Aeolus is building a planetary defense system for extreme weather using atmospheric manipulation

just like move because you know you're in Typhoon Alley like what's the whole point of you know having to stay there but it's like something so deeply ingrained right it's something you know that you accept that you you know kind of just accept as as feat and I was talking to the environment minister and the and the infrastructure minister of Japan recently and they were saying that they they've spent billions of dollars on you know disaster infrastructure like they would build these like you know columns under Tokyo right so like when it floods the water can go in yet like they were like oh shoot we haven't actually solved the the problem which is the intensification um of these events but back then I you know I was 11 and I did not know that you can you know actually start to fight these natural disasters um but I started like building you know robots when 3D printers were really popular started selling them um and then the money I made From that I put into Bitcoin and then uh going to like hedge funds in high school when I started a a small one.

Um and then and then in college I I came across this um uh startup which was building AI weather forecasts. Um as like not many people think about like you know weather forecasting as like this this huge expense that you have to take on.

But when you look at you know these larger countries they spend about $2 billion per year on supercomput essentially like you know a big chunk of the world's supercomputers are used on running partial differential equations to be able to calculate the weather right it's like something um that you don't really think about and with machine learning what you're able to do is reduce costs by 99% and increase resolution by 100 times and when you see that you know like we started out you know very small like five people but you know I was splitting time at Berkeley and then um at the startup and we became like the national weather forecast for the Philippines where it's you know was a lot um more affordable and I started to think about okay wow like our understanding of you know not only just earth itself but like being able to you know model ga gather data and then predict has gotten so much better I start to think about like what's the next natural step right like the whole point of gaining you know more intelligence is so that you can act upon it.

And that's what I said to think about. Can't really specifically tell what we're tell you what we're doing um in terms of, you know, the the application layer, but this is kind of like the the transformation we're thinking about. So, is it magic? Is it spells? Are you going to bomb the hurricane? Is that is it a bomb?

I've heard that pitched before. I I I wonder where you heard that from. It's a It's a 10,000 times the power of a a hurricane. Hurricanes are interesting, right? Because um the power the the amount of energy that goes in there is is so strong.

Like if you look at Google Research, they used to have this program where they're trying to harness energy um out of a hurricane. And if you can, you know, harness it properly, you can power the United States for for months and months just from one hurricane. That's incredible.

Which is like a crazy thing to think about. So, so you're going to make them and then you're going to control them and you're not, you know, and then there's going to be a bunch of sci-fi offshoots where K be nice to me. Koke had a good thing going.

He was powering a lot of data centers with his his homegrown hurricanes and then one got out of control, you know. Yeah.

Uh but yeah, I mean talk to me about uh weather forecasting and kind of the data sources because obviously yes massive uh kind of revolution in AI and machine learning generally but you slipped in that you're doing the weather forecasting for the Philippines. Yeah. Like you guys are like a data provider already.

The previous company I I worked at during um school. Uh that's what we're focused on. Um and and a lot of it is coming up with with models, right? How can you compute these partial differential equations more efficiently um with the same amount of data you have?

Um and so you look like Deep Mind or you know Nvidia coming up with their like frontier models, right? Like being able to you know run diffusion based models to to forecast better. But for for us what what we think is you know fundamental is collecting data that no one really has.

like software has come has become so commoditized that like now it's just like quad and then you know an X AI and then like open it's just like a never- ending battle for us it's like how can you collect data and how can you like see the world in a different way that allows you to to you know act upon it and how can you you know causally um act upon u like a weather phenomena right yeah uh you you mentioned diffusion uh is there a distinction between weather forecasting using diffusion models versus transformer based models or token prediction models like is is that a meaningful distinction or are you kind of using the term broadly?

It it gets really complicated uh in a sense that I think you know both of them are good at you know kind of milking um as much you know information you know looking at okay what what are the latent parameters how can you you know figure out you know somewhat unnoticeable um characteristics within within weather models.

But I think the interpretability um of weather models also become very important, right? Like if you can like characteristically like define okay, which parameter within a a high dimensional vector, right, is creating a lot of bias, how can you, you know, tweak that so that you can you know get back on track.

And it's like weather is so chaotic that like even one small pertabbation um in an initial condition can really mess up the forecast. And that's why it's really difficult to forecast like a week out. And so being able to like assimilate data as you go.

Um and that's why like we're focusing on okay, how can you, you know, collect data that that no one really has. Um so that you can, you know, make sure you're not like debiased. Yeah. Too much. Uh talk to me about the data sources for this. Uh I'm familiar with like Doppler radar systems that like the local news has.

Um, but are satellites playing a role now in weather prediction? Is the actual data that we're gathering to build these models on top of improving? And what are the key vectors that are driving uh better data going into these models? Yeah, satellite data is interesting.

Um, you know, a lot of reanalysis data has always been uh used and it's, you know, made by Noah or ECMWF. Unfortunately, Noah is not doing uh too well now.

What we're doing is building sensors and then literally like flying um into you know storms or into clouds to measure um within them right it's it's a complex convective process that you really have to you know just send something in um to measure autonomous vehicle I assume um no flying the plane personally our our um our focus is on how can you build infrastructure right like infrastructure in a sense that Okay, models are a part of the infrastructure, but how can you collect data um for you know a targeted you know vehicle?

So for example like hurricanes are interesting right like I'm sorry like I can't you know say everything in detail but in terms of hurricanes it's like okay if you want to see how convection is happening how what about you know um the the density of clouds what I don't know grapple columns and and water droplets exist you really just have to you know um fly like a a plane in there like it's hard to just do like drones because it's too turbulent or these like sensors are you know too big and so thinking about that um is is an interesting problem as well.

But I think fundamentally right what we're pushing against is that many of these you know reactive methods of of disaster if you look at you know insurance right like there's about like $140 billion per year in insurance premiums issued in the US uh for like tropical cyclones um and yet like insurance premiums have doubled every single year for the last you know four or five years in many coastal cities and when you look at like another catastrophe happening like last year hurricane Helena and Milton was was bad, but it could have been a lot worse.

There weren't even counter fives and Congress had to give out $100 billion in aid, right? But when a next Hurricane Katrina happens in, you know, Florida, for example, then insurance is just going to um be insolvent in many ways. Isn't there a market for better weather data by itself?

I mean, it sounds like fighting the hurricane is important, but uh we talked to somebody who works in data brokerage loosely and said like hedge everyone talks about, oh, we'll sell it to hedge funds, but there's only like hedge five hedge funds that buy anything and they're kind of stingy apparently and they don't really buy that much data.

You think about the insurance market. Yeah. $140 billion dollars. But if you're an insurer and you know that, you know, a whole bunch of houses are gonna be destroyed in two weeks or you know it in four weeks, you're not really going to be able to adjust premiums that fast to make up for that.

So it feels like and then and then you look at the government. The government might want to know, but even then it's like are they really a good buyer? I'm just just to continue spitballing in in California when there's certain communities that are threatened, they'll have like a private fire response.

I guess that makes sense. Basically, even an HOA is that a good is that a reason I'm just thinking about like the the dynamic right now where a a hurricane is effectively forming. Yeah. And the response is to watch it and then tell people to evacuate based on where you think it's going to go. Yeah.

And it's kind of insane that we don't have any sort of proactive response to say, "Hey, this thing is progressing to being more and more intense. Can we, you know, sort of stop it? " Yeah. So, yeah. Yeah. Data market. I I don't know how much you can talk about it. We might just have to have you back on when you No. No.

When you when you look at insurance, it's pretty insane. Like they look at, you know, okay, what does, you know, hurricane damage look like in the last 40 years? Okay.

let's, you know, aggregate average and then kind of, you know, add bias for like some of the intensification of these events or the increase um and the exposure and that's all you do.

And so a lot of the interesting financial assets come up when you look like catastrophe bonds like catast they're like certain hedge funds that literally just trade on catastrophe bonds, right?

It's just about like estimating what is the the arbitrage in terms of how much you know wildfire is going to wipe out you know a certain city or you know same for hurricane as well. Makes a ton of sense.

Uh well we'll have to have you back when you can talk more about the product and and where you're going with the company. Uh but this was fantastic. Thank you so much for hopping on. This is a real pleasure. Excited to hear more and congratulations. We'll talk to you soon. Thank you. Thank you for the Bye.

And that concludes Teal Fellow Tuesday. We have interviewed looking forward to next Tuesday. Next Tuesday we'll do a year from now. A year from now. Um also I mean we we mentioned Figma a few times but we forgot to mention that Dylan Field Teal fellow. Uh so go to figma. com. Think bigger build faster.

Figma helps design and development. You are addicted to Figma ads today. Great products together. I love ads baby. I love it. That is the I think the third that you've run today. I love it. Anyway, uh I like this post from Andrea. This bumper sticker. It says honk if you like the taste of venture capital.

And I just wanted to say and I love honk honk honk honk air horn honk honk and it's graza and lollipop. I thought that was very funny. Is that is this is this an anti- venture capital bumper stick. I don't know. I I don't know how anybody could be anti- venture capital. Iiculous.

My my lovely wife was one of the first investors in Braza. So So if you're honking you're honking and they're absolutely crushing. Um, what what else is interesting to cover today?

I mean, uh, TBPN was in the news with Ben Thompson chiming in over the weekend saying, "I love how it never occurs to the media that maybe they're just being outco competed instead of the subject of some sort of conspiracy. Literally, anyone can make a podcast. Sorry, they don't want to listen to yours.

" I think to to add to here the frustration from the media is that oftentimes entrepreneurs and investors don't even want to talk to the media anymore. So, it's not that they don't, you know, they maybe will listen. Yep.

But they don't even want to go and and um this is something we talked about with Abe, the guy who wrote the profile on us, is um he was curious, you know, why why why do you feel like people are so uh comfortable, you know, talking with you and and we don't need to go actually respond to this.

He said, "I think it can be both true that non-traditional outlets like Ben's and TBPN are outperforming many traditional ones and that some of tech distrusts the traditional media, creating some of the audience for Ben and TBPN. " And I love that because no quotes, no quotes around TBNN. Abe got the memo.

We're in the mainstream now. We are the mainstream media. We are the traditional media. Yeah. I mean, we're wearing suits. What's more traditional than wearing What's more traditional than wearing suits? Highly traditional. How could we be more traditional than wearing a tailored suit while you report the news?

While you newsmax, um I'm newsmaxing, John. What what what else is interesting to talk about? Um there's some news in Poly Market. The odds of a US recession have dropped significantly. I'm very excited to hear that. Love that. I I threw up a market. Will Coinbase acquire Circle before September? That's gonna be fun.

Um, somebody else was writing, see if I can read through this. Uh, this guy, I I don't actually know this guy, but he wrote up, uh, an article, and I guess he was at Coinbase for some time, which is interesting. So, uh, he's giving some background. I spent years in the crypto industry.

First at Coin Fund, then at Coinbase, helping scale its venture strategy. Everything in this post is based on publicly available data from circles S1 and Coinbase's public filings. No inside information, just analysis anyone could replicate, but most don't. Um, he's breaking down the USDC supply.

He says total USDC equals Coinbase's USDC plus circles USDC plus everything else. um platform USDC refers to the percentage of stable coins held in a party's custodial products or managed wallet services. And so he's basically saying that Coinbase has 23% of the total USDC in circulation.

And they've also been on a little acquisition spree. They bought that other big company. Uh I think it was a billion dollar acquisition at least. forget exactly the name, but we'll have to talk to Brian Armstrong from Coinbase about that hopefully tomorrow.

He goes on to highlight that USC is Coinbase's number two revenue driver. Yep. And contributed to 15% of Q1 2025. Could be more than staking. Could be a beautiful partnership. We'd love to see. So something Well, we'll cover it here and we will track it. What What is the poly market actually at? Let's see.

Relatively low volume right now. 5% 50% 10%. No, no, it was at it was sitting at 40% over%. Somebody took a big no bet and so it's sub 10%. Sub 10%. Okay, we'll see. Well, we'll see. If you uh if you have an opinion, hop on Poly Market and check it out. Uh I think it's good to close this.

Everyone's loving the Dyson keynote. Scott Bellski said, "Agree. Found it riveting and seriously considering getting a Dyson pencil vac. " Uh Sam said, "So, what is the pencil vac? " That that Dyson keynote was so wholesome.

Just an excited inventor vacuuming the stage and showing off his new toys all wrapped up in under nine minutes. It's a it's a feature basically uses a laser to expose Well, that's been on Dyson vacuums for a while. Uh I think I have that on mine. Oh, well, you're living in poverty, I guess.

Well, I have a Madic now, so I don't need Yeah, that's true. Madic is is pretty incredible.

But um but Dyson, yeah, they've had the laser for a while, but I think what they did was they re they restructured it because it used to be the internals, like the heavy battery part and the actual vacuum part would be at the handle. And so the handle was always a little bit heavy.

It seems like they've kind of shifted a lot of the weight, the center of gravity much lower. And so you can use it much more just like a walking stick. If you see here, you'll see that what he's actually holding in his hand, if you zoom in, is is very thin.

And so the whole thing is this very thin tube and that means it's a lot easier to vacuum. And so uh get this for your in-laws. Get this for your parents and grandparents. They're going to love it. We talked about lesson was talking about how this is this could basically be the beginning of the e-ane.

This is the smart cane. This is the smart cane. No way. Yeah, he was talking about that. Imagine if everybody was walking around vacuuming all the time. The world would be so clean. The world would be so clean. This is the future. This is the future. And then you throw an LLM in that bad boy.

You start talking to slap some AI in it. Slap some AI in that thing. We could pitch so much AI in this thing. Intelligence and clean streets to meter. Well, that's a great place to end it, folks. Stay tuned. Tomorrow we are doing crypto day. Uh we have a whole bunch of absolute killers from the crypto ecosystem.

It really came together well. The lineup is fantastic. Very very exciting. It's actually I I I wanted to before tomorrow try to estimate the the the total volume that that the people joining tomorrow oversee on a daily basis. It's definitely in the in the billions. It could easily be in the Oh. Oh. Daily volume.

Massive. Massive. Anyway, why don't you take us Why don't you take us out with a massive gong hit, Jordy? I would love to, John. Um, wide is so good. I love it. Swing my microphone. Swing the mic. I don't think I almost broke the uh Anyways, goodbye everyone. Thank you for watching. See you tomorrow. Have a great day.

We will have a fantastic evening tomorrow. Leave us a fivear review. Leave us a five star review. Thank you for watching. Deal fellow Tuesday plus a bunch of other haters. You know, it's great. We'll see you tomorrow. Goodbye. Cheers.