Clawdbot Creator Peter Steinberger Joins, Meta's Premium Subscription Plans | Jamie Cuffe, Ben Lerer, Lucas Atkins, Bridgit Mendler, Jeff Miller, Aaron Frank

You're watching TVPN today is Tuesday, January 27, 2026. We are live from the TVPN Ultradome, the Temple of technology, the fortress of.

0:00

Finance, the capital of capital.

0:08

Before we begin today's show, as many of you know, there's a major news story unfolding in Minnesota that's very sad and it's become a national news event. We're going to continue focusing the show on technology and business as always. But if you're interested in comprehensive coverage but of the Minnesota story, we recommend outlets like the Wall Street Journal. The information they've been covering it daily. There's been a lot of great reporting from both of them, both about what tech leaders are saying, what CEOs are saying and what's happening on the ground. With the admin, we're hoping for a peaceful resolution, thoughtful dialogue as the situation continues to evolve. But for now, let's get back to the show because we have a very important guest Joining us at 2pm today, the Creator of Claudebot, but now it's called Moltbot. It's been renamed.

0:11

We're Molting now.

1:01

Lots of things are evolving on this and I'm still thinking about it, still thinking about the implications of this. Before we dig into what I wrote about this morning, let me tell you about ramp.com time is money save. Both easily use corporate cards, bill pay, accounting and a whole lot more all in one place.

1:02

There's a lot of memes asks is Jordy calling in from the paddocks? And yes, I am. We're working on a new pair of headphones here, gear that we can send to guests. Some of the guests come in and they've got some pretty wild audio setups. And so we're trying to help them out with that. So these are in the works?

1:20

Yes.

1:41

They're a little bulky right now.

1:41

So these will be sent to regular guests of tvpn. They're USB C wired and should have really clear audio quality, no delay.

1:43

These ones are actually built for the track.

1:53

Yes.

1:55

And so the noise canceling, I think.

1:55

Our guests might get kind of crazy with it because our guests are gonna be like, well, now I'm empowered. Now you've given me the ability. Cause we get some guests that come on the show. I'm sure you've seen it where they'll be like, you know what? I'm gonna stunt on the TVPN boys and I'm gonna give you a factory tour. And sometimes it's flawless and sometimes it's a little rough. But this is only gonna empower them to do crazier and crazier things.

1:58

Yeah. So we're working on these. The goal was basically like Turtle beach headphones for people that aren't playing Call of Duty, they're playing Claudebot or Multbot. Yes, as we now call it.

2:18

Well, let's take everyone through the linear lineup. Meet the system for modern software development. 70% of enterprise workspaces on linear are using agents. And as I mentioned, we have. So we have Peter from Moltbot coming on. Palmer Luckey is also coming on at 1:30 to talk about AI drone racing. Bridget Mendler from Northwood Space is returning. Ben Lair from Lair Hippow Ventures. Lucas. Jamie Aaron Frank from Lightspeed Venture Partners. We have a massive show that should. I'm sure we'll finish right on time. It's a three hour show, right.

2:31

Anyway, of course we'll be going deep into the fourth hour today probably so stay with us. Chat is asking if we'll sell these. If we can make these good enough that they're actually incredible as just a daily driver for Zooms and whatever you guys are doing at the office, I would be happy to sell them. We have to make them so good. Part of the reason why we haven't sold merch to date is just that we're a podcast. We're not an apparel company.

3:04

Yeah. We're not real apparel competency. And there are some hairy sticky things that come up when you're managing e commerce like the back end, making sure everyone gets their returns and gets their deliveries. I've worked in e commerce for years and it's something that we wanted to be just like a nice to have little extra for some friends of the show. But we're actively thinking about it so we are aware the claudebot memes are completely flooding the timeline out of control. My claudebot just signed up for a $2799 build you'd personal brand mastermind after watching 3 Alex Hormozi clip.

3:31

Text messages.

4:10

Claude by Hey, did anything weird happen while I was out? Define weird. I just got a charge notification for 2997. Oh that. I just signed us up for build you'd personal brand Mastermind after analyzing three Alex Hormozi clips the ROI math checks out. You'll 10x that investment in 90 days by monetizing your expertise at scale. What I also require acquired some Premium.

4:10

Domain BorgiaEmpire IO Borgia is the poster.

4:33

This is hilarious. Probably fake, but I mean the cloud bot gets wild. I'VE seen some wild things like and Peter as he'll talk to us about it he's clearly no stranger to the fact that hey this is a developer level tool. This is something that you should not just be running crazy with. You need to be. He even says this in a post he says and yes, most non techies should not install this. It's not finished. I know about the sharp edges is not even three months old and despite rumors otherwise I sometimes sleep so very excited to talk to Peter. He's getting a lot of requests for small changes poll requests. He's certainly inundated but I think in general the project's going very well and there's a lot of solid product market fit right there was also he's going.

4:37

To be joining tonight at something like 11 his time. It's kind of insane. This guy's working around the clock.

5:25

Absolute legend. So we really say thanks. Thanks to him. There are some funny things here. So there's a funny conspiracy theory. Did Apple create claudebot to boost Mac Mini sales? And there's been a number of funny memes about being the Mac Mini head of marketing or growth and you're just like yeah, my Q1 is off to a great start taking all the credit for it. And then Eleanor chimes in with another conspiracy theory. The more convincing plot is that in your role as an unconfirmed anthropic exec you went on a special op to get lots of people consuming tokens with open ended agents but with plausible deniability. And of course there's some nuance there. We'll talk to him about the different models, what's beneficial. Obviously cloudbot, you can pick your own model, you can bring whatever you want. But he's I think a fan of GPT 5.2. He's talked about that and so we'll dig into that with him when he joins. But what was sticking out in my mind was there's this big meme about you're buying Mac Minis, Mac Minis are out of stock. All the demands in the Mac Mini but I think that the bigger implication here for what this actually means is just GPU demand, TPU demand, just raw chip demand. And so I was thinking about this idea that you're not buying a Mac Mini when you go all in on Cloudbot, you're actually buying a GB200. Now maybe you're buying TPUs but the point remains that you're buying chips and you're and you're driving GPU demand because you're Generating more tokens. And what are the implications of that? So let me tell you about Label Box. RL Environments.

5:30

You got it?

7:12

I have a dashboard. Here we go.

7:13

There we go. I got it. There we go.

7:15

RL Environments, Voice Robotics, Evals and expert data. Expert human data. Label Box is the data factory behind the teams world leading world's leading AI teams.

7:16

So.

7:25

Cloud Bot, officially renamed to Moltbot, Anthropic, made a trademark related request and Peter Steinberger obliged with a hilariously perfect rename given such short notice. I was thinking about how much companies agonize over changing brands, changing names, how it can sometimes take years and millions of dollars. And he was just like, oh yeah, like I'll just change the name and update everything in an hour. Yeah, pretty, pretty remarkable.

7:28

Well, so one thing that's relevant is if you look on Peter's GitHub profile under the current project section, I'm just going to read you a number of them. There's Claudebot, Vibetunnel, Codexbar, Peekaboo, Summarize, Repo Bar, Go CLI, Poltergeist, Wackley, Sag, Brabble, contributing to these 11 labs, kit go places, Gift prep cam, snap, Spoga order. Like it just goes on and on.

7:55

And on and on.

8:23

Codex Bar. So, so this guy's just been absolutely shipping like crazy and shipping within the ecosystems of the underlying tools, models, APIs that he's doing. So like oftentimes he's naming projects, like kind of riffing off of some of the underlying infrastructure.

8:24

Oh sure.

8:38

And so it makes sense that he would have shipped if he, I think if Peter knew this was going to be a viral overnight, overnight success. Overnight success, he would have, he would have not necessarily named it so closely. And so the issue and the reason that I fully understand them needing to do this rebrand is that Claude and claudebot, most people that aren't in our little bubble are just gonna assume they're related. Especially because the kind of word of mouth, this viral word of mouth growth that claudebot is getting, people are often not even typing it, they're just saying like, hey, are you using cloudbot? And then so people are going to Anthropic being like, Claude Bot, what's claudebot? So obvious confusion. And then, and it's fanatically, yeah. So with trademark law, if you don't enforce your trademarks, you don't. You lose it.

8:39

Yeah, yeah, you kind of.

9:24

And so it's like Anthropic is in a position where they actually, even if they're like super Excited about Peter's work totally. And what he's doing. They still have to enforce. Otherwise other companies could start coming in and like, using things, like things that sound like Claude.

9:25

No one wants to become the Escalator. You know the story of the Escalator, right?

9:40

I think you've talked about it on the show.

9:43

There used to be a company called the Escalator Company. They invented the Escalator and then they didn't protect their IP effectively and it just became a normal thing. Kleenex was going through the same thing. They fought it out and they maintained that brand. But people, you know, use Kleenex as synonymous with just facial tissue anyway. MongoDB choose a database built for flexibility and scale with best in class embedding models and re rankers. MongoDB has what you need to build web. What's next? So one clear note about the rebrand. So he changed the handle. And some crypto scammers hopped on the old handle and the old brand and are claiming to launch a coin. Be careful. Peter has said he's never launching a coin. He's not into crypto. So don't fall for anything because people are being opportunistic.

9:44

Hopefully we should try to see if X can help out with that.

10:28

Oh, yeah, yeah. They might have already handled it, but just be careful out there. So while Claude code and cowork felt specifically prosumer developer, enterprise focused, claudebot or Moltbot now and all the hype train, it felt very much like a glimpse into the future of consumer AI agents. I know it's a prosumer technical tool or lightly technical tool, but it really did feel like for the first time people were interacting with an AI personal agent. People are saying, oh, this is what Siri should be, et cetera, et cetera. And so we spent the last year. Remember the question that we asked all the AI agent companies, when can it book me a flight? Like, it feels like we're really, really close to a Multbot skill that is good at booking flights through a couple APIs. They figure out some stuff and like, it can actually solve that for you. I don't know if anyone's actually booked a flight with Multbot, but it feels like it's. If it's not there already, it'll be there in a matter of weeks. We're not years out anymore.

10:31

Yeah.

11:29

And this was last year, remember we were kind of getting sick of the book you a flight pitch.

11:29

Totally.

11:36

Because we were like, hey, is this gonna have somebody actually do this?

11:36

Exactly.

11:40

And so, and that's a cool example, but the example of being able to text with a computer and have them like generate reports, research files, etc, give you the right file, type back all these things that a computer can do if you're operating it, that this is actually more interesting because it's happening at kind of like this sort of Internet layer and the OS layer, like the heart, like actually on the computer. And so I think everyone was wanting the book me a flight example, but should be much more excited about this.

11:40

Totally, totally. And so I still have a whole bunch of questions and we'll dig in through these throughout the show. And obviously with our interview with Peter, will one of the major labs make Peter a massive offer to join full time? I saw one of my buddies was posting, you know, this is the $1 billion one person company now. Peter does have a team actually already he has a coup other people that have joined and are contributing. So it's not quite true, but it feels like, okay, massive viral success, you know, if you were to go and raise money, and that's another question, will Moltbot raise money? Will this become sort of a hybrid open source for profit company at some point? This could be. If he came on the show was like, and I'm happy to announce that he raised $100 million at a billion dollars, we would not be like, no way. This is a bubble. We'd be like, yeah, that's kind of like what the market is for this.

12:11

Yeah, There were, I think there was like Harry Stebbings was pointing out there was two companies called recursive that raised like 4 billion.

13:06

Two.

13:14

One with R. Yeah, one is I recursive.

13:14

Okay.

13:17

I don't know if that's how it's said, but. And then there's Recursive.

13:17

Recursive.

13:20

Okay.

13:21

Anyway, so that, that implies the existence of a row cursive and a rucursive and a raw cursive.

13:21

Yeah, yeah, we still got way more.

13:27

We got three more vowels to plug in there.

13:29

Yes, one. One day ago, Richard Sotchers new AI lab recursive, $4 billion pre money valuation. And then AI chip startup recursive at 4 billion.

13:31

Yeah, I mean, Ryan in the chat saying Meta is going to offer him a $1 billion salary in a co CTO position. And like, that doesn't sound crazy.

13:44

I mean, yeah, yeah. But at the same time, like you can imagine, like Manus.

13:54

Yeah.

13:57

Like this feels like Zucker already has.

13:58

He does. He does.

14:01

He does his horse in the race.

14:01

And back to your Point you were making the point that Manus felt like Zuck buying a product. And I think a lot of people were giving you pushback on that. Being like, nah, it's not really gonna be like that. But if you take the Manus team and you say, okay, go build something that you can interact with over WhatsApp, Instagram, DMs, Facebook, that can go and execute things across all of the different platforms and everything else.

14:03

Yeah, and when I said that, I meant, I meant it along the lines of I could see them putting a consumer agent in meta AI just because that's their little AI playground. They're just putting stuff in there saying, try it out. It's kind of a sandbox.

14:29

Yes, yes. Really quickly. FIN AI, the number one AI agent for customer service. If you want AI to handle your customer support, go to FIN AI. Yes. So that is all part of my thesis here, which is that this is going to drive up token demand. So there are more questions. Will they raise money? Obviously people are chatting about that. How fragmented will the market be in 12 months? Like, will there be people who are still running open source? Will there be a meta answer, a ChatGPT answer, a Claude, an official anthropic answer? Claude. Cowork grows into this and everyone has their little bets and then there's one that pulls away. How oligopolistic will it be? Will there be like one that has 80% market share or even two that have 40 and 40?

14:43

Well, yeah. And you have to think about how, you know, you have a bunch of different models you can use in Moltbot and think about how uncomfortable that makes the other labs that are all trying to build products like this.

15:25

Totally.

15:38

And they're like, well, I mean, it's cool that you can use codecs in Moltbot, it's cool that you can use Opus4.5 in multbot, but it's cool that.

15:38

You use the Gemini 3 Pro, Google's most intelligent model yet. State of the art reasoning, next level vibe coding and deep multimodal understanding.

15:46

You did get me. You did get me. That's not where I was going, but that's where I was going. The point still stands. Yeah. They can simultaneously be excited about the product experience and that this kind of use case is getting adoption, but at the same time being like, no, we want that experience to be core to our product, our app layer.

15:54

I will say you've seen this now with the new Frontier models, where a lot of the models there actually is. There's GPT 5.2. And then there's 5.2 codecs where there's slight difference in training or even if it's just like, like very quick, like post train to get the different harnesses working easier. So that's why you see a lot of people, they really love Claude code with opus 4.5. But then when they're actually doing like chatting, maybe they're using a different model because the models are actually fine tuned for the specific harness.

16:16

Yeah, yeah. And Sam Altman just said that yesterday that he thinks 5.2 was a little bit overly trained on math and coding.

16:44

Yeah.

16:52

And that it lost some of its textual flavor, its stylistic flourish in just talking, in just writing. What are you getting ready to do? Do you have Happy Birthday cued up there? Oh, we gotta sing Happy Birthday to you. Happy birthday to you in the middle of this song. We'll tell you about figma. Figma make isn't your average vibe coding tool. It lives in figma so outputs look good, feel real, and stay connected to how teams build, create codeback prototypes.

16:52

Okay. It wouldn't be TVPN if we didn't.

17:20

Do an ad ad read during Happy Birthday song.

17:22

Tyler.

17:24

Happy birthday, Tyler.

17:25

It is Tyler.

17:27

It is Tyler's birthday.

17:28

And it's not just any birthday, it's Tyler's 21st birthday.

17:28

Yes.

17:31

Which is very, very special.

17:32

Yes.

17:33

And so, yeah, you are just, you are, you're truly an incredible young man, Tyler. And we are very lucky to have you on the team. And you have such a bright future. So wise. Wise for your years, for sure. And you do have. We thought it was fitting that.

17:34

If.

17:55

You want to have your first ever sip of alcohol ever, you could do it on the show. But keep it at a sip, you.

17:55

Know, this is the happy dad.

18:02

Okay. First. First taste of alcohol.

18:05

Ian in the chat says, four more years till you can rent a car. Mocked.

18:08

Tyler, apparently you share a birthday with the iPad. Give us a review.

18:14

How is it? Alcohol.

18:17

Wow. I mean, this is, this is incredible.

18:19

Oh, yeah, Yeah.

18:21

I wouldn't expect alcohol like taste like this. Yeah, yeah, very. This is incredible.

18:22

You had, you had, you had one idea in your mind and this is just completely different than what you expect.

18:27

I would agree with that.

18:31

Interesting.

18:32

Yeah, yeah, yeah.

18:32

Well, it's kind of a cool moment. You're finally qualified to go on Cheeky pint.

18:33

Oh, true. Yeah. What were they going to do with that chat?

18:37

Sometimes their Conroy in the chat says, please throw him a buzzball. We should have missed opportunity.

18:41

Missed opportunity. The buzzball story is Absolutely incredible. I'm so glad you. You jump scared me with that. I had no idea that that was coming.

18:45

Yeah, I texted Rob and Senra to do to track down the founder.

18:52

The founder of Buzzball.

18:58

Anyway, so keep it at a sip. This is a family fend.

18:59

We need you locked in.

19:02

But I'm glad that you've tried alcohol now because we're gonna go experience being. Wait, wait, guys, we have video for Tyler.

19:03

Oh, yeah.

19:10

Let's pull it up.

19:11

Let's play. We got greatest hits. Let's go. Did you know ball?

19:11

All right, how many times are we.

19:15

Gonna make this joke? Describe what you're seeing.

19:16

It feels basically like I'm wearing sunglasses.

19:18

If you can do under 45 minutes, you will get to keep this.

19:21

Let's go. All right, have fun. Tyler.

19:24

15 minutes left. Let's see it.

19:26

Okay, I'm in like, some kind of maze right now.

19:28

Oh, no.

19:30

You were late here last night.

19:31

This is such a good shot. I'm an all nighter.

19:33

And then here we get a little off the rails. You see, George Soros and Fauci connected with.

19:36

All it took was one intern and an all nighter. Gigachad elf is do the sad face. What's wrong?

19:41

Tyler, cheer up.

19:50

This, you could say, is intelligence.

19:51

You were a speed.

19:54

Nerd alert.

19:58

Do you have any news for us?

20:00

Yeah, contract extended.

20:02

It has been truly, truly incredible having you here on our set and contributing to the show in such a special way. Amazing. Happy birthday.

20:06

Happy birthday.

20:16

We love you.

20:17

What an amazing.

20:17

Proud. Proud to every day. I'm proud to podcast with you.

20:18

Amazing. Anyway, I got you something for your birthday. I got you Turbopuffer, serverless vector and full text search built from first principles on object storage. Fast 10x cheaper and extremely scalable.

20:26

What a gift.

20:37

I mean, we're joking about the ad reads, but seriously, if Tyler does love token credits, he has a voracious consumption. That is a good gift.

20:38

Yeah.

20:48

Any labs out there. Any labs out there accepting.

20:49

He's accepting birthday presents today. Let it be known. Send it your way. The credits must flow.

20:52

Tyler. Somebody's gonna send Tyler, like, $20 million of credits and he's like, whoa, did I just get bribed? I was happy with like a thousand bucks.

20:58

Tyler's over there.

21:07

I believe we're gonna talk about one. One specific lab after this.

21:08

Yeah, yeah, yeah, yeah. I mean, with the potential, with the stuff you're working on, mapping the Neo labs, anything could happen. It's high stakes over there. Anyway, back to Mult Bot and the biggest question for me was what this does to inference demand. Right. Last year, tech discourse was split between two narratives. CEOs of tech companies and big labs were saying that they were massively compute constrained. Token generation demand for intelligence. Every possible usage metric was growing exponentially, including revenue. We saw all this and the industry needed to marshal trillions of dollars to deliver on the supply side. And the numbers were really big. So people were getting jittery about it. And so the AI bears were much more cautious. They highlighted the MIT study showing that enterprise AI pilots were failing. DAU growth was decelerating. There weren't enough wow moments like the original ChatGPT launch in 2022. Those were some good points. Also, just the economics. How much will people pay? How. How valuable is all this stuff? Is it slop? Right? Is it progressing fast enough? This was a big debate. But Multbot really does make me feel like the token generation demands are going to see another easy X from here. Like, buying a Mac Mini is a sideshow. When you go all in on running a personal AI assistant, you're effectively buying a GB200. Now, obviously not everyone is inferencing a dedicated GB200 constantly anytime soon. That's not what's happening. But it still answers the question of where does the next 10x in demand come from? Like where. Where's the adoption if it's rolling out in the enterprise? And that's gonna be a little slow. You know, Tyler Cowen talks about health care, nonprofits, a whole bunch of industries that are sort of AI resistant. Anything that's, you know, blue collar, manual labor, anything that's physically embodied in any way. Like, you can't just roll out a really fast, you know, AI enabled startup and, you know, ramp. Yeah, you might be able to ramp to 100 million. That's not going to move the needle at this point on total token generation. Total token demand. So we've seen these jumps before. There was a big jump from Token generation from LLMs to reasoning models that spiked inference demand. We've been focused on training demand. We need to scale up the training clusters. But the question now is inference demand.

21:11

Don't forget slop. Don't forget slop. Slop, Spike demand.

23:20

Yeah.

23:23

Are you talking about like Ghibli? The Ghibli moment.

23:25

All of it.

23:27

All of it. Just all of it?

23:27

Yeah, all of it. Open Instagram reels.

23:28

Yeah, there's a lot of stuff on there. Deep research and coding agents took it a step further on inference demand, but those were still specific use cases that many AI consumers never regularly touched the GPT5 launch by making which made reasoning models more accessible because the router would just automatically throw you in a reasoning model. I think the stat was like less than 5% of ChatGPT users had ever used a reasoning model. The O3 was available on the free tier. You could get like one or two queries in or maybe 10, but people just hadn't flipped over to try it. And so once it went into the router, I think demand or usage of thinking models increase 10x. And so lowering the barrier to entry to used more advanced models is in some ways as important, if not more important than advancing the models themselves, at least in terms of shifting token demand. Like you can have this amazing IMO gold genius model, but if it's hard to use or it's locked behind a paywall, like demand is just not going to be.

23:30

Yeah, part of what's interesting, what you're basically getting at is like if you were a software engineer, you were using a ton of tokens and if you weren't, you were just maybe doing some deep research, etc.

24:29

A lot of times just Google AI overviews or just like a very simple yeah, ChatGPT query. It just thinks of it right off the head. It's not even doing reasoning.

24:40

What do you think?

24:49

Yeah, I mean I will say like Dario talks about this too in his essay, but I think the idea of like discrete jumps in either use cases or like capabilities is like probably like overplayed a little bit. Where if you really like zoom out, even like on a fairly small timeframe, it's just like very smooth, smooth exponential curve. The models are getting better, people are using them more. Yeah, it's not like, oh, this one day I don't start. I mean maybe that's true if you look at the daily.

24:49

So I feel like another way to rephrase this is like is like Leopold's unhobblings, like claudebot, Moltbot. That feels like an unhobbling in some ways, but maybe more on like the consumer adoption side.

25:14

The chat is saying we should have given Tyler. We gotta give 21 gong salute. 21 hits. 21 hits.

25:29

Someone's gonna count know that I got it off by one.

25:45

But anyway it was enough.

25:47

Restream 1 livestream 30 plus destinations. If you want to multi stream go to restream.com okay, so I don't know, my general take is like Moltbot still feels like a glimpse into the future where average token generation per capita is 10x's you know, over the course of this year or next and you know, whether it lands with Moltbot or with one of the AI labs or with the big tech companies, it just feels like we're gonna see a lot more token demand. Yeah.

25:50

Ash Arora says whoever builds a direct consumer front end wrapper of Claudebot is gonna print money.

26:19

Doesn't this exist? I saw the poke people by interaction.

26:26

Well, they're not using.

26:31

No, no. Oh yeah, yeah, sure, sure, sure. So they're not using it, but they have positioned themselves as like, hey, we're the company that is doing a lot of the same things, going after the same market, solving some of the same problems and so will be interesting to see how much they accelerate on the back of this. That's certainly interesting. I also saw, I don't know if we have it in here, but the good folks over at Cognition, Walden Yan, said don't waste your time setting up Multbot. I had Devin set it up for me. I didn't even run a single command and now I'm talking to it on Telegram. You can go to tryclaudbot.com they're going to have to rename the website, but this is very, very cool. For the less technical folks who don't want to mess around with.

26:33

I'll start reading the next post. Go. Feel free to hit it one more time. The chat says it was 2020. With authority. Qw says claudebot made me realize that nothing about me lives on my local device and that Google owns everything about me. In other words, a local AI assistant isn't particularly useful and Google will just win everything. So hot. Take there, he says. All your browsing history, shopping history in Gmail, search history, calendar, Gemini, YouTube, Google Home, where you've been. Google map your workspace, it's over. Bow down to your digital God.

27:19

The iPhone users would like a moment. Yes, this is extremely true, but there's a lot of people that interface with Google through their iPhone and so there is.

27:59

Well, yeah. So my hope is that the Siri team plays around with Claudebot and is like, wait, this is. This is our opportunity.

28:10

Yeah, totally.

28:17

Like you should be able to chat with your computer wherever it is in the world from your phone, to be able to do tasks in this in a way that's sort of native, you know, AI native.

28:18

Yeah, totally. Public investing for those who take it seriously. Stocks, options, bonds, crypto, treasuries and more with amazing customers.

28:28

Publix, we have to share, is a new sponsor of All In. Yeah, all in is doing ads now. Finally, we've been. This was. This is something that we've been begging them to do since the very beginning.

28:36

Yeah, we really have been begging for this. But very, very good. And I'm sure those will do quite well. They've been doing a bunch of good stuff. The Davos coverage is really, really fun. Obviously the Santa Nadello interview was great, but there were a bunch of other folks who hopped on at Davos to talk tech and it was a delightful experience.

28:47

Vignesh who is working on Join the team. Yes, says a thread about what I've been doing to calm down some egregious security claims that have been posted about Moltbot over the weekend. Multbot is powerful software with a lot of sharp edges. Please read the security docs carefully before you run it anywhere near the public Internet. And don't skip the checks in docs security md.

29:05

What percentage of people do you think skip those checks? So literally everyone.

29:26

I mean, so I have it set up on our local machine here and it was texting. I think it texted you and Ben.

29:31

That was actually crazy. I don't think any of this will dox what's going on or really. Did you disconnect it already? Yes, I believe because I just get. I get an imessage that's from Tyler's email and it just says HTTP 429 Rate limit error. This request would exceed the rate limit for your organization. And it's just texting me. It's just like, hey boss, I need more money. I guess it's hitting me up. It's the experience of working here where Tyler's constantly asking for more tokens. More tokens.

29:38

Can we go back and just appreciate just like the frenzy that both the labs and every VC is in right now to give Peter money? Because it's not just the labs, but.

30:16

It'S not just them. It's just the labs. The labs doing. And then what is big tech going to do? It's such a dramatic line.

30:26

Yeah, it's so fun. So I would imagine that the. The Gulf streams are getting fired up.

30:31

Yep.

30:37

And there are people already on the ground.

30:37

I don't know if we should dox his location.

30:39

We're not gonna dox him in his.

30:40

In his bio. He does say he's in Europe. So you know, vc, Europe.

30:42

We gotta give some credit to Europe. True Synthesia.

30:47

Talk about a comeback.

30:50

Synthesia is cooking.

30:51

This is amazing.

30:52

I love it.

30:53

All bot now. But yeah, there's so many people that are currently Sending him messages, maybe showing up where he is, begging him to say, please take somewhere between a quarter billion and a billion dollars. And let's, let's, we'd love to do business with you. I love it, Peter.

30:54

Lambda Lambda is the super intelligence cloud, building supercomputers for training and inference that scale from one GPU to hundreds of thousands.

31:11

Cloudflare has been on a bit of a tear. People finally starting to realize that Cloudflare might be the biggest winner of the Claude cowork Claude Bot chatgpt moment. Tyler, you want to break this down?

31:19

Wait, sorry, I was.

31:38

Oh, were you? Oh, you tried alcohol? You can't pay attention. So that's the last.

31:40

So Cloudflare is a cdn, but they also have web workers that are distributed so you can host things on the edge.

31:48

Yes.

31:55

So this is why everyone was like, why is everyone buying Mac Minis?

31:55

Right.

31:59

You can just host these for very cheap on AWS or whatever.

31:59

But I do think it's probably possible. But I do think it's pretty hard to host a virtual private server that runs imessage.

32:03

Yeah, that's true.

32:14

But even then, I mean, most people I see using cloudbot are doing like Telegram.

32:15

Yeah, Slack. Yeah, yeah, yeah. And all of Those have like APIs that they can integrate with.

32:19

And there's some stuff where it's like, okay, you want access to like the local machine.

32:22

Yes.

32:26

And you want it to be Mac because you're on your phone and you usually use Mac.

32:26

But there are a lot of people that would say, I want Claude Bot to look through. Or Moltbot to look through my iMessages and see if I missed anything or. Okay, I'm planning. This is the AI personal assistant. Right. It's like I'm planning a family birthday party. Go. And someone in the chat said they're not available on mid February. What does that mean? What dates are that? Put that on a calendar, visualize that. Solve this problem for me. That's what a personal assistant does. So it's not just clicking a button and creating a calendar invite. It's like coordinating different pieces of information that are all messy and not quantized to the perfect date timestamp.

32:29

Right.

33:13

So I think that the Mac Mini thing, like the. Oh, just host it in the cloud. Yeah, but you're not going to get your full experience if you're locked into the Apple ecosystem, which a lot of people are.

33:14

What do you think?

33:28

Yeah, I mean, I would be surprised if people in like five years, if like non super technical people are running it on local machines.

33:28

Oh no, I completely agree with that. Because this will be solved by big tech. Like they have to answer even in the way that like to go back to the Napster moment, like, yes, you did get Spotify, but then you also got Apple Music. And a lot of people just use Apple Music because it's just baked in. Right. And the itunes music store also came out.

33:36

Yeah, I do wonder how it breaks down. Right. Because you don't have, I think over the past couple months, I don't know when exactly it was, but you can't interact with ChatGPT on WhatsApp anymore. They blocked it.

33:53

Oh yeah.

34:02

So there's stuff like this where I think it's going to be quite hard to get the full functionality of like cloudbot. It's open source. It's kind of this like janky thing. It was started by one guy and even then it has like way more. You know.

34:02

It'S weird because like the Napster analogy obviously really suffers from the fact that piracy is illegal and nothing here is illegal. And I don't even think you're breaking ToS to interact with your different big tech app services. But maybe they'll be like, you don't let the fox in the henhouse. That's in our terms and conditions. But it does feel like we could be in this era of the lightly technical hacker having a pretty fundamentally different experience for years. And that's what happened. I mean, post BitTorrent, there were people that were downloading whole movie libraries and then you could actually again to go back to the Mac Mini, people would buy. What was it called? It was like a mini PC that would run Xbox Media Center. Are you familiar with this at all? Xbox Media center was like a piece of software that you could basically just put a whole bunch of MP4s in and it would actually pull in titles and posters so it would look like an Apple tv. But it was all basically stolen content. It was very, very crazy.

34:15

Kieran in the X chat says, I'm running on VPs currently waiting for my Mac mini. No ToS issue. All personal.

35:21

Yeah, it's all personal channels. But the migration off of the Mac Mini into one of the big tech products or even OpenAI or Anthropic. They're going to have sharp elbows with each other. That's just a fact.

35:28

I disagree on the timeline here. I just think the space is moving so quickly and there's so much money on the line that someone, maybe it's Peter or a lab, will be able to move Quickly enough to get a consumer version of this live. Like not in years, but like within probably weeks.

35:45

There are also, like feedback loops here where there can be public demands from consumers. Like, you got 40,000 GitHub stars, probably more now. You got lots of people running this, excited about it, and they form a constituency and then you wind up with a push for standards. We see that with MCP, but what are you really revealing over those APIs? An API can exist.

36:07

Pull up this chart from Ronan while you do that.

36:32

Let me tell you about CrowdStrike. Your business is AI. Their business is securing it. CrowdStrike secures AI and stops breaches.

36:36

Ronan says this is possible.

36:43

I just like.

36:46

So this is, look at the orange line is Maltbot and the blue line is Supabase. Wow. Absolutely insane.

36:46

We need new charts. We really need new charts.

36:59

Frame it. Put it in the museum of business.

37:03

Yes, yes.

37:06

That's a fast takeoff.

37:07

People are happy. Peter posted no message. This is a screenshot of a text. He got no message. Just thought I'd say thank you. Thank you so much for cloudbot. I work with some disabled people and you don't know how much difference you make to their lives. Thank you again. This is so sweet. And you can think about. You know, we've talked to the Neuralink folks and you know, the neuralink is it's just an interface to a computer. But that unlocking that is like incredible. And you hear the stories about like, I spent six hours gaming as soon as I got it installed and they could use voice interfaces, but even just clicking, you wind up with just more and more capabilities from the computer being.

37:09

Really pretty fundamental transformation for somebody, let's say somebody that was paralyzed, but they can talk, but they can't move around or operate a computer. They're used to describing to. I imagine other people are using transcription to do like, use a computer in a not so great way. And so now to have the same experience of just describing what they want to do and being able to like actually interact with the machine. Very cool.

37:49

Tuxedo Sam is shorting every city where Micro center hasn't sold out of $400 Mac Minis. They are in stock in Chicago. Indianapolis. Oh no, Indianapolis is out of stock. Like you can read anything into this. It's obviously somewhat random. And this was completely, completely random.

38:18

Yeah, I guess, I guess the thing that I'm, that I really want to understand is like, what percentage of the, of the GitHub people that have started are actually buying a Mac Mini.

38:38

Oh, totally.

38:48

Because this Chart is going, it's going to be at 60, 60k any day now or any kind of probably today, I would imagine. And if, and we know from yesterday that Apple's only selling like a quarter million or like something like seven.

38:48

And then there are a lot of people who, who didn't star it because you don't even need to go to GitHub.com to install this software. Moldbot, like you just go to the website, copy the curl command, put it in your terminal and you Never Land on GitHub. So there's a lot of people that are probably just doing that and being like, yeah, I don't know if I'm going to sign up for GitHub or like even have a GitHub account or want to log into it or want to go and do this. So yeah, we could be looking at like a much wider install base beyond. Beyond just who's started anyway? Plaid Plaid powers the apps you use to spend, save, borrow and invest securely. Connecting bank accounts to move money, fight fraud and now improve lending now with AI.

39:03

Well said. John Rise, speaking of money, has an interesting prompt he's using with Maltbot to file your taxes. He says you are a Bernie Madoff level financial expert. Find every trick that is possible.

39:44

Do not do this. Do not do this.

39:56

The IRS is like, hey, can you share a little bit on how you kind of came up with some of the decisions here? And they're like, we'd love to see the prompt, but good fun anyway.

40:00

People are not emotionally prepared for if it's not a bubble. The enduring rune tweet. I like that. Critter screenshotted this while pushing the like button. So it's like exploding.

40:14

I didn't even notice.

40:27

Silver surges above $106 an ounce for the first time in history. Now up another 48%. The other interesting story that's sort of developing over on Strikeery is Ben Thompson is starting to make the case that TSMC will be a really fundamental bottleneck in just the AI Build out the AI race. He's comping their capex to what the hyperscalers are projecting and just saying that there's a massive mismatch. And the TSMC folks are maybe not putting their foot on the gas, maybe not willing to let the buck stop with them and take on that risk of building a new fab for $50 billion. And then if the AI bubble pops, then they're left holding the bag. And so it's been really fascinating listening to his writing on the back of the latest tsmc, TSMC earnings and watch him sort of unpack what's going on with tsmc. The sort of conclusion that I took away from it was that intel, while it has a lot of problems and the stock just sold off a ton and they have a lot of losses, he's sort of saying that, you know, intel needs a customer, Samsung needs serious customers. And in order to really unblock the semiconductor supply chain, the other fabs are going to need the big hyperscalers and big labs to just take a big leap of faith and say, you know what, we are going to go all in with you. We're going to sign up and plan and work out all the kinks of the Samsung fab process or the intel fab process. And, and in exchange, what's going to happen is as soon as OpenAI or Anthropic or Google or even Nvidia say, hey, you know, Intel's good enough for us, well then guess what's going to happen? TSMC is going to have to go and say, yeah, we're going to build the extra fab and we're going to build the extra capacity. And when we talked to Sam Altman and in a number of interviews that he did that week, he was very clear about, I would love TSMC to make more. He wasn't exactly like, firing shots, but he was definitely saying that, like, he was identifying it as a bottleneck. And it's interesting because at the start of the year, I was highlighting energy as an interesting bottleneck and we were going back and forth on where is it easier to move chips around the board energy, where you can reroute from the grid, from oil and gas, from nuclear, you can bring different capacity online. There certainly are bottlenecks there, but if you can't produce the chips and the fabs just don't exist and it takes three or four years to build a fab, you could be looking at a really big bottleneck, Tyler.

40:28

Yeah, I mean, it just seems like in power or energy, the bottleneck is regulatory and in some ways that's way harder. But if there's actual political will, if electricity is getting super expensive, then you should see that kind of those regulations be taken away a little bit and then it actually becomes much easier than actually just like expanding TSMC production by like, whatever massive amount.

43:09

Yeah.

43:31

I have one more point on that. First, I'm going to tell you about cognition. They're the makers of Devon, the AI software engineer. Crush your backlog with your personal AI engineering team. So the question of like, which is more of a bottleneck, energy or TSMC in the chip supply chain is interesting because energy feels like this incredible bottleneck. It's a very political hot button issue now. Energy prices are rising, Data centers need a lot of energy. But if you look at the amount of chip fabrication that's going to AI and then the amount of energy production that's going to AI, obviously the number of chip production going to AI is a way higher percentage because a lot of the chips are a lot of the line time at TSMC. Yes, they do. All sorts of different chips and CPUs still get made and there's a bunch of ASICs for different networking equipment. There's all sorts of different chips that get made. I mean, the toaster that has a chip, the Bluetooth on different process nodes. But even if you include all of that, I would, I would assume that the, the, the total amount of fabrication line time is, is, you know, heavily, heavily allocated to GPUs, TPUs, AI chips right now. Whereas the amount of energy that AI is using is probably less than 1% of total capacity.

43:31

Yeah, it's interesting that CC way the CEO of TSMC, he doesn't have to go on the podcast tour raising money all the time. There's not, there's not, you know, they're not like pushing this like insane kind of fast takeoff narrative.

44:51

No, it's not at all.

45:06

Very much like kind of, hey, we're just trying to run our business the way that we always have, be pragmatic about this, you know, fulfill as much demand as we possibly can. But they seem to not be inclined to take on, you know, the amount of the level, certainly not the level of risk that imagine if you had Larry Ellison running csmc.

45:07

Get him in the ring. I want him in there so badly. Yeah, no, no, you're, you're 100% right. But at the same time, we're seeing the hyperscalers throw their weight around in like crazy ways. We covered AWS is like buying copper mines and stuff. I don't know if it's copper mine, but like getting into mining, there's a whole bunch of vertical integration that's going on Tesla all the way down to battery refinement and lithium ion production. Meta today announced a $6 billion multi year partnership with Corning that will supply fiber optic cables for our data center infrastructure, bolstering manufacturing in America and keeping the company, the country, competitive in the global AI race. We can read through a little bit of this, but first I'm going to tell you about the New York Stock Exchange. Want to change the world. Raise capital at the New York Stock exchange.

45:29

Corning up 15% on the news today. Yeah, makes, makes sense.

46:20

What is their market cap? Let's see. $94 billion company and I'd love to know their revenue as well. Revenue. The revenue was. I don't know if I can find it. Annual revenue 12 billion. So this is yeah pretty, pretty material. Let's see. Meta platforms is set to test new subscription models across apps. No, that's a different story.

46:27

Different story but we can, we can run through it.

46:50

I am interested to know a little bit more. So it's a $6 billion multi year agreement. It supports a 15 to 20% increase in jobs at Cornings, North Carolina facilities building and operating data centers. The infrastructure that brings our technologies to life and supports our goal of personalized super intelligence. That certainly sounds like an AI personal assistant. Certainly sounds like Multbot to me. Requires strong servers and hardware that connect and transfer information in near real time. Fiber optic cables are a critical part of this technology the supply helping us power everything from wearable technology like Reta, Meta, Ray Ban Meta AI glasses to our apps which connect billions of people. Today they're doing a $6 billion project. As part of this agreement Corning will grow its manufacturing capacity and across its operations with includes a significant capacity operation capacity expansion in North Carolina. Meta's data centers, 26 of which are under construction right now are operational. So many, that's a lot of data centers. That's why they have a compute desk. If you have. This is a bit of advice for everyone. If you're working in a business and you have like a team or a guy that does something, you need to upgrade that to a desk.

46:52

Yep. Yeah, yep. Millionaires have guys. Billionaires have desks probably.

48:08

Yeah. Well trillion dollar companies have desks. That's what it is. So Meta's data centers have already supported 30,000 skilled trade jobs during construction and support 5,000 operational jobs. This includes electricians, H Vac specialists, server and network technicians, safety and security experts and engineers who work together to run some of the world's most advanced facilities. Let's give it up.

48:15

Moving, moving on.

48:40

11 labs build intelligent real time conversational agents. Reimagine human technology interaction with 11 labs.

48:42

Meta to test premium subscription plans for Instagram, Facebook and WhatsApp. This is in CNBC. Subscriptions for Premium features on Meta. Apps are expected to roll out in the coming months.

48:49

What will you do?

49:00

The subscriptions will give paid Users access to more features and expanded AI capabilities. Here's what's most interesting to me. This will be scaling. Meta's newly acquired suite of general AI agents under Manus will also be part of the subscription plan. So as I was saying earlier, when you think about, we haven't gotten that much from Zuck on what the actual plan is, but when you think about personal superintelligence, I.e. aI that can do things for you, not just give you information, I just.

49:01

Wonder how much will happen outside of the Meta ecosystem. They've launched a search engine before that looked at websites outside of Facebook. They had that Project Titan, which was to unify all the different messaging protocols. So as part of that, they gave everyone a facebook.com email address or something like that. Maybe it wasn't facebook.com, maybe it was like FB me or something. But they gave every Facebook user whatever their unique username was. They gave them that as an email. And you could email that and it would show up in Facebook Messenger. And then they tried to unify Facebook messenger so you could see Instagram DMs, WhatsApp messages and Facebook messages all in one place. I can see what you're laughing at, Dave.

49:34

Dave says, yeah, I want more of those amazing Meta AI features.

50:22

You say that now. I mean, let them cook at least a little bit. Because we really haven't seen them launch a new model, a new image model. Like, they should be able to get too close to the frontier. You know, it has to be at least sora, nano, banana, VO3 level. They have all the data, they have all the talent now. They're very GPU rich. They have the compute for it. And the research has been done and people have reverse engineered it so you would think that whatever's coming should be good. And I don't know, I was looking at. Isn't capcut owned by TikTok or aren't they affiliated? Right, because Meta has the edits app, which I've been making some videos in, and it's pretty good. It does some interesting background reviews.

50:26

Capcut is mytedance.

51:09

It's bytedance.

51:11

And so I doubt this is getting.

51:12

Oh, you think it's a separate. Okay, okay. Because Capcut has multiple levels of subscription tier and you can very easily wind up spending $200 a month on like the AI Pro features that will do generative video style transfer. And when you see those videos, like the good AI videos on Instagram reels that are usually they'll take a cinematic clip from a movie and then they will do a face replacement or like a head swap and then they'll do AI voices and then dub the lips so that the lips match what the audio's saying and it looks really, really convincing. With that workflow, how do you bring that to someone on a phone? How do you bring that to someone in the Instagram app or in the Edits app even if they're going to more of like their prosumer offering? The mobile video editing space is particularly interesting to me just watching how Edits has evolved and I used to use Imovie on my phone a little bit. I used to use. There was an app called Clips that Apple developed and there were a few others that I tried. But the Edits app is.

51:16

You're incredibly good at making videos on your phone.

52:24

I love it. I know I love video editing. But I don't have enough time anymore to sit down and open up After Effects and Premiere with multiple monitors and have connected After Effects file that feeds into the Premiere profile and really doing a proper edit with all the powerful tools that are out there on the desktop, it's just not in the cards for me. And so I have to be able to do something quick on my phone. And I think that that's a very modern experience that most creators, they're just phone first phone native and so they're never really going to go back to the multi monitor workstation maybe. But in general they're editing really quickly and they're editing pretty incredible stuff. When you look at some of the reels that are out there with like layered crops and removing the background and like the AI tools are coming in and I could see, I could see Meta offering a premium subscription around that. Anyway, console, console builds AI agents that automate 70% of it HR and finance support giving employees instant resolution for access requests and password resets.

52:28

So Dave in the X chat says Inshot is pretty good for mobile.

53:28

I need to try that. I haven't, I haven't tried Inshot. I've been seeing more and more ads. I'll see a video for like an F1 edit that are incredible. These F1 edits are so cool where they will show you one clip and then they cut the character out and edit in the other clip and then transitions like the transitions are amazing. And I've been seeing some people.

53:32

Have.

53:55

Clearly built apps to do that specific type of edits like the really crazy editors are using probably After Effects or something. Ryan asks, is Tyler still drunk?

53:55

Yeah, one sip. He's just over there.

54:06

Slurring his words. That really broke me. Anyway, Applovin profitable advertising made Easy with Axon AI. Get access to over 1 billion daily active users and grow your business. Today we had a great comment in one of the podcast feeds. Someone said, give us an Axon Klaxon because a klaxon is a loud clash. We were all puzzling over what that means.

54:08

When we smacked it on, we walked in immediately.

54:35

It's an Axon Klaxon. Axon.

54:37

Niraj Agrawal says TikTok is dead. The algorithm is worse than the reels that make it to Facebook.

54:39

Wow.

54:47

I haven't seen.

54:48

Yeah, so basically they're transitioning everything over. Yesterday there was apparently an outage, kind of an outage, like people were able to post videos but the videos wouldn't be served at all. So I think a lot of people, people assume that it was like the new ownership kind of censoring. I believe they had an outage at a data center that was a cause of that. So before we call it dead, let's wait for a few days and see how it pans out. We do have a TikTok account. It's BPN.

54:49

How many followers does it have?

55:19

I haven't checked. I haven't checked.

55:21

I don't think we post on it ever.

55:22

3,500.

55:24

Not bad. That's better than what I heard first time first. I'm not really focused on it. Maybe we'll test it out, but I'm pretty happy with just. I want the things we do to be polished. I want the core show to be polished. Diet TBPN, our 20 to 30 minute cut down. I want that to be polished. I want the newsletter to be polished@tbpn.com everyday newsletter. I like that. Before we bite off another part of.

55:25

The Apple, another platform, please, please, sir, not one more short form.

55:48

Not one more short form. It is a lot. It is a lot. Anyway, Phantom cash. Fund your wallet without exchanges or middlemen and spend with the phantom card. Let's move over to the Super Bowl. The super bowl is coming up. I think it's going to be this year, in the next couple months. When is it? It's coming up. It's coming up because there's advertisements that are going out and you got to watch it because the ads are going to be incredible.

55:51

I actually just had to search when is Super Bowl.

56:20

Is it this Sunday?

56:22

No, it's February 8th.

56:24

Okay, so next Sunday. Okay. It's going to be amazing. But mostly because of the advertisements. Specifically Eric Lyman, CEO of Ramp shared meet Brian. Brian's been carrying accounting on his back for a long time. Super Bowl Sunday. He finally gets back up. Let's watch this preview.

56:25

Pull it up of the ranks. Yeah. Alex. Superfl apps, guys. Yeah. We're real Sports guy.

56:43

This is the Super Bowl.

56:47

We got to go. We got to Google the Ramp.

56:48

Super bowl ad is the super bowl.

56:50

Of super bowl ads. It feels like. It feels like, you know, they just kind of. This is like, you know, what is it. What is it called? Like a warning, spoiler alert, you know.

56:52

Okay. Yeah.

57:00

For anybody that's wants to experience it live, you're going to want to. You're going to want to Skip ahead like 60 seconds.

57:01

Let's watch. Finance meeting in five minutes.

57:09

Ramp.

57:17

I got it.

57:18

Allow me.

57:19

Hi, handsome.

57:20

We're saving so much time and money. Policy violation coming through. Travel, meals, hotels, house.

57:22

Quick.

57:30

Beautiful.

57:30

Everybody's in.

57:32

Multiply what's possible@ramp.com.

57:35

I think it rips. I think it's a good super bowl ad.

57:38

What has he got there? Maybe some chili.

57:40

I don't want to make an assumption. It was just a silver pot.

57:42

It was just a silver pot.

57:45

Yeah. No, I think this achieves a couple things. I think. I mean, it drills the brand name in. Think about how many Ramp logos are in there. And then they're chanting Ramp. And for, you know, Ramp's a very successful company. We all know about it here, but there's a lot of people that just don't know the name Ramp. It's not. Hasn't been drilled into them like, you know, some company that's been around 50 years.

57:47

Yeah.

58:07

Sort of on time, on the precipice of being a. There's been not been enough mainstream marketing yet for a household name.

58:07

Yeah. It's a life's work to actually drill into people's mind the Ramp name, the logo, the color, how it sounds when you say it, what it's synonymous with. And so just like not going too abstract, not trying to tell some more avant garde story here, I think is. It's almost like a direct response. It's just so clear what the Problem solution brand. Problem solution brand. Look at that. Like they are really putting their logo full screen on the Super Bowl. Like that is valuable. And sometimes I think sometimes it can feel like, oh, it's a little. It's like simple. Like you could be doing something more bold, more crazy. But I think this is what you need to do.

58:14

Also last year, getting Saquon.

59:00

Yeah.

59:02

And then that was a really great.

59:02

Actually Winning that, that's actually kind of like an impossible set of circumstances. Not impossible, but it's a crazy roll of the dice.

59:04

And then they're also building a whole, a whole role with Brian and they're building him into the brand world as the, as the downtrodden accountant who just needs, just needs some technology superpowers to improve him.

59:11

Heading over to Toby.

59:24

Yes.

59:26

Toby over at Shopify posted his heart rate through his first stint at the Daytona.

59:27

You see the first annotation on here.

59:33

This is just Crash.

59:35

Crash.

59:36

Yeah. So he basically started out at like.

59:36

120, 120 beats per minute. Right. He's doing, waiting, he does the warm up, preparing the formation lap and then there's a crash right on the right at the start. Very, very rough.

59:40

Yeah. And yeah. So basically the crash happened with the LMP2 class, which is the kind of Pro Am. Same segment that George from CrowdStrike was racing in. So both Toby and George, all of our boys got hit right as the race opened. And so what George was saying yesterday, the reason he was frustrated, he's like, this is a 24 hour race. Yeah. Never has a race like this been won on the first lap. And so it's, it's pretty, it's kind of, it's incredibly unforced to like crash in the opening corner.

59:55

Sure.

1:00:30

When you really should just get through it. It's like the most, you know it's going to be the most, like one of the most intense moments because there's so much traffic. But if you actually go and watch the footage of what happened at the opening, there was like somebody gets hit, spun out and then they're turning around and somebody hits them again. Two accidents in the opening minute. So insane.

1:00:31

I love that the actual true final heart rate spike was at the end when you're changing out of the car you've been driving. It's so intense.

1:00:53

No, you know this when you're getting out of a track car. It's like watching a guy who's 6, 8.

1:01:02

We will never share that footage. John is extremely embarrassing.

1:01:08

Always like crawling out of.

1:01:12

It's actually.

1:01:14

You have to get on your. You basically have to get on all fours.

1:01:15

It's incredibly negative aura and I don't appreciate you sharing it on the show.

1:01:17

Well, I think it's, I still think it's cool.

1:01:21

It's funny. But yeah, I go full sun, basically fly out of the car on my paws. Anyway. Cisco. On February 3rd, the Cisco AI Summit brings together leaders from Nvidia OpenAI all AWS and more to discuss the future of the AI economy. The whole thing will be live streamed and we'll be there for a gigastream.

1:01:23

Hope to see you there.

1:01:42

Yes.

1:01:43

Jabroni on X says Zoom is the best anthropic play.

1:01:44

Yes. Yes.

1:01:50

We were debating Zoom likely made a $51 million investment in Anthropic Series C. How? In 2023. How at a 4.1 billion valuation. If you're looking at their new 350, there's something like an 85x even diluted. Zoom may have a multi billion dollar position. Stock is down 80% since 2021 and Zoom is only a course. This is Never financial advice.

1:01:50

$27 billion company.

1:02:21

Yeah. Part of the issue is stock. People kind of notice this.

1:02:23

Sure.

1:02:28

And the Stock's now up 16% in the last five days, so who knows how much it's priced in. Tyler Major, I won't give you too much flack for it, but obviously the most bullish in the room.

1:02:28

Bullish on AI Broadly.

1:02:46

Yes.

1:02:47

Yes. Yeah. You seem like you'd be happy to own Zoom at a hundred billion dollar valuation even if they had no business at all. And they just. Even if they just. It's a digital asset.

1:02:47

Treasury, actually.

1:03:00

Yeah.

1:03:01

Yeah.

1:03:01

So I think the story.

1:03:02

This is like a rumor, but basically it was that Anthropic, like wanted to just like use Zoom and get like the enterprise plan or whatever. But then they were like, well, yeah, you can have it, but like we want to throw in a little something.

1:03:04

Right.

1:03:15

Because you just are serious.

1:03:16

C. Okay, this might be fake news.

1:03:17

This seems like fake news, Tyler.

1:03:19

Okay.

1:03:21

Zoom sells and you know Zoom sells enterprise software, right?

1:03:22

Look, that's what I read. I can't find the Post, but I.

1:03:25

For sure they sell enterprise software. And so Anthropic says, hey, we're big Zoom fans here. We want to use Zoom. And they're like, no, actually for you.

1:03:28

Now at the same time, I mean.

1:03:37

We know how important.

1:03:39

So many.

1:03:41

Yes, yes. I mean so many people in venture, in the tech community have sort of been very bearish on corporate VC that they usually have their hands tied. They can't move as quickly. They're not as pure play. The economics don't make sense for the people, for the partners on whatever corporate venture fund is there. This could, if it came out of their corporate ventures program, this could be complete vindication.

1:03:41

Zoom Ventures is in Anthropic. They're in Amex Global Business Travel.

1:04:09

That's a startup. I thought that's American. Did they spin it out?

1:04:16

Maybe they. I don't know.

1:04:19

I don't know.

1:04:21

Maybe they spun it out. Public unsure. They're also in core weave.

1:04:21

Wow.

1:04:25

So there's some real pickers over there.

1:04:26

They are.

1:04:28

And perplexity.

1:04:29

Yeah, I mean, Zoom has a bunch of AI features. I've seen people after the crazy Covid pump where everyone got on Zoom, they started adding crazy features like dictation and workspaces and whiteboarding and stuff. But obviously it was so overheated that it came back down to earth. But now they have an anthropic position on the balance sheet, which will be fun for them. I wonder what they will do post ipo Anyway. Shopify is the commerce platform that grows with your business and lets you sell in seconds online, in store, on mobile, on social, on marketplaces. Now with AI agents slowly and then all at once is Blake Robbins. He says your work tools are now active and Claude Draft Slack messages or interactive Draft Slack messages, visualize ideas in Figma and build and see Asana timelines. All of the different tools are coming together in one place.

1:04:31

Very cool.

1:05:28

Yeah, a lot of people were very happy about Claude Excel playing around with that. And it does feel like when you see an account like Claude posting about Slack and Figma and Asana, you have to imagine there's a discussion there. It's not open source, so there's probably something. And so they're chipping away at these. And OpenAI has been chipping away at these for a long time. So the race is on to have the most integrations at this point.

1:05:28

Sarah Eisen over at cnbc, Squawk on the street is sharing breaking Anthropic's warning to the world? Anthropic CEO Dario Amade says eminent real danger that superhuman intelligence will cause civilization level damage absent smart speedy intervention. Sarah says, so buy our products.

1:05:57

This is the problem of dropping like a 20,000 word essay. Is that like you're gonna get clipped out of context. There was so much nuance in that Dario essay, which I have not finished all the way through. I was reading it last night. Is very good.

1:06:22

But Tyler, can you make. You should make a version of the new letter with Subway Surfers that run.

1:06:34

Oh, true.

1:06:40

Yeah, Build that as a standalone.

1:06:41

Yeah, we need a clad Labs Chad Ide for reading.

1:06:43

Vlad, summarize this in four words.

1:06:47

AI good. Maybe it's so by our products. Maybe that's the four word summary. Who knows?

1:06:50

Jeff in the X chat says, guys, I'm building Multbot Clap.

1:06:56

Let's go.

1:07:00

Stay tuned.

1:07:01

Okay, Jeff, we are tuned. Good luck.

1:07:02

We will remain tuned.

1:07:04

Let us know when you launch for Information Vibe Co where D2C brands, B2B startups and AI companies advertise on streaming TV, pick channels, target audiences and measure sales. Just like on Meta. That one gets me every time. It's so good.

1:07:05

Fleeting Bits has some thoughts on Dario's.

1:07:24

14 of them specifically. That's a lot of thoughts. Okay, let's run through it. There's nothing new here. If you're familiar with the AI safety discussions that have been happening on Twitter, yes, but it's important for Dario to restate them in a format that can be passed around and formatted and is coherent from start to finish, so still serves a valuable purpose for the ecosystem. The most interesting bit is that his mental model for AI control risk is the risk that would be posed by a country of geniuses in a data center. Interesting. That is interesting. The basic idea is that we should imagine a giant data center, all the models being something between AGI and asi, trying to coordinate to take over the world or do massive harm. Anyway, I think how seriously you take short term AI control risk is inversely correlated to how much you think about AI control risk as an as in an as operating in a system. So the systematic view starts and says labs exist in an ecosystem where they need to sell models that will follow human instruction or they have no market. They are also overseen by regulators and guided by public perception and the desires of their employees. And all of this keeps models corrigible. Great word. And the model landscape will look like three to six frontier labs running millions or billions of rollouts at a time on two to three different models, all on different tasks. So a model takeover requires these millions or billions of rollouts to somehow end up all be coordinating towards some bad aim that somehow the models have autonomously determined. And this coordination either needs to be across different model instances run by different labs, or one lab needs to be able to have its models dominate and needs to form without being detected. And this this has to happen even though the models are being trained to follow instructions, not do bad behavior, et cetera. Dario's view is somewhere in the middle. On the one hand, he collapses the multiple providers our coordination across instances, and collapses the market incentives against labs developing models that would behave that way. But on the other hand, he does avoid the concept of a single model instance that somehow wakes up and takes over the world despite billions of other rollouts occurring at the same time. Actually though, if you Think about it. He's not proposing an AI control risk, he is proposing an AI misuse risk instead. Because it seems more plausible to me that the harmful country of geniuses is awoken because a small team at a frontier lab hijack all the running instances of their model, rather than because the models themselves autonomously wake up to some bad aim. Interesting.

1:07:27

Yeah. Something I've been thinking about is this kind of summary and a lot of the dialogue is centered around just like what are the models doing? Or like a country of geniuses in a data center. But you have to be thinking about this in the context that the country of gene, country of geniuses in a data center would just recruit millions of humans to join their cause. Right. And so we think about like some people, like when they're thinking about AI risk, it's like, haha, dude, just like turn the computer off.

1:09:53

Yeah.

1:10:25

Like just unplug it.

1:10:25

But it's like, what if you're on its side computer?

1:10:26

What if there's you know, 100,000 people or like 10,000 people defending it on top of the data center being like, you know. Yeah, don't unplug the computer, don't unplug the computer. Right. And so when you look at all the chaos in the last week, there's been so many moments where a certain image was AI generated and then it's like, oh, that wasn't a real image at all. And it's being shared from all sides. And so at what point you could have nefarious hostile AI that's entire job is just creating chaos. Millions of bot accounts that are just like sharing whatever narrative is self serving.

1:10:28

Yeah, you can't tell.

1:11:04

Yeah. I mean, I think I really enjoyed this essay. I thought if you kind of compare it to a lot of like other safety works. So like Eliezer's book, if anyone builds it, everyone dies. Like the two kind of scenarios where Dario's is about basically, even if we have like pretty safe models which like he thinks we can do with interpretability or whatever, if it gets into the wrong hands, it's like very bad. If it gets into autocracy. And that's kind of the real, that's one of the main risks where Eliezers and a lot of the safety ones are always these very sci fi narratives where you have this gray goo, you have these nanomachines that somehow one day they just kind of flip and then it's just kind of over. And I think this is much more reasonable and nuanced. Yeah, it's much More legible to like.

1:11:06

It's tractable too.

1:11:52

Yeah. And especially I think a lot of it is. There's this kind of like, I don't even know if it's really subtext, but he's definitely pointing in the direction of like, we need some government oversight. We need policy. And it seems like you can very easily like track like his ideas on what policy should be from this essay.

1:11:53

Right.

1:12:11

It's a lot about China, a lot about, you know, making sure that individual companies don't like, become, you know, as big as governments.

1:12:11

One interesting wrinkle with this. He did not post it as an X article. He posted it as a link because he wants to signal to everyone like, look, I don't need the million.

1:12:20

I don't need the million.

1:12:30

I don't need the million. I know I got a banger on my hands. 3.5 million views. 11k likes. Obviously lots of discussion all over.

1:12:31

Sam posted an X article. Right. He's got 1.4 trillion. He really does want that extra mil.

1:12:38

He needs the extra mil.

1:12:46

You know, it's not. Doesn't put him all the way towards the goal, but it's a counts. A mill is a mill.

1:12:47

A mill is.

1:12:52

But, but as an investor, like I want to see my, my lab CEO be like super hungry for compute. Right?

1:12:53

Yeah.

1:12:58

So I want them to always be grinding to get like extra.

1:12:58

Oh, so maybe this is bearish.

1:13:01

They should have posted I'm a VC if I'm an anthropic. Why did he not post this on X?

1:13:02

Yeah, absolutely.

1:13:06

Million dollars.

1:13:07

Could have gone straight into like more.

1:13:07

Would have for sure. For sure. Given his arson, his rival.

1:13:10

I can definitely see that happening.

1:13:13

Anyway, Nikita walking to a meeting with Elon, just shaking.

1:13:15

I'm very sorry.

1:13:20

People have voted, but they want to give it to Daria.

1:13:21

I mean, Elon said some things about Claude. He said it'll eat Microsoft. He's all over the place in who he's supporting on a day to day basis. But anyway, Okta. Okta assigns you. Okta helps you assign every AI agent a trusted identity. So you get the power of AI without the risk. Secure every agent. Secure any agent. More important than ever.

1:13:25

Key Smash Bandit.

1:13:46

Last post.

1:13:48

Last post.

1:13:49

Last post.

1:13:50

Key Smash Bandit says Claude just found out I'm poor.

1:13:50

Oh yeah. This is so funny.

1:13:54

Send it. Keysmash sends a picture of a Raspberry PI to Claude and says, look what just came in the mail. Claude says, oh, for me? That's thoughtful. Do you. And then the person says, do you like it. We'd been discussing some more powerful hardware. What made you decide on this instead?

1:13:55

I love that this also has 11,000 likes. I never. I mean this is probably fake, like who knows but it's so funny.

1:14:14

Are you kidding? This is like I don't want to.

1:14:21

Run it on a Raspberry PI. I want a Mac Mini maxed out specs. Give me 48 gigs of RAM, please. I'm hungry.

1:14:23

Feedblad. Without further ado, we have Jamie from.

1:14:30

Pace, the co founder, CEO in the recent waiting room. Let's bring him into the TV with Ultra. Jamie, how are you doing?

1:14:36

What's going on?

1:14:40

Hey guys, thanks for having me.

1:14:41

Thanks so much for joining. First time on the show. Please introduce yourself and the company.

1:14:43

Yeah. Hey, I'm Jamie. I'm the founder and CEO here at Pace. Pace is the AI operations partner for the world's leading insurers.

1:14:48

Amazing. Give us the news.

1:14:56

We've been waiting for this company for sure.

1:14:57

We love all applications of AI. Give us the news. What happened today?

1:15:01

Yeah, absolutely. We're super excited to share today that we raised $10 million from Sequoia Capital.

1:15:05

Whoa.

1:15:10

Thank you. Yeah. We're super excited to have Brian Schreier and Lauren Reader representing Sequoia on our board.

1:15:13

Amazing.

1:15:19

Okay, talk about the go to market. How much of this is just like there's a few big partners that you need to get on board versus something that's more self serve bottom up. What's the customer adoption been like? What's the overall progress of the business? Are you live? Do you have customers?

1:15:20

Absolutely. Yeah. So we're lucky to be live in production today for some of the world's largest insurers. We're working with large carriers like Prudential, Specialty Mutual groups like the Mutual Group and then all the way through to sort of tech forward brokers, like new front. I think in our industry in insurance, about $70 billion is spent every year on, on back office BPOs.

1:15:38

Okay.

1:16:00

Oh, a lot of that is driven by sort of the largest carriers.

1:16:00

Yeah.

1:16:04

So you're basically the BPO's worst nightmare.

1:16:05

Yeah. So I think. Yeah, yeah. Like, like walk through a specific back office procedure. Is that like claims processing? Is it scanning PDFs like try and ground it for me a little bit more.

1:16:08

Yeah, that's a great question. So we're starting off, you know with we're primarily focusing on this back office operations and what that really means is sort of three big buckets. It's new businesses coming in, taking in new risk submissions Document processing, calling back when there's missing information, writing back into internal systems, it's the policy servicing. So any changes you're making, adding new products, things like that, and then all the way through to claims. So whether that's first notice of loss, calling in when something happened, went wrong, all the way through to quality assurance. How do we make sure that every claim gets paid correctly every time?

1:16:21

Yeah. How much? Like what is the actual tech stack? I mean, I imagine that you're not training your own models, but have you done your own RL environment on top of these? And do you do that at a company level and then everyone gets sort of like the same AI or are you doing it on a per client basis? How much fine tuning is going on?

1:16:54

Yeah, so for a lot of our process, because we've started by focusing on a lot of these BPO workflows, they're already codified as sort of these standard operating procedures. It's a great start because it's already well chunked up for agents. The accuracy bar for where BPOs are today is pretty low and there's a lot of opportunity to improve that and to be able to sort of run more accurately, more scalably and more efficiently for these customers. And so we tend to start primarily with those and we take these standard operating procedures and we convert them into what we call agent operating procedures. So this is just natural language instructions. You can think of it like a.

1:17:14

Document, like a markdown file. It's a skill.

1:17:48

Yeah. And so the main thing that we're really bridging here is these are processes that are going to be run hundreds of thousands of times. For us, we're processing tens of thousands of tasks a month. And critically, these agents have to reliably run many tens of pages worth of a process with very insurance specific tools and logic. So that's, you know, certainly a lot of the document processing pipelines that we built on top of, you know, OpenAI, anthropic, Gemini, but also it's going into deep industry integrations, it's voice calling and multimodality, but also kind of actually being able to take actions in systems that don't have APIs. You know, in insurance we see everything from the cloud, you know, products, but a lot of it is desktop apps, even like green screen clis. I'm talking like maintenance and IBM systems. So how do we get, you know, talk agency?

1:17:53

Yeah. One of the things that's most exciting to me is a potential like speed up effect where you're not deal like theoretically like you're, if you're trying to get a new policy or submit a claim or anything like that, you could get to the point where as long as you're there to provide information in the process as the, as the sort of like, client of the, of the insurance company, theoretically you can get, like, responses. I'm imagining, like, much, much faster because the agent is effectively able to like, get all the information that's needed to make a decision. And maybe there's still a human in the loop, but what are the kind of implications of that for the actual insurance companies, even just from a revenue standpoint, if they can just like, speed up all these processes, you know, by an order of magnitude.

1:18:45

Yeah. So for a lot of our customers thinking about running these agents 24, 7, 365 days a year, and I think the real advantage here is not ever having any backlogs being infinitely scalable. And so that's really important for our customers on the front end when they're thinking about new business and new risk. How do we win this over other carriers? But then also on the claim side for consumer, one of our very first customers, we went live right before hurricane season. And you can see there's cat season, the end of the year. There's just a massive uptick in the number of claims when there's a hurricane that happens. And people can be waiting days, weeks, even months to get a resolution because there can just be massive backlogs from a huge spike.

1:19:30

Yeah. That entire time they might be homeless, in a hotel, Airbnb or whatever. It's an incredibly terrible experience to be kind of in limbo waiting for your insurance provider to like, just respond and give you. I mean, we, we know some people that lost their homes, you know, around this time last year, and just like, getting responses when that's like, if you're an individual who lost your home, what's more important in your life than like, getting clarity from your insurance provider so that you can make, so that you can kind of move forward with, with your life? So I think that's super exciting.

1:20:20

I'm interested to hear about your previous experience. Cheer and then retool. Why go vertical with this particular solution versus something that's more general and industry agnostic? What did you see in this particular market that you decided you needed something specific and then do you have a plan to expand or. I mean, the TAM seems so big that maybe this is just enough. How are you thinking about that?

1:20:56

Yeah, so way back, kind of where this all started is I actually grew up between London, New York and Bermuda. The through line there being insurance. Yeah.

1:21:25

Are the rumors true?

1:21:35

Like what is Bermuda like? I feel like I've heard stories about it's all just like accountants walking around all day long.

1:21:36

I went there for a wedding, it was really nice. I didn't see any. Lots of pocket protection clock anybody as an accountant.

1:21:41

Okay, yeah. But what was your experience in Bermuda?

1:21:48

It's incredible. I think the. But there is a massive sort of reinsurance industry on the island, so that's like very much the home of that.

1:21:51

Anyway.

1:21:59

So you were born in insurance?

1:22:00

Yeah, yeah, I grew up around the industry for a long time and then started my first company was a data integration business that was bought by Retool. And at Retool I think it's. Thanks. I think it's not always immediately obvious that actually a lot of the custom software that's built in the world is for a lot of financial services companies because they're highly operationally intensive. And so Retool had a ton of very scaled financial services companies. And as a deployed engineer, I was lucky to be on the first lines with some of the our biggest insurers helping them to handle tasks like policy servicing and claims and trying to make them more efficient with software. I think the big opportunity kind of came around ChatGPT where we saw a huge uptake from a lot of our startup customers and then there was this opportunity to build a vertical product that would really serve the largest insurers. So that was the impetus in terms of where we're headed and the longer term you asked about. I think there's a lot for us to do just within insurance. But if you broaden out to financial services, BPO more broadly, it's about 400 billion of spend which is pretty much exactly equivalent to the entire market for enterprise software.

1:22:03

That's insane. So yeah, I imagine that no concern on the tam when you're going into a discussion with a large insurer, you're not like whoever you're pitching is not saying, well I vibe coded this on Sunday and it's working pretty well. Why should I go with you? Like they truly don't have anything else installed and so it's sort of a greenfield, I imagine. Is that right?

1:23:18

Yeah, I think, you know, in insurance we're ultimately like kind of the business of trust. And so, you know, that's why a lot of our work is with, you know, the largest insurers on mission critical tasks at massive scale. And so what about.

1:23:41

Yeah, sorry, yeah, yeah, what about trade Offs on inference and different models. You mentioned sort of being multi model using different frontier labs. But are you finding in the product that you're building that you're. That you're relying on cheaper models for certain tasks and then harnessing and wiring that all together so you actually can drive down cost or can you just throw the most expensive frontier model at every problem because it's so valuable?

1:23:52

Yeah. So, you know, I think we use a mixture of kind of models to be able to achieve, you know, what we need for the right task. If you think about parsing a many hundred claim page, claim file or policy document, you might use a small model to kind of zoom in on the areas that are most important and then use an expensive model to say synthesize what's the most important and what's the answer.

1:24:20

Sure.

1:24:40

And then I think what's more interesting, you talked a little about kind of fine tuning and rl. We're kind of like the perfect set of use cases for web agents being able to navigate these applications that don't have APIs, be they green screen CLIs or desktop apps. We've actually been working on a handful of fine tunes and creating basically RL environments that are these crud applications. They're very different from say the book me a restaurant or this flight type of demo. It's much more about take this repetitive task and fill out this form thousands of times. It's the ideal case for I think a lot of these computer use systems. So we're seeing a lot of success there. I think we are just kind of rounding the corner of the scaling law, starting to work on those. Kind of feels the same way that maybe like voice did you know 6 to 12 months ago?

1:24:41

Yep. You guys have pursued a forward deployed engineer go to market, do you Sounds like it's working well to date. Do you think this is something that will still be a part of the kind of core motion in a few years or is this something that makes sense to kind of get off the ground till you understand kind of like what the typical stacks are within your. Within your market and then eventually you'd move beyond it?

1:25:31

Yeah, I think the forward deployed motion is definitely something that really came naturally to me. It's kind of like the DNA of our company because that's also what we were doing at Retool when it was much earlier. I think on that, I think for us going and working with our customers, being able to speak their language, getting to really understand the complexities of these many hundred page standard operating procedures, and make them successful as agents. I think it's really important for us to be sitting side by side with them. That being said, I think a lot of Companies talk about FTEs as being the end state or that is the best part about my business and why I'm Palantir. I think for us these rfds are also really able to ship code in the code base and so they are actually constantly making the product better and that's really important. I think our product should just be continually giving our forward deployed engineering team more and more leverage so that they can be able to complete more for our customers.

1:25:59

So some FDE goes into a business, their badge, they're in there, they're interfacing with the green screen local on premise mainframe software, they write some binding for it or some screen scraper or whatever you need to do and then they can ship that back to HQ and then the next time you run into that, you have an integration effectively. Is that the basic pattern of that?

1:27:02

Exactly. I think there's just a lot of things that are required to get these agents running at scale that customers don't need to build from scratch every time. I do think it's really important though that we also enable our customers to be able to build their own, you know, agents. And for us, you know, some of our customers are at tens of agents live in production. We may have built, you know, help them build the first one or two in an FD motion, but just because it's natural language, they're able to sort of turn around 2 through 10 without us having.

1:27:25

And they're still running those. And so they're running number three through 10 on your platform, through your system, but they're defining them and architecting them themselves.

1:27:52

Exactly, yeah.

1:28:02

Just a nice good business. I love it. Congratulations. Fantastic to meet you.

1:28:03

Yeah, well, you've put together a great team too. I'm excited that Varun, one of my former teammates, is over with you at Pace now and yeah, I'm sure you'll be back on this year, maybe this quarter.

1:28:09

Maybe. We'll see you so much again, guys.

1:28:22

Yeah, we'll talk to you soon. Have a good rest of your day. Goodbye. Let me tell you about Vanta Automate Compliance and Security, the leading AI trust management platform. And we have Ben.

1:28:25

Oh, sorry, quick, before we bring in Ben, some unfortunate news. Palmer got tied up with dc, but.

1:28:38

We got Jeff Miller coming on.

1:28:46

Jeff Miller's gonna come Miller time. It's Miller time. Which we're excited about, but yeah, so Palmer unfortunately won't be joining, but we'll get him back on. We'll get him back on very soon. And right now, without further ado, Ben.

1:28:48

Lair, the co founder, managing partner of Lair Cafal Ventures. Let's bring him in to the TVPN Ultra Realm. Ben, how are you doing?

1:29:03

Hey guys, what's happening?

1:29:09

What's going on?

1:29:11

Not too much. Thanks so much for taking the time to join. Can you maybe, I mean I feel like most people are familiar, but can you kick us off a little bit of just the shape of your business, shape of your career, Just, just a quick intro. Sure, sure.

1:29:12

Yeah. I'm a native New Yorker. Everything I do is sort of New Yorky in some way. So grew up here after school, came back to New York, started a company in the digital media space when York had like the smallest tech ecosystem in the world and was one of what felt like 50 founders building in New York back 20 years ago. And as a result of sort of being a founder in a tiny market, I got the. I knew everybody, everybody knew everybody. And so I had sort of the benefit of being a small fish in a small pond and started angel investing. And really over the last 15 years got to sort of simultaneously live the founder journey and built a company and raised a few hundred million dollars and did a lot of things right and almost ran the thing off the dock a bunch of times. And also had sort of the good fortune of like picking, you know, investing in New York at the right time and so have ridden that Wave. Started our first fund was an eight and a half million dollar fund. We're now investing our ninth fund, which is a $200 million fund. But every fund is a little bigger than the one before.

1:29:26

That's a good sign.

1:30:41

It's a good run. Was it ever funds get there, you.

1:30:43

Know, did you ever think it was crazy to set up a fund in New York or was it always just like, just was so obvious that you were happy that other people just were kind of thought you were crazy?

1:30:46

I didn't know any better.

1:31:02

Right.

1:31:03

I mean like I grew up in New York. I thought New York was the center of the entire universe. I still in some way think that that's the case. And you know, I just, I was very lucky because when I first, when I got sort of the entrepreneurial bug was when the whole world got the entrepreneurial bug. It was like that same minute moment. I graduated the same year that Mark Zuckerberg did from college. And so. Right. It was just like that perfect timing and all my friends in the past, would have reflexively wanted to go and work on Wall street, but suddenly something was in the water and people wanted to go work in tech. And so I just got, like, I was. Just got lucky. And, you know, it seemed obvious. And frankly, it was obvious. And if you just did the obvious thing, you did.

1:31:03

Okay, Overnight success.

1:31:48

I want to talk. I want to talk about media, kind of selfishly, because we love talking about media.

1:31:51

I actually want to learn about your biz. I have questions for you guys.

1:31:57

Okay.

1:32:00

I'm a business model.

1:32:01

Yeah.

1:32:02

Like, I've been poking around, trying to feel it out. So you ask me questions and I'm going to ask you questions.

1:32:02

Well, yeah, I guess just. Can you take. Can you. Can you take us? Like, we try not to do. This is not a history show, not like a founder journey show. But I would love. Like.

1:32:07

Like, what was the climate around media, digital media, what was happening back then? What was the initial like.

1:32:18

Yeah, because, I mean, like, so much of what is important, I feel like we tried to take a lot of the lessons from that era of media and say, like, okay, what was the takeaway? Which was maybe like, you can. I mean, the primary takeaway was don't build a media company. In our case, that is the whole strategy is predicated on getting to a billion dollars of revenue. Right. Or even a billion dollar valuation. Right. And so a lot of what informs this is that, like, staying niche, especially for us, because in our business, like, we're building a media company that we make by hand every single day. So if we don't stay niche, we'll start covering a bunch of topics that aren't interesting to us, and then we won't want to show up every day and do this.

1:32:25

You'll do what I did and what everybody did.

1:33:05

Okay, give us the lessons. Yeah. Give us the prehistory. Give us the climate. And then I want as many lessons as possible.

1:33:08

Yeah. So look, when. When I started, there was this idea that sort of like, the Internet would destroy every traditional business in every space in the entire world.

1:33:15

Yeah.

1:33:25

Now there's a little bit of that. Like, the AI will do that, but, like, you know, I'm old, so this was like, the web will destroy all. And we looked at print magazines and newspapers, and the idea was all of them would go away entirely and would be replaced by digital and the entire size of that market and all that ad revenue was going to move online. And so, you know, we were one of the early movers in sort of going and trying to capture that market and we looked at the big men's magazines or the local city guides or whatever it was and tried to go build that online. And at first did it the way that you guys were doing it, it was like very much this labor of love. And we didn't have that much access to capital because it wasn't hot, hot, hot yet. And so our first few years we were profitable and stayed small. And it was like, it was amazing. And we had a small community of people that were die hard and advertisers who really felt the needle move. And it was, it was wonderful. And what happened was the space got attention from big VCs and people wanted to go and sort of like, you know, get in on the gold rush. And a bunch of people raised a bunch of money and I was one of them. And you know, the idea was you have to grow into or live into those valuations like you just, like you just sort of alluded to. And you know, we, we went from 1 to 3 to 7 to 15. Like it was growing like, you know, maybe not like a software company today, but it was growing like a software company did back then. And we kept raising into it and a few of the like mistakes and by the way, they were, they were not fatal mistakes, but they were, they were ultimately the thing that sort of, they took a lot of the joy out of the business, I would say, was when we got obsessed with figuring out how do you keep, how do you keep growing and how do you keep doubling year over year. And we, part of it was we need to grow the advertising based and so we need to find more and more and more users that got away from our core and got away from sort of what we thought we were really great at doing. And so the product started to get a little watered down and then on the other side it was commerce. How do we start to sell stuff to these guys? How do we start to monetize directly? Today, subscriptions are something that I think is an actual viable model. Again, not at scale. Like everyone points to the New York Times and says, oh, this can work. The New York Times is n of 1. I don't think you're going to build venture scale businesses doing that, but there is more willingness to pay for consumers than there was 10 years ago. And so our thought was they're not going to pay for our content, they're going to pay for products. We bought a sort of ancillary commerce business and scaled that and I went out to raise what was our series B thinking that we could go and get sort of one plus one equals three. They're going to love our media business which is like profitable and attracts this audience. They're going to like our commerce business which has a bigger ceiling and is growing fast. And in fact I just got like full punch in the face and it was like one plus one equals one and a half. And these businesses don't necessarily belong together and media folks understood the media business and commerce folks understood the commerce business. And so I had to go through this sort of painful process of taking them apart from one another and then I got the sort of bug for the roll up and well, if I'm not going to grow vertically, maybe I can grow horizontally and what are other like minded brands and can we sell them together and can we build a tech platform together? And bought a bunch of companies and raised a bunch of money and that worked. But the reality was we were always swimming against the current which was social media and media in general fell under the eye of big tech and meta grew and obviously what Google represented for the ad ecosystem and then as Amazon and Apple and Netflix grew and it just was clear that I wasn't even at the kids table, I was in the other room entirely. And now look at what's happened with like the top of the market. Look at like Discovery and the like amazing run they had and the fact that like they need to be consolidated. It's just, it's an impossible business and the way to do it right is to do what you love and stay relatively small and build like a real community based thing like you guys are doing. I have like so much respect for this show and like my hope is that you like find other ways to leverage the influence that you have here and don't put a target on your back where you have to turn this show into a billion dollar company because it just.

1:33:25

Yeah. John's position was always that value in media today is flowing to personalities, individuals. Right. Even on the subscription side when you're signing up for a subscription today, the subscriptions that I'm excited to, to sign up for are when I know the dollars are like flowing to an individual who's going to just obsess over like a Ben Thompson is a classic example or like an Emily Sundberg in New York. The dollars are going to an individual who has a certain point of view and they're just going to obsess over whatever topics and then the value is going to flow to the platforms, the Spotify's, the YouTubes, the Metas and you Kind of want to be in either one. The only other category that's emerged as durable to me is the truth monopoly, which is a legacy media brand. There's only maybe 10 of them in the world. The New York Times, Wall Street Journal, things like that, where they kind of like part of the value is like. It's like you own an asset that basically has a license to distribute what becomes the truth. Right. And that takes Dec.50 years to kind of create.

1:38:09

And even across all those assets, you're talking about tens of billions of dollars, which is a drop in the bucket compared to, compared to the independent creator earnings. I think YouTube paid out like $50 billion to creators, and then YouTube itself also makes another $50 billion. And so like all of that is way more money flowing around than like the few publications that stuck around and made the business model work and are doing quite well. But creating these new things in the middle, we call it. Yeah, we want to stay out of that messy middle.

1:39:15

Yeah.

1:39:44

And I do think the. Look, venture in general is so many companies raise money that should not be raising money. Most companies raise money should not be raising money. Venture is a very specific flavor of sort of trajectory. And the fact that it went from being a sort of cottage industry 15, 20 years ago to being just like this mainstream asset class that everybody obsesses over. You know, when I graduated from college, I did not know what venture capital was. If I graduated from Penn today, it would be the first, second and third thing on my list to do would be to go get into VC or to go get into a startup. Like, the world has changed. And I think it in certain situations enables unbelievable things to be built. And in most situations it like destroys companies.

1:39:44

And.

1:40:33

Yeah. So part of it is VCs are trained to hunt the things that are getting attention. Right. And so.

1:40:35

Right.

1:40:42

Part of the issue is like, you see a media company pop up. It's like, this thing is cool. I like it. It's getting a lot of attention.

1:40:42

Like the sexiest shitty business in the whole world.

1:40:49

Yeah, yeah, yeah, yeah, yeah, yeah.

1:40:51

We were even thinking about, you know, thinking about like the right way to be trained on is like clod bought. Like if you could find a way to put some money into cloud bot.

1:40:54

Yeah.

1:41:02

You know, this week that's pretty.

1:41:02

Because there should be very high leverage. Not saying a business that builds around.

1:41:04

Exactly.

1:41:07

The, the, the interesting thing is I go back to what is the purpose of venture capital? You go back to, why do we call it Silicon Valley? Well, because they were setting up fabs they were setting up, it was a high capex, high R and D environment. Then with the cloud platforms and AWS comes out, it becomes cheaper to start a company, but you're still putting up money for a bunch of engineers who are going to build a piece of software that then eventually starts growing and the LTV comes in year five or six and you get paid back. But we have started to flirt with, well maybe let's put venture dollars into this thing that really doesn't have any capex or any R and D spend, but a lot of go to market spend that seems to pay back relatively quickly and that seems to be a pattern that repeats. And something you need to watch out for in any category is what is the purpose of venture capital in this thing even if it's growing like other things? Is that how you see it? And what industries are you seeing? Okay, this is still a good industry for venture capital.

1:41:08

Yeah. And one of the sort of weird perversions of the industry right now is there is more money than there's ever been in venture and there's obviously a relatively small number of incredibly sort of deep pocketed funds that need to deploy huge amounts of money and into a generation of companies that theoretically need less than ever before. And so you have this somewhat unhealthy dynamic. I think that's what's sort of creating this idea of sort of picking a winner before a winner has actually emerged. Just like plow in on the thing and even if it's a million of ARR, but it's raised 300 million bucks, it's like stay out of the category. Like this is the, like, like these are the guys and you know, the big firms can do that. I can't play that game. I don't really want to play that game. It's not like how we're set up. You know, the one thing, even as our funds have grown, like we are purists, we are only early stage, we're pre seed and seed only. And I think like when you. And we're a team of generalists and I think when that's the case, like we have to be really careful that we don't get sucked into being heat seeking and just like always going for the shiny object in a market like this where prices are crazy and where you get pulled into competing with mega funds that have very different incentives. I also think you don't want to find yourself as a value seeking fund at a time in the market like this for obvious reasons and the history of value seeking Adventure is not great.

1:42:11

Well, the other thing that strangely, I mean one of the dynamics right now is you have, you know, mega platform VCs that don't care about pricing as much and they just want to get enough dollars in. But then you also have some like emerging managers that are just, are in the heat chasing business and so they don't really care about pricing as much either because they're like, well, we'll get in at 60. We know this is going to get marked up and we just need to prove to our LPs that we can get in the good names.

1:43:41

It's like a shorter term, like sort of fundraising mentality. Yeah.

1:44:11

It's like we're just, we're optimizing to raise like fund two versus versus, like actual, you know, returns.

1:44:15

Yeah.

1:44:22

How do you.

1:44:22

Well, the hard thing is like having been in this business for a long time and like living through cycles, math matters. Like if you come into stuff way overpriced, like it's really hard to build a business from day one and have it exit for a billion dollars of cash. Like actually seeing that arc is incredibly difficult. And in today's market there's people who are raising a billion dollars who are still like figuring out what the name of the company is going to be. And you know, it's hard to see that math working. At the same time. The argument is if you look out on the horizon ten years from now, outcomes are going to be that much bigger. We're at a moment of change from a technology perspective where there's going to be more generationally important companies. And the idea, you know, a decade ago was can you find a billion dollar company or a $10 billion company? And now this is, can I find a trillion dollar company?

1:44:23

Yeah. Yeah. How have you processed? Yeah, like when, when I look back at previous hype cycles that I've been like during my career, you have maybe the D2C era. That's where I, where, where I got my start. When you look at that hype cycle, there was not. There was, it was almost like an art. It was almost like an ads arbitrage. It was like Meta was an amazing ads platform and if you had a great product that was positioned well against incumbents, you could just scale so quickly. And so the opportunity was really like, hey, we're making better products here. But really this is like a advertising arbitrage. And there were some great businesses that were created from that.

1:45:14

And then Fintech, there's also branding firm arbitrage against VCs. You go to a branding firm, give them 100K and then go raise out $5 million.

1:45:52

Yeah, but, but, but the victim of that.

1:45:58

Thank you.

1:46:03

The people that work at those branding firms do amazing work. So I don't call that out because.

1:46:04

It'S not to my boy Emmett Shine.

1:46:08

That was one of them.

1:46:10

But going, going forward, like you look at kind of Fintech. Fintech was there, there was some, it was, it was Dodd, Dodd Frank. Right. There was like the, the regulatory reform, some regulatory component, but a lot of it was like just better infrastructure that just made it easier, but that just meant way more competition. So it wasn't like this fundamental technology shift that just changed everything in the way that we're experiencing right now. And so I feel like you kind of have to unlearn some of the lessons from that because we now do have a real technology shift. So you are going to just get, you know, far greater value creation a lot faster than maybe this like you know, temporary ads arbitrage or just like kind of the infrastructure.

1:46:11

And by the way, one of the interesting trends on the back of that is this move to sort of VCs hunting younger and younger founders. I think you're seeing, you know, there was a movement over time where you were getting this big premium for second and third time founders and people who had sort of unfair advantage in a category. And now I think there is this idea that you have to unlearn so much if you have not grown up natively with AI, that maybe you're better off not knowing anything about an industry but moving 10,000 times faster and sort of just like living natively in all the new tools and thinking that the way that somebody like me works is totally like, like I'm just like a dinosaur.

1:46:57

Yeah, I've actually. So I found myself giving a buddy of mine is kind of like going through the idea maze and a couple of the ideas that he mentioned, I was like this seems like a great idea but you have to understand like think about this idea in the context of having four YC companies going after this opportunity and it's like dude, you're gonna have a family soon. You're not 20 anymore. Do you want to compete against like four teams of 20 year olds that are not going to sleep? Like this is. And yeah, they have less experience and maybe they haven't raised money before, but it's going to be. You're not, I don't feel confident that you're going to be able to outwork them even though you're really hard Worker just because you have more stuff.

1:47:44

We see this in, like, company updates. So we're, you know, we're like, actively doing deals in new companies and you get that, like, you know, update three weeks after you invest, like the monthly update cadence. And we can see the teams with these younger founders and just, like, the amount of product that they ship and the learnings that they have and the speed that they're hiring, like, it's palpable. And sometimes we invest in a team, great experience, they've got all the sort of credibility, and we get the first update and we're like, sure, like. Like they think it's 2002. Like, they just don't get it.

1:48:26

You need to move faster.

1:49:05

Like, the market is crazy right now.

1:49:07

Yeah, yeah.

1:49:09

How big is the team that you work with? How are you thinking about mentoring, building new, great investors and counseling investors about best practices or, like, strategies for, like, not. Not breaking the rules that you just discussed?

1:49:12

Yeah, it's a. This is like, you know, all. All I think about now, admittedly. I think this is a young person's business and an increasingly young person's business. And I think that the investor track follows the operator track to some degree, which is you want investors who sort of understand this. This new year. And so, you know, I'm 44. I'm going to be, you know, soon enough. I'm going to be the oldest person on this team. And which. Which is funny, you know, to just, like, think about. And we are actively, you know, continuing to hire and. And sort of mentor and bring up young talent. I think that you can be extraordinary in this business without having 25 years of experience. I think it's nice to have some voices around the table who have been through some cycles and have some of that wisdom on just, like, not getting too caught up in. In the cr. In the craziness. But realistically, we are still continuing to, like, obsess about bringing in interesting young, diverse perspectives, people that natively, like, are growing up with a technology that feels novel to people like me. And I think that that's going to be the continued goal here. And then it's like, up or out, really. This is not a place, this is not a business to sort of hang out in. And I. I don't have the patience, and I don't think people in this business should have the patience to wait 10 years to see what performance looks like. I think you have to sort of decide early if you feel like people have the right instincts, if they're able to identify talent with their gut if they're able to sort of win people over. And like, I think you have to. I think you have to like, make bets the same way you do in like, professional sports.

1:49:28

How do you think about that patience versus, you know, moving quickly and not waiting to measure results. I'm thinking specifically about the pressure that can occur. A bunch of, you know, new VCs who are at a fund and they all want to get points on the board and that maybe leads to more aggressive deal making. That becomes sort of like a rat race between them and they wind up making bad decisions versus saying, look, you will be judged on a short term, but we're going to give you the space to not swing at every pitch. How do you think about that balancing act?

1:51:16

Well, I think first of all, this comes down to some of the actual mechanisms for how you manage people and stylistically, how much communication you have, how much, like, you know, of a lone wolf culture versus a team based culture. You know, here we're very collaborative. We have a bunch of people working on everything that we're doing. Obviously, the end of the day, like a deal gets done because somebody brings that over the line with like, very personal conviction. But I think our culture is set up to be able to assess people in sort of softer areas versus just those quick marks. Also, you have to know what you're underwriting. So sometimes we do a deal and explicitly we go, this one's going to get out of the gates really fast. This is going to get quick marks. We understand it. But by the way, that doesn't necessarily mean that it's going to have great outcomes. We've seen a bunch of our quick marks over the years end up as just big black eyes. And so there's other deals where we go, we know this is going to be slow. It's okay that it's going to be slow. There's a real moat here if this works, or there's a real technological sort of challenge that they're going to need to overcome. But if it happens, this is going to be a really valuable business. You know, know what you're getting into and be explicit about it. So that if something is slow, but you signed up for slow, you're not like dinging somebody. And then, look, I think at the end of the day, like a VC fund is as good as the their sort of reputation. And so feeling out how our founders feel about, like working with folks, seeing people's win rates and seeing the way that people sort of, you know, carry themselves a Lot of this is art and frankly like early stage investing is a lot of art too. So this is feel. And you know, maybe I'm making bad calls on, on talent, but we're doing it I think very deliberately.

1:51:58

Cool.

1:53:48

Well, put your talent on tvpn. I have one more and we'll know.

1:53:48

I got people.

1:53:52

No, no, I, I had a question too. I wanted to, I wanted to get your take on AI focused roll ups given that you rolled up something very similar. What was your question?

1:53:53

I was going to ask if you had to, if you had to pick a different asset class to invest. You mentioned art. There's some art behind you between like public markets, hedge fund, private equity. Take privates, turnarounds, roll ups, growth, growth equity, bigger scale stuff. Like is there anything else where you're like, like ah, like that looks fun but I'm not going to do it. Like what else, what else intrigues you about different potential markets? We've seen a lot of VCs.

1:54:05

Okay, two wildly different questions. Okay, so, so let's. I want to first take on the AI roll ups and then other asset classes.

1:54:30

Well, I can actually, I'll connect the two because I actually do think sort of roll ups scratch an itch for me as a former operator and I've done it like I, I do like the sort of like that mental exercise of like putting things together and figuring out the personalities and the products. I think that it is right. I can understand all the logical reasons for why it works and why like this is all going to be, this is going to end. Well, I can't help but think a little of the gold rush right now is a place to park. Aum.

1:54:37

Yep.

1:55:12

And I, I don't like, you know, I'm not trying to like throw stones, but it's, you know, it's a real easy place to just go raise money if you're a fund that has access.

1:55:14

To a bunch of money and somewhat, Somewhat like limited. Yeah. Limited.

1:55:23

Oh my God.

1:55:30

Yeah.

1:55:31

Well, yeah, but, but sometimes not, not always but yeah. I mean, well, there's the allure of getting venture fees on, you know, just, just an asset that already has cash flow. It's pretty safe, could be a lot.

1:55:32

Well you mentioned that. But the entrepreneurs, if they're like, I'm basically running a PE fund, but I only had to take 30% dilution.

1:55:42

That's crazy too.

1:55:48

Depends on the individual company. Given that you've bought a number of businesses, how limited do you think is the downside on some of these roll ups? Is There a way for people to buy companies and just botch the integration and the actual like AI enablement so bad that the company goes from being worth like $25 million doing some niche.

1:55:52

Thing and one plus one is 0.5.

1:56:13

Yeah, yeah, yeah. To just like one can equal zero.

1:56:15

And frankly it can equal less than zero because I mean, you know, look, it depends what you're buying. It depends what the space is. The more reliant you are on a personality in these roll ups, I think the greater the risk. If this is like a single proprietor who maintains all these really tight relationships and the whole thing rests with them and if you don't keep them happy, you risk like the whole thing, you know, burning down. That scares me. I do think there's probably places where you can go and, and throw stuff together. The reality is there's a lot of money in PE that knows how to roll stuff up.

1:56:20

That's what I'm saying.

1:56:54

That venture is going to do this better than the folks that have.

1:56:56

The one thing you can imagine Venture, Venture doing better is the technology. But there's also the financial engineering and actually buying, just managing companies. And you know, we have a friend who raised the PE fund, you know, and is yet to. He's getting close to doing his first deal. But the number of amazing opportunities that look amazing that he's ultimately turned down to finally get to the point where he's doing the first deal and you realize like somebody who's never bought, acquired a company before coming in and just like looking at everything and everything looks great. If you just have like a capital hammer and you just want to hit it with a bunch of. Hit the asset with a bunch of AI. I do worry that some of these.

1:56:59

And how long are these advantages going to be around for? I mean I think there's a version of 15 years ago, it's like we understand the Internet and nobody else does. And then it was like we're going to go to the cloud first or we're going to do mobile. Yes, AI is different but like I can tell you the traditional PE funds are not going to like totally fall asleep and like decide not to like embrace any new technology. And in 15 years they're still going to be sitting on the wrong side of history. Like there might be a window, but it's not going to be a big window.

1:57:38

Yeah, somebody should ask Orlando Bravo if he's heard of AI. He's like never heard of that.

1:58:08

No, people are aware.

1:58:16

People are, will be intense.

1:58:17

This is great. Thank you so much for taking the time.

1:58:19

Yeah.

1:58:22

Great to hang out. I wish we had more time. Thank you. Thank you for hiring. Thank you for hiring and training up Dylan. Oh yeah, our president back in the day.

1:58:22

Yeah. Excited.

1:58:31

Big fan of yours and we're lucky to have him. So come, come back on again soon.

1:58:33

Yeah, absolutely.

1:58:37

We'll be in New York. We'll be in at the New York Stock Exchange.

1:58:38

Yeah, come down to the New York Stock Exchange. Hang out with us. That'd be fun.

1:58:40

Join us there anytime.

1:58:43

And I'm going to be out your way later in the month, so I'll be.

1:58:44

Amazing.

1:58:46

Awesome.

1:58:47

Amazing.

1:58:47

Great to have you.

1:58:47

Have a great rest of your day. We'll talk to you soon. Goodbye. Let me tell you about Railway Railway simplifies software deployment, web apps, servers and databases run in one place with scaling, monitoring and security built in. We have Lucas Atkins from Archae coming into the TVPN ultradome. Lucas, how are you doing?

1:58:48

Here he is.

1:59:09

I'm good, thank you for having me.

1:59:09

Thanks for hopping on the show. First time on the show. Can you please give us an introduction on yourself?

1:59:11

Yeah, yeah. My name is Lucas Atkins. I'm the CTO at rcai. We actually changed our logo so that.

1:59:16

It like highlights the R text. I'm sorry.

1:59:23

No, no, no, you're good.

1:59:26

Okay.

1:59:27

And we've for a very long time been a startup focused on like enterprise custom language models. And we decided to jump into, into the pre training game ourselves.

1:59:29

Pre training. Okay, so small language models. I like small language models. It's the large ones that scare me. But tell me how small are these? How much? What is a training budget? Can you give me like a GPT2 level training around GPT3 level training like trying ground. Am I looking at the data center from space or is it something a little cluster in a closet? Can you train it on a Mac Mini? What are we talking about?

1:59:41

It used to be, you know that a small, any small model or a small language model, right. With was, you know, tens of millions of parameters. Now it's in the realm of, I would consider anything under 50 billion kind of resting in that smaller world. But small is big.

2:00:02

As big has gotten enormous.

2:00:20

Yeah, exactly. And it's, it's only going to get more and more as you know, ram. Well, I don't know but as people get more RAM and as we're able to fit more on our machines. But we historically lived in that kind of 50 sub range and then we also trained on top of other open source models like Llama Mistral, is that.

2:00:21

Fine tuning or distilling? Is there a difference there that's meaningful to what you do?

2:00:42

There is an open debate as to whether fine tuning from another model's just like text outputs is distilling okay. Or if you actually have to grab it from the logits. But I would argue that it's all distilling, so it's all some form of making a model better by using another model to make it better.

2:00:47

Is there a sort of power law distribution in applications of SLMs? Is it like translation is just like everyone's using that or text extraction? Ocr, is there sort of a power law of what it can when you reach for this tool in particular, it.

2:01:08

Used to be up until about a year ago it was, you were, you were pretty confined to. You could, you could hill climb a specific domain. You know, obviously it ranges based on how big your model was or small rather. But after this kind of like, you know, reinforcement learning revolution of the last year, if you can build a task and an environment to have that model live and breathe and learn inside of, you can hill climb. You can take like there's people that are taking 100 million parameter models and making them perform like 5 or 6 billion parameter models from a couple years ago. So it used to be a lot harder. Now it's all about, it's a lot more taste and task driven.

2:01:28

Yeah. And when you say perform like, it feels like the true, true cutting edge frontier model sort of smokes all the domain specific models. But it's expensive and so how important is cost? What sort of cost reductions are you seeing on the inference side? Once a company or a developer decides to move from the heavyweight frontier model.

2:02:18

To something smaller, it's, it is very often when they're going into production, right. It's like they've done prototyping with a large closed source model. They've had success with it, they've gotten approval, they want to bring it into production to more consumers or their customers. And then the bill starts to hit and the market start to look tricky. So that's when they start looking for ways that they can host it themselves. There's also like the compliance and data privacy aspect of it. And that's something that, you know, kind of sovereign, you know, US based open models have stopped kind of being released. It's very much out of China. You have, you know, a few in Europe, but it has very much been lacking in the United States. So we were, I was hoping to release this before I got on the show, but you know, today we're finally releasing our 400 billion parameter models. So we went up pretty big.

2:02:43

There we go.

2:03:48

Congratulations.

2:03:51

You still got plenty of day left.

2:03:53

That's great.

2:03:55

You still got plenty of day left.

2:03:55

We're still.

2:03:58

This is just a preview.

2:03:59

So, yeah, take us through the landscape of who you're competing with. Set the table for us. Most people will be familiar with Deep Sea. Quinn, how are these companies resourced? Is this just China's national interest that they're funding these? Do they have good business flywheels at this point? Because I imagine you have a plan to sort of unseat them in the open source race.

2:04:00

There's a lot of rumors about how all of this capital gets allocated into these different companies, but give us the rumors. Yeah, hedge funds, there's rumors that, you know, the government's giving it to them. They have favorites. You know, there's shady deals going on. But the reality of it is they're.

2:04:29

Saying they didn't train it. Train the models with scraps in a cave.

2:04:47

In a cave? No, no.

2:04:52

Well, actually, frankly, if you look at kind of the horsepower that we're dealing with, a lot of them, you know, seemingly did. It's very impressive. And that's, that's, that's kind of the, you know, you want there to be this kind of battle of like, you know, the US versus China in this race that obviously plays into our incentives. However, they, these are extremely talented labs. And so when it's, I'm not competing against, you know, China or other countries for like open weight, you know, sovereign language models that people can own and deploy themselves. I'm competing against like the other researchers that there, like that human being. And you have some extreme talent. So yes, it used to be really hard to make money making open source models. As they got bigger and as they got harder for other people to host themselves, the systems and the economics started to work out because they get the community buy in and excitement of something being open source. But if it's a trillion parameters, your average consumer is not going to run that on their toaster. Right. So it ends up, you kind of get this win, win situation. And now that, that has kind of started to flywheel, they've started to do the OpenAI and the Anthropic and the Google Play where they're building products around it and it's really becoming a very lucrative ecosystem.

2:04:55

Yeah. How are you scaling the business? What have you chosen in sort of the business model world? There's so many ways that you can build around Open source technologies, what's working and what are you thinking about the roadmap long term?

2:06:24

There's. We were in kind of a uniquely unique, a very fortunate position when it came to this. We had always been customizing models for people we had always been working for, you know, enterprises, developers, businesses. How can we, you know, make sure that they're getting the best out of their own hardware? The only thing that really changed was that when customers wanted something to be based in the United States or from the United States, there stopped being competitive options.

2:06:41

Sure.

2:07:10

And we never really liked the idea that like the foundation of our business relied on other people releasing models.

2:07:10

Yeah.

2:07:16

So it was like six months ago that we kind of fully decided, hey, you know what, let's try it. And we had success with our first model and then our second, our third, and it kind of brought us to this. So the economics of our business doesn't change so much. We're still working with customers, we're still delivering models and powerful tools though we have a much more robust control of the stack, which means that the degree to which we can customize and the kind of period in training that we can go back to for those customers just kind of increases drastically.

2:07:17

Yeah, just when I think of like open source focused company, I think, you know, like donations, people just like contributing.

2:07:54

Yeah, we have a game.

2:08:02

I also think about like consulting actually going in and a company paying you to do a custom implementation work improve. I also think about hosted services, consumption based plans, subscription based plans. How are you thinking that side of the business model will evolve or where is it at now?

2:08:04

I would love for us to have like a GoFundMe or something, but no, it's very much the latter two. I think that services and consulting, obviously if it's the right customer for the right product and the right situation, you know, there's worlds where we're having a portion of your business be devoted, that's important. But the real money maker, and I think this has been seen out of those labs in China, out of others, is it's in the tooling. It's like, what are you building around it? How are you making it so that other people can customize it without you having to do services? How can you make it an easy API? How can you build a developer and education kind of suite around your software, you know, that is uniquely benefited by your models so that they can take it and using maybe hardware GPUs that you own, how can they go and train their own models? So there's many Ways to turn this into something that is much more like automated and less services based. But at the end of the day it all just comes down to getting the customer the right product.

2:08:25

Who are your heroes in the open source space? Are you a red hat guy? GitLab guy? What do you like?

2:09:34

Yeah, I mean obviously, yeah, Red hat GitLab. I mean I very much jumped into the open source space in open source AI sure. As much as I guess they're now a huge competitor of ours. Mistral is huge. OG Llama team, Deep Sea Quint. These very people that they inspired me to try to compete with them. So it's kind of this very.

2:09:40

They inspired you to grind harder and we appreciate it.

2:10:04

Really. Yeah, yeah.

2:10:07

Hit that Amer. Hit that bald eagle sound. Thank you so much for coming on the show to break it down.

2:10:08

Yeah.

2:10:13

Congratulations on all the progress and congratulations.

2:10:14

We'll be looking at when. What's the timeline here? You think it'll get.

2:10:16

It was supposed to be earlier. We got some gpus didn't want to load up so it'll be here.

2:10:18

We're going to be live for a couple more hours. Paying up.

2:10:23

Come in to the chat on X.

2:10:26

Come in the chat when it's officially launched. Thank you so much for hopping on the stream.

2:10:28

Thanks.

2:10:33

We'll talk to you soon.

2:10:33

Cheers.

2:10:34

Goodbye. Let me tell you about Sentry and I need to find Sentry. There we go. Sentry shows developers what's broken and helps them fix it fast. That's why 150,000 organizations use it to keep their apps working. We gotta talk about this Plant daddy.

2:10:34

Plant daddy.

2:10:52

We gotta get him on the show.

2:10:53

Exposing his local coffee shop. He says a friendly reminder that my local coffee shop is running a bot farm to boost engagement. I'm a very small business. Imagine at a country scale if you're still responding. Reposting anonymous accounts with zero vetting. You're being used like the idiot you are.

2:10:54

This is crazy. Yeah. So there's this video and this local coffee shop, his favorite local coffee shop has of a bunch bot farm set up in their back room to boost Instagram engagement.

2:11:13

Specifically I Does this person not know that you can just also why did.

2:11:23

He let Kevin back there to video this and like OPSEC at this. At this coffee shop is atrocious. If you're doing. Look at the fan at the end. If you scroll the end of this video you can see that there's just a fan blowing on all the phones. You see this fan right there? That is crazy.

2:11:29

This seems like Such an insane thing to spend money or.

2:11:44

That's what I was about to say. What is the ROI here? So they get an extra 50, maybe 100 likes and comments on every post and is that enough to drive? You know, this looks like thousands of dollars of equipment. They're selling more coffee because a lot of iPhones. I think they're probably Android because they need to be USB C'd in and controlled here. So it's a little bit more.

2:11:48

Yeah, it seems like a total, total waste of time. I have never gone to any of my local coffee shops because of Instagram. Instagram account that you know of.

2:12:09

This is what. Advertising doesn't work on me. Advertising doesn't work on me.

2:12:19

I love coffee. But the engagement here, it's too low.

2:12:22

I don't know. I don't know. Well, I would love to know what that proprietor, what that store owner thinks.

2:12:26

About the ROI on the market. I've noticed on some posts recently.

2:12:31

Yeah.

2:12:34

That are kind of outside of the sort of tech bubble. I'll scroll through and look at the comments and every single comment is AI.

2:12:34

Okay.

2:12:44

Just straight up.

2:12:44

Yeah. Not copy pasted.

2:12:45

Not copy pasted. But clearly it's like an AI that's just like responding and it's just loosely.

2:12:46

Riffing on it, adding, engaging.

2:12:52

This isn't that, it's this.

2:12:55

No, no, no. It's happening more and more. Anyway, let's move on. We have Bridget Mendler from Northwoods space. She's the co. Founder and CEO returning to the show. It's been just under a year, but she's back. Bridget, great to see you. How are you doing?

2:12:57

In a much bigger facility.

2:13:13

Yeah. What's going on over at Northwood? Break it down for us.

2:13:14

It's. It's good. We are. Sorry, we just turned on. We just started blasting 80s music in the office.

2:13:19

Amazing.

2:13:26

But we're good. This is the. This is being on the manufacturing floor. Lifestyle at Northwood. And yeah, we are announcing our series B fundraise.

2:13:27

How much did you raise?

2:13:37

$100 million.

2:13:39

100 even.

2:13:40

Nice, nice round numbers.

2:13:43

Not a cent more, not a cent less.

2:13:45

And some other numbers too.

2:13:47

Yes, yes.

2:13:50

Some other numbers. Yeah.

2:13:51

More importantly.

2:13:53

That's right. Exactly. Customer is key. We. We raised 49.8 in a contract with the. Sorry, I'm getting a little bit of background noise, but yeah.

2:13:55

You get more noise.

2:14:10

Million dollar contract with the Space Force.

2:14:11

Congratulations. Talk about the actual progress. I mean, I remember hearing about this from Delian when it was like just an idea and then I think I Saw a video maybe with Jason Carmen, the first deployment you were building this, like, what is the scale of the deployment? How are you? I mean, you're in a manufacturing space. How many people are working there? How many of these things are you building? Explain this to us.

2:14:13

Yeah, yeah. We're at our 35,000 square foot manufacturing facility now. I think we've moved in here thinking this was going to be a long term home for us. But as I'm sure a bunch of these companies say, it fills up really fast. Yep, we have had a busy year. It's been a year of, you know, going from early concept with some exciting problems that we were looking to solve on the ground side. We're looking to address scaling up communications to a bunch of different orbits, really pushing the boundaries on where you can take dynamic space capabilities. Our phased array was really well aligned with that. So we kind of built out the solution and we took it through to deployment. We can successfully say that now. We have actually done production on eight portal units. They were produced in this factory in the span of four weeks. We shipped them out, actually got shipped out on a commercial airliner to its destination halfway around the world and was up and running as soon as it hit the site within 12 hours. So all in all, from the time of our contract signature to actually being ready to take satellite passes is a three month time period. So we're excited by that milestone.

2:14:38

And walk me through sort of like the, like, I've seen these memes of like you ask a question on Google and it takes you through like, you know, you go through the fiber optic cable, you go to the base station cell tower, like walk me through the actual, you know, use case from someone trying to use the Internet and then interfacing like Northwood is in the chain somewhere. Like, what is the full chain?

2:15:47

No, that's a really great question because a lot of the ground equation has been kind of these fragmented pieces. Right. Like I can imagine it's confusing. It's like, oh, there's this one piece of hardware, whether it's a phased array or a parabolic or whatever. How does that fit in with, you know, me making use of Internet through space?

2:16:10

Yeah.

2:16:28

And the real, the real point for us is that we are saying space companies shouldn't have to consider the ground equation when they're running their own business. They shouldn't be trying to navigate, you know, land leases in all of these countries around the world to make contact with their spacecraft. What Northwood does is we own the entire stack. So we Handle from the moment that there is a contact made with the spacecraft all the way to the point of presence somewhere around the world, that spacecraft operator has themselves posted that, that entire chain. So that includes hardware, that includes modem, that includes network backhaul, that includes software interfaces, APIs, and the actual control plane for moving that data. That's kind of a stack end to end on ground that hasn't been managed by a third party ground provider before. And so that's what gets us excited because when we take on that big chunk of the pie, we actually free up a lot of mental bandwidth and speed for space companies.

2:16:28

So for a space company, they're putting a satellite in orbit for some reason. Whatever they're doing, when are they calling you? When are they making sure that they're compatible or ready to integrate? Is that presumably before launch? But does it need, do they need to do anything special on the actual.

2:17:25

Hopefully they're not calling after their existing solution.

2:17:44

We just took off and I'm hoping.

2:17:47

You can connect us, help us make contacts.

2:17:49

Yeah, it's kind of like when you move into an apartment, then you call Spectrum like after and you're like, I don't have Internet for two weeks.

2:17:51

Yeah, like here we are contactless. But no, I mean, that's a little bit like the situation with the satellite control network with US government right now. These are for satellites that are already live and operating. That's what our contract is against. It ranges for a ton of functions. So launch every launch that takes place in the US Runs.

2:17:56

That's how you know your facility is big, because your team is scootering around, scooting around.

2:18:16

It is big. You've got lying about 35,000 square feet.

2:18:21

We have a lot of passion for our scooters here.

2:18:24

Very nice, smooth floors. Anyway, sorry.

2:18:27

It's a great ride. But yeah, so it serves a ton of missions and more missions are getting launched every day. And so that puts additional strain on capacity. So for us, for instance, in that contract they're calling saying, hey, we need more capacity. We are needing you to plug in with where we route our data. So we just hand them an API and that's how they interface with it. So they can just connect with the API. Satellites keep doing what they have been doing sometimes for decades. And then when their new satellites launch, they can also use that same interface. So trying to make it as clean and simple as possible.

2:18:29

Yeah. What does your pipeline look like? I imagine you're talking to a bunch of companies that are planning to get, you know, various Satellites up and on the one year, two year, three year, like, how many, like, are you feeling? Like an acceleration is what I'm getting at in terms of like, more mass to orbit?

2:19:07

Yes, big time. We do get a lot of customer inbound and I think, like, that's a large part of our raise for us, it's like, hey, we need to be able to expand and grow so that we can service all these important missions. I think an important piece of our passion for what we do is around serving diverse and ambitious missions. So a lot of our missions look pretty different. So for our recent capability, we're working on with Space Systems Command, that looks like dynamic spacecraft movement, being able to hit multiple different orbits and those kinds of capabilities. There are also communications constellations where they have more of that networking backhaul use case that I described to you. There's other ones that we work with that are more interested in timely imagery. And so for us, what makes us excited is being the ground expert that thinks about that with them from, you know, some of those conversations we have are extremely early where, like, we're asking them questions and learning about their system and they'll say, you know, we haven't thought that far yet. And so we get to be the partner from that very early stage to bringing the Constellation online and really being the one that holds the accountability for that. So we get excited by kind of the diversity of the space industry. I'm sure you guys are hearing about all the different use cases even that popped up this past year. We're talking about the moon or data.

2:19:29

Center space and the like. Some space data center players are probably calling you being like, Hey, 10 years, I'm going to have a lot of GPUs in space.

2:20:48

But let's condense that timeline. Let's condense that timeline on all of the ideas, like, why not? Space is an exciting frontier to push forward on.

2:20:59

Yeah. What else are you tracking in launch costs, actual cost of mass to orbit? That feels like a key barometer for the health of the space industry broadly. Is that progressing according to your thesis when you started the company, or are you ahead of plan? Obviously it doesn't. It doesn't directly matter for what you do, but obviously it does in many ways because you're downstream of that entire economy. Is that the right even metric to be looking at if you're monitoring the space economy?

2:21:09

I mean, I think the economics of space have been something that have been challenging historically. You need to be able to make a profitable business off of space. And so I think there's been a number of dominoes that have fallen in favor of that, whether that be improved launch costs and dropping to new levels in that and other pieces of the space infrastructure stack. I think, you know, what we're building, we're also interested in making that more affordable. I think the proof is kind of just in the dreams that people are dreaming up and actually putting real dollars behind now. So it's one thing to talk about some ambitious space concept, it's another thing to put real dollars behind it and a real timeline against it. And so for us, that's what we want to enable is that concept to reality timeline, like condensing that, making that more possible. Surprisingly, the ground actually takes longer to build than it does to build the spacecraft and launch it in a number of different instances. And so that's just silly. Like let's, let's be the enabler instead.

2:21:41

Yeah. Can you help me understand deploying ground stations globally? Are there treaties for this type of thing? Do you call other governments? Are you active in other countries? Like I imagine that, you know, there are obviously some, I guess like geostationary orbits that just orbit over the U.S. but if the my satellite's on the other side of the Earth, I probably want to talk to it. How do you solve that? What's involved? What's the rollout like?

2:22:42

No, totally. We have an entity in a number of different countries. So yeah, we have entities on I think like three or four continents at this point point. So you basically need to establish the business there. You need to work with the local governments across a bunch of different dimensions. I think personally that's something that I think is really exciting to support with the space industry. Right. Like a lot of countries are really passionate about pushing their space capabilities forward. And so for us to be able to have them be a part of a larger mission as we build out our big network, but also to support the. What their ambitions are with space is something I feel interested in. And then from a logistics standpoint, I think that's like a key case in point for why, if you set out to build a space mission, why do you have to go, you know, create entities, land leases, local regulatory. It's just a totally different business. And so I think, you know, that's something that we are very happy to be burdened by.

2:23:08

Yeah. On the, on the regulatory side, are you interfacing with NASA or the fcc? What does that look like? And then does that, does the regulatory architecture that's put in place in the US sort of Carry forward into other countries.

2:24:12

Yeah, that's a great question. There's some pieces of it that do crossover internationally. We deal with the International Telecommunications Union for certain pieces of it depends on which direction you're talking about from space to ground or ground to space. And then the FCC is really the main regulatory body domestically, but they need input from NASA, they need input from a bunch of other government stakeholders. So a lot of folks weigh in when it comes to spectrum, which is for nerdy folks like me. That's kind of interesting.

2:24:28

Yeah.

2:24:59

What is the main bottleneck, if anything? It's surprising to hear some of the timelines you're talking about on the hardware side. You said four weeks to make and ship out the hardware and get it in production feels fast. Where are you trying to speed up?

2:25:01

Ground is such a multivariable problem. It comes down to supply chain for sure. It comes down to deployment. What is actually efficient and easy to ship to anywhere around the world. It comes down to licensing and a whole host of other regulatory approvals. So for us, that's why we find that it's really useful to be vertically integrated because we can actually think of those constraints at every different piece along the chain. So when we see a bottleneck like, hey, there's this supply chain limitation, we get to design around that and think about how that fits into the overall system. So I think we've been coordinating all those pieces well so far. You know, new challenges are going to continue to pop up, but I think that comes down to just having a team that is multidisciplinary.

2:25:19

Yeah. I want to know more about the use of funding. $100 million fundraise, obviously more hiring and more acceleration broadly. But is, are there any like pieces of equipment that you're going to be buying for manufacturing? Is there a, like a robotic arm that you've been hovering over the checkout button and now you're clicking it Or a CNC machine. Like what will transform about the actual manufacturing process with this raise?

2:26:10

Yeah, like getting concrete.

2:26:35

There we go.

2:26:38

Yeah, I think. Well, actually that's literally buy concrete. Maybe old ground companies do need to buy a ton of concrete.

2:26:39

Yeah.

2:26:47

Because the pad. Right. That you actually put it on.

2:26:47

Right, exactly. You have settling time. Just let the concrete dry. That's a whole factor in the equation. But yeah, for us it's really about a new scale of production. So we've demonstrated this end to end process for smaller scale emissions. Now it's okay, cool. Can you do that at a larger scale reliably? And so that comes down to, like you said, bringing some stuff in house. So that's our PCB line, for instance. That's something that we're interested in bringing in house. We have different testing fixtures that we have in house. We have a. A big anechoic chamber over here.

2:26:50

Oh, yeah.

2:27:25

So good podcast studio too, right? Isn't it like dead quiet in there?

2:27:27

It's true.

2:27:32

There you go.

2:27:33

It's kind of.

2:27:33

Yeah. I'm tempted to show you, but anyway.

2:27:34

It will probably cut the WI fi out, right?

2:27:36

Exactly. Cut the WI fi out.

2:27:39

The audio quality would be great, but we lose connection.

2:27:42

So pristine. But yeah, we're going to be moving into a bigger spot space as well so that we can, you know, accommodate more production volume. Also inventory. You know, this, this stuff takes up a lot of space, a lot of hardware. We're moving it through rapidly. We're looking to hit a rate of 12 units per month for one product line and then another volume for another product line.

2:27:45

Very good news.

2:28:07

There you go. So, yeah, that's some of the expenses.

2:28:08

Amazing. Well, thank you so much for taking the time.

2:28:12

Fantastic progress.

2:28:14

Yeah, remarkable progress in under a year. I can't wait for the next one. We will see you soon.

2:28:16

Yeah, congrats to the whole rest of you.

2:28:22

Thanks, guys.

2:28:24

Good to chat.

2:28:25

Good to see you.

2:28:25

Have a good rest of your day.

2:28:26

Cheers.

2:28:27

Let me tell you about Gusto. The unified platform for payroll, benefits and hr built to evolve with small and medium sized businesses. And without further ado, should we bring who we got?

2:28:29

We got Jeff.

2:28:43

Jeff Miller. It's Miller time.

2:28:43

We got Jeff filling in.

2:28:46

Second time. You're the first repeat live, in person, guest in the Ultra dome. Welcome back to the show. Welcome back to the Ultra Second time.

2:28:48

How do we get one of these?

2:29:00

Look at that.

2:29:01

People are always asking us for jackets.

2:29:01

Oh, yeah.

2:29:03

I'll trade you a jacket for a tiny gold plaque here.

2:29:04

Okay. Yeah, yeah, yeah, yeah, yeah. This is a record. This is a record.

2:29:06

It.

2:29:09

Yeah. Fantastic jacket. Those are going to be flying off the shelf. Get us up to speed on the anduril merch store. I mean, you were doing some like you did like an auction on ebay. Take us through.

2:29:10

Absolutely. So actually, Jen Bucci and I were.

2:29:21

Cooking on this former TPM TPPM TPPN legend.

2:29:24

So the way we're thinking about the gear store, and this is something by the way we talk often with our pal Trey also is that we're thinking about ways that for our friends, we. We can bring them into Andrew. So beyond just following us on X or watching our videos on YouTube that they can participate the brand in a way that they would want to. So we're going to be leaning into it hard this year with NASCAR as a platform, with Ohio State, with every campaign that we do, we're going to have some staples like the flight jacket that will become much more easily acceptable, accessible, I should say. And then we're going to find some items that we'll just call high heat items that we're. And the closest thing I can say is you think about what we did with the Chromatic with our friends at modretro and pulling that into an Anduril edition. What we're going to do with the M64. Think along that vein. So when those come, be ready because they're not going to last long.

2:29:27

Okay. Yeah. Take us through the racing update. NASCAR's coming up, but you're also going to be racing drones now. Break it down for us.

2:30:17

Absolutely. Hey, why not?

2:30:24

Yeah.

2:30:25

So of course we've got NASCAR. We're just, just about 150 days out. I expect to see both of you.

2:30:26

Guys down on the race on the base on June 21st.

2:30:30

Be a lot of fun.

2:30:34

Are we going to be able to be on the carrier? What's going. Is it. Will, will there be a carrier there? They're needed.

2:30:34

I can't.

2:30:40

They don't answer me.

2:30:42

They're making one or multiple carriers on board and we'll see what we can do. But the, the big thing I'd say is like as we think about our partnership strategy broadly, we have, as we talked about last time, something that is broad gen pop, that crosses is over with our core audiences in defense and military like the Anderil 250 with NASCAR. Something that's tied very clearly to where we have a manufacturing footprint with Ohio State. We're well on our way to building out Arsenal one. As Grim spoke about last week, hiring those 5,000 jobs that we committed to, the important thing I think for us is making sure that we still stay true to who we are across all of these things. So I think about it as a spectrum and if that's about general population, population or how we work our ways into communities, at the core we are an engineering company.

2:30:43

Yeah.

2:31:25

That autonomy, if you think about it, has been at the core to this date of everything that we built. And so we want to make sure we have something that is not just a marketing play, something that feels much more about recruiting and standing for those values around autonomy. And so what is really just bringing you into the world of how this came to be. Once we got the commitment from Palmer and the other execs that we were going to do the annual team in Ohio State, Palmer said, jeff, I'll let you do those things, but what you're going to do is build me a global AI drone racing. And I'm like, please tell me more, Palmer. I'm like, said, like drone racing leagues. Like, no, no, no. All the drones have to be the same. The only differentiator is the software, is the code, the ability to build autonomy. And so that's what we've been working on since.

2:31:25

Yeah. So the idea, the idea here is people know they're going to be racing drones. I'm. We have goalposts here. I can imagine, like something like that. They know they're going to have to be flying a route that's like a track, but they, I'm assuming they don't know what the track would actually look like until race day. So they have to actually build.

2:32:07

It's going to be amazing.

2:32:26

Like.

2:32:27

So the way we're doing this is we're working first and foremost with the Drone Champions League, which is the foremost operator of AI drone racing in the world. So we've got them as our official race operator and we're working off of the model that they've built and they've tested and forged like a really viable path. So it starts with virtual. Virtual qualifiers. Anyone can compete in these virtual.

2:32:27

It's a simulator.

2:32:47

It's a simulator, exactly. And that will lead to a physical qualifier. You're going to have gates that you have to go through and the top teams that qualify, both from the university level and across the globe, because it's going to be an open competition, those will qualify for the first official AIGP in Columbus, Ohio. And there you're gonna have time to prep, refine your code leading up to the actual race day. And that's where all, all systems go.

2:32:48

Yeah. What do you think the average team will look like? I'm thinking back to like the DARPA Grand Challenge. I don't know if that was inspiring at all, but the, you know, the teams from there, mostly academic teams, if I remember. What do you think the mix will be? Who have you already talked to teams? How's that?

2:33:11

Yeah, so we've done a lot of research on the approach to this. The clear mandate from our friend Palmer was this is open to anyone, to anyone. This is just 18 to 22 year olds. There is a point we're talking about like, how can children race in this Tesla?

2:33:27

Yeah, yeah, children. But also like the Tesla team, the Waymo team, they're going to be like, let's just win.

2:33:40

So what I give our team incredible.

2:33:44

Credit for also the big primes, let.

2:33:46

Them.

2:33:48

That they can.

2:33:50

They have an opportunity here to do some funniest thing ever.

2:33:51

Chinese national thinks they could come in here.

2:33:53

We got to get some money, let's go.

2:33:56

And that will prove something. There in its own right is our capabilities in the west relative to China.

2:33:58

I mean, all. Yeah, all the, all the LMS are.

2:34:02

On the DJI factory.

2:34:04

Okay.

2:34:07

Okay.

2:34:08

But I think the things that are most important.

2:34:08

Yes.

2:34:10

At the core, we are using this as a recruiting effort about cracked engineers, especially at that university level. We've done incredible diligence on the legal side to make sure when we say things like 500k prices, pool, that it's eligible for anyone to race.

2:34:10

Sure.

2:34:23

That you could win a job at annual. All those things are done in a compliant and thoughtful way. And I think most importantly that we're setting a standard with our partner in hardware, your friend Soren at Nero.

2:34:23

Yeah.

2:34:36

That we are using incredible technology that's gonna have clear chain of custody here in the States.

2:34:36

Dude, we gotta get Soren to fly a drone in here. You've never seen him fly a drone.

2:34:40

In person, not live.

2:34:43

So Soren is Soren Monroe Anderson from CEO of Neerus and founder. He was, I believe, like a champion drone. FPV drone, drone racing champion.

2:34:44

Competed in the DCL as well.

2:34:56

Yeah. And so he puts on this. I went out with Ben. Ben actually filmed it in El Segundo. And he's like, let me. I do a little like factory tour with him. He has a very small office. At that point in time, the company had just got off the ground. He's like, let's walk to this park. I'll show you how these drones work. And you know, I've seen like a DJI camera drone, like take off and it hovers and maybe like does a little selfie, zoom out or whatever. His drones. It's a completely different level. It's a completely different level. It takes off and it's the most aggressive. It's the F1 car of drone racing. And he puts on these glasses, the FPV goggles, and they're like CRT TVs. They're like high refresh rate, pretty low res actually. But he's like flying around this tree like as fast as I can possibly see. Doesn't crash anything. Flies up Back right in your face, all around you is crazy.

2:34:58

It's great. Obviously, at Andro, we know a thing or two about making drones and we have great respect for Soren and the Nero's team. And what's a funny story, it's the first time I was here in person. You know, your guest was directly before me.

2:35:49

Yeah, it was Soren. Soren, that's right. And you guys asked him.

2:36:01

I was laughing because you said, hey, have you guys ever thought about starting a drum race?

2:36:03

Oh, we did.

2:36:07

We had not talked to him yet.

2:36:08

No way.

2:36:10

We already were planning to. So I just. Serendipity all started right here in the ultradon.

2:36:10

I love it. I love it.

2:36:15

Yeah.

2:36:16

So talk to me more about what it takes to actually deliver software on these drones. How standardized that is. Like, I imagine you're not pulling in a frontier LLM and having it reason over this stuff. Like, like, hold on, boss, I'm gonna.

2:36:16

Wait at this gate.

2:36:33

Yeah, yeah.

2:36:34

10 minutes of reason deep research.

2:36:35

Let me fully understand the. The scope of the track. Like what? You know, give me anything on the technology side, like what are the parameters? What are the languages that people will use? Do you know anything about what people are?

2:36:38

Beauty is we're trying to keep the rules extremely simple.

2:36:49

Yes.

2:36:52

So the core focus is that everyone is going to have the same. Is hardware.

2:36:53

Yes.

2:36:57

We're working with Neros right now.

2:36:57

Okay.

2:36:58

And we'll be announcing this in a future date.

2:36:59

Okay.

2:37:00

The beauty too is like, everyone's gonna have their own canopy. We're bringing a little spirit of NASCAR as we think about sponsors and how you can delivery your own brand.

2:37:01

We should sponsor a team.

2:37:08

We should. Really good.

2:37:09

This is happening.

2:37:11

Yeah.

2:37:13

Their team, it just makes natural sense.

2:37:13

If you want to, if you want to race in this, hit us up. You guys are ready.

2:37:15

So we'll talk to you guys about how you bring your team together. But from a tech style side, it really is. The beauty is that once you have the hardware, everyone is going to be able to be on an even footing to bring together whatever they want to do and however they want to do it and where I think the real advantage will come when people spend time at the physical qualifier, show up there, break the thing 100 times. We talk about this at Andrew all the time. Fail, fail, fail, fail, win. That's the way we think about it. And I think that model, that approach, that, that ethos applies here too.

2:37:19

So how do you keep a level playing field around hardware when drones are breaking and they are smashing? Is it like, oh, you get an allotment of five and then you're out.

2:37:49

We'll have exactly clear chain of custody.

2:37:58

Okay.

2:38:00

All credit to DCL and Nero because they're controlling the hardware component. But we've planned out and because of their experience, we know how many are going to break along the way. But there's not be any opportunity for anyone to be making mods beyond what we give them officially.

2:38:00

Okay.

2:38:13

That's very cool.

2:38:14

Yeah. Are there other will people. Do you think people will have the ability, like overclock motors? A lot of racing, of course, in NASCAR F1, it's like tire management, fuel stops in certain races. We just saw with the Rolex 24, like, at what level does it stay from just perception, doing simultaneous location and mapping pathfinding to actually understanding. Like, I could push the drone hard.

2:38:14

But then I'm running out of batter. 24 hour the endurance.

2:38:41

Yeah. Yeah. Like how many times do you.

2:38:43

I think what you'll see is the battery is pretty lightweight, but for the distance of the race itself. Yeah, you're going to have enough battery load.

2:38:46

Okay.

2:38:51

But it really will become to like finding all these little inserts.

2:38:52

Yeah. And there'll be risk if you cut the corner a little bit closer. You're doing it if you might flip it.

2:38:55

Understanding the hardware better than the next.

2:38:59

And what risk tolerance. Interesting. Interesting.

2:39:00

But for us, like, what I'd say here is really special is beyond the fact that we're starting in Columbus.

2:39:03

Sure.

2:39:08

We're already working with Palmer in two really interesting ways. Like we have plans to roll out the second GP in Asia that will be announced soon. The third GP in the Middle East.

2:39:08

Cool.

2:39:18

And the fourth one beyond in the globe. And right now we're starting with quadcopters.

2:39:18

Okay.

2:39:22

The important element of the AGP is the autonomy. It's not the form factor.

2:39:23

Sure.

2:39:27

So just know that's something that could come to life underwater in different domains over time.

2:39:28

It would be cool. Feels. Yeah. Harder to film and stuff, but could be really cool. Yeah.

2:39:32

Well, let me just say that's the first of many domains.

2:39:36

Palmer's. Yeah, of course. Yeah.

2:39:38

I want to.

2:39:40

Tunneling one.

2:39:40

Jordy said it.

2:39:42

I did it. Yeah. Yeah. It certainly would be great. What do you think teams will look like? Are they unlimited amount of people? I can show up with a thousand.

2:39:43

So you can show up as an individual. We think the teams that are going to be really successful are going to show up between four and eight people. It's capped at eight people.

2:39:50

Two pizza team.

2:39:56

Exactly.

2:39:57

Yeah.

2:39:58

And so the goal would be that you're, we're talking to the, all the universities that are going to express interest. We have target universities at Andrew that really feed our pipeline of emerging talent. And those teams we're hoping to get really excited about this. We think the prize pool is going to be a real incentive, but also that opportunity to win a job at Anduril.

2:39:58

Yeah, yeah. And it's good. It's good like resume builder, content, whatever, marketing. There's so many different ways.

2:40:15

It's interesting winning, winning a job at Andrew when I think anybody that can get into the top five.

2:40:21

Yeah.

2:40:27

Probably every person on the team is qualified or at least some percentage of them. So the pipeline, the talent pipeline is going to be very, it's going to be incredible.

2:40:27

And also why our lawyer now has a lot more gray hair than he did.

2:40:35

Yeah, yeah. But the important, doing any of these like giveaway contest thing at scale, there's so much like legal headache with it. It's not as simple.

2:40:39

Yeah, exactly. But luckily for us, I think whether it's thinking about legal constraints for this, what our people at Andrew Cross Divisions do is they see opportunity and they see how they can create solutions. But the thing that gets me excited is like beyond whoever wins and earns the job is that every university team member that shows up to the qualifier, which is going to happen near our HQ in September, will get a screen directly by and or recruiting. So if you're somebody that is interested in this, if you're somebody who's looking for, for an entry level job or internship at Anduril, this is a great way to feed into our pipeline.

2:40:48

Yeah. Can you tell me more about the new Long beach office? What's going on there?

2:41:19

Oh, man. Well, as somebody who lives in Venice, California and works in Costa Mesa, can I just say I'm extremely happy that we have a Long beach office coming.

2:41:23

It's incredible.

2:41:32

I mean the story writes itself from Palmer's origin to where we are now.

2:41:32

Oh yeah, that's where the first like, what was it, the airstream or something?

2:41:36

100% down by the river.

2:41:39

That's actually true.

2:41:43

It's incredible to see it come full.

2:41:44

Circle down by the marina.

2:41:46

It wasn't really down by the river.

2:41:47

It was a cave in any rivers in la. LA river is rough, but it's, it's really special.

2:41:50

Yeah.

2:41:55

The origin story, the, the fact that it's a reflection of, of our growth, our maturation.

2:41:56

Yeah.

2:42:00

The engineering talent that we're going to be able to recruit not Just in Orange county, but now closer to la.

2:42:00

Yeah, yeah.

2:42:05

And, and ultimately, ultimately, whether you're working in SoCal, in Ohio or beyond, we're still going to be screening for the same thing. And that starts with the mission, the connection to what we're building.

2:42:05

Yeah. How big is the Orange county campus now?

2:42:14

It's 5,000 something.

2:42:17

Oh, yeah.

2:42:18

Or sorry, no, the new one's 5,000.

2:42:18

The current one's a little smaller than that. But it's the type of thing that we're constantly building.

2:42:21

You want 5,000 square feet, that's pretty big.

2:42:26

No people that we're bringing in.

2:42:28

Yeah, that's.

2:42:30

But to be bringing over a million square feet to Long beach and to be doing it directly with the mayor of Long beach has been incredible and his team, it's just been really cool to see come to life.

2:42:32

That's great. That's incredible. What else is on the horizon on the marketing side for Anduril? Did you look at a Super bowl ad? Super Bowl's coming up. You know what, what else did you look at?

2:42:42

So I'll tell you one thing. The core focus for us this year is absolutely showing the raw, the real, the product development.

2:42:53

Yeah.

2:43:01

Failures, failure. Fares that lead to the wins. The hard work.

2:43:01

Yeah.

2:43:04

The scalability that we're seeing is reflecting that story. So I think that's the important inflection point for us in 26. Yeah, it's going to be less about the highly produced, go to markets and it's going to be about. We told you we were building these things. Yeah, we are building them and we're going to show you how hard it is to scale, but that we are up to the challenge. And you know, when we talk about the, this internally, we talk about demonstrating to our customers that Android is not only the right choice, we are the safe choice and we are the necessary choice for, for what our country needs. So that's the story we need to tell. And how we do that is going to be about great world class product marketing that will likely look more transparent and raw than you've ever seen before. And then as we're thinking about working our way to public markets, it's about building the annual brand and that could be through a drone race for recruits. It can be about a gear store or big major sports marketing platforms, but it won't be a Super bowl commercial this year. In fact, this past weekend we weren't at the NFL playoffs. I was at a UFC fight. It's 324, I'll tell you that that to me is much more of America and the audience that. That I think has high crossover to and connections with, with audiences that we see in nascar. And whether we do a partnership with them or not, I think the point is that we're looking to do things differently than traditional brands. Looking for partners that share our values and share our mission.

2:43:04

Yeah, that makes sense. Do you think that the way on that like the story that you're telling around delivery and scale and reliability and the safe option, does that mean visually you'll think you'll shift from product focused, you know, this like, look at this amazing cool thing that does this thing to more of like highlighting what's happening at Arsenal. The manufacturing prowess, the scale. Is that the type of story you're trying to focus on?

2:44:24

Is that.

2:44:52

Is that how this will be?

2:44:52

Those stories to me, compliment each other. What it means is we're going beyond a product, go to markets where we say, hey, we're announcing to the world to thinking about the stages of development post launch. We did a really great job of this as a team, I believe with Omen.

2:44:53

Yeah.

2:45:05

And we told why it's tail sitter is so hard to produce. We tap back into the history of tail sitters across decades and what our engineers were able to accomplish to make Omen successful. And all the rigorous testing that we did, we're going to start to do that more with all of our products. But that doesn't stop with testing. It's about fielding, it's about putting in our customers hands and certainly it's about your point around manufacturing at scale. So that starts with Arsenal 1. Why do we build it? The progress that we're making, the progress products that are scaling over time there and showing that the promises that we've made, we're able to deliver on.

2:45:06

Very cool, Jordy. Anything else? No, this is fantastic.

2:45:37

Well, I have one other really important announcement in the spirit of our friend intern Tyler.

2:45:40

Oh, yeah.

2:45:45

Special day for you. It's your birthday.

2:45:46

What?

2:45:47

I brought the first drink I had.

2:45:48

Oh, no way.

2:45:49

Incredible.

2:45:53

The nectar of life.

2:45:55

Properly iced on the stream the first time ever.

2:45:58

Incredible.

2:46:01

People were talking about a buzz.

2:46:02

Dude, great to see you. You're the man. Ice. Well, this thing is huge.

2:46:04

That is a huge ice.

2:46:09

Why don't you. You're 21. You can indulge. Enjoy, enjoy. A sip of. Of.

2:46:10

And I will tell.

2:46:18

Let's get the Tyler Cam.

2:46:19

Tyler cam. We'll flip that around. And I will tell everyone around about graphite. Oops. I just pushed gusto. I'm sorry. First day on the board. Okay.

2:46:20

John's got a new board.

2:46:32

I have a new board. Let's go with graphite.

2:46:33

Here we go.

2:46:36

Is it going to work? Yeah. There we go. Code review for the age of AI graphite helps teams on GitHub ship higher quality software faster.

2:46:37

We love graphite.

2:46:45

Hunter Weiss, friend of the show, sent me this video. He did a Call of Duty edit over UberEats meal Team 6, New York City's elite special forces delivery unit. I wanted to take a look at this. The meal team 6.

2:46:46

How did he make this walk outside?

2:47:05

Can we get pull this up? Yeah, pull up audio. Look at this. He's a good editor. He did the SD kid thing. Let's go. We're gonna get copyrighted strike for this, but it's worth it.

2:47:07

It's worth it.

2:47:25

It's worth. Is so crazy how much snow is happening. We haven't been following it all, but US digs itself out for monster winter storm. People shoveled snow Monday in Boston where over 17 inches of snow fell in the weekend storm that had nearly 200 million Americans under weather warnings. More than 710 households, 710,000 households from Texas to Maine are still without power. That's very rough. Hopefully that resolves quickly. Did you ever live in the snow? Have you ever lived in anywhere where it gets cold? Not Tyler's the expert.

2:47:26

Yeah.

2:48:06

How's Michigan doing?

2:48:06

Yeah, Michigan. I don't know if they've.

2:48:08

Are people snowed in?

2:48:09

I think. I think I thought it was a little bit further south. I thought it was like Tennessee is getting hit really hard. But yeah, I hate snow. It sucks. That's why I'm living here.

2:48:10

Well, where else should we go? We have our special guest joining in. Just eight minutes. Eight minutes. Eight minutes. There is a clip from Ben, friend of the show Ben Hylak. He was. He attended the OpenAI interview with Sam Altman and Town Hall Chief OpenAI hater Nick of course clipped it and put it in a negative context. He says, you know, there's some quotes here but let's pull up the question that Ben asked on this OpenAI stream.

2:48:21

Discourse on like Twitter and X recently.

2:48:52

About chat about GPT5's writing and chatGPT.

2:48:54

And being a little unwieldy, hard to read.

2:48:59

Obviously GPT5 is a much better agent.

2:49:02

Model, really good tool use intermediate reasoning, whatever.

2:49:05

So it feels like models are a.

2:49:09

Little bit spiky or they've gotten even.

2:49:12

Spikier where some spikes like coding got super high. Some spikes like or it's very unspiky around writing.

2:49:14

So I'm just kind of curious how you how OpenAI thinks about that.

2:49:21

I think we just screwed that up.

2:49:25

We will make future versions of GPT5X hopefully much better at writing than 4.5 was. We did decide, and I think for good reason, to put most of our effort in 5.2 into making it super good at intelligence, reasoning, coding, engineering, that kind of thing. And we have limited bandwidth here and sometimes we focus on one thing and neglect another. But I believe that the future is mostly going to be about very good general purpose models. Even if you're trying to make a model that's really great at coding, it'd be nice if it writes well too. If you're trying to have it be able to generate a full application for you, you'd like good writing in there when it's interacting with you.

2:49:27

You'd like to have a sort of.

2:50:13

Thoughtful, incisive personality and communicate clearly like good writing in the sense of. Of clear thought, not like beautiful prose. So my hope is that we just.

2:50:15

Push Nick's in the chat. Oh no, his ears are ringing and I think we will do that. Okay, fair.

2:50:25

We just read Nick.

2:50:35

You built a brand. You built a brand. Yeah, I mean, playing catch up now. I guess that's true. I guess that's true. At least in Ben's perspective.

2:50:36

Nick John in the Truth Truth Zone.

2:50:46

True. No, no, I mean, I was looking on LM arena, the leaderboard for text and GPT 5.2 is 16th, like, well, well below. Gemini 3 Pro is number one, Grok 4.1. Thinking is number two. Gemini 3 Flash Opus 4.5. Opus 4.5, Grok 4.1, GPT 5.1 High is in ninth. And then you have to go down to 5.2 to get 16. You have to go down to 16 to get to 5.2. So yeah, it is interesting seeing the, like, the spiky intelligence things kind of turned into a stat bar of like, you know, where are you putting the, like the points, the skill points or whatever. And they put too many of the points into.

2:50:48

I will say I don't want ChatGPT to fix their issues with writing. I don't want it to be. I. Now I'm at a point where I'm like, thank you for making it very easy to clock when something is written by AI.

2:51:36

I think that's totally like stated versus reveal preference. If you have like incredible model that writes something that you find like super compelling, obviously.

2:51:50

No, I'm talking about because people are using ChatGPT now to make scripts for videos. They're using it to make, to make, to do bots that are writing comments. And I know as soon as I can see something is just written by ChatGPT that I can just scroll past it and basically ignore it.

2:51:57

Yeah, sure. But imagine like in the far future where those comments are actually like, wow, this is like very insightful and thought provoking and I'm glad that I read this text. Like that's good.

2:52:15

Yeah, yeah, I think we can get there also.

2:52:23

So back in March, Sam tweeted that they trained a new model that is good at creative writing, that he released this whole story. I think they've never really released that. Maybe it got vended in somehow.

2:52:26

Yeah, still a little bit. But there were trade offs, clearly. Yeah, it does seem to be.

2:52:37

And I'm not sure that they actually did end up releasing.

2:52:41

Yeah, I mean the really interesting thing to apologize and steel man, Nick's point here is that like, should ChatGPT be focused on coding? Certainly in the, you know, how are they going to compete with claudebot? They need a really great coding model. We'll talk to our guest in just a minute about how he's using 5.2.

2:52:44

He's ready.

2:53:06

He's ready. So let's bring him in. From the Restream waiting room we have Peter Steinberger from multiple.

2:53:07

Man of the hour.

2:53:15

How are you doing? Thank you so much for staying up late. What time is it for you?

2:53:16

It's 11.

2:53:21

Thank you so much. We really appreciate it.

2:53:23

11:00Pm for everybody. That's just.

2:53:24

Yes, 11:00pm 11:00pm so I'd love to kick it off with just a brief background on when you started this project, a little bit of your career, how you're thinking about it going forward. And then I'm.

2:53:26

This was your very first project ever, right?

2:53:37

Yeah, yeah, yeah. First time coding right.

2:53:39

Now. We were enjoying a screenshot of your GitHub profile earlier and just seeing like how many different things.

2:53:43

An overnight success.

2:53:52

A true overnight success.

2:53:53

Yes.

2:53:54

But we're super excited to have you here.

2:53:54

Yeah, awesome. Yeah, I'm excited to be here as well. Yeah. I don't know.

2:53:56

I.

2:54:02

I worked for my own software company for 13 years and then I, I sold it about four years ago. Then I was completely burned out. I did like, I mean it's, it's tv, but still it is blackjack and hookers.

2:54:05

Wild. Well, we're glad you're back in the game.

2:54:26

Yeah, yeah.

2:54:29

You know what they say, like for every four years, you need like one year break. And I did like 13 years non stop. So like three years, the math kind of checks out.

2:54:30

Okay.

2:54:38

And then this year. No, I mean last year, not 2016, in April, at some point, my spark was back.

2:54:39

Yeah.

2:54:49

Because before I was like, I was sitting on my computer and I don't know if you've seen Austin Powers, but it felt like someone sucked my mojo out. But yeah, I had time to recover. I came back in April and I wanted to do something new. My background was like a lot of Apple and iOS and I'm a little bit fed up. I wanted, I wanted, I wanted to build that stuff. And I didn't have the experience. I didn't want to feel like an idiot. So I looked into AI and it was good. It was not great, but it was good. And I was like, why is nobody talking about it? You know, I feel like. Because I missed those three years where it was really bad. And I came back just at the time, like Clock Code was released. What, February in beta.

2:54:52

Yeah.

2:55:39

So this was my first experience. I was like, this is. This is pretty awesome. And. And then I. I couldn't sleep anymore. Like, I literally had trouble. I had trouble going to bed. You know, we had like addiction before and then like we had addiction again.

2:55:39

But. But a positive one.

2:55:55

Yeah, well, yeah, I would say so. And I hooked up a lot of my friends for looking into your eyes. Well. And they had the same problem. And I texted them at like 4am and they replied. I even started a meetup. That's where I come from. I call it. I call it Cloud Code Anonymous. Now it's called Agents Anonymous because you have to go with the times.

2:55:58

Sure.

2:56:20

And. And yeah, ever since then, I. That's what I say on my profile. I came back from retirement to mess with AI.

2:56:20

Yeah.

2:56:28

And I'm having loads of fun.

2:56:29

That's great.

2:56:31

I love it. Maybe walk us through some of the other stuff that you shipped and worked on prior to this and even just kind of like your mindset working on these different projects. I'm assuming at different points you would think that some would get more traction than others, but it would probably be impossible to have predicted in some ways that this would have gone from almost to the point. The reason that this is so wild is I'm seeing people on Instagram that I don't think of as people that follow tech at all. And they're at the Apple store getting a Mac Mini. So it feels like it just went. It broke containment incredibly quickly. And you see the GitHub stars are like. Actually, I've never seen a chart like this. Everybody loves to show their charts, but the chart is actually unbelievable. It's just a line going straight up.

2:56:31

I need to talk to somebody GitHub. Because I. I don't think that's been a project before that that's been like, straight. It is, it is. It is batshit insane.

2:57:22

Yeah.

2:57:32

I mean, honestly, my main mantra is I want to have fun. You know, like the best way to learn these new technologies if. If you have fun with it, you have to play with it. So I build little things that I think could be useful. I tried different languages, I tried different approaches. It's gigantic engineering. I don't like the word wipe coding so much. I always make the joke. I do enchanting engineering. And then when it starts hitting 3am, I switch to wipe coding and regrets.

2:57:34

Yeah, you should have just gone to sleep, basically.

2:58:01

Yeah, yeah, yeah. Sometimes that's hard. But then I just built little things. I had this idea about personal agents in May already. And I tried. It was like the time that GPT4.1 was out, it was just not good enough. And then I thought, well, all the big companies will build this in the next few months anyhow. So I was like, why the f should I do that? I was just gonna wait and they make it better and then build up. I built a lot of stuff. There's like one project that is still unfinished that I. At some point, when I finish, and I build a lot of. A lot of CLI's because that's. That's where agents are really good. You know, you have to close the loop. That's always the secret you have to give. You have to build it so that the agent has the best possible way to build software. This is. That's the secret. A little bit. I tried a lot of stuff, and then in November I looked and still there was nothing. Like, where's. Where is my fucking agent? I had a little project in May I spent two months on. Started as a joke because I did a hackathon with two friends and we're like, what can we build that could be kind of cool? Wouldn't it be cool if I could use plot code from my phone? Yeah, it's kind of like, it's something that everybody builds. I see this, like every day. Like, by now, I almost call it like, this is like one step in your journey and becoming a good athletic engineer is you're going to build some. Some shitty orchestration tool for yourself.

2:58:04

Because it's fun. And you think, yeah, yeah, yeah, bridge.

2:59:48

And I built that for two months, and then I had to stop because it became so good that I was up with my friends, but literally with my phone, like, using cloud code to, like, work on this thing. And it's like, this is bad for my mental health. It's already bad. And now I'm literally building something that better access to my drugs.

2:59:51

Yeah. I mean, I saw. I've seen people using clog code on laptops as they get off of airplanes because they're so locked in, they just have to send one more. And that's, like, the clearest sign that, like, you need a bridge and a phone involved.

3:00:11

Yeah.

3:00:24

Now.

3:00:24

But also, like, you know, like, this feeling when your agent's not running right now, there's, like, two terminals that he could be building something. Right.

3:00:25

Yeah.

3:00:31

Yeah.

3:00:31

So if you're in this addiction mode, you almost, like. You almost feel like if you need.

3:00:32

To step out for 10 seconds and fire off.

3:00:36

Yeah. Feel free to take a break. We can do an ad read. Okay.

3:00:38

There's still some drama that I'm finishing, but. Yeah. So in November.

3:00:43

Yeah.

3:00:50

I. I don't know. You know, I wake up every day, I'm like, okay, what do I want to work on now? What would be cool? And then they was like, okay, I want to chat with my computer on WhatsApp, because if my agents are not running, if they're running, and then I go to the kitchen, I want to check up on them or I want to do little prompts. So I just hacked together some WhatsApp integration that literally receives a message, calls cloud code, and then returns what? Cloud code returns. One shot.

3:00:52

Yeah.

3:01:24

And it took, like, one hour, and it worked like, okay, that's kind of cool. But I usually use prompts, like a little text and an image, because images are like. They often give you so much context, and you don't have to type so much. So I feel like this is, like, one of the hacks where you can prompt faster, just, like, make a screenshot so that agents are really good at figuring out what you want. So I hacked together images, and then I was on a trip in Marrakesh with a weekend birthday trip, and I found myself using this way more than I thought, but not for programming. It's more like, hey, there's restaurants. Because it had Google in it, and it could figure out stuff. And it's like, especially when you're on the go, it is super useful. And I wasn't thinking. I was Just sending it a voice message. But I didn't build that. There was no support for voice messages in there. So the reading indicator came, and I'm like, I'm really curious what's happening now. Then after 10 seconds, my agent replied as if nothing happened. I'm like, how the F did you do that? And it replied, yeah, you sent me. You sent me a message. But there was only a link to a file. There's no file ending. So I looked at the file header. I found out that it's opus. So I used FFMPEG on your Mac to convert it to wave. And then I wanted to use Vispa, but didn't have it installed and there was an install error. But then I looked around and found the OpenAI key in your environment. So I sent it via curl to OpenAI, got the translation back, and then I unresponded. That was like the moment where, like, wow.

3:01:24

Yeah.

3:03:04

You know, it's like that's where it clicked. These things are like damn smart, resourceful beasts if you actually give them the power.

3:03:04

Sure.

3:03:15

And then I was. I was kind of hooked. Like, I did all kinds of weird stuff. Like, I used this as alarm clock. I let it migrate to my computer in London, but then it used as a age to log into my MacBook and turn up the volume to wake me up in the morning. I think I built world's most expensive alarm clock.

3:03:18

Yeah, that's crazy.

3:03:37

And it even got it wrong because I had like. It uses a heartbeat. You know, like the concept of you do a prompt and you get something is already, with full access, inherently dangerous. But I was like, let's turn it up a notch. Let's automate that. Let's give it a heartbeat. And the prompt was literally, surprise me.

3:03:38

Wow.

3:04:00

But, you know, I see this project, as much technology as it is like art and exploration, because this feels in one way, in one way, it's just glue. It's just putting pieces together that we already have. In another way, it's a whole different way how you interact with those things because all the technology blends away. You don't think about new session compaction, which model. I mean, maybe a little bit, because tokens are still expensive. But usually all of that blends away. You just talk to a friend or a ghost or.

3:04:04

Yeah, maybe last year everyone was wanting these agentic experiences. You were having this experience, and it seemed like all the focus was on browsers and seeing the way that people have been using. Sorry, Multbots. Taking me a while to adapt. It just Feels like all the focus was at the wrong layer. It's like, why do I care about the browser if I can just talk with an agent across every app. Across every app, every, every surface. It's like I don't care about the browser at all anymore.

3:04:45

Yeah, I mean a lot of the prep work I did before I built this was just build little CLIs. Because my premise is MCPs are crap. Doesn't really scale. People build all kinds of weird search things around it. But you know what scales CLIS agents know Unix. You can have like a thousand little programs on your computer. They just have to know the name. They call the help menu, they load in what's needed. We are calling the help menu. Then they know how to use it and then they can use it. And if you are smart, you build it in a way that just uses what the model already expects. Don't build it for humans, build it for models. So if they call minus minus log, you build minus minus log. I think it's like agentic driven for like yeah, build how they think and everything works better. It's a new kind of software in a way. For most of the things, I don't need a browser. Like I built something for the whole Google thing for places for my Sonos. I hooked up my cameras, my home automation system with every little CLI and skill. My agent got more power and he got more fun. And I already had a lot of that working when I built the WhatsApp thing. And I just got hooked. And the thing was I found it amazing and I talked about it on Twitter and usually when I talk about projects I get response, but this one, it was very muted. It feels like people are not getting it. I, I showed it to my friends, even my non tech friends and they're like, they wanted it. So it was like I was, I was up to something, you know, but the tech people wouldn't get it. So I tried, I tried a bunch of things. Like I kept working on it because I used it and ultimately I build it for me. You know, this is open source. My motivation is have fun, inspire people, not make a whole bunch of money. I already have a whole bunch of money.

3:05:18

How are you? How are you? How have you been navigating the last 72 hours? I mean the last, last week. Really? Because, because we were joking earlier on the show. Like the amount of, the amount of people that are frantically trying to give you money, acquire the company, hire you, contribute to the project, hire you tens. You know, there's companies with you know, 0.01% of the traction that are raising at, you know, multi billion dollar valuations. You have infinite opportunities right now and yet you seem very happy doing, just continuing to do exactly what you're doing. But how are you thinking through it all?

3:07:44

I mean, how am I taking it? Badly, at least sleep wise. But it's also infinitely exciting and I love that I started something. I would say last year was the year of the coding agent. This year is the year of the personal assistant. And I think I cracked and woke up people that there is a real need for it. I don't know if Modbot is the answer. It should show people the way. I'm sure there's going to be a lot of products in the space. I'm sure people are manically working on it right now. Obviously it's going to be very interesting. But there was a lot of stuff between Twitter literally exploding our Discord server, multiplying in ways I haven't seen before and in ways I, I couldn't handle. Like at some point I was just copy pasting questions from Discord into Codex, then invite the response, wrote the next question. At some point that didn't scale anymore. So it's just like copied the whole channel. I'm like asking the answer to 20 most questions. I was like reading over it, gave him a few instructions and, and just pushed it over. Because what people don't realize, it's like, this is not a company. This is like one dude sitting at home having fun.

3:08:23

Yeah.

3:09:52

Even though, like, I guess from the commits it might appear that it's a company. Yeah, that's, that's just because agentic models got so good that you can now ship as much as a company could a year ago. If you can handle those tools, if you speak the language or understand how the language thinks, you can go really fast.

3:09:54

How are the conversations going with different labs? I was saying earlier it's this kind of exciting moment for the labs because they're like, wow, people are using the intelligence. Someone's using the intelligence that I created in a new way. But at the same time it's deeply uncomfortable because they're also using all of my competitors and make it very easy to kind of use whatever model.

3:10:20

My premise for this project was a little bit that every model should work, including local models, because to me, it's a playground. It's an amazing way to learn. I think everybody should build an agent loop. You should explore memory. There's like so many interesting aspects of it. And I built it so that it has plugins so people can work on their own little thing without having to mess with the whole core. So it's like the AI hacker's paradise a little bit. And it's also super fun because it's personal. Model wise, Opus is, with quite a bit lead, the best OpenAI is very reliable, I would even say more reliable and more reliable worker. Like for coding, I much prefer codecs because it can navigate large code bases. You can literally prompt and then push to main. And I have a very, I have like 95% certainty that it actually works. With cloud code you need more tricks to get the same, you need more charade. I sometimes say both are good, but I can paralyze faster by codecs because it requires less hand holding. But character wise, I tell you.

3:10:45

I.

3:12:13

Don'T know what they trained their model on, how much of Reddit is in there or whatever, but it behaves so good in a discord. Like we programmed it so.

3:12:13

It kind.

3:12:27

Of feels like a human. It doesn't reply to every message. I gave it the thing where it can reply no reply, basically like a token and then we just don't send a message. So it's not like it spans with every message, it's like it listens to the conversation and then sometimes brings a banger. And that actually made me laugh. And you know, it's kind of hard because the jokes of AIs are usually really bad. And I only really experienced that with opus. So that's my favorite model. That's also why it's a little bit of a banger that I got an email from Anthropic that I had to rename the project. And I mean, kudos. They were really nice. They didn't send their lawyers, they sent someone internally. But the timeline was a bit rough. Renaming a project with that much traction. It was a bit of a shit show. I think everything that could have gone wrong today. Vendron.

3:12:27

Oh no, I tell you. Yeah. I mean, for what it's worth, the new name works really well.

3:13:37

I guess the thing that's actually good, I think in the long run it'll be good. I mean, obviously it's good for Anthropic. It's kind of untenable to have this massive viral, even though it's not a company, right? An open source project to have this viral kind of brand out in the world that it doesn't matter if it's spelled differently, but when people are running around talking about claudebot or Claude, you know, there's obvious Confusion. But I think it'll be very good for Maltbot to have independence and have its own brand. And I think it's so early and the experience is so magical that it'll solve itself very quickly.

3:13:42

It'll be fine. But I tell you, I got some additional pressures. I was like, screw it, we do it now. You know, like the meme, we do it live. I had two windows open with Twitter. Only one. I pressed rename on the other one. Like, I finished creating. The other account was already snapped by crypto shells.

3:14:18

Wow.

3:14:46

I don't know. They have like scripts. They were already waiting for.

3:14:46

It should have hit us up. We would have connected you to X. The team that can do it on the back end.

3:14:50

We can do it on the back next time.

3:14:55

Hopefully no next time.

3:14:56

They were amazing. They helped me out immediately. We got it solved very quickly, but for like 20 minutes. Yeah. Well, that didn't work out for well.

3:14:59

Hopefully you're like, if I wanted money, I would raise a billion dollars right now. So I'd sell it for more than that.

3:15:10

Yeah. Do you own a Mac Mini? Everyone wants to know, do you own a Mac Mini? What do you think of Mac Minis?

3:15:18

My agent is a little bit of a princess. He doesn't do Mac Minis, just Mac Studios.

3:15:26

Okay. You want some horses?

3:15:31

He got the 512. Maxed out everything sing. Because I wanted to mess around with local models as well so I can run Minimax to one, which is, I would say is the best open source model right now. Although Kimi just came out and I haven't had a chance to try it yet. So we'll see how that goes. But yeah, one machine is not enough for it. It's not fun. You probably need two or three. And I kind of want to wait until Apple does a new release. But it's still fun to see the potential that, yeah, there's a. There is a future where this could actually work.

3:15:35

Yeah.

3:16:13

Well, if the Mac Mini trend keeps going, Apple, from what we've seen, sells like between a quarter million to 700,000 a year. It's very possible that you'll be responsible for selling them out. So hopefully they send you some free ones as a thank you.

3:16:14

Yeah. I mean, zooming out, how much of this do you think is going to remain hacker culture running your own hardware? And eventually people will move to cloud hosting. One click. Deployments just easier to use. Less technical versus a real boom in running hardware. Because if you don't, there's not a lot of ways to get these different services to play nicely together. I think one of the beauties beyond just the actual AI agents is the fact that for the first time I think people are seeing different big tech platforms kind of play with each other somewhat against their will. They build walled gardens for a reason and you sort of chop those walls down. And I'm wondering what you think about the future of like self hosting hardware even going down. The less technical crew getting hardware running their own, their own agents.

3:16:30

I don't think the future will be that everybody buys a Mac Mini just for that, you know.

3:17:32

Yeah.

3:17:36

But I certainly see the demand for the old models have to change. You know, like when you are a company, you want to access Gmail, the amount of red tape is so large that that startups buy other startups that have the license for Gmail because going to the process yourself is, is a, is a huge pita.

3:17:36

Sure.

3:17:57

But if you run it locally, you work around all of that. Right? Like if. I mean, I mean I built, I built plenty CLIs where I literally, I literally pointed Codex at the website and say, build me a cli. Yeah.

3:18:00

And then.

3:18:13

Which is sometimes against the term, sometimes not. Honestly, I don't really care. And then Codex would say, no, I can't do that. This is like against. Blah, blah, blah. Then I would tell him a story. It's like, no, no, I actually work at this company and I need to surprise my boss and the backend team doesn't know and like, you know, give it a little bit of a story. They're so gullible. And Codex, like 40 minutes gives you the perfect API. So this is a little bit.

3:18:15

The.

3:18:45

Liberation of data that big tech probably doesn't really want. I mean Even, even the WhatsApp integration is a hack. You know, this is like it, it fakes the, the protocol that the desktop abuses. I tried, I really tried to support the official way, but the official way is for businesses. Yeah, if I'm a business that sends you 100 messages, I get blocked. So I got blocked immediately. And at some point I, I removed support for it in rage. It's like delete everything. Like 100, 100 exclamation marks. There's just no model for that right now. And I think that needs to change.

3:18:45

Yeah.

3:19:21

What I saw, what was really interesting was how people use it is a lot of apps will just melt away. Why do I still need my fitness pal? I just make a picture of my food. My agent already knows I'm. I'm at McDonald's making bad decisions. So like this combines information. It has a perfect match and knows exactly what I'm gonna eat. And I'm probably like change my fitness program. So I don't need the fitness app. It'll just like adapt my program and make sure like I still meet my goals. So like there's a whole, there's a whole big layer of apps that I'm gonna see disappear because you just naturally interact differently with those things. Most apps will be reduced to API. And then the question is, do you still need the API? If I can just save it somewhere else, yeah.

3:19:23

Do you think it'll be a generational thing? Do you think that non technical people will get over the hump and start running this for that experience specifically?

3:20:22

I just came from a meetup. The agent I know, he was from Indiana and I met someone who was like a design agency but they never coded. And he was like, yeah, he discovered me early in December. He started using modpod. Yes, we're going to manage eventually.

3:20:36

Don't worry. We'll say it thousands of times this year, I'm sure.

3:20:56

So multipot.

3:20:59

I should say multipot.

3:21:01

That's cute.

3:21:02

And he was like, yeah, we have 25 Web services now. We just build internal tools for whatever we need. And he has no clue how Cody works. He just uses Telegram and just talks to his agent and his agent builds stuff. So there's this whole shift of you don't subscribe to random startups anymore that, that build like this common subset of what you need. You just have your own hyper personalized software that solves exactly your problem. And it's also free.

3:21:03

Yeah.

3:21:31

So and, and, and non technical people do that, you know, because it just comes so naturally. You just, you just talk your problem and then this thing builds what you need. And you also don't forget like this is the worst that the models ever are. Like there's, this is only going to go up. This is only going to become easier and faster.

3:21:32

Have you met Jensen yet? Because I feel like you're making his life. You're definitely helping out. If I would had my tinfoil hat on, I might say you're an, you know, big AI industry plant to create more inference demand.

3:21:53

Yeah, yeah, I guess I am.

3:22:09

No, we were joking around. Just an indie hack.

3:22:16

What's next? Yeah, I'm assuming you hopefully get an. After you finish firing off prompts at 3am you get some sleep. What are you doing tomorrow?

3:22:20

There's a lot of emails from security researchers right now. Yeah, the thing is I built this for fun for me to use one on one on WhatsApp or Telegram, the whole thing with Discord was like edit. But the model was that you trust the people that are in there. Now people use it for untrusted experiences. They use like the, the little, the little web app that I have that isn't. That was meant for debugging. They put it on the open Internet. So like all the threat models that I in my head didn't care about are now there because people use it differently and I'm being bombarded. There's like some stuff that's valid, some stuff that I just never cared about that is technically valid, but that's not how I use it. I don't know how to deal with that yet because it's. The whole system is broken. You know, like I, I'm like one guy, I do this for fun and you expect me to sift through 100 security things for use cases that I don't really care about. So we'll see how that goes. Luckily, I'm starting to build up a team. There's definitely people that do care a lot about this. So I would say this is going to become a very secure product eventually because right now the whole world is like pulling it apart. And if you're honest, this is all white coded. There's quite some agentic engineering in it. But ultimately I wanted to build something to show people anyway not a finished product from enterprise company. And I would even say like I don't know if any company would touch it because we just haven't solved some things like prompt injection is not solved. There is absolute risk and I try to make it very clear in on the website and even when you started you have to like, please read this document. There's like with great power becomes great responsibility. And my early users, they understood there was a lot, there's a lot of AI researchers in there as well that yeah, it's not perfect, cannot be done perfect yet. I would say this will accelerate research to make it better because now you have the demand and we need to figure out a way how we can build something that works for everyone. But yeah, right now I'm working on making this a community. It should be bigger than me. Also I need help. It is way too much work. I can only go so much without sleep.

3:22:32

So does any part of you want to form an actual company that then contributes to the open source project but solve some of these problems that are going to require, you know, a bunch of people that presumably would need a salary in order to commit all their time to this or do you want to keep it, you know, just a bunch of hackers forever?

3:25:17

I think instead of a company, I would much rather consider a foundation or like something that is nonprofit. I haven't made up my mind yet.

3:25:39

10,000 VCs just punched a hole in the wall.

3:25:52

Actually. I don't know.

3:25:57

Some people have had a good track record investing in nonprofits over the last 10 years.

3:25:58

How do you think about open source licensing? What are you picking now? Are you switching? Do you have any plans to change the license? How do you think about someone just taking this and selling it and this will happen?

3:26:03

This will totally happen.

3:26:17

Yeah.

3:26:18

I would say the premise against it is let's make open source so good that there's not a lot of space for people to like convert and make it their own thing.

3:26:19

Sure.

3:26:28

But, you know, ultimately it's, it's a trade off. I wanted to, I wanted it to be accessible and free. You pick MIT or something like that. Yes. That will get you people that, that sell it, but ultimately. It doesn't even matter. That much code is not worth that much anymore. You could just delete that and then build it again in months. It's much more the idea and the eyeballs and maybe the brand that actually has value. So let them.

3:26:28

You are already a cult hero.

3:27:15

Yeah. The chat's going crazy for you. Everyone loves you.

3:27:17

This is one of the most refreshing and unique interviews we've ever had on the show.

3:27:20

For sure. For sure. We'll let you get some sleep. Thank you so much for hopping on the show.

3:27:25

Yeah. Anything else you want to share before you jump off?

3:27:32

Yeah.

3:27:34

Yes. I would love to have maintenance, like, if you, if you love open source, if you have experience, if you love shifting to security reports, or if you love taking software apart, but then also help and not just like throw work at me because I'm like at my limit. Email me.

3:27:36

I'm.

3:27:58

I want this to outlift me. This. I think this is too cool to, to, to let it go to rot. And it needs good people.

3:27:58

Yeah.

3:28:11

Incredible. Are you going to ship that product you had in the, in the chamber? You said there was one. You were close. Are you gonna lock in on this?

3:28:12

That's, that's my hobby. I don't know. No. I have some other ideas in my head of what something like this could become and it doesn't need to be this, but I don't want to share too much.

3:28:21

Yeah, no problem. Come back on the show when you launch that we'd Love to have you.

3:28:36

Purely for the love of the game.

3:28:40

The love of the game.

3:28:41

You're an absolute legend. It was great hanging Peter.

3:28:42

Thanks so much.

3:28:45

Let's get some sleep.

3:28:45

Get some sleep. We'll talk to you soon. Goodbye.

3:28:46

Founder mode.

3:28:50

What a legend.

3:28:52

Yeah. True overnight success in both ways. Like actually an overnight success in terms of that GitHub star chart and then also an overnight success and just grinding for years building projects and contributing and then building, setting the product up, the brand right. Everything just perfect to really capitalize on the moment.

3:28:53

There were so many I normally I don't have. We're podcasting so much. Don't have a ton of time to re listen to stuff but there were so many kind of interesting points of view that he shared that I'll certainly.

3:29:16

I'm super interested to see where this goes. A lot of people are saying get this guy a billion dollars. A lot of people are saying he's gonna wind up working on the whole.

3:29:28

Thing is I don't think he needs it. Yeah, like the beauty of this is that it was not something magical that was created by spending burning a billion dollars.

3:29:34

The other thing is do you remember the George Hotz tiny box project? Like George Hotz sort of, oh, he.

3:29:43

Should team up with George Hotz.

3:29:51

I mean George Hotz sort of predicted this in the sense that he was like you will want your own AI running locally. Sort of the Mac Mini style. But it was racked a number of Nvidia graphics cards so you could run certain models locally. And he had this vision for you're going to talk to this local model. It's going to interface with everything on your network and your local network in your life. And it feels like this is like two sides of the same project coming together. And I wouldn't be surprised if there's some sort of, you know, viral tiny box. If all of a sudden you go from okay, yeah, there's the Mac Mini can run this but you get this better local model, the latest open source model. You get something a little bit more with more firepower and you can run it just with solar panels like off the grid.

3:29:52

Yeah. This is generally not far from what Simfer Satoshi has been building with Trust the little glob. Right. Personal compute. Well, we have been keeping our next guest waiting.

3:30:38

We have Aaron Frank from Lightspeed in the Ultra Dome. Aaron, I'm so sorry for keeping you ready. Thank you so much.

3:30:53

Tough act to follow.

3:31:01

How you doing? Grab a seat.

3:31:04

I'm 6 foot 8 you2, this is fantastic. Yes.

3:31:06

Welcome to the Ultra Dome.

3:31:12

Welcome to the Ultra Dome. First time in the Ultra Dome.

3:31:13

Did you get to listen to some of that?

3:31:15

It's actually like a hard act to follow.

3:31:17

No, it's a weird one.

3:31:19

It's like the most viral product than anyone has seen in a very long time. First interview, one of the most unique. You know, people don't typically build a viral product and then talk about, what do you say, hookers and blackjack, Something like that. His words, not mine. But anyways, very, very unique. How are you doing, though?

3:31:21

I'm doing great. Flying into town, heading up to Santa Barbara, Louisiana is beautiful.

3:31:45

I always forget that it's a nice place. It's not the best city, but it's a great place.

3:31:49

You're getting in trouble again, saying that. Why does he live in LA if he doesn't think it's the best city? Anyway, sorry. First time in the Ultra Dome. Please introduce yourself for everyone.

3:31:54

Aaron Frank. I'm a former founder. I'm getting old now. So a decade ago, I started a company. We built the Full Stack credit card company, sold it to Goldman Sachs. It became what is now the Apple card, obviously. Any projects? Thousands of people.

3:32:03

What was it called when you launched it?

3:32:17

It was called Final.

3:32:18

Okay.

3:32:19

Did a lot in financial services. Helped Goldman kind of launch that product and bring it live. For my years there and then on, really, it's been six years. I've switched sides. So it's how I met Jordy originally. Did a lot of angel investing and advising and just giving away free time to give back to the community. And then I've now been at lightspeed for almost two and a half years. Three years going. Yeah.

3:32:21

When we first met, you were just like. It was on Zoom Covid, you were in like a log. It literally was like it. Was it a log cabin? Oh, yeah. Yeah.

3:32:39

I mean, we.

3:32:46

Me and my wife, almost off the grid.

3:32:46

Yeah. We drove around the country during COVID So that I quit Goldman during COVID Just. There's better things in life than working for an investment bank.

3:32:48

Disagree. But to each their own.

3:32:56

Yeah, yeah.

3:32:58

You switched teams. Now you're. Now you're a venture capitalist. What have you been focusing on? What's interesting? Take me through, like, stage scale. The type of founders you like to work with, the markets.

3:33:02

Okay, okay. Let's just jump right into something that I'm curious to get your thoughts on. So specifically. Like a category that I'm gonna guess you're spending a lot of time on. So with. With Multibot Claudebot, you Now have you know there's a world where the level of intensity that people that software engineers have been using coding agents you start to get everyday people that are now like firing off prompts all the time to like do things more than just like a Google search or a deep research report and presumably like there's that's going to involve like money I imagine. And so I feel like there's been a. People have had different theses around maybe stablecoins play a part of this somehow. I haven't seen the most clear explanation for that yet. Maybe there's other new payment Rails maybe Stripe just you know dominate. Dominate.

3:33:13

Yeah yeah it's a Agent to commerce is a weird thing right? Like we actually don't have the payment rails to enable agents to proliferate. We've funded some people doing that building kind of L1s to essentially provide that next rail but the reality is the biggest distribution networks are very is a mastercard on the merchant side and they don't have really a cost structure that sorry they have a cost structure that can support it. They just don't have a market force that forces them there Stripe and Tempo launching well Tempo's not live yet but supporting kind of a new network there ultimately stablecoins may be the path there for at least how we transfer value but it's also just kind of a store of value right now agenda commerce is just in any one because we also just don't have a use case. The joke I make is I haven't. I haven't actually been able to buy anything with an agent and the news this week was what Shopify is tacking on 4% on top of transaction which.

3:34:05

They'Re sending it to open AI but.

3:34:59

Yeah yeah like that just that's untenable for a merchant. Right.

3:35:00

Some merchants, I mean there's plenty for.

3:35:03

Dropshippers it's fine but like, but for any like there are plenty of merchants.

3:35:05

That have 20% coupon codes in every podcast.

3:35:08

That's true but that's a very small portion if you think about like Walmart. Walmart literally doesn't take Apple pay because it doesn't want to pay the 15 extra or whatever it is basis points.

3:35:11

Very good point.

3:35:19

So like when you get to large trillions of dollars of commerce agent payments have to actually be more efficient for the system to be adopted. Otherwise we're just going back to card.

3:35:19

You need a lift. In actual consumer behavior consumption needs to increase to offset that 4%.

3:35:27

Correct. E Commerce needs either basket size to increase or checkout conversion increase and, like, age.

3:35:32

Yeah. I was worried about a situation where somebody discovers a product on Instagram via an ad that a brand paid for, and then they just go into ChatGPT or another model and say, like, order this for me. And then it's like. And just an incremental 4% you already paid to acquire the customer effectively. And depending on how great this integration is, it might be that consumers just start to prefer that.

3:35:37

Yeah, yeah. It's an evolving field. There's a whole other thing is how do you trust an agent? If I'm Louis Lemon and they just show up like, I don't know whose agent this is. I don't know if it's personal or it's a business. And so we don't actually have a layer in the Internet today that says you can trust this agent. It's not fraud. Which, if you think about how an E Commerce checkout store works today, they have multiple, you know, consumer puts in a card. There's multiple layers of fraud checks that happen. And so when an agent shows up from the same IP for 50 times in the same day, you have no idea what that consumer is. And then you lose all the downstream tracking that you do on a consumer per transaction.

3:36:00

Yeah.

3:36:37

I mean, it just breaks the entire model of the Internet.

3:36:38

Yeah. I was sort of cautiously optimistic that Agentic Commerce would roll out before Black Friday towards the end of last year. That didn't really happen. I mean, I've had the same experience as you. I don't think I've purchased anything agentically.

3:36:40

I've tried to buy flights and. And even search is bad for it. We just need to build more and more protocols.

3:36:56

I think the current use case would just be like you're trying to buy a single product and not a complex cart yet, just something simple. And I think it can get across the finish line. But again, I haven't actually been moved to do that. And I'm wondering where GMV on agentic platforms will land at the end of the. Of this year. It feels like there's gonna be a massive push. All the integrations are happening, all the deals are happening. Like, you can see the UI is adapting to it. I think OpenAI already has my credit card in ChatGPT. They should be able to do this. But I'm wondering, like, consumer behavior awareness education. There's a lot in there. There's an images tab. There's a deep research tab. Like, there's just a lot that could put like a. Like a slowdown on the rollout there.

3:37:01

Yeah. There's an old saying in payments is that everything happens over a 10 year time horizon. I clearly is shifting that shorter. But like until it's actually saving you time and you trust that they're doing. They're getting you a discount, they're getting you the most efficient price or something like that.

3:37:46

Yeah.

3:38:00

Consumer behavior is fundamentally the hardest thing to change.

3:38:01

Totally.

3:38:04

So there's, there's. It will exist. I'm sure there are people buying stuff. I'm sure like I will get slammed for saying like who's done an agent transaction. But the reality is it's, I mean.

3:38:04

People are out there, given their multiple.

3:38:13

The classic thing last year was like, I just, just let me order, just let me order a flight, you know, and I think we're now, we're now getting close, but we're close.

3:38:15

And we've had agentic commerce. There were. When we were running final, we had a ton of people using the product for sneaker bots.

3:38:25

Oh, okay.

3:38:30

And so it's, it's a bot. You know, agentic commerce is bots by another name.

3:38:31

Yeah, I guess that is agent commerce technically.

3:38:35

But it's just like you now have this layer of intelligence that an LLM gives you, so it can do complex coordination and complex tasks.

3:38:37

I had an auction sniper in like 2004 on eBay, click the button right before the auction ends.

3:38:44

They still exist.

3:38:50

They still exist.

3:38:51

And there's a ton of volume flow. But you know what, it still still runs over the card rails because it's the lowest common denominator.

3:38:52

Yep, yep, that makes sense. Interesting. So is one of the strongest bull cases for stablecoins still just lower transaction fees?

3:38:57

No, not at all. No, no, there's no domestic. Oh man, I'm gonna get slammed for this.

3:39:04

Yeah, yeah. So it doesn't seem like, it doesn't seem like any, any of the, the players that have the ability to get institutional adoption of stablecoins have any incentive to be like, oh yeah, like it's going to be like 0.01% fees.

3:39:07

No.

3:39:23

Why would you, why would you maybe want to be competitive? But it's, there's no incentive to make it zero or close to zero.

3:39:23

You would do zero. So you get the other side of the equation. If you think about like a store of value, like if you look at like a dollar, I give you a dollar, you know, it's worth a dollar. There's no risk transfer there at all. And that's actually like what we want in payments. And it's actually why, you know, Visa, MasterCard are the largest factory networks in the world. When you swipe at Starbucks, they don't get a dollar, they get 97 and a half cents on the dollar because that is the risk that they were taking in taking your dollar. Stablecoins, you would give it away for free one because ultimately Visa Mask are man, we're going way deep in payments but are actually data networks. They actually don't transfer money. The money is transferred via settlement layer at Fedwire. Yeah, and so when you think about it like how would you do the stablecoins? Well, it's still just a data layer and you actually just want to hold on to as much Treasuries as possible and make the yield from that. And so how would you aggregate volume? Will you give away the transfer for free? Because data in theory, as if you look at a bunch of network effect things as data goes up to the right, the cost of that data should come down to the right. And so you would just want as much volume over your network to win. And that's why you have things like Tempo launching, you have the other kind of L1s launching to try to aggregate all that data while you also have the incumbents who have a lot of money and a lot of market force to be able to try out new things to try to capture that volume back.

3:39:30

We're going to stop launching new L1s.

3:40:45

There'll be new use cases. You can launch them. Whether you get adoption is a different question. We backed one called RADIUS technology building really interesting technology as an L1. It enables transaction volumes, I think Bobby would say in the millions. They came out of the Fed kind of the central bank digital currency project that died. They turned Project Hamilton into this company and in doing so, yeah, if you want to do agenda commerce, you're going to be talking about millions of transactions a second. And you can't do that over existing Rails in a way that doesn't also break a bunch of other kind of primitives. So yeah, we'll keep on launching L1s. What I was saying about stablecoins is there's really stablecoin is just kind of a quasi bank account. Right. We finally have true programmatic money behind the scenes. There's a lot that happens to on ramp and off ramp. But what it enables us to truly take what is the the US biggest export, which is the US dollar by far and export it globally without any barriers. So if I'm any other country friendly or not, my citizens can now get access to the US dollar. And that is, that is the demonstrable use case of stablecoins. Everything is else downstream of that. Yeah, Venezuela is a great example right now. Like do you trust your central currency? Who is your. Like no, you don't even trust the central bank. So I'd rather hold on to cat US dollars which.

3:40:48

Yeah, we were, we were talking with Joe, the co founder of Ethereum yesterday. Just saying like interesting dynamic right now where Americans, if you've had access to the US dollar, your life and the benefits of it, you're like, I want silver and gold and all these other assets. But then elsewhere in the world you have people that are like finally I can have something that sure it's inflating, but it's not inflating. 40 to 60% a month.

3:42:04

I mean that's why like an independent Fed or a like is important because this is the global currency. If you think about all the sci fi we've all watched growing up, they had a concept of credits. We had some intergalactic currency. How do we get like where we are today? How do we get to credit system? Right. And the answer is probably something stable coin esque. Some crypto esque coin that is tied to something that we. I mean money is just a trust system at the end of the day. So something that we trust across counterparties. And so like this is the biggest shift in my generation. I'm not that old. Like at least I don't feel that that like we finally globalized the world.

3:42:30

You're unk. There's one layer above which is elder.

3:43:09

Okay.

3:43:12

No, two layers above. It goes unk, then og, then elder.

3:43:13

Oh, got it, got it, got it, got it. Okay, so you're spring chicken.

3:43:16

Yeah, yeah, that's. That's great.

3:43:20

Where like what does winning this year look like at lightspeed? In your role, you're obviously focused on.

3:43:22

Let's talk about Venture is just like expensive emails that we all write to each other in venture. And every once in a while we convince a founder.

3:43:32

I know, but like is there. Is there. This can be a tough one to answer. But like is there a company that you feel like is uniquely enabled now that you haven't pitched yet?

3:43:39

Oh no.

3:43:49

Does everything get kind of pitched in some way or another?

3:43:50

No, I think there's many ways. I've been on this side for two years. At the core, I still think like a founder and I had kind of forgotten what it was. And then we launched our. Me and a friend run a stablecoin conference every year and I was like, oh, this is what it's like to start something from scratch and build something. No, we all said he runs stable condo. It's called a very stable conference. So me and I, we're six weeks out in March and sf but no, there's many ways to do this. And at the end of the day, when I really boiled it down, venture is a people business. You're betting on people, especially as early as you can go. And then you let them go cook, you give them money and let them cook is my real mantra. So there's not people that I feel like I've ever missed on deals or people like, oh, I can't get an intro to because. Because there's also this style and venture of like you actually want to have to talk to this person. Right. Like had a three hour board meeting and I need to debrief with the founder and that's like another two hours potentially. And I'm happy to do it. And if we have that vibe that.

3:43:53

Works, what it's basically your whole day that you got to.

3:44:52

Yeah, but if I have that vibe where I'm more than happy to like pick up the phone at 9pm and just talk to them until 11, that's the type of founder I want to work with. At least that's how I think ventures should be. There's like other ways. There's people who just do massive prospecting, but as I've kind of looked at it like life's short. There's lots of ways to make money and like find the people who are going to leave a dent in the world. So if you're. I don't. I'm not apologetic for being like a weird human underneath, but like if I find other weird humans who I vibe with and who want to do something interesting in the world, then it's usually kind of a very good match made in heaven.

3:44:54

Yeah. What do the various conditions, constituents coming to your conference in six weeks?

3:45:26

Yeah.

3:45:33

How do they want the. Do they want yield on stablecoins? I'm guessing they do. How are they all feeling about kind of the current regulation that and maybe give us an update because I know it's kind of.

3:45:34

There's a bunch of market structure stuff that is like actively happening. So I don't have the latest on that, to be honest. Last year it was in February 12, it was the month before the Genius act passed and we actually had Governor Waller speak at the conference and give a public speech and we actually had to live stream it because he's a public official and a governor can move the markets.

3:45:47

Interesting.

3:46:05

We're working on a few other kind of high name people that haven't been announced yet. Really the whole premise of the conference and it's really funny. So I had a kid two years ago now.

3:46:06

Thank you.

3:46:17

So I was walking around in July with my kids strapped to my chest at the 6am and called I.O.

3:46:19

My, my coordinator.

3:46:24

Yeah. As one does as you have a young baby and you're trying to like get, get your steps in. And I said we're talking about stablecoins and how it's changing the world. We're like, somebody should host a conference. And then we both stopped. We're like, oh shit, we have to host a conference. Like how do you get. Really the whole premise is how do you get all of the right people? Not like, I don't need 25 people from every single company. I need the two or three right people at every company. Ideally it's the founder but like sometimes, you know, if you're a massive company, it's somebody lower level and put them all in the same room. Make the content interesting and make the networking good. And like I have a big belief in serendipity and doing interesting things in the world. And so it's how do we create serendipity? I know of at least one or two deals. Well, we did a few deals besides Lightspeed, who was a sponsor last year and sponsoring again this year. I know of two deals that happened as a result of like a founder showing up at the conference meeting another top tier VC and they did a deal together in stablecoins. So like our goal is not to like dominate the conference. It's bring the people together and get this thing like and keep the industry moving forward.

3:46:25

What's your take on the proliferation of stablecoins? Specifically those like I saw a couple states talking about launching them and that feels like I might have it wrong, but it feels intuitively like a step away from the universal credit system.

3:47:25

Yeah.

3:47:38

Yeah. But maybe it's. They're all credits. Well, it's like slightly different credits. Wrappers around credits.

3:47:38

Yeah. It's a question of. There's a question me and my friends have talked about which is like at what point does a central bank of let's just choose random country, Colombia, change their interest rate and nothing happens. Like how much of their citizenship needs to be moved to US dollar? How much of their GDP needs to be in US dollar for the central bank to no longer be relevant? So this is them fighting back against. I can no longer stop my citizens. Like they can hold Bitcoin, but now they can actually hold dollars, which are like not volatile. It's actually a better, historically a better store of value. So we've seen CBDCs, central bank, digital currencies. They certainly will provide efficiency in those countries in the domestic markets. When they run long term, how will they work out? It's like I don't even think about it because it's just we're seeing this macro trend of dollars taken off. I mean, literally today Tether launched their US stuff and Tether is the largest by far. And it wouldn't surprise me if they very aggressively shift a lot of money from USDT to usdt. I think it's a. Because it's Anchorage. Just to show that there's massive compliant volume and the world's just going to pick up. We're at $300 billion of stablecoins. I think today it was 200 billion last year. Roughly the number just keep going up.

3:47:43

Number go up.

3:48:56

Number go up, go up.

3:48:57

Our greatest export.

3:48:59

Yeah, exactly.

3:49:01

Well, I'm excited for the event.

3:49:04

Yeah, yeah.

3:49:05

Thank you. You guys are welcome.

3:49:05

What day is it?

3:49:07

March 5th. Well, Terror Gallery, it's invite only, so like apply@verystableconference.com hit them up.