Here are some things I think I am thinking about.
1) Something Big is Happening.
Matt Shumer wrote a mega viral article this week about the impact of AI. Drawing from his experience as an AI investor and founder, he argues that recent breakthroughs in models like GPT-5.3 and Opus 4.6, have moved AI from a helpful tool to a system capable of autonomous, high-level professional work, effectively replacing the technical requirements of many white-collar jobs. He highlights the “intelligence explosion” occurring as AI begins to assist in building the next generation of itself, predicting that up to 50% of entry-level professional roles could be eliminated within five years. It’s interesting and I’ve noticed a big change as well as the AI systems I use seem to be probing and responding in a much more human manner. I also notice, on the coding front, that the improvements are dramatic. But how worrisome is this and what will the impact be?
I wrote some high level thoughts on X about this:
“Let’s say AI turns out to be the humongous job killer that some think it will be. That would result in a huge decline in aggregate demand and negative wealth effect across large swaths of the economy. GDP stagnates, but doesn’t necessarily go down. Unemployment goes up enough to get a technical recession. The wealth accrues to fewer and fewer firms and people (the stock market could, paradoxically, go UP in this recession, after initially collapsing). Yields collapse. The Fed stimulates. QE to the max. Automatic stabilizers fill some of the void. The deficit explodes. Calls for a UBI become more prominent. This gets worse and worse until the govt is filling the aggregate demand void entirely (which they’ll be able to afford because inflation will be so low).“
That’s your outlier scenario. It’s not a great one, in large part, because social upheaval will be tremendous. If you think people are mad about inequality now just wait until this materializes. But I don’t think this outlier scenario is the most likely scenario. Or at least it’s not coming as soon as Matt might think. I think the much more likely scenario is that AI continues to improve dramatically, but that adoption remains relatively slow. I believe the result of this is more of what we’ve seen so far where firms are essentially in a hiring freeze where they try to soak up more from existing workers without necessarily leveraging more into new labor. So the labor market remains soft, not because firms are firing workers, but mainly because they’re not in a rush to hire as they cautiously navigate the AI evolution. This cautious approach to AI adoption will make the labor market appear weak, but I don’t think we’re going to see a widespread loss of jobs unless we get some sort of recession and economic shock. But AI isn’t going to be the near-term cause of this.1
Anyhow, let’s stop doom mongering and start prepping. What can we do to mitigate this risk in our personal lives? For me it basically means a few things:
- You need to own capital. You need to own assets. This is an extremely exciting time for capital because it will be the direct beneficiary of any potential decline in labor. Sorry, Karl Marx.
- You need to be creative and entrepreneurial. AI is the ultimate decentralizer. Individuals who can creatively leverage these tools will build enormously valuable personal brands and businesses. Don’t shun this technology. Lean into it.
It’s not all negative. This is the most exciting time to ever be alive. But it will also be challenging in unknowable ways.
2) The Duration of Equities
I am getting A LOT of questions about the Defined Duration strategy in the new book. The consistent feedback from readers is “how did this not exist already?” And also, “how are you quantifying duration exactly?“2 My answer to the first question is that time weighted asset-liability matching strategies have been around for a very long time, but they just haven’t been available to financial advisors and retail investors. Or, at least the ones that have been available are clunky or simplistic “bucketing” strategies. I actually learned the basics of this from bankers I used to work with after the GFC. The problem was the more formal strategies were mostly confined to fixed income instruments where the outcomes are virtually certain. You see, a bank can’t go around matching long-term liabilities to equities, for instance, because it’s either against the rules or would expose them to unnecessary risk. But your average person doesn’t have these constraints. It’s one of many reasons why I think asset-liability matching strategies are actually better suited for retail investors than institutions.
As to the second question, the answer is relatively simple. I am quantifying an assumed sequence of returns risk. For instance, let’s take our expected return simulator and plug-in a global CAPE ratio of 25 and an expected inflation rate of 3.5%. You get a 5.07% expected annual real return. Now let’s apply some sequence risk and assume the stock market falls -50%. If that happens then our “Defined Duration” of stocks (the break-even point using this assumed return) is 14.02 years. I am being overly conservative in many ways, but that’s a reasonable time horizon over which to judge the sequence risk of the stock market. Given this we can now apply specific time horizons for asset-liability matching purposes. For instance, you wouldn’t match a 24 month liability to this 14 year instrument. But, what if you took an 85% Tbill instrument and blended it with a 15% stock allocation inside a multi-asset instrument? If Tbills have a duration of 0 and stocks have a duration of 14 then this instrument has a duration of about 2.1= (0.85*0 + 14*0.15). Now this is where things get interesting because now you can start matching liabilities with multi-asset instruments and potentially boost returns without sacrificing significant temporal certainty.
This is especially interesting to me when we start getting longer out on the duration curve. For instance, if you had a 10 year liability and stocks have a Defined Duration of 14 while bonds have a Defined Duration of 5 then you could choose one of three options:
- 10 year T-note yielding 0.5% real returns (current 10 year yield of 4% minus our estimated 3.5% inflation).
- 10 year TIPS yielding 1.8%.
- A 50/50 stock/bond instrument that has a Defined Duration of about 10 (14*0.5 + 5*0.5).
Options 1 and 2 give you certainty of principal when you need it and also give you a pretty meager real return. Option 3 has a huge asymmetric difference though. While it was positive in real terms in 87% of all rolling 10 year historical periods its average real return was 4.5%. In other words, in 87% of outcomes you had a huge asymmetric positive payoff that dramatically beat 10 year bonds. As I often say, long bonds are terrible long duration matching instruments when compared to equities even though equities have more embedded uncertainty.
The difference between 4.5% and 1.8% might not sound like much, but $100K compounds to $155K vs just $119.5K in these scenarios. Of course, if you needed 100% certainty of the outcome then option 1 or 2 is what you have to go with. But if you don’t, and you’re able to bear some temporal risk, then option 3 is the vastly better option in most cases. So you see, this is where the Defined Duration strategy is really interesting in the context of financial planning and asset-liability matching. By embedding equities into the strategy you’re not only enhancing potential returns, but you’re doing so without sacrificing a lot of temporal certainty. It’s kind of similar to what a Target Date Fund is trying to do, but we’re actually quantifying the time horizon more precisely. And our approach is vastly more customizable than a TDF approach.
This is all especially interesting in the context of the news that Google was issuing a 100 year bond. Isn’t that crazy? Who would buy this thing other than traders? Imagine doing ALM strategies with a 100 year corporate bond. Ha. The only thing worse than a sovereign 100 year bond (which is loaded with inflation risk) is a 100 corporate bond (which is loaded with credit risk and inflation risk).
3) Why stocks and bonds are your core assets.
Over the course of the last 15 years we’ve heard numerous narratives about how the “traditional finance” approach was going to be replaced by crypto, robo advisors and now AI. But so far none of these things have actually replaced much of it. Crypto, for instance, is still a very small and fringe component. The crypto space is where the narrative appears to have really lost momentum in recent years.3 As Bitcoin slips in price in recent months I think people are finally starting to ask – “it’s been 15 years of Bitcoin now and where are the widespread use cases?” I don’t mean to imply there are no use cases. I often write that BTC is a no-brainer hedge if you live in a third world country, for instance, But in places like the USA where you have large and developed payments systems the use cases are still negligible and the traditional financial system has left crypto in the dust. For instance, since 2008 when Bitcoin was created the traditional stock market in the USA alone has grown from about $12 trillion to $70 trillion. Meanwhile, BTC has grown from $0 to $1.4 trillion. Not nothing, but the traditional finance system has grown 40X more than the BTC ecosystem over this period. More importantly, Bitcoin’s impact relative to traditional payments systems is virtually non-existent.
It all goes back to basics in my opinion. Firms can fund their spending in a few basic ways. They can retain earnings, issue debt or issue equity. They can’t issue BTC and they don’t even like retaining BTC for payment because its FX price is too volatile in relative terms. This doesn’t mean it has no role in the monetary system or portfolios. But the reason it hasn’t become a more central component of the financial system is really rather basic – it doesn’t have the properties that allow it to be a primary funding instrument.
The point is, there are fundamental reasons why stocks and bonds remain the core instruments in our portfolios – they’re the primary instruments that correspond to how firms fund their cash flows. And no matter what happens with technology, AI and crypto this is something that doesn’t appear to be changing any time soon.
I hope you have a great weekend. As always, stay disciplined out there.
1 – I’ve been saying for years that the really scary scenario is when the robots finally come online. One day we’re going to start seeing robots navigating construction sites. And that’s the day where this becomes legitimately scary for labor because once AI can replicate difficult manual labor none of our jobs are safe.
2 – I did have a good laugh (or was it an angry scream?) at one Amazon reviewer who complained that the book didn’t provide any “novel” strategies. Well, this person clearly didn’t read the book entirely because there are at least 4 strategies in the book that I personally created. Defined Duration, Flying Ladder, Countercyclical Rebalancing and Future Cap are all original, sir!
3 – I say that, but the it’s the robo advisors who really ended up having no impact here. In fact, they all turned into banks, high yields savings firms or brought humans in to service their clients….