对冲基金经理 David Orr：投资 99% 靠天赋

Your performance still blows me away. 50% annualized, Sharpe 2.1. What are you using AI to do in different parts of your investment process?

My whole way of working has changed completely on the company-by-company level. Lean into this new thing, learn how to use it really well and efficiently, and you probably have a gigantic edge over these guys. Every year, I’ve done a few hundred independent bets. That’s more reps than most hedge fund managers get. I run a 400% levered portfolio, so I have to care a lot about the downside risk on my longs especially.

You were up 9% in July, right?

We just tend to do well when things are blowing up.

David, thank you so much for coming back on the pod.

Yep. Thanks for having me again.

Your performance still blows me away. 50% annualized, Sharpe 2.1. You’re having a great year so far. What are you using AI to do in different parts of your investment process?

I think the risk system’s probably been the thing that’s moved the needle the most. I don’t know if I mentioned this last interview, but I already had this system. It would automatically translate the filings into English and then summarize them, so you’d save a bit of time that way. But now we’ve built the tool, and it’s way better than before. The AI will actually rank the different filings or news items of the day, so now I only have to look at the top five or 10. By the time you get to the middle of the pack, it’s just noise. I can save a ton of time there, too.

What does that risk system entail?

Well, it’s a few layers. My business partner, Sam, basically went out and bought one of the professional ones, which tests it a few different ways. I don’t like that one as much, but the one Mike made is like a custom factor. We can look at correlation to the traditional factors, but we actually have AI now. It’s building these custom baskets of stocks, so it categorizes people’s stocks in the portfolio. It’s funny, because it’ll give them funny names. It’ll be like “Banks Done Right”—those are a long portion—or “Yield Traps” would be a short. Somehow the AI is figuring this out in its own way.

Talk to me about those custom factors that the risk system includes. What does that actually mean? What are you trying to do there?

I’m trying to… Correlation is key, and correlation can change pretty fast. So what this does is update and recalculate the basket every day. The basket size can actually grow or shrink without the portfolio itself changing at all, just because the underlying correlations change. If I’m trying to track this on 10 different portfolio managers, that’s not possible. It’s already hard in my own portfolio. I got pretty good at it, and I got more conservative over time, so there was a big safety buffer. It was basically fine.

But with the new portfolio managers, I need to understand: maybe they thought they were fine at, say, 25%. It’s kind of a big bet in a levered portfolio, but it’s fine. But if that spiked up to 45% because the correlations changed in a week or two, I need to know that quickly.

It’s so crazy to me. How are you able to recruit so many PMs that you perceive to be better than yourself? Your skill is… You’re clearly very skilled. How have you found great PMs? What’s been the process for finding those guys?

The first two, Mike and Rodrigo, were pretty easy. Rodrigo was just posting his results on Twitter, so it was out in the open. I don’t know why nobody else offered him a job first. It was like, okay. Mike reached out to me in 2019 because of my Seeking Alpha posts. I was living in Thailand back then, and he was traveling to Chiang Mai, so he just wanted to hang out and drink beers. We met up and talked stocks. He knew all the stocks, so even back in 2019 I could see he had a real passion. Those two were kind of easier.

For the newer ones, we’ve set up these nets to say, okay, maybe they’ll find me on my Twitter feed, or maybe they’ll find me on the website. I think Sam’s finding them another way, too. I don’t fully understand how. Basically, these track records come in, and 95% of them are total junk. Then, in 5% of them, there may not be fraud in the sense that it’s fake, but we only want live track records. Somebody might be sending it as if it’s back-tested, but it’s real. Another thing we’ve seen is that they’ll have 30 track records and cherry-pick the best one, and somehow Sam’s good at detecting that.

Then all we do is look at the numbers. Once we realize it’s real, we ask them directly, “Hey, you’ve done a 2.1 Sharpe in your personal account for three years or whatever. All we really care about seeing at this point is: is it real? Could you actually pull up your account? We want to see.” We found one guy recently—I’m actually having lunch with him tomorrow—who turned $1 million into $9 million in his personal account. I think he just reached out to us because being part of the team is a lot better than trying to go at it alone.

How correlated do the different PMs tend to be?

The cross-correlation’s 0.15 with the core portfolio managers. So far, even with all the trials, if hypothetically they were equal-weighted, it would still be below 0.2, which is kind of crazy. I don’t know if that’ll hold. It’s not that big of a sample size yet, but it seems like it’ll probably be below 0.3 for sure, so that’s way better than we would’ve expected.

Talk to me about this. You see a track record come in. Say it’s real, on a decently significant amount of capital. What does your vetting process look like for PMs? Is it just running the numbers and then saying, “Okay, this guy’s clearly great. Let’s give him a little bit of capital and see if he can continue to perform”? What does that look like?

We also test a few other things. We test for tail risk, to see if a guy’s just selling volatility, martingaling, or using some fake strategy. We haven’t really run into that yet. We just give them a small sleeve. If anyone hearing this wants to run big capital, please apply to us. We’ll give you a shot, and if you can do what you said you were going to do…

We don’t even ask that many questions about the strategy. One of the new guys described himself as a mid-frequency trader. I don’t really understand mid-frequency trading that well. I guess it’s not high frequency, which is milliseconds. I guess this is more seconds or minutes. But he has some system. We actually have to set up a whole broker for that guy, too, because Interactive Brokers isn’t fast enough for some reason. As long as the sample size is there, we’ll just give him a shot.

Give me your spiel. Let’s say I’m an individual PM, and in my personal account I’ve gone from $1 million to $10 million in not that many years, with not that much correlation to the S&P. Why should I keep running money? Why should I run money with Militia rather than just compounding my own personal account?

We’re paying you—not super early on, but after you’ve proven that things are real for a year, you’ll get paid in carried interest, which means your earnings will basically be tax-deferred. The mid-frequency trader is a perfect example. I don’t even know if this guy realized this. He’s a really sharp guy, so he might have. We actually have a lot of capacity to defer tax with our more fundamental, longer-term strategy. If he’s joining us, we can basically shield his gains. Mid-frequency trading is all short-term capital gains, to be clear. We probably have space to shield all of his gains within the fund. So instead of him paying tax every single year on his gains, we can defer it at the fund level, and then his carry will also defer the gains for years. I think that’s one of the big underappreciated things.

Another thing I learned from poker is that edges go away. You could be a fantastic player in whatever the meta game is—or, in markets, what we call the market regimes—but things change, and whatever somebody was killing it with for five years might go away. I’m not going to kick them out. Eventually we’ll have to return all outside capital for the hedge fund, but the very last people that would happen to are the retired portfolio managers, of course.

If we’re in this engine of guys who are all killing it together, and one of them has to drop out because his edge goes away, I’d have to fire the PM, of course, but he’d keep his capital in this machine, which hopefully would compound it at a super high rate. It’s almost like an insurance policy you’re getting for free, too. There are a lot of reasons. Another reason to join us: we’re building these tools.

We actually have this team, and we’re going to hire more developers. If you had an idea like, “Hey, I want to build this thing, but I don’t want to hire people or manage them,” we can figure it out for you. Managing people can be a little annoying. People are difficult, and you don’t want to do that. So fine, just join us.

Can you break down how that tax deferral works mechanically, real quick?

AQR put out these really good research papers describing how, if you run with both high leverage and high alpha, it makes it very easy to defer tax, for whatever reason. It’s shockingly easy. I’m not even really thinking about it, and since inception we’ve deferred almost all the tax. I could even push it further. If we wanted, we could give up just a little bit of edge, and I could create a tax loss for LPs even while they’re winning money.

What happens is, if you’re long-only, you get stuck in a winning position, and you want to cycle into something new, but you have to pay tax. In a long-short strategy, let’s say a short is losing pretty badly. This happens on a weekly or monthly basis, and you can harvest the loss there and cycle into a very similar short. But if the short is winning, if it’s down 90%, you just don’t have to cover it, maybe ever. That means you don’t have to pay tax until it goes bankrupt and gets delisted and the stock is canceled. These things can take 10 years to play out.

What’s the current size of the hedge fund?

480.

What do you think is the maximum capacity you could run?

It just depends on the number of PMs we can find.

So it’s you and 2 others right now, and then you’re testing a couple more?

7 more. Yeah.

7 more. And do you think each of these guys could run pretty big sleeves independently?

Yeah, at least a few of them, hopefully. I’m guessing at least 1 or 2 will have a big edge in scale, so presumably that makes our capacity 30% higher with just 1 more.

How do the economics work when someone comes in and you give them $5 million or whatever it is to start? How does the scale-up work? I’d love to hear that.

It’s based on sample size, risk, and correlation. The mid-frequency trading guy will have a big sample size pretty quickly. That’s the nature of that strategy. With a longer-term strategy, it might be 6 to 9 months before I get comfortable upsizing for the first time. Then you’d be up to $10 million or $15 million. Another 6 to 9 months after that, and this is sort of the riskiest part for the hedge fund, you might have $30 million or $50 million. At that point, the edge looks pretty good, but the sample size isn’t super robust yet. We still upsize before that happens.

At the beginning, we have to tailor the deals a little bit because each strategy is different, and each volatility profile, for example, is different. So each PM gets a slightly different deal based on that. On average, you get 15% above the risk-free rate on your P&L. That’s roughly what it’ll be.

It seems to me that your model is working. What do you think you understand about building a multi-manager business that others don’t? I ask because whenever you read about these CIO types who try to recruit PMs and raise a ton from allocators, most fail. The most famous one is the former CIO of Millennium. He’s still running money for Millennium, but it’s not what he set out to do. What do you think you understand about your model that a lot of others don’t? Is it your model, or the way you think about things? I’d love to hear it.

1. What Makes the Model Work

I was meeting with another PM in Austin, a guy I respect who’s done really well too, and he asked me the same question. He said, “How are you doing it? I don’t understand.” I think people aren’t focused on the simple answer: What’s this guy’s edge? They’re not willing to look outside and pick unusual people.

One of our guys won a TV game show. I can’t get into specifics because I don’t want to out who this is. He doesn’t have any credentials at all. He didn’t work at any of the big shops or banks. I’m fine hiring him because his personal account Sharpe ratio was 2-point-something for a few years, so we’ll give him a shot. I don’t think this Millennium guy is going to do that. You could be a delivery guy, but if you’re killing the market, I’ll give you a shot. It could be anybody.

Would you say it’s also because you yourself are a compounding engine? A lot of these CIO types just try to recruit PMs and build the business, so to speak, and they themselves tend not to run their own sleeve. Do you think it’s partly because you’re building the business and have room to run, and that can bankroll all the other quant experiments?

Maybe I just understand edge better than the guy who’s just trying to play manager. I mentioned earlier that one of the reasons to join us is that you’re earning carry in a really high-return vehicle. But let’s say they’re just trying to mimic what Millennium does from the outset. If you’re a really talented guy, why wouldn’t you just go join Millennium? Everything else is worse for you, so all the best people go to Millennium instead.

To me, their whole plan is crazy. It’s just like, “Oh, yeah, we’ll become small-scale Millennium.” Who the fuck wants to work for that?

When you recruit PMs, do you pitch them on being able to funnel their own money into the full compounding vehicle? Or do they still like to compound their own capital and run it that way? How does it normally work?

So far, Mike and Rodrigo just want to invest in the fund. Given the tax benefit, I’m guessing any carry that someone earns in the fund is a no-brainer, unless they’re tax-free. I don’t know yet because we don’t have other people. I haven’t even gotten good at pitching this yet. I don’t know what some of these guys have realized on their own or not.

I plan on making a more official pitch and trying to advertise it this way to get talent. There are lots of good reasons to join us, but it definitely hasn’t been my pitch so far. So far it’s been, “Hey, guys. If you want backing and you can’t get backing anywhere else, just come to us.”

There’s something you mentioned in your last answer that I want to zoom in on. What do you think you understand about edge that other people don’t?

I just have a ton of at-bats. Every year, I’ve done a few hundred independent bets since 2019. That’s more reps than most hedge fund managers get. In a single year, I’m getting more reps than a lot of hedge fund managers get in a decade. That’s a lot of experience seeing it. I just got it kind of quickly.

Joe Patter wrote that book where he describes what you’re looking for in a portfolio manager: a guy who consistently wins. These people are out there. You just have to go find them. But the bigger firms seem to give them only a small allocation.

At one of the firms, there’s a guy who’s short-only, and he’s won 40% a year, short-only, for years now. The guy’s a killer, and somehow he gets a tiny allocation at the fund he’s at. If you do that at my fund, we’re going to give you a lot of money. So come join us instead.

40% a year, short-only?

He’s maybe the best short seller in the world, quietly. I heard it doesn’t scale super well, so maybe that’s why they don’t upsize him. It’s unclear.

To give you an even better example, I know guys who are doing a 12 Sharpe ratio, but it works on $800,000 or whatever. Is that the same thing? Not really. So I don’t know to what degree that example applies. But given what he’s done, why don’t you just let him keep doing that and give him a way bigger sleeve, where let’s say he even only wins 10%? I’m sure there’s more opportunity there, given the headline 40% sort of thing.

Do you think the multi-managers right now are making a lot of mistakes by passing up on these guys?

Well, they have a different business. Their business is marketing. I mentioned this in the last interview: They don’t actually care that much about the return. What they want is something that’s extremely friendly to institutional investors. If they can show that all of their PMs had a pretty traditional path, maybe at most there’s 1 oddball. But even then, why would they bother? If we can get a humongous management fee on our...

I think Millennium’s at $100 billion. I actually think Millennium’s one of the better ones, too. There are some pretty crappy multi-managers out there. But yeah, if they can collect $2 billion a year, that’s it for them. That’s mission accomplished. They don’t care about the return going up to 14 or whatever.

What does your talent strategy look like? Obviously, someone can bring up adverse selection and all that. Why would a really good PM join you? Are you exclusively targeting talent that wouldn’t be able to land something at the multis, or is it also people who want to compound capital faster and not get locked up? What’s that part of the spiel?

One of the guys recently did go shop around, and he picked us anyway. Maybe he understood this. I’m going to meet this guy tomorrow, so I’ll talk to him. I’m going to ask him, “Why did you think to join us instead?” This guy probably could have gotten a foot in the door at a bigger shop, so I’ll have a better answer on that.

A lot of the good track records really poured in even in the last few months, so I don’t know that well yet.

Are there any edges you’re happy to discuss, or edges that used to be present that you used to exploit? What do you do well discretionarily that you think others miss?

2. The Edge in Simplicity

I just keep doing what works. When I first started short selling, I was trying to short everything. I was even shorting Tesla in 2018 or whatever, and pretty quickly realized this short just wasn’t working as well as the other ones, so I stopped. I just keep doing more of what works.

It’s kind of hard because you have to go against your nature. I used to want to do more deep research on a stock, too. It’s interesting and almost fun as recreation. But there was no edge, and I noticed it actually created bias.

The key variable in an investment thesis can change. I’d already poured 100 hours into this stock, Gogo, the in-flight internet company, which I’m still short. It’s now $2, so fundamentally I was right. But it went up to around $20 in the middle there, and I got hit on the way up because I knew the fundamentals of the company were bad. I stayed wrong because I’d put in too much research time, even though there were short-term reasons I probably could have escaped better.

How much research do you think one needs to do on a name to actually decide whether it’s going up or down?

Early on, maybe it’s worth diving in a bit more to train yourself. Hopefully you start picking up on what matters and what doesn’t. This is more advice for experienced guys who’ve already put in 1,000 or 2,000 hours of that. I’d say try to decide within an hour or two. If you can’t decide in 2 hours, just skip it. Why bother? There’s something easier out there.

I always think of it like this: I’ll wrestle with something for 3 hours, finally say, “Okay, I guess it’s too hard,” and give up. Then I’ll go to the very next idea and decide in 2 minutes. I’m like, “Oh, great. That’s the bet.” And then the next one, you know? You’ve got to really think through how you’re using your time.

It’s great when you decide on something in 2 minutes. If the stock starts going against you and you start seeing a reasonable-sounding case for why it’ll keep happening, you just quit. You don’t even think about it then. It’s another 5 minutes or whatever.

Today, when you look at a name, how long does it take you to go from not knowing that much about it to deciding it’s a long or a short?

Yeah, within an hour every time.

How many positions do you hold?

I have 500 in my partition.

How do you have the mental clarity to understand what’s happening with all of them?

On shorts, I actually forget what they do sometimes. I’ll be reviewing my shorts, and let’s say the best short is the one that’s down 40%. I don’t even remember the company or what it does. I don’t remember the description, but the short is still on.

Then I’ll remind myself for a minute or two what they’re doing. I’m like, “Oh, okay.” Then I’ll see that earnings did go down, or whatever bad thing I thought would happen did happen, so it sort of refreshes in my brain. Usually I’m like, “Okay, I’ll short some more.” This whole thing took 2 minutes, and I’ll probably have forgotten again by the next time I review it, 6 months or a year from then.

The worst shorts are the ones that just keep going up every day on no news at all. I look at it and think, “This thing sucks so bad.” There’s not even news. It’s not like management’s being promotional or whatever. Every day the thing is grinding up. I’ve gotten better at quitting those, but once in a while I still get stubborn because it’s so bad. Sure enough, before you know it, the thing’s up 250% on no news at all. It’s crazy. Why wrestle with it? The market said I’m wrong; I’m wrong.

Do you have a writing practice to keep tabs on everything? With 500 positions, do you have notes on every single name?

No notes.

It’s all in your head.

Yep. That’s it.

Is there anything you’ve learned, any processes or rituals from the other PMs you’ve worked with that have yielded dividends for you?

No, we try not to get correlated, so we deliberately don’t talk about that at all. It was nice; I got to see them recently. Rodrigo flew in from Spain, and Mike lives in Austin now. We all flew up to New York together, so that was a really fun, enjoyable experience.

We really avoided talking about this even while we were there, because keeping the correlation down is way more important than our edge being good enough or whatever.

Since the last time we spoke, what’s one lesson you’ve learned?

Relying on other people has helped a lot. I normally did my own thing completely, but watching these other two guys, these other core portfolio managers, they’re making systems that make what I do way easier.

Mike made this risk system this year, which is saving me so much brain damage, even in my own portfolio. Now we’re backing 7 new trial portfolio managers, so it would be completely impossible to track all 7 guys’ risk without this tool. I’ve realized that actually using a real professional tool built for what I do has been really helpful.

How do you keep mental focus on building the business, hiring, and doing stuff like this, and investing? Five hundred positions occupy a lot of mental space.

Yeah.

How do you do it?

Sam takes care of almost all the hiring, and we’ve got Luis, who handles a lot of the brain damage, accounting, and stuff like that. It’s actually become easier now because Mike made that risk system I mentioned, and he did that with the staff Sam had hired.

Sam hired great people, and Mike and this guy Andrew would work until 1:00 a.m. They did that for about a month straight, building this risk system. At the end, it was done. It was great; this thing was saving me a lot of time compared to before, even for my own partition.

Before, I’d have to manually track things like the old media stocks or the fiber-optic stocks, or whatever the short category was. Some of them were straightforward and didn’t change. Charter would be bucketed in with the internet thing. But some were more of a gray area. I don’t want to talk about specific companies. Some would jump around between categories, or maybe some would randomly become more of a rate-sensitive play because the market decided.

Now I don’t have to guess or move them from bucket to bucket. That used to be part of my process: I’d flip through the charts and say, “Okay, this one’s, for whatever reason, trading with this other thing now.” Our new system just does that for me.

Do you tend to like buying on the way up versus on the way down?

Yeah, I think you want to buy things that go up. Rodrigo—I mean, it’s very counterintuitive, but that’s what Rodrigo does. He looks at stocks making new 52-week highs or 52-week lows, and then he’s like, “Oh, okay, I could see why this will keep going up. Buy. Ah, yeah, I could see why that’ll keep going down. Short,” sort of thing.

It’s a judgment call, because obviously you can’t just do that as a bucket. That doesn’t work. You need to get the best 5% of each of those buckets.

So this is just one of his idea generations. He has a bunch, but that's one of them. I do the same thing. I don't like to. Mostly, I get stubborn on a few. I'm in this Hikari Tsushin that's been going nowhere. I think it's even down year to date, and it's my biggest long. But I really understand the fundamentals on that, and it's not like a risky stock at all. So am I making a mistake there? I don't know. I've done this only a few times before, and the few times I've done it, it went really well. I think it's an incredibly cheap stock, but mostly I'm not in stuff like that. If I am in something like that, I'm almost saying, “I'm so confident that the market's wrong, I'll actually bet bigger than normal.”

I asked the question about buying on the way up and buying on the way down, or shorting on the way down, because in a previous podcast, one of my guests, Alix Besquet, said that the best stocks are always making new highs. Another thing he mentioned was older investment books. Do you think a lot of that stuff is good for a beginner starting out, or is a lot of it out of date, or can you learn the wrong lessons, so to speak?

Yeah. It's like poker books too. They're the same thing. They're all out of date. Maybe if you read a Peter Lynch book or whatever, that's okay. Most of these books, I feel like the ideas are just really bad. Who even knows how well the anomalies they talk about worked, even when they wrote about them? There's that funny joke: somebody will be killing it for five or whatever years, and then they'll write a book about it, and then of course it starts getting hit. And then not only that, but these things are already very old too. I think books are mostly a waste of time. I think reading business books is good, but that's not an investing book. I like Shoe Dog. It's about the history of Nike.

It's a great book.

Great book. You can see how the mindset of a great CEO at a growth company is just constantly pedal to the metal. You learn about really subtle things in these books. Part of what made him such a great CEO is he didn't let the banks limit him, right? He found creative ways—

Mm.

Right? That's the kind of thing to ask: Is my CEO that I'm betting on doing something like this? Is he thinking outside the box? Is he willing to push the bankers, push back, and shop around to 30 different banks? That's the kind of subtle detail you should try to pick up on in those books, and then try to pick up on in your longs too. Or shorts, for bad businesses.

Do you have a reading process or a learning process? How regimented are you? Are you just going based on vibes, following your curiosity? Would love to hear.

I read not even that many books. Maybe 10 or 15 a year. I'm pretty careful about which ones I pick up. If I'm not getting it or enjoying it in the first chapter, I just put it away. I'll usually just give it away. I tend to do it in bursts. I'll not read a book for six months, and then I'll read four back to back in one month or so. I don't know why. There's not a deliberate reason for that.

What are the specific—do you only read business books? Do you read fiction? What's that look like?

I don't like fiction. I like history books. I guess mostly you could say they're still sort of about business. I like Kitchen Confidential, for example. That's sort of about the restaurant business in some ways. There's a chapter where he describes the really successful restaurant runner in New York City, and the kind of neuroticism, the crazy details he would attend to, like not even believing the shipping label and weighing the thing himself. That's the kind of dedication it took to win at the restaurant business. But for non-business books, I don't really... I'm trying to think. I can't even remember the last non-business book I read.

A thing that comes up when I speak to people, particularly young people, and I would agree with this, is that no one reads. Everyone's addicted to their phone, and no one really thinks. There's all this stimulation. I think you've been able to use all the tools clearly very well for your process, and you're still able to think very clearly. What would you say to young people who, in my opinion, lack analog training for thinking very deeply because they've grown up in a slightly different world? I know you're not that old, but how would you think—

Mm.

—about building up the brain to be good at picking stocks, investing, or just business in general?

3. Learning in the AI Era

I don't know. I don't even know if the old way is better. Maybe the new way is better. Let's say somebody suggests a book to me. If it's a person I trust and it seems like a great book, I'll still read it. But nowadays, if the guy suggests a book and it's not that interesting, you can go to ChatGPT and ask it, “Can you summarize the book in bullet points?” Then you can pick out the key bullet points that seem interesting and relevant to what you might learn for investing, and you can have a back and forth with this machine about it. You can maybe learn the key points of a book this way in an hour instead of reading a whole book. I'm not sure if that's better or not. I don't think anybody knows.

I know that there's always been an anti-technology bias. Even when the alphabet first came out, there was a moral panic about what this alphabet thing was going to do to people's brains. It's crazy, but it turns out they were dead wrong. Reading's one of the greatest things you can do. In my lifetime, there was the internet. People were like, “Oh, the internet's so bad. It's not even real. What, people on the internet, as if they weren't the same people in person?” It's just a better way to communicate, a better way of doing things.

But it's kind of early with AI. I don't know. I think you should still read some really great books and try to balance it somewhat. I don't really have an answer.

That's actually interesting you say that. I forgot which book I was reading, but it was written in the '50s or '60s. The way they were talking about the radio was very similar to the way we talk about Reels—

Mm-hmm.

—and short-form content.

Yeah.

Which I do think short-form content tends to be poison a lot of the time. But—

Yeah, I think the TikTok thing or whatever, that's hard to make an argument for.

Yeah.

The same way with AI, you know? Those are really trashy videos. And what they're doing is they're not even aiming to make good content. They're just getting the right hook, right? Apparently, they study the first second. How do you even get somebody locked into this thing? That's what they're optimizing.

Yeah.

There's nothing there. That's empty. But for ChatGPT, I'd say things are way more likely than not to be positive for learning, and I would probably lean in. It's also a way to differentiate yourself. Let's say you're competing with someone. I always grew up with technology, but let's say some of these people out there have done a lot of reps with this more analog, slower way of thinking. Do you really want to compete with them, especially if your heart's not in it? No, you're not going to. You're going to lose.

What you have as an edge as a young person is to lean into this new thing, learn how to use it really well and efficiently in an honest-to-yourself kind of way. Focus on that, and you probably have a gigantic edge over these guys.

How do you think about building personal edges today? This is something I talk a lot about in podcasts: building moats. In another episode, I had a guest, Lihong Wang, and he was talking about two ways to think about it. There's moat, and then there's just growing super fast and being ahead of the curve. How do you think about building an edge in either one of those two things?

Today, you just want to focus on growth. I wouldn't even focus on the moat thing. Nobody has a clue what's going to be sustainable. This whole thing is changing way too fast to even be thinking about that. So, really, you want to be pragmatic: How can you use this technology today to make things better now, or improve on it within a month or two? That's the key thing to focus on.

I’m talking to some of my friends, and they’re like, “Man, the thing it does even now compared to 3 months ago is jaw-dropping. This is changing so fast.” So just keeping up with that and being the best at it. If you’re incredibly good at AI and you can show someone, I’m pretty sure you could approach some of these giant companies and be like, “Pay me $5 million a year or I’m gonna go to your competitor.” Eventually they’ll probably say yes if you’re really great at it. Then you have retirement money in 2 years or whatever, and who cares if you even—You know what I mean? At least you’ve taken out your downside first.

What’s the craziest use case you’ve seen?

I don’t know. Do you mean from my own use?

For your own business processes or investment process, what’s the highest-value thing you’ve gotten out of it?

Just being able to take my normal, follow-the-news kind of flow and save 90% of the time. Now my brain gets tired because it’s so much information so fast, whereas before it was more busy work, going through the motions. The fact that my whole way of working has changed completely at the company-by-company level is pretty insane. Compared to 2 years ago, it was completely different.

What are your thoughts on current AI stocks?

4. The AI Trade’s Risk

I don’t know. It’s got a kind of binary outcome. You have the hardware producers, like Taiwan Semiconductor and NVIDIA, or the memory companies. You have the hyperscalers or the neoclouds. Then you’ve got the consumer side, which today is mostly a lot of software companies.

I like the hyperscaler piece the best. It’s got way less upside in the good outcome. But the clever thing they did was structure the deal with the end buyer—mostly software companies or whoever—to say, “Okay, we’re gonna have a 35% return on capital at the price, and you’re gonna lock in about half the compute use on average.” If you do the math, there’s basically almost no downside. Even though these guys are buying a ton of compute and it looks like they’re taking a huge risk, it’s actually their customer that’s locked in. It’s almost like a take-or-pay contract, right?

If there’s a big bust, if way too much compute gets built and the price of renting compute drops by 80%, which could easily happen, it’s not the hyperscaler that’s really gonna lose. I think they kind of break even on the deal. It’s their end customer. All those guys will probably go bankrupt, or if they have debt or whatever. That’s a hell of a liability, right? You said, “Okay, we’ll lock into buying $5 billion or $10 billion worth of compute.” The value of that goes down 80%, and their business is hosed on top of it. That’s it. That’s lights out.

The people who make the compute, of course, make the most money in the bullish outcome. If the memory thing keeps getting tighter and the price of memory keeps going up, they’re gonna print an insane amount of money. But they have a lot of downside too. If the whole bust happens, then their revenue’s gonna drop 80% or 90%.

When I look at the overall landscape of what to bet on, my intuition is maybe own a little bit of that first bucket. I like the hyperscalers. I like Google, Amazon, maybe Microsoft, and then maybe shorting the ones taking a huge risk. But even that’s tricky. Maybe some of them are improving their businesses a lot. They get to fire 90% of their workers and keep this wide-moat business, or who knows. A lot of it just seems too hard.

But the hyperscalers are in a position where I feel like it’s heads I win if, let’s say, compute stays tight for 10 years. They’ve only locked in half the contract with that 35% return on capital. But if the other half of the contract is open-ended and the price keeps going up, they’re just gonna have a 50% return on capital. Remember, they’re investing hundreds of billions of dollars, so they’ll make a ton of money. I like them the best because I don’t like stocks with huge downside risk. I run a 400%-levered portfolio, so I have to care a lot about downside risk on my longs especially. But if you’re in an unlevered or lower-levered portfolio, and you have a strong opinion that memory will stay tight for 3 years, that could be the better bet too.

And your correlation to the AI factor is—

Zero.

…zero, right? Was there ever any moment while the run was happening where you thought, “Hold on, the way my fund’s allocated, maybe I could take a punt here as well”? What were your thoughts then?

I had a long position in Taiwan Semiconductor on the way up, and before that I played some other AI stocks on the way up. But it wasn’t a huge bet. I’m not betting the farm on this.

Before April, it was a way different risk-reward. Taiwan Semiconductor was about 20 times earnings. It’s a pretty good business. Now the earnings have gone up, but the multiple hasn’t gone up that much because earnings are growing so fast. It has all this new earnings growth too, so if growth reverses, there’s way more downside than in March. I sort of cut all that exposure in June. The risk-reward changed in the AI hardware space. I did ride it up, and I think there’s still a very credible bull case, but it’s just not my game now.

You were up 9% in July, right?

Yeah, the fund was up something like that. I forget exactly.

Can you break down how the fund was positioned, and why that was such a good month when it was awful for everyone?

We just tend to do well when things are blowing up. Really, we’ve been fighting the tension all year. This is a nightmare market for me. This is like a 1999 kind of market. When I posted about inflation-protected TIPS, when those are going up fast, that’s really bad for what I do. For some reason, this goes up during a technology shift. It went up during the dot-com era too. The whole environment’s been a really bad headwind for what I do. But when it’s reversing, of course, we should kill it.

I wanna talk more now about the broader business of what you’re trying to build. We’ve touched on the multi-managers. Talk to me a little bit more about the broker business and your vision for Militia.

5. Building Militia’s Business

The broker business is as simple as this: we run perpetual, high short. The first step is just hiring a guy who goes out and finds a way to borrow stocks for us to use internally. I don’t even know exactly what that takes, but I know that paying this guy can’t be that expensive. I’m pretty sure if we can cut out the broker on some of the volume we have, we could probably save a million dollars a year. That’s a pretty rough guess. I don’t know if that’s exactly right, but it seems hard for it to be a bad hire even today.

Given our growth, if our AUM doubles, then that savings might be $2 million a year. The guy’s salary might go up a bit, but it’s not gonna double. This is just scaling the profits of our hedge fund, basically. Also, that’ll come out of the management fee, which is good for LPs because the cost is coming out of that. I have this ego with my hedge fund. I just care about the track record, so it’s a way for me to make the track record way better that way.

The net-of-fees track record.

Yeah. I care about the net-of-fees track record.

Are you closed?

Yeah. The thing is, if a few of these new PMs work out, then we’ll probably need to find a way to get them some money. We have a long waitlist and everything, but if you wanna join that, you can reach out to the fund. We really don’t need capital, though. We sent one email to existing LPs earlier this year. We needed to raise capital to get onboarded to a broker. A few brokers said, “We want you to get your AUM higher, and then you’ll be one of the really big ones.” I was like, “Okay.” A $300 million fund is still kind of small for these huge global brokers. So I said, “Okay, fine. We’ll go raise some money.”

We sent this email to LPs that said, “Hey, we wanna raise $100 million,” and $150 million flowed in from that single email. Fundraising has been shockingly easy. I don’t think there’s any trouble there for us.

$150 million after a couple emails?

One email.

Wow.

Yeah. I think if I focused on raising money, we could go raise $1 billion. But I don’t want to. I’m focused on getting the track record as good as possible, so I’m optimizing for that.

At some point, you’re gonna get huge too. I mean, if you maintain your current rate of returns—

I don’t think we will.

I think we must have gotten lucky. I’m not claiming that we’re going to compound at 50% after fees a year. No, we must have gotten at least somewhat lucky.

What do you think is the natural cap for a business like yours, with the current number of PMs you have and the ones who are being looked at?

I don’t know, because more strategies keep popping up. One of the guys who joined recently is doing a managed futures kind of thing, and I don’t really know what to expect, but so far, so good. The guy’s running at 10 vol, and he seems to just keep grinding out a win. He’s betting on all sorts of weird stuff. I don’t even think about this. His correlation, of course, is going to be zero. So far it’s actually showing as negative, but I’m guessing it’ll be zero long term.

How many more people like this will show up? I have no clue at this point.

And you take any strategy whatsoever?

Yeah. We care about the sample size of bets. Do I believe that these are actually different bets? That’s a really key focus we have. Sometimes you can say, “Oh, this is all ultimately the same factor bet.” That’s not going to count. But if a guy’s like, “Hey, here’s 300 bets, and I keep winning,” we’ll say, “All right. Well, we’ll assume you’ll keep winning, and then if it ever changes, we’ll reevaluate.”

And what vol do you typically like the PMs to run at?

I don’t actually care. I care more about correlation, so in that case, I’d probably ask him to double the gross exposure, or at least. But while he’s tiny and I’m just seeing what the hell is going on, I sort of like that he’s demonstrating, “Hey, I can control my risk,” however he sees fit. Not that I’m even controlling it exactly, but to me it shows that he’s very risk-aware, which is probably why he’s even doing it this way.

He told me, “If you want me to double the vol, double the risk, I can do it tomorrow.” It’s like, “Okay.”

What’s been the most shocking track record you’ve ever received, in a good way?

In a good way?

Just something that blew your mind.

I don’t know. There are a few. This guy Grant runs a fund called Latham. I actually tried to convince him to join us. He was before Mike and Rodrigo. I was trying to convince him to join us, and he just decided to launch his own fund instead. Part of it was because he had a ton of friends and family money, and there were some other reasons.

He launched his fund, and I told my friends to invest in it. I think he’s actually open, by the way, so you’ve got to reach out to Grant if you’re listening to this. He launched after I tried to get him to join us, and three years in, he’s done a 2.1 Sharpe. He’s up 400%.

Wow.

And he’s only using 200% gross. I use four, he uses two. That guy’s a killer. He’s so good.

Do you have an insane network of PMs now, just after putting yourself out there and saying you’re open for anyone who’s very talented to try it out? What’s that like?

Yeah. People just reach out. Some of them are not as good as they think, because we actually just care about your total return. Let’s say a guy’s reaching out with a pretty good short track record. It’s only losing a few percent a year for the last few years, but I’m thinking, “Well, the return on equity here still sucks.” We actually want to generate a return, and we could figure out how to balance his short book with some longs and do that thing. Maybe then it would be more like a 20% a year return, but we’d be having to wrestle with half his portfolio for him.

It’s kind of sad to me when people box themselves into these styles. It’s almost like they’re doing it because it’s more marketable to professional allocators. But I’ll stress, if you didn’t go to Harvard or whatever, they’re not going to give you money. They’re going to lead you on for years. I know people who’ve generated a return like that for years, but they’re not going to—yeah, so it’s just a waste of their time. It’s kind of sad to see. I’m like, “What are you doing?”

What are the typical… Do you need to understand what someone’s doing specifically, like the ins and outs of their process, or are results enough for you?

I need to understand the risk, too. Usually that’s pretty straightforward. You can see—let’s say they’re long orange juice futures or whatever. I don’t know anything about this, but I can pull up the chart and see, “Okay, how volatile has this been? What’s the craziest thing that’s ever happened in history?”

Let’s say that happened in his partition, and it happened on a few of his commodity bets at once. In that case, you can reasonably understand what the risk profile is. I don’t need to understand anything about cattle futures. I just need to understand that his risk management is obviously good.

On a daily basis, what are you monitoring with the new PMs who come in? Is it their cross-correlations? Is it whether they’re behaving out of sample as they did in sample, based on what they’ve shared with you? What are you looking at?

I care that they’re doing what they said they would do, or what their track record showed before. If there’s a mismatch, it’d be like, “Okay, this is BS. You’re fired.” But that hasn’t really happened.

I don’t actually follow it daily or anything. Once a week, I go check our correlation matrix. It shows each portfolio manager’s correlation to every other portfolio manager. I’m just looking at it. You can adjust the timeframe on that now, which makes it really easy for me, so I make sure that it looks pretty good. Maybe not even weekly—maybe every couple of weeks.

When I first get a PM on, I try to go through the positions. The first day or two, I’m trying to understand them. I’ll follow up after a few weeks to make sure they’re not dramatically changing things. Then, at random, in another couple of months, I’ll do the same thing. After a while, I get comfortable with it.

The big risk thing I do is check the fund-level P&L. I try to look at the biggest loser on the day and what’s going on there, not just at the fund level, but at each partition level. Does he have some horrific 8% loss or something? That hasn’t happened, but I’d look for something like that in one position.

Are there any personality traits you see in the PMs you work with that are predictors of success, or are they all sorts of different people?

6. What Makes Great PMs

One thing I was talking to Mike about recently is that if a guy’s looking for outside affirmation, looking for someone to tell him, “Oh yeah, what you’re doing is the right thing,” that’s a kind of tell for groupthink. People who do this tend not to perform well.

In hindsight, I can tell my mentor was testing me. He’d push back in ways that were very rigid, like, “No, this is wrong,” and “You can’t do this this way.” I was always thinking, “I’m going to do it my way. I don’t really care what you think,” when I was sure.

He was a great teacher because he’d teach me the rules of the market, or tell me, “This is a terrible risk mistake,” early on. I could understand that. He’d explain why, and I’d get it. But if you push back in a way that even kind of made sense, that was common sense—like saying nobody had been able to do it the way I’d done it—that didn’t mean it was wrong. You just had to ask, “Well, why has nobody done it?”

Who’s your mentor?

Rob. He lives in New York.

What are some of the most important things you’ve learned from him?

Risk management. I’m a better large-scale short seller now. He hunts these niche, small things. He has a really high win rate doing this kind of stuff. His way of doing it is investing in lots of other funds rather than growing his own portfolio endlessly. He’s kind of doing a multi-manager approach that way, with his own capital super diversified.

He taught me that bet sizing was key. He taught me not to focus too long on any one stock. He taught me that narrative would drive me crazy. I’d be like, “Man, this thing sucks so bad. The thesis is totally on track. Why did the stock go up 10%?” He’d be like, “Man, revenue’s going up. It’s so good for the company.” And I’d be like, “No, it’s not.” I know now that he knew it sucked, but he was just so matter-of-fact.

“No, no, it’s great for the company.” The point he was making was, “Look, revenue’s going up, man. The stock trades on a revenue multiple. It’s not gonna stop just because you figured something out. Why don’t you wait for it to break down from that, and then short it? Why let the market run you over?” That kind of thing.

If there was a guy who studied math, didn’t know anything about markets, but found out he was extremely passionate and genuinely loved this game—not just because he was interested in prestige, this, that, and the other—and he was in front of you, how would you mentor him to make him an exceptional PM? What would the process look like to teach someone?

I’d be really hands-off, mostly. I’d say, “Look, just generate an idea, and give me a new idea every day, at least. Maybe twice a day. Just turn out ideas over and over and over again.” I’m probably not even going to want to comment on a lot of them, but maybe I’ll notice something in some of them, like, “Oh, okay, here’s what you’re missing: the whole revenue growth thing.” Or, “Oh, that’s a really particularly good idea,” and I’m going to copy trade you. Good job.

That’s what Rob would do with me. He’d be like, “Oh, I’m gonna put that one…” Actually, he didn’t even tell me when he was putting it on most of the time. He’d just tell me after the fact, like, “Oh yeah, I totally joined that one too.” So yeah, I think just letting them learn through experience.

How do you make someone generate real ideas, instead of just seeing a news article or a tweet? I’m not saying you should never get inspiration from other people, but how do you build the mind to come up with its own stuff?

I would just tell them a few times, “Look, you gotta go find your own thing. I can’t keep repeating it, so you’re either going to or not.” A lot of people don’t even have the self-control.

One thing I noticed is that a lot of short sellers mostly short crowded shorts: the high short interest stuff, or maybe the short interest isn’t even super high, but it’s just obviously bad and trading not on fundamentals, let’s say. They can’t seem to get over it. They can’t get over how bad and dishonest management is, or how terrible the business model is, and they can’t learn to focus on the key variable.

I just met a pretty well-respected guy on Twitter who does a lot of his own thinking and never copies people. I had dinner with him a few nights ago, and he said the same thing. He said to me, “Your ability to hone in on the key driver of things, that’s your edge.” He told me, “I don’t think it can even be learned.” So I think a lot of it is innate. I don’t know if that’s sad or not. It’s just how it is.

Do you think success in most domains is, like, 98% talent?

Yeah, especially at the outlier level. You’re not gonna train somebody to be as good as Michael Jordan. That’s just not possible. He worked really hard and everything. He’s quite a bit better than he would’ve been, but 99.99% of people won’t make it to that level anyway. It’s not possible.

If you were to break down what percent of the variance in investing is explained by talent versus, say, reps or diligence?

I’d say it’s almost all talent for the big alpha people. Maybe for the pod shop, they’ve quantified it better, I guess, because they actually do kind of work like a talent factory. They bring in a new grad or whatever and say, “Here’s how to be an analyst. You’re gonna be an analyst for five years or whatever. Then we’ll give you this small sleeve of capital, and you’re gonna…”

They’ve sort of idiot-proofed it so the guy doesn’t even have to learn risk, almost. It depends on the shop. Some of these shops are pretty bad, or even their risk model’s bad, but that’s a whole other story.

They just kind of idiot-proof it. It’s like, okay, you can use the analyst skill set you’ve trained to do. This kind of alpha’s, say, 3% unlevered, and then they’re like, “Well, we’re gonna lever it six times,” so now you’re at 18% alpha. Maybe that one’s teachable. I think probably. I bet a lot of people still fall off, but it’s not as dire as I’m saying, where it’s one in 10,000 people. Maybe it’s more like one in 50 or something.

Last time you were on the podcast, you talked about how, if you were to start investing, you would just start making bets and see if you could actually do it for 14 hours a day.

Yeah.

Right. I want to ask a question like that, but framed differently. If you really wanted to learn investing and you had the pedigree—let’s say you’re the Harvard guy, and you have the opportunity to go to a pod shop—do you think it’s a great place to learn? Is it a place to grow your skill set? What are your thoughts?

At that point, yeah, I think it’s worth trying, as long as you can keep an open mind. Let’s say you don’t make it after two years and they show you the door. If you can mentally reset, you could always just, you know, now you’re 24 or 25 or whatever, go try doing it my way after. That’s fine.

A pod shop’s cool because, let’s say you get your foot in the door, you work, they make you a PM, and you’re generating tens of millions of dollars. It’s pretty good. I’d say you probably have a better shot their way, to their credit.

The question’s kind of weird, because I swear I got started investing late only because I believed my economics professors that the market’s efficient. But let’s say you know that this is wrong and you’re 17. You could just have an account at 17 and already be doing it my way before you even get the opportunity to work at a pod shop. So just start then. There’s no reason to wait.

With the PMs you recruit, what’s their typical path? Is there a typical path? Is it all different? I’d love to hear it.

It’s all different. They’re all self-passionate. That’s the typical thing: they all really care about whatever style of investing.

One of them has a more traditional job at a financial services company. I’ll keep it kind of abstract, but he has a pretty good job, and he realized, as he describes it, “This is not a serious place. These people don’t even care about the investing returns.” They’ve got billions of dollars to allocate. He does his job, but luckily, as part of it, they let him run another account. So he developed his own strategy on the side while realizing, “Well, these guys don’t even want my strategy. They don’t even care.” In that case, he was able to hold down a very traditional job and run his personal account at the same time.

The Grant example: he races cars. He’s into car racing. Rodrigo is the government worker in Spain.

Wow.

I was a poker player. Mike had a traditional background. He worked at a hedge fund. Who else? The game show winner. One of the new guys is a programmer. He’s just a really geeky programmer kind of guy, in a good way, and for some reason he thought, “I’ll get into this game.”

One of the guys has a really bizarre background. He’s a European guy. He was on vacation in Spain, met this American girl there, and they got married and lived in Spain. She got kicked out after a year for visa reasons. He flew to America, and now he’s in America. It’s a totally bizarre life path, and this whole time he’s just trading his personal account, killing it. I take that whole weird story arc as a positive. I’m like, “Great, the guy’s totally different.”

No, I agree. It’s the same for you. When someone has a differentiated path that doesn’t make sense, you know the result is due to skill.

Yeah.

Right? It can only be explained by skill.

Yeah.

I guess this is a final question, summing up everything you’ve learned and everything we’ve spoken about. It’s about building personal differentiation. It’s a question I like to ask a lot, because investing is ultimately about the quality of your ideas. You need to have differentiated thinking, truly, like you were saying. What are the ways people can build personal differentiation today, when narratives are constantly flying around and it’s very easy to think like everyone else? Any tips?

I just unplug more, because you don’t have to… On a site like Twitter, you can actually choose who you’re following. Or you can try to skip that whole For You tab thing, in this example.

Just like, okay, these are 100 insightful people. They're not just part of some dumb hive mind. Don't go on Reddit. That's a really bad place. Skip TikTok and everything. Find things that are really interesting to you, that you're really passionate about doing. Then put it through the filter of, well, what's some way I can make money at this kind of thing? There has to be a money component in capitalism.

And then the final component, which ties in with the money part, is: how is this going to help somebody, right? So I look at your podcast. Somehow you've found a great show. You've found great guests. You're running the best investing podcast today. I've turned down a lot of podcast requests, and I think your show gives the best guests, so somehow this seems to be your passion. You seem to be doing good at it. That's pretty much it. Whatever the ikigai, the Japanese thing: the mix of what makes money, what you love doing, and what was the other one?

What's valuable.

Yeah, what makes money, so—

Yeah, what you love, what you're good at, and what makes money.

Yeah.

No, I fully agree. Thank you for the compliments about the podcast.

Yeah, it's a great show.

And I've been personally very surprised that you can literally interview people who you would've wanted to speak to anyway and make very good sponsorship revenue. It's been crazy to me. And this is crazy: we're going to have dinner right after. You're someone who I wouldn't have been able to meet otherwise.

Yeah.

But thank you very much for coming on the pod for the second time. This was awesome, David.

Yeah, I had a good time. Thanks for having me.