The Harsh Truth About Investing Edge with Adam Butler

[00:00:00] Welcome to Excess Returns, where we focus on what works over the long term in the markets.

[00:00:04] Join us as we talk about the strategies and tactics that can help you become a better

[00:00:07] long-term investor. Jack Forehand is a principal at Validia Capital Management.

[00:00:11] The opinions expressed in this podcast do not necessarily reflect the opinions of Validia

[00:00:15] Capital. No information on this podcast should be construed as investment advice.

[00:00:19] Securities discussed in the podcast may be holdings of clients of Validia Capital.

[00:00:22] Hey guys, this is Justin. In this episode of Excess Returns, we sit down with our good friend

[00:00:26] Adam Butler. When we published our episode with Med Faber a few weeks ago that focused on

[00:00:29] things Med believes about investing that most of his peers would disagree with, we asked our

[00:00:33] followers on Twitter to respond with some of their own. The most thoughtful response we got

[00:00:36] came from Adam when he wrote the idea that an edge is indistinguishable from noise over a typical

[00:00:40] investing lifetime. We found that so interesting that we wanted to have Adam on to discuss it.

[00:00:44] We dig into that question in detail and cover many more interesting topics,

[00:00:47] including return stacking, the impact of passive investing, whether AI will boost economic growth

[00:00:51] and a lot more. Adam always brings a thoughtful perspective to the topics we

[00:00:54] discussed with him and I think you will find that he certainly delivered that in this

[00:00:57] episode. Thank you for listening. Please enjoy this discussion with Adam Butler.

[00:01:00] Adam, welcome back to Excess Returns. I love coming on with you guys. Jack always sends me this

[00:01:07] list of topics and questions in advance, which is very thoughtful

[00:01:12] and they're very thoughtful questions. So I'm excited to dig into what we're going to

[00:01:17] talk about today. That's why we love having you because we can almost throw anything at

[00:01:21] you and we know that we're going to get some thoughtful insights and maybe some

[00:01:25] contrarian views on certain beliefs or things that investors might be thinking or believe in.

[00:01:33] Today we're going to tackle the idea of alpha and edge. We'll get into return stacking and what

[00:01:41] goes into that. You guys are seeing some nice growth and the ETFs that are utilizing

[00:01:45] that investment strategy. Then hopefully toward the end, if we have time, we've had a couple

[00:01:52] guests on where these interesting things are coming up in terms of market structure and

[00:01:58] what's influencing the markets. Hopefully we can get into to sum it out with you because

[00:02:04] I think you'll have some really interesting thoughts there. But we wanted to start with,

[00:02:09] so we had Meb Faber on recently and the whole entire podcast was about going through a set of

[00:02:18] beliefs that he has that most investment professionals or most peers would disagree with.

[00:02:24] And in response to that podcast, which was great, you threw this up on Twitter,

[00:02:29] and I'm not going to read the whole tweet, but it's around an edge in investing.

[00:02:33] And the first sentence of that tweet was, the edge is indistinguishable from noise in your

[00:02:39] lifetime. So let's just start there about investing edge and what you meant by that.

[00:02:44] Yeah, so edge is I guess skill and it probably correlates pretty strongly with what people

[00:02:54] typically think of as alpha. But it's the idea that you're able to identify

[00:03:02] mispricings in markets and generate profits in excess of

[00:03:08] what you can get from just investing in standard benchmarks. And yeah, my point with that is that

[00:03:17] for any given edge where, and let's say we'll call an edge in stock picking,

[00:03:27] you know, there's maybe an edge in trying to emphasize stock to your portfolio with

[00:03:33] high cash flow yield or high shareholder yield or positive momentum, that sort of thing.

[00:03:42] That the size of these edges are so small relative to the noise around those edges that we

[00:03:52] experience every day in terms of their relative gyrations to the underlying indices

[00:04:01] that it's just very difficult to make informed choices in advance

[00:04:09] that you can have high confidence. In other words, you're going to choose which edges you want to

[00:04:15] allocate to in your portfolio that you're going to have a high confidence

[00:04:19] will have outperformed, you know, a random set of other edges that you might have or strategies

[00:04:27] that you might have selected instead. If you were to observe the performance of this,

[00:04:34] the portfolio you've curated versus the set of kind of random portfolios of equal size

[00:04:44] that you might have selected instead over the next 10, 20, 30 years of your investment

[00:04:50] paris it. What do you think that means then for everyone out there that is trying to pursue

[00:05:01] this idea of generating alfie? I mean, how can investors have confidence or should they not

[00:05:08] have confidence in these strategies that are trying to seek out some type of edge and how

[00:05:13] how would you address that? Well, there's a sort of distinction between

[00:05:20] alfies and strategies, but let's just call them sort of things that you can allocate to in a portfolio.

[00:05:28] Right. So what can you allocate to you can allocate to

[00:05:32] indices like the S&P 500 or the Bloomberg ag or a commodity index or you could or you

[00:05:40] can allocate to foreign stocks or you can wrap all global equities up in a single allocation.

[00:05:47] So these are things that people typically think of as core holdings and portfolios.

[00:05:52] What else can you allocate to? Well, you can allocate to if you kind of want to

[00:05:56] could separate out the distinctive characteristics of a stock portfolio that emphasizes

[00:06:05] stocks with high momentum or strong value characteristics or high profitability or these

[00:06:11] types of characteristics that have historically been associated with long term growth above what the

[00:06:21] market has provided. Or you can try to allocate to those those strategies in a long short context

[00:06:30] and sort of think about what in the academic literature we might call alpha beta separation,

[00:06:35] which is kind of what we try to get at in our return stacked ETF complex.

[00:06:42] But the idea being that you've got this momentum strategy and you've got this equity strategy,

[00:06:48] most people typically kind of commingle them they buy a long a long only portfolio

[00:06:55] of high momentum stocks. So what do you get there? When you sort of packaging up market beta,

[00:07:01] you've got kind of this long exposure to equities in general and you've got some amount of exposure

[00:07:07] to the momentum strategy or factor because you're trying to emphasize high momentum stocks.

[00:07:13] So alpha beta separation says that well, ultimately, ideally what you want is

[00:07:20] to get pure exposure to that momentum factor via a long short portfolio where you're long

[00:07:27] the highest momentum stocks and short the most negative momentum stocks. And so you're kind of

[00:07:33] market neutral, right? And so that's not contaminated by any long exposure to equity beta,

[00:07:40] so to speak. And then you're going to stack that on top of your long equity exposure, right?

[00:07:47] So the idea is you want to get access to these other strategies, but I don't think about

[00:07:56] this momentum overlay or this momentum edge or strategy as being any different than

[00:08:04] an equity beta strategy, right? So in your portfolio, you can hold equities

[00:08:09] and you can hold momentum. What else can you hold? You can hold value,

[00:08:13] you can hold size, you can hold profitability, you can hold trend, you can hold carry,

[00:08:20] you can hold convertible arbitrage or merger arbitrage. You've got all these different

[00:08:27] kind of alphas or edges that are mostly uncorrelated with one another and with

[00:08:36] the factors or allocations that we all know and love like equity beta and bond beta.

[00:08:43] And ideally, you want to add these all to a portfolio. But then the question becomes, well,

[00:08:50] there's a near infinite variety of ways you can define value strategies or momentum strategies

[00:08:56] or trend strategies or whatever, right? And so investors are left with both, well, what

[00:09:04] general ideas or concepts do I want to allocate to in the portfolio? And then beneath that,

[00:09:11] what methodology to get exposure to those factors or strategies do I want to choose, right?

[00:09:19] And so what I'm trying to get at with my point here is that it is very difficult

[00:09:26] to look at a set of potential strategies that you could allocate to,

[00:09:31] including equities, including bond beta, but also including trend strategies, carry strategies,

[00:09:39] value momentum, etc., and make a choice ex ante of the portfolio that you want to create from those

[00:09:48] that is likely, but that doesn't include all of them, that is likely ex ante to outperform a

[00:09:57] strategy of just choosing all of them, right? And because the edges are so noisy that over

[00:10:07] an investor's lifespan or investment horizon of call it 30, 40 years, there's no way that you'll

[00:10:15] be able to make a strong statistical case, even after 30, 40 years, that your the specific portfolio

[00:10:24] that you chose was better than just trying to get access to all of the different strategies in your

[00:10:34] portfolio that you think that you reasonably believe it, right? So let's just find as many

[00:10:44] reasonable exposures that you can get into your portfolio and add as many of those as is possible,

[00:10:52] given what is actually available to you, given where you are in the investment universe, right?

[00:10:59] Obviously, institutions have access to a much wider variety of strategies than a typical retail

[00:11:06] investor, though that gap has closed dramatically over the last five or 10 years as asset management

[00:11:15] shops have launched various ETFs and mutual funds that attempt to approximate the type of strategies

[00:11:23] that institutions have always had available to them and make them available to retail investors

[00:11:30] in a package that they can buy and that they can mostly understand at a reasonable price.

[00:11:35] So that's a real win for all investors. Well, and I think you were, you mentioned before you

[00:11:41] jumped on live here that you were in the process of like developing or coding something up that

[00:11:47] kind of demonstrated this. So maybe it's not totally ready yet, but how are you tackling this

[00:11:53] like mathematically or statistically? There's some really great databases that some academics

[00:11:59] have made available. One of them is on the QFactor site. She and Zhu, I think, who wrote the paper

[00:12:11] on the QFactor model as an alternative to say Fama French. And they made available about 180

[00:12:22] strategies grouped into different ways to run momentum or value or profitability or trading

[00:12:30] frictions or liquidity, that sort of thing. So you've got kind of 180 odd strategies with

[00:12:39] daily returns that have all been written about and demonstrated to have statistical

[00:12:47] significance by the academic community over say the last 30, 40 years of publications.

[00:12:54] And so an interesting experiment is, well, let's sort of randomly, let's examine the performance

[00:13:04] of these 180 different factors from say when most of them start in the data that's provided

[00:13:11] in the late 1960s until say, you know, the late 1980s or the late 1990s. So over sort of a 20,

[00:13:19] 30 year horizon, right? And then based on the performance of all of these different strategies,

[00:13:27] now you're charged with choosing a portfolio of 10 strategies out of those 180 strategies after

[00:13:37] looking at 30 years of performance, right? So, you know, typically people might look through

[00:13:44] that data and say, well, you know, I'm going to pick the ones with the highest sharp ratio,

[00:13:53] the top 10 with the highest sharp ratio or the top five with that delivered the highest

[00:13:58] cake or something like this, right? There's a variety of different metrics you might use.

[00:14:02] Maybe you'll evaluate it on some Pareto frontier of a combination of different

[00:14:06] metrics, right? Whatever. And then you hold that portfolio over the next 10, 20, 30 years

[00:14:15] and examine how that portfolio that's sold that you selected performed against all of the other

[00:14:23] portfolios that you may have selected, right from that group of 180 potential strategies.

[00:14:30] And, you know, can you identify any set of metrics that you can use as evaluation criteria

[00:14:40] that will allow you to choose a basket of all 180 portfolios that will go on and have a high

[00:14:49] probability of outperforming any random basket of strategies that you might have chosen instead

[00:14:57] of the one that you chose, right? And, you know, evaluating where the ones that you selected lie

[00:15:07] on that distribution will give you some intuition even after, keep in mind, you've got a 30 year

[00:15:14] evaluation horizon here and you're going to go ahead and hold it for another 20, 30 years,

[00:15:19] right? So, a lot of people I think might intuit that, well, I've got 30 years of history.

[00:15:26] I think I can be pretty confident that the ones that have the highest sharp ratios are going to,

[00:15:32] you know, go on and be in the top half of the all potential strategies that I might have picked.

[00:15:41] And what you see in fact is that that's not true, that it ends up being, right,

[00:15:47] pretty well in the median, right? So, if I'm going to select 10,

[00:15:52] and then I generate a thousand alternative portfolios of 10 different strategies,

[00:15:59] well on average if I run this simulation many, many times,

[00:16:03] you're at about the median. But why am I choosing 10? You know, why wouldn't I choose 20 or 30? And

[00:16:11] it turns out that if you choose 20, randomly choose 20 instead of curating 10 based on their

[00:16:18] historical performance, then your curated 10 on expectation will perform worse than

[00:16:27] or your random 20. And the random 20 you'll perform, you know, the curated 20 will perform

[00:16:33] worse than your random 30, right? The message being there and then again this is like a bit

[00:16:39] of a constrained simulation but it does demonstrate that trying to make decisions

[00:16:43] ex ante even over long time horizons, you end up not outperforming a simple strategy of just trying

[00:16:53] to get to allocate to as many different strategies as possible and take advantage

[00:16:59] of the diversification opportunities that are available because your skill in selecting

[00:17:04] strategies in advance based on even very long histories of performance is pretty close to

[00:17:13] zero. It would seem to me, correct me if I'm wrong about this, but it would seem to me like

[00:17:17] this is why the pod shops operate the way they do. They've got a bunch of people who think they

[00:17:20] have edge, they put them all together, they manage the exposures around that and then

[00:17:25] they have a much better chance of producing alpha over the long term doing it that way.

[00:17:29] I think the pod shops, so I think there's something to that. I also think there's

[00:17:34] something to the fact that the pod shops are really looking for people that genuinely have

[00:17:40] alpha. So I think it's useful to kind of distinguish between what we might call sort of

[00:17:47] systematic factor strategies and alpha, right? So I would distinguish it as alpha comes from

[00:17:55] somebody who has various particular niche insight or information or experience within a fairly

[00:18:05] narrow domain of the market. So for example, we have a client who allocates to a municipal bond

[00:18:14] manager. Now this manager, he's got a hard cap at about a billion dollars. The team that runs it,

[00:18:21] spun out of what used to be the largest muni market making desk, worked there for 20, 30 years.

[00:18:29] What did that give them? Well, it gave them access to knowledge of where all of the flows

[00:18:38] from muni bonds, all of the issuance from the muni bond sector coming from the different state

[00:18:45] governments. Who are the decision makers there? How can they get inside information on

[00:18:52] what type of an issuance is coming down the pipe? And then being at the center of flows

[00:18:59] of the muni market, which is a very niche segment of the market. I think that's just one example,

[00:19:07] but there are many examples for example, somebody who works in for 20 years in the electricity

[00:19:13] markets. Electricity is a very nuanced pricing market with a very small number of key players

[00:19:24] and is largely driven by changes in regulations at the state level and the county level.

[00:19:30] Having very specialized knowledge of that from having worked and having experience inside the

[00:19:35] sector gives you a real edge. These are the types of strategies and people that the pod

[00:19:45] shops are looking for. Now, these typically tend to be fairly in liquid strategies. You can't

[00:19:53] have Elliott managed with a $71 billion firm that's running just running an electricity

[00:20:01] niche electricity strategy or a muni strategy. But the goal is to find hundreds of people who are

[00:20:10] all running these niche little strategies that will all require liquidity to take advantage

[00:20:19] of opportunities at completely different times from one another and putting them all together

[00:20:25] in a diversified basket. Now, I'm sure there are also very scalable strategies and there as well

[00:20:34] that maybe are running more liquid equity strategies or options, strategies or whatever.

[00:20:40] I fundamentally believe and my insights from knowing people at those shops is that the majority

[00:20:47] of the alpha that you can't get anywhere else at scale comes from the assembly of many different

[00:20:58] less liquid small niche players that are all operating together in an ensemble within these

[00:21:04] pod shops. I want to ask you a couple other things about this idea that there is a lot

[00:21:08] of noise even in very long term returns. And one of the things people like us have been

[00:21:13] struggling with is this idea is does the value factor still work? And I think this is kind of

[00:21:17] a related concept in the other direction because we're trying to say it doesn't work.

[00:21:22] And I think Corey showed in Factor Fimble Winter, the amount of time we would need

[00:21:25] to show that is longer than our investing lifetime. So this does become about faith.

[00:21:30] It becomes about the person running the strategy and their ability to make a decision because

[00:21:34] data's not going to tell them that. Yeah. I agree. Value is a great example because

[00:21:40] if you look at a cross-section of potential ways you could define value, whether it's kind of

[00:21:45] based on book value or cash flow yield or what have you, more than half of those

[00:21:54] academic definitions of value have underperformed, have negative alpha over the last 20, 30 years.

[00:22:05] And so what ends up happening, of course, is that both practitioners and academics

[00:22:11] end up kind of modifying their definitions of value over time in order to sort of preserve the

[00:22:21] veracity or faith in the strategy. And I have no specific opinion on value.

[00:22:29] It has just as compelling evidence as any of the other strategies that you might consider typically

[00:22:37] to hold in the portfolio alongside other strategies. But there's a really good case for

[00:22:45] where if you looked at the... And we've got good data on value portfolios.

[00:22:52] I think that goes back to the 1800s now, but certainly back to the 1930s. We've had

[00:22:57] that for a couple of decades. Even 80 years where even your medium value strategy in terms of all

[00:23:05] the different potential definitions had a sharp ratio in the neighborhood of 0.4,

[00:23:11] like that really high information ratio. And then sort of post-1990 it basically flatlined.

[00:23:18] Right? So you've got 50 years, 60 years, 70 years of... Like the T-stat is just ridiculous.

[00:23:27] And yet over what I think most people would agree is a time horizon that's about 10 times longer

[00:23:36] than most investors have patience for an underperforming strategy. Typically investors

[00:23:43] like 90th percent dialed patience is like three years. Right? So you've got like 30 years of

[00:23:50] underperformance, just no investor is going to stick with that. And the only reason why anyone

[00:23:57] is still investing in the value strategy except for a handful of academics and people

[00:24:03] who have actually done their own research on it I think is because for the most part it's

[00:24:07] been commingled and kind of diluted against a broad equity beta. Right? So because most people

[00:24:15] were getting their value exposure by holding kind of like a value tilt on a broad, larger, mid-cap

[00:24:23] equity portfolio and the equity beta itself has done so well. Right? So people don't kind of notice

[00:24:30] so much that the value tilt hasn't added any value over the last 30 years. Again, I'm not

[00:24:36] saying it won't go on to add tremendous value over the next 30 years. In fact, you know,

[00:24:42] gun to my head, I think value is probably poised to maybe have a renaissance because so many people

[00:24:49] are now disillusioned by it. Right? You've just got most of the flows over the last 10 or 15 years

[00:24:56] have been out of concentrated value funds and into either cap weighted beta or some other

[00:25:03] factor strategy that has been doing something for people more lately. Right? And as a function of

[00:25:10] that what we've seen and I think our friend Ned posted a really good chart from GMO recently that

[00:25:17] shows that the really deep value stocks in the US are still showing extremely good value relative

[00:25:26] to their history while pretty well every other value quintile is expensive. Right? So those who are

[00:25:34] really big believers in value, this is probably a really good time to double down on it. The problem

[00:25:42] is a lot of people who people listen to for good reason have been saying people should double

[00:25:48] down on value like every couple of years for the last five or 10 years and it except for a surge in

[00:25:56] kind of 2022, it's just been a real disappointment. Right? So it just takes a lot of faith and that's

[00:26:03] why I say you really can't make any decisions over your lifetime related to or based on the

[00:26:12] performance that you observe. Rather, you should take advantage of that free lunch of diversification

[00:26:21] and create as much diversification in your portfolio as possible. And just allocating to these kind

[00:26:29] of value tilted or momentum tilted coal mingled equity portfolios, you get very little of

[00:26:37] diversification that you might seek into these other types of strategies, which is why this alpha

[00:26:43] beta separation I think is so important and why we think this return stacking concept which

[00:26:51] institutions have been taking advantage of for so many years is an idea whose time has come for

[00:26:57] retail. Yeah, on that idea of diversification, I wanted to ask you about, I had Jason Buck

[00:27:01] on the podcast and I interviewed him recently and we were talking about this idea, he was

[00:27:04] challenging the idea of stocks for the long run. And we talked about the long term data and I think

[00:27:08] over 100 years or whatever it is, the US has something like 6.5% real returns, global stocks

[00:27:14] has something like five, I mean, I might be a little bit off on that but I'm probably in the

[00:27:17] general area. So a lot of people will look at that and say that's 100 years of data.

[00:27:21] Like I have to expect that that's telling me a lot about what's going to happen in the

[00:27:24] future. I mean, a period that long, but you don't think that's necessarily true, right?

[00:27:28] Oh, I definitely don't. I mean, I think the best estimate of any single stock market.

[00:27:37] So let's take the US for example, right? How would I estimate the expected return of

[00:27:42] US equities? I would probably go back and run a large variety of simulations

[00:27:53] of all portfolios that cumulatively sum up to approximately 50% of global market cap

[00:28:04] at any point in time and then observe the median performance of those portfolios.

[00:28:11] And then I would shrink that toward the global average. So I would probably be somewhere in

[00:28:19] the neighborhood of somewhere between 3.5% and 5% annualized excess return for US equities,

[00:28:30] unconditional. So if they were trading near their median historical valuation, but they're not

[00:28:37] trading anywhere near their median historical valuation, they're trading, depending on how

[00:28:44] you measure it somewhere between 50% and 100% above their median historical valuation.

[00:28:50] And so even if you make the case that the average valuation over time should rise

[00:28:59] to reflect the fact that markets are more liquid, you've got lower trading frictions,

[00:29:05] you've got a longer history, so you can gain more confidence in the expected premium, etc.

[00:29:10] So I could buy the argument that maybe you should have a slow creep of equilibrium PE

[00:29:22] expansion over time, but even if you run a regression on that increase in PE ratios going

[00:29:32] back to the early 1800s, you're still at about a 50% overvaluation relative to even that adjusted

[00:29:42] PE ratio. So adjusting for the fact that we expect PE ratios to be higher today than they were in

[00:29:48] the 1800s, for example, right? So condition on the fact that markets are expensive,

[00:29:55] I personally would expect US equities to probably underperform their long-term average

[00:30:02] over the next kind of 15 to 20 years. I say 15 to 20 years because if you look back through history

[00:30:08] that tends to be the horizon over which we do see this kind of cyclical reversion,

[00:30:14] there's a huge variance around that horizon and we've only had 10 or 15 year

[00:30:26] non-overlapping periods through history to gain any statistical confidence on this.

[00:30:34] There's a lot of noise here, but a decent guess would be kind of a reversing to

[00:30:42] somewhere near the expansive adjusted PE ratio over the next 10 to 20 years.

[00:30:52] At the same time, global equities maybe are trading a little bit below their average valuation.

[00:30:58] Some of them, some markers are trading well below and those are likely to outperform US

[00:31:04] equities over the same horizon. That leads really well in the return stacking,

[00:31:07] which you already introduced, but I wanted to dig into more. But first I wanted to

[00:31:10] define some terms. There's a term that's been around for a long time, portable alpha.

[00:31:14] Is that the same thing as return stacking? I thought it was pretty similar,

[00:31:16] but then I asked Claude who could potentially be Hennu hallucinating and told me all these

[00:31:20] differences. Are those basically the same thing? We certainly think of it as being very similar.

[00:31:27] The whole idea of portable alpha for institutions, there's two ways to run it.

[00:31:35] One is you can take your core portfolio of equities and bonds because many institutions

[00:31:45] don't suffer from tax issues. They're tax-examined. They can allocate to derivatives

[00:31:57] for their core exposures. For example, allocate to S&P 500 futures and various treasury bond futures

[00:32:06] to get their core equity-rich premium and duration premium. That doesn't consume very much capital.

[00:32:16] Let's say that an institution has got $100 in capital that they need to deploy,

[00:32:22] getting an exposure to a 60-40 portfolio of equities and bonds using futures consumes less than 5%

[00:32:32] of that capital just because of the way that you get leverage in the futures markets. You've only

[00:32:37] got to post a small amount of the total notional value of the futures that you're buying

[00:32:45] in order to get that exposure. Now you've got $95 that you can allocate to things like

[00:32:52] private equity and real estate and hedge funds as an example. That's one way to do it.

[00:33:02] That's a fairly common way to do it. Another way to do it is to have a core allocation to your

[00:33:10] equities and bonds by buying cash equities and owning a credit portfolio, cash equity portfolio

[00:33:19] directly, and then using that portfolio to collateralize. In other words, you're borrowing against

[00:33:26] that portfolio, either borrowing against it directly almost like a margin loan so that

[00:33:31] like a Goldman Sachs will say, well, we'll accept this, we'll accept $70 of that $100 of equity and bond

[00:33:45] collateral as collateral against, will lend you $70. Then you can take that $70 and you

[00:33:54] can go to play at the private equity or hedge funds or what have you. Or you can run your portable

[00:34:01] alpha overlay as you're holding cash equities, cash credits, and then you're using those to

[00:34:09] collateralize a futures portfolio in which case you don't need to go to the bank and say,

[00:34:14] hey, bank, could you loan me money against my equity and bond portfolio? Instead, you just

[00:34:19] carve off a small amount, maybe you're only $95 of your $100 is invested in cash equities and cash

[00:34:27] bonds. You take that 5% and use that to collateralize the futures trading overlay. The latter variety

[00:34:37] portable alpha is closer to what we are doing in return stacking.

[00:34:44] You talked a little bit about some of these things before, but how do you think about

[00:34:47] what to stack on top? What are the characteristics of the types of strategies you'd want to pair with

[00:34:52] stocks and bonds? In a retail context, in general, what you want to stack on top are strategies that

[00:35:03] you expect to have low correlation with the underlying beta component of the portfolio that

[00:35:12] you are allocating to. For example, you've got a return stack portfolio where the

[00:35:20] beta of that portfolio is trying to give you up 100% exposure to US equities. It's called the S&P 500.

[00:35:29] You want to stack on top a strategy that you expect to have a very low correlation

[00:35:36] with US equities over time. When you're adding that on top, you're implicitly using leverage.

[00:35:43] You've got nearly $100 that are already allocated to holding, let's say, an S&P 500 ETF.

[00:35:52] That's fully invested. Anything that you're running on top is necessarily providing,

[00:35:58] is done using leverage. If you were to just stack another S&P 500 exposure right on top of that,

[00:36:07] well, now all you've done is double your risk. You may be double your expected return,

[00:36:12] that's not quite right actually because of the volatility drag, but you may be

[00:36:16] approximately doubling your expected return, but you're basically doubling your risk and

[00:36:19] not getting any diversification benefits. That's not very helpful. What you want to do is

[00:36:25] stack a strategy on top that in expectation has a very low zero or negative correlation to the

[00:36:33] US equities component. If you're stacking bonds, you want to stack on top something that

[00:36:37] you expect to have lower negative correlation to the bond component.

[00:36:43] Now, that could be anything. An institution could go to a bank and say, I want you to run

[00:36:51] a portfolio of long, short US equities. A value, market neutral, value factor portfolio,

[00:37:03] or market neutral, momentum factor portfolio, or what have you. To run that directly in an ETF,

[00:37:10] you run up against borrowing constraints. The ETF structure will only allow you to borrow

[00:37:17] directly up to 30% of the value of the cash that's invested. There's $100 invested in the ETF.

[00:37:28] You can borrow $30 and you can use that $30 to run a strategy, but that doesn't give you much

[00:37:33] latitude if you're going to run a market neutral equity portfolio. Typically, the

[00:37:39] ball on that portfolio is extremely low. The way that you generate, you make it attractive is

[00:37:45] by levering it up very substantially. You just can't run that kind of strategy in an ETF

[00:37:52] as a stacking component very easily because of this direct leverage constraint.

[00:37:58] In contrast, with futures, we could take 10% of the total value of the portfolio.

[00:38:09] Let's say we've got a stacking product that wants exposure to the S&P 500

[00:38:17] with something stacked on top. You hold cash equities to the tune of, say, 80 or 90%.

[00:38:24] And then you hold 10% or 20% in T-bills. You make up for that missing equity exposure by buying

[00:38:34] and rolling equity futures that have that 10% or 20%

[00:38:41] notional exposure to US equities, the top ups. And they are getting your 100% US equity exposure

[00:38:47] that you want. But then you're also using that extra T-bill collateral to collateralize

[00:38:54] a futures trading strategy. And for the purpose of the regulatory constraints,

[00:39:02] the leverage on trading the futures is not considered the same as directly borrowing

[00:39:10] against the portfolio. And so you can effectively run a future strategy at a meaningful level of

[00:39:17] expected return and risk on top of a full equity portfolio or a full bond portfolio

[00:39:26] under the regulatory constraints of a typical 40-act mutual fund or ETF structure. So

[00:39:34] that's why we tended to gravitate toward futures-based strategies or strategies that just

[00:39:39] don't require a lot of leverage. And that point on uncorrelated assets is a really

[00:39:43] important one because the idea of Portable Alpha got a bad rap in 2008. But I think what a lot of

[00:39:48] people were doing back then is they were stacking hedge fund strategies that had correlation to the

[00:39:53] market and also were very illiquid. They were stacking that on top of stocks and bonds,

[00:39:56] and that could be very problematic. Yeah, that's a really good point. So remember I said there's

[00:40:01] kind of two ways for institutions to run a Portable Alpha. One is to get your core

[00:40:06] exposure through futures and then to use the excess collateral to buy hedge funds

[00:40:11] and private equity and real estate and all that kind of stuff. So in that model, let's go to 2008.

[00:40:18] So you've got your, let's call it 60% inequities via futures, 40% in bonds via futures,

[00:40:26] and now you're stacking kind of $80 to $90 of private equity, real estate, and hedge fund

[00:40:37] exposure, maybe credit exposure on top. Coming to 2008 credit craters, equities crater,

[00:40:49] you're kind of hoping that your private equity, equity, your real estate equity,

[00:40:57] and your hedge funds, which kind of have a 0.8, 0.7 beta equity are going to hedge your

[00:41:07] equity and bond exposure that you've got through futures. And of course, we learned that in a

[00:41:13] credit event or a major liquidity event that most of these strategies are not really designed,

[00:41:21] they all kind of hurt you when it hurts most to be hurt, right? When your

[00:41:28] portfolio is in the gutter, sadly a lot of these other exposures are in the gutter at the same time

[00:41:36] for the same reasons, right? Which is why it's so important that if you're going to run a portable

[00:41:42] alpha strategy, the leverage that you're taking to generate the excess return on top is not stacking

[00:41:51] risk in the same way that it's stacking returns. You want to stack, call it a unit of returns,

[00:42:00] but only say half a unit of risk or a third of a unit of risk, right? And the way that you do that

[00:42:09] is by stacking strategies on top that are structurally designed to have very low correlation

[00:42:17] to your core stocks and bonds allocation during crisis periods like 2008, where it really hurts

[00:42:27] to be hurt if your diversified overlays are going in the same direction as your core portfolio.

[00:42:36] So there's two more for me before I hand it back to Justin. I want to ask you about

[00:42:38] managed futures, which is I think the first thing you guys did when you stack something

[00:42:41] on top. One of the things I've, whatever I know, I know whatever I have you on,

[00:42:44] I can always ask these tough questions that I can't figure out the answer to. And this is one of

[00:42:48] them because I've thought about this a lot. Like if I'm running a managed futures strategy, and

[00:42:52] I believe you guys are replicating an index, you can correct me if I'm wrong about that. But

[00:42:55] in general, if I'm running a managed futures strategy, and I'm coupling it with a stock and

[00:42:58] bond portfolio, I go back and forth on do I want stocks and bonds in the managed future

[00:43:03] strategy or not? It seems like there's arguments in both directions, and it seems like

[00:43:06] smarter people than me have argued both ways. So I'm just wondering how do you think about

[00:43:10] that? Well, in an ETF structure, you want to be able to take advantage of the basket redemption

[00:43:20] creation policy that the market makers provide, which shelter you from accruing taxes within

[00:43:31] the ETF structure, if you're trading cash equities in particular. So if you've got cash equities

[00:43:42] as your core data that you're going to then stack managed futures on, typically you want

[00:43:48] to have as much allocated to cash equities as possible so you can take advantage of these

[00:43:54] create redeems to mostly eliminate the tax distributions that you would otherwise get

[00:44:03] every year. With bonds, it's a little bit less clear and the structure of the ETF

[00:44:10] and the types of bonds that you're trading make a bit of a difference. Futures are taxed as 6040.

[00:44:20] So actually, if you've got bond futures and you're accruing gains from bond futures,

[00:44:28] most of the gains are more tax efficient than if you were to hold a cash bond because of course the

[00:44:38] interest payments from the cash bond are taxed as regular income. Whereas the gains from rolling

[00:44:44] bond futures are taxed as 6040. So 60% long-term capital gains, 40% income. So you've got a nice little

[00:44:53] tax pickup there. It just depends on the type of investor and the type of structure. Like if you're

[00:45:03] in a mutual fund structure, then your equities are going to accrue capital gains or going to

[00:45:09] get distributions because there's no create redeem mechanism like there is for ETFs. So you

[00:45:16] can't play fun games with taxes that way. So there's a little bit less flexibility.

[00:45:23] But yeah, I mean obviously if you're a taxed and exempt investor then if you can achieve alpha,

[00:45:29] beta separation, whatever the most efficient way to achieve that is best. If you're a retail

[00:45:36] investor in a taxable account then it matters. And we try to structure the ETFs so that they're

[00:45:44] maximally tax efficient for retail investors. And the second thing you guys added after

[00:45:50] managed futures is something to stack on top of, something I think less people will be familiar with.

[00:45:54] It's this idea of futures yield. Can you just explain what that is?

[00:45:58] Yeah. So I think it's also important to clarify that a lot of people equate the term

[00:46:04] managed futures with a strategy you get, let's say most managed futures run within their fund,

[00:46:13] excuse me, which is trend following. So trend following in futures that you've got somewhere

[00:46:20] between call it 25 and in some cases a couple hundred different global futures markets across

[00:46:28] equity indices, bond indices, many, many different commodities and global currency markets

[00:46:35] and some exotics sometimes. And you're effectively kind of, you're leaning into the phenomenon

[00:46:43] whereby when futures markets have gone up over the last kind of three to 12 months,

[00:46:51] they're more likely than not to continue going up over the next few days or weeks.

[00:46:56] Right. And so if you buy futures markets that have been rising and you sell short futures

[00:47:04] markets that have been going down, this is called trend following. And historically,

[00:47:09] this has generated a very nice premium over time with near zero long-term average correlation with

[00:47:16] equities and bonds and commodities and currencies, but most people don't hold

[00:47:22] those in their portfolio, so they're less relevant. The managed futures, so managed futures doesn't

[00:47:28] always just mean trend following is kind of what I was going. So our carry strategy is also a managed

[00:47:33] future strategy where it is using a different signal to decide which futures markets are

[00:47:41] more attractive or likely to go off over the next few days and weeks.

[00:47:45] We're not using trend, we're using something called futures yield.

[00:47:49] And so I always like to say that carry or futures yield is just the futures market's way

[00:47:59] of ensuring that there is no free money lying on the ground

[00:48:07] that banks or sophisticated traders can pick up without having to take any risk.

[00:48:14] And the reason I say this is that, well, carry in cash equities is the dividends

[00:48:21] that are paid out. You might argue it's the shareholder yield maybe, right? If companies

[00:48:26] are buying back stock, that's a distribution that shareholders accrue, right? But either way,

[00:48:33] that's, you know, let's call it dividend yield just to keep it simple. So you've got

[00:48:37] dividend yield that's paid out and you get that dividend regardless of changes in the

[00:48:43] equity prices themselves, right? So it's the same with interest payments on bond, right? Bond prices

[00:48:49] go up and down but they pay a fixed interest payment. So carry is just the return you expect

[00:48:58] to get from holding an asset if the price doesn't change, right? Regardless of what happens

[00:49:04] at price, you're going to get paid the dividends, going to get paid the end.

[00:49:07] So a future that is benchmarked and transferable into the S&P 500, for example, at maturity,

[00:49:19] needs to price in the fact that if you hold the futures instead of holding the cash equities,

[00:49:26] that you're not getting a dividend yield in the future where you are getting

[00:49:30] in cash equities. So if the futures didn't price in, the fact that you could get this if you hold

[00:49:37] held the cash equities instead, somebody could own the cash equities, short the futures,

[00:49:43] earn the dividend yield and be assuming no price risk on the S&P 500, right?

[00:49:49] So to assure that doesn't happen, markets are generally efficient in this way,

[00:49:54] the futures need to trade at when you buy the futures, they need to trade at a slightly lower price

[00:50:03] than the cash index that they track so that over time as we approach the period when

[00:50:10] the futures are directly transferred a little into the cash index at maturity,

[00:50:16] that both the futures and the cash index have provided the same return, including the dividend yield,

[00:50:26] right? So when futures are priced slightly below the level of the cash index, they are expected to

[00:50:32] then creep up over time all things equal so that those returns on the cash index and the futures

[00:50:40] are the same over the same time period, right? It's the same for bond futures. Bond futures are

[00:50:46] deliverable against a bond at maturity and those bond futures need to provide approximately the

[00:50:53] same return as cash bonds over that time, including the distributions that were paid by those cash

[00:51:00] bonds, right? Well, commodities and currencies also have these carry characteristics, right? So

[00:51:08] a future tends to be priced either above or below the cash spot market. Why? Well, maybe for

[00:51:17] example, let's see, you've got copper futures and you've got someone who is looking to build a

[00:51:26] copper mine and they need to get financing, they need to raise a few billion dollars to go

[00:51:32] and build this copper mine. The copper mine may not be on stream for 10 or 15 years, right?

[00:51:37] If they did not sell the expected production from that copper mine forward,

[00:51:43] so they walked in the price that made sure that the copper that they're going to mine from that mine,

[00:51:51] 10 or 15 years from now, made that mine profitable, made the investment in that mine profitable,

[00:51:58] it would be more difficult for them to raise the capital to build the mine and they either

[00:52:03] wouldn't be able to raise it or the cost of capital that they would pay in order to get the funds to

[00:52:11] build the mine would be unreasonably high. Whereas when they lock it in, they're able to go to

[00:52:17] lenders and say, all right, the expected ROI on this mine is X. I know this because I've sold

[00:52:24] forward a good chunk of my production and the people that are providing capital are then comfortable

[00:52:33] enough to give them a much lower rate on that capital, right? So the producer who's building the

[00:52:41] mines wins. Their shareholders win because they're able to go about their business operations with

[00:52:48] a lower cost of capital because there's a much lower variance in their expected cash flows

[00:52:56] because they walked in the prices for the things that they're going to sell in the future, right?

[00:53:01] The question then is, well, who is insuring those prices for producers? And how much do

[00:53:09] they get paid for that? Right? Well, they get paid the speculators like people that run

[00:53:17] carry strategies on commodities. They provide this price security for the producers. They

[00:53:26] basically sell this price insurance and they burn an insurance premium in the form of this

[00:53:34] futures yield, all right? And there's a few other things that go into this for different

[00:53:40] commodities. It's, you know, there are different dynamics of play crops are affected by

[00:53:45] weather and oil is it is impacted by storage availability and demand and all kinds of stuff,

[00:53:53] right? But you know, there are the point is that the futures in the future trade at a different

[00:54:01] price than the spot price. And you're able to then take long positions in futures that are

[00:54:08] trading above the spot price because all things evil were expecting the price of those

[00:54:14] futures to fall towards the spot price over the time to maturity. And you want to buy futures

[00:54:23] whose price is currently below the spot price because you expect them to rise toward the spot

[00:54:28] price over that maturity, right? Now, importantly, that spot price is changing over time, right?

[00:54:36] So this is far from a guarantee that if you're buying a future that's below the current

[00:54:42] spot price today, that you're going to earn a positive return on that because you're still

[00:54:47] taking price risk, right? But on average, that yield bears out, right? So now like on average,

[00:54:56] that you do earn a positive dividend yield, you don't do earn a positive interest rate, right?

[00:55:02] And so we're able to take advantage of this across equities, bond indices,

[00:55:07] of righted commodities and currencies in order to generate this return on top of stack on top of

[00:55:15] the core beta component, like stock equity beta or bond beta to generate this kind of return stack

[00:55:27] profile. So just one last kind of point on this is you guys run five different ETFs at

[00:55:34] the current time. And one of the things that's, I think that that's impressive given we had an ETF

[00:55:38] and how hard it is to kind of grow assets. I mean, at this point, I think you guys are,

[00:55:43] you know, north of 750 million in terms of total assets. And the other cool thing is

[00:55:48] it's a lot of times you see these ETFs families and it might be like all the assets are maybe

[00:55:53] one or two strategies. But from my viewpoint, it looks like, you know, it's a nice mix.

[00:55:59] Investors are embracing different strategies within these ETF wrappers and the assets like

[00:56:05] aren't all on one thing. It's kind of in, you know, across the board in the mix. So that's,

[00:56:11] I think a testament to what you guys have built and how these kind of stand out in what is a crowded

[00:56:18] space with ETFs. Thank you. I mean, the primary recognition here, well, I guess there were two,

[00:56:23] but the primary recognition here was that it's actually very hard for retail investors to get

[00:56:31] diversifications free lunch. Because traditionally, in order to add diversifying investments to a

[00:56:39] portfolio, you need to sell down your exposure to your core stocks and bonds. So, you know,

[00:56:49] you're adding maybe sleeves of the portfolio that are uncorrelated to your stocks and bonds,

[00:56:56] the overall risk of your portfolio is maybe going down. But your return is also going down, right?

[00:57:05] So in other words, your portfolio may have a higher sharp ratio or a higher,

[00:57:09] it may be more efficient. But, you know, there's this old saying you can't eat sharp ratio,

[00:57:16] right? And so that is true or has been true historically for most retail investors.

[00:57:22] You can increase your sharp ratio, but only at the expense of lowering your risk and lowering

[00:57:29] your return. So what we said is, well, we want to provide investors an opportunity to be able to

[00:57:36] eat their higher sharp ratio by allowing them to keep full exposure to the core

[00:57:42] stock and bond exposures that they love so much. And then just stack these

[00:57:51] alternative uncorrelated strategies on top, so that you can have your diversification

[00:57:56] and you can eat it too, right? So it's, you know, like the icing on the cake,

[00:58:03] you know, you're not having to like cut off, cut out a bunch of pieces of cake,

[00:58:07] the slide, something else in. Now the whole cake, and then you've got this lovely icing layer on top

[00:58:13] that legitimately does give you much higher diversification. And for every extra

[00:58:23] percent of return that you expect to get, you're getting far less than a percent of extra risk

[00:58:32] in your portfolio, right? So you're able to, you know, expand your efficient frontier and eat

[00:58:38] that diversification's free lunch. Well, I have a good idea. Let's get, let's, the new tagline can

[00:58:44] be, you can eat your cake and have it too. And then we can put Corey on a roller and see how many

[00:58:51] calories he can burn in 60 minutes. That's exactly, yeah, I like that. I would, but he

[00:58:57] can burn a lot of calories in 60 minutes. So probably a little narrow fuck.

[00:59:00] Rowing crazy guy over there, but okay, okay. So good stuff, Adam. We want to just sort of at the

[00:59:06] end here, maybe just get your thoughts on some of the things that have come up recently on excess

[00:59:10] returns with various guests. And the first thing we wanted to get your, and I mean, maybe we've,

[00:59:14] I don't think we talked about this with you, which is, you know, Mike Green,

[00:59:17] you're familiar with Mike Green obviously at Simplify. And I think maybe he's even been

[00:59:20] on your podcast. And, you know, his work around passive and how passive flows are influencing

[00:59:28] the overall market. And then kind of coupling that with, I think it was David Einhorn who was

[00:59:34] on Patrick with Shaughnessy's and Best Like The Best, where, you know, he was making the point

[00:59:38] that, you know, fundamentals because of what some of Mike's greens work, fundamentals are

[00:59:43] mattering less and less in the market because of these flows. And he's kind of changed

[00:59:47] his investment strategy to look for stocks that are returning capital

[00:59:51] to investor through dividends or buybacks and things like that, rather than trying to

[00:59:55] invest based on the market recognizing the fundamental value of companies.

[01:00:00] Right. Yeah. I mean, so it has been a fundamental tenet of the academic canon

[01:00:08] on markets that markets will trade in equilibrium as a function of the expected discounted cash flows

[01:00:20] of the investments. Right. And, you know, Ben Graham I think put it really well with,

[01:00:27] in the short term, the market is, oh, what is it now? In the long term it's a weighing machine

[01:00:33] and the short. Voting machine in the short run.

[01:00:34] It's a voting machine in the short run in the long term. And I think, you know, that the thesis

[01:00:40] that Mike Green and Dave Einhorn and others are espousing here is that the constant

[01:00:49] woes during positive employment cycles like we've expected, we sort of experience, call it 17 out

[01:01:00] of every 20 years are now effectively short circuiting the weighing machine. Right. And so

[01:01:09] you only get to work with the voting machine. And I've had occasion to chat extensively

[01:01:19] with Mike about this in private and in various venues and with others, you know. And

[01:01:26] I am unusually convinced that this is a very substantial effect in markets. I think it is

[01:01:38] amplified by regulation and policy, for example, rules that set default investments for 401k plans.

[01:01:53] New rules introduced by the SEC that make it more difficult for investors to

[01:02:00] invest in diversifying strategies because you've got a, there's more rigorous hurdles,

[01:02:08] regulatory hurdles you need to overcome and able in terms of justifying why investors are paying

[01:02:14] high fees. And also policies that over the last 15 years and arguably over the last, you know,

[01:02:21] 30 odd years has favored equity investors. So continue to kind of bail out an over leveraged

[01:02:32] financial system, maybe improved in equity managers and equity investors to constantly

[01:02:43] support an equity market because it now has become sort of de facto everybody's retirement plan.

[01:02:50] And because other things are linked to it, like, you know, there's now a very highly reflexive,

[01:02:58] if not inverted relationship between markets and the economy. It used to be, of course, that

[01:03:04] markets responded to growth or a lack of growth in the economy, right? Which provided for

[01:03:16] recessions that were markets corrected due to an expectation of lower earnings. I think it's

[01:03:23] reasonable to argue now that the market lags, the economy is tailed, right? So, you know,

[01:03:32] so much of lending now is a function, it is a gain assets instead of being against income

[01:03:39] that when asset markets crater, lending markets or even, you know, go down a little bit,

[01:03:46] lending markets seize up, which is a de facto tightening of the financial system makes it

[01:03:52] harder to get loans and therefore tightens the economy as well.

[01:03:57] So, you know, for all of these different reasons, policymakers have needed, felt like they needed

[01:04:04] to step in and support equity and credit markets to a greater and greater extent over time.

[01:04:11] And so therefore, there's all this sort of, you know, mindless flow on the long side

[01:04:19] into U.S. equities. And of course, they've also had a performance run that is way outside

[01:04:26] expectation over the last 10 or 12 years as well, right? So, all these factors

[01:04:31] mean that there's a huge proportion of the market that is just default allocating to equities over time.

[01:04:38] And if you read some of the literature on the price elasticity of demand for equities,

[01:04:48] it's clear that, you know, if you've got a stock with a market cap that is 10% of

[01:04:56] the market, like we've had, well, not quite 10%, or let's say 5% of the market,

[01:05:02] then 5% of all automatic flows into the market are going to go into that stock, right? Do a cap

[01:05:10] cap weighted stock. So, if you know, the expectation is that a stock that has 5%

[01:05:19] of the market cap of the market will also have sufficient liquidity or liquidity

[01:05:26] equal to in proportion, its market cap proportion. And what they found is that in fact

[01:05:33] that liquidity is not equal to what is expected by a market cap weighted index where people are

[01:05:43] just sort of automatically allocating to it on the long side. And as a result, every dollar that

[01:05:50] goes in is just proportionally raising the prices of these stocks that represent a large

[01:05:57] portion of the index and that this is a compounding effect over time that in fact,

[01:06:03] there is no expectation that the excess pricing that accrues from this phenomenon

[01:06:10] will equilibrate, right? And so this has distortionary effects on markets which may reverse during

[01:06:21] recessions, right? Because a lot of these flows are coming from labor income. If labor income

[01:06:26] at the margin declines, then we may see actually these flows reverse at the margin. And

[01:06:35] I think Mike Green has posited that this might result in kind of a go to zero moment for

[01:06:43] equity markets, right? But it doesn't need to be so awful as that as to imply that we

[01:06:51] probably should expect more enlarger shocks as a result of these kinds of flows and the

[01:06:59] fact that the elasticity of demand for these biggest stocks in the index are not what is expected

[01:07:05] by equilibrium theory. One other thing I wanted to ask you about because I know you're an expert

[01:07:10] in this area and it's really interesting to me like we just had a swat the motor in on the

[01:07:13] podcast and he's been talking a lot about AI and he's a big believer in AI and what it's

[01:07:19] going to be, I think as all of us are, but he had also said like on a net basis,

[01:07:22] he doesn't think it really means anything for GDP growth. He thinks it'll kind of be what

[01:07:26] it's always been. There'll be winners and there'll be losers. And I thought you were a great person

[01:07:30] to ask that question. Do you think this is something that stimulates GDP growth over time?

[01:07:34] Do you think it's something that is deflationary and refuse like overall and decrease inflation

[01:07:38] over time? Like when you think about the overall economic impact of AI, how do you think

[01:07:42] about that? I'm pretty convinced that it's going to have a very substantial impact on productivity.

[01:07:52] We're mostly seeing those impacts now are in the developer community. So there are new tools on

[01:08:04] the market like cursor and Ater and Zed or just the chat GPT interface or claw.ai interface where

[01:08:18] you can give a model instructions about what you want to do in code and it will write the code for you.

[01:08:29] And then maybe it takes a couple of turns to kind of refine the code that is produced, but

[01:08:38] with a sufficient number of turns, it will eventually produce the full code working code

[01:08:45] for whatever application you are looking for. Now it's not yet at the level where it can write

[01:08:54] enterprise level code. The challenge with enterprise level code is often it comprises

[01:09:00] millions of lines of code and the context window or the amount of information that current models

[01:09:09] can pay attention to at any given time to make decisions is just insufficient yet for the largest

[01:09:17] code bases. But I mean, Google's Gemini 1.5 Pro model has a 2 million token context

[01:09:25] window so call it 1.5 million words about 10,000 lines of code. So you're like we're starting

[01:09:35] to see that and there are new architectures that are coming out Samba and Mamba, other types of

[01:09:40] architectures that are demonstrating orders of magnitude, larger capability in terms of the

[01:09:48] size of the context that it can source from in order to make inferences or solve problems

[01:09:55] or write code. So it's just a matter of time before we get there. The challenge with

[01:10:02] extending AI into other domains at the moment is that the AIs are currently call it the best ones

[01:10:13] are they kind of get tasks right about 95% of the time, right? Which sounds great. But for complex

[01:10:24] tasks, what typically happens is you've got an agent that goes and runs a step of the task

[01:10:31] and it returns some information or some inferences. And then it passes it along to another agent.

[01:10:38] An agent is just like a call to chat CPT or Claude, whatever. And where the agent typically has a

[01:10:47] built to have a specific function, right? So maybe one agent is really good at

[01:10:52] figuring out search queries and going out to Google and bringing back

[01:10:56] information. And that's giving it a hand it off to another agent that is very good at identifying

[01:11:03] out of the large set that the original agent brought back, it's very good at identifying

[01:11:09] which segments of that set are most relevant to the problem. And then it's going to hand that off,

[01:11:16] though that smaller set off to an agent that's going to do some work that requires

[01:11:22] some information, right? So whatever, now you've got kind of three steps. The thing is,

[01:11:27] after kind of 10 steps, you've got a 50% failure rate because that 5% failure rate

[01:11:35] compounds at each step, right? So what you kind of need our agents that are able to have kind

[01:11:42] of like a 99.9% success rate on each task. And that way, as it's passed along, there are ways to

[01:11:53] ameliorate these, there are some unbelievably complex agent models that have some deterministic

[01:12:01] steps and some agent like steps, we're able to get really long way on complex tasks.

[01:12:09] Despite the kind of 95% success rate, for example, if you kind of know what the desired state is at

[01:12:17] any given milestone along the way, then you can course correct. But anyway, there's a bunch

[01:12:22] of different directions here, but eventually you're going to have agents that are going to be

[01:12:27] able to be much better than humans at, for example, fund accounting, tax accounting,

[01:12:33] accounting, evaluating legal contracts, developing cases for litigation or defenses for litigation,

[01:12:44] performing financial analysis for the purpose of portfolio allocation of security selection,

[01:12:52] etc. And the progress in robotics, which is increasingly using a lot of the same

[01:13:01] learning architectures that we've discovered are most effective for the building of our language

[01:13:07] models, we're seeing unbelievable progress in the robotics domain as well. And so,

[01:13:16] once we get robots that are able to effectively empty the dishwasher or do the dishes,

[01:13:25] prepare a meal, once you've got cars that are able to go and drive your kids from school to

[01:13:32] activities or what have you, we're very close. When I say very close, we're sort of like

[01:13:39] on bad or maybe one or two years from seeing tech that is able to effectively take on these

[01:13:49] kinds of tasks at or better than and more reliably than humans. And this is just going to

[01:13:56] extend at an exponential rate because we're able to use AIs and robots to build better AIs and robots.

[01:14:07] See, you eventually end up on this sort of double exponential curve of development.

[01:14:13] And so this is the type of change that I think that we all need to be prepared for. And why I

[01:14:21] think that the most important skill for people who are going to be active in the labor force

[01:14:30] at all over the next five, 10, 20, 50 years is going to be the ability to

[01:14:37] have self-long learning be self-curious and self-motivated to innovate, make improvements,

[01:14:48] to have ideas and work with AIs to bring those innovative ideas into operation and production

[01:14:56] because we're no longer handicapped. Right now, we're no longer handicapped by not knowing

[01:15:05] front-end coding skills or coding skills to prototype applications in most

[01:15:14] development languages. Right? That is no longer a barrier. It's no longer really a barrier

[01:15:21] to write complex documents. Right? I mean, a group out of Japan has now

[01:15:27] dropped an open source repo where they demonstrate how to write scientific papers.

[01:15:34] Right? So test, generate hypotheses, write the code, gather the data, run the analyses to test

[01:15:46] the hypotheses, generate summary tables and graphs, charts, make inferences from the tables,

[01:15:57] charts and summary outputs, and then actually write a complex scientific paper and then do that

[01:16:04] iteratively. So you've tested one hypothesis, you've got an outcome, what other hypotheses

[01:16:12] are inferred or implied by this new information that we've gathered. Let's go test these other

[01:16:22] hypotheses and show how this compounds over time into automated scientific research and discovery.

[01:16:30] Right? So we're already at this stage and it's not long before this extends into virtually every

[01:16:36] area of the labor market in my opinion. Yeah, it's interesting. When you hear the really

[01:16:40] good VCs, I've listened to some of them on podcasts, that robotics thing is something they're talking

[01:16:43] about a lot. They're saying it's coming faster than you think, it's going to be bigger than you think,

[01:16:48] like it's going to be a huge part. These robots are going to be doing things that we don't think

[01:16:51] are imaginable right now in a shorter time period than we think is possible. Oh yeah,

[01:16:55] completely agree. The other thing I thought was interesting though is at the end of the podcast

[01:17:00] with Aswath, he does those detailed valuations of companies and one of the computer science

[01:17:06] professors at NYU basically made a bot of him, took everything he's ever done because all his

[01:17:11] classes are online, took every class, everything he's written on his blog, took everything and

[01:17:15] they're competing. He's competing against five of the students right now to see can the bot do

[01:17:20] better valuations than he can. And he wrote this blog post Aswath did. He's asking himself this

[01:17:26] question all the time now, what can I do that my bot can't? And like that stuck with me ever

[01:17:30] since the podcast. Like it's something we probably all should be asking ourselves

[01:17:33] because he's thinking like, this bot's going to be able to do this. Like what does it mean for me?

[01:17:38] And I'm thinking the same thing with everything I do. Like we probably should all be asking

[01:17:41] ourselves this question, like what can I do that my bot can't do? I completely agree. I love

[01:17:48] that Dr. Demotorin is doing this. It's the kind of experiment that only a tenured professor

[01:17:56] can run, right? Where he's not threatened by AI. If he's made obsolete, whatever,

[01:18:06] he could do almost anything and still get paid for it. So only a tenured professor can really

[01:18:12] run that kind of experiment. But it is astonishing the number of tasks that we are able to get

[01:18:21] mostly done in our area of expertise that we can get done by AI now using fairly simple Asian models.

[01:18:34] And in some case, just a single call to the language model. And that scope is going to

[01:18:41] widen at a double exponential rate. So yeah, absolutely. We all should be asking the question,

[01:18:49] what can I be learning about doing or thinking about being innovative for

[01:18:55] that the bots are unlikely to catch up to me on? And how can I best be making use of this tech

[01:19:03] to give me the time to think about and innovate and create new ways of doing new things?

[01:19:11] Well, Adam, thank you. As is usually the case when you come on, I could have gone for three

[01:19:14] hours here and we wouldn't have run out of things to talk about, but we have to cut

[01:19:17] at some point. Thank you so much for doing this. If people want to find out more about you,

[01:19:21] about return stacked ETFs, where can they go? I am at at gestalt you dot com on or at

[01:19:28] Gestalt you rather on Twitter to find out more about resolve invest result calm. We are a co

[01:19:36] manager of the return stacked ETFs learn all about the return stacking concept at return

[01:19:43] stacked dot com and learn all about the ETFs at return stacked ETFs dot com. And feel free to email

[01:19:53] me or anyone else on the team if you have any questions or want to figure out how these can be

[01:19:59] useful in your portfolios. Thank you, Adam. Really appreciate the time. Thank you. Thanks so much

[01:20:06] for the great questions. This is Justin again. Thanks so much for tuning into this episode

[01:20:11] of excess returns. You can follow Jack on Twitter at at practical quant and follow me on Twitter at

[01:20:17] at JJ carbono. If you found this discussion interesting and valuable, please subscribe

[01:20:22] in either iTunes or on YouTube or leave a review or a comment. We appreciate