Weekend Reading #384

This is the three-hundredth-and-eighty-fourth weekly edition of our newsletter, Weekend Reading, sent out on Saturday 3rd October 2026.

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

*****

What we're thinking.

A clear divergence between stocks and bonds this week as markets held up, with the Nasdaq reaching new highs, while bond yields continue to spike in the US, driving other sovereign yields. There are two camps at present. The panickers (who sometimes are right!) and those who are a bit more relaxed and even bullish. Either stocks ultimately follow bonds lower or bonds follow stocks higher. Our take is that despite the weakness in yields, at the index levels stocks are holding up well with some stocks even taking out highs. We don’t really mind if most stocks are going down and breadth is bad because there is plenty of opportunity and there is always a bull market somewhere! That being said, like we mentioned last week, keeping an eye on the floor is important as it can get ugly at any moment! Our base case is that we have new highs in short order especially in AI linked names, which have churned and not really gone up much over the past 3 months or so. The argument was at one point in recent months that high bond yields are bad for AI capex but now bond yields are even higher and most genuine AI names are well back in bull market territory, Even the memory names are up. Leon Beressi sent this great meme to us.

Our response was well when price goes down its cyclical and when price goes up its a 6 p/e growing 100% AI revolution. And we readily admit that we have been on all 3 of these branches over the past three months as price fluctuates. We are not conviction merchants. We are price merchants!


What we're reading.

This paper just reminded me once again how negative for us uncontrolled use of our phones has become. And the key for me is that the epidemic is led by adults. The kids are downstream of the problem with adults and addiction to devices has major effects for ADULTS according to this study. It really is quite simple. DC

It seems like the AI story is taking another interesting turn, equal measures disturbing as potentially desperate, with this article in the NY Times, which tells of how Anthropic tried to lobby religious scholars, including the Vatican, into recognising the personhood of AI and deeming it conscious, even trying to get the Pope to alter the text of his encyclical “Magnifica Humanitas”. This comes on the heels of Anthropic and OpenAI’s apparent efforts to engineer a crisis around AI and get regulation imposed (pretty much the FTX playbook in crypto before) where they are named the gatekeepers, and instead finding themselves talked down by their biggest supplier Nvidia effectively saying that agents running out of control is something that a secure sandbox, not monopoly-granting regulation, solves. They then went on to deliver the sandbox at chip level. To make things a little worse, the FTC seem to be on the case as well too. The realisation is making its way through the market now that making the best LLM is an expensive but otherwise moat-less endeavour. The AI use isn’t going away, and the trade seems to be back looking for the next leg of the story, but it seems that making another better LLM is no different from another website builder back in the 2000s (Who remembers Geocities?), especially with open source continuing to deliver for free a reasonably good baseline level of reasoning.
 
On the jobs front, the initial data seems to suggest that the jobs apocalypse likewise predicted by OpenAI and Anthropic (presumably as part of the same campaign of fear above) hasn’t happened. That’s good news, for now, although I continue to be of the view that already as a small business there are functions that would have previously required hiring a junior analyst to do which simply don’t exist anymore as a standalone role. Even considering my use of fully paid credits on Grok via Perplexity Computer (my weapon of choice for the moment), a tedious data crunching task probably costs around $10 in credits to do a proper job with huge context in a reasonable time (aka ASAP, or 15-30 mins). That’s more than a good deal considering no training required, and availability at any time of the day (or night) which no human would ever reasonably sign up to – not for $10. Without the time constraint, the work goes straight to Qwen (the uncensored version of course) and it’s just a matter of paying for electricity and a couple of hours of overnight running. Where does this leave humans in terms of roles? My guess is “liability”. As OpenAI and Anthropic are finding out now, AI can do what AI does, but the legal system isn’t happy putting a block of code into jail. A human needs to take the fall – and that’s what the humans are going to be for. EL

What we're watching.

This is such a cool interview with Tom Cruise of all people. And the piece written by Rory Sutherland about it is awesome. Mainly because I feel like the way he describes relationship with movies is how I feel about mine with the markets.
 
"The thread running through the conversation is his unwavering commitment to learning about the world around him. He starts by describing his childhood drives, his curiosity and his dreams. In a touching exchange he says he stopped talking to adults about his big dreams so they wouldn’t be crushed. He committed to keeping them secret but never letting them go. He talks about how he never wanted to be famous, that his connection to movies began because they were his way into exploring the wonder of the world, the conduit for his interest and his fascination with people. It sounds to me like they were reflective of his joy of creation."
 
"Movies are “not what I do, they are who I am” he says, and explains that he didn’t go for fame, he went for competence, for skill, “whatever level that was.” He articulates a commitment to curiosity and craftsmanship that resonate strongly and bring to mind the incredible inspiration that I got from Professor Roger Kneebone on a call recently, and from his book Expert. DC

What we're listening to (and doing).

A couple of weeks ago I started using Grok Bot for a new bits and found it useful for some things but not others. As I am not on Facebook or Instagram etc and also live in Cape Town where Muse is not downloadable yet, I haven't had a chance to experiment with it. But this week thanks to Stephen Mcbride of the outstanding Rational Optimist Society, I got an invite to it. It is actually pretty life changing. Admin that I had been procrastinating due to its manual complexity and my sloth I handed straight over and my word did it work miracles. I cannot recommend this product enough. It works entirely through WhatsApp and (currently) is free. I had also listened to founder, Noah Shinn, on Invest like the Best earlier in the week which I also suggest everyone listen to it. Amazing product and great listen too.  DC