During their recent episode, Taylor, Carlisle, and Seth Cogswell discussed:
- The Passive Investing Problem No One Talks About: Hidden Risk
- Why Apple and Microsoft Could Fall 50–70% and Still Be Overvalued
- The AI Paradox: Why Tech Burns Cash While Other Sectors Capture the Value
- AI As An Existential Crisis
- The Forgetting Curve, AI, and the Future of Human Memory
- Consistently Not Stupid: The Three-Pillar Strategy Behind Running Oak
- Why Predictability Is Undervalued: Boring Compounders
- The Hidden Gap Between Expected and Realized Growth
- Why Traditional Valuation Metrics Fail—and What Works Instead
- Avoiding the 50% Trap: Thinking About Downside Risk
- The Momentum Machine: How Passive Flows Inflate the Biggest Stocks
You can find out more about the VALUE: After Hours Podcast here – VALUE: After Hours Podcast. You can also listen to the podcast on your favorite podcast platforms here:
Breaker
Overcast
RadioPublic
Stitcher
TRANSCRIPT
Tobias: Which one’s better?
Jake: I don’t know.
Tobias: We’re debating when we start livestreaming. I think we’re live. My dashboard says, we’re still setting up, but apparently, we’re broadcasting. I’m Tobias Carlisle.
Jake: As surprised as anybody. [laughs]
Tobias: This is Value–
Jake: I think–
Tobias: Let me try again. This is Value: After Hours. I’m Tobias Carlisle, joined as always by my cohost, Jake Taylor. I was going to say co-star. Good grief.
Jake: Ooh. Costar? I like it.
Tobias: Our special guest today is Seth Cogswell of Running Oak. How are you, Seth?
Seth: I’m great. Thanks for having me.
Tobias: Thanks so much for joining us today. Let’s talk a little bit about your investment strategy and your philosophy. Let’s get going. What is your investment philosophy, and how is that implemented in your strategy?
Consistently Not Stupid: The Three-Pillar Strategy Behind Running Oak
Seth: So, if I’m going to describe our investment strategy in three words, it’s consistently not stupid. [Tobias laughs] That’s based on a line by Charlie Munger. Charlie Munger attributed much of Berkshire’s success, not to being geniuses or the smartest guys in the room, is more just being consistently not stupid. And it fits our strategy so perfectly. So, when I read that, I was like, “All right, that’s it. I’m stealing this.”
Our strategy is clearly not stupid, and that it’s built upon three very simple commonsense attributes. So, the first is maximize earnings growth. Because if you’re investing in companies that are growing in value, great, there’s no arguing that that’s not a positive. The second, though, is being very disciplined around valuations.
If you have a company that’s worth this in the real world, it’s operating in the real world, and then you have a stock, which is– The price of the stock is determined by supply and demand, which hopefully at some point is equivalent to the company in the real world. We want to buy when the stock is undervalued, we want to sell when the stock is overvalued. So, again, maximize earnings growth, discipline around valuations, it does not pay to own assets that should go down, let alone a lot over the long run.
And then, the last is a focus on downside risk, for the very obvious reason that you drop 50%, you have to double your money to get back to flat, which is very hard to do, as opposed to if you experience a 30% decline, which is never fun, you only need a 40% return versus 100 to be making new highs.
So, again, it’s not stupid. It’s super simple, obvious common sense. And then, that consistently our process is rules based. So, we do the same thing over and over which makes for a reliable portfolio or strategy that again any client or advisor can plug into portfolios and expect– They’ll have reasonable expectations for precisely what we will be providing over time. So, again, consistently not stupid.
Tobias: Let’s talk a little bit about growth. How are you thinking about growth? How are you diagnosing it? What are the conditions for growth?
Seth: Growth is tough, because it’s forward looking. One of the reasons why our strategy is a little more quantitative in nature and rules based, is because my father who created our strategy started out in the 1970s at a large bank doing discounted cash flow models which anybody that’s ever had a pleasure doing that, you’re basically trying to predict the revenues of a company three to five decades into the future, despite not actually knowing what those products are. So, good luck on any of that.
And so, the strategy was largely a result of going a completely different route and really trying to focus on what’s known versus unknown. But again, growth is looking forward, so it’s certainly unknown. And that ends up being a smaller part of our approach. But regardless, the way we look at growth, and there’s a number of ways you could look at it, but we focus on earnings.
Earnings can be manipulated. So, maybe one could argue that cash flow is a good metric to look at. We choose earnings, because again, that’s really what’s coming back to the shareholders over time. We measure growth in a number of ways. We look at the one year expected return, three and five. The goal is, again, it’s inherently faulty, but we want to position the portfolio in a way, so that it is consistently on average creating more value for clients over the long run. So, historically, our portfolios average 12% to 13% earnings growth versus 6% to 7% for the S&P. So, that 6% gap, we would expect to provide value over time.
Why Predictability Is Undervalued: Boring Compounders
Jake: It sounds like you’re looking often for predictability in the business. Do you think that the market tends to overprice or underprice, predictability?
Seth: I would say underpriced, actually. Predictability is boring. Really the sweet spot that we’re looking for the inefficiency is companies that are, again, growing at a good rate. They’re high quality, which I describe as just well run. So, they’re profitable, they’re not taking on crazy amounts of debt, which I think will matter a lot in the near future, in that you can get them at an attractive value.
Usually if you’re getting something that’s relatively undervalued, it’s because they’re not exciting. You’ve got money flowing into the most excited names, whether it’s the Mag 7 right now, which everybody knows. Everybody either owns the product, or uses the product, or at least knows the company. Their hair stylist probably told them they should invest in it, just the other day.
And so, you’ve got all this money flowing into these names, not because of numbers. It’s flowing into these names because of popularity. And that popularity takes those companies up, it takes the index up. And then, the companies that are less popular just sit there, because they’re not receiving an appropriate amount of money. And so, that creates that undervaluation.
But again, growth is inherently unpredictable, trends are kind of unpredictable. That’s not to say that we shouldn’t pay attention to them, but be hard to predict. And so, the things that are most predictable, such as, let’s say like industrial companies that are just rolling along doing their thing, nobody cares about them, and so it creates that relative undervaluation.
Tobias: Do you have any idea– Have you modeled what growth you’ve actually received versus what growth you’ve expected? Do you find that you’re on the money or does it fall short, or you do sandbag a little bit and you find that you’re surprised? Have you tracked it?
Seth: I wouldn’t say that we’ve tracked it directly, but I will say at least from 1989 to 2013, before– I launched Running Oak in 2013. Again, my father ran the strategy prior to launch. I actually ran it for five or six years from 2007 to 2013 or so. But regardless, from 1989 to 2012, 2013, portfolio outperformed the S&P by roughly 3.5% before fees annualized. Just top one percentile returns or so.
But again, what we have tracked, is that on average our growth rate was 12% to 13%. So, there is a gap there between if we’re outperforming by 6% on a growth as far as growth goes, but then outperforming by 3.5%, which is still great, clearly, there’s a gap there. So, I don’t really know how to answer your question other than to say there definitely appears to be a little slippage there.
Tobias: Yeah, a little compression, which is about what you’d expect to see. That’s what I would have expected to see anyway. Let’s talk about valuation a little bit. How do you think about valuation?
The Hidden Gap Between Expected and Realized Growth
Seth: From what I’ve seen, it’s a big world out there. There’s a lot of people in our industry. We value stocks very differently from most. Even just talking to you before the podcast, I would imagine you’re taking a much more, maybe discounted cash flow really, but certainly a much more fundamental approach or a deep dive, what we do as a relative valuation versus the S&P 500.
You can think of as two ratios where we measure the wealth creation of a company versus the wealth creation of the S&P. And by that I largely mean earnings growth. Again, this is five decades. Over time, we have found that price will follow wealth creation, or relative price will follow relative wealth creation.
So, if we see a company that’s creating wealth or earnings at 2X the rate of the S&P, we would expect over the long-term for the stock price to do the same. Again, because stock prices are driven by supply and demand, they’re swinging around that true value. One of the things that’s really illustrative to me to give me confidence, just that we’re moving in the right direction as you can see that relative price swinging around that relative value. So, again, it’s different. But we’ve been doing it for a long time, and it seems to provide value or have merit.
Tobias: Is it like you’re looking for a market PE with a higher growth rate? Is that the way you think about it, or better than market PE with a higher growth rate? Is that kind of the mix?
Why Traditional Valuation Metrics Fail—and What Works Instead
Seth: Kind of. PEs have been proven to be absolutely worthless, as far as predicting returns. But PEs are also a helpful snapshot. They’re backward looking, but it gives you a chance to see where things lie right now.
Forward PE, you could maybe say is a little bit better. But again, it’s a snapshot. It’s just one year looking forward and inherently it’s faulty also, because it’s based on prediction. Yeah, so, that’s a good way to sum it up. It’s different, but that’s still a good approximation.
Tobias: So, do you favor a DCF type approach then? You don’t like the DCF, or you do favor a DCF?
Seth: DCF is almost entirely prediction. And so, our strategy is intentionally built on characteristics that we feel that we can rely on. So, again, companies that are increasing in value intrinsically in the real world. Never mind the stock market, because that’s going to do whatever it’s going to do based on emotions and FOMO, especially right now. But then, also being where does that relative valuation of the company lie as far as the stock price versus what we determine the value to be in real life. And then, again, we focus on a number of things that I’d say clearly lead to greater downside risk.
But again, those are three very simple principles or metrics. Again, the valuation part of it is a– I don’t really like the term proprietary, because it sounds a little stuffy, but it’s definitely like a metric or a calculation that I’ve never seen anyone else doing. It’s not DCF. But again, it’s a more thoughtful way than looking at PEs or looking at price to book and allocating just based on whatever has the lowest PE or the lowest price to book.
Tobias: Okay. Talk to us about downside risk. What sort of things are you looking to avoid, or what red flags are you looking for in your downside risk assessment?
Avoiding the 50% Trap: Thinking About Downside Risk
Seth: So, over time, if you go back to 1989, again, our portfolios had basically 50% of the average drawdown of the S&P, which for me, I feel average drawdown is the best metric to really measure or gauge the discomfort that clients experience when a market goes down, because it measures both how much you’re down as well as how long you’re down there. So, for me, average drawdown is a much better way to really, again, measure how a client feels risk than say, downside capture or standard deviation or Sortino or anything like that.
The biggest driver of that, in my opinion, is avoiding overvalued companies. Valuing companies is hard to do. A lot of people do it in different ways. A lot of them have merit, some probably don’t. But regardless, at least philosophically, you can’t argue that owning a company or owning a stock at $100 when the company is worth $50 isn’t begging for trouble at some point.
Why Apple and Microsoft Could Fall 50–70% and Still Be Overvalued
There’s a few great examples right now. There’s probably a lot of great examples right now. Apple, Apple’s an awesome company. Two of my closest friends work there. I wish it well. I love their products. But its current PE is $39 or so. Long-term, it’s $14 to $15. That’s its average where it’s at for a long time. That means that Apple could drop 50% tonight and it would still be really overvalued relative to its historic norm.
Again, great company. But same with Microsoft. Microsoft has a price of sales of almost $14 now. One of my favorite quotes within the market is a quote by Scott McNealy, who is the former CEO of-
Tobias: Sun Microsystems.
Seth: -Sun Microsystems. I believe he was getting a hard time for losing people’s money. I’m sure he was sued for people investing at the peak of the tech bubble, and then getting mad at him. But he was like, “Look, I’m running the company. You’re the ones that paid 10 times price of sales. That meant that over 10 years, we had to return every single penny in revenue just for you to break even. Not make money for you to break even, which assumes that we have no employees.
We’re running a big company. We had a lot of employees. It assumes that we have no cost of goods sold or R&D. We’re a tech company. We have to constantly innovate. And also assumes that you don’t pay taxes. If you guys don’t pay taxes, you go to prison. So, don’t come to me about you paying 10 times price to sales.”
The thing is, both Microsoft and Oracle are 40% beyond that. So, 40% beyond the very peak of the tech bubble. Now, we’re in a dynamic time. I don’t know how this plays out with AI. I have a sense of how it plays out, but regardless, I don’t know how it’ll play out. But the main thing to answer your question, is that’s just risk. 40% beyond the tech bubble, maybe it works out, but what if it doesn’t? And so, that’s our goal, is to really avoid that what if. Because what if almost always comes to pass. It does very often.
And so, Microsoft, again, awesome company, seems to be leading the AI battle at the moment. But it could drop 70%, and it wouldn’t be a big deal. It would not be a steal if Microsoft drops dropped 70%. Maybe on a PE basis, it would be a lot more attractive.
So, again, that focus on valuations and avoiding companies that should go down, let alone a whole lot, is probably our biggest, our greatest value add. But another area where I feel will be very important over the next decade is debt. We all know that if we max out our credit cards to go buy a boat or something, that it’s probably not going to work out very well. We’re headed for some pain.
In the last decade, companies took on more debt than any time in history, not to build better companies that could then pay that interest, pay down that principal, and make profit on top of it. They just bought back stock. Buying stock has some benefits. But if you really think of it as most simplest form, basically, these companies mortgage their futures because they got to pay that back at some point, and then they handed that cash to discerning individuals who were selling their stock.
Tobias: [laughs]
Seth: Again, it’s neither good nor bad buying stock. It could be both ways. But if you’re mortgaging your future to do it, that’s taking on risk. And they did that when interest rates were quite low. Interest rates are now a lot higher. So, more than likely, these companies are going to have to a lot of them are going to have to refinance at higher rates, which will hurt profitability.
Profitability is one of the main arguments for better or worse. I actually think it’s a dumb argument. But one of the main arguments for higher PEs or higher multiples. So, if you see profitability decline and you see multiples decline, now you’ve got like a double whammy and contraction. And that just assumes the company is performing just as well as it always has. That’s not assuming any adversity. That’s assuming just reality. And so, avoiding that, I believe, will be a significant value add going forward.
Tobias: You’re starting to sound like us, Seth.
Jake: [laughs]
Seth: I think that’s as a complement.
Tobias: Fitting right in.
Jake: Other than that, though.
Seth: Let’s go.
Jake: Yeah. [chuckles]
Tobias: What a crazy bet.
Jake: Yeah. [chuckles]
Tobias: Let’s talk a little bit about passive. You’re not a fan.
Seth: [laughs]
Jake: [chuckles]
Seth: I’m trying not to position it in that way to say I’m not a fan. I think one of the issues that drives me craziest in the world today, both in real life and investments, is this tendency to see everything in black and white. It’s either my team or the other team. Most aspects of life are in the gray, and passive is very much that. So, I’d say that John Bogle had a very positive impact on the industry.
Prior to Vanguard, many fund managers were charging massive fees providing no value whatsoever. That didn’t benefit anyone other than, I guess, a handful of managers. But in the end, any industry or any company that’s not providing service or value to the end client shouldn’t exist. They should just die quickly. And the investment management industry didn’t seem to be providing a whole lot of value for the most part.
Bogle provided an alternative. He was like, “Look, if you’re not even going to get market returns, or if you are and you pay a high fee for it, why don’t you pay little to none for it?” Fine, there’s no arguing that paying a lower fee for the exact same thing you were paying a high fee for isn’t better. And that also had the benefit for investors of bringing fees down across the board. So, that’s a positive.
The other thing is a lot of people made a lot of money in the last 15 years being invested in passive, which is a good thing. Now, the issue is, why did that provide so much value? And that’s my concern, is that people aren’t really considering context. So, again, lower fees? Good. Investing in a strategy that to a certain extent makes no sense, that’s bad.
When I started in the industry, passive and active, that’s really when this battle was really raging. And the idea of investing in a company simply based on size just seemed like the dumbest way to invest. I mean, if you’re looking at two houses, you’re not going to pay 25% more for a 5,000 square foot house than you would for a 4,000 square foot house without at least walking inside the house. You’re probably going to do a home inspection. There’s no investment where you would just be like, “Ah, this is bigger. I’m just going to pay more. I’m going to put more of my money in it.” But that’s the way that passive works.
Now that said, as time has gone on, my view has evolved. One of the things that I have become particularly [chuckles] upset about and worried about, is whenever a manager, anybody in investments as well as advisors, whenever we speak with a client about investment, we speak about two things. What are those two things that we would talk about? The two main things.
Tobias: Oh, that’s a question for us. [laughs]
Seth: [unintelligible 00:22:05]
Jake: Returns and fees?
Seth: You’ve got one of them. No. Return? The fee matters. But what’s the other one?
Tobias: Risk. I don’t know. [laughs]
The Passive Investing Problem No One Talks About: Hidden Risk
Seth: Yes. Which one matters more? I would argue that risk in the long run matters more, because that money matters. People have forgotten that risk matters, because we haven’t experienced a true downturn in 15 years. But when you need that money that you’ve worked so hard for and it’s not there, that is a whole lot worse than if you have more money than you needed.
And so, passive has somehow gotten away with completely overlooking risk. The only way that the passive narrative addresses risks is they say, “Eh, market goes up on average. You’ll be fine.” That’s absolutely ridiculous. What if you’re retiring in the next three years and we go through a lost decade, which the odds look pretty high of at the moment. You’re having to pull out money after a big drawdown which absolutely destroys your cumulative returns over the long run. There’s a lot of data points. There’s no arguing that.
The other thing is that’s conveniently overlooked is the behavior gap. I am not immune to making dumb decisions when I’m emotional. I’ve made plenty of stupid decisions. When we, as people, are stressed out, if we see our net worth drop 50%, we are inclined to panic and sell, because we don’t want to experience any more. So, yes, we’ve got the people who sold us passive, we’ve got the S&P saying, “Oh well, it’ll be okay. Just hold on.” Are you going to want to pile more in? No. And you might sell. And that destroys returns. So, conveniently, completely overlooking that also bothers me.
Sorry, I’m getting a lot long winded. I clearly am a little passionate about this. But the other thing that I recently had an epiphany on, is we’d have this active passive debate and everybody brings up the data, which I would argue as, henpecked as most data is, I don’t think a lot of it’s valid. But regardless, if you break passive into its simplest form, so if we just look at it on an individual stock basis. Again, we’re going to use Apple, because everybody knows it.
Apple is a company, it’s operating in the real world, it’s selling products, it is worth this amount of money. If it’s a private company, we would work with number of auditors to come up with a value that somebody would buy it for. It’s a company. It has a value. Now, the stock is driven by supply and demand. So, if there’s a lot of demand, it goes up. If a lot of people are selling, it goes down, whatever. It’s swinging around. Let’s say at one given point in time, we used to think the market was efficient, that many would argue that it’s just broken now. But regardless, let’s say at some point, the market is efficient.
The price of the stock is equal to the value of the company. Somebody bought an iMac, two months ago, they’re doing something on it, and they’re like, “Man, my iMac is the greatest. I love it.” And so, they turn around and they buy, let’s say, a million dollars of Apple. They’re pretty wealthy. Buy a million dollars Apple. That demand is going to push the price of Apple up. Again, nothing changed with regard to the company, because they bought that iMac months before. But now, the stock price is higher. Because the stock price is higher, it’s now a higher percentage of the S&P. And anybody that invests in the S&P from here on invests more in Apple, because it’s overvalued.
Then we were talking earlier about momentum. I can’t imagine how much money is in momentum strategy at this point. But regardless, a momentum manager says, “Hey, Apple’s flying. We need that.” And so, more is invested in Apple that’s going to push Apple up even further, that demand. So, now, because Apple is even more overvalued, it gets an even higher percentage of the S&P, which means that everybody invests in Apple now buys even more, because it’s overvalued. And that keeps going.
The Momentum Machine: How Passive Flows Inflate the Biggest Stocks
We hit the 99.8th percentile in history in momentum last year. That dynamic of paying more and more for companies, the more and more overvalued it gets, is historic. And the thing is that the S&P or passive never sells. They don’t sell unless it gets to a certain point where the committee that runs the S&P says, “You know what, we should kick this out.” So, it’s literally the exact opposite of buy low, sell high, which as a value investor or as anyone that’s even remotely thinking critically about investing– Everyone’s hopefully looking to get good value on the buy and to sell when the value is positioned to maybe work against your view or you’ve realized that value, passive is the exact opposite.
So, again, passive is neither good nor bad. It’s good in that it brought fees down. It’s good in that it really simplifies investing. It’s certainly better than chasing mutual fund returns or day trading. But I do believe that there are far more thoughtful, far better approaches out there. And so, it’s in middle.
Jake: I would say to the abdication of responsibility of ownership of all of these businesses is actually adds a fragility to the future cash flows of those businesses. [chuckles] So, you’re creating your own problem that you’re going to eventually that chicken has to come home to roost.
Seth: That’s certainly a concern. You’ve got the investment side of it, and you’re holding a momentum portfolio when momentum just hit the 99.8th percentile in history and not recognizing that you own a momentum portfolio that just had the hottest period in history and not selling, if anything, people are piling more and more, that’s one thing. That’s a lot of risk.
The other thing to your point, is the real-life implications. Capitalism, we can thank capitalism for the massive gains in our quality of living over the last 50 to 100 years. At the heart of capitalism is capital being allocated intelligently into companies that have a lot of upsides and allocated away from companies that should disappear, and that mechanism has been completely broken.
Now, part of that I would say is passive. You got a lot of money going into companies for no reason other than size or no reason other than popularity in this construction. You also have a lot of zombie companies out there that should have died a long time ago, but have been kept on life support, thanks to massive fiscal stimulus and 0% interest rates. But yeah, I mean whether it’s on the investment side or in real life economy, it likely has some very significant negative implications going forward.
Tobias: Yeah, I couldn’t agree more. I thought one of the interesting things that I looked at over the last– I’ve had that running deep value. I’ve had the stuffing kicked out of me for about a decade now. I’m quanty, too. So, one of the things I like to do– I’ve run data back to 1926 in the French data. You guys can do this, too. It’s just a French size series.
If you run the top decile against the market, which I guess is about the fifth decile, or you run the top decile against the bottom decile, you clearly get massive outperformance by buying small over large. It’s about 0.8% a year relative to the market 1.7% a year from largest to smallest since 1926. So, that’s a century of data.
In that century of data, there are these six periods of time where you get this massive outperformance for size. We’re in one of them now. We’re 10 years into one now. So, I did a count. There have been something like 10 that have been three years long, 7 that have been five years long, and there’s 5 that were 10 years long.
Jake: What’s the longest?
Tobias: The longest one was 1999. I think it was 13.5 years and we’re currently– Or, maybe it was over 15 and we’re currently in one that’s 13.5, something like that. This is the second longest.
Jake: So, don’t worry about it for a couple more years?
Tobias: [chuckles] I’d say, you’re good to go. [Jake chuckles] Close your eyes. JT, we’re coming up to the top of the hour. Let me do a shoutout to all the folks playing at home, then we’ll do-
Jake: Some veggies.
Tobias: -veggies. Tomball, Texas. Petah Tikva. Bethesda. Cottage Seven. That’s a new one. Welcome. Philly. Lewes, Delaware. Thanks for the pronunciation last time. Toronto. Valparaiso. Is there someone else in Valparaiso besides Mac? Limerick, Ireland, what’s up, Colm?
Breckenridge. Tallahassee. London. London, UK. Tampa, Florida. Dead Cat Gully, New South Wales. Me, too. New London. Mayfair, London. Milton Keynes. Menatwork. Nice, Les. You got me. That’s in Australia. Madeira Island, Portugal. Hunter Valley, New South Wales, what’s up? Boise. It must be daylight saving in Australia. The Aussies are on.
Jake: [chuckles] Still pretty early there.
Tobias: Philly. Transylvania, what’s up? Is that real? Paresh Patel, Ridgewood, New Jersey. Transylvania. That’s a new one. All right, that’s a good spread.
Seth: Popular with the vampires.
[laughter]Jake: We score well.
Tobias: They’re the only ones who’ve seen value working. [Jake chuckles] JT, hit us with some veggies, make benefit, glorious nation of Value: After Hours.
The Forgetting Curve, AI, and the Future of Human Memory
Jake: All right. So, centuries ago, Aristotle wrote a short essay called On Memory and Recollection. And in it, he described memory as an imprint, like a signet ring pressing into a soft, heated wax. “Experience,” he said, “leaves behind a trace, not just an image, but a kind of echo in the soul.”
When we remember, we’re perceiving that echo again. And when we recollect, when we actively try to recall, we’re searching for the echo in the wax. It’s very elegant, but it’s fatally flawed in that the wax tends to melt. Time, distraction, new sensations, all of them deform that original imprint. Memory fades, because life keeps pressing new rings into the same surface. So, that was the philosophy. And for nearly two millennia, memory stayed as this mystery that was wrapped in a metaphor that Aristotle gave us.
But then, in the late 1800s, a young German psychologist named Hermann Ebbinghaus decided to do something radical. He started measuring memory itself, and he turned this philosophical fog into actual data. Ebbinghaus was born in Germany, and earned his doctorate at the University of Bonn. He was inspired by this emerging precision of psychophysics that was bringing tangible rigor to all of these mental processes. He ran exhaustive self-experiments, and gave us the first quantitative map of forgetting, the curve that now bears his name.
So, Ebbinghaus wanted precision, and he wanted to quantify what happens between learning something and then forgetting it. So, he became his own little lab rat. And for two years, he conducted hundreds of experiments on himself. And to strip away the meaning and the emotion of the memories, he invented these little nonsense syllables like Z, A, K, W, I, D, T, O, V. They didn’t mean anything, but he was just trying to memorize them. And that way, it prevented him from having these associations that might influence the memory.
So, he memorized these long lists of all these random things, and then he’d retest it at 20 minutes, at one hour, one day, and one week. And he was measuring the rate of decay in the loss of the memories. What he found was that would become one of psychology’s most enduring images, this sharply descending curve that’s steep at first, and then it gradually levels out. And this is Ebbinghaus’ forgetting curve famously now. He discovered that within 20 minutes, we forget about 40% of what we’ve learned. After an hour, over half is gone. By the next day, two-thirds of the material is evaporated. And after a week, we retain only about a quarter. And that pattern is this exponential. But the same shape has replicated in study after study now over a century. So, it’s actually probably pretty good science.
It’s not just really a grim reminder of how fragile our memories are. It’s also kind of a map to guide us on how to resist that decay. So, he found that if you review the material not consistently, but strategically at certain intervals, you fight that forgetting curve. Each repetition and each active recall strengthens that trace in a memory. And he called this principle his savings method.
So, even if a list seemed forgotten, relearning it took less time, there was like a trace that remained that you could pick up the thread. So, this insight became the seed of what’s now called spaced repetition, which is a very common learning tool where you just learn things at certain intervals to recall before you forget. So, what’s actually happening when we forget? Ebbinghaus could measure forgetting on himself, but he couldn’t really explain it. We need neuroscience today to give us a much fuller picture.
They’re a few different theories on how this works. Trace decay suggests that memory traces, these neural patterns, form during the learning. They simply just fade over time, like footprints that are washed away by the tide. There’s also interference theory which says that new memories disrupt old ones. Sometimes that old knowledge blocks new learning. Other times, new information is overwritten by the old. There’s also retrieval failure, which is you have that memory, but you can’t quite find the right key to access it. It’s like having the file in your brain, but maybe the label peeled off of it.
And then, there’s motivated forgetting. That’s the mind’s protective reflex to very painful or useless memories. Sometimes forgetting is not really a flaw, but more of like a defense mechanism. So, like mothers forgetting about the birth of a child, then the pain of that, or value guys eventually forgetting the last 10 years.
[laughter]So, anyway, let’s fast forward today. We live surrounded by infinite storage. Every photo, every note, every message is archived somewhere in the cloud. We can search through these archives with a keystroke, but that’s not really the same as carrying the knowledge in your bones, like, we forget faster than ever. We outsource recall to our devices. We rely on technology to remember for us, but that changes how we know things. We may have access to more information than any generation in history, but maybe perhaps less internalization of that information.
So, to remember something is to really integrate it and to feel the pattern and not just recall the fact that. AI systems are becoming increasingly more of this external memory, capable of infinite recall for us. They index, and sort, and retrieve without fatigue. When we set these algorithms to hold our memories, we’re trading retention for convenience, perhaps. So, we know where to find things, but we don’t really know what they mean as much as we might have.
So, the real danger of AI-assisted thinking is not amnesia, but really atrophy, like the slow erosion of the interpretive, subconscious work that connects memories to meaning and understanding. We have to be careful in how we use these new AI crutches. At least, that’s the common refrain that we hear today. I’m going to try to play devil’s advocate and posit that perhaps man plus machine might unlock an immense intellectual bounty for all of us.
So, imagine an AI thought partner that doesn’t just store facts, but really trains your brain and your memory and your judgment. It watches your projects, builds personal knowledge graphs, schedule space rehearsals when forgetting risk is highest. It turns notes into targeted retrieval prompts, surfaces, contradictions and base rates at opportune times. Running quick simulations, proposing alternative hypotheses for you to consider. Raising red flags before mistakes are made. It’s like a prosthesis for your attention. Not a replacement for thought, but raising perhaps both the floor and the ceiling of what a single mind might be able to hold.
We, humans, will bring the aims, the taste, the ethics, and perhaps the risk appetite. And the machine brings this relentless recall, audit trails, sanity checks, some guardrails for us. And every decision that leaves a record has inside of it these assumptions and probabilities and rationales. And later, we can take those outcomes and score them against calibration of improving and feedback loops tightening.
So, I think teams perhaps could become even more antifragile working together, fewer forgotten lessons, faster learning cycles, more disciplined interactions. We don’t outsource our thinking, we use AI to help provide a scaffolding for it. Humans get to choose the ends and the machines really strengthen the means for us. So, AI doesn’t erase Ebbinghaus’ forgetting curve, but it turns it perhaps into a more viable training plan.
Tobias: That’s brilliant, JT. Is that Journalytic? Is that what’s coming?
Jake: That might be something related there. Yeah.
Tobias: That’s a cool one. I read once that if you sever a long-term relationship, like you get divorced or something like that, you stored a whole lot of information in your spouse’s brain. Like, you just don’t know where things are because you know that they know where it is. So, you already use the external hard drive, so that’s good, timely advice.
Jake: You getting divorced?
Tobias: [chuckles] No. I don’t know why I said that.
[laughter]Tobias: Not that I’m aware anyway. Yeah, you’ve thrown me there, JT.
Jake: Oh sorry.
Seth: No, I love that. It was awesome. It was a both optimistic and I don’t know, humanistic way to look at AI.
Jake: I don’t know the right answer, but I think it’s fun to try to just think through the positives and the potential negatives.
Tobias: Yeah, I’m an optimist for the AI. I love using that. I think it’s incredible. Chat and the ones that I use. We’re talking AI with Seth. It’s one of the topics that he suggested we take a look at.
Jake: Yeah, let’s hear your– [crosstalk]
Tobias: You think that most people are thinking about it in terms of first order thinking. You’ve got a second-degree view of AI that you think’s being overlooked?
Seth: Yeah. I didn’t see it going that route that Jake just took, [Tobias chuckles] but I loved it. Yeah, one of the things that I would like people to consider that I feel that they aren’t and what led to this, is we reconstitute our portfolio several times a year. We did it, let’s say, three or four months ago. And the result– Again, our process is rules based. It’s driven by numbers. It’s not our decision or subjective.
The result was we ended up having less tech exposure than I ever remember us having. We had more industrial exposure and some other things. But I was very uncomfortable with having less tech exposure during what appears to be a tech revolution. And since then, tech, obviously, almost every dollar that’s going on the market, it seems like is piling into tech-
Tobias: I think it’s more than every dollar at this point.
Seth: -and not into other things.
Jake: By the way, just real quick. I’ve heard funny– Apparently, now some allocators have– They call it a completion portfolio, which is maybe you have all your allocations and then you’re like, “Oh man, I’m like not long enough big tech probably to keep up with my benchmark. I need a completion portfolio.” [chuckles]
Tobias: Little SPAC filler filling the gaps-
Jake: Yeah, you just need to fill in gaps, so you don’t [crosstalk] confirm too much. [chuckles]
Tobias: -packing peanuts.
Seth: Yeah, yeah, I’ve spoken with a number of firms that are maybe regretting not having that.
Jake: Sorry, Seth, I didn’t mean to interrupt.
Seth: No, please. Right now, if you look at just a quick snapshot of the costs and the revenues of AI, it’s nuts. Let’s say, a somewhat conservative estimate, is that a trillion dollars will be invested infrastructure and AI in 2025. The native AI companies, so the company is really specializing on bringing AI to the market, currently have roughly $20 billion in revenue. I think that’s annualized. So, we’ve got a trillion dollars in costs which aren’t– Those are just sort of infrastructure. We’re not talking about the actual cost to generate the service, like the massive amount of electricity and whatever else. So, we’ve got trillion dollars versus $20 billion in revenue. You’ve got this massive disconnect.
The AI Paradox: Why Tech Burns Cash While Other Sectors Capture the Value
What I feel that many are completely missing, is that $20 billion in revenue has a reciprocal cost. So, companies are paying $20 billion for that service, which costs over a trillion dollars or so to provide. So, the interesting thing about AI, is it can be applied to pretty much anything to improve it. It’s not a tech thing. You can apply tech to financials, industrials, real estate, utilities, definitely healthcare. You can apply it to almost any approach to ideally, if you do it efficiently, to actually provide value.
Again, the companies that are using, that are paying very little while the companies are providing are paying a whole lot. So, everything around tech actually has the opportunity to immediately take advantage of AI and the services that these companies are providing, and they get this massive value gap. And so, people are piling into the things that are burning cash like nothing seen in history, and then they’re completely avoiding the things that can most immediately benefit. It’s pretty nuts. Who knows? Maybe that’s philosophical and I’m talking my own book, but it’ll be really curious to see how that plays out. For me, it seems obvious. But then, again, I’m not always right. That’s one of the things I’m really trying to get people to contemplate– [crosstalk]
Jake: Everyone’s going to have pink slips by Christmas time, or where’s the cost savings or productivity increases that are supposed to be justify all this spend?
Seth: Look, again, who knows? I mess around with AI. Tobias just mentioned that he’s a big fan in a number of ways. But MIT did a study. I’m going to mess this up, but it was 90% to 95% of companies polled have seen no return investment in AI. How many silly pictures are being created with AI? I wonder if that’s 50% of the usage of AI at this point is silly pictures or memes or whatever. It’s great. I guess, that improves life if you get some extra laughs out of it.
Jake: Or, targeted advertising. Hooray. [laughs]
Seth: Sure. Great. That’s what we need more of. So, I’m less positive on AI than I’d say Tobias is, as far as the impact on humanity. I think there’s positive Implications I do worry a lot about– This is getting really philosophical. But frankly, Jake, I blame it on-
Jake: That’s fine.
Seth: -you went straight on philosophy. I mean, you’re quoting a philosopher or a scientist. Regardless, I worry that– There’s a phrase by Eckhart Tolle that’s “Once a human surpasses survival.” Meaning, becomes extremely important. I would say that more people have surpassed the needs of survival than any time in history, let’s say, in the last hundred years. You’ve seen depression rates significantly increase. We’ll just completely overlook smartphones at this point. But just even in the last hundred years, you’ve seen that.
I think it’s because people are looking for meaning. But what if you start even then sucking out all their time that’s used to be productive. They go to work, maybe they do something of hopefully of value, and now that’s gone. Or, it’s even more efficient, so they spend less time in that. I don’t know, Jake, I love the way that you like position that in a way that could be positive, that maybe enables people to find more fulfillment lives or maybe increase their capabilities, so that’d be awesome.
Jake: If you had to think about that meaning and lack of meaning perhaps, would you imagine a world with UBI would have greater or less guillotine risk in it? So, everyone’s getting paid. Everyone’s can stay at home and watch Netflix or whatever. The machines are creating enough quality of life for all of us to survive, and yet most people have no meaning in their life. What are they going to do then? Are they angry about that or are they docile and happy to be in the zoo?
Seth: I think that’d be horrible. I have kids, but even pre-kids, I could probably quote Disney movies better than any grown man. Wally–
Jake: Let’s not unpack that one.
Tobias: Pixar. Pixar.
Jake: Yeah. [laughs]
Seth: Yeah, I love some Pixar– [crosstalk]
Tobias: Pixar gets surpassed.
Seth: Wally. You get to a point where everybody’s just hasn’t gotten up out of these like floating chairs where they’re just being fed whatever. It certainly seems like that is not out of the realm of possibility. At least in that, they were perfectly docile and not necessarily anarchists.
Jake: Storming the best deal.
Seth: Yeah. But if people don’t have a purpose, life is empty. So, that’s a major concern of mine.
Jake: Mm. Toby, how you think about that these days?
Tobias: Yeah, I couldn’t agree more. You need a purpose, otherwise it’s– There’s a lot of things wrong with the way that we’ve set up. Before we go too philosophical, I just want to go back to–
Jake: [chuckles]
Tobias: There’s a few things on AI. There’s a little chart that’s been doing the round showing there’s a whole lot of layoffs that coincide with the rise in AI, and then that seems to be reinforced by the fact that there’s a lot of excuses given by companies when they fire people saying that it’s an AI related firing. But there was a little research report out today that says that’s not true, which is the perspective that I have that they’ve already been.
They’ve over hired during the little 2021, 2022 stimmy, sugar high, and now they’re just letting people go, and we’re getting back to where we should be in terms of long-term unemployment. Do you have a view, Seth? Which way do you lean on that one? Is AI winning the race, or is it something else?
Seth: No, I completely agree without really numbers to back it. I think that there’s some numbers I guess that back it. It was next to impossible to hire people in 2020, 2021. Wage inflation was skyrocketing, because it was just so hard to get really good people. I do think that firms very clearly over hired then. Meanwhile, they did so when profitability was at pretty much all-time highs, which was partially driven by globalization, which is now reversing.
And so, if we see a recession, if we really see a meaningful change in globalization, if we see profit margins begin to decline, then you’re going to see the opposite of– I mean, you’re going to see unemployment rise. By pretty much, by many metrics, it appears like we might already be in a recession unless of course, you look at the most accepted metric which would be like GDP growth. I think half of the growth that we’ve experienced this year is strictly due to basically-
Tobias: [crosstalk] spending.
Seth: -CapEx investment in AI. Keep in mind, we have no idea whether that pays off anytime within the next 10 years.
Jake: Well, the chips won’t be. Thery’ll be gone by then.
Seth: Yeah. Well, I saw a number recently where they estimate that maybe 80,000 jobs were truly cut due to AI. So, yeah, to echo what you were saying, Tobias, it seems like that number, especially if you consider. Again, the MIT study, 90% to 95% of companies, they’re messing around with AI, but they don’t really have any idea how to add value. So, the odds are companies are making that decision. It’s an expensive decision to let people go. So, the odds that people are making that significant decision without actually having a plan in place seems doubtful.
Tobias: I saw that Mike Burry of The Big Short fame. He released his– I don’t know if it’s come out of his 13F or he tweeted maybe. I think he tweeted that something that– I think Jim Chanos might have said something like this six months ago. I don’t actually know where I got this idea from, but it’s something that I’ve been aware of for a little while that if you look at the rate of CapEx which is not then expensed, of course, it’s depreciated.
So, there’s a little bit of lag when it gets recognized. We’re now at the point where the depreciation, amortization, whatever, it’s going to run through the income statement and the numbers are quite– They’re hard for them to overcome even in terms of revenue, let alone in terms of–
So, we’re just not paying back the AI spend. You can already see it free cash flows falling off for a lot of these bigger companies. They’re very, very expensive. And at some point, it starts running through the income statement, so that’s how you get sort of a normalization. I don’t think it’s going to collapse. I don’t think it’s all over for those companies. I think they’re almost certainly bigger in 10- and 20-years’ time. But I do think it’s an interesting data point. Do you have any view?
Jake: Just shift the useful life out another 10 years, and that’s not as big a deal. 50-year mortgages, not a big deal. This is just-
Tobias: Kick that can, baby.
Jake: -extend and pretend everything.
Tobias: Kick that can. That’s the problem with the chips, though. Like, the useful life is so short. I think I saw some people who’ve been installing them in the centers and they say even three years is aggressive, because they run at such high heat they seem to break down a little bit faster than people have been expecting. It’s hard to know how much is like FUD, the fear, uncertainty, and doubt, and how much is like a real take on these things.
I think at face value, you’ve got to accept that the useful life is as the companies are saying that it is. But even then, the hurdle is too high for the revenue generation, they’re going to be some diminution in profits anyway over the next few years. Do you have a view, Seth? Is that something that you’re encountering as you’re looking at companies?
Seth: I’m really excited that you mentioned that, because I think it’s very pertinent. So, Burry actually also, he had a post yesterday, I believe. It laid out the period of time or the life expectancy for these infrastructure investments. The life expectancy has been growing. So, if you track it from 2020 to today-
Tobias: Yeah, [crosstalk] pushing it out,
Seth: -a lot of these companies, the life expectancy has gone from three years to six years, which means that as far as their income statements in the earnings are reporting, at least according to him, they might be 30% higher than they really are. So, he’s arguing that they are based– I don’t know, if this counts as fraud, but certainly that they’re playing–
Jake: Aggressive accounting.
AI As An Existential Crisis
Seth: There’re some financial shenanigans going on. What really backs that up, exactly what you just mentioned, Tobias, although I’d say it’s actually a little crazier, where these chips are improving in capability so rapidly. At least what I’ve read the life expectancy is probably a year and a half to two years. Now, what you mentioned, as far as the heat, that’s a whole another thing. It’ll be fine. I’m sure that–
One other thing to think about is how many towns are going to be perfectly fine with waking up one morning to turn on their water and there’s no water, because a datacenter is sucking it up. I find it hard to believe that that’s going to be perfectly fine for everybody. Regardless, there’re so many things that seem very unsustainable about all of this.
Again, it seems like especially according to what Burry said, he clearly knows more about this than I do. What companies are doing seems– I guess, unethical is probably not the right word, but there’s definitely some tweaking of the numbers. I had someone the other day ask me how this all ends? It’s really just a realization at some point, or it’s slowly and then suddenly all at once, right?
It’s the realization by many that this isn’t sustainable, that there are shenanigans going on, which it appears there very much are. Again, the life expectancy according to their income statements has doubled even though we know that it’s actually shrinking. And that by saying it doubles, it has a very significant positive impact on earnings when we know there’s actually a negative impact on earnings. It’s crazy. So, I think you’re right on. I think that it’ll be really interesting to see how this all plays out.
Tobias: It’s entirely possible we go through a dotcom style. We’re just too far ahead of ourselves and we crash. But the next 25 years, clearly, the dotcom businesses have gone in directions that nobody– We’ve done so much more than people, I think even could have conceived in 2000 to get to this point. I’m sure that AI will be the same. We’ll be using it for things that we cannot even conceive of now in 25 years. But that doesn’t mean that in the interim that there’s not a lot of volatility or that it all accrues to the companies. Like, consumer surpluses are a thing. It’s entirely possible it’s a huge consumer surplus.
But I will say that if the AI is going to consume all of that energy and fresh water, then we need nuclear, because there’s just no other way to generate enough power and enough fresh water. Although Amazon has come out with a– Andy Jassy had a tweet today where he showed– They’ve had some innovation in the backend of the datacenters where they figured out a way to cool the chips a little bit beyond, but they’re getting cold directly on the chip and that separating the water away from it, so it circulates. It’s quite water efficient. Sounds like a really impressive thing.
Seth: That’ll be good. But keep in mind how much money has already been dumped into datacenters. So, now, we have to redo all of them to potentially provide that. It’s constantly evolving, which again goes to your point that the life expectancy is getting shorter, not longer, because of innovation. We’re not at a point where there’s a steady state. We’re not at a point where we’re not improving efficiency. Efficiency and innovation brings the current life expectancy down because of that trend.
I was at a conference a few weeks ago, and a gentleman was speaking that is very much in the know. Like, he’s in the AI, in the very thick of it. He made a comment that I haven’t heard anybody else really focus on, but it really stuck with me, which was these– Let’s say, the biggest companies, the Mag 7 companies, in particular, see AI as an existential crisis, that’s what he said, which means life or death.
If you are backed in a corner and you feel you’re in life or death, is there anything you’re considering other than just purely survival? You will do anything to survive. And that if you actually take that perspective into account, then that begins to actually justify or help these silly numbers make sense. Because if these companies believe that, it’s we either are all in or we lose and we’re done, they’re going to be all in, and profitability and all these other factors don’t matter. The only thing that matters is survival.
And so, I just don’t think any of these numbers, for the most part, make sense in the short term. And the problem is valuations are crazy. Valuations mean higher risk or more downside, further to fall. And so, it’s a confluence of a number of factors that I wish people were paying more attention to, because it seems almost inevitable how this ends, but maybe I’m wrong.
Tobias: On that cheery note, Seth, [Jake chuckles] that’s time. If folks want to get in contact with you or follow along with what you’re doing, what’s the best way of doing it?
Seth: Feel free to email me at seth@runningoak.com. You can also check runningoak.com, runningoaketfs.com. And then, I’ve begrudgingly been far more active on LinkedIn, so you can hit me up there as well. I’d love to talk to or hear from anyone.
Tobias: JT, Any final words?
Jake: Just-
Tobias: The excellent– [crosstalk]
Jake: -embrace our AI overlords.
Tobias: I, for one, welcome our new AI overlords. Seth Cogswell, Running Oak, thank you very much. Everybody, we’ll see you folks next week, same bat time.
For all the latest news and podcasts, join our free newsletter here.
Don’t forget to check out our FREE Large Cap 1000 – Stock Screener, here at The Acquirer’s Multiple:
