Albacore raises $6.5M seed for undersea loitering munitions targeting Indo-Pacific threats

Do we have this uh we have Dante? Should we start early? Oh yeah. Yeah. Yeah. Let's bring in Dante. Speaking of drones, underwater drones, we got Dante from Albaore coming in. Amazing name. The Yeah, great name. Dante, welcome to joining us. Fantastic. Fantastic name.

I thought we were out of incredible names, but Albaor is strong. Albaore is a great one. It's amazing. I like Lord of the Rings stuff was taken, so we had to get creative. Go to the tuna section. Do you know the uh the the headline that uh is is in the the the group chat is YC backed Albaore swims into $6.

5 million seed round. I love it. Hit that gong. We've heard it all. We've heard it all. We love making marathon funds. Congratulations. It's great stuff. Uh yeah, take us who's in the round. Yeah, so lots of folks. Um it was led by uh Outlander. Uh we also have our first institutional Thank you, Paige and AJ.

Um yeah, our first institutional backer was D3, which is a very drone focused fund. Um they're sort of mix of US and and Ukraine um investments. Um but they they've backed a lot of folks like Nearos. I've got the they made two sequels to D1 Capital. We're on D3 now. No, we're never going to run out of names there.

You just go to D4 and then D5 and eventually you go to D10, D11. Yeah, A17Z is going to be a ripper for sure. Uh, who else? Sorry, we got we interrupted. Oh, yeah. No, all good. All good. Um, yeah. So, Brave Capital, uh, which is a bunch of exinkel people. Cool.

Um, uh, R squared, which is a national security focused fund. uh managing partner there is a Navy Seal commander, Road Scholar who's on multiple national security councils. That's great. Um uh a lot of different a lot of different investors. You got Box Group in there, New York. Different stuff. Yeah.

Uh they've they've done a lot of defense. I think they they also did uh Mock Industries. Oh, cool. Pretty recently. Uh yeah. Um how did you get into this? Uh yeah, so prior to this I actually started a a battery tech company and this was like while I was in college.

Um and really quickly identified we were working on like power systems that were high performance but um would make batteries for like really sort of uh complicated use cases um last longer. Um and so uh that led us to the aerospace and defense market.

Did a lot of work with drones and started thinking about the power problems there. Right? You know, the the issue with drones is you're really limited in terms of how far things can go. Um and that problem just gets harder and harder um when you start thinking about the oceans, right? Because uh oceans are really big.

Um you know, the people talk about in the Indoacific, the tyranny of distance. And so I was at a Shabbat dinner, ran into this guy, John, who I'd known for a while through mutual friends, brilliant engineer.

can't share too much about what he was doing before this um but he's worked on some major unmanned systems uh production type projects um very knowledgeable about the sector and I was like John why why do underwater drones have just like such niche use cases right um they're very focused on um short range you know very human in the loop type of operations there's nothing like an undersea loitering munition that we have right now um and uh we really need like longer range capabilities for anything that's happening in the Pacific so that's what we're working So is that like distinctly positioned against like the dive LD program from Anderole?

Like how do you see like that seems like a bold company to go up with against uh you know it's a great company. Uh it's funny that you found right. Yeah. I mean so actually the the founders of dive were were also investors in us really early on. Sam Sam Rouso and Bill Lebo.

Um they're super knowledgeable about the space obviously. U I was actually just with them earlier this week in Boston. Um, yeah. So, so the dive platform is focused on going really deep and doing survey type work. We're working on loadering munitions. We're we're working on stuff that blows up.

So, you can fit a 250 pound explosive payload on our vehicle. And that's enough to do really severe damage to a large surface vessel and if you're lucky enough to find a submarine to to sink that submarine. Wow.

Um, and what are the kind of um I don't know how much you can share, but like what what's the goal in terms of for like the timeline of a loitering munition?

you want to go out effectively, park it somewhere strategic, and then it's just kind of playing a waiting game before you would want to activate it and have use it for for a certain use case. Like, how long do you want one of your drones to be deployed without any sort of like human or machine intervention? Yeah.

So, uh we're targeting 30 days uh for endurance and a thousand nautical miles. And to kind of put that in perspective, the dive vehicle, which is, you know, a couple thousand pounds, um, that's less than a week and, um, a few hundred nautical miles.

And as you get to the the size of vehicles that that we're building, you know, our vehicle is like twoman portable. And, uh, our our lead software engineer claims that he can bench one of our vehicles. So, yeah. Well, he's he's really strong. The best news I've heard all day. That's amazing.

Well, he actually he just got one of his fingers cut off and so not related to his work here, but um once his fingers are healed up, he says he's ready to do it on actually cut off. Yeah. Yeah. He works on a tall ship and so uh yeah, log fell on his hand. It's a long story. Brutal.

He can still type faster than anyone in his in the office. So that's you lose a finger though. That's that's 10,000 aura points for sure. That's like eye patch level. Yeah, you're we docked his salary by 10%. Oh, stop. That's so rude. Stop it. That's very dark. What uh what what's your path to program of record?

How are you attacking? Yeah. I mean, so uh contracting is is changing really really fast, right? The current administration is doing things very differently. And you can get pretty far without a program of record, but um I'm not sure if you're familiar with the defense autonomy working group that that just got released.

Um there's a lot of things that are that are moving through programs like that with with SOCOM.

Um, you know, Indopayom has various plush funds that are available and we're, you know, we're focused on on really connecting with the operators and making sure that, you know, from the start of, uh, founding this company, we were really talking with with end users, the people who are going to be at the tip of the sphere um, in the Indoacific who'd be using this.

Um, but you also have to go once you have an understanding of those requirements, you have to go to Congress, too. And so, um, we we were actually lobbying Congress ourselves uh, within the first couple of months of starting the company. We've since moved on to we work with a firm now.

Um but uh yeah, you have to basically approach it from all angles and um you know there's also a lot of you know FMS FMS interest as well.

So, um, can't get too much into specifics there, but there's a lot of companies that have been really successful, um, with a a route to commercialization that consists of a number of different buyers and, um, you know, largely speaking, DoD can be, um, really appreciative of of FMS as well, um, because that means you can reach larger scale and, um, you know, potentially even battlefield test your systems.

Are you guys based in the Gundo? We are not. We are based in Philadelphia actually. Is isn't there is there great robotics program out there? Is that is that why? So, um there's a lot of great stuff about Philadelphia.

We're super close to Washington DC where a lot of the um you know customers that we'd be speaking with are based. Uh so it's a very quick trip. We can get up here on you know very short notice which is important. Um there's also a lot of great engineering and industrial talent in Philadelphia.

Uh and so I know there's not like too many hard tech companies that are based here. There's one they got pretty big called Ghost Robotics. They make like robot dogs military. Yeah. Um but we're we're based in the Penovation Center right now which is where Ghost Robotics started.

Um and we have lots of access to brackish water just um right next to our offices. Um and yeah, I mean Philadelphia has a really rich ship building history and um you know recently the the uh naval yard it's it's being um re reindustrialized. Um but it's it's actually not owned by a US company anymore.

Uh and so we want to build the sort of next generation of maritime capabilities in Philadelphia. um and you know help sort of foster more of that industrial labor force here that's still you know really strong. Amazing. Does your office have a pool? So we're getting a giant fish tank and also a turtle.

Um but we don't currently have a pool. We have limited limited space. Okay. What kind of fish tank do you need to make sure that like a presumably metal drone doesn't doesn't smash through the glass? It's going to be a pretty big fish tank. I don't know. I'd have to look into the specifics more. Yeah.

You got to call Keith Rey. He's He's got a big fish tank. Oh, yeah. He does. He's famously fish uh fish tank build. Uh well, thank you so much for coming. Great to meet you. I'm sure you'll be back on soon. Uh and uh congrats on on the milestone. Yeah. Yeah. Thank you guys. Been a pleasure. Talk to you later.

Have a good one. Cheers. Bye. Uh and if you're interested in some hardware for your wrist, head over to get getbzzel. com. Your bezel concier is available now to source you any watch on the planet. Seriously, any watch. Um, Andre Carpathy launched a new repo. Nano chat repo alert. New repo alert.

Uh, did you see this, Tyler? Yeah. Yeah, it's super cool. This is among the most unhinged he's ever written. Um, 8,000 lines of code to actually build basically chat from scratch. Is that how I should understand it? Yeah. So, um, he he's done he's done this whole like course lecture series on YouTube before. Yeah.

Um, where he goes through like uh different parts of how to train a model. Yeah. Um, and then this is basically kind of all of them together plus some extra stuff. So, it's training the full model.

It's doing the pre-train and it's also doing the RLHF to get it into like from predict internet text to be like a chat model and it's just using open source data sets basically for Yeah. Yeah. I forget the exact name, but there's a a good like RLHF data set that you can use to to get it to become like a chat model.

It's pretty remarkable. Is this so this is mostly for uh like educational resources? Yeah, I I think mostly I mean um like presumably you'd have to run this on a much bigger scale to get like a really kind of capable model. Yeah. Um but yeah. Uh what what was your interpretation of the hot take that somebody asked?

Curious how much of the code was written by hand? And Andre Carpathy says, "Good question. It's basically entirely handwritten with tab autocomplete. I tried to use claude/codex agents a few times, but they just didn't work well enough at all and net unhelpful. possibly the repo is too far off the data distribution.

Um, that's kind of interesting. I think uh part of it is that like the end product of this is like the repo itself. So you want very like nice code that is very readable.

It's in his style and he can like explain everything super easily which is like a little different from if you're building like actual product or or just like a vibe coding thing where the endpoint is just like the interface. It's what you're actually like using where this is like the code itself is the product.

So you're it's like you need super nice, you know, formatting and all this like kind of custom stuff.

Do you feel that like the like the the whether or not the task is in the data distribution when you're firing off cloud code or codeex agents like are you noticing that it's like okay well clearly uh you know there's plenty of data around how to build like a mobile a modal or you know some sort of like rest interface and so it's really good at that.

But if you're trying to do something that's um you know probably a unique problem that there might not be like a direct open- source example of like you're going to try and code it from hand. Uh yeah, that's probably true.

Um like I I find it's it's it's really good at writing React code and like Tailwind CSS components. Um and then when I try to do some like custom, you know, AWS flow, it's it's a little harder. Um and it just takes longer, but I still find it's it does it very well. Yeah.

Uh Jordan, did you see that uh Tyler Cowan put the timeline in turmoil taking shots at Europe? So Tyler Cowan, we talked about this yesterday.

He said that uh the European Union is run by uh people who have incredible depth of knowledge in a bunch of different areas and if they ran the world, the world would be growing at negative 1% growth and uh and is pretty pretty hot take. Uh Cassie Pritchard said, "Never understood why people liked this.

" quote, "The EU does exist. It is run by people like this and it does not have a negative 1% growth rate. " I'd say that's about the typical sharpness of analysis.

for Tyler Cowan though taking shots at Tyler friend of the show and young macro says difficult to overstate the extent to which you need to be underexposed to macroeconomic data to think that Tyler Cowan diagnosing Europe's growth as anemic is some sort of contentious statement or that his negative 1% hyperbole is anything but mild and contextually entirely suitable.

Young Macro continues and says I will defend Tyler Cowan. I will defend Tyler Cowan with my life. uh difficult to overstate the extent to which one's large audience must skew uneducated for this take to pass through with broad-based support. Nothing to do with pedantry.

10 out of 10 pedantic domain experts would say Tyler Tyler's quote is quite fine. Tyler Cowan, there's no way a regular guy would lose to a hobbit in a fight. A regular guy is what? Six feet. No chance.

person who has come across height as a concept as insufficient number of times an insufficient number of times to be up to speed with baseline magnitudal inst intuitions ecstatic uh at the gotcha after googling the average male height is 5'9 actually having a lot of fun on the timeline wow and yeah Cassie Pritchard's really going back and forth the young macro one of the greatest upand cominging posters um should we watch this clip from Rolof Ba from uh uncapped um unhinged the uh he is the steward of uh of Sequoia Rolapota.

Here let's hear it. One of the greatest stewards of capital in history. It doesn't support the numbers.

So there was a lot of analysis back in the 1970s and 80s with you know the capital asset pricing model and people figured out that there's this asset class that supposedly has uncorrelated returns and a bunch of asset managers deem that they need to invest a certain percentage of their endowment or foundation or pension fund into this thing called venture capital.

If you look at the data they're basically 20 companies per year on average over the last 20 30 years that have ended up being worth in realized exits a billion dollars or more. just 20 companies.

Despite a lot more money plowing into into venture capital, we haven't seen a material change in the number of companies that are outcomes that are that large. And I think part of that is that there's a lot more talent than really interesting ideas or interesting companies to be built.

And I think we're spreading a lot of that talent thin right now, similar to what happened in 1999, by the way. Yeah. So when you look at the data, the amount of money going into venture capital right now in America is in the order of $250 billion a year. And the numbers, you know, these are all estimates.

Let's just say it's $250 billion a year. Need a lot of exits. Well, let's just do some very simple arithmetic for a second. $250 billion going in every single year. If you assume that the firms generate 12% irr net of fees and carry, which isn't that great, by the way.

I mean, over the last three or four years, the NASDAQ has compounded at 16 17%. Let's just say 12%. Not spectacular. You basically just average performance. you'd need a 3. 7x roughly on 10 years on a seven-year exit horizon. So, I'm being a little bit aggressive. I mean, maybe it won't even be that good. So, 3.

7x on 250 billion. That approximates to a trillion dollars a year coming out. Coming out, by the way, that means and and that's what the investors own. So, let's say that the investors own two-thirds of the company to make the arithmetic simple. That's 1. 5 trillion annually in company exit value. Yes.

Just think about that for a second. Where's that coming from? Well, Figma is worth what? It's a lot but 40ish billion. I mean that's worth you know. 03 trillion. So you know if you start thinking in trillions and figma gets you 0. 03 trillion. You need 30 40 50 Figas every year. Yeah. To make that arithmetic week.

It just I don't see that many companies of that scale every year. So the only thing breaks is the return assumption doesn't hold. Yeah. And so venture is a return-free risk. Not a risk-free return. It's terrible. Yeah, you are better off investing in the index or holding t-bulls honestly.

And so I don't think venture is an asset class. Asset classes scale if you add more money. You can build more real estate. There's a lot of equities. There are, you know, trillions and trillions of bonds to be purchased. Venture capital doesn't scale with more money.

I mean, the most important I mean takeaway here is Bobby's question in the chat. What's on the wrist? Risk. Risk check. We'll work we'll work on that. We will we will get to the bottom of that. Venture is returnfree risk is a crazy line. Yeah. Um taking shots. But uh there was some good followup from Hari. Yeah.

Ragavan. He said obviously Roloff's a legend and one of the smartest people in the valley, but I'd love to dissect this add a bit more nuance to this. I completely agree with his directional point that most of VC everyone but top deskile sucks as an asset class.

But I think the same is true of hedge funds and the top cortiles less uh risky asset classes like real estate etc.

But the nuance is that the true venture capital segment formation through series A or modest series B is probably 20 billion a year maybe 25 billion because the remainder of the 250 billion is really private equity being deployed by a VC firm.

When you take it down in order of magnitude, the exit expectations make a lot more sense for true VC that is uh over let's say 8 to 10 year horizons 4 to 5x. So 100 to 125 billion in investor ownership at 150 to 200 billion in exits for the PE growth equity class theoretically of 1.

7 to 2x clear similar hurdle return because the hold periods are say 3 to 5 years from series D E etc uh works out to about 350 billion. So the total now comes out to 400 to 600 billion. Nothing to sneeze at but not one and a half trillion. And then you don't need 50 figs a year, but 15.

And more importantly, if a unicorn is a home run, a decacorn is a grand slam. We're seeing more of the centacorn, which triple crown. Let's go for the horse race analogy. Switching out of baseball into horse racing. Um, and he says because you could have 15 fig or literally one open AI. A centacorn.

If a if a unicorn's a home run, a decacorn is a grand slam, a centacorn should be like a world a world series clean sweep, something like that, like you won the whole se the whole season. Um, he's saying I'm not saying we'll have one open eye per year. That'd be crazy.

But we could see per year one, stripe, data bricks, etc. Two to three or sorry, three to five figma scale outcomes. Uh, 20 times Wealthfront scale outcomes. Wealthfront obviously was recently acquired by Gusto.

All we have to do is adjust that there are 20 unicorns a year for inflation and that's very plausible and that gets us to 400 billion in exit value. Markets are so much bigger than they used to be 10 years ago. We had zero trillion dollar companies.

Uh he says maybe I'm just too much of a techno optimist and Naval comes in with the reply and says nailed it. Correction on Wealthfront. Wealthrun's independent. You're thinking of a different company. Uh Wealthfront's the robo adviser. Oh, gusto guidelines, sorry, guideline somewhat different.

Um, yeah, the the interesting like nuance here is how weird open AI is. It just breaks the venture model because yes, it is, you know, maybe going to put up a trillion dollars of value, but as we saw the venture ownership Roloff was saying it'll be 50% ventureowned. Uh, and that's not the case for OpenAI.

And then also it just completely breaks the whole model of like who's founding these companies. Sam Alman's like much more experienced, much older than, you know, some uh Gen Z needle in the haystack founder that you can pick out and get, you know, 20% of the company for for $10 million. That just wasn't the nature.

Like the one of the first venture deals we talked about with Thrive was what $150 million at like a 12 billion valuation. it was like, you know, building like a 1% ownership in the firm. And so, uh, it's, uh, it's it's like even it's such it's just such an outlier. But this, that's the game of adventure, right?

That's the game of venture. Um, OpenAI just hired prizewinning black hole physicist Alexis Loop Loops Lucasa as the first member of its new opening eye for science team for Axios. Uh, good timing here because they are uh, leaning more into science.

Uh, Loops Saska is a Vanderbilt professor known for his work on black hole photon rings and now he will help shape how GPD5 tackles advanced math and physics problems and guide open AI's push into scientific discovery.

VP of science Kevin Wheel said GPD5 can already perform limited novel scientific research, calling it the worst AI model you'll ever use again. Um, I got to meet Alex earlier this year and I said, uh, OpenAI should hire you.

Uh, I I want to I'm hit him up after this and say, uh, I'm sure they he he he, um, he said something like that. That sound that that sounds like it'd be fun. I'm sure they were probably already talking at that point. So, um, not the most original idea, but uh, awesome guy and super excited for him to be over at OpenAI.

Yeah. Um well, we should prep for our next interview uh with uh Iso Kant uh from uh Poolside. Uh the news that we'll be digging into at some point during the show is uh a giant new AI data center is coming online to the epicenter of America's fracking boom.

We talked about this uh yesterday, but uh uh Cororeweave is partnering with them to build a uh a massive data center on more than 500 acres of land that sits on a sprawling ranch in West Texas.

The site is owned by the Mitchell family, which has run on oil and gas companies, which has run oil and gas companies for decades in the state and is located in the heart of the