-startup/id958230465 Follow Harry Stebbings on X: https://twitter.com/HarryStebbings Follow Aravind Srinivas on X: https://twitter.com/AravSrinivas Follow 20VC on Instagram: https://www.instagram.com/20vchq Follow 20VC on TikTok: https://www.tiktok.com/@20vc_tok Visit our Website: https://www.20vc.com Subscribe to our Newsletter: https://www.thetwentyminutevc.com/contact ----------------------------------------------- #20vc #harrystebbings #ai #perplexityai #aravindsrinivas #founder

Aravind SrinivasguestHarry Stebbingshost
Jun 15, 20261h 35mWatch on YouTube ↗

EVERY SPOKEN WORD

  1. 0:00 – 1:25

    Intro

    1. AS

      I have nothing to lose. I came from nothing. I never even imagined myself to be doing all this.

    2. SP

      A $20 billion company. 45 million users.

    3. AS

      Over a billion searches a month.

    4. SP

      Built in three years by 400 people.

    5. AS

      If these numbers like doesn't motivate me. It's hard to get motivated by wealth. You wanna get motivated by impact.

    6. SP

      This is Perplexity with Co-founder and CEO, Aravind Srinivas.

    7. AS

      No one's ever in a comfortable position. No one can relax. They forced Google to redesign their homepage.

    8. SP

      Then bid $34 billion to buy Chrome. More than their own valuation.

    9. AS

      Perplexity changed google.com more than any product manager at Google has ever done. Now you look at AI mode, it looks exactly like Perplexity.

    10. SP

      He doesn't do defense. He doesn't do comfortable. His words-

    11. AS

      Attack, attack, attack. That's my motto. Go all in and try your best. Be on the offense all the time.

    12. HS

      You know what I hate with podcasts? When people sit on the fence. Aravind has really strong opinions in the show today. He says that Micron will be more valuable than Meta. He says that the resistance to data centers will continue and get worse. He says the biggest problem today is a lack of power. He claims that Perplexity has changed Google more than any Google PM. You want opinions? This is the show for you. Ready to go? [upbeat music]

  2. 1:25 – 3:04

    From Lower-Middle-Class India to a $20B AI Company

    1. HS

      Aravind, dude, I am so excited that we get to do this. We've done one remote, and then we did one at Founders Forum last year, so thank you so much for joining me in person.

    2. AS

      Thanks a lot, Harry.

    3. HS

      Dude, I- it's a weird start, but just roll with me on it. I ask this of the best founders that I meet. Are you motivated more by the fear of failing or by the thrill of winning?

    4. AS

      Thrill of winning.

    5. HS

      Why?

    6. AS

      Because I have nothing to lose. I came from nothing. Like I, I, I never even imagined myself to be doing all this. So my life has already been extraordinary, uh, beyond any level of imagination. Um, I was just in India like, you know, doing my undergrad and, you know, just, just training neural nets with graphics cards that people in the labs were using for playing video games. It was all for fun and, um, you know, my path led me all the way here. It was never like a mo- my, my... For my mom, just getting a job was success because we were not-- we were financially lower middle class in India, which is not even like lower middle class in UK or the U.S. And so from there, all we wanted to do was get a job in Google. Being an engineer at Google was considered a win. And so I'm, I'm already doing remarkably well compared to that ambition we had as a family. So there's really nothing for me to lose. That's why any time I try to act like I'm trying to avoid failure and being on the defense, I remind myself that like that's the stupidest thing to do. Like, you know, you, you... It's be- better go all in and try your best. Be on the offense all the time. Attack, attack,

  3. 3:04 – 4:06

    “Attack, Attack, Attack”: Aravind’s Founder Mentality

    1. AS

      attack.

    2. HS

      When you review then, what are you not being aggressive enough on today?

    3. AS

      Maybe in the early days, we'd be very, very loud on social media talking about Perplexity versus Google, and I, I used to do that myself a lot. So, and some people don't li- like me for having done that. Today, I'm a lot more measured in how I talk about our products, our competitors, and stuff like that. But it's not a lack of aggression or anything. Um, it's just that like it's... That, that is boring. People already heard that enough from me.

    4. HS

      Do you regret the being so bold in your messaging?

    5. AS

      No.

    6. HS

      So it's not a nuance and maturation of message. It's a, "That's stale and I need something new."

    7. AS

      Not just that. I, I kind of don't think it's a relevant framing anymore. We worked on Search. Perplexity started out as Search. We built the first answer engine in the world that people know Perplexity even today. If you mention the name Perplexity, people think, "Oh, that's an answer engine." We built a lot more things after that. We built a lot of agents, browser agents, deep research,

  4. 4:06 – 8:27

    Why Perplexity Forced Google to Change Search Forever

    1. AS

      computer. We built so many products after that, but we're still known for that first product. And, uh, the mark has already been made. The-- We changed the roadmap of Google. You could argue that I or the company Perplexity changed google.com more than any product manager at Google has ever done.

    2. HS

      Make that argument for me.

    3. AS

      Well, they had never... Nobody ever wanted to ship an answer engine at Google. Nobody. Like nobody wanted to tinker anything on, on the interface that made them $250 billion a year. And, uh, and then now you look at AI mode, it looks exactly like Perplexity. There's, there's not even any difference like the font, the citations, the specific bolding of inline text, inline hyperlinks, um, suggested follow-ups. The whole experience is literally looking like Perplexity, except it's still not as good. And so-

    4. HS

      Is that bad or good for you that they learn from you and adapt?

    5. AS

      It's, it's, it's, it's both good and bad in the sense, you know, you have to obviously... We-- I knew this like around end of 2024 this was gonna happen, so it never caught me by surprise at all. Um, it was just a matter of time. I still am, am surprised that the quality is still not there 'cause I, I regularly test every product out there. And, uh, but I'm, I'm happy that honestly, uh, they, they, they changed Google to be what it should be. And, um, I believe that the frontier is where the money is. The frontier in AI is not about answering questions anymore. It's about actually going and doing work for you. We, you know, like we still have the state-of-the-art deep research in the world, and that's actually where people subscribe to pay for our Pro or Max products. It's not for getting answers In, in the, in the traditional way. They're asking for sophisticated research reports. They're asking for agents to go and do things for you. And so we wouldn't have been able to do all that if we were sitting in 2024 thinking we have everything settled here. We're, we're good and comfortable. No. We- It w- the answer engine was always a lead gen for the frontier products we build. You need something, right? Like, think about it. Every company needs to have one successful product to build the next set of products. And in AI, nobody can sit comfortably thinking they have it all sorted out, including Anthropic. If Anthropic thinks Claude Code is already a win, in 6 or 12 months from now they won't even be around. And so that's the uncomfortable... It, it's, it's, it's, it's an uncomfortable fact about the whole field. Would you argue today-- You just told me, i- if you don't mind me quoting you here, um, you just told me before we started that you think OpenAI isn't ready for an IPO. Um, would you have b- believed you would be in a position to say this two years ago when nobody ha- wanted to deal with any product other than ChatGPT? Think about it. So anyone, even in such a massive advantageous position, can be in a pos- can be put in a position where they're no longer the kings. They're fighting from behind, right? So that's the state of the field. That's the s- it, it's less about Perplexity or Anthropic or OpenAI not having modes or having modes-

    6. HS

      Can I push back on you there?

    7. AS

      Yeah.

    8. HS

      I would, I, I would stand by two years ago, even when they were a dom- and they are still a dominant consumer product, but I would stand by it because I don't think they are financially ready.

    9. AS

      Okay.

    10. HS

      When you look at the, the balance sheet of that business-

    11. AS

      Okay. Maybe, maybe I'll decouple that. I'll-

    12. HS

      Do you see what I'm saying?

    13. AS

      I'll decouple that. Let's decouple that being, like, financial readiness for an IPO-

    14. HS

      Mm-hmm

    15. AS

      ... versus perception of a dominant leader.

    16. HS

      Yeah.

    17. AS

      Do you perceive them as a dominant leader right now?

    18. HS

      Yes.

    19. AS

      In what?

    20. HS

      Consumer search.

    21. AS

      Well, except there's no money there, right? Because it's been commoditized. So it's, it's always a lead g- Like for example, why, why, why are they going all in on Codex? 'Cause that's where the money is. And, uh, we're doing the same on computer. Anthropic's doing the same on Claude Code. Google doesn't yet have a product in this category, but I'm sure they're gonna come after that. Meta's

  5. 8:27 – 12:05

    OpenAI, Agents & Where the Money Actually Is

    1. AS

      trying to launch Hatch for $200 a month. You, you see that, you see what's happening, right? So nobody has-

    2. HS

      But it- there, there has to be more money than just code, Codex, Claude Code.

    3. AS

      It's not about, it's not about code. That's the main thing. The, the money, at least in non-advertising. I'm, I'm not talking about advertising revenue. In non-advertising subscription or usage-based revenue, the money is in whatever is the frontier. And today the frontier is about doing, going out there and doing things for you. And, uh-

    4. HS

      Do you not think, then, that there will be a $100 to $200 billion advertising business for OpenAI?

    5. AS

      Yet to be proven. Let's, let's work through the categories of advertising. Um, who's the number one advertiser on Google? Amazon. So number two? Booking.com. Number three or four I think is Expedia. So, um, how much do you think Booking.com spends on Google? 16 billion, something like that. Something s- some, some, some crazy amount like that. Um, and, uh, um, how do you book your hotels or flights today? Do you book it on ChatGPT or do you book it on Google?

    6. HS

      Google.

    7. AS

      Why is that?

    8. HS

      Uh, for me actually, like, discovery. I would like to-

    9. AS

      Exactly

    10. HS

      ... see the options.

    11. AS

      Exactly. Right. So the interface. The interface is less about conversations and more about exploration. So when, when the decision-making is more subjective and vibes-based, you don't need an objective answer engine. A- and so it's, it's, it's... And, and, and you think about the other category of advertising, direct consumer products. Fashion. Where is most of that advertising budget going into? It's going to Meta, Instagram. Because you're just br- browsing, you're just, like, doom scrolling or whatever you call it, right? And so, uh, the chat interface doesn't capture that user intent, that user behavior right now, which is why it was never a great fit for advertising. And, uh, it also fundamentally corrupts the trust that people have when they go into a product and they want the accurate answer, which is what, you know, Perplexity is known for. Um, and then you're like, "Hey, by the way, you know, you asked for the most, um, highest, like, like, mo- best protein shake, but by the way, these are good protein shakes you, you can check out." Like, it, it, it kinda, like, hurts the trust that people have in your platform and your product. And so, um, that's another ca- reason why... I- if you think about it, like, like, what Meta or, like, I think some other companies in the past have tried to put ads inside, um, messaging apps and emails and it's never really worked out. Um, it, it works out in China in WeChat because there's no other way for them to fund the whole thing, you know? So the, the, the whole economy and, and, and user sen- user behavior has been optimized around gamifying. It's not how things work in America. So I, I'm, I'm, I'm, uh, bearish on advertising to really take off in, in the chat interface. I, I'm happy to be proven wrong there, but I'm bearish on that.

    12. HS

      There, there are two areas that I wanna unpack there. The first one, just taking them kinda chronologically in how you said them, money's in the frontier. The more I hear this, kinda the more I question it because I think that we dramatically overestimate how important frontier models are to do quite basic work.

  6. 12:05 – 18:42

    “The Model Is Not the Product”

    1. AS

      Yeah. So f- frontier doesn't mean a frontier model Frontier just means whatever is the, the frontier outcome you can have right now with AI. Greg Brockman recently tweeted, "The model is no longer the product," right? Um, and it's funny because you, you know that as, as a leader of a frontier lab, he has all incentive to say the model is the product. And, and that's what Google people tell-- I think one of the Google people keeps tweeting that model is the product. I forgot who. Um, and so the reason Greg is right is because, um, if you take Codex or com- Perplexity Computer or Claude Code, what is that? It's, it's an orchestration system, right? It, it takes a model, pairs it with an agent harness. And what is an agent harness? Think of it-- The, the simplest way of describing it is like rules for how the agent loop should run. What are all the skills and sub-agents and connectors and tools it accesses? And, uh, without the harness, you don't necessarily capture and convert the intrinsic intelligence in the model into valuable output tokens. The output tokens, if you're, if you're literally just a reseller of model tokens, you have no business because the model will get commoditized. So even if you're a model builder, you don't have a business. As an infra layer, you have some business on serving those output tokens. But as an application layer or a model builder, you don't really have a business if you're just a reseller of tokens that come directly out of the model. You have business if you know how to take the model, ground it in valuable context, orchestrate it with a really good agent harness, um, connect it to the right set of tools and connectors, y- whether it's personal connectors or business connectors, and provide the experience to people in one single unified system. And, uh, the way we differentiate ourselves at Perplexity is we don't just orchestrate across tools and files and connectors, we also orchestrate across models. That is the differentiation that Anthropic and OpenAI cannot claim because you wouldn't find GPT-5 Phi inside the Claude Code harness. You wouldn't find Claude Opus 4, 7, or 8 inside the Codex harness. These are competing with each other, right? Whereas you would find both these models inside Perplexity Computer. And that way we can bring down the-- we can increase the token value per watt per user. If you assume that, if you assume that whatever decides the dol- like the price, the dollars is the power watts fundamentally, 'cause that's, that's the thing that nobody else can subsidize other than the government, um, you know that whoever pro- pro- you know, provides the most valuable output tokens with the least amount of power expended to produce them generates the greatest value to the end user and has the most pricing power, has the most value. And so that, that, that is the orchestration problem to solve. Who-- The one single-- The s- the most important metric in AI is token value per watt per user.

    2. HS

      What does it mean for the value of OpenAI and Anthropic if model is not the product and it becomes a utility, something you can switch into and switch out of?

    3. AS

      The interface. Everyone thinks we're all building the model layer or the race. We're not actually. Um, I would even argue that building models is a way to stay at the frontier, but you have to own an interface in which valuable AI output tokens are generated. The most valuable tokens. It doesn't have to be the product. This is the single most important thing to like, you know, unlearn for most founders, a- and I had to do it too, which is to be successful in AI product layer, whether you're a model builder or not, it's not about building something that gets a billion users. That mentality has to completely shift. There are a few power users who are propelling this token economy right now. If you look at like all these, um, crazy stories of how there's this one engineer who got Amazon to spend like half a billion dollars in a month because of some stupid way they set up like Agent Loop inside Claude Code. Okay, maybe that's a mistake, but there are real engineers in Meta, in, in other companies spending like 10 million a year per engineer on, on, on, on these, you know, coding tools. There are users in Perplexity Computer. Um, there's one user I think who spends upwards of like $10,000 a month, something like that. Crazy. And, and, and not like wasting it. They're not wasting money. Their business runs using Agent Loops that are running inside these harnesses, and they use these products in sophisticated ways that I, I, I couldn't even conceive when we were building the product ourselves. Even internally inside our own company, there are some people who've set up this kinda like multi-agent hierarchy and Agent Loops that looks like its own software architecture. And I often just ask these guys to come explain to the rest of the company, "Hey, like what are you doing with these tools? Like you clearly are consuming it way over, you know, what we thought n- the average person in the company would do." And the single biggest differentiation between those who use agents a lot and those who don't is whether they run repetitive cron jobs. Like whether you use AIs as one-off tasks, you just delegate a task and then it gets done. That's like kinda using it for like deep research or like whatever, right? Like one single task. Versus the AI is like continuously monitoring something for you The AI's continuously like i- triggering based on certain events and going and doing certain things, giving you alerts. Like, you set up workflows that keep running for all the time. Every time you get an inbound email, like it triages, or every time there's a latency spike, it has to identify which part of the code base caused that. It has to go and do the root cause analysis and then identify the right engineer. All these things, th-

  7. 18:42 – 27:03

    AI Agents Will Generate More Revenue Than Google Ads?

    1. AS

      th- th- this is where the frontier is. And, and so, uh, going back to my main point, these products are not gonna be used by, uh, you know, 100 million people. But they will generate revenue that's gonna be higher than the advertising revenue of Google or Meta. It's gonna happen.

    2. HS

      Completely understand what you say there. I do just wanna focus in on a specific element there when you were saying, like, the power users, because I think one of the core numbers is actually Marc Benioff said they spent 300 million on Anthropic, which works out to be about three-

    3. AS

      Yeah. It'll be interesting to know from him if that 300 million came from, you know, what is the distribution across employees?

    4. HS

      So it what it sounds to be is that was on developers within Salesforce-

    5. AS

      Yeah

    6. HS

      ... so it's about 3.8% of developer salaries. What percent of developer salaries do you think will be spent on tokens in 24 months' time? 'Cause that fundamentally changes the value of OpenAI and Anthropic. If it stays at 3.8%, they will not be $5 trillion companies. But if it's 100% like Brandon at Macaw said it will be in a year, they will be $10 trillion companies.

    7. AS

      Well, um, I think they can certainly be $10 trillion companies, whether it's gonna be, um, a full percent of the developer payroll today or not, because there's a lot of non-developer work that'll also be done with a- agents. Um, and that's actually what we focus on for Perplexity Computer. We're not going after the developer market. We're going after anything that developers don't-- m- non-developers do, basically. Um, your, your finance department or your corp dev or your, like, um, sales reps or your data science teams, um, your, your research analysts. I think that's actually even bigger market that n- it's, it's, it's not even-- Like, th- think of it as, like, Claude code multiplied by 10. That, that's the size of that market.

    8. HS

      If I push you on developer salary spend, what percent of token spend as a p- portion of salary do you think we'll see in 24 months?

    9. AS

      It's hard to say. I think the costs are gonna go down. That's why it's hard to say.

    10. HS

      Do you think the costs will go down? 'Cause this is the, kind of the challenge that we've had. We thought when we went from chat to agent that costs would go down and token costs would go down. They've gone up.

    11. AS

      Yeah, for now.

    12. HS

      Help me understand that and how that changes.

    13. AS

      I think in software, um, you kinda, you kinda wanna pay for the frontier. Um, it's kinda like if you know some engineer is awesome, if you know you have, like, uh, the next Jeff Dean, would you rather hire that person and not hire five, five people who are medium engineers, but not Jeff Dean level with the same amount of budget you have? Yes, right? Let's say you had a million dollars. You could hire five people who are at 200K, or you could hire one Jeff Dean and pay them a million. What would you do?

    14. HS

      The one Jeff Dean.

    15. AS

      Yeah. So I think you would pay for the frontier. Um, but what stays frontier keeps changing. Um, in, in 12 months from now, let's say, thought experiment, there is an open source model as good as Opus 4.8.

    16. HS

      Mm-hmm.

    17. AS

      And, um, you still have to pay for inference. You know, nothing is truly free. But it's gonna be like, let's say, 10 times cheaper than Opus 4.8. And when you pair it with the right agent harness, you know, g- and all the connectors, GitHub, everything, uh, all your developer workflows work fine. Why would you, um, assume that the token spend is gonna be still high? It's not gonna be for the same things you're doing today. It's not gonna be. But there might be a different set of things you might do with the frontier that you're not conceiving today. Uh, my prediction would be so- a- agents that are, like, completely autonomous software engineers. Today, I think we, we're all using tools like Claude Code or Codex to write code, but not as literal software engineers.

    18. HS

      There is a large sway of the people that is now bearish on your frontier models-

    19. AS

      Yeah

    20. HS

      ... who OpenAI's and your Anthropics because they're realizing that you can actually do a lot with open models for-

    21. AS

      Yeah

    22. HS

      ... a fraction of the price.

    23. AS

      Yeah.

    24. HS

      What you're saying is actually that is true, but-

    25. AS

      Yeah

    26. HS

      ... we will still pay for the frontier-

    27. AS

      That's right

    28. HS

      ... and so they will still accrue great value.

    29. AS

      That, that's right. And, and, and I think this distinction, the-- it feels like a contradiction. It's not, though. It feels like two things cannot be true in- simultaneously, but, uh, that, that's not quite the case. In fact, I would argue that the frontier is increasingly gonna be a thing that, um, very few individuals might even want. Like, you could argue that after a point, like, it's not even interesting that AI can write software. You, you, you-- we've normalized it, right? Let's say, let's say that that's gonna be the case. Instead of companies being built with, like, tens of thousands of software engineers, unlike the past, there'll be a lot more companies with smaller software teams, and each of us will be using a lot of AIs. So, um, that's actually good for the world. We'll be seeing a lot of different businesses. We'll be seeing allocation of software labor in places that was never even possible. And, and, uh, whatever is the frontier is gonna be things that, kinda like AI is going and designing chips, AI is designing drugs, AI is figuring out how to build robots, AI is figuring out how to cure cancer. These are applications where you don't have, like, 10 million users. It's like a few companies. But the effect of that work will touch a lot of human lives. I think to me that- that's where the frontier is headed. Um, you could also see that from the moves that frontier labs are making. Anthropic bought, um, a wet lab. Could be for the talent, could be for the infrastructure to run like wet lab experiments. But imagine taking all those tokens and putting it in the mid-training instead of just tokens from GitHub, right? Um, so then that's gonna produce something interesting.

    30. HS

      Don't laugh. Is there an asymptote to frontier problems to be solved? I know that sounds ridiculous, but if you are continuously on the chase for the next frontier problem, you get to cancer, you get to climate change, and my word, I hope they solve both in like heaven. That's a huge amount to solve. But if you're on the treadmill of continuously solving... Is there an asymptote to that? Do you see what I mean?

  8. 27:03 – 32:50

    The Future of 24/7 AI Agents

    1. HS

      Can I ask you, you mentioned about agent usage, and you said if you do repetitive tasks versus one-off, say, cron jobs. You know, uh, I think Sam said it's, "We're gonna have 24/7 AI," and, um, yeah, they've talked about a hardware product that's gonna come out.

    2. AS

      Mm-hmm.

    3. HS

      Do you think we will have continuous agents running-

    4. AS

      Yeah

    5. HS

      ... in everything?

    6. AS

      I do so. And I think that's kind of why I believe the orchestration problem I s- I talked about maximizing the token value-

    7. HS

      Can you just help me understand? Sorry.

    8. AS

      Okay.

    9. HS

      When you say the orchestration problem-

    10. AS

      Yeah. So, so, okay. So there are like four objectives. Um, accuracy, intell- intelligence and accuracy, and then privacy and cost. You know, these are all competing with each other. So you can-- you could argue that, um, you could max out on intelligence and accuracy by building giant, giant data centers and spending a lot of power to r- uh, you know, run them. And, uh, you could miss out on privacy and cost 'cause everything will be centralized and, and, and you're gonna be paying a lot. Um, you could argue that everything can run locally, and so that'll be good for privacy and cost, but may not be frontier intelligence, may not be frontier accuracy. So the solution is to figure out a sweet spot. You know, use local models when necessary, use server-side models when necessary, and orchestrate across local models and server-side models, uh, grounded in valuable personal context. Sometimes the intelligence might already be there, but the system might not work th- because the harness isn't grounded in the right set of tools, right? So build a world-class harness that can even make an okay-ish model appear great, and be able to use the right model for the right task and, uh, right part of the task, sub-agents. And, and even like utilize the compute we all have in our own devices all the... With that, that, you know, doesn't need to be always on a server. That is an orchestration problem, a router, an awesome router, a master orchestrator router. Now, um, if you do that, you can realize the vision of a 24/7 AI without people freaking out about going bankrupt. Because no one's gonna be able to afford a 24/7 AI, frontier AI running on a server. Imagine you s- turn it on and you n- you could never switch it off unless something crazy happened. Um, here the, the, the, the, the thing that most people worry about those AIs is like, "Oh, what if it does something crazy?" But the real concern actually is the cost. Um, nobody's gonna be able to afford it-- a, a cron job at the fidelity of few seconds, you know, um, that, that runs all the time. And so, um, the bottleneck there is actually orchestration and local compute. And so I believe like, like, um, one needs to build a continuously learning local model, um, that can save you on like compaction, context windows, so you... And, and, and try to preserve as much compute locally and rely on the server-side frontier only when necessary, and keeps learning, keeps adapting, keeps evolving. And that model, um- It's not just a model. It's a model plus the harness, plus the local chip and the compute and the ecosystem of devices it controls. That system, um, is gonna be your own intelligence. Essentially, the data center moved to your local device, and you, you get to control it, you get to own it. You don't get to worry about somebody, like, you know, spying on you or looking at all your tokens, very valuable personal tokens. Imagine you have, like, very sensitive deal materials. Let's say you're doing a deal, um, and then, um, a Frontier lab has all your tokens that you use to, like, write a memo. Uh, imagine somebody could hack into that server and steal your deal from you. You wouldn't want that, right?

    11. HS

      I'm gonna be honest, I think there's much more valuable things for people to steal from [both laughing] a London-based VC.

    12. AS

      Sure.

    13. HS

      But yes, I completely-

    14. AS

      I mean, you're, you're not just yet another London VC. You, you have a $400 million fund last time I read it.

    15. HS

      Yeah, yeah, yeah.

    16. AS

      So imagine, like, you know, you're already making your moves for the $4 billion fund, right? So, so e-everyone has certain levels of, like, you know, uh, sensitive stuff. And, and so I think that's where I believe that the 24/7 always-on agent is gonna be realized by the company that wants to play the role of the orchestrator, not the model builder, not the frontier model builder, but the orchestrator. And, uh, and I think that's what, that's what we wanna do. Um, Computer has, has been positioned explicitly as the agent orchestrator. The, the musicians in the orchestra are these sub-agents that utilize these different models. Think of them as the instruments. And, uh, the tools, the connectors, the models, these are all the instruments. And the musicians are the sub-agents, and the symphony is the work, and the system is the orchestra, and, and Computer is the orchestra conductor. That, that, that's how it's been positioned, so ve- what it orchestrates keeps evolving, right? It, it changes. It changes from, you know, models to files to tools to chips to devices. But, but it doesn't even matter. Like, you don't care as long as it orchestrates things correctly and, and, and, and maximizes the token value per watt per user. If you can solve this problem, you will capture the most economic value in AI long term. Short term, it might look like, oh, like, this other lab's revenue is growing, you know, exponentially, this, that. But long term, this is the one objective that truly matters.

  9. 32:50 – 34:57

    Why Perplexity Thinks It Can Become the Ultimate AI Orchestrator

    1. HS

      Who is best positioned to do that?

    2. AS

      I believe it's us 'cause you have the incentive of not token maxing. You have the incentive of delivering the most value to the user. Like, we, we-- every time any part of the AI stack improves, our product improves. Um, since the beginning of the year, Anthropic's models have made tremendous progress. But what's also true is that our revenue has more than tripled since the beginning of the year. Tripled since the beginning of the year. And, uh, we-- and a lot of thanks to model progress made by Anthropic, and we also brought our burn down thanks to OpenAI competing with them and bringing down the cost for the same capability. And now, with progress in open source and local models and local chips, we're gonna move some of the inference back to the local devices and bring down the cost even more. So every time any part of the AI stack, whether it's chips, models, harnesses, any of these gets better, our system improves tremendously. And if our system improves tremendously, our users love it and they pay more. They spend more, and so our business grows. So I think to your question of who's best positioned to win in that world for that objective of being an orchestrator is the one whose product or business benefits from other people's progress at any layer of the stack. And so if Jensen produces a better chip, it's great for us. If Dario produces a better model, it's great for us. If Apple produces a better device, it's great for us. And I c- I, I love the fact that we are able to be a very positive sum player at every layer of the stack and not have to rely on any one person to win.

    3. HS

      When we look at the different providers that-- we said kind of server side versus, you know, uh, on device. When we look at server side, a lot of people have won an AI infrastructure bubble, which I think is funny, stupid, and moronic. To what extent do we have a data center supply problem today from what you see?

  10. 34:57 – 43:18

    The Biggest AI Bottleneck Nobody Can Ignore: Power

    1. AS

      I think the biggest problem is actually in power. So wha-- l-let's break down. What is a data center? Is it, like, that you just buy, like, a bunch of chips from Dell or Supermicro? And, uh, no, that, that's just one part of it. You actually have to go secure land, or you have to lease something, lease a property, and, uh, you have to s- buy a bunch of turbines to generate power, or you have to work with, like, power suppliers, grid suppliers, and, uh, you also have to work on cooling. So there's a lot of other work you gotta put in that is far, far slower. You have to get permits to do all these things. And, uh, and so usually what's happening is, um, there's a lot of lead time doing this. And, um, the models that are, um, already in use today, these have been trained in the Hopper generation. So the Blackwell generation model, I think the first model that's Blackwell generation category is, um, Mythos, and it's already scary. Like, people are already, like, freaking out about it. So imagine that, um, everyone pre-trains a model, um, on like a million or, like, you know, hundreds of thousands of Blackwells. Now, those models are gonna be far more powerful than what exists today. And then the Vera Rubins are coming next year in full capacity. Like, like, all the data centers of Vera Rubins will be in next, uh, you know, used next year. Th-that model will be even more powerful. So I think- We-- There is a certain physical build-out time that always bottlenecks frontier capabilities. That's why there's a value in that layer. Whoever knows how to do this puts it-- puts together a bunch of GPUs and chips and networking and power and cooling, and actually, like orchestrating all this software layer on top and c- you know, is able to convert that into frontier output tokens. That, that, that vertical integration has a lot of value. So that's why the markets are pricing infrastructure companies with a higher, uh, P/E ratio-

    2. HS

      Mm.

    3. AS

      -than companies like Meta, for example. Even though Meta builds a lot of infra, it's valued as a software company.

    4. HS

      When we see like, you know, Meta's CapEx spend and it wanting to increase in the last few days and thinking about raising more and more money to increase CapEx spend, I get it with a lot of the AI providers like your OpenAIs, your Anthropic, because they aren't making money from their AI products. For Meta, the CapEx spend correlates to increasing accuracy on ads, which is like a 6 to 8% bump in revenue. I, I get it.

    5. AS

      Yeah.

    6. HS

      But for the CapEx spend, it doesn't make sense.

    7. AS

      Well, um, I, I, I believe like they're, they're, they are understanding what the market's saying, you know. They-- I don't think they're dumb to not see what, what, what's being said. I think they're introducing a lot of subscription products, um, from what I'm reading. So they're definitely gonna like... Basically, the company needs to not just be a social platform, maximizing engagement and turning that into ad revenue, right? And I think, um, that requires them to launch a lot of like agents, subscription-based products, and maybe even a cloud, Meta cloud that, that rents out servers like what Elon's doing at SpaceX. And, and maybe once they do that, the, the, the narrative might change, right? But, um, to go back to my point, it might not be inconceivable that, um, Micron, the supplier of HBMs, might be more valuable than Meta in the next six to 12 months. It's already at like a trillion, and Meta is like 1.3 to 1.4 trillion.

    8. HS

      Can you help me understand that? 'Cause, uh, memory is already a massive bottleneck. It's increased 5X in price in terms of the COGS.

    9. AS

      Right.

    10. HS

      Um, but people are going, "Wow, Micron is fully priced at this point." Why is it not fully priced?

    11. AS

      Because it's still the bottleneck. Whatever is the bottleneck will command the price. Um, AMD is doing really well because CPUs became a bottleneck again. Agent loops, agent harnesses are all running on CPUs. The tokens are produced by the frontier models on GPUs, but whatever work-- Like, like let's say, like Claude generates a coding script that decides to download 500 files from different websites and then, you know, munges a lot of data and transforms it in certain ways and generates a plot and then hosts it on a website that you can, you can share with your people. All that compute is running on CPUs. Agents are using CPUs more than humans, right? And so suddenly there's a rise in enterprise CPUs, and the beneficiaries of these are like Intel and AMD. So then they get to be the bottleneck. Like, like whoever is gonna be the bottleneck will win. And, and so infra is the bottleneck right now because there's a lot of demand and we just don't have the supply. And so whoever supplies memory, SSDs for storage, CPU compute, suddenly these are all like interesting. Like, um, they're more important than companies that are just building data centers and not knowing how to turn that into a valuable outputs.

    12. HS

      Do you believe your Nabis and your CoreWeaves will be a sustainable multi-hundred billion dollar company in the future, or is it solving a short-term supply problem?

    13. AS

      Um, I certainly think they can be sustainable.

    14. HS

      Yeah.

    15. AS

      Um, I think there are some-- I don't... Like look, I don't know particularly which of those is gonna win, and there's also other players like Crusoe and, um, Firebird, and there's a bunch of companies. It's all about being resourceful. You gotta take power from areas where there's a lot of natural resources, and the cost to bring up the data center is pretty cheap, and the time to bring up the data center is cheap, and your service is reliable. Like if somebody commits to buying 100,000 GPUs from you, um, the service should be pretty good. Um, and, uh, you should be able to secure the supply ahead of time, plan well. Um, and I think some companies are even innovating at the power layer. You know, um, generating their own power is one way to bring down the margins. Uh, and so I think there's certainly like value in that layer because, um, it's hard to replicate work. That's how I see it. You could argue that OpenAI can do all the work that CoreWeave was doing, and that's kind of what they wanted to do with Stargate. But why is CoreWeave more successful at building data centers than OpenAI?

    16. HS

      It's hard to do.

    17. AS

      That-

    18. HS

      It's operationally intensive.

    19. AS

      Yeah, operationally intensive. You gotta focus. You gotta like spend most of your time, um, securing permits, like figuring out power, figuring out like bottlenecks in the supply chain here and there, um, and constantly plan ahead and like test all these systems carefully, deal with like random physical issues that, you know, arise in like, you know, running a data center. There's something called TCO, you know, cost of operations.

    20. HS

      Mm.

    21. AS

      You gotta factor that in. So, uh, that said, I, I, I don't think there's value, um, if you're just like a server renter. If you're just a GPU server rack renter, if you're just leasing it to different companies on certain hourly pricing rates, there's not a lot of value. You have to actually build some software on top, kinda like how AWS did. It's called Amazon Web Services, not Amazon Servers, right? So, um, you have to have some software orchestration on top that allows you to get software margins on top of what you're doing, and I think that's why you're seeing moves like Nabius, um, b- like, like going for the AI model inference, like, you know, taking open source models or hosting your models and, and, and, um, that's a business model of certain other companies like Fireworks and, you know, um, Base10 and all that. But you could imagine Neo clouds just going for that business.

  11. 43:18 – 48:57

    Can Inference Companies Become the Next $100B Giants?

    1. HS

      That was exactly gonna be my question. So I just had the co-founder of Nabius on the show, and, uh, the really clear takeaway was the, the challenge that he has, which is there's a huge amount of money that wants just capacity and compute-

    2. AS

      Yeah

    3. HS

      ... with the awareness that he needs to build a full stack product if he wants to have a long-term sustainable business.

    4. AS

      Yeah.

    5. HS

      That was the core realization for me. When I look at the inference layer, like you said, Fireworks or Base10, how do you think that plays out? Do we have standalone $100 billion companies in inference alone, or do we see that commoditized?

    6. AS

      Possible. I mean, y- y- it just, it's all about working backwards, like what does it take to build a $100 billion company? Assume like a 10-

    7. HS

      $10 billion in revenue.

    8. AS

      Exactly. $10 billion in revenue, 30 to 40% gross margins, good amount of net income, good cash flow. Okay, $10 billion in revenue, um, is not that inconceivable for, for a company that can both do AI hosted inference and server capacity and data center build-outs very operationally well. It's all about like, you know, there, there are some factors beyond their control, like open source models continuing to be awesome. If open source models stop to actually be good, where the gap between them and the frontier is like more than 12 months, or like 15 months, 18 months, then I don't think these companies really have a business model because, um, they're not gonna be able to host... They, they're only gonna be able to rent capacity to OpenAI or Anthropic. And so, um-

    9. HS

      That's exactly what Roman at Nabius said. He said if consolidation happens and there's Anthropic and OpenAI or, or two or three dominant providers, that is the biggest threat to Nabius.

    10. AS

      That's correct. Yeah. And so, um, but y- y- you gotta make a leap of faith assumption that, you know, like the, the models from China or NVIDIA's making good progress on their models and Nemotron. Um, you got-- So there, there, there's gonna be enough factors in the market to keep, uh, consolidation as an outcome from, from like sort of stopping from happening. But you don't control your own destiny if you're those companies. That, that, that's basically the problem.

    11. HS

      Totally get that. Okay, so we can have standalone companies that are $100 billion in inference alone. Sorry, [chuckles] I'm just pillaging you for your knowledge. When we look at the model selection companies like an OpenRouter or like Factory AI just released their kinda model selection or model routing product-

    12. AS

      Yeah

    13. HS

      ... which did very well on launch. Is there $100 billion companies in the model selection and routing business?

    14. AS

      Probably not. Um, I think you can't just be a provider of router. You have to use the router to produce something meaningful. Um, actually, most of the business value of OpenRouter is less than the router, even though the product is called OpenRouter. It's not routing across models there. It's actually just routing across different endpoints of the same model. So, um, wha- okay, so maybe I- let, let's, let's a- let's ask this question. If you wanted to use Claude Opus or, um, I don't know, like GPT-5 as a developer, why would you not wanna just use it with your own API key versus using it inside OpenRouter? Number one argument. The single simplest argument as to why you would wanna do that is model fallbacks. Sometimes your API keys might not have the rate limits, or even if you have the rate limits, it might, there might be an error on OpenAI servers that, you know, don't guarantee you the response time you need to run your application. And, uh, OpenRouter would go and earn the sup- uh, you know, they would pay for a capacity for like one year ahead, uh, with the funding they have and secure the rate limits and multiple endpoints across multiple different providers of OpenAI models, be it Bedrock or Azure or OpenAI themselves. And so that routing is valuable. It's essentially an infra problem they're solving, which is reliable token supply. It's not actually, oh, like they're lowering the cost by deciding if this prompt should go to like GPT or Claude or something like that. That, that's not what they're actually selling to the developer. That's not actually the business model. And, uh, and then for a, a lot of these Chinese open source models, there's not... You, you probably don't want, um, your API tokens from going to like... Let's say you don't want your API tokens going to China. Um, and so y- and, and let's say you don't have the bandwidth to work with like different inference providers or verify who's good and, you know, who's not. You're just trusting OpenRouter to take care of all that, and then, you know, they're gonna like supply the tokens to you. So it, it's, it's routing not at the level of like, oh, like deciding which model is cheaper or task. It's more like, um, a reliable token supply. And I think there's some value in that layer, definitely. Uh, otherwise they wouldn't have these many, um, users and these many trillions of tokens being routed a month. But it's, um, it's not like, you know, high gross margins business. It's, uh, the way the business model works for them is actually, um, they would secure a discount from the model providers by guaranteeing a lot of supply, but they would still charge the user listing price on the API, and that difference is their margins Do you understand?

    15. HS

      I, I, I totally get you.

    16. AS

      Yeah.

    17. HS

      We spoke about bottlenecks, and you said about HBM, high bandwidth memory-

    18. AS

      Yeah

    19. HS

      ... and Micron and the value that, you know, uh, they have to say and what it can be. What bottleneck will, will we have in three years that we're not discussing today?

  12. 48:57 – 54:04

    The Next Massive AI Bottleneck (and Why It’s Not Models)

    1. AS

      I think power will remain the bottleneck.

    2. HS

      Yeah.

    3. AS

      I think it's-- It feels like that to me, unless something dramatically changes in the way data center build-outs happen. Uh, I actually believe that there'll be a lot of resistance to building data centers. It's because people incorrectly think that data centers consume a lot of water or eat up a lot of power, which isn't-- both, both are untrue. Satya even made the statement that it, it's like a can of water or something, uh, in terms of how efficient these companies are. And-

    4. HS

      Do you think that's why they're putting up resistance to them? I don't. I think it's 'cause it's a symbol of job losses, uh, increasing wealth inequality.

    5. AS

      It's a lot of things. It's a lot of things. Um, it's a lot of apprehensions, um, fear about, like, what's gonna happen, channelizing in so many different ways. Um, sometimes it's channelizing through hatred for wealth inequality and, like, wanting to tax people. Sometimes it's channeling through, like, concerns for the environment and, like, climate change. Um, sometimes it's, uh, channelizing in a way where, um, you're all like, "Oh," like, "the price of the grid is going up because you guys are building all these data centers." And then-- Or like, "I'm paying more for my phones and laptops now because the RAM prices have gone up because you guys went and bought all of it." So I think there's a lot of different ways in which it's getting channelized, but the common sentiment is, um, like, like a pretty bad sentiment about AI.

    6. HS

      Do you think it will be meaningful to the development of those data centers? I think right now forty out of a hundred are not being developed because of public resistance.

    7. AS

      Yeah. So, um, tha-that's where the power bottleneck is, and, um, you could see maybe certain countries seize the opportunity for this and, um, um, allow these model builders to go build data centers there. Um, Elon's going to space to do that. Um, so that's gonna be an interesting experiment, um, because there's a lot of energy from the sun that can be harnessed there, and, uh, there's a lot of natural resources in other countries. Regulations might be more friendly. So we're still gonna see data center build-outs. It might not happen in the US. And, um, but, but the fact that you have to solve physical problems, like you actually have to deal with the supply chain, the permits, securing power, like making sure, like, things work and getting the lead times lower and lower. You're not solving problems like cloning some SaaS apps here, right?

    8. HS

      [laughs]

    9. AS

      You're-- Or like you're building a go-to-market team or like, um, doing better marketing against the competitor's products. Like, yes, those are also hard problems, but these are, like, much harder problems where, like, you're not in full control of your destiny, and you need a lot of capital and connections and, like, the right people, uh, sometimes even, like, political help to unlock progress. And so that's why this will continue to remain the bottleneck, in my opinion. And there's a lot of risk as well because if you do encounter another DeepSeek moment here where there's a vastly more efficient model that's been built with a very different vertically integrated architecture, and you built out all this capacity and you're like, "Damn, that's-- I overbuilt. There's something far more efficient that can run on, on, on people's local devices, their MacBooks, their Windows PCs." Yeah, like you're probably freaking out then. And so you hope that-

    10. HS

      How likely do you think that is, though?

    11. AS

      Uh, it's probably like twenty percent, thirty percent chance. The reason I think there is some possibility is that because of the export controls, um, you are-- So the DeepSeek is not building with the NVIDIA stack. They're building with the Huawei stack. And because there are export controls on not just the NVIDIA GPUs, but also on HBMs, these, um, architectures that, that DeepSeek's building are far more, like, memory efficient. They've made innovations on the KV cache to be really small enough that you can host it on the SSDs, and you don't need high bandwidth memory for inference time. And they're gonna have a completely different architecture for inference, completely different architecture for storage, 'cause they're not allowed to use the 3D NANDs. So their architecture is gonna look-- It's not just the model architecture. The model architecture is already pretty different. They've made innovations on the attention layer. They've made innovations on, like, the, the training algorithm so that it doesn't consume a lot of interconnect capacity. So they, they made a lot of-- They, they-- Basically, their whole stack is getting vertically integrated to their hardware and their chips and their fabs and so on. And so that's a very different bet from what America is making.

  13. 54:04 – 58:27

    Did U.S. Export Controls Accidentally Make China Stronger?

    1. HS

      Do you think the export controls have helped or hurt us?

    2. AS

      Jury's still out. Short term, it's helping because the only reason I-- my belief, the only reason where-- why, uh, there's even like a development gap between open source and frontier is export controls. It's definitely helped and, and, and definitely, like, companies like Anthropic lobbied very hard for it. But, um, there is a chance that because of that, they now get really good at the physical layer. And one advantage they have is they can actually build data centers a lot, lot faster. Power is not a problem. Permits are not a problem. People are not a problem. Labor is not a problem. Expertise is not a problem. And so by forcing them to go out there and build all this, you're converting them to a far more like potent competitor.

    3. HS

      Do you think we still dramatically underestimate China's capabilities?

    4. AS

      I think so. Because if AI is like not just digital, there's also physical AI, you gotta build fabs, robots, chips, and harness the energy really well and, um, package it into local devices. I think they have a lot more advantages than America.

    5. HS

      How important is it that we have our own TSMC in the US?

    6. AS

      So TSMC is actually-- There is a fab of TSMC in Arizona. Like not a lot of people talk about this, but TSMC is investing like $150 billion into that, into, into building American fabs. And, um, they've already invested $40 billion or something like that, sixty billion last time I checked. So there is a TSMC in Arizona that's coming up. There's also, um, Intel, and that's why, you know, American government owns ten percent of Intel. Uh, Nvidia and SoftBank own five percent each. So there is a lot of investment going into Am- an, an American fab, as well as TSMC is investing into its American fabs. Elon's building Terafab. Like I think it-- people have woken up to the importance of building fabs. But, um, this is also why China is, uh, particularly very, very competent because-

    7. HS

      Given the capabilities of China that we just mentioned that really articulately, I know it's a ridiculous question, but sort of, um, if I were to say to you, your job is to make sure America stays competitive, what would you do to ensure that you retained competitiveness in an increasingly strong China?

    8. AS

      I think, I think take physical infrastructure a lot more seriously and continue funding it, um, and not like have all these... You know, I, I wouldn't say meaningless. It's more like not propagate fake news around data centers, um, about how data centers are polluting and contaminating water or like they're sucking up all the water, um, and, and, and actually be fact-driven. And so, you know, I hope our product helps there. Like you, you can, you can go to Perplexity and ask any question and get fact-checked on your assumptions. But yeah, like it's very important that we educate the public, um, about what's actually going on in a, in a language they easily understand and not fearmonger. Okay? Like not be like, "Oh, all their jobs are gonna go away," like this, that. Like there's gonna be lots of amazing companies that are gonna get built with far fewer people getting multi-billion dollar, multi-hundred million dollar valuations with like 20, 30 people and propelling like a trillions of dollars of new GDP. Like let's talk about how to enable that. Let's talk about how to build that and create a more positive future together, right? Uh, instead of, "Oh, like 90% of the jobs are gonna be gone. Like you're all gonna get screwed over by our, our models," and like, and, and it's, it's our, it's our moral duty to tell you all this. Like blah, blah, blah. Like that doesn't make any sense to me. Like you can't win by saying that and also like complaining about not being able to build data centers fast.

    9. HS

      Do you think we've done a complete disservice by having the marketing message that Dario has had that all jobs are going and it's all doom and gloom?

  14. 58:27 – 1:01:06

    The AI Jobs Narrative Is All Wrong

    1. AS

      Yeah, I think so. I mean, I think, you know, they have contradictory messages in their own like p- uh, different social engagements so far, where the most recent one I heard was there is no evidence that AI is taking over jobs. But so I, I, I think there, there needs to be a consistent communication around this. And I also think that, um, very little is being spoken about how AIs can help you build companies in a very, very different way. Like the current AIs, where agentic AI. It, it's already true that so many things you would hire people for, you can do it with agents. But one way of looking at it is like, oh, like what happens to all the jobs? But the other way of looking at it is like, hey, like I can-- I never had the chance to go build out a company on this idea that I've been having all this ti- all this while, and maybe me and a group of friends can come together and build this. And can you guys figure out a way to give us compute credits or... You know, Amazon gave a lot of compute credits to a lot of startups. Like when we started Perplexity, we had like around $200,000 worth of Amazon credits and GCP credits and Azure credits, um, that almost like together cumulatively this was worth like a million dollars in compute credits. Now, the, in, in today's world, it's gonna be like a million dollars of computer credits, and we are doing that. Like we, we are funding this thing called the Billion Dollar Build, where we're giving a million dollars of computer credits to any group of people who have a credible path to building a billion-dollar company. And I want like thousands of such companies to be built lately.

    2. HS

      What did you think of Sam Altman giving $2 million of tokens to YC companies initially?

    3. AS

      I think we should do more of that. Yeah. That, that's the right thing to do. Like we should do a lot more of this because you want new companies to be built. Um, and, and, and even if they're worth multi-hundred million dollars, right, it's good. If there are thousands of them, like that's a lot of new GDP.

    4. HS

      I, I, I spoke to Anne Bordetsky before the show, and she said how AI-pilled the team is for you. How big is the team today?

    5. AS

      It's like 400 people.

    6. HS

      400 people. How big will it be in two years' time?

    7. AS

      I don't know. It's hard to say. Maybe 800 or 1,000

    8. HS

      So will companies follow the same headcount trajectory that they have always followed, and we will just solve new problems? Or will they be dramatically more efficient with a much fewer

  15. 1:01:06 – 1:13:07

    Why Future Unicorns Will Need Far Fewer Employees

    1. HS

      number of people?

    2. AS

      Definitely they'll be dramatically more efficient, right? And, and that's why I, I, I am a believer in building a lot more efficient companies n- and, and being an example for all these companies ourselves. Like, like, people should look at Perplexity and be like, "Oh, like, with 400 people, um, you can build like a multi... Like, like, I don't know, like 20 billion, $20 billion company." Um, and so that means with, like, 40 people, I could probably build a billion dollar or $2 billion company, you know? And that, that, that's totally doable. Totally doable. And, um, and, and so for us, maybe that means is with 4,000 people we could be worth 200 billion. We, we could be worth $2 trillion with, like, 10,000 people. You know? I think, I think that doesn't mean it's bad for all the, um, 100,000 people we did not hire for a typical $2 trillion company. I would rather have those 100,000 people be split into groups of, like, 100, 1,000 groups like that, and each of those 1,000 groups are worth a few billion dollars. That's awesome. And I, I think a lot more people need to be entrepreneurial. Um, there are people who would be bad employees in any company because they're just, like, difficult to work with. They, they don't listen to, like, instructions, so, like, they don't follow, like, roadmaps. Uh, or not, they're not, like, easy to collaborate with. But maybe the, the flip side of that is those, those are the kind of qualities that founders typically have.

    3. HS

      Aravind, there is a population and a very large population that are not AI native people, that are not using AI to improve workflows, improve efficiency. What would you advise them?

    4. AS

      Get started. First steps, get started. And, and, and channelize your curiosity, right? Um, you don't need to use AIs to do your existing work. If that's... If your existing work is boring to you, you probably won't enjoy it even if you use AIs to do it.

    5. HS

      You got a lot of heat for saying, uh, uh, people don't like their jobs, so...

    6. AS

      I, I didn't say... If you actually listen to my interview, I did not say that. So people want clickbait articles, and they take something I said in one sentence and out of context and make it into a headline.

    7. HS

      What did you say?

    8. AS

      I, I specifically said this: "Hey, like, there are a lot of people who don't enjoy their jobs." Does any... Like, by the way, the fact that, that, that, that thing went viral is not because I was completely wrong. I think a lot of people resonated with the fact that I was actually honest in saying a lot of people don't enjoy their jobs, and that has nothing to do with your economic position or standing in society. You might even be, like, really wealthy, but doing a job that you completely don't enjoy and, like, destroying, like, the peak years of your adult life working on something that is horrible or, like, like, depressing. So, like, my point is that if that's you and if the reason you could never leave your job is because you were always worried if you... How, how would you build a company from scratch? Like, there are all these things to figure out. How... You have to hire a lot of people. You have to, like, set up an office. This, that. Like, that's changed. For the first time in history, you can get started on an idea with, like, one or two other friends and, and, and, and maybe have a real genuine shot at building a billion-dollar company.

    9. HS

      I totally get that. Everything that we've discussed today has been on the back of unprecedented demand up and to the right. We need more memory. We need more data center supply. We need on-demand and server si- Everything's, like, up and to the right. Seeing some cracks in and Uber's saying, "Hmm, I'm not sure I'm getting the productivity gains that I thought." Microsoft aligning with them, putting a $1,500 token budget. Uh, do you think we will have a continuous up and to the right acceptance, that productivity gains are unwavering, we have to do this, or will there be falterings along the way?

    10. AS

      I mean, I'm sure there's gonna be falterings along the way, and people are rightfully freaking out about token maxing, which is why I think you need some form of hybrid agentic inference. You need some amount of inference compute to run locally that you're not paying for tokens on. Um, unmetered intelligence, essentially.

    11. HS

      How will the best companies of the future structure token budgets?

    12. AS

      My hope is that they don't have to understand that. They will be able to work with an orchestrator who does it for them. It's not gonna be easy for you to constantly keep track of, like, which models are the best at what things, and how do you allocate, "Oh, this is the budget for coding. This is the budget for finance." So, like, like, how do you even understand, like, which models are good at each of those things and, like, how much do you spend on each of these divisions? You're not gonna be able to keep track.

    13. HS

      I, I, I had a friend on the show the other day say that Google will be the token king. They can produce the lowest cost tokens out of anyone. They own full stack TPUs, data centers, networking, power, procurement. Do you think that's true that they will be the lowest cost token producer?

    14. AS

      They have advantages, all, all advantages one needs to have to be that, but they underestimated the importance of coding models, and so they're far behind the frontier right now. So again, they could catch up. Totally capable. Totally competent team. But today they're not quite at the frontier.

    15. HS

      I was shocked the other day. I saw the Cloudflare announcement that, um, now agent traffic has overtaken human traffic-

    16. AS

      Mm-hmm

    17. HS

      ... for them.

    18. AS

      Why, why are you shocked?

    19. HS

      It was quicker than I thought.

    20. AS

      Okay.

    21. HS

      Personally. I thought that would happen, but in two years, maybe not now. How does the world change when agent traffic far exceeds human traffic?

    22. AS

      I think people are just gonna have a lot more agency. That's it. The, the, literally, literally the-

    23. HS

      But do websites go away? Does design not matter? Does the advertising model of the internet die completely?

    24. AS

      No, it doesn't. Because my belief is that the advertising model around, like, travel or shopping or, like, um, fashion are not getting disrupted by agents because the judgment is not objective. Any- anything where the judgment is objective, the transaction is based on objective judgment, that's gonna get disrupted by agents. Anything where the transaction is more subjective, like the decisions are more subjective. Like, like what is the best piece of furniture inside this, this, this spot? Like, why this particular table? Or like those kind of things. Probably for the mic you would buy an objective decision.

    25. HS

      Mm-hmm.

    26. AS

      The table, you probably are caring about the aesthetics of the room. I think, I, I think that's kinda how I, I feel the world will split, and subjective things will still be ad-based. Objective things will be agent-based.

    27. HS

      I watched your commencement speech on the back of speaking to Sam at Accel, and he said I had to watch it, so obviously I watched it. Um, and one of the points you made was the defining skill of the era is asking better questions.

    28. AS

      Yeah.

    29. HS

      What question is no one asking today that maybe everyone should be asking?

    30. AS

      I think people need to ask more about, like, okay, assuming I have a lot of agency a- available to me, what do I do? Imagine, like, I gave you a head count of, like, 100,000 people or 10,000 people and, and, and, you know, enough compute credits to run those agents. What would you do? Like, let's say I ask you, Harry, like, you know, let's say I g- you have suddenly, like, 10,000 agents at your disposal. What would you do? Like, I, I remember you telling me, or, or not me, b- but, but in some episode of yours where you said you only, you only did this podcasting because you felt like you didn't have an arbitrage to go win deals.

  16. 1:13:07 – 1:19:44

    Wealth Inequality, AI & The New American Dream

    1. HS

      more likely.

    2. AS

      Yeah. That's true actually for even people Like, it, it's m- way more likely for a person with $100 million in, in liquid net worth to become a billionaire than someone with $10 million.

    3. HS

      Are you not worried about the wealth inequality? I- I mean, if we, we're being blunt, we both are very lucky now to live in kind of nice worlds and rarefied airs. Are you not worried by just how much money a very small number of people have, and how fucking hard it is for everyone else, and that gap is getting bigger?

    4. AS

      I think the way to, like, ensure that that's not, that, that doesn't remain the case is to distribute the benefits more widely. You got, you gotta let any... By the way, the people who are using our tools... Like, I had an Uber driver. I- I'm, I'm not even, like, making this thing up, so, um, as honest as this, as it can get. An Uber driver in San Francisco, um, once told me that he watched one of my, uh, YouTube interviews where I explain how you can build a product or a web app with an AI from scratch. Went on to do it, and, um, um, used AI to add, like, billing and all that, and that makes more passive income for him than, um, r- driving Ubers. And so he actually reduced the amount of time he's driving Uber because he, he loves live coding new apps. And, uh, n- that, that already tells you that for the person with agency and a positive outlook for the future, anything is possible. And so if you keep communicating all the negative things you can about AI and wealth inequality all the time, and that's the only thing news, uh, and press writes about, I think it'll perpetuate, and people will only think the bad things. And so it's, it's, it's very essential that if you think you're already doing well, it's very essential that you talk about what are all the things that can go well, and give hopes to people who were once upon a time like you. Like, even you were, you didn't... You, you started this, um, podcasting circuit, like, when you had nothing, right? So-

    5. HS

      Nothing

    6. AS

      ... exactly. So it's possible. So you gotta, you gotta talk more about that than be like, "Oh, I feel so guilty that I made it, and now I'm like, uh, you know, what about all these people who haven't made it?" Like, you can also make it. Like-

    7. HS

      I, I think I have a more pessimistic view of actual general public, which is I don't think that many people have agency. I think a lot of people have victim mentality.

    8. AS

      Yeah, but you gotta, you gotta help them. Like, I think that's, that's the most important thing.

    9. HS

      I think they gotta help themselves.

    10. AS

      Sure. But people will help themselves once they see that, okay, like, I kinda wanna be like this guy. Let me, let me work hard. You need an example, right? Um, it's not like nobody can become, um... get in shape. Like, it, it takes discipline.

    11. HS

      No.

    12. AS

      Takes discipline. You gotta get rid of bad habits. And so, um-

    13. HS

      And now is the best time ever to change your life in 12 months.

    14. AS

      Exactly.

    15. HS

      Like, the ability to go from nothing to, to actually billionaire in 12 months is now possible-

    16. AS

      Possible

    17. HS

      ... in some respects.

    18. AS

      Yes. And so I... Look, I'm not saying everyone's gonna make it and everyone's-

    19. HS

      Yeah

    20. AS

      ... gonna be worth a billion dollars.

    21. HS

      Isn't that the caption from this, this show? Aravind, everyone's gonna make it. [laughs]

    22. AS

      Anyone has the potential to make it.

    23. HS

      Mm-hmm.

    24. AS

      So it's, it's, it's as likely for Perplexity to become worth $2 trillion as, as a founder who's yet to secure your funding to worth, be worth a billion dollars. So it's, it's equally hard. Equally hard. And, um, and I think you just have to give yourself, you know, shots at the goal, and, um, be curious. That's, that's the message from the commencement speech. Be curious.

    25. HS

      We have SpaceX, we have Anthropic, we have OpenAI going public. It feels like someone's kind of shot the gun and the race is on. Is there enough money to fund three such large IPO-

    26. AS

      There will be some reallocation, for sure. Like, um, there, there might be some holders of, like, SaaS stocks who would put it into Anthropic or something. Let's say you believe that enterprise AI is gonna take off. You might wanna hedge between having a lot of Microsoft stock and Salesforce stock versus, like, putting some of that into Anthropic. So let's say, like, Vanguard or BlackRock own like, you know, cumulatively they own, like, $200 billion of Microsoft and Salesforce. They might be like, "Okay, I'm gonna take 30, 40 billion of that and put it into Anthropic." Fine, you know. Not a bad bet to make.

    27. HS

      What happens to all the enterprise SaaS companies that are public, growing, eh, fine?

    28. AS

      They have to, they have to weather the storm.

    29. HS

      Is it a storm or is it a continuous precipitation?

    30. AS

      I think you have to bring down the costs and produce new value. Salesforce has done well because they always went and bought the next thing. If you're just selling the same software, you're probably not gonna be around. IBM is still around because they went and bought Red Hat and HashiCorp, and now they're buying Confluent. So there are ways for these companies to stay alive and extend their lifespans and stuff. It's obviously gonna be hard to preserve a brand that's as relevant. Like, like, I don't think the IBM brand is that relevant anymore in terms of, like, evoking an emotion in people to go use their products.

  17. 1:19:44 – 1:21:41

    Why Perplexity Is Training Its Own Models

    1. HS

      two to three years?

    2. AS

      We're training our own models, post-training it on top of amazing open source models, and that will bring down the cost that we currently spend on frontier model tokens. We expect to continue to use frontier models for designing new experiences and new capabilities that do not exist today in our products. But whatever exists today in our products right now, we expect it to completely re-rely on, like, models we own and serve ourselves, and that's al- gonna be the best way to bring down the costs and increase our margins.

    3. HS

      Will the largest enterprises in the world all be fine-tuning open models to have tailored models that are much more specific to them?

    4. AS

      Absolutely, because it's in your incentives to bring down the cost.

    5. HS

      Does that not provide another bad case for the large frontier model providers?

    6. AS

      Frontier model providers will only remain relevant if they remain at the frontier. If for six months you're not seeing a new capability, it's bad for them. And so that's the uncomfortable nature of this field. You-- no one's ever in a comfortable position. Like I said in, in the start, no one's-- no one can relax. This is-

    7. HS

      Fucking hard. [laughs]

    8. AS

      Yeah.

    9. HS

      In business, it's got harder. Like [laughs]

    10. AS

      It, it's gonna get even harder. And, and, um, that's the nature. It's just the, the price is too big. Like, you've never seen... Uh, like, like, take Anthropic. I think it's, um, worth, like, one to one and a half trillion, some-something in that range. That's basically the valuation of Meta. And, and, and this all was created in, like, six years. Meta took, like, 20 years to build. So the price is so big, and so no one can, um, no one can be comfortable. And, and, and, and, and anyone who's winning today can lose tomorrow, including, including the model providers.

  18. 1:21:41 – 1:23:15

    “Perplexity Was Voted Most Likely to Fail”

    1. HS

      Pre this year, there was, like, a three-month period where people were like, "Oh, Perplexity. What's happening with Perplexity? Oh." Do you pay attention? Do you give a shit?

    2. AS

      Of course I pay attention to all of that.

    3. HS

      Do you care? There was one in particular in San Francisco, do you remember, where they were like, "Oh, what's the company you'd short?"

    4. AS

      Yeah, we were voted the most likely to fail. Cursor was voted the second most likely to fail. OpenAI was voted the third or something. Um-

    5. HS

      You didn't give a shit?

    6. AS

      I, I feel like we're all doing well.

    7. HS

      [laughs]

    8. AS

      Cursor, I think, is getting sold to SpaceX. OpenAI is, uh-

    9. HS

      Going public, baby

    10. AS

      ... going public soon. We tripled our revenue since that, since that, uh, judgment was made, so brought down the burn by more than 50%. So I, I don't know. Like, my, my, my sense is that, um, I also feel most of those people who w- sit on these, like, meetups and vote don't actually build anything useful.

    11. HS

      [laughs]

    12. AS

      Well, yeah.

    13. HS

      Uh, okay. We're gonna do a quick fire round 'cause, uh, I could talk to you all day. Um, first one: What's one widely held belief that you think is completely wrong?

    14. AS

      I think a lot of people are obsessed about, like, you know, identifying a moat in the first year or two of their company. But, um, I think, like, the only shot you have is to move fast. Like, velocity-- In, in my mind, like, moving fast is a way of expressing humility 'cause you- you're constantly making contact with the world and trying to question your assumptions all the time.

    15. HS

      Where are you still moving too slow internally today?

  19. 1:23:15 – 1:25:38

    Turning Perplexity Into an AGI-Powered Company

    1. AS

      I think we can be even more AI-built. It's insane I'm saying this because we are building some of the most interesting AI products, and internal adoption of our own products or competitors' products can be even higher. And, and, um, this is despite us being extremely, um, agent-built internally and trying to delegate as much to agents. Yeah, that's where-- That, that's, like, a big area for... My, my hope is that we can turn this company almost into an AGI and, and, um, have-- That doesn't mean no hu- no humans work here. But there will be an AGI that has all the context it needs to run different divisions of the company in a semi-autonomous way, with some scaffolding provided by humans here and there. And, and that's not-- That's gonna feel-- That's not gonna feel scary at all. We'll normalize that, that feeling very fast. It's just gonna feel like 10-- you know, uh, the 10 X engineers running certain aspects of the company.

    2. HS

      If I gave you unlimited money, what would you do today that you're not doing?

    3. AS

      I would build data centers.

    4. HS

      You would?

    5. AS

      Yeah.

    6. HS

      In space?

    7. AS

      I don't have expertise to do that, but I would, I would start with land on Earth. You know, I think there's a lot of land and, and maybe you can be resourceful in securing permits and power in different countries, but I would start there. I think it's... You know, I, I-- Like I said, like, I think physical infrastructure build-outs is, like, the return of the industrial age again. Like, like, the, the, the forefathers who built the Industrial Revolution, um, oil pipelines, steel bridges, factories producing cars, all these things that we take for granted today were built by people who, you know, spent a lot of time thinking about how to scale these things in a cost-efficient way. And so, um, we need to do that a lot for AI and, um- Yeah, that's what I would do. Of course, you, you, you cannot just be building infra. You need to be able to utilize all that infra to producing valuable output tokens to user. But, um, we're already good at doing that, so infra is the thing I would focus on.

  20. 1:25:38 – 1:32:44

    SpaceX vs OpenAI vs Anthropic: The Best 10-Year Bet

    1. HS

      You can buy and hold for 10 years SpaceX, Anthropic, or OpenAI, the three IPOs coming in the next few months. Which do you buy and hold for 10 years and why?

    2. AS

      SpaceX.

    3. HS

      Why?

    4. AS

      It's an n of 1 company. Like Anthropic and OpenAI can claim they do whatever each other does, but, uh, um, SpaceX is the only company building space infrastructure for connectivity. Have you been on a flight with Starlink?

    5. HS

      No.

    6. AS

      You should. You will hate being on a flight without Starlink after that. Imagine we can record this-- I can watch this podcast while flying on a plane. Starlink lets you do that. That's just one aspect of the business. That's just one aspect of the business.

    7. HS

      One small aspect of the business.

    8. AS

      Yeah, like there's a lot of g- Like I'm excited about possibilities to travel from Australia to like, uh, San Francisco in like 30 minutes. You know, all this feels like sci-fi, but I'm excited about all these possibilities.

    9. HS

      What job does not exist today that will be incredibly common in five years' time?

    10. AS

      I think it already exists. So th- if-- The forward deployed engineer is definitely on the rise. I guess like people with a really good sense of like, um, quality control. Maybe a better, uh, way to answer this is like most jobs that exist, like valuable jobs that exist are usually like reincarnations of something that already existed. Like, so I don't think we're gonna see completely new things. It's just gonna reincarnate in different ways.

    11. HS

      You can advise your little sibling who's finishing university today and just done a computer science degree one thing. What would you advise them?

    12. AS

      Stay curious. Don't, don't, don't give into like FOMO and trying to max out on something here in the short term. Like, don't go to Twitter and feel like a loser that people on, on, on one of frontier labs are getting so rich and like you-- everything feels hopeless to you or something. Like, there is so much more to build. Like, we are just getting started. Like the, the, the, the application layer era or like infrastructure build-outs, th- there's like a lot of opportunities.

    13. HS

      We are seeing more spin-outs from OpenAI, Anthropic, you name it, every single day. Do we have hundreds of these neo labs and vertical models?

    14. AS

      No. Not, not a big believer in too many of them. I think you gotta produce some differentiation. That's the most important thing. Um, like, like I, I-- If-- Would you call DeepSeek a neo lab?

    15. HS

      No.

    16. AS

      Why?

    17. HS

      I think very stupidly for me, I don't call it a neo lab because I, I attribute neo labs to like spin-outs from larger labs-

    18. AS

      I see

    19. HS

      ... and, and kind of verticalized, which is probably wrong on both pa- pa like-

    20. AS

      Yeah

    21. HS

      ... axes, but it's horizontal and it's not a spin-out.

    22. AS

      Yeah, I mean, I, I kinda like the idea of labs taking a differentiated bet. Okay, if somebody really questions the transformer architecture itself, or somebody really questions needing to build on Nvidia GPUs or something like that, like, like foundational bets makes... Or, or somebody questions or somebody bu- goes out and builds robotics models, I think, I think that's like somewhat uncorrelated and different, and that makes sense for a lab. But I feel like there are like just labs for the sake of being lab, and I don't think they're gonna make it.

    23. HS

      Can you paint for me-- I, I do like this one. What is the most plausible story where Perplexity becomes a trillion-dollar company? What do you do then? The orchestration layer? The...

    24. AS

      I mean, accuracy and orchestration is, is, is like two goals that have been consistently true since the beginning of our company. So I think we'll continue to do that. We'll be orchestrating across devices, chips, models, tools, files, connectors, everything, right? So what would I do once that happens? I don't know. We'll, we'll, we'll chart our path to 10 trillion.

    25. HS

      Are you happy now? Like are you, are you enjoying this?

    26. AS

      Of course. I mean, I wouldn't-- Like there are so many things I could be doing if not for this. And, um, I think the process is w- is, is what motivates you. So you asked me, I think, somewhere in between, you need to give me a number of where you want. I don't, I don't work like that actually. I, I-- Like for example, like the, these numbers like getting to 2 trillion or 20 trillion are, are exciting, but like that, that doesn't motivate me. It's, it's hard to get motivated by wealth. You, you, you wanna get motivated by impact.

    27. HS

      Who's the smartest person you've met? Final one. You've met Jensen Huang, you've met the best of the best. I've been fortunate enough to. Who's the smartest?

    28. AS

      Um, people are smart in their own ways. It's hard to compare. Like I've met Jensen, Elon, all these guys, and l- like Bezos.

    29. HS

      What was it like meeting Elon?

    30. AS

      Amazing. I mean, Elon's like a very, uh, focused person. Like, like, uh, he, he might not appear that way on Twitter, but you know, with a lot of like random tweets, but he's extremely laser-sharp focused on whatever he's doing at that moment in time. Actually, the, the one skill that as an entrepreneur that I would really like to ta- like build from somebody like him, like take from somebody like him and have it for myself, is that ability to just zone out of all the other things that's happening in your business or other businesses and just focus on that limiting problem right now, like the bottleneck problem, and, and, and, and ignore everything else. It's very hard to do. Like l- even within Perplexity, I, I cannot just focus on like one part of the business alone. It's very difficult. Like, I'm, I'm always looking at other things simultaneously. And, um, his style is to just always look at the limiting problem and just ignore everything else. And, uh, that's very hard to do because you, you actually have to be really good at concentration. You, you have to be really good at ignoring even important things which are distractions to your core objective right now.

  21. 1:32:44 – 1:35:05

    Elon, Jensen & Why You Should Never Retire

    1. AS

      learn. I think, uh, there's one aspect of like, you know, being comfortable where you are, thinking you made it. Uh, you know, that that feels good to get here so far. But these guys are not stopping. Like, I, I don't think Elon wants to stop at, uh... If you look at his pay package with SpaceX, it's structured around, um, creating a colony in Mars with a million inhabitants and, uh, building enough compute in space. So that's why it, it's not like motivating to be worth a $10 trillion in net worth or something. You know, if, if he does these things, I'm sure he's gonna get there. But, um, it's more motivated around, like, making the impossible things happen and, and, and having, like, that long-term outlook. Like, you, um... I think that has been the biggest thing to learn from maybe these two individuals in particular, is, um, a lot of people view this, like, entrepreneurship as like, "Oh, if it wins, if, if I win and have a great outcome and I sell my company, I would have, like, generational money. I don't have to work ever again." And then what? You end up, like, like, just staying at home and, like, your, your, your kids will obviously have, like, trust funds, and they're not gonna get inspired watching their dad-

    2. HS

      Play padel?

    3. AS

      Yeah. You know? You're not gonna set the right example for them. They're not gonna be able to take your wealth and multiply it 'cause they a- they didn't watch somebody who actually did that. You did it, you did it before they were, they, they were, like, adults. And so, um, I think you always need to be doing something. Um, like, like Jensen said some- recently that he hopes to die on the job or something like that. Like, that's the attitude you need to have. Like, you got... You, you need to work forever.

    4. HS

      I was so upset though when Jensen said, "If I'd known how hard it was going to be, I wouldn't have done it." When he did-- I don't know if you saw that interview. I was like, "Oh." Like-

    5. AS

      I-

    6. HS

      ... you know?

    7. AS

      Yeah. I think it's pretty hard, but you, you don't do it because you do it despite that. I think, I think that's how it works.

    8. HS

      Aravind, listen, this has been so fantastic to see you. I so appreciate you taking the time while you're in London, so thank you so much for joining me.

    9. AS

      Thank you, Harry. Appreciate it.

Episode duration: 1:35:16

Install uListen for AI-powered chat & search across the full episode — Get Full Transcript

Transcript of episode OxFyVcO1Yow

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

iOSAndroidClaudeChrome