The a16z Show - Network Effects, AI Costs, and the Future of Consumer Investing with Anish Acharya on The Kevin Rose Show

Episode Date: April 19, 2026

This episode originally aired on The Kevin Rose Show. Kevin Rose speaks with Anish Acharya, general partner at a16z, about how AI is rewriting the rules of consumer software, the defensibility of network effects in a world where anyone can spin up an app in 48 hours, and why the real threat to consumer founders may be the cost of inference, not competition. They also discuss model pricing, the future of the four-day work week, and peptides.   Resources: Follow Anish on X: https://x.com/illscience Follow Kevin on X: https://x.com/kevinrose   Stay Updated:Find a16z on YouTube: YouTubeFind a16z on XFind a16z on LinkedInListen to the a16z Show on SpotifyListen to the a16z Show on Apple PodcastsFollow our host: https://twitter.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.

The idea guys are sort of having a moment. In fact, it's funny, like, I'm like looking for new ideas to work on. You know, in the old days, somebody would give you their app idea and you'd be like, oh, here we go again. And now I'm like, cool, how about I build it for you? What has changed for me is, you know, someone that dropped out of computer science because I just couldn't keep up with everyone else. I always had the creative ideas, but my ADHD was too bad that I just couldn't remember all the syntax,

and I was like thumbing through manuals back in the day and, like, you know, trying to like, I had like the C++ Bible or whatever they called it, you know? When you're a different, you know. child, nobody tells you, Kevin, you're bad at drawing or you're good at painting. The thing that I think we need more than UVI if we ever get to that place is universal basic purpose. And the way you actually get the French revolution is less that people don't have enough money, though that's part of it, and more that people don't have something important to work on. Everybody's got to feel like they're on a hero's journey. I was talking to my wife

and I was like, we're getting a lot of arguments about X, Y, and Z. What if we just had a conversation with a model that built out these frameworks for us that understands what we like, how much we care about certain things. And then we just have our models go and like duke it out. And she looks at me and she's like, this is one of the worst ideas for her. Yeah, I'm just like, shit. This episode originally aired on the Kevin Rose show.

When anyone can build a Slack competitor in a weekend, what actually makes a consumer startup worth back in? For decades, software moats meant engineering effort. When Kevin Sistram and Mike Krieger were hand-coding Instagram's filters, copying them cost you months, That window is now 48 hours. But a niche argues the moat was never really the code. When Hipsomatic and a dozen others launched alongside Instagram,

it still wasn't obvious which would run away until it was too late. That pattern may hold, but the cost structure has shifted. One founder told Anish he'd need $25 million just to reach 100,000 monthly actives because AI inference isn't free. So the real tension isn't whether great consumer products get built. It's whether venture economics can survive a world where the best companies skip early rounds altogether. Kevin Rose speaks with Anish Achari, general partner at A16Z, focused on consumer investing. Anish, we're back.

Nothing has changed at all, so I don't know what we're going to talk about. Exactly. How long ago did we do that episode together? God, four or five months? Everything has changed. That's so amazing. Dude, great to have you. Real quick, a primer for everyone.

Your partner, General partner, Andrews & Horowitz. That's right. He focused primarily on consumer investing. That's right. Anything else to mention on that front? I mean, I focus on consumer, but I'm a programmer. I mean, I grew up the same way that you grew up. We worked at Google together.

We worked at Google together. Yeah, yeah. So I've got consumers my area of investment focus, but I've got a ton of personal interest outside of that. Love it. All right. We got a lot to talk about. We figure this would be like fun little variety show, cover all the things, AI. Yeah, yeah, yeah.

Let's get this out fast. Get people, you know, thinking about what's coming. You've got a list. I've got a list. Do you want to start first? Well, I want to hear more about what you've been working on because I experienced one of your products this morning. You've been obsessed with the models.

You're just like, you can't stop programming. Yeah. I mean, tell me what you're working on and then tell me how much you think it's real productivity versus productivity porn. Yeah. Well, I'd be curious to see how you define productivity porn. But in terms of what has changed for me is, you know, someone that dropped out of computer science

because I just couldn't keep up with everyone else. I always had the creative ideas, but my ADHD was too bad that I just couldn't remember all the syntax, and I was like thumbing through manuals back in the day and shit, like, you know, trying to like, I had like the C++ Bible or whatever they called it, you know. And so now there's none of that. And I finally realized that six months ago I was going and manually looking at the code and thinking to myself, like, oh, I should just at least look at it to make sure I kind of see what it's doing so I understand best practices.

And now I just realize I don't ever have to leave. the code again. Yes. Because the next, it doesn't matter what bugs I'm creating right now. If I discover them,

I can squash them and the next model is going to be better and it'll rewrite any bad functions or anything that I have going on. Yes. There is,

there is a world, what's so weird right now is you're taking a look at all these companies that are SaaS businesses and they're realizing that the moat is no longer there. Like,

I can spin up anything, my own personal CRM app, a Slack competitor, anything that has definable outcomes for software. You know, I'm not making gene editing software, but anything like a SaaS-like business, very easy to replicate. It's easy to say, I want to go build my own workout app, right?

And like little things like that, are those businesses? No. But like the stuff I'm working on is personal, passionate areas where I believe I want to put my own time and attention. And for me, I've always cared about social news, you know, starting dig back in the day. and I believe that we're entering into a really interesting time where information is going to be platform and app agnostic so that it should be able to traverse and meet you where you're at.

So if you care about the latest AI coding tools, I'm just going to pick that up. If you want it in a podcast form when you're coming into work, custom tailored to you, if you want it in a video form that you tune into an Apple TV app and it says, Good morning, Oniash, and it welcomes you by two hosts that you want. Yep. If you want it in a daily newsletter, if you want it inside of your Claude Code experience,

or your Claude Co-Work experience, if you want it in your Obsidian, so there's like a little markdown file that you read every day. Yes. It doesn't matter. It is just going to, because all the connective tissue is finally happening, information can freely flow amongst these different, diverse platforms, which I think is beautiful and I'm super excited about. You know, it's so interesting.

So the Obsidian founder is one of the most interesting people to follow on. Nice guy, too. Oh, really? I've met him. Yeah. Kipano, right? Yeah. So he's got this whole theory of how apps are ephemeral.

Sort of apps and intelligence are these infomeral things, but files are permanent. And there's something very elegant and beautiful, and it sort of speaks to what you're describing, which is a file is just the basic unit of information. And then it can get expressed as a video, as a pod, as all of these different things. Right. And the app with which you consume this information may actually change from time to time. So I don't know if that's right, but something about that direction feels spiritually interesting. Well, for me, what I'm doing is when I work with coworker or work with any of these apps,

I'm always saying write this and markdown. Right. And I want it in Markdown because Markdown for me is kind of like the text, basic, lowest atomic unit of what is possible and portable, you know? And so I know in the future, anything will be able to ingest markdown with ease. Yeah. And so for me, that's kind of how I'm thinking about my file structure of all things. You know, it's interesting.

If you hear Mark talk about a bunch of the design choices they made when they were designing Netscape, One of the most controversial choices they made with HTTP was actually have it be plain text on the wire because at the time it was like, well, that's not secure, that's not safe, that's dangerous. How can you just have plain text on the wire? So all the wire protocols were encrypted, and that was considered the best practice.

And I think Mark and Ben and the team said, let's actually just make it plain text because it'll be easier to work with, easier to debug. And that turned out to be a brilliant design decision because it drove a ton of HTTP adoption. So there's something that sort of mirrors that in what we're seeing right now. All these file formats, all the cloud storage, like, it's just too much heaviness.

And we're in this moment of beautiful interoperability, you know, bashable, all the things that you guys aspired to in the late 2000s. Yeah. It never really came to fruition, and now we're getting it. Yeah, absolutely. And I think that there's something about having this, like, if I put something inside the database right now, I'm kind of locked. I'm not locked in there because I could tell any agent to go and read the tables and understand the schema and, like, have it. But for me, I just want ultimate flexibility. And the context windows are big enough and search is decent enough.

And we see the compound engineering and some of these other protocols where markdown works just fine for now. It depends. Obviously, if you need to do large datasets where you need to do vector embeddings and you need to actually find similar documents like there's other bigger tools and bigger hammers to use in those cases. Potentially, but I'd say that those are the raw information should be stored in the most interoperable fashion possible.

So just, I mean, one of the great things about open clause is that all the memories are just flat files, right? They're markdown files, I believe. So now you've got a memory file per day, plus you've got persistent memory, which is like preferences and things that don't change about you. Right. And it just means you can use it in a thousand different ways.

You can try different memory architectures. Right. You just, your information is yours in a way that it hasn't been for 15 or 20 years. Yeah. Are you an open claw user? So I love open claw. It's very, very interesting. I guess I haven't been as at the edge of OpenClaas, others.

I'm just spending so much more time in cloud code and codex. Yeah. For some reason, it's just that form of creative direction is so much more satisfying to me than the kind of productivity 10X you get from OpenClaw. It's incredibly important and really cool architecturally. I'm just not as obsessed with it as, you know, Chris and some of our other friends. Yeah, yeah, yeah. How about you?

I'm in the same boat where I obviously installed it. I got a Mac Mini like everybody else did, and I wanted to start playing with it. And a little hesitant in the sense that I have sensitive data that I don't want out there. So I create all new accounts and all that. Yeah, same. But I have yet to see the one use case where someone has come to me and be like, ah, this is the game changer for me. You know, like, it does this.

And I'm like, wow, I want to do X. Like, that sounds awesome. Yeah. And oftentimes just little tiny things, little tiny productivity boosts here or there. And sometimes I got one friend that is calling him with little updates and stuff, and they can talk to it on speakerphone. Yeah. And those are fantastic demos.

They're fun. Yeah. But I just don't see yet. And granted, it's early days. But I like directionally, I think it's right. I think we're going to have a personal assistant likely just built into the larger models that just has their own storage. I mean, we're seeing this now with these agents that Anthropic rolled out.

They now have their own storage in own containers where they can keep these things over and make them durable over long periods of time. Yeah, cloud agents. Yeah, cloud agents, yeah. I mean, I think a lot of the magic of OpenClaw is just the ergonomics of it. And I'll give you a specific example. So when you use a website or a mobile app, you obviously expect it to be perfect synchronous. Like you press a button, there's an immediate response. That's your user expectation

from 20 years of using apps. When you actually text someone, you don't expect me to reply right away. You're like, oh, Anish might reply in five minutes or it might take him two hours or whatever. The magic of OpenCla, I think, is because it's in a mobile messaging channel, your sort of subtle expectation is it won't reply right away. So it has to go off and do something more ambitious. It can do it without feeling like this delayed experience. The other way that shows up, If you look at the model switching settings in chat chb t, for example, it's trying to find this balance between like instant and deep thinking, which is a little bit of a user expectation contour that they're trying to meet

versus just being like, hey, I'm going to go do some stuff and I'll make the tradeoffs and I'll let you know when I'm ready. Right, right. And OpenClaude does that really well. So I don't know that there's any one thing. It's just the whole way it's put together, I think is very elegant. Yeah. Back to that your original question, though, around, you know, this coding, what we're up to,

what we're building. Yeah. You're seeing a lot of entrepreneurs that, and you called it, what would you call something porn? Productivity porn. Productivity porn. And it's because everyone,

you, me, everyone in this room, actually, has produced some type of application now. Yeah. Some of that is trash. Yeah. Some of that it's going to be awesome. Yeah. I don't know how much of it is defensible.

Yeah. So as someone is doing consumer investing, yeah. Like what scares you? Well, first of all, I think that we need to, this whole mindset about defensibility and modes, and it all feels so heavy. Like, there's this beautiful creative direction in what you described and how you don't have to

thumb through manuals. And now the cost of doing something that fails is zero, right? So you get to try every idea, and there's so much information in exploration. So I say productivity porn to be provocative, but I don't actually think it's that. I think that we have this new way of learning, which is just trying things. And, you know, the idea guys are sort of having a moment. In fact, it's funny, like, I'm like looking for new ideas to work on. You know, in the old days, somebody would give you their app idea and you'd be like, oh, here we go again. And now I'm like, cool, how about I build it for you?

Right. So I think there's this beautiful exploration that we're saying that's never been possible before, and there's a lot of value in that. And then when it comes to consumer products, I still think that consumer modes are as good as gold, right? Things like network effects, they're not trivially reproducible by models. Even if you can reproduce a software, this is what, like, it's not like Instagram has got like software mode. Yeah, exactly. Yes, but okay, so that the incumbents are locked down.

That's fine. Yes, I'm not moving from Instagram anytime soon. But when you're sitting there and you see a new consumer product, you know, five years ago and you're like, damn, that's a good idea. You write a check for $10 million. You bet on an entrepreneur. You wait for the next five or seven years and you see what happens, right? Yeah.

Now knowing that, damn, that's a good idea could be replicated by anyone else if it is a good idea. Yeah. Does that freak you out at all? I don't think so, Kevin. Because even in 10 years ago, the good ideas weren't obviously good at the time. Like what Systrom was up to at Instagram wasn't an obviously good idea until the network really started to take. And then it was too late to reproduce the software. So I don't think that the moat has ever been. It's like really hard to reproduce the software. I think the moat is in

part, every consumer idea is embarrassing to work on until it's obvious. You know, it's like it seems trivial. It doesn't sound important. VCs don't know what you're talking about. Your family doesn't know why you're working on it. You just feel this. pull in a direction. And ideally, when it works, the network runs away before people can replicate it. But we live in a time where information is propagated like instantaneous, right? And so an open clock comes out or whatever they call it when they first launch, I remember now. And you see 30 different clones of it, right? And then all of a sudden, everyone's moving in different directions. And I just feel like that wasn't the case. It wasn't as easy to reproduce something like that. And you see a

hit than it is today. The time has shrunk dramatically. Like, for example, if I launched Instagram today, I remember when he goes by Mike now, but back in the day he went by Mikey Krieger and Kevin were working on this. And, you know, it took them, I remember talking to them about developing the filters. Yeah. The basic filters for Instagram. It was real work. Real engineering effort, right? Yeah. And so even if you saw Instagram start to hit escape velocity, If you were a startup, you had to go try and copy them and catch up to them. Yeah. Those are a few months of work.

Yeah. Now it might be 48 hours of work, you know? Yes. Even at the time, there was so much noise in the ecosystem, right? There was Bourbon, later Instagram, there was Hipsomatic. There was a dozen apps that all did roughly the same thing. And it wasn't clear why Insta was special until it was clear in retrospect. So I don't know.

I think that if you were to build Insta today, it would still be non-obvious until it was obvious in a way that it was too late. Right. That's fair because there's lots of things, lots of examples today of like, Wordle, for example. Yeah. You and I can clone that within 20 minutes. Yeah, yeah. It's like, but it's not, we're not going to take over Wordle

because they have that built-in base. Even OpenClaw, right, which was originally called Quad, Quad-Bot, I think. Yeah, Quad-Bot. So even though there's been a thousand forks of it, it's literally open source, open-quah itself still dominates conversation. Yeah. Right. We're not talking about the 10 open-claws that have equal market share.

We're talking about OpenClaught. Yeah, yeah. So there are these compounding effects that are sometimes, you know, less intellectually satisfying the network effects, software modes, whatever. Yeah. The thing I worry the most about with consumer software is that the costs are too high to have a free model, really for very long. Right. I was talking to Signal who's launching a product in the next couple of weeks in consumer product, and he was saying, hey, dude, I need to raise like $25 million if I want to have $100,000, you know, and by the way, the money will go quickly. So the fact that you don't have this sort of zero marginal cost of distribution benefit, which is really advantaged consumer founders in the past,

I think is a major drag on the ability of consumers to scale, consumer founders to scale things to a lot of people. So let's talk about some of the actual models on how people are building these things. It seems, and you guys are investors in pretty much all the bigs, right? Like Anthropic and Open AI. Are you in all the major model providers? We're not in all of them, but we're in Mistral, we're an open AI.

Anthropic or no? We're not an anthropic. Okay. So when you think about what's happening with the moment that Anthropic is happening right now.

Yeah. And you have Open AI, which has been largely, I mean, not totally quiet, but they're cooking. Yeah. Definitely cooking.

Obviously, this isn't win or take all or is it if someone hits escape velocity? Yeah. Well, I think it's, you know, it's easy to over-rotate on some of the conversations on X.

If you sort of zoom out and you look at Open AI, first of all, they've got, I don't know, 950 million weekly actives. Like it's the biggest AI consumer product by many orders of magnitude. It really is. And it's the first product to hit that kind of scale in that short of a time period in the history of technology.

So they've got a sort of unassailable advantage in consumer. I think that's one. And then if you just look at the model quality, I think we both experience this. Like the codex models are excellent. I think the codex harness is a little behind cloud code. And I hope they'll fix that. But I think that that's like,

that's like a hill climb that they can do. On model quality, they're right there, and we're hearing that the new OpenAI model, which is going to hopefully come in the next few weeks, will be just as good as the sort of super secret, you know, my-host model that's too dangerous to release. Do you buy that too dangerous to release? I don't buy it.

I mean, maybe, but look, I think that if you actually have a model like that, there's like four independent things that I think about. So first of all, let's even assume that it's six months ahead of our geopolitical rivals. The first thing you probably do is use it for offensive, capabilities, right, to hack others. The second thing you do is to lock down

all your defensive. Exactly. I think the third question, which Mark pointed out, and I think strategically wrote about this first, is do they even have the compute for it? Like, it might just be that they have a GPU shortage, which is also why they sort of nerfed Opus 4-6. I don't know if you followed this.

They took down the default thinking level from high to medium. People are complaining that it's nerfed, it's nerfed, because they may be trying to free up GPU capacity. So it may be an infrared problem. And the last bit is just like the aura farming of them having a model that's too dangerous for the world to handle.

Like, how can you have better marketing than that? There's no better marketing than that. So it's unclear to me that it's actually too dangerous. There are a lot of other reasons why they might actually be holding it back. There are rumors that they haven't coded anything internally since January, I believe. Yeah.

Do you buy that? I mean, Boris has said it a few times. Like, we don't write any code on Claude Code. Like, Claude Code writes Claude Code. So I do buy it. And like, how can that not be true? It's your and I lived experience as well, right? Right.

Yeah, I just, I wonder if you have a model of that capabilities, if you aren't, like, one of the things I've been really impressed about with Anthropic is this, the rate of shipping. It's like every freaking week, something new drops. Every day, did multiple things, yeah. And so you have to imagine, if you're the first one to have the model, that is, let's just call it 30, 40% better than what's out there today. Yeah. You have to be like, okay, let's make our tools like dope as hell as fast as possible. Of course, yes.

So you turn it internally and get all that tooling and just ship, it feels like that's what's happening because of the rate of shipping. You're an open AI investor. I think you're an anthropic investor as well. Okay, so what's your kind of view? I guess you're the most even-handed having been in both. Well, I mean, I think that Open AI, listen, they have the, it's hard to count anyone out at this point because, you know, even Google, they've got their, they got their

tensor processing units or their self-training model. They've been eerily quiet. And like, those folks are just sharp as hell. Like, don't fuck with Google. Like, they're going to come back and get you. So I just have a hard time believing there's going to be, at least from someone that's doing software engineering with the models, in allegiance to anyone. Like I will go to Chad GPT tomorrow or Gemini if it's a better model. I don't care.

Yeah. Like my credit card just transfers to whatever model, then it starts working again, right? Yeah. So on the consumer side, you know, when I talk to family members and people that aren't deep in tech like we are, it's all chat GPT. Yeah. because that is the household name. That's right. And I think it doesn't,

it does everything that they want it to do. Yeah. And it's not like they're seeing anything and we're like, oh, that co-work feature from Claude is so good. Like they don't realize that yet. Yeah, exactly. Yeah, you mean they're not using remote control on code?

Exactly. They're not connecting their phone and leaving their laptop open. It's so, Carpathie talked about this this week. And it's so funny because he's like, look, the people who have used chat GPT in a somewhat cursory way, they're like, I don't know about this AI thing. Maybe it's over hype, like, you know, maybe the Instagram audience. Right.

They sort of don't quite get it. So I think it's like, and for the ones of us who are really deep and we're using it to make software, and we're seeing all the model progress. Right. And we're not talking to each other. Right. You know? Exactly.

Like every time I talk to someone that's not in this space, I try to like red pill them and I set them down. And I said, let me show you what this can do for you. Right. And then within 20 minutes, they're like, oh, my God, I need you be spending more time here. Yes. But it is a lift to get people up to that. I think it's a multi-year lift. We live in a world where we're trying everything,

but the average consumer is going to try a couple new things per year, you know? Yeah, yeah, yeah. So it's interesting. What about model pricing, like the max that we're seeing these max plans for $200 and we eat all the credits and all of a sudden we're out of credits and people are complaining about that? Yeah. Do we see a $10,000 plan?

Do we see a $5,000 plan? Yeah. I think it barbells, so I think, I'd love your view on this, but it feels like the most powerful models are gonna be more expensive than ever, and yet the price of a token on GBT40 is down 100x since the model was released, and what was that 18 months ago?

And when GPT40 was released, it was this incredibly cutting edge, it had this beautiful personality, like it seemed inconceivable that it would be cheap, and now it's very inexpensive on a per token basis, plus you have open source, right? So it feels like there'll be inflationary effects on the edge models,

and maybe they'll even be unavailable as APIs, and there'll be huge deflation effects on everything that's not cutting edge. Yeah. What do you think? Yeah, I mean, I think there's, listen, like, even when Google announced these models that now run on phones, on device, there are so many instances where a lighter weight model will do most of what you need done.

Yeah. And so I love that that exists because there is a ton of things that I'm building right now, which are kind of like backfill and data kind of fetching and, like, just very non-reasoning tasks that six months ago, a year ago, would have cost me thousands of dollars to do what I'm doing. And now I'm like, oh, that was $12. So I do love that that is happening. I think that the API pricing makes sense. The weird thing for me is, I believe there will be a higher tier. We're already starting to see this where you can pay per poll request and pay Anthropic to review it with a better, more finely tuned model for those.

bug squashing. It is a little bit weird. It is a little bit of that Spider-Man point in a Spider-Man situation where like, I coded it for you, but I'm also going to fix the bugs for you. For like $12 a bud. Exactly. But the, so is the opening eye is security stuff is actually quite good as well. It's kind of a bunch of vulnerabilities that would have never found. Amazing. So, yeah, I don't know. But what's, what's up on your? Well, actually, let me ask you related question, though. When you think about your budget, I'm assuming that you don't have like a ceiling on how much you spend on the models. Right. But how much if you had to make a trade-up, would you trade off against, say, your entertainment budget?

Like, would you skip a movie or a bottle of wine because you wanted to spend the money on tokens? Yes, 100%. Well, but I'm building towards something. Are you, though? How many of your things are important and have to be durable versus are satisfying to work on? 80% are important and durable.

And the other 20% are just for fun, just because, like, I just want to see it exists. Do you know which is which at the outset? Yes. Okay. Okay. 100%. Well, largely because, like, you know,

I step back from my venture role, it's my investing role there, and because it's too fun to build right now. Like, why not build? And so when I'm tackling a bigger idea, the hope is that hundreds of thousands of people will use said product at the end of the day. Yeah. And if it is not that, then it's just going to be a little tiny thing that I think should exist.

And I'll put a little bit of time and effort into it. And if there's a thousand people that love it, God bless. Like, at the end of the day, I think we, I don't want to optimize for just financial outcome or usage. I want to optimize for, do I enjoy doing this? do I enjoy building this product, you know? I've seen some of your vibe-coded stuff. It is largely because you enjoy doing that, right?

It is fun to work on. And that's why I think your language is interesting. You didn't say it's too important of a time. You said it's too fun of a time. Yes. Right? So there's so much beauty in the process right now. And that's why I think that it will eat into what would otherwise be called

entertainment budgets because there's just, we're discovering this part of ourselves that has been asleep. Right. Both as technologists, but even as humans, right? Like when you're a child, nobody tells you, Kevin, you're bad at drawing or you're good at painting, whatever, you're actually just good at anything you're interested in, and we're able to get back to that place again for the first time in a long time. I know, and it meets you

where you're at, too, which is so awesome. Like, there's the number of things that I now know versus a year ago in terms of, like, how to fix, like, faucets and shit. And, like, the stuff where I'm pointing at chat chip-chipati at, I turn on video mode. I'm like, hey, how do I do this? What does this mean? Yes. And, like, I'm up-leveling my knowledge across an entire spectrum of different things. I've got to tell you this crazy story. So there's a guy that we met, came in and presented to us yesterday named Mike. So Mike is a personal injury attorney who won the global anthropic ClaudeCode hackathon. Okay. So he started using Claude code to solve a bunch of his problems because he's running, he's like a startup guy, sort of running a startup law firm,

and he couldn't afford a paralegal. So he used Claude code to basically start coding a bunch of things that would otherwise he'd have to hire someone for. And then he went so deep in the tooling. He came in, he showed us the software he's built. It's extraordinary. It's like what a whole venture-backed SaaS company would otherwise do. He's so deep in it. He's using whisper flow. He's talking about slash commands. He's like, okay, I've got a slash loop that's always looking for security fixes. And this is not the guy you'd expect, you know, he's like a jujitsu, square-jod, sort of personal injury attorney. Right. So I think there's just, this is my white pill. There's going to be so many beautiful stories of like digital homesteading or something where these unexpected people build

these really incredible things based on personal need or interest. Yeah. And, and, And I think this is where I had a company in a similar vein where they were spending something like 80K a month on SaaS subscriptions. And the founder was like, well, I'm good enough to where I can build this in-house, kind of what I want, a little bit more custom tailored to me. It took him about a month and a half and then you just cut that entire SaaS budget out. So here's my only counterpoint on the kind of SaaS is over thing, which is, yes, conceptually, maybe you could build a CRM. But have you built a CRM? I have, actually. Real?

Jesus. All right, fine. Well, I guess that's over then. I was going to say it's so fucking boring to work on all that software. I wanted a really simple personal CRM, so I just built it. It was a bad example. Okay, fine, fine, fine. Okay, have about payroll.

No. No. Payroll is safe. He just built some tax accounting software the other day. Richard, come on, man. I know. All right, so what else is on your list? I know you had some really spicy, fun topics to get into.

Yeah, dude. Well, I wanted to actually get your take on, you know, obviously there's a stuff that happened with Sam's house. There's a bunch of attacks on his house, which is, you know, super tragic, because whatever you think of AI politically, like, you know, he's a human running a company who has the best of intentions. What's your kind of view on the conversation we're having as a society about this new technology? I like how opening I came out and said, we're going to have to heavily tax us and provide some type of safety net fund. And we have to work that into the legislation at some point. so that when we do displace all these jobs, which it is coming, in my belief. You know, it may be five, seven, ten years from now. But I believe that we're going to have to have some of that flow back to the public in some way. I think when traditionally, and course correct people,

I'm not answering your question correctly, but traditionally when we talked about, you know, 10 years ago, universal basic income, I was like, well, how are they going to fund that? Like, how are we just going to, like, print money to give people money. And now it's clear to me that if the bigs continue to dominate and disrupt and just crush jobs across every single vertical, we're going to need some way for those profits

to flow back into everyday consumers so they are just everyday humans so that they can live and thrive in a world where they might have been displaced from a job loss. You know, it's interesting. So one of my good friends is an executive at Google. And, you know, I talked to him about this and said, hey, how much? are you guys actually laying off people as a, because you're obviously, you know, you're at the edge of the new models,

plus you have all the TPUs, so presumably they have all the tokens that they want and need internally. And he said, look, we actually haven't laid off anyone. We're just ripping through our backlog. So we're doing like 100x more than we ever did before. So it's interesting the kind of point.

A lot of people have made this point on like, hey, where are all the job's going to be? But I think there's no ceiling on human desire. There's no ceiling on company ambition. And so far we're seeing it as like companies doing more versus, hey, let's do the same amount with less people. Right. So we'll have to see.

I also think on the UBI point, like the thing that I think we need more than UBI if we ever get to that place is universal basic purpose. And the way you actually get the French Revolution is less that people don't have enough money, that's part of it, and more that people don't have something important to work on. Like, everybody's got to feel like they're on a hero's journey. Right. And even if it's a little, I mean, look, arguably we both have fake jobs right now. So even if it's a little bit contrived, I think that we have to make sure people have a sense of purpose

in addition to the means to kind of go achieve it. Where do you think that's going to come from? Again, I think there's no ceiling on human desire. I think people are going to want vacation homes on Mars. I think people are going to like, you know, companies are going to be doing more ambitious things. We're going to have 100x more skyscrapers. Like the history of all technology is it increases productivity

and it also increases human desire. Right. You've heard this thing of like luxuries become commodities, right? Therapy was considered a weird luxury 50 years ago. Yeah. Now therapy is an expectation. Right. with Instagram culture, so many bottle service

used to be something that was unheard of 20 years ago. Right. And now it's only bottle service. Nobody wants to go to the club unless they're getting bottle service. So I just think we're underestimating how much human desire will continue to grow and productivity. Yeah, flying was a great example of people

used to get dressed up and put ties on and suits on just for their flight because it was such a luxurious thing that happened. I wouldn't mind if we went back to that. Yeah. I'm big on manners. But yes, exactly. So I don't know that it's like the end of jobs.

And, you know, the other thing, I'd love your taking this since I know you're big meditation and sort of self-awareness is it feels like this technology is a tool to explore our emotional spiritual self in a way that no technology ever has been like the fact that you can talk to it like a human and you can explore things and you can get into flow state when you're programming it has all of these attributes that are just so different from you know spreadsheets and word processors and even the creative tools of the past which are very point and click do you think that's overstayed in the case i mean what's your take as a zem practitioner i mean i i think

Listen, I think that the one thing that's exciting to me, well, two things. Certainly human connection, real friendships, spending time together. If we can have more time to do those types of activities. Yeah. And we do lean into that versus just being hooked on our phones. Yes. I think it will lead to better and stronger emotional states, better family units at home, like a whole slew of benefits from that side.

There are certain pieces that, you know, I'm not. going to want a digital meditation instructor in terms of like, you know, I want a real human teaching me. It might come in digital form, but I want a real human that's behind that, not just some AI. Therapy, maybe not so much. Maybe I do like the idea of having a digital therapist. We'll see. So it's not all doom and gloom. I hope that we start to shift to celebrating craft and finding, in celebrating that if you find, find a niche that you are into, then you're the best at that particular thing,

that we can hopefully, you know, put some value behind that. And Japan is quite good at this. I mean, they've done it for 700 years, right? You find these small little artisans. Yes. They're the best. And there are people, like, I've gone to coffee shops in Tokyo

where it's like one guy doing aged coffee beans and there's like a line down the street to get in. Yeah. And it's just like, he doesn't want to become a billionaire. Isn't that beautiful? He's just happy to do this thing that he's found this very small little niche that people love. I hope there's more of that.

What's your version of that? Probably Japanese woodworking, a little bit more meditation, more time with friends and family. Getting off of technology, I found that I need breaks. We mentioned how it is so fun to do coding right now.

And because it is so fun, I can literally wake up and kind of like out of my trance of coding and be like, wow, I've been on the computer for 11 hours today. It doesn't feel like 11 hours, but that can't be healthy, right? So I just need to figure out what is that balance. Am I really, truly finding the time to hit the gym?

Am I trying to do the things that are going to benefit my body over the long term? Am I staying connected with my kids, my family? You know, I think those are very important questions I ask. And I worry a little bit that AI makes it more of a trap to get kind of get sucked in and avoid some of the... I don't know. Like, maybe it's barbell, you know? So maybe you spend these really productive, satisfying, fulfilling time on your computer.

And then you don't have any of the overhead that you typically had, right? All the meetings, you don't go to meetings, your agent goes to meetings for you, right? And makes decisions and handles conflict and sells customers and all this other stuff. And the rest of the time you're, you know, practicing meditation, hanging with your kids, touching grass. Right. So I think there's the potential to have this like really productive, satisfying technology time. And it's really productive maybe or satisfying sort of fulfilling human connection time.

why can't that be the shape of our life going forward? No, I hear you. And also, I'm, I've, I've, I've, I've danced around the edges of this in a dangerous way. Oh, God. So I thought about, I, so I was talking to my wife and I was like, okay, listen, we're getting a lot of arguments about X, Y, and Z. And what if we just had a conversation with a model that built out these frameworks for us that understands what we like, how much we care about certain things?

And then we just have our models go and, like, duke it out and be like, actually tonight we go out to dinner because historically Darya has had the right to go out to dinner and X members, I'm just making this up. Yeah, yeah, yeah, yeah. And she's like, and I explain this whole thing. And I'm super proud of the idea and I think it's going to like, save us a ton of time effort. And she looks at me and she's like, this is one of the worst ideas for her. I'm just like, shit, you know?

But so I worry about outsourcing too much of that negotiation and conversation and just, you know, saying like, oh, the agents are just going to handle it. That said, around certain things like policy. or conflict resolution between nations and things like that, there might be a little bit more of a steady hand to have two agents discussing their and getting to a resolution faster. I don't know.

It's interesting. So, well, I'm not sure about the family setting, but if you think of the corporate setting, you know, there's this old saying, like, the best way to compete with an incumbent is to pick a product area that lives between two VPs. Because the idea is the VPs hate each other so much. They won't possibly be able to collaborate

to actually, like, handle the competition in this area of crossover. Yeah. And that's like an area in which I think, look, if each VP had an agent that went in, I mean, we'll have to see. And I guess there'll be agent politics, corporate politics. But you could just imagine conflict resolution feeling so much less personal. Yeah. I think so many issues at work come from this like misattribution of professional decisions to personal feelings where it's like, hey, the boss went against me because they don't like me.

It's not because they just had a professional judgment that was in a different direction. Right, right. And I hope that this technology helps to less than that. Yeah, absolutely. What is the coolest thing you've seen in the last few weeks? Because you must see so much stuff in terms of pitches. And what has you excited to be an investor still? Because for me, early stage seems like,

I'm not saying that you're squarely and early stage. Yeah, yeah. Because obviously your fund does a whole gamut. Early stage seems pretty dead to me. Yeah. Just because it's like people, entrepreneurs are just going to skip that round altogether. Well, I think that there's venture exciting and then they're sort of like human exciting. Yeah.

And when I tell you a story like the one I told you about this guy, Mike, who built this incredible software and he's like, you know, a clod-pilled and all this other stuff, that's one of the most hopeful things I've seen in a long time. I mean, we should put Mike on billboards. I actually pair that with a lot of the stuff that Elon's doing. Like when you see the, you know, the effort to get to the moon, when you see SpaceX, when you see Waymo, like, oh, my God,

that's what we met when we said technology, you know? How about you? I mean, what do you think? No, I'm in the same boat. I worry that the kind of early stage, like, sassy world, the little bite-sized pieces of consumer stuff is going to be largely dead. Hardware is still very capital-intensive to get off the ground.

I'm very bullish on actually devices that encourage us to kind of disconnect and have more shared reality together. Have you seen Ten Can at all? Yeah, my son has one. Yeah. It's amazing. Yeah, so my, my,

You should describe what Tin Can is this little device that is a phone, like an old school phone that you plug in the USBC and it connects to the Wi-Fi. And then as the parent, you can tell the app which kids they can call. Yeah. And so it was so funny because my daughter picks up and she goes, she listens to it. She goes, someone's calling us. And she hands me in the phone and it's a dial tone. And she'd never heard a dial tone before in her life.

Yeah. Why would you? And she was like, someone's calling. It's ringing. It's ringing. It's ringing. I'm just like, no, that's what you had to listen for before you could dial the numbers. Incredible. But now when that rings across the room, they get so excited. They don't know who's calling.

And they go running to go pick it up as fast as possible can. But I want 10 can for grownups. It would be so cool to have a little phone that is only with your friends. And you don't know who's going to call, but you're going to pick it up because you know it's a trusted friend. So it's little bits of technology like that that I think I just going to provide these little moments of delight and real connection. that I'm excited for. It's very funny. So my son has one as well.

And so he geeks out when he gets a call, right? Yeah. So he gets, he geeks out when he gets a call, right? So he'll race his darts over to the phone. Just sort of like, uh, just grunting a little bit. And then they'll hang up. And how was your call? Oh, it was great.

It was really good. It's amazing. Yeah, but he's still young. though, too. It's not like they have like... A ton to talk about, I know. It's just such a funny little microcosm of like male relationships, you know. Are you looking at more hardware these days? Yeah, we're open to everything actually right now.

One of the most interesting, coolest pieces of hardware I saw recently, we're not an investor is this company Pocket. You know, a bunch of companies are doing things like this, these sort of like passive recorder devices. I know you, Sanbar, I think you guys have invested in one that you're excited about. Yeah. It's just, it's really elegant hardware design. And also just love that founders are being ambitious enough to try hardware again. Hardware has all of these, like, you know, it's just a challenging place to operate, but founder ambition is higher than it's ever been. So, yeah, we're definitely open to hardware. Yeah, that's awesome. Yeah, I've been looking at the always on recording

stuff for me has been a little bit like, eh, I don't know if I want that in my life. I know, I know, I know. Still, it's just cool that people are building things in those directions, you know. Yeah. I also think that you're too skeptical on consumer. Like, consumer feels like dead, dead, dead, oh my God, it's on fire. Right. Yeah, I guess the, the question. question is I think what the reason I am a little bit skeptical on it is I think it'll be dead dead dead oh my god it's on fire and then when it's on fire it's on fire with three engineers and they've you know hit something and all of a sudden the first round has done it at a $500 million pre yeah and I'm just I worry that what we're going to see is just a decimation of

all the early stage funds because they won't they'll skip all those rounds altogether and go straight to B or C yeah in which case and Drieson will do fine but like all these other funds are just going to be completely wiped out. So I have a history question for you. When you started dig, what did it, what did the environment feel like in terms of consumer? Like, does it feel like a good or a bad time? Who's dead?

Okay. Yeah, it was largely dead because we were coming out of the dot-com burst. Yeah. And then the thing, it was a technology enabler, which was Ajax that allowed for dynamic content to be refreshed, that enabled a new wave of kind of consumer apps to be built on. But it wasn't obvious at the time when you guys started tinkering. No.

But it was going to be like a global social media. Like, you know, I mean, it would be regulated. It would like potentially influence elections. Did you have any sense of how important it would be? No, there was none of that. And I don't think it was just, it was a very, it was a time where people said what if and then they tried it. Like Flickr came out and it was the first time people actually shared their photos publicly. Right.

Because that was a weird thing. You're like, oh, there's a picture with my family. Why would I ever want to put that online? Yes. Someone might see that. Those are my private photos. Photos were always kept in albums in your house. Yes.

You know, and like some family would come or look at them, you know? Yes, yes, yes. And so there was a lot of that going on. Like, you know, 4Square was like, oh, I'm checking in somewhere. Is it weird to tell people where I am right now? Yes. You know, like Twitter was like, what am I up to? You know, like there was a lot of strange things.

Yeah. And we're seeing kind of very similar experiments happening right now. And I do believe we'll see, you know, multi-billion dollar companies on the consumer side come out of this wave. Yeah. I have a feeling it's going to be structured very differently than it was in the past. Maybe. I mean, I don't know about venture economics, but I feel like if there's a big new set of consumer categories, venture will be fine and consumers will benefit. And I love that you talk about the photo sharing thing because I feel like anytime you have

new culture and new technology and they both happen at the same time, it's amazing. Like, think of the photos thing. Photos went from hyper-private to being public, to being public and seeming so important that you would never want to delete them, to snap coming around and being like, no, now they're disappearing. Right. Which felt like, whoa, why would I want it to disappear? Right. Ten years after, you're like, whoa, why would I want to share it? Right. But then it went back to Instagram saying, I actually want my photos to look cooler and stick around. Yes, yes. And so it's kind of all... And now, if you actually, if I shared an AI generated photo of myself with you,

you'd be like, this is slop. Yeah. So it's just, there's so, these cycles, there's a location's there's this whole moral panic around privacy and security and will people share location and location based ads, like this whole topic. Now if you look at it, Gen Z, they share their location with friends, family, exes, like everyone, right? They have zero expectation of location privacy. There's no more like sort of distorting effect on culture than AI right now, right? Think of even AI companions and friendships and psychosis and productivity porn and there's so many topics here that are going to drive culture, not just technology.

How can that not result in interesting new consumer products? Yeah, I agree. So let's wrap things. I have one final question, but if you have any one want to touch on me before you go, let me know. I know you got a hard out soon. I've got a little bit of time. I mean, I, uh... You got a good one.

Let's hear it. Okay, so let's like explore this and we can cut it if it ends up not being interesting. But we were sitting here getting ready for the pod and we were sort of trying to get our mics on the right way. And you were saying, oh, you've got to wrap the cord around the mic in a certain way. Yeah. So there's this sort of knack for getting things done, small things and big things in this world. And it's just in people's heads, right? Nobody has sort of written that kind of thing down.

You've called that dark data, right? So what's your kind of view on that type of information? Does that have value? Does that have more or less value than traditional information that's written down? And how does that show up in the future? Yeah, I mean, there is, I think that is a huge opportunity. I don't know how you monetize that, but it's an opportunity to unearth it, save it, and then work it into the models that we're doing.

You had a great example off mic where we were talking. talking about, you know, the one person that knows how to adjust a carburetor in a certain way for a certain model of car. That's right. And if that person dies, that knowledge is then lost, right? That's right. And so I believe that this exists kind of all around us in terms of little tiny micro things of dark data that we just have yet to work in and record in little tiny preferences and nuances that just never make it their way into these models. Yeah. So, you know, for example, this table right here, one of the things I wanted to do is I wanted to get a very Charlie Rose-esque table, which was an old, he basically was very famous for doing these interviews where he could like lean in and talk to someone.

Yeah, yeah. And I actually had, there was no stats or dimensions for his table anywhere in the models. Right. And so I provided the model with like five different screenshots. Right. And then set it on the heaviest working mode. And then it calculated the size of the coffee cup. Okay.

And then estimated the dimensions of the actual table. Wow. And so I could actually buy a similar table in terms of size so I could fit up the four guests sitting around this table, as Charlie did. Yes. And that was an example of something that the data was not in the model, but I had to back my way into it, right? Yeah.

And so I just think there's a lot of that. Proprietary data sets, I think, are really interesting. Yeah. In some sense, there's probably a market to go raise a fund just to acquire proprietary data to resell it. This is what Mercor and others are doing, you know? They're getting human data generation, some of which is like coding and math and all that. But I imagine in the future, some of which will be this type of tacit knowledge you're just going to be. Yeah.

So that to me is very interesting. It's not a business I would want to go build, but certainly I think we're going to get a lot more of that unearthed as robotics and other sensors come online. Yeah. I mean, I think Google Maps was probably the earliest thing going and mapping this kind of dark data. Yeah. But, yeah, it's models will, this is where I think when you talk about the layoffs, you're like, oh, none of my buddies are getting laid off at Google, if anything, they're just accelerating. But that doesn't account for the models getting better.

Because when the models are, you know, three times as performance as they are today or five times or ten times, then you really don't need as many humans. Your backlog quickly dries up. Yeah. And, you know, one person can orchestrate, you know, 20 projects or 100. Yeah, but dude, then Sundar says, like, I want to be the first $100 trillion company. You know, like, what is the... Sundar's not going to say, you know what, we should, like, keep our market cap, but just, like, be more profitable.

No CEO thinks that way, right? They want to shoot for the moon. So I think Sundar's ambition will grow at least or faster, as fast or faster, than model progress. Do you think government regulation steps in there and says you can't run everything? Because there is a world where just it's open AI, Anthropic, and Google, and they're kind of just everything to everyone. But we just, it's just not the world we're living in, right? Like, meanwhile, here on Earth, there's hundreds of open source models. There's all kinds of distilled models.

The model price, token prices for what we're cutting edge models three months ago are going down. Yeah. There's edge devices actually having models embedded in them. So that could happen, but I think the case for sort of outlier, you know, N of one, outcomes was much stronger in social networks than it is in foundation models. We just don't see foundation. You said yourself, right? I've got my credit card.

I'll use Opus one day and Codex and X-D. Yeah, that's fair. All right. Any predictions for the next six months? Let's hear the craziest wildest prediction that you have. I mean, weird or better, faster, man. Weird or better faster. So, okay.

Give me a hard prediction. Open AI IPO. I mean, I'm not supposed to talk. Look, I think all these, fuck, look, I think all the big model companies are going to go public. Yeah. You think that's like pretty well now? I mean, I don't know exactly, but I think that it's going to happen in the next 12 months with no inside information. That would be my prediction. I agree. I think that, look, okay, maybe like my strangest prediction is that I think we're going to get to the four-day work week in a uniquely American way. And let me tell you what I mean by that. So I'll tell you a little bit of a story. So if you think, back to the conversation we were having about diet and nutrition 30 or 40 years ago, there's two schools of thought, right? There's the European school of thought, which is we should eat less sugar. There's the American school of thought, which is we should eat less fat. And, you know,

maybe they were sort of influenced by industrial lobbies and whatever else. So we ran the experiment for 40 years, and Europeans stayed pretty slim and Americans got really fat, right? Now, probably what Americans should have done, and a lot of people try to make this better is said, hey, we need to just change our habits as a society and, you know, moderate a little bit and sugar intake Of course, that didn't happen. Instead, Americans did this uniquely American thing, which is we invented OZMPIC, GLP-1s. We're like, with this breakthrough new technology

to solve this problem that we really should have just solved with, like, perhaps different choices as a society and as individuals. I think the same way the Europeans have aspired to the four-day work week, and they're like, this is better for human flourishing and families, and I think that's right. But Americans have to do it the American way,

which is we create this incredible new technology that makes us so much more productive that we can actually reduce our workload from five days to four days and maybe even three and a half days. So there's some sort of a parallel here. And I think we're heading to a four-day work week. That is how the 20% productivity increase is going to show up in my view. And that's my stranger's prediction. Okay.

My strangest prediction is that in the next three years, we will probably discover and produce and create and get into humans another 50 to 100. peptides that alter our longevity in a meaningful way. Can you tell me about the peptides thing? Give me the quick take. So peptides are just small chain amino acids. They're very easy to produce in the lab. And they kind of like, think of them as upstream regulators. They can then trickle down and cause other things to happen in the body.

So they're not as powerful as like a straight up hormone. And they're not, so they're not as heavy as a hammer. They just kind of like cause things to happen. Yeah. And so the most famous peptide is, insulin. The GL-1s are peptides. But there are

20 other candidates that are out there that I've seen people experimenting with that are producing wild results. There's one called the Wolverine stack where literally I have had friends that have had back injuries and knee pain and elbow pain. I had tennis

elbow really bad in my which I didn't even play tennis which fucking kills me. But I had tennis elbow with my right arm really bad and I took the Wolverine protocol for like two weeks and I don't, I feel amazed. Really? Yeah. Literally. Yes.

And so I believe and there's a lot of people working in the space that they're going to identify these peptides will be able to easily compound them and create them. And the FDA is backed off a little bit. They were banning them for a while and now they actually have approved quite a few.

Okay. One for muscle mass in AIDS patients. and so they can do like give them this peptide and they'll like gain more muscle mass. There's another one that actually, I think, I believe it's Eli Lilly that is this three agonist, GLP1 that you don't lose muscle with now, which is really interesting. That's in the pipeline. It's phase three trials right now. Okay.

So as AI allows us to do a lot of this modeling in more real time and hopefully candidate discovery, I think peptides is just going to be a fascinating space to watch. So, okay. So I have two questions. One is, what's the basic stack for somebody playing with peptides? Like, what would you recommend? What's the right amount of risk to take? I would say go listen to Huberman's two podcasts.

He has two podcasts on peptides. Yeah. And they go in depth about each one. Some of them are for sleep. Some of them are, you know, they really increase your deep sleep and you sleep better. I have one buddy that says he sleeps like he's 18 again. He just like wakes up like fully rested and do these peptides. I have one buddy that's removed all of his visceral fat and is done by a dexas scan,

which we know visceral fat is like really the bad type of fat that you don't want to have around your organs. And yeah, there's almost a peptide for everything now, which is just really insane. And these are largely, you got to be very careful, though, because as human will talk about, there's grain markets, there's contaminants that can get in there. You got to work with like a high quality company when you're thinking about the stuff. So my second follow-up question is, you know, if we radically extend human life, do you think we're sort of emotionally, spiritually, societally prepared for people to live to 100? 120, 150, 250, 200.

I don't think so. No. Sadly, not in America. We just don't have the support infrastructure for the elderly nor the care, which is a huge bummer. But what if we controlled for that and said people live, you know, their full most flourishing 25-year-old lives? I agree. The thing you don't want to do is have, extend dementia for an extra 20 years because they're taking peptides. You know, it's like, so that's the tricky part.

But, you know, there are candidates like clotho and some of these other. That's a protein that looks like it's in the pipeline. for a potential treatment for a lot of dementia. So it's a very exciting time on the science front. Yeah. So that to me is, and then I would say my other prediction by the end of the year is that I just don't think we'll ever look at code again. Yeah.

I think that Elon talked about how, why are we actually writing languages? Yeah. Like writing coding languages. Is you just, if the AI is smart enough to understand the inner workings of every single CPU, GPU, and every piece of silicon that it's going to touch, it can just compile directly to binary and just work, which is wild because then that rewrites all of how we think about infrastructure in general. You don't have to pick databases anymore.

You don't have to think about any of these things. That, I believe, is going to be in the next five years, there won't be weird questions like, should I be using a graph database for this? Or, you know, it'll just kind of be solved, which is strange. I think it'll be interesting to see what happens when agents make these choices on our behalf. Right. Because there'll be technology with tradeoffs.

Presumably, they'll be marketing, but not to humans, to agents. Because ultimately, it's going to be codex that chooses the database. So how does that marketing show up? What are the trade? It's very wild. I don't think it'll have to be marketing. I think it'll just go spin up in instance, run it, benchmark it for its use case, come back and be like, yeah, I just test that out.

This is the best one. And then actually, Chimot had a really interesting point about it that a lot of people were calling bullshit on, but I kind of believe, which is he said that these agents will just shop around. Yeah. They'll go and they'll be like, okay, that graph database that was serving us well six months ago is now more performant on this type of database, on this infrastructure for this cost. Yeah. I'm just going to do the migration.

Yeah. And so you have to imagine that just happens like seamlessly at some point. Dude, it's so cool. So when I worked at Amazon when I was a kid, which was 2003. Delivering packages? Yeah, exactly. Were you? Going door to door.

No, hell not. I was a programmer. I was a programmer. I was a programmer. So Jeff used to say this thing. He said a lot of things, which were, you know, forward-looking. But one was we should assume that we're going to live in a world where our buyers have perfect information, right? And if you think of like how much of industry and commerce is based on either buyer laziness, apathy, or this asymmetric sort of information between sellers and buyers?

What if that goes away? Yes. What if like you have to win by product quality always 100%. There's zero value in brand. There's zero value in marketing. Like isn't that just a better world to be in? 100%. And so many of these areas, I think financial services is another big one.

And we were talking about Hero before the show, right, which is a company that was acquired by OpenAI yesterday, really talented founder Ethan, who started digits prior with this sort of consumer fintech mission that many of us had, which is like, hey, why can't we just make this system efficient and work for the everyday consumer, right? And it's like, why should they have to know how the credit system works and how to play the game and how to do all of these things? And I think with agents, a lot of that just goes away. Your agent goes and makes great financial choices on your behalf, given your constraints. Yes. And I think that's going to happen in a lot of product categories and it'll be a much better sort of world in society as a result. Yeah, absolutely.

Well, always a pleasure to have you on the show. Thank you so much, Kevin. Glad we did this. Come back. Let's just keep doing these, man. I love these little roundtables. It's so fun. Especially because every four to five weeks, everything changes all over again.

I mean, we can do this once a week. I know. We run out of time. Ran new topics talking about. Awesome. Thanks for having me, man. Good seeing you, brother. Super fun.

Cool. Thanks for listening to this episode of the A16Z podcast. If you like this episode, episode, be sure to like, comment, subscribe, leave us a rating or review, and share it with your friends and family. For more episodes, go to YouTube, Apple Podcasts, and Spotify.

Follow us on X and A16Z and subscribe to our substack at A16Z.substack.com. Thanks again for listening, and I'll see you in the next episode. This information is for educational purposes only and is not a recommendation to buy, hold, or sell any investment or financial product. This podcast has been produced by a third party it may include pay promotional advertisements, other company references, and individuals

unaffiliated with A16Z. Such advertisements, companies, and individuals are not endorsed by AH Capital Management LLC, A16Z, or any of its affiliates. Information is from sources deemed reliable on the date of publication, but A16Z does not guarantee its accuracy.