I think, seeing recordings of people using LLMs to accomplish non-trivial tasks would go a long way.
I’d love to watch, e.g. you Simon, using these tools. I assume there are so many little tricks you figured out over time that together make a big difference. Things that come to mind:
- how to quickly validate the output?
- what tooling to use for iterating back and forth with the LLM? (just a chat?)
- how to steer the LLM towards a certain kind of solutions?
- what is the right context to provide to the LLM? How do it technically?
I believe Simon has full transcripts for some of the projects he’s had LLMs generate the code for. You can see how he steers the LLM for what is desired and how it is course corrected.
I don't generally recommend the Intelligent Investor. I think it's almost always a mistake for ... 99% of new investors to bother with Benjamin Graham. His material is far too dense and often advanced for anyone that isn't quite an experienced investor. I'm sure there are exceptions, however I've found it's a huge turn-off for most new or newish investors, it delays / stunts their learning process, it's an obnoxious book to try digest if you're starting out. It'll make you hate investing or think that value investing is difficult (it's not, it's simultaneously the best approach for generating consistently high returns over time and very easy to learn).
Here is what I point new investors to:
- Buffett: The Making of an American Capitalist, by Roger Lowenstein.
- Margin of Safety, by Seth Klarman
- The Little Book That Still Beats the Market, by Joel Greenblatt
- Common Stocks and Uncommon Profits, by Philip Fisher
- Business Adventures: Twelve Classic Tales from the World of Wall Street, by John Brooks
- Peter Lynch also has a couple of optional books that are decent and very easy to digest for a new investor, very common sense oriented.
- Also optionally, Buffett's various writings are often excellent, however they're all over the place in focus, so it's hard to pick one. His annual letters for example can be acquired on the Kindle or from Berkshire's website and many are worth reading (if somewhat boring for most people I suspect).
The single most important thought in investing, in my opinion, is to always be cognizant of price vs value. What you're paying, what you're getting in return. Then always be aware of, always estimate as best you can, what your moat is for the investment at the price you're paying (what Klarman and others have called a margin of safety). How much can go wrong with your investment before you drown? How much room for error is there in the price that you paid? I like the Buffett book I reference above, because it pounds home that concept while introducing how Buffett came up, how he thinks (I don't particularly like his book, The Snowball, for that).
Also, Margin of Safety is out of print. However, there is a certain Archive site with a time machine, that if you were to put this url into it:
You'll find an archived copy of the book in PDF format. Alternatively you can put that file name into Google and find some other copies of it floating about still (Klarman refuses to put it back into print and had been having the PDF copies taken down).
Bless your dear soul for the Margin of Safety link. I've been meaning to read it for quite awhile, but you can guess why not. All your other commentary is top notch, although I still think the SP500 is fairly valued (won't have great 8% returns going forward, but won't be 0% stagnant).
Some if it is time and experience. Seeing markets come and go, seeing valuations come and go. You could perhaps study historical markets to gain some of that, but there is no better teacher than going through it (including taking some beatings along the way, along the process of learning and instilling discipline).
The absolute easiest things to look for (things most anybody can do), is growth vs valuation, along with having enough of an understanding of the business you're buying part of, to know whether they have an enduring position in their market, whether they have a moat or edge that isn't going to easily vanish. It's important to understand the context of the business you're buying into. Ultimately if you're going to self-manage, you have to decide what kind of ratio on growth vs valuation you're willing to accept, what's too high. These are largely subjective decisions, there is no right or wrong answer in most cases, only answers that entail more or less risk (the worse the ratio (eg high valuation + negative growth), the closer you get to it being an objectively wrong, dangerous answer though). Personally I prefer a balance on that equation. I'm a value investor, and it's commonly thought that value investing is a conservative, low return approach, however that's not inherently the case, that's up to the investor's approach (Benjamin Graham was a cigar butt value investor, looking for a last puff on cheap stocks, and his approach produces far lower returns over time than Buffett's approach to value investing). If a stock has a medium or high valuation, it can be a great value proposition if the necessary growth is there as a proper offset.
For example, if you go back to December 2018, Facebook was trading for around a $370 billion market cap, around 14.8 times operating income for that year. Why didn't more investors see that opportunity? It should take only a few minutes at most to run a basic extrapolation of a modest growth rate and see how that Facebook value proposition would end up a few years into the future. And yet, sentiment on the stock was irrationally bearish. That's nothing more than emotion, herd behavior, and it's extremely common. It's investors listening to other low-value opinions, it's Wall Street money being the typical followers that they are. People are generally terrified to stand apart on any decision, much less an investment. Many of those investors end up being the desperate sellers you want to buy from as they sell you FB at $137. They have no idea what they're doing, they have no idea when to buy or sell. And that's true of most of the professionals on Wall Street, they're almost all clowns (very highly paid clowns, because they're operating in a protected cartel). Was Facebook's social monopoly - their moat, their edge - going to vanish soon circa December 2018? That was an absurd, silly, borderline stupid premise, and yet it was a commonly floated notion; there was wildly bearish sentiment going around at the time - and yet back in objective land, where you always want to remain, their business was still firing soundly. So if an investor could brush away the irrational people spewing their emotional bias about FB, you could focus on the actual business and what it was actually doing, and run some very straight-forward projections into the future (even being conservative about it).
I'll give you another obvious example from my perspective. In March I posted here about which stocks I thought were interesting during the crash. I argued about Square and why I liked it. For that stock I looked at a few things. They did $4.7b in sales in 2019 and their market cap was down to like $17-$19 billion during the crash, so it was trading for around four times trailing sales with a solid growth profile (and, in theory, a lot of growth ahead of them, as they were still relatively early into their growth phase). Their operating profit/loss picture had persistently improved as well, they weren't at any risk of bleeding to death. Now, that's a laughably obvious value opportunity, that's dirt cheap value. That's a margin of safety or moat of safety that is massive, a lot of things would have to go wrong for Square to not be a great buy at that price. The calculation when you input all of that, including its business context (putting in risk for the pandemic), is that they could probably see their sales fall by a lot during the pandemic and their growth rate plunge and it'd still be fairly valued at near a $18b market cap with maybe some further downside risk of 1/3 from there if the damage were really bad. And then you'd have to believe they weren't going to resume growth, that they'd never get back up, to think the position would be really dire. That easily goes into the low risk camp.
Most stocks and situations aren't so obvious, so easy. The principle is the same though, it just takes more effort, thought, measurement, well considered projection, to reach a conclusion about the price you're paying vs the value you're getting. Start by asking: what value am I getting for my money? What value is represented behind that price. That includes the business comprehensively, its prospects, its growth near-term, its potential or likely growth longer term, its risks; and often it can include externals, depending on the business (is this business depending on a trend sustaining, on a commodity price remaining high/low, on politics in China, on hurricanes in Florida, and so on).
Getting really good at rapidly extrapolating with multiple outcomes is a great skill to acquire. Take a stock, absorb its P&L statements from the last 3-5 years or so, learn some about its business, and run some scenarios in your head (or write it down, whatever works). It's a training exercise, and over years you get to mentally compare your projections vs reality, and you can adjust your mental models if you're missing by too great of a margin about how you gauge such things, how you extrapolate forward. Over time you learn some things - risks, potential upsides - to look out for in businesses, to put into the calculations, that you maybe didn't know in the beginning. Did you include a variable and apply a discount for the potential that corporate income taxes might increase (and is this corporation's earnings primarily domestic, like an airline or bank)? Etc etc. Eventually you can look at the profile of a company, its valuation and P&Ls, and get a great sense almost immediately whether it's in your risk wheelhouse, on what value its presenting at the current price.
Run a value calculation on a bubble stock. What value is Snowflake presenting to me at a $82 billion market cap? That when it finally gets around to $3 billion in profit, it'll have a 27 PE ratio and probably low growth to match, all just to get it to where it's already at. What do I think its growth will look like in the century it finally gets there? Why should I pull returns so far forward for that stock, what extraordinary value am I getting to make up for the hyper price I'm paying? When Salesforce had a $82 billion market cap back in 2017, they were set to do $8 billion in sales that year. Snowflake presents something around a 20x worse value proposition than 2017 CRM (when you include their horrible burn rate and dramatically lower sales). Who the hell in their right mind would go anywhere near that? It has disaster written all over it (or at best, extended mediocre returns). And 2017 CRM wasn't a steal, that was a rich valuation. Snowflake requires that you pay an extreme price for growth. That's where an investor has to decide what ratios they think are acceptable; place those settings in the wrong place and you increase the odds of getting crushed, getting trapped in a lost decade, or taking a haircut you have to potentially spend years to recover from. Do I think I can buy Snowflake in the future at a steep discount to the $429 highs it previously hit? I wouldn't be surprised if it can be picked up in the future below $100, possibly far below that. Do I think this market's hyper valuations will persist forever? No. And you can go from there. Inevitably something will crash the stock, something won't go right, growth will be weak for a year or multiple quarters, any number of a thousand things, and it'll tank the stock hard, and they're priced for perfection. Snowflake has to execute to perfection for the next decade, non-stop, to prevent that stock from tanking at some point, and even then the market may kick their feet out from under them regardless. Do I think it's likely they can execute like that for a decade? No, hell no. They'll have their Qwikster moment too, and if they're lucky they won't go out of business (classic dotcom bubble scenario, high burn rate, market crashes, economy tanks, business vanishes; the supposedly impossible happens).
Why are people buying Tesla at $780 when they could have had it for under $180 a year ago? Where were these people a year ago? It's a great buy at four times the price and wasn't back then? Wild irrationality, a complete failure to understand anything about Tesla, a failure of due diligence, people generally buying things having no idea what they're doing. The market is always filled full of investors looking to buy your high priced stock at the wrong time, and later they'll be begging you to buy it from them at 10% or 20% the price. Rinse and repeat, it never stops happening, it will never stop happening, it's standard issue human behavior, and you can generate consistently great returns taking advantage of that fact, so long as you can control yourself and maintain discipline about your behavior during times of manic greed or intense panic (because the other people sure can't).
Thanks for all the detailed responses. Very much appreciated food for thought.
I have a specific question about the calculation of the fair value. My understanding is that there are two steps:
1. Project the future earning (that is the difficult part)
2. Discount the earnings and get the NPV - the fair value of the company
Currently the interest rates are so low that the NPV will be quite high and not that far from the current valuations. Indeed, this seems to be the rationalization by many for the sky high valuations of today. But this seems wrong to me. What would you advice? Pick some other interest rate, maybe an aspirational rate of return for ones investing? Or don't try to compute the NPV and think in terms of multiples like P/E and P/S?
When I say fair value, it's what I consider to be fair value. As an investor you always have to ultimately make those decisions for yourself, or you have to defer to another person's judgment on the matter (whether a talking head on TV, or pump & dumpers on Reddit, or newsletters, etc). I'm not basing that on something some guy put into a book 70 years ago about how to value a stock, even if some textbook'ish knowledge can be worth learning to use as you go about coming up with your own valuing formulation (as in the case of Ben Graham). It's based on my past ~26 years of experience with stocks and what I look for in investments. You'll find with experience as an investor, if you're self-educating and or managing some or all of your own investing, you'll come up with your own tests for investments, your own way of valuing what you're buying & selling (or you should anyway). You can take pieces here and there from others and assemble it based on how you like to invest, inevitably over a lifetime it no doubt becomes an amalgam from what you learn.
So for example if I think the fair value for Coca Cola (KO) is 30% to 50% lower than where it's at today, that's not based on a textbook valuation approach. I base it on what I'm willing to pay for growth, and Coca Cola is a pathetic non-growth machine (not to mention a giant sugar liability). I look at Coke's financials and, with some understanding of their business, I ask: what am I willing to pay for zero or negative growth across time? China's boom has come and gone and Coke's growth - as a global business - has recently been stagnant, mediocre, so what are their prospects going forward? I don't like that picture at all. I might be willing to pay somewhere between 8 to 15 times earnings for zero growth (depending on context; I might pay less for a financial firm than a tech firm, and so on), if there is something I like about a company. Coke's multiple is closer to 27-33 lately. Why would anybody ever pay 30 times earnings for zero growth and bad prospects for growth? Coke is a very easy fair value calculation as far as my personal judgment is concerned, their persistent growth problems make that a super fast decision. I'll look elsewhere. McDonald's is in a similar boat as Coke, it's a horrific value proposition, 30+ times earnings for a business with very little (or negative) growth. I might pay 12-15 times for MCD or KO, maybe. Personally I tend to really dislike companies with no growth or weak growth prospects going forward, it's a giant negative in the margin of safety calculation (growth is a first-aid kit for problems that inevitably crop up in a business over time, random messes, it applies a bit of a balm, helps as an offset in the value calculation; if you don't even have growth, inevitable problems are that much worse when they happen).
Fair value means I've looked at the stock in a way that I prefer to approach a stock and I've made a determination for myself, for my investment purposes, as to how much I think it should be worth. And I may come up with a few versions of that, one for an average market (with typical multiples), one for a slightly bubbly market; typically I disregard trying to come up with a value based on a mania, I'm not a buyer at that time in most cases. Those variations, models, are meant to inform myself as to the flex in my investment. If valuations merely go back to where they were in 2012 or 2016, how might my investment perform if its multiple is reset 1/3 lower? Will I get killed on the price I paid? It's modeling.
Interest rates will absolutely distort the context of deciding what something is worth, that falls into the variations, models, you build for different scenarios. The point of doing that is to check / prepare your position against a bad outcome. People claim that low interest rates will keep stocks inflated, so there's nothing to worry about; I like to point out that multiples were far lower at numerous points in the past decade when interest rates were at zero and we also had QE going on. How about if we just roll back to where multiples were in 2014 when rates were zero (and our economy was better positioned in 2014 than it is now, although our headline unemployment rate was similar)? If I were a buyer today I'd absolutely be running that simulation for myself whenever I buy.
Yeah they don't tend to talk very openly about who is doing the buying for what. Occasionally in the last few years Buffett will indicate if it was him specifically buying something, usually outsized positions.
I think either Ted or Todd is likely is doing the buying on Snowflake (at least instigated the premise), perhaps with Buffett's sign-off (given it's an increasingly large position). I also think one of the other buyers is likely responsible for Berkshire's position in VRSN.
There's a high risk that Snowflake position will humiliate Berkshire Hathaway. I think it's a mistake. It wouldn't be Berkshire's first mistake in dabbling in tech but it looks like it could be the biggest.
I have great respect for Buffett's historical performance, however I consider him mostly done at this point. He's conservatively managing the end game of his career now, he seems to be intentionally avoiding doing anything that might stretch beyond his lifetime now (I don't think he wants to do anything that might need a decade to manage that he might have to offload onto the next person). I think he should have taken advantage of KHC's weakness for example, to strip Heinz back out of the conglomerate and break it up and sell off the rest of Kraft, the stock was so cheap at times you could have almost gotten Heinz for free in the process. Instead Buffett is sitting on a hundred plus billion dollars yielding squat (KHC was down to near half the market cap it currently sports, which was the general time to grab it to try to get the Heinz business, and then sell off the lesser pieces, either after you repair what was ailing KHC or immediately depending).
Once a security has entered bubble territory, by definition its price has become disassociated from any underlying fundamentals - bubbles form when rising prices drive positive sentiment which pulls in fresh buyers who buy into the rising price story, driving the price higher which in turn draws in more buyers in a self feeding cycle that only ends when the supply of fresh buyers dries up.
Because this is a self feeding cycle, the price can double & redouble from almost any point if the "story" is good enough. So you can be dead right about a given asset being in a bubble, but if you short it the odds are that you're going to get painfully stopped out long before the bubble collapse finally happens. As Keynes said, prices can remain irrational longer than you can remain solvent: The history of investing is littered with individuals who were right, but lost their capital because they were simply too early.
Obviously BTC can be in a bubble & whilst still having real value - the two things are not necessarily connected. (Although having a good future value story is helpful for a bubble to form in the first place I would argue.)
Part of the problem is that you don't have time to think deeply about things, because there is to much to think about. The solution is simple: drop almost everything except what is most important to you and allocate plenty of time to for it. (This solution is simple but not easily implemented.)
A good read on this topic is "The Shallows: What the Internet Is Doing to Our Brains" by Nicholas Carr
I’d love to watch, e.g. you Simon, using these tools. I assume there are so many little tricks you figured out over time that together make a big difference. Things that come to mind:
- how to quickly validate the output?
- what tooling to use for iterating back and forth with the LLM? (just a chat?)
- how to steer the LLM towards a certain kind of solutions?
- what is the right context to provide to the LLM? How do it technically?