Weekend Reading #353

This is the three-hundredth-and-fifty-thirdweekly edition of our newsletter, Weekend Reading, sent out on Saturday 28th February 2026.

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

*****

What we're thinking.

And the march goes on. Commodities and EM lead the way as tech continues to sell off all-round. And for good reason! Nvidia results came and they had their obligatory beat, but the stock sold off and the broader selloff has accelerated once again. What markets really hate is UNCERTAINTY. Usually, a reason comes up, and markets sell off and people panic, and in the end, it resolves as the uncertainty is removed one way or the other. The problem this time is that there is no way to know how the uncertainty around the impact of AI will resolve and when. We will only know through the passage of time. So, we do not think that there will be a resumption of the tech led bull market like every single time before. Some are saying that tech is now attractive because valuations are lower. Well then does that mean one shouldn't have participated for the past 15 years or so when valuations were higher? In public equities valuations are a nice to have but, in our experience, they are but one small driver of price action. Generally, when things are going from highly valued to lowly valued you want to be OUT. The derating goes on and we suspect as our base case that we are in a longer-term shift from a US tech-led to an EM-commodity led regime. What is really getting our attention though at present is the private credit space where things are really hotting up. It really is reminiscent of the 08 crisis with everyone wondering who is holding the lemon. Once again, we don't know how far this can spread but it has the INGREDIENTS to be systemic given the interlinkages to insurance and off-balance sheet entities. And it's all hidden of course by those beautifully private monthly performance marks! How wonderful it all was when one could just see the numbers going up for all those years and be thankful one didn't have to deal with the hassle of public markets monthly marks! We all wait still with bated breath to see whether the US will attack Iran. And also, this week apparently it became clear that Bitcoin collapsed because of Jane Street. Yippee, we have the culprit that caused it to go down! We can all aim for all time highs again now that they are out the way. Right? Right? So much good stuff going on we hope you are all having as much fun as we are.

What we're doing.
 
Following the excitement around clawbot/openclaw, I decided there was no better way to learn than to try it out myself. However, downloading a pre-built package that has in recent weeks been shown to have huge security vulnerabilities, not to mention this famous episode that saw Meta Superintelligence’s Director of Alignment (what a title) face off with a rogue clawbot instance that wiped out an entire email inbox. As an aside, clawbot was by no means repentant – another issue to think about, LLM morals.
 
So, after reading the docs on openclaw, I took to Git, and with a little help from Perplexity, replicated some of the basic functionality of openclaw, minus the ability to wipe out my entire PC. Moral of the story: the bulk of the supposedly fascinating capabilities of openclaw (or any of its peers) actually already exist in modules like node.js, and what LLMs bring to the picture is really a dynamic replacement of the “if… else… else…” functions that would need to be programmed by a human – a tiresome and all-consuming task if a human were asked to contemplate and spec out all the possible permutations of what a potential user could do. Put differently, the automation that we are witnessing become increasingly widespread now had always been possible – it was just the scenarios being unable to be written in a definitive enough fashion which was the bottleneck. LLMs provide the final piece of the puzzle, a logical “formula” that enables the deployment of natural language logic and supposedly “common sense” (or “consensus sense”) into code.
 
And no, I did not purchase a brand-new powerful Mac Mini to run these – deploying node.js on a local dev server via Git takes very little by way of computing requirements, and sending triggers (e.g. whatsapp messages) via API to the likes of Perplexity or Claude needs only an internet connection. Of course, if the information being handled is proprietary or confidential then it’s best processed locally on a local model deployment, but the truth is for everything else (i.e. if it’s something you’d be comfortable asking on the web version of your LLM of choice) you don’t need crazy powerful compute. And no, you don’t need a full powered Mac Mini.
 
Everyone can do these – it just takes time, focus and a good dose of curiosity. EL

What we're listening to.

Jeff Currie, formerly of Goldman Sachs, appeared on the Macrovoices podcast this week and gave his view of the commodity supercycle. It really is a compelling case. Is he right if he called for it in 2020 though a few years early? It really doesn't matter in our view. What matters is what is happening now. He believes we are in the first inning. Can you imagine how high some of these prices can go if he is even a little bit right?
 
Josh Wolfe interviewed VC giant, Bill Gurley, this week. Gurley has a new book out apparently trying to help young people choose which careers to enter into. It is called Runnin’ Down a Dream and sounds great. I'll probably buy it. This is a really cool chat. I'm generally not into VCs interviewing other VCs but this was really thoughtful and has some good tips on a topic that is top of my mind at all times. In the age of AI, I spend almost every minute (when not managing the fund of course) thinking about the best way to prepare my kids for a very different world to the one I grew up in. I wonder if my parents felt the same. I somehow do think though that it isn't the same. This is a very big change we are undergoing and the uncertainty is enormous. DC

What we're reading.

After taking a while to warm up, I really began to enjoy the first book in Joe Abercrombie's First Law fantasy trilogy, entitled "The Blade Itself". It's a richly described universe with cracker characters. Very reminiscent of Game of Thrones in fact. Sinister and dark but only subtly. Of course, lots of violence and weird stuff though as usual. Cracker. DC