Weekend Reading #387

This is the three-hundredth-and-eighty-seventh weekly edition of our newsletter, Weekend Reading, sent out on Saturday 24th October 2026.

To receive a copy each week directly into your inbox, sign up here.

*****

What we're thinking.

With the AI names in free fall this week there has been much discussion about a rotation. Rotations come in many shapes and forms in markets and this one is interesting. Any time names go up while a formerly great group of stocks collapses (usually tech) it is called a rotation. Funnily enough no one ever calls it a rotation when money flows the other way! In this particular iteration, AI (tech) has been cratering after having gone crazy to the upside for the last few years. On the other side we have the likes of Berkshire Hathaway, US food retailers and some friends. The rally in these names being rotated into is anywhere around 5-10% whereas the collapse in the AI names is anywhere from 20-30%. Option one is we have a full rotation, and a new bull market is underway in the other 480 or so names in the S&P500 which everyone was complaining about a few weeks ago showed bad breadth as a precursor to a crash like (insert year of choice). This could easily be validated by collapsing AI usage costs thanks to open source. The implication being that we are entering a broad-based productivity-driven bull market as these tools proliferate and help most companies cut costs and improve efficiencies. Case in point here is banks. Bloated, boring and very much ex-growth mostly, simply collapsing the cost of system maintenance and IT would make a big difference to margins. There are examples like this in almost every "boring" sector. Option two is we have a short rotation before it all comes down. The thinking being that an AI bust will collapse growth and despite tools resulting in Option 1 in the real economy, it takes a long time and the resulting effect of the bust on credit and growth all round results in a collapsed economy. Falling commodity prices may be alluding to this. If you are looking for the answer, guess what - we don't have it. Time will tell and price action will tell us where this is going. What we will say is that we have seen many times before Option two - where excitement builds around a rotation into value type names, only for it to fizzle out when the sh*t hits the fan. Is this time different? It might be, given we have never had AI before. But are you willing to bet on that yet?


What we're reading.

This is a fantastic profile on one of my favourite comedians, Larry David. It turns out he really wasn't ever anything special until he was. There were no signs he would do anything remarkable at all! Excellently written and full of fun too. DC

A Fusion lab based in San Francisco called Marathon Fusion this week dropped an interesting announcement: that they’ve managed what alchemists of old have dreamt of, turning mercury into gold. The process involves bombarding mercury atoms, knocking out neutrons and allowing the destabilised mercury to decay into gold. Costs aside, the main issue is that this created gold takes about 13 years of storage for the radiation to wear off. This has of course attracted the commentary of the gold vs bitcoin crowd, on one hand the argument that even if new gold supply comes on, you at least own your gold; on the other, the argument that “you can’t software update gold”, which unfortunately I do find relatively less compelling.
 
Earlier this week a vulnerability was found in PHP, one of the key foundational server-side technologies used in everyday internet applications. No doubt everyone would have come across a page ending in .php at some point in time – that’s it. The source of the disclosure seems to be still unclear – by some accounts, a cybersec account called Jλckλι published the vulnerability without giving the world enough time to patch; in other accounts, the vulnerability was found using an AI model prompted to find exploits in PHP. Either way, it seemed to cause sufficient panic globally as governments rushed to shut down websites to give themselves time to patch. The key risk was unauthorised access by a third party to restricted data – could large amounts of cash be stolen? Probably not, given the degree of centralised control over networks like SWIFT. Ironically, stealing electronic money doesn’t work as well as stealing physical money. But could it lead to phishing attacks, impersonation emails or sensitive information being used, manipulated and weaponised? Or small amounts of cash taken out of bank accounts that we don’t notice? Possibly. If it’s true that all it took was asking an AI to “find the problems with PHP”, we’d better hope it’s the good guys doing the checking.
 
Here’s an interesting thought experiment: assume the hypothesis that AI is simply regurgitating and rearranging past data – does it mean that AI can’t come up with a new idea, or something creative? Or is creativity simply the stacked up recombination of old ideas in novel ways? A methodology was proposed to test this: semantically index game descriptions of all the games on Steam, including tags and categories, and represent that index in vector space, so games that are similar are clustered together, forming a cloud of ideas representing existing games. Then based on the parameters of that vector space, start picking random spots outside the existing cloud of ideas, which would necessarily represent ideas that currently don’t exist. Are those new ideas? Someone put this methodology to the test, and it turns out that some pretty fascinating game ideas get tossed up. For example: “Pay Zone: Reading the Rock” – A petroleum engineering sim where you learn to read well logs, seismic sections and production curves. Pick the pay zone, plan a directional well through layered rock, set a budget, and see whether your reading of the data was right. 99.9% of Steam games have a closer twin than this idea. That’s pretty creative if you ask me. EL

What we're watching and listening to.

One of my other favourite comedians is Sacha Baron Cohen, of Borat and Ali G fame. Ali G is making a comeback with a new movie coming out soon. Here Cohen sits down for an interview about his life and career. If you are a fan of his stuff, it's a cool watch. Not funny though, its a serious sit down.
 
Citadel's Ken Griffin had a great Bloomberg interview this week in which he speaks about AI, Trump and lots of stuff. He thinks it is a bubble. He also loves Miami. He does most of the time speak a bit like a politician as he clearly has candidates he backs and doesn't want to step out of line. The fiasco with Mamdani some months ago is also discussed. He is becoming more public in his speaking. Maybe he wants to be a politician. Probably not. As a major market player every interview is compulsory viewing. This one was worth it.
 
Here is a real treat. Joe Walsh, my former colleague at Capricorn, is launching his new hedge fund. I met Joe when we brought him in as an intern at the tender young age of 16. Even then he was an absolute mine of information. Now in his early thirties, he appeared on the Odds on Open podcast to talk about his investment philosophy and markets in general. There is absolute gold here. An obsessed student of markets, I can't see quite how he won't succeed spectacularly. Really valuable and most importantly HONEST insights in here and hit me up if you want to learn about his fund. 
 
Finally, the new Grand Theft Auto is coming to life over a decade since the last one. Grand Theft Auto 6 looks like more of the same - incredible but hopefully updated for a new world. Video games seem to have lost their sex appeal over the past few years. This will be a fascinating launch. A more anticipated game we could not ask for as a good test case. DC