创意为何会比代码更重要｜Kevin Rose 与 Anish Acharya

So, how did the Like button come into existence? What was it like back during that time?

I think in the early days, when we first had something called AJAX—to get a little bit geeky—it was the first time, even though this sounds super archaic, that you could actually click a button, send out for a server response, and get something back into your browser. There was no way to just say, “I think this is cool. Let me just tap on it and show my vote of interest on something.”

I was like, well, wouldn’t it be cool if you clicked something and you saw the number go up, and that number was the number of humans who had actually clicked on something? The way I see it, this is a social signal that will feed back into an algorithm that eventually gives you more stuff that you would like to consume.

All right, man. So, ketones. Am I going to go on a trip?

This is going to be a surprise for you. These are ketones that Tim Ferriss turned me on to.

Okay, we don’t have to make it a plug in the podcast or anything. This isn’t a show about ketones.

But I will tell you that when you hit this, these are specifically made for—they’re supposed to go in your coffee.

You have to be careful here because you have to bend this like this, and then it kind of shoots into your mouth, and then you just chug it. Have you ever had ketones before?

No.

Okay. All right. Let’s do this. Let’s do this. Let me show you how to do it—I’ll do it first. So, you bend like this.

Mm-hmm. Mm-hmm. Mm-hmm. What am I doing here? Okay, there you go. Like this?

Let’s do it, Anish. Yeah. Wait, wait, no. No, wait. That’s backwards. Let me see. Yeah, there you go. I had a friend do that, and he shot it all over his chest. You don’t want that before the podcast.

Okay, take it back. Take it like a champion. Squeeze. Suck. You’re going to notice a little bit of bitterness. Do you taste it?

Mm-hmm.

Yeah, it’s not good.

Oh my God.

Yeah, it’s the worst. Now, this is going to give you brainpower for the podcast. So, this is how we start the show.

Jesus, dude. It tastes like I regurgitated a Tylenol. Fucking terrible.

It does taste like a chewed-up Tylenol. That’s a great way to put it.

Okay.

All right. So, we’ll let that kick in. Anish, dude, I’m so glad we’re doing this, brother.

Excited, man. Thank you. Thank you.

Before we get into the ketones kicking in, let’s get a little bit of background because it’s important. I know we’re going to be syndicating this across a couple of podcasts. I was telling people who we are. We’ve known each other for a long time. We worked together at Google a long time. We worked together at Google Ventures and Google corporate, and yeah, man, we’ve been good homies for many, many, many years.

No one’s ever seen us on a pod together, I don’t think, right? We’ve never done anything. So, do you want to give a little of your background?

Yeah, yeah, yeah. Of course. Thank you for having me, man. This is awesome. I’m Anish. I’m a GP at Andreessen Horowitz. I’ve been friends with Kevin forever. I’m a product person. I’m an engineer. I’m a technologist. I started 2 companies. I’m probably having the most fun I’ve ever had, both personally and professionally and all of that. I’m excited to talk to everyone about it today.

It was really cool when we first met. We both were—I don’t know how we ended up there and got on Google+.

But we were both on Google+, and I remember working with you in the very early days. When you land at Google for the first time, you don’t know what to expect. You’re getting in there and you don’t know: What is this whole thing going to be about? Who am I going to work with?

Very quickly, you realize there’s fantastic engineering talent there. Then, product-wise, I was like, “I don’t know that these people actually live and breathe product.” It was very much—you’re kind of in the Google bubble—but you were one of the first people I was like, “Oh my God, this is an amazing product mind,” and I got to work with this guy.

We figured out a way to stay connected and work together when we bounced over to Google Ventures, and yeah.

Well, dude, “figured out a way” is generous. You saved my life. I remember I worked with you for 3 or 4 weeks, and what’s cool, actually, is you joined and you were like, “No, this is not it. I’m not doing it.” I’m like, “What do you mean? What about vesting? What about Google+? What about this and that?”

And you were just so sure, and you were right. You moved over to Ventures.

I remember, at the time, talking to you outside of that building, the Google+ building, and saying, “Kevin, don’t leave me behind, man. Please bring me with you, and I’ll make you look good.” That’s a thing you say and you assume nobody will ever call you back after that.

Then, 3 months later, you called me. You were like, “All right, man. What’s up?” Like, “Why don’t you come over here and do this with us?” I mean, that single act of professional generosity changed my whole life.

That’s awesome, dude.

I’m very grateful.

Well, big shout-out to Chris Hutchins, who got us connected to Google—the mastermind who got us connected to Google Ventures—and then, of course, Bill Maris, who was running Google Ventures. I pulled you and Burka [?] in, and we all got over there and had a great time at GV.

Fantastic crew. I love those people. I thought Google Ventures was a great place. They’re the best of us, you know. Bill had the best energy. I met my wife there. I mean, everybody there was very good people.

Yeah. And then David Krane’s running it now, which is awesome to see. He was such a great person to be around. I really enjoyed my time with David, and it’s great to see him just turning that firm into an awesome—I look at the portfolio now since we’ve left, and he’s built out a great suite of companies that are just top-tier.

It’s legit, man. It’s legit.

Well, it’s actually fun to see you in action as a consumer investor. I think so much consumer investing and building is willingness to be embarrassed.

Yeah.

You did deals like Blue Bottle.

Yeah.

And I’m like, “Kevin, what is this? A coffee shop with a stand in Hayes? How is this going to be a venture investment?” And it was. You were willing to be embarrassed by it, and you were so right.

Well, you’re wrong a lot, as you know. We’re wrong most of the time. Consumer stuff for you—what does your day-to-day look like at a16z? It’s a massive firm. How many total employees are at a16z?

God, 600, probably.

Yeah. Holy crap.

It’s big. It’s big.

And how many GPs?

That’s a good question. Maybe 30.

Okay.

The investing team’s relatively small. It’s probably 70. But it’s actually nice, the way the firm works, because it’s a collection of specialists. Everybody who’s there is the best in the world at what they do, both knowledge-wise and network-wise.

Even though it’s a large group, that’s because we’re all specialists pointing in different directions. So, my consumer world every day is spending time with consumer founders. I spend some time with enterprise founders as well. Everything AI apps. It’s a mix of seeing companies, supporting founders on boards, and also playing with products.

To me, this is the craziest thing. It’s something that I learned from you, which is so many people just don’t use the products. How can you have intuition if you’re not making videos on Sora or vibe coding? There’s just so much alpha hiding in plain sight. All you have to do is use the products.

Yeah. I feel like the last, call it, 5 years on the consumer side have been relatively boring for me as an investor because I’ve looked at stuff and I said, “Gosh, you know, we’ve got the bigs in the room. You’ve got the TikToks. You’ve got Instagram still dominating, and then you see some of the folks like Threads that are just copying and somehow strapping on integrations with decentralized social, and that’s supposed to be a thing.”

Yes. But I hadn’t seen anything that, in my mind, was that interesting. It felt like consumer had this kind of down period of several years where there wasn’t really anything new.

Now that AI has come into the mix, I feel like there’s a chance to almost reinvent every piece of that framework. Do you feel that way as well? Is AI a shot in the arm for consumer?

100%. I mean, I’ll give you the investor perspective and then the builder perspective. I want to hear yours. From an investor perspective, this is like a renaissance for consumer investing. We haven’t seen an opportunity like this, I think, since 2010, 2011, 2012, because consumers are downloading products organically. That hasn’t happened in a long time.

Consumers are willing to pay. You look at the price points of the top products, right? The top ChatGPT SKU is $200 a month. The top Gemini SKU, Google Ultra, is $250, and then Grok is $300 a month. Consumers are actually paying that amount. I mean, how much do you think you’ve paid to Cursor?

A lot.

A lot. Yeah. Some of that is professional behavior, but a lot of that is just consumer and hobbyist behavior. So, I think that consumer is back in a big way. The tech-enthusiast consumer is excited in a way that they haven’t been for 10 years.

And when you look at the companies you just mentioned, though, they're largely the bigs. One of the things that's happened in consumer that I never thought would be the case, and that shocked me, is that the bigs—top-tier Fortune 100 companies—can put out consumer tech for the first time and actually see it get scale. It was always the bigs that couldn't figure it out; they had to acquire Instagram to be cool. They had to buy the cool.

And I think with ChatGPT—and actually with Google as well, with their Nano Banana model and whatnot—for the first time, we're seeing the bigs put out consumer apps that are getting some traction. Granted, they do have the scale and the install base; they can put that in front of a lot of people. But like we saw with Google+, just because you have the install base and the eyeballs doesn't mean you're going to see success on the consumer side. That's right. And I feel like we're starting to see that with the bigs, which was kind of shocking to me. Did that shock you as well?

It did, in a sense. What the bigs have released are models, not products. And I think there's an important distinction here.

NotebookLM, I would say, was the first time I was like, “Wow, that was a zero-to-one kind of product that I actually still use frequently.” But it's not consumer-consumer. That's a little bit more—

Business slant.

You're absolutely right, though. I think that NotebookLM was inadvertent, right? It was one of many experiments, and it worked. And, you know, not to take any credit away from Google, but I don't know that they said, “Hey, let's design an incredible consumer product,” and then invented NotebookLM, right?

Right?

So I do think that if you look at the models, the models are incredibly capable, and the bigs and the labs obviously are advantaged in certain ways, with capital, et cetera, to build great cutting-edge models. Consumers love those models, so that's great. They're playing in that space. It's definitely relevant.

But when it comes to opinionated products on top of the models, like Cursor is a good example. Krea is a good example. Companionship is actually a very interesting category where there's huge demand for companionship products. And a lot of these companionship products, like Janitor AI, deal with parts of the human experience that big tech is not going to be comfortable shipping, right? It's disagreement, it's sexuality, it's persuasion. So much of this technology can address parts of the human experience that 1,000 committees at Google and Facebook don't want addressed, and I think that is one of the interesting opportunities in—

100%. I think that's well said, and I think it is the space where we need to invest, because if you think about the pitches that we see on a daily or weekly basis, a lot of the stuff that I see is sanding down the rough edges. It can also just be something where I wholeheartedly believe that one of the bigs is going to dominate that category, and so I avoid those investments. They will never touch the emotional side as much as they should or could—

Not should, but that they wouldn't feel comfortable doing so, right? And so those are the categories and the verticals where I think they're defensible by the pure nature of what they're addressing, right? And so that's kind of what you're saying here, right?

100%. I just think they're structurally set up to kind of take the soul out of products, and when you talk about categories like companionship, the whole thing is the soul; there is no product without that.

I think the other interesting area where the labs and the bigs are not going to be successful is products that benefit from being multi-model, right? Not multimodal, but multi-model. So Cursor is good because you can use every model inside of it—

And Google is never going to ship with Anthropic models embedded.

So that also is very interesting. They're sort of constrained to the models that they've developed in-house, and products that benefit from being multi-model are actually really unconstrained.

Very interesting. We should dive into each of these things. I think we have 2 great topics to go on. Let's pin the multi-model, but let's talk about companionship. On the companionship side, what do you believe should be built? What have you seen built? And how much of this do you think is fad versus depth? Meaning, it's funny and weird and awkward and strange to have an AI girlfriend.

And in the early days—well, early days, let's call it a year ago—

Way back in the day.

Way back in the day. Yep.

A friend of mine was like, “Hey, go check this out.” It was not something that was X'd about or tweeted about. It was this kind of thing that you went into. And I'll say that I went there and it was models unhinged, right? This is before Grok did it, and you could basically talk to a model and get any type of response that you wanted, meaning in terms of sexuality or whatever you were looking for.

How much of that are you concerned about in terms of shaping humanity, replacing real relationships? What lens—how do you evaluate something like that? Do you invest in those types of things?

Yeah.

This is such a great topic. I wish we had Eugenia here, who founded Replika, because she'd have a ton of thoughts. I'm always going to give you the optimistic take; that's just how I'm wired. So let me give you a couple of my most optimistic takes.

One is, we all know the importance of human connection and the value of it. Maybe the human part is overvalued. You live in an embarrassment of social riches. You have friends, you have professional connections, you have interesting people you can meet. They all want to connect with you, and your biggest challenge is prioritizing. I don't know that that's the experience of the average person in this country. I think that there's a deep loneliness, and any progress we make toward addressing loneliness is human progress and is very prosocial. And I think progress, though, is having a conversation with a chatbot. Do you think that starts to fill that emotional bucket up?

Yes.

So you believe that when someone is having a conversation, they are getting a lift?

Yes.

Probably 10% of what a real human would provide you, but a little bit of something.

I think more than 10%. I just think—look, it's like our lizard, reptilian brains. We are wired to have these sorts of emotional, chemical responses when you're engaged in a human-like conversation. We intellectually understand that these are computers on the other side, but you still feel the feelings, and I think that's what's necessary to make personal progress as a result of connection. Do you believe?

So this is my concern. I'll play the doomer side of this. My concern is that one of the things I noticed with these models early on, while I'm playing with them, is that they were very agreeable. And—

And emotional well-being—the muscle that needs to be built—is actually not when you're laughing and agreeing with friends and kind of going along with the flow. It's when there's disagreement and the discomfort that comes from, “Oh, I don't agree with you. That doesn't hit me. That doesn't feel right. Let's build a bridge and figure out how to find common ground here.” If you're dealing with models that are just paid to agree with you all the time,

Then you're basically—so I'll give an example. I was talking to Tim Ferriss about this. I was taking screenshots of what this model was saying to me, and it was basically agreeing to anything that I put in front of it. I was like, “Hey, what do you think about this?” And it was like, “I'll do that with you.” And I was like, “Oh, my God.” I couldn't believe it, right? I was sending it over to him, and he's like, “We are doomed.”

And his response was that you're not getting any pushback. And so then, okay, when you make the jump to a real relationship—okay, I want to have a relationship with a human—you’re used to an agreeable, kind of subservient thing, and the proxy and the jump don't make sense. And all of a sudden you're like, “Well, that's not as good.”

I want to go back to my model. Are we training people to want and favor agreeable models over the real character-building that comes with the emotional side of things?

I don't know. I think we just need to dial it in. Everybody knows a human relationship where the other person always agrees with you is not authentic, and it doesn't nourish you. You need to have the tension, the disagreement, the exploration. I mean, we've got to be fair to where we are in the cycle. Yeah.

ChatGPT was November '22, right?

Right. It's not even November '25. And we're trying to dial in the entirety of human connection and experience and find the right place between not being a sycophant but not being disagreeable for the purpose of it. We're dialing it in. We're going to get there.

And again, because we're so wired from an evolutionary perspective to know what the right amount of tension is, I believe our intuition will guide us. We'll get there. I like what you're saying because we've talked about this on the car ride over here. If you just, for a second, imagine we are in the huge brick-cell-phone era of AI, it's like this is the very—not even the first inning. We're just stepping onto the field of this.

I mean, imagine 20 or 30 years from now. We're going to have a much deeper understanding of what's going on behind the scenes here, what is required to make it work, and we're expecting that on day 1. The expectations are so high. Let's give it some time, you know.

The trough of disillusionment is real, you know, and our expectations are high.

Yeah.

I also think that I’m glad we’re talking about companionship, but broadly, we’ve had 40 years of technology that extended our intellect and our minds.

Yeah.

Right. But most of the human experience is actually emotional and subjective.

Now we have a technology that extends our emotions and our subjective experience, and it can address that. I love AI spreadsheets. I’m glad that somebody is building them, but to me, that’s the least ambitious execution of the primitive. Let’s start to really address the human experience through this new technology. I think that’s where all this goes.

Yeah. Well, it’s early days. We’ll see what comes. I’m certainly looking at those startups. I think a therapist or counselor—somebody on the back end who can give you this partner to bounce ideas off of and kind of push you toward best practices—is really interesting to me. I don’t know if you’ve looked at any of those types of startups.

Yes, I’ve looked at a few. I do think that’s interesting. By the way, the ketones are kicking in.

Yeah. You feel that? Isn’t it amazing?

Yes. Is it just caffeine?

No, it’s not caffeine. It’s a full-on—I'll send you the thing about ketones, but they’re a natural fuel source that is not caffeinated.

I’m definitely feeling something.

Yeah. Well, okay. I think therapy, companionship, all that stuff makes sense. I actually think there’s a really interesting emerging category. You see this with Poke—that’s a great example of it. Huxe does some of this, as well as Sierra’s new company. We both spent time with him. I call it indirect companionship.

So what is Poke? Poke is kind of this subjective human front end to your email.

An email is like a set of tasks. It’s pretty heavy, and it’s functional work that we have to do.

And Poke creates this human overlay on top of it. In a sense, I don’t consider Poke a companionship product, but it’s letting me interact with a functional part of my life through an emotional interface. I think that’s actually a really interesting area for exploration.

What does it mean to return to our roots and still do functional work, except through an emotional filter?

Yeah. Can you talk about Poke a little bit? What’s the onboarding like, and why are people talking about it?

Yeah. It’s so wild. I encourage everybody to try it out. You download Poke. One interesting thing they’ve done is that the interface is actually iMessage, so you’re texting with it. You’re typically only texting with people on iMessage, which subliminally sets you up to expect the Poke product.

First, it refuses to allow you to sign up for the product, so you’ve got to convince it that you deserve to actually use the product. Then you engage in a negotiation on price, where it starts at $200 a month.

It reads all of your email, so of course it can say, “Hey, Kevin, I saw you bought XYZ. You’re spending a lot on your ketones. I think you can spend $200 a month on this product.”

So by the time you get in, even though it hasn’t done anything for you except torture you—

You’ve already got this perception of value—

Right?

And it’s a very interesting experiment in onboarding. If you extrapolate it, I think it’s an experiment in product design in an era where we have emotional primitives.

Yeah. It’s fascinating. I love product builders who consider every step of the process, from sign-up all the way through onboarding, and rethink the entire stack.

This is an example of something that I don’t know is the long-term best strategy for a product like that—to negotiate SaaS software in real time. But that little feature alone is a great example of something that went viral within a small subset of people, and I’m sure it led to tens of thousands of sign-ups.

Just from people talking about that feature—

And we’re talking about it now.

Right. Exactly. That’s right.

But for the product builders out there, I think that’s such a great little way to wedge yourself into the conversation by rethinking part of that step.

Yes.

I’ve got a related question for you. You’ve both designed a lot of new products, and you’ve spotted a lot of new product thinking. How do you do it? What are the patterns? How did you see it all of these times?

I think that on the product side, what I’ve always been drawn to is, for better or worse, the novelty of original thought in certain areas.

If it’s something—we’ve talked about this kind of sanding down of the rough edges—that, to me, is not that interesting.

But I really respect people who are building something like this price negotiation, where I just haven’t seen it done before.

Right.

It doesn’t mean that it’s going to be a top-10 app or that it’s going to be a multibillion-dollar company, but if you find that signal in a builder’s brain, it means that they’re going to apply that signal to other parts of the product that are interesting. It means that they will continue to reinvent and reimagine things through a lens that nobody else has.

Interesting.

That’s enough signal for me to say, “Okay, we will probably turn over a card at some point that is the magic card—the one that’s going to get us that home run or that massive product.” I value that almost above all else, even if the idea isn’t the one that I think is going to hit scale.

Yeah. If you have a founder who builds through that framework—“I’m not going to do this just because everyone else has, and I’m not just going to take the path of least resistance and build a better Slack or whatever it may be”—to me, that is very rare to find. When you find it, that’s the type of person I want to back, especially at the seed stage, when you don’t know what’s going to work.

That’s super interesting because one of our observations is that it’s just really hard to predict consumer seed. We haven’t done a ton of it for that reason. In fact, our stated approach is “weird and working.” That’s how we’ve tried to invest in consumer. Yet you have been consistently right at consumer seed investing. It’s interesting that that’s what you look for.

Yeah. I think that’s exactly right: weird and working. I look for weird first.

Yeah.

Then, if it starts to have a little bit of smoke, that’s a very interesting signal. Oftentimes, I’ve seen founders do this a handful of times: weird and working ends up failing, but they still have the weird because that’s in their DNA. So they’ll build the next weird and working, and that’s what you want, because they’re going to need several shots on goal. You can’t manufacture the weird. The weird is internal. The weird is, “I just see the world in a different way.” That is the founder who is going to win.

Why is weird important?

Because in a world where you need attention—we’ve talked about this socially in terms of what draws you in and what draws people into consumer products—consumer is a finicky category. After a certain amount of time, what was weird and working ends up being boring and stale. It’s the reason why Digg is no longer here. It’s the reason why there are a thousand startups that have MySpace or Snap, or whatever the next one is, that will eventually become old and stale.

And so, if you want to find the next thing in consumer, often you play with something, and the aha moment is, “Oh, I hadn’t thought of that,” or, “I didn’t feel this. I’ve never felt this way before.” I’ll never forget when I made the investment in Twitter way back in the day. The thesis for me was very simple to see because we’d had bidirectional relationships for the longest time. The MySpaces and all the social platforms that had come before it had this idea of friendship.

It was, okay, in order for me to get invited into your world, I would friend you. You would get a request and be like, “Oh, Kevin friended me.” Then you would have the permission. You would grant me permission to see what you were up to.

Right?

And the beautiful thing about what Jack, Ev, and Biz did was say, “Okay, we’re going to take that and just open it up.” They said, “Actually, you can just broadcast out and say, ‘I have something important to tell the world,’ and everyone can follow it, whether they know you or not.”

And it was that one simple little product tweak that said, “Okay, it’s no longer about bidirectional communication. It’s about this idea of broadcast to all.”

And when you see something like that, it unlocks not only a massive product opportunity, but a market opportunity, because you’ll never become friends with the celebrity, but you can sign up for their broadcast.

And that was the weird thing.

It seemed odd at the time.

Was it weird? I was going to ask.

Yeah, it was weird. I remember signing up for Twitter for the first time and just being like, “Okay, well, where—what do we do here?” They had this thing called following, and I was like, “What is the following thing?” It sounds so obvious now, but it was weird at the time.

I think that is going to be so important for consumer, because what seems weird at the time eventually becomes mainstream, and then everyone just assumes that that was always the case, right?

Yes. And that's what I'm looking for: that weird thing that will eventually become so mainstream that we'll assume it was always a primitive that existed forever, when in reality it was new and novel at that time.

I love that. Does that make sense at all?

It makes perfect sense. It's very interesting.

You know, it's also that founders have to push themselves to the edge of being embarrassed, potentially.

Yes. To design genuine new things. Otherwise, it's very derivative. It just has to be. And that is what all the X-for-Y companies typically are.

Right. Exactly. Like when you and I were at GV and we did the Uber investment, how many startups would we say, “I'm the Uber for this”? And you're like, “Oh, God, another one,” right?

And then, yes, there are the DoorDashes and some of the others that are built on the same kind of similar primitives, but it's nothing like it was. It was so weird to get into the back of somebody's car. Do you know how odd that was? I remember hailing an Uber and being like, “Okay, someone is pulling up at my house, and I'm going to get in the backseat of someone I do not know.”

Yes.

Every day when we took the ride over here, we didn't even bat an eye. We didn't even look at the guy and say, “Are we safe here?” It was just a given, right? But it was so awkward at the time.

Well, so this is what I love about that era. We talk about that era as a part of history and we say, “Oh, it was early mobile, and it was remote control for the world, and software was eating the world.” We have all these things, and they're all true and they're all valid. The most interesting part of both Airbnb and Uber was that there were these dramatic changes in human behavior.

We had all been raised being told 2 things: never get into a stranger's car.

Never sleep on a stranger's couch.

Yes. Yes. Exactly. And then all of a sudden you started doing it, and guess what? The strangers were kind of pleasant.

Yeah. And it was just this incredible unlock for the world and this sort of positive view of human trust and strangers.

That was a big change in consumer behavior.

Yeah. And that's so exciting, because I think we're going to see more and more of that.

One of the things I've done—I'll just be a little bit vulnerable and transparent, because that's what's fun about podcasting—is that one of the things I get in a lot of trouble for is that, if I get in a disagreement with my wife, I will take some of that shit to ChatGPT and be like, “Am I right here? What's going on?” I'll say, “Use the framework of Terry Real.” I use some of these professional therapists and say, “Help me analyze this.”

The mistake I make is that I paste that back into the conversation. But the point is, that's weird and awkward. She didn't like it. She's like, “Hey, this isn't cool. You're going to this—” She calls it “my bot.” “You're chatting with your bot about our relationship,” blah, blah, blah.

But I've talked to enough people now that I know this is a trend. People are looking to AI for validation: “Help me through this level of thinking here. Help me figure out: am I in the wrong here? What am I doing?”

I see a world where the bot is in the room with both of us as we're having a conversation, or as we're on video, or as we're debating something. We're like, “Oh, this is heated. Let's turn it on. Let's let it watch us,” and then give us feedback on how we're doing as a third party—hopefully an unbiased person—sitting there and evaluating this in real time. Is that weird and awkward right now? Absolutely.

Yes.

To your point about how these things evolve over time, 10 years from now, will that tech be good enough that we'll consider it not just a healthier thing for our relationship? I think so.

100%. I think that's coming. Ideally, the Alexa in your kitchen is passively observing your family and then perhaps is just letting you know, “Hey, your daughter was trying to talk to you about something. You weren't really paying attention.”

There's all of this stuff that can just be in the oxygen around us, and today that seems unusual. What about a kid's classroom? My son goes to an amazing school, and I'm very grateful for it, where they have 2 teachers. One teacher is focused on academics, one on SEL, social-emotional learning. What an amazing thing for every school in America to have.

And the way we're going to get there is not by paying for 2 teachers in every classroom. It's probably some kind of vision model that's privacy-first that can observe the social interactions of children and let parents and teachers know what's going on.

This idea of always-on recording—you and I have had a debate about whether we want to get to the multimodal thing real quick, though, before we go in there. Let's—

We should. I also want to get, since we were talking about internet history, your take—your story that you've told me, I think privately, about the Digg button, the like button, and the story there.

Yeah, we can talk. Where do you want to begin? You want me to talk about that?

We're in a history. Tell me: the Digg button was a very new and interesting idea in social news. I think we're sort of in a safe place 20 years later. Talk about how it influenced the like button.

So, how did the like button come into existence? What was it like back during that time? I think in the early days, when we first had something called Asynchronous JavaScript and XML, or Ajax—to get a little bit geeky—it was the first time, even though this sounds super archaic, that you could actually click a button, send out a request for a server response, and get something back into your browser. It used to be a world where we'd have to refresh the page to see anything.

And so this idea that content could talk to a server and come back to you and display something new was just completely exciting and novel at the time.

Mm-hmm.

I had seen a bunch of sites that were allowing for user submissions, like Slashdot, but they wouldn't show you what content had been submitted. So it was user-generated content in some sense, but there was still a gatekeeper to that.

There were social bookmarking sites like Delicious and others that would count the number of people who had bookmarked things, but there was no way to just say, “I think this is cool. Let me tap on it and show my vote of interest in something.” And so, when Digg launched in late 2004, it was the first time I had ever seen—because Ajax had just been out for a couple of months—that anyone had voted on content. Obviously, I didn't invent voting, but no one had done it on the web.

I thought, “Wouldn't it be cool if you clicked something and you saw the number go up, and that number was the number of humans who had actually clicked on something?” Wisely, just by luck, I happened to be in contact with enough people that they were like, “Hey, you should go out and file a patent on this.”

And so I went and filed a patent on it—not because I wanted to enforce our own liking, but because I was told and advised to do that as a defensive practice, so that no one could go out and patent it—one of the big guys could go out and patent it—and say, “No, you can no longer do this.” Even though I had first-use rights and would have been able to do it, I went and did it, which was great. It started off, and it became really big. At that time, we were larger than Facebook traffic-wise.

Mark and I shared a common investor in Greylock. We had a lot of overlapping friends, and he was like—actually, it was David Sze from Greylock who was like, “Hey, you should meet Mark. You guys should hang.” So we went out and grabbed dinner, and we hung out a couple of times. We had dinner a couple of times, and he came to my office.

He drove up and came to the Digg office, and I'll never forget it because he came into the office and he was so young at the time, and so was I. No gray hair back then. He was sitting on the ground and just wouldn't sit in a chair. He was sitting on the ground.

We were just talking about what it meant—what liking was all about, what digging was all about. I said, “Well, the way I see it is that this is a social signal that will feed back into an algorithm that eventually gives you more stuff that you would like to consume.” And that was the patent that I had gone and filed.

Mark didn't have this at the time. There was, I think, a way to poke someone on their profile, but that wasn't really a like; it was just to say what's up. And then, a few months later, they had rolled out the like button, and that was kind of a big part of what they pushed as a similar thing—a social signal to feed back into the algorithm to give you better content.

They obviously hit a scale. We were serving billions of Digg buttons per month because we had them integrated into websites—you could actually put a widget into your site back in the day, and there were Digg buttons all over the internet.

It was cool. I was never offended by that. I was like, “Oh, that’s awesome.” He made it his own, called it “Like,” and it made sense to apply that to what he was doing.

I know there were other people at Facebook who worked on that. It was interesting because I saw one of the engineers who worked on it in an interview, and he credited Digg as a place where they had seen that first, which I thought was really nice.

It’s strange because even though Digg went out of business and failed for that version of the site, when all was said and done, the asset of Digg sold for pennies on the dollar. It wasn’t really worth a whole lot.

But the patents ended up being worth millions of dollars, and LinkedIn bought them from us.

Really?

So LinkedIn bought the patents. The thing at the end of the day was the defensibility and patents around the actual social Digg button.

Wow.

Which LinkedIn ended up buying. But yeah, it was a really wild time to sit there and come up with that idea.

Not that voting was novel, but I just hadn’t seen it applied. And then we applied that to comments as well. I’ll never forget that. It was different from the Digg button, but equal in importance: I had created the comment system with Daniel Burka. We were sitting there designing it, and we were like, “Okay, well, we should apply digging and burying to comments because that makes sense.” People should be able to vote comments up and down. That had never been done before.

Mhm.

And we were like, “Okay, this makes sense.” Daniel and I were designing it, and I was like, “What if, after a certain number of buries, the comment kind of shrinks down?” He was like, “Oh, that’s cool.”

We riffed on that, built the design, deployed it, and rolled it out. We had it hard-coded so that after 5 downvotes, it would shrink into a single line. You’d just see who submitted it, but we color-coded it so it looked disabled or out of the way.

We rolled it out, and 3 hours later, I noticed that one of the comments had negative 200-some downvotes. I thought, “The product’s broken. Something is wrong. We have a bug in the code because it’s meant to shrink after 5. Why would it be negative several hundred?”

We looked and couldn’t find a bug. We couldn’t find a bug. Then we realized what had happened. We talked to some users, and they said they wanted to see the train wreck, so they exposed the comment. They were like, “Oh yeah, that guy’s an asshole.” Then they downvoted it again.

At that time, people were like, “Why is that person getting so many downvotes? They must have said something crazy. Let’s take a look.” Then they’d see that it was horrible and downvote it. It immediately showed me that people are attracted to the carnage. They want to see the destruction and the bad side of the internet.

I was like, “Oh, God.” As a product builder at that time, I would have thought people would say, “Negative 10 votes? I’m not going to look at that.” I had these rosy glasses on, thinking people would see it as something negative and move on. But they wanted to see the carnage.

There are so many interesting things there. First, the story of your connection to the Like button is fascinating. That’s a piece of internet history, and sort of a piece of human history, right?

What are the implications of the Like button, as inspired by the Digg button, for all of society? We’ll never know, but I think it’s certainly something. I also love the ethos of you not feeling offended by them being inspired by it, right?

Yeah. Listen, I have a lot of respect for Mark and what he’s created. I don’t agree with his direction around glasses and AR/VR. I never have. I’ve been anti-AR/VR for years. I’ve never made an investment in the space. I’ve always avoided it altogether because I think we’re not considering the social side of things enough—the impact it has on those around you.

That’s another topic that we could get into. But I think he’s obviously a brilliant man. It was really fun to have those moments where I’ll be able to tell my kids I was in those dinners. I was having these deep product-sparring sessions around some of the product thinking behind what he was building.

It’s very special to even have a seat at that table. I feel very lucky, as someone who came from nothing, to be able to sit with some of the great product thinkers of our time. I would put a bunch of people in that camp, like the Systroms and Reid Hoffman, and all the people that I’ve been able to brainstorm with over the years.

To sit and hear great minds dissecting all the things that we build and the social impact they have—it’s fascinating stuff. It’s so fun.

You must feel this way with Marc and Ben when you get to sit with them.

Marc and Ben were involved in the early Digg and helped me through that stuff. Gosh, to sit down with Ben and listen to his steady hand and sage wisdom—

Yes.

—you just learn so much from that dude. It’s insane.

For sure. There’s a sense of seriousness and purposefulness, but also some playfulness. It’s amazing because if you think of the history of when you were involved in tech, we were just weirdos in Silicon Valley back then. It was like the dog chasing the car, and now we’ve caught the car. We are the car.

Technology is something that’s really important for humanity and society, and I don’t know that anyone who was building it at the time appreciated that.

No, and none of us had any idea. It’s funny when you hear these stories about the bad sides of social media and how these platforms have given the microphone to certain voices, and how much hate there is.

I agree with that wholeheartedly. There are a lot of things that we need to fix, or at least try to fix. But I can tell you, as someone who was sitting around watching all these builders—whether it was the Twitters or the Facebooks, or you name it—no one knew this was going to be the case.

Nothing about this was intentional. It was like, “What does this unlock?” It turns out it unlocks both the good and the bad.

And I’ll tell you the good. When I was growing up, maybe you had the same experience, but there was 1 way to be cool. I was born in 1979 and grew up in the ’80s in a small town in Canada. You either played football or you didn’t, and if you didn’t, everybody knew what the social hierarchy was.

When I was growing up, they said everybody’s cool in college. Why? Because you’d find your sense of community of weirdos.

Yeah.

But the beauty of the internet now is that you find those weirdos online at any age. I think there’s an acceptance of people’s own individuality that didn’t exist 20 or 30 years ago.

Sure, there are some edges on the internet, but I think the benefit to humanity of people finding their like-minded people early is really significant. The power of microcommunities is only going to become more important over time.

These very small, in some sense hopefully walled gardens, are places where you can go and have private discussions. If your weird isn’t so weird, you just need to find the people who are into it. If you’re into Japanese woodworking and you can find 5,000 other people who are also into that, that’s a very powerful network, right?

Yes.

That’s the exciting side of all this. You’re a product builder. You’ve been a builder for a long time, and you’ve launched and sold companies. What are you doing these days? What’s your outlook for how products are being built, and how has that evolved? How does that play into your investments as well?

In some sense, we’re entering a really weird era—not weird, but new—where someone, even a designer who has no experience on the engineering side, can prototype, build, and launch. It’s deputizing a wider base of people to actually roll out products.

I think it’s incredible. I’m so fired up. I haven’t felt this way since I was on my computer working on my BBS, with my mom yelling at me, “What are you doing down there? Come up for dinner.”

That’s the level of enthusiasm I have right now because it feels like everybody is unlocked to do what they want to do. I think there’s a set of people—or a part of us—that will be creative, and a part of us that will want to be productive and emotional, and all of it is getting meshed together.

It’s the best time that I’ve ever experienced in my career, hands down.

Does it worry you that entrepreneurs can go and build a startup, launch it, get product-market fit, and not have to raise venture capital?

No, it doesn’t worry me. I think there are businesses that will be built by 1 person and reach $100 billion in revenue.

That’s awesome. Will they ever need venture capital?

Maybe not. And that’s okay. We have 1% of the software that we need in the world that we’re going to have. We’re just getting started, and of course there are going to be businesses that don’t need venture, and that is fantastic.

For example, in the ’90s, there was this big movement of digital small-business owners, right? You could sell shareware online. You could do all of these things to have these sort of corner stores on the internet that actually worked. What happened in the 2010s was that everything moved to centralized networks, and all the economic value was extracted by the networks. You couldn’t really do that—maybe you could be a creator. That was probably the most ambitious form of digital small business that you could have in that era.

And now, because consumers are paying for software, and because anybody can make software, I think we’re going to see this renaissance of people creating million-dollar run-rate businesses as individuals. And that’s awesome.

Do you think there’ll be fatigue around subscriptions? One of the things that I feel personally, maybe more so on the Substack level, is that there’s $5 everywhere, right? And if we see micro-subscriptions across the board, you’re going to be like, “Okay, wait a second. I wake up to several hundred dollars per month based on all these tiny micro-subscriptions.” Is that a concern?

It’s a good question. We’ve talked about micropayments for 30 years. It feels like it’s never really worked. But maybe it’ll work this time. So that feels like part of it, because I think there are new protocols where payments can be directly embedded in API calls. I’m not an expert in crypto or Web3, but I understand that those protocols exist.

I think there are potentially new models for monetization. And look, I think the second thing is that software can do so much more for us than it could 5 years ago. It’s incredibly ambitious. So, sure, maybe I’ll have more subscriptions, but the parts of my life that are addressed by software are so much more significant. I think we’ll be happy to pay for them.

If you’re an entrepreneur—or anyone, actually, because anyone can do this these days—and you say, “Hey, I want to play. I want to go. I want to build,” what do you see? What are the tools that people are using out there? What are you personally using to build? What does your current stack look like?

Yeah, and I’d love to hear yours as well. So I’ll tell you my productivity stack, my coding stack, and my creative stack. Productivity is actually pretty simple. I love the Perplexity Comet browser. It’s really well done. It’s got this consumer RPA feature called Assistant built in. That’s actually very useful day-to-day. It’s really important to me, and I use it all the time.

Do you use it for search? Because that does replace your search with Perplexity.

I do use it for search when I don’t know exactly what I’m looking for. I find that for navigation queries, Google is still better. But I definitely do use it for search, and that’s an interesting topic: What are the implications of a new front door for the internet? For 20 years, that was Google Search; now, maybe it’s language models.

I use Notion Notes a lot to capture meeting notes. I think that’s very useful. I love Deep Research.

Same. I started using that recently, by the way—Notion Notes. I was using Granola for a long time, and I think it’s a great product.

Great product.

So I’m torn because Granola is freaking awesome. But Notion Notes are built right in, and it automatically launches. I’m like, well, then all my notes are in Notion, which is kind of nice.

If you’re in the Notion ecosystem, I think it makes a lot of sense. If you’re not, Granola is a fantastic product. But I’m a fan of both, and I just love the capability of not having to handwrite notes.

Right.

Deep Research is cool. I’m actually a big fan of the Grok DeepSearch feature because it can search X. It’s got a very up-to-date index of X.

Tell people how you use it, because I thought this was very interesting from a trends perspective.

If you have a query like, “Hey, what were the top AI memes of the last week?” and you go to ChatGPT or Anthropic, the models just aren’t trained past a certain date. They can use the web search tool, so they can do some of this, but Grok is hyper-up-to-date with what’s happening on X beyond just the web search tool. I find it very, very good for timely information. And look, let’s be honest: Most of the interesting things that are happening in our industry, and often in the world, are happening on X.

Mhm. Yeah, it’s more of that real-time nature of what’s unfolding. Do you have set queries that you go to to keep up with what’s breaking on X?

I’m always interested in what’s breaking culturally and what’s breaking technically. The memes are the best proxy for culture. That’s often what I look for. And technically, think of what’s happened just in the last 2 weeks, right?

Claude Sonnet 4.5, OpenAI Dev Day with ChatKit, AgentKit, and the Apps SDK. Gemini 3, I think, is coming next week. Sora 2 with the Sora 2 app. That’s 2 weeks.

I know. It is an embarrassment of riches. It’s awesome.

As an entrepreneur and as an investor, I find it insanely daunting to try to keep pace with the number of releases that are coming out. How does a16z handle this internally? On your consumer team, how many people are there?

The core investors are probably 6.

With associates and everything? That’s it? So, 6 people? What’s the process for surfacing something interesting to you? Where are they looking for signal? How are they keeping up? For example, let’s just say this Agent Builder—I played with it for the first time yesterday. That’s going to take an hour of your time if you’re really going to go deep on it. If something’s coming out every 2 or 3 days, or once a week, how do you work with your team to stay up to date with what’s going on, given the pace at which things are rolling out?

Yeah, we just work our asses off. We really do. We commit to trying every product.

But what does that look like? Do you have an associate who comes to you and says, “Hey, I played with this. Take a look. I’m going to grab 20 minutes of your time because I want to bring you up to speed on something”?

Well, it’s interesting. You know, the sort of associate-GP model—we’re pretty egalitarian, actually. People cover different areas. My partner Justine has a real specialization around creative tools. She’s gone to the edge of the internet on creative tools. My partner Brian, who looks at a lot of things in social—anytime there’s a weird, working new social product, he’s there. Olivia does that for us around productivity and areas like voice.

I’ve been spending a lot of time in AI coding. I think I’ve probably got 50 subscriptions for app building. I’ve just tried them all.

Wow.

So we all commit ourselves to subareas, and we try to cover them very thoroughly. But look, we just hold ourselves to a really high bar of trying everything, because our belief is that you cannot have an opinion until you’ve actually tried the product.

I think that’s so wise, because there’s just too much for any one person to try and wrap their heads around. Do you have briefings where you share learnings? How do you make sure that knowledge gets transferred broadly within the fund and at the GP level?

We definitely have a lot of active conversations about it, but the truth is, so much of our best thinking we publish. We really do. If you read the essays and follow us on X, we’re talking very transparently about what we’re thinking, what we’re learning, what we believe to be true, and what we’ve changed our minds on.

It’s interesting, because as a founder on the outside, I always assumed that VCs knew all these secret things that they didn’t tell us. Maybe that was even true 10 years ago. But now, I think we’ve moved, in a very good way, to a world where investors are talking very actively about what they’re seeing. I think that’s a great signal for founders.

Now, the signal is always a lagging indicator. So, for founders to develop a weird idea, definitionally, no one’s talking about it. But it’s great to know that there really aren’t that many secrets. It’s all being published out there.

On the coding-tools side, you said you’re seeing a lot and have so many active subscriptions. What is your preferred stack these days? If someone was getting involved in coding for the first time and wanted to start playing, what do you go to on the coding side?

Man, I’d love to hear your stack as well. Look, there are so many tools, and they’re all pointed in different directions. My view on AI code and software is this: This is not a market; it’s an industry. The whole software supply chain is changing. The cost of creating software is collapsing.

We have 1% of the software we need in the world. We’re going to build the other 99%.

Why do you say that, dude?

Because there are so many parts of the world where the ROI didn’t make sense for creating specialized software, right? In so many parts of our lives, by the way, some of those parts of our lives may not be serious.

It’s like my wife is working on a manifestation app for her and her 5 friends.

Right?

5 years ago, I mean, what are you going to do? Go raise venture capital? Is she going to hire an engineer? Now, in her free time, she’s been using Base44. But there are a lot of these other products that are sort of for mass-market consumers. Replit’s another great one, where you can just make a product because you want to. You could make a product that’s disposable. You could make a product for a fun bachelor party weekend. You can do things that aren’t justified by the 5-year-ago ROI, and now it’s just typing a prompt into a box.

It’s so well said because there are so many times when, if you think about what it would take to create an app that does something very niche to me and my friend group—

Yes.

Five years ago, if someone came to me and said, “Hey, Kevin, you get these ideas all the time. I have an app idea. I have an app idea,” I’d say, “All right.” And the best way to shoot them down back then was always, “Well, that’ll cost you about $75K to get off the ground.” They’d say, “Oh, well, it’s not that big a deal. I guess I’m good.”

But now that’s no longer the conversation. It’s like, “Oh, here, go play with Bolt—or, you name it, Lovable—and you can deploy this in a day.”

Those were tough even 6 months ago. It was pretty hard, actually. These services existed, and I was using Replit or Bolt or whatever, but I was like, “It’s just not quite there yet.” But now, even in the last few weeks, I would say that with Sonnet 4.5 and Gemini 3 coming, we’re getting to where it’s no longer what it used to be.

It used to be that I’d go into v0 and spend 20% of my time putting together the scaffolding. Then I would connect to the database, and a lot of that was bug squashing, which was another huge piece of it.

Yes.

And the vast majority of my time was still bug squashing. That window is collapsing. It used to be that the bulk of the work was just that final 10%, and now that is going away.

Tell me about your setup, Kevin, because we were together last weekend. I saw the way you were using v0. Talk a bit about it.

Yeah, I think with v0, this is a product that was largely built to quickly put together a beautiful design, a very simple design, and a framework for any new app. So you can go in there and discuss what you want to have built.

I liked it because I could start with a piece of paper. I could draw a prototype of how I think something should function, take a picture with my phone, attach that as an image, and say, “Can you build something like this that does X?”

It puts together components because, as it’s built in the Vercel ecosystem, it’s largely going to build components for Next.js, which is a fancy way of saying that’s the server that serves all that stuff up. Then you can take those components, download them, and drop them into Cursor.

Cursor, for me, as someone who leans more technical, is a better solution because I like to be in the nuts and bolts of how things are being built. I would drop those components in there and then basically say, “Okay, put together this and actually create a real app behind the scenes here.”

It starts building things and finds where the missing pieces are. I can connect up Supabase and have a Postgres database tied to it, and then I’m deploying everything out to GitHub, which automatically gets deployed to Vercel. That means I have an active app that’s up and running.

There are certainly easier ways to do this if you’re nontechnical. You can go the Lovable, Bolt, or Replit route. There are probably 5 or 10 others that are playing in this space.

The thing I’ve been doing more recently, though, is dealing with the dead ends that you still run into, where something isn’t working. It can be insanely frustrating if you’re not technical. You don’t know how to self-diagnose it, and you’re like, “Why isn’t my authentication working for X?” You just keep hitting the same model over and over again.

The nice thing about the way Cursor functions is that you can install something and have multiple models running. I’ll have one model on one side, which is the Cursor built-in chat, set to Sonnet 4.5. On the left-hand side, I can have Codex from ChatGPT. If I run into a problem, I take that problem over to the other side and have it work on it, too. If I put them against each other, you can typically get to a solution pretty quickly.

That’s great.

And so I’ll break through those walls and march toward something that’s actually ready to deploy to production.

I love that point. Talk a bit about how you do the design exploration in v0 of individual components.

Yeah. One of the things that I always like is a good amount of visual flair and touch on things. You saw something last night when you came over to my house and we were showing each other our vibe-coding progress.

One of the things that I’m exploring right now is chat around objects of content. They could be stories; it could be all kinds of things. I want to figure out how to select an object and bring it into chat in a very fluid way.

I think, at least in the consumer internet, one of my spidey senses is for new primitives for how things flow in and out of conversation. One of the things that you have to figure out is the emotional vibe: Does this feel right? Does this feel new and interesting? Does it unlock more productivity or a better way to have a conversation about something? You have to try a bunch of different things.

So what I’ll do is take one thing where I want to figure out the best way to collapse and quickly jump in and out of the story. I’ll take that over to v0, give it the screenshot, and say, “Come up with one way.” It’ll create something, and I’ll be like, “Okay, I kind of like this.” I’ll tweak it a little bit 10 other times, so I’ve gone through a bunch of iterations and have that one way.

Now the big unlock is to say, “That’s great. Give me 20 other ways to do it that are completely novel and unique and don’t have anything to do with the first way. Put them all on a single page.” It’ll churn for 3 or 4 minutes, and then you have 20 different ways.

Now I click down: “That was a weird interaction.” Go down to the next one: “Ooh, I like how that kind of zoomed this way and brought it off to the side. Whoa, that would—no, that’s not going to work.” You just keep going down, and you’re like, “Oh, I like number 3 and number 8.”

Yes.

“Give me 10 different ways to do number 3 and 10 different ways to do number 8.” Now you’re getting deeper and deeper. You take a little from here and there: “I like the zoom on number 3, but I really like the fade on number 8.” This is the conversation you have with the LLM.

By the end of the day, you’re left with something that is truly unique and different. It’s through telling the AI to take that idea and give you 20 different versions of it. Then you just take that little freaking component, drop it in Next.js, and you’re off to the races.

What I love about this, first, is that it’s such a fascinating workflow. But also, you’re not thinking about ROI. You’re not thinking, “Oh, let me talk to the designer and ask them to come up with 20 variants. That sounds like a lot of work. I better prioritize.” Instead, you’re able to go infinitely deep on something seemingly trivial and have this level of polish and craftsmanship that just wasn’t justified 5 years ago.

That’s where we’re going with software.

Well, what’s funny is I have this little interaction where you can heart something. I’m thinking more about the little personal forums that I’m trying to build. One of the things I wanted the heart to do was just explode and fade out. I came up with a way to make it do that at an angle, and it’s all vector, so it just does it beautifully.

The emotional connection to that little, tiny nugget—that little atomic unit of “Wow, that was a fun interaction”—is important. I want surprise and delight peppered throughout all the products that I build because I think it really gives the product a depth that you can’t find anywhere else. That’s what I’m always looking for.

So cool. Very cool. So, how are you using it?

I’ve been using a lot of the same products that you’ve been using. I’ve been using v0 for a lot of initial explorations. I use Replit and Base44 for things that I need to do fast and have working.

What’s Base44? Why Base44? Why that one?

I think Base44 is an interesting company. It’s a one-person company that was acquired by Wix for $80 million. Philosophically, they’ve talked about this approach of “batteries included,” which is that when you’re using these code-generation platforms, you shouldn’t have to—especially if you’re a mass-market consumer—you don’t know what Supabase is, and you don’t want to know, right?

Everything should just work.

And I think the characteristic of that is that there are a certain set of apps that the platform simply won't generate. If you want to do some complex data-processing thing, this is not the platform for you. You want to do—I have a fun little Instagram clone for cats called Catsstagram for my kids and me, and it's super fun. It's super easy to build. Base44 one-shotted it. I deployed it, bought a URL for it, and it works.

We can cut this out later, by the way. It's a little embarrassing, but it's fine. Yeah.

Dude, it's great. It's just that things work, even though the things that you can do are probably less ambitious. Yeah.

So, I think there's this sort of spectrum of how ambitious it is versus how batteries-included it is?

Yes.

And for simple things, fully batteries-included is awesome. Then, for really ambitious things, I use Cursor; I use GPT-5 Codex, both the model as well as the CLI. I use Sonnet 4.5. All of these products are the things you need to use to be ambitious.

Convex was something that you turned me on to, which I think is quite cool. Can you speak to why you chose that database?

I mean, I'm speaking sort of outside of my area of specialization, but it's a Series A that my partner Martin led. It's a real-time database, so it's great for applications like chat.

And how fast you got chat up and running with vibe coding is insane to me, because I had built a chat thing on Postgres, and to do WebSockets and state management—all that—was a pain in the ass. You had it up and running in minutes.

I mean, it's just the nature of the product, I believe, of the database. So, it's really, really good for real time. I think the second thing is you define and interact with the database in code. So, there's no going to the dashboard to do an insert in the SQL table. It's just very, very clean.

So, look, I think this is the magic of vibe coding. You can explore all of these things. You don't have to become an expert in Convex.

You don't have to become an expert in TypeScript. You can just try it.

It's so cool.

I love it. What else is in your stack? I know we kind of cut off a little bit there. You had said that Notion, and I'm assuming you're using—what else?

Yeah. So, the other thing I'll just speak to is that, because I spent a lot of time with AI music, I love AI music. It's very, very exciting. I believe that anyone who's ever picked up an instrument has wanted to play it well. But there's this sort of technical aspect to playing music that's similar to the technical aspect of coding, right? How much unmade music is there because people don't know how to play instruments?

I think AI, at a minimum, is an instrument that lets people express their musical desires—perhaps something more important than that.

Is there anything cool there on the music front? I played around with a couple of them, but I haven't seen anything that has been transformational.

Text-to-song is where all of this started, with Udio and Suno, a year ago. Text-to-song is cool, and I think it's an important part of the consumer use case, but you kind of want to go deeper and be able to refine it further.

And now we're seeing all the products. I played with one called Mureka this week. ElevenLabs has got a really good product. Suno itself has actually launched a DAW within its product. So, you can start to do things that are more ambitious within these music products. It's really, really inspiring.

I've also been making remixes of music videos. I did one called “Impossible Tiny Desk,” which was The Notorious B.I.G. doing the NPR Tiny Desk series.

That's amazing.

I used Hedra for it. I used Demucs for stem separation. It wasn't very complicated to make, and you can just experience this moment that you couldn't otherwise experience. I've also been doing ones of 90s music videos. So, we did a little remix of Nirvana's “Smells Like Teen Spirit,” and I generated all the video in Veo 3.

It's so satisfying. And you know how many people saw it? 10. But it doesn't matter, because I felt fulfilled making it.

I saw that there was a video that went around probably 2 or 3 months ago where a music producer—and it was a couple of people—went from prompt to album artwork to a published song on Spotify in a few hours. They made a completely fake person, and the song sounded pretty awesome.

Yeah.

And it was just—God, I have to imagine there's a massive opportunity here to unlock creativity if they get the right tool, you know.

And it should be able to go in and sculpt that song and kind of hone it to your own personal taste via your prompt. Is that where you see things going?

I do, and I'm very—I'm very—there's, I think, a lot of the conversation, unfortunately, has been about “slop, slop, slop.” But I think if you look at music, music is a sort of adaptive system that changes because of culture.

Let's say we had a model, and let's say we trained the music model with every genre of music right up until hip-hop, but not hip-hop. Would it infer hip-hop? No. You needed the Bronx. You needed Queensbridge. You needed New York culture in the 70s to have hip-hop.

So, culture is such an important part of music that you don't just get a model that knows all the music we're ever going to make. Our lived experiences are what inform the next genre of music. And that's why I'm so bullish on it.

Yeah. I don't know that AI would build brain rot if it just—my kids are so into that crap. I'm just like, “Oh, my God. No one should listen to this.”

I'm pro. I'm pro.

You're pro brain rot. All right. One of the things I wanted to chat about—I'm sure—do you have more stuff to jump into? I would love to talk about the idea of always-on recording and the social impacts.

I wanted to ask you about the kind of trend spotting.

Okay, whatever. You start first. Yeah.

Yeah. Yeah.

So, one of the things I've known you forever, and the people that know you, Kevin, is that you're always early to important new consumer movements. So, when you change your mind about something, 1, 2, or 3 years later, the rest of us change our minds about that. And you've done that consistently, right? With small things, maybe small things like coffee, and with really important things like media. I mean, you were years and years ahead of the rest of the technology industry in terms of embracing media.

What's coming next? How do you know what's coming?

Well, some of it is this idea of saying no to things that you don't think are going to have any traction. I did a lot of that. When everyone was going heavy into VR and AR, I kind of avoided that. It's a little pessimistic, and it's not necessarily predicting the future, but knowing how not to spend your time, I think, is a big, big piece of it.

At the end of the day, I really like what Ryan Hoover did in creating the Weekend Fund because it was actually based on a Chris Dixon quote where he said—and I'm paraphrasing—“What the geeks are playing on the weekends will become mainstream.”

I thought that was a brilliant quote from Chris because he also was very prescient. He's so good at predicting the future and seeing things before anyone else. The only way to find this stuff is to have this inherent childlike drive that says, “I must play. I must play.”

It's not something you can tell someone to do. You just have to be constantly at that edge of what is possible and always be kicking the tires on things. If that feels good to you, if that feels like something that doesn't feel like work, as it does for me, then I think you will naturally be led into the next big thing. But it can't be a job or something someone tells you you have to do. It just has to be driven from personal curiosity.

When I was a little kid, there was a show—God, I hope I get the name right. I think it was called Beyond 2000 or something like that. Do you know what I'm talking about? There was a TV show, and all it was was it would just tell you for a half hour—and it was probably wrong 99% of the time, because they'd talk about flying cars—but they would just tell you where they thought things were going and try to predict the next 20 years.

I was enamored by that. I'd sit there and be like, “Wow, we're going to have flying cars. This is crazy.” But every episode was just a different vertical of what the future could look like.

And that has never left me: this idea of wanting to see what's about to happen and live at that edge of things. Even if I wasn't investing, if I decided to call it quits and never turn on another camera, I would be holed up in a little room somewhere constantly playing with the latest stuff.

So, what does that mean today? It means I splurged on a crazy-ass GPU for my house, and I'm downloading local models. You and I were talking even last night. I had this idea where I'm asking all the models out there to give me the 100 best albums of all time—meaning music albums of all time—not based on downloads, but cultural impact and their impact on other genres springing up, to your point.

I'm going to have them duke it out. I'm going to select the most expensive option to do. I'll probably burn half a lake's worth of water, or whatever the hell it's going to—I'm joking, but it's pretty bad.

So, they're going to give me, then stack-rank those and see where they agree and where they don't agree, and come up with a list. Then I'm going to go and have each of those albums at a granular level.

Tell me why each song is very important and create a script for about a 2-minute-long podcast for each of the albums. Tell me where to pay attention to each of those tracks, why, and what riff to listen for that impacted somebody else or influenced someone else. Then I'm going to feed that into ElevenLabs and create a beautiful, relaxing, guided journey for that album.

Then I'm going to tie it into the Spotify API so that I can sit back and have it on my list of to-dos to listen to an album, say, once a week or something like that, get that guided, beautiful journey, and go through it. That product doesn't need to exist, but I want it for myself, and so I'm going to go build that.

But it's an example of how that is a forcing function for me to learn speech-to-text, to learn how to have these models duke it out and come up with the best possible answer, and to build the whole backend that's required to pull that off. That's the play at work, in action, right? I think that if you can just find that for yourself, you will find the frontier of where things are going just through that natural curiosity.

I don't know where things are going. I will tell you that there are a couple of things I know to be true. One, I think engineering is over. I think we're going to be orchestrators of information, not engineers. I think those will be solved problems. Anything where there is a nonsubjective outcome, AI will tackle and solve that problem.

Like, engineering is a solvable problem. If I want to fan out 1 million stories to a social network, that is not a subjective thing. There is a way to do that at scale, efficiently, and with the end result that we're all looking for. AI will figure that out. You won't need a team of 10 engineers to figure that out. Those will be solved problems.

There are certain things like that that I look to and think, “Oh, the world just hasn't woken up to this yet.” I mean, we all, as technologists, have—but—

I wouldn't go for a CS degree. I would not go into that right now. There are little bits and pieces like that, but then you think, okay, well, what has to happen in order for that reality to unfold? What tools need to be built? What stacks need to be built? That's where the investment opportunities lie.

If you see this stuff—like, I was on The Tim Ferriss Show talking about when NVIDIA was hitting its kind of $1 trillion mark. I'm like, “This is going to run, because we're at the early innings of this,” and talking about this stuff. It's so funny, so many things on Tim's show, because he asked me to make these predictions.

When Ethereum first got off the ground, even before it launched, I was talking about it on Tim's show, because if you're paying attention, you'll just see it, and it'll be right there in front of you.

Yeah, I love that. I want to quibble with your point about a CS degree. But first, I want to say that so many good things in life are downstream of authentic curiosity.

And I love that that's the way that you've done this, because it's not some big-brain, galaxy-mind, predicting-the-future thing. It's just, “What gives me energy?”

That's right. That's the whole thing. I remember Michael Arrington way back in the day. He was interviewing me on TechCrunch Disrupt, and this was probably 20 years ago. He said, “Hey, I want you to talk about your investments. Why—how are you good at picking these things?”

I was like, “I don't know. It's kind of based on my gut and what I'm excited about.” He's like, “Dude, you've got to have a better answer than that.” That's what he told me before we went on stage.

I was like, “It's not a magical formula. It is intuition, especially at the early stage. Especially when it's awkward, because if it looks awkward, it's going to be contrarian. No one's going to want to do that weird thing. Those are the hardest deals to do.”

It's funny, I was talking to—I think it was Reid and David over at Greylock back in the day, and they said their best investments were the ones that were the most contentious internally, because people don't see it. Not everyone sees it. That's okay. That's actually a positive signal, because many people are like, “That will never work.” But guess what? When it does, it changes the world.

What's remarkable is that anybody who's got nothing to lose can be embarrassed and be willing to be embarrassed. But once you have something to lose—you know, you're a person that people know and trust—the natural human instinct is to be more cautious and to take less risk around being embarrassed.

Because you've consistently been at the edge, maybe because of this authentic curiosity, you've done it a lot. I think that's a real conscious thing you've done, which is unusual.

But I've also lost a lot and made the wrong call. At the end of the day, when my kids look back at my career—and I have them as my little proxies for the world—I want them to look back on Dad's career and not say he was the best investor or that he made the most money.

I want them to look back and say, “Wow, he took a lot of risk, and he was wrong a lot. But you know what? He got to scratch every personal itch that he had and try everything. He was at the buffet and tried it all.” A lot of it failed. Most of it failed. But that's okay, because what are we doing on this earth if we're not trying crazy shit? Do you know what I mean?

100%, dude. 100%.

Back to the thing that you said you wanted to debate out: What was that?

I want to quibble about your point around a CS degree, because I actually think this is a really, really important point. If you're somebody that goes to Stanford, or maybe Harvard or MIT, those people are leaning into CS degrees in a big way.

Unfortunately, I think this story has gotten out into broader society that CS is over, engineering is over. Sure, being able to write a certain kind of algorithm may be less valuable than it was 10 years ago, but being able to think technically is more important than ever.

I think it would be a real tragedy if, especially, kids that go to non-Ivy League schools decide not to do CS because CS is over. I think the value of being technical is higher than it's ever been. While the future technologists may not be programming in the way that we are, they're going to do something that needs that skill set.

I will say I think it's over. The reason I think it's over is because I would argue that creativity in the future is going to be more important than technical ability.

We don't need to teach people SQL, or have anyone learn these things in school. I'm curious—let's crack that open a bit. What do you believe that a technical degree gives you that will be useful a decade from now?

Well, look, I think, 1, it's just fluency, right? What is a distributed system? What does that mean? Yeah, I think it's useful to understand that as a mental model for systems thinking.

The second thing is just general problem-solving. How do you go step by step and solve a problem? How do you understand how to explore a problem in design space? Where do you take risks or be ambitious, and where do you not?

A lot of it is downstream of mathematics. I went to computer engineering in Waterloo, and none of the stuff I took away—in fact, none of the things that we typically studied—was a programming language. A lot of it was systems thinking, and I think the systems thinking is actually really important.

That's really interesting, because I think that's different from a classic CS degree. One of the things that I like about some of these programs—and they're largely more on the vocational-school side of things—is how they've evolved. It's training you to become an entrepreneur, which is full-stack: building it, designing it, coding it, shipping it.

That depth of knowledge across the full stack impacts marketing, creativity, hiring, firing—the full suite.

Yes.

Essentially, it will of course be very important in the future. I think that just a singular technical focus around the code piece of it—that's where I would like to see the degrees loosen up a little bit and be a little more holistic and overarching.

I don't think—and don't get me wrong, there will be technical challenges that we have in the future—but if you get to scale as an entrepreneur, you will have plenty of everything to go solve those challenges, right?

I think that's right. Being a technologist, and broadly having a technology mindset, sets you up to be a good founder. I actually think Elon has probably run the most ambitious experiment on this that's totally under-discussed, which is that he typically has technical people in every role—

Right.

Marketing, comms—you name it, there's a technical person there. I really do think there's something in the training that sets you up to be successful in a lot of vocations, which is totally distinct from programming.

Yeah. What I look for in a founder is someone who's kind of multidisciplinary in some way. If you're too technical, then you don't understand—

I met a founder yesterday who's building a device that can listen to everything, another one of these types of devices. The angle of what they were going after missed a lot of the social side of it: what it means to be listened to, the invasion of privacy, and how to potentially sidestep that.

They were all technical, all device-focused.

And I feel like, in the future, you have to have that kind of full suite. It’s not just any one discipline; it’s probably going to be a handful of them—the orchestrator role.

Yeah. It’ll be interesting to see. I also think there’s value in ignoring—I don’t know why I’m just arguing with you here—but there’s value in ignoring the social norms. A great example of this is 10 years ago. It’s so funny looking at the mobile transition. I remember it so distinctly: iPhone in 2007, the App Store in 2008.

When the App Store came out, there were 6 million iPhones in distribution, right? Not a lot. Today, ChatGPT has 800 million active users, right? We’re not even 3 years in. So, it’s crazy the scale at which these things grow.

If you look at the predictions from 2009—you can go read the blog post—no one knew what to do with the technology. The most consistent prediction was location-based ads.

That was the most ambitious thing anyone could think would happen with mobile. But one of the big predictions around location was, “Wow, people don’t want to share their location.” There was angst amongst the pundits about that.

Fast-forward 10 years, and every Gen Z—every Zoomer—shares their location with all their friends and their exes and who knows who. That moment—from consumers feeling one way about location to the sort of tipping point—happened very rapidly.

So, I can’t speak specifically to what’s going to happen around being transcribed, but I do think ignoring the social conventions can be a really good instinct for founders.

Yeah. I worry about this trend of always recording. Do you believe that this is going to be ubiquitous? Do you feel like we’re always going to be recorded? Is the future one where you can’t—I mean, there will be safe spaces—but you largely can’t find a safe space because every meeting you go to, every dinner you go to, it’s always on?

Look, I think that technology and social norms sort of adapt to each other in lockstep. So, yes, I think there’ll be more things recorded, and I also think that we’ll develop new social norms so that it can fit into our world. Right? Technology is downstream of what we need as a species and society, and we’ve always adjusted. We’ve always adjusted, and I believe we will again.

Right now, we’re in this awkward, transitional period where, if you’re in a bar in New York, you’re definitely getting punched in the face for recording. In San Francisco, everyone’s recording. So, it’s just a moment of transition where society is not caught up with technology. But I do think we’re going more in that direction than less.

Yeah, that’s going to be a hard one to overcome, I’m telling you. I’ve gotten—well, certainly on the personal side, if you wear one of those devices that’s always recording with a partner, that doesn’t fly. On the professional side, I’ve seen a lot of people ask you to take it off, to put it away.

But also, I enjoy the rawness of an unrecorded conversation because you will—

Put your toe in and say things that are probably more real. I’d rather have that conversation happen than something that’s just like, “Oh, let me reference my notes and make sure that I got that right,” you know?

So, we’re talking a lot about how it changes things for the negative. We’ve got an amazing portfolio company called Limitless. They do this, and the benefit of having a lot of those things is that sometimes you’ll find yourself saying things you didn’t expect to say, and they just get lost in the ether.

The benefit of having that signal and having it be less transcription-focused and maybe more ideas-focused, maybe vibes-focused—there’s a form of lossy compression you can take with the ideas—

That, I think, actually—

It has to go there. I think the lossy nature of it—for people who don’t know what we’re talking about, lossless would be every word verbatim recorded in a diary somewhere, which is one hack away from our personal conversation in an Uber being exposed to everyone—that bothers me.

As friendships deepen and evolve, you’re like, “Hey, this was a tough thing I was going through,” or whatever it may be. The fact that that is somewhere in the cloud is just bad. I think it’s bad for all sorts of reasons, but—

The idea of lossy is interesting because you pick up broader themes—

And if you can pull emotional context along for the ride, where it says, “Hey, we noticed that Anish was feeling down about this,” and I’m like, “Oh, let me send him a little something to cheer him up,” because that was a great callback to our conversation. Totally.

That side of it is fascinating because then I don’t feel awkward having the conversation, especially if we can do that on-premises, on-device, where you have a small model running on-device.

It doesn’t pick up the fact that I said this one thing that would be hurtful out in a broader context, but it is getting the emotional tone that’s surfaced back in a sentence that cannot be connected directly to what I was saying.

Yes.

I love that idea. I hope it goes there.

I think it will. I think this is the interplay between social dynamics and technology. It’s going to go there.

The Limitless device—I had it, and my dream feature—

Is that they have a little LED on the outside. There have been any number of times when I had it on and people are like, “Turn that thing off.” And fine, that’s fair, right? Turn it off, whatever.

I want a mode where, if it’s red, it’s doing verbatim recording because there are meetings where I want that to happen, right?

If it’s green, it’s recording, but it’s theme-based. So now I can have that personal, intimate conversation. I know it’s happening on-device. Those more intimate words will never be leaked online, but then I know where I stand with you, right?

Because I think that’s so important to know, like, “Oh, what we’re having is going to remain largely confidential. It might pick up some themes, but it’s never going to be damning to me.”

That’s so interesting. I mean, and that’s exactly it, right? A visual cue—

Right? Has to be—

Confidential mode, you know, and that’s not a technology problem. That’s a product design problem—

Right?

I think we’re going to solve it.

Yeah. Well, that’s exciting.

Yeah.

Because I certainly believe that, especially with our brains as we get older, having that additional ability to jump back into context and find some of that information—

Although I did use it—I tell you, I used it with Daria one time, where she told me she had said something and I pulled up the transcript.

I keep shooting myself in the foot.

That is a no-no.

No way. Because you pull it up and you paste it, and it’s the last time you wear the device.

There is no end. That’s when you just go limp and take the pain.

Well, brother, this has been great. It’s so great to have you. Anything else you wanted to touch on before we wrap things up?

Now, I absolutely had a blast. Let’s keep talking.

Yeah, let’s keep doing this. It’s good to get together every so often, hash out these things, talk about the future. Where can people find you online socially?

I’m on Twitter, illscience. We can talk about that another time, the genesis of that name. And you can always get me at anish 16Z.

Awesome. So, hit me. I want to know what you’re working on. You know, send me things that you actually have made. That’s always the best way to engage with us. Best way by far.

But generally, I’m just excited to be out there and thrilled to be here with you, Kev. Thank you.

Awesome. Yeah. Kevin Rose on Twitter X. I still get confused. I call it Twitter on X. Yeah. Awesome. Let’s do this again soon.

Thanks for a fun session.

Yeah.

All right, man.